Datasets:

disi-unibo-nlp

/

SciLay

like 0

Follow

DISI UniBo NLP 11

10.3390/ani12070905

PMC8996860

Campylobacter species are the leading cause of foodborne bacterial enteritis worldwide. Recently, extensively drug-resistant (XDR) and multi-drug-resistant (MDR) Campylobacter spp. have caused several global crises. Therefore, the present work aims to detect the prevalence and antimicrobial resistance patterns of Campylobacter spp. from various chicken sources in Egypt, and to investigate the efficacy of a mixture of eugenol and trans-cinnamaldehyde on the performance and immunity of challenged broilers and also to assess their effects on C. jejuni load and virulence gene expression in an in vivo model. Our results showed a high prevalence of campylobacter isolates (67.3%). Of note, 25.7 and 74.3% of campylobacter isolates were XDR and MDR, respectively. Interestingly, a mixture of eugenol and trans-cinnamaldehyde had significant enhancing and antimicrobial effects through improving the growth-performance variables, minimizing the C. jejuni fecal loads, and decreasing the C. jejuni virulence genes (flaA, virB11, and wlaN) expressions in broilers challenged with C. jejuni. Moreover, the mixture of eugenol and the trans-cinnamaldehyde had immunostimulant and anti-inflammatory activities. In conclusion, our findings suggest that the utilization of the mixture of eugenol and trans-cinnamaldehyde has a growth-promoting role and can be considered as a better replacement of the antimicrobial agents for the control and treatment of campylobacter infection in broiler chickens.

Campylobacter species (spp.) are one of the most important causes of human bacterial gastroenteritis in foods of animal origin. Recently, with the spread of multi-drug-resistant (MDR) and extensively drug-resistant (XDR) Campylobacter spp., natural alternative therapeutic methods are urgently required. Phytogenic active principles have gained considerable attention due to their proficiency to enhance gut health and, thereby, performance of broiler chickens. Thus, the current study aims to determine the prevalence and antimicrobial resistance of Campylobacter spp. of different chicken sources in Sharkia Governorate, Egypt, and to assess the growth-promoting, immunostimulant and antimicrobial effects of a mixture of eugenol and trans-cinnamaldehyde in an in vivo approach. A total of 101 (67.3%) campylobacter isolates was identified, according to both phenotypic and genotypic techniques. Moreover, all of the campylobacter isolates were resistant to erythromycin, trimethoprim/sulfamethoxazole, and ampicillin (100% each). Of note, a dietary supplementation of the mixture of eugenol and trans-cinnamaldehyde led to a significant improvement of the feed conversion ratio and body weight gain and a decrease in the cecal C. jejuni loads in the broilers challenged with XDR C. jejuni. Additionally, eugenol and the trans-cinnamaldehyde mixture had protective activities via the down-regulation of XDR C. jejuni (flaA, virB11 and wlaN) virulence genes and proinflammatory cytokines (TNF-α, IL-2, IL-6, and IL-8), and the up-regulation of anti-inflammatory cytokine IL-10. Thus, we recommend the usage of a mixture of eugenol and trans-cinnamaldehyde as an alternative to antimicrobials for the control and treatment of campylobacter infections.

1. IntroductionFor more than a century, there has been an evolution in the knowledge about the public health importance of human campylobacter infection. Campylobacteriosis is a global foodborne bacterial gastrointestinal infection and the majority of human campylobacter infection cases are caused by Campylobacter jejuni (C. jejuni) and, to a lesser extent, Campylobacter coli (C. coli) [1]. Since 2005, Campylobacter species (spp.) have been recognized as the most common cause of human foodborne enteritis in the European Union [2,3]. Furthermore, campylobacter infection is hyperendemic, mainly in young children and infants, in developing countries [4,5]. Campylobacter spp. are ubiquitous microorganisms and they are commensal bacteria in the gastrointestinal tract of poultry and domestic animals. Human campylobacteriosis cases usually occur by the consumption of contaminated water, raw or uncooked meat, mainly poultry and its products and unpasteurized milk [2,6].The majority of campylobacter infection cases are self-limiting and antimicrobial therapy is not indicated; however, antimicrobial treatment may be necessary under specific clinical conditions, such as immunocompromised individuals and patients with persistent or severe symptoms, prolonged or intense enteritis, extraintestinal infections, and bacteremia [2,4]. In these specific circumstances, the treatment may be complicated because of the emergence of multi-drug-resistant (MDR) and extensively drug-resistant (XDR) Campylobacter spp., due to the excessive uncontrolled utilization of antimicrobials in the poultry industry, livestock breeding, and veterinary medicine [7,8]. The drugs of choice in the treatment of human campylobacter infections are erythromycin (macrolides) and ciprofloxacin (fluoroquinolones (FQs)) [9]. Recently, Campylobacter spp. has become resistant to multiple antimicrobial classes, such as macrolides, FQs, aminoglycosides, tetracyclines, and beta-lactams, which result in increasing the infections with MDR and XDR Campylobacter spp. [10].For public health significance, it is important to identify the pathogenicity markers in campylobacter isolates, which are detected in food. Campylobacter spp. have several virulence factors, such as adhesions, acid resistance, heat and cold stress resistance [11], toxin production, hemolysin, capsule [12] and the flagella, and its secreted proteins [13]. The Flagellin A (flaA) gene is an important virulence marker in Campylobacter spp., which is responsible for the flagella formation and, consequently, bacterial adhesion, invasion, and motility [14]. Additionally, virB11, a plasmid-encoded gene, is correlated to the host cell invasion [5] and the wlaN gene is associated with the production of lipo-oligosaccharide and a β-1,3 galactosyltransferase. The wlaN gene products imitate the myelin sheath ganglioside structure on nerve cells, which result in the development of Guillain–Barré syndrome, an acute peripheral polyneuropathy, after campylobacter infection [15]. Chickens have been identified as the main reservoir of Campylobacter spp., and it is responsible for up to 80% of human campylobacteriosis cases. Human infection usually occurs through the handling and consumption of chicken meat and its products that are contaminated during the slaughtering and processing of carcasses [4,15]. Therefore, there is an important demand for effective protocols to control Campylobacter spp. at the farm level through minimizing the prevalence of Campylobacter spp. in poultry products, which will result in reducing the contamination of products and the prevalence of campylobacteriosis in humans [16]. Additionally, reducing campylobacter colonization, adhesion, and invasion of the intestinal epithelial cells through minimizing the production of their virulence genes and enhancing the immune response, might control the campylobacteriosis in humans [17].Recent studies reported the capability of natural antimicrobials in controlling Campylobacter spp. in chickens, due to the consumers’ increasing demands for safe and natural products that use the least preservatives [18]. Herbal plant extracts, such as phytochemicals, have generally been used since ancient times as food flavoring agents, dietary supplements, and food preservatives to avoid the spoilage of food and for public health improvement. Furthermore, phytochemicals, such as essential oils (EOs), have antimicrobial characteristics [1,17] and are capable of modifying the pro- and anti-inflammatory cytokines and virulence genes’ expressions [19,20]; thus, they are used in herbal medicine for the treatment of various diseases [17]. Additionally, EOs can be utilized in poultry nutrition as feed additives because they enhance poultry growth performance, feed efficiency parameters, and meat quality by improving digestibility [19,20]. Eugenol is a polyphenol complex, which is considered as the primary antimicrobial active ingredient found in clove (Syzgium aromaticum) essential oil, whereas trans-cinnamaldehyde, an aldehyde found in the bark of the cinnamon (Cinnamomum zeylandicum) tree, is the principal component of cinnamon EO. The aforementioned EOs are classified by the Food and Drug Administration (FDA) under “Everything Added to Food in the United States” and as Generally Recognized as Safe (GRAS) with fast biodegradation and least cytotoxicity, and they can be used in food as good substitutes for the antimicrobial agents [17,21].As the problem of resistant foodborne bacteria rises, infection with MDR pathogens will be substituted by XDR strains. Thus, we predict the worldwide spread of XDR foodborne bacteria, especially Campylobacter spp., soon. Eugenol and trans-cinnamaldehyde were reported as effective therapeutic options; however, to our best knowledge, no reports evaluated their activity against multi-virulent XDR Campylobacter spp. in vivo. Moreover, when a blend is utilized, the active principles might have synergistic effects influencing their modes of actions [22]. Hence, the present study aims to (i) detect the prevalence and antimicrobial resistance patterns of Campylobacter spp. from various chicken sources in Egypt; (ii) determine the virulence profiles of XDR campylobacter isolates; and (iii) assess, for the first time, the in vivo efficacy of a mixture of eugenol and trans-cinnamaldehyde, in comparison to the most susceptible examined antibiotics on the growth performance and expression of cytokines-related genes in broiler chickens experimentally infected with XDR and multi-virulent C. jejuni, in addition to investigating their activities on the viability and expression of the virulence genes of the infecting C. jejuni strain.2. Materials and Methods2.1. Ethical StatementThe experimental protocols were conducted following the regulations and approved guidelines of the Institutional Animal Care and Use Committee of Faculty of Veterinary Medicine, Zagazig University, Egypt under the reference number of ZU-IACUC/2/F/263/2021.2.2. Sample CollectionA total of 150 different samples were collected from recently slaughtered broiler chickens (Ross 308) of 6 weeks of age during the period ranging from August 2019 to October 2020 (14 months) from different areas in Zagazig city, Sharkia Governorate, Egypt. A total of 11 samples were obtained per month from the 11 main chicken processing plants in Zagazig city, Egypt (n = 11, 14 samples from each outlet), including chicken luncheon meats, liver, breast meats, cecal parts, and cloacal swabs (n = 30 each), and 1 site was randomly tested per bird. Each sample represented a single bird (the samples were completely independent and randomly picked). The cloacal swabs and 25 g of each sample were transported directly into 225 mL of supplemented Bolton broth (Oxoid, Cambridge, UK), leaving a headspace of about 20 mm in the tightly capped tubes to generate microaerophilic conditions [23]. The obtained samples were aseptically transferred into an icebox and to the laboratory as soon as possible for the isolation and identification of Campylobacter spp.2.3. Isolation and Identification of Campylobacter SpeciesFor Campylobacter spp. isolation, the obtained specimen in Bolton broth was incubated at 42 °C/24–48 h in darkness in a microaerophilic atmosphere (85% N2, 10% CO2, and 5% O2) utilizing an anaerobic jar (Sigma-Aldrich, St. Louis, MO, USA) and CampyGen sachets (Oxoid, Cambridge, UK). Subsequently, 0.1 mL of the inoculated Bolton broth was inoculated onto the surface of supplemented modified charcoal cefoperazone deoxycholate agar (mCCDA) plates (Oxoid, Cambridge, UK), then the inoculated plates were incubated at 42 °C/48 h in a microaerophilic atmosphere. For further purification, the suspected campylobacter colonies were cultivated on 5% sheep blood agar plates (Oxoid, Cambridge, UK), then incubated at 42 °C/48 h in a microaerophilic atmosphere. Next, the suspected colonies were presumably identified through their culture characters on mCCDA and blood agar, Gram’s staining, motility test, susceptibility to nalidixic acid and cephalothin, and biochemical identification procedures, including oxidase, catalase, rapid sodium hippurate, and indoxyl acetate hydrolysis [23].2.4. Antimicrobial Susceptibility TestingThe standard Kirby–Bauer disc diffusion method [24] was utilized to test the susceptibility of all the recovered campylobacter isolates to 18 antimicrobials. Briefly, a few (3–10) colonies from a single sample were suspended in sterile physiological saline and compared to the 0.5 McFarland standard solution. Then, Mueller–Hinton agar (Oxoid, Cambridge, UK) plates contained 5% of defibrinated sheep blood were cultivated with the prepared suspension. Next, the antimicrobial discs were placed onto the surface of dried plates and incubated at 42 °C/48 h in a microaerophilic atmosphere. A total of 18 antimicrobial discs (Oxoid, UK) that fall into 10 various antimicrobial categories, which were usually utilized in veterinary and human medicine in Egypt, were used: chloramphenicol (C, 30 µg); clindamycin (DA, 2 µg); colistin (CT, 10 µg); linezolid (LNZ, 30 µg); tobramycin (TOB, 10 µg); gentamicin (CN, 10 µg); amikacin (AK, 30 µg); azithromycin (AZM, 30 µg); erythromycin (E, 15 µg); doxycycline (DO, 30 µg); trimethoprim-sulfamethoxazole (SXT, 1.25 + 23.75µg); nalidixic acid (NA, 30 µg); aztreonam (ATM, 30 µg); imipenem (IMP, 10 µg); cefoxitin (FOX, 30 µg); sulbactam-ampicillin (SAM, 10 + 10 µg); ampicillin (AM, 10 µg); and ciprofloxacin (CIP, 5 µg). The susceptibility of the examined isolates was measured via the determination of the diameter of the inhibition zone of the antimicrobial discs, and the results were explained following the guidelines of the Clinical and Laboratory Standards Institute (CLSI) to classify each antimicrobial agent as either susceptible, intermediate, or resistant [25]. The XDR was defined as the non-susceptibility of an isolate to all antimicrobial agents, except two or fewer antimicrobial classes; meanwhile, MDR was described as the non-susceptibility of campylobacter isolates to at least one antimicrobial agent in three or more unrelated antimicrobial classes [26]. The multiple antibiotic resistance (MAR) indices were determined for the campylobacter isolates through the following equation: MAR = a/b, where (a) represents the number of antimicrobials to which the examined isolates were resistant and (b) is the total number of antimicrobial agents utilized [27].2.5. Conventional PCR AssayThe PCR assays conducted in the current work were carried out on the highly resistant campylobacter isolates. The DNA extraction was performed utilizing the QIAamp DNA Mini Kit (Qiagen, Germantown, MD, USA), following the manufacturer’s instructions. Conventional PCR amplification procedures were carried out for the detection of the 23S rRNA, ceuE, and mapA genes of genus Campylobacter, C. coli, and C. jejuni, respectively. Additionally, three significant virulence genes (wlaN, virB11, and flaA) were also detected via PCR assays. All PCR protocols were performed, in triplicate, using the Emerald Amp GT PCR Master Mix (Takara, Mountain View, CA, USA), following the instructions of the manufacturer. The primer sequences for the tested genes in all the PCR procedures are presented in Table 1. All protocols of the PCR amplification were carried out as previously described [28,29,30]. Subsequently, ethidium bromide staining (Sigma-Aldrich, St. Louis, MO, USA) and agarose gel electrophoresis for the PCR products’ visualization were performed [5,31]. In all the PCR procedures, negative controls (no template DNA) were PCR grade water and positive controls were the reference strains of C. coli (NCTC11366) and C. jejuni (NCTC11322).2.6. In Vivo Assessment of the Efficacy of the Eugenol and Trans-Cinnamaldehyde MixtureOne XDR and multi-virulent C. jejuni strain was selected for a further challenge experiment to assess the efficacy of the mixture of eugenol and trans-cinnamaldehyde, in comparison to the most effective examined antibiotic on the growth performance and the expression of the cytokines-related genes of broilers, and to also investigate their activities on the viability and expression of the virulence genes of the infecting strain.2.6.1. Plant-Derived AntimicrobialsThe eugenol and trans-cinnamaldehyde used in the current experiment were purchased from Sigma-Aldrich (St. Louis, MO, USA). Eugenol is a polyphenol complex that is considered as the primary antimicrobial compound found in clove EO, whereas trans-cinnamaldehyde is an aldehyde found in the bark of the cinnamon tree and it is the main component of cinnamon EO.2.6.2. Experimental Infection by XDR and the Multi-Virulent C. jejuni StrainAn XDR and multi-virulent C. jejuni strain were used in the current experimental trial, in which the challenge was conducted orally at 23 days of age via the crop gavage with 1 mL of 108 CFU/mL of the bacterial inoculum. The bacterial infection was checked via the re-isolation and identification of the infecting strain. Additionally, the re-examination of its antimicrobial resistance and virulence genes profiles were carried out to make sure that the isolated strain corresponded to the infecting one.2.6.3. Experimental Broiler Chickens, Design, and Feeding RegimeThis experimental trial was conducted on 400 1-day-old broiler chicks (Ross 308) that were purchased from a local commercial poultry hatchery. On arrival, the chicks were weighed separately and randomly assigned into 4 groups in floor pens (100 birds in each pen, with 5 replicates/group and 20 chicks/replicate). The chicks in the negative control (NC) group were kept unchallenged, while the chicks in the positive control (PC) group were challenged. The chicks in the NC and PC groups were fed the basal diet without any supplementation. The PC and the other 2 treatment groups were challenged with XDR and a multi-virulent C. jejuni strain at 23 days of age. The EOs treatment group was fed a basal diet supplemented with a mixture of eugenol and trans-cinnamaldehyde (1:1) at a concentration of 400 mg/kg each from the first day of life as a prophylactic. Moreover, in the antibiotic treated group, all the birds were treated intramuscularly with 50 mg/kg of cefoxitin after the appearance of clinical signs (lethargy, depression, and decreased body weight) and re-isolation of the bacterium 3 days post infection. All chicks were kept under completely hygienic environments, according to the Ross Broiler Management Guide [33]. All the chicks were permitted free access to feed and drinking water throughout the 37-day experimental period. All the birds were fed coccidiostat- and antibiotic-free diets in the mash form for the starter (days 1–10), grower (days 11–20), and finisher (days 21–37) periods, according to the Ross broiler nutrition specifications [33], as shown in Table 2. All the feed constituents and diets were examined chemically for the determination of crude fiber, crude protein, moisture, and ether extract, and these tests were in accordance with the Association of Official Analytical Chemists [34].2.6.4. Monitoring the Growth Performance of the BroilersThe body weight (BW) and feed intake (FI) were recorded for calculating the cumulative body weight gain (BWG) and feed conversion ratio (FCR) over the entire experimental period (1–37 days), as previously described [35,36,37,38].2.6.5. SamplingA total of 5 chicks from each replicate were slaughtered and sacrificed, and the cecal contents from each chick were aseptically removed and kept frozen at −80 °C in sterile tubes for the re-isolation and identification of the infecting C. jejuni strain and further quantification of the campylobacter populations through a quantitative real-time PCR (qPCR) assay and the analysis of the mRNA expression of campylobacter virulence genes via the reverse transcription quantitative polymerase chain reaction (RT-qPCR) assay at 30 (7 days post-infection, dpi) and 37 (14 dpi) days of age. Additionally, the spleen was rinsed with sterile phosphate-buffered saline and used for investigating the expression of cytokine-related genes using RT-qPCR assay 7 and 14 dpi.2.6.6. Gene Expression Analysis by the Reverse Transcription Quantitative PCR AssaySplenic tissues were utilized for detecting the mRNA expression levels of cytokines-related genes (tumor necrosis factor-alpha (TNF-α), interleukin (IL)-2, IL-6, IL-8, and IL-10), and the cecal contents were utilized for the subsequent investigation of the expression levels of campylobacter virulence genes (flaA, virB11, and wlaN). The total RNA was extracted utilizing the QIAamp RNeasy Mini kit (Qiagen, Hilden, Germany), according to the instructions of the manufacturer. The concentration of the extracted RNA was determined at 260 nm and the RNA clarity was determined utilizing a Spectrostar NanoDropTM 2000 spectrophotometer (Thermo Fisher, Sunnyvale, CA, USA). A one-step RT-qPCR assay was conducted, in triplicates, utilizing a QuantiTect SYBR Green RT-PCR Kit (Qiagen, Hilden, Germany) via the Strata-gene MX3005P real-time PCR identification system (Thermo Fisher, Sunnyvale, CA, USA), following the guidelines of the manufacturer. A melting curve analysis was used for the verification of all PCR amplifications. The transcript’s expression levels were normalized using glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and campylobacter 23S rRNA genes as endogenous controls. The primer sequences of the virulence and cytokines-related genes used in the RT-qPCR assays are shown in Table 1 and Table 3, respectively. The relative mRNA expression data of the tested genes were assessed using the 2−∆∆Ct method [39].2.6.7. Quantification of the Campylobacter DNA Copies by Quantitative Real-Time PCR AssayDNA was extracted from the cecal contents using a QIAamp DNA stool Mini Kit (Qiagen, Hilden, Germany), according to the recommendations of the manufacturer. The absolute quantification of the campylobacter populations was conducted via qPCR assays, in triplicate, utilizing the Stratagene MX3005P RT-PCR machine and QuantiTect SYBR Green PCR Master Mix (Qiagen, Hilden, Germany), following the guidelines of the manufacturer. The primer sequence of the 23S rRNA gene is shown in Table 1. Ten-fold serial dilutions of the extracted DNA from pure campylobacter cultures were conducted to generate the standard calibration curves for the qPCR. The number of target genomic DNA copies were measured and the campylobacter concentration was expressed as the log10 CFU/gram of the cecal content.2.7. Statistical AnalysisThe results were examined statistically via the SPSS Inc. software version 26 (IBM Corp., Armonk, NY, USA). The chi-squared test was used to analyze the variations in the incidence of Campylobacter spp. from various chicken sources, and to assess the variations in the antimicrobial resistance profiles and the prevalence of virulence genes in the recovered isolates from different sources and between C. jejuni and C. coli isolates. Additionally, the homogeneity and normality between the experimental groups were conducted using Levene’s and Shapiro–Wilk’s tests, respectively. The differences between the data of the experimental trials were expressed as the standard error of the mean (SEM), and the ANOVA and Tukey’s test were used to assess the significant differences between the mean values. If the p-value was less than 0.05, it was considered statistically significant. All graphs were produced utilizing the GraphPad Prism software Version 8 (San Diego, CA, USA).3. Results3.1. Prevalence of the Campylobacter Species in the Broiler Chickens in Sharkia Governorate, EgyptAccording to the phenotypic identification results, a total of 101 (67.3%) campylobacter isolates were recovered from 150 various samples obtained from broiler chickens in Sharkia Governorate, Egypt. The prevalence of campylobacter isolates in the recovered specimens is presented in Table 4. Campylobacter spp. were highly distributed between chicken cloacal swabs (90%) and cecal parts (86.7%), while chicken luncheon meats were campylobacter negative (Table 4). Furthermore, C. jejuni was the most common Campylobacter spp. (54%), followed by C. coli (13.3%). Campylobacter jejuni was highly distributed among chicken cecal parts and cloacal swabs (70% each), followed by breast meats (66.7%). Meanwhile, C. coli isolates were more prevalent in the chicken livers and cloacal swabs (20% each), followed by cecal parts (16.7%) (Table 4). Additionally, there were statistically significant variations (p ˂ 0.001) in the incidences of Campylobacter spp., C. coli, and C. jejuni between the different sample origins.3.2. Antimicrobial Susceptibility Testing of the Campylobacter SpeciesAll 101 recovered campylobacter isolates were examined for their susceptibility against the examined 18 antimicrobials, as presented in Table 5. Interestingly, the examined campylobacter isolates were completely resistant to trimethoprim/sulfamethoxazole erythromycin and ampicillin (100% each). Additionally, the majority of campylobacter isolates were resistant to clindamycin and nalidixic acid (95% each), followed by azithromycin (91.1%), aztreonam (84.2%), and doxycycline (83.2%). Meanwhile, the lowest resistance rates were detected against cefoxitin (33.7%), amikacin (34.7%), and imipenem (42.6%) (Table 5).Regarding the isolates’ sources, the antimicrobial resistance patterns of the campylobacter isolates from various chicken sources were different (Table 5). It was observed that all the campylobacter isolates from the chicken liver and breast meat samples were 100% resistant to nalidixic acid, while the campylobacter isolates from the chicken cloacal swabs were 100% resistant to clindamycin. Additionally, statistical significance variations were observed in the resistance patterns between the campylobacter isolates from the different chicken sources to nalidixic acid, aztreonam, and cefoxitin (p = 0.03, 0.018, and 0.048, respectively). Additionally, there was a higher statistically significant difference in the linezolid resistance rates among the campylobacter isolates from the different chicken sources (p = 0.001) (Table 5). Meanwhile, there were no statistically significant differences (p ˃ 0.05) in the resistance percentages among the campylobacter isolates from the different chicken sources to the other examined antimicrobial agents.According to the species level, our findings present that C. jejuni isolates are more resistant than C. coli isolates to the tested antimicrobial agents, except for clindamycin, colistin, linezolid, doxycycline, and cefoxitin (Figure 1). Moreover, statistical significance variations were detected in the resistance rates among the C. coli and C. jejuni isolates to doxycycline and sulbactam-ampicillin (p = 0.021 and 0.012, respectively). Additionally, higher statistical significance differences were observed in the resistance rates between the C. jejuni and C. coli isolates to cefoxitin, aztreonam, and nalidixic acid, (p = 0.008, 0.003, and 0.005, respectively). However, no statistical significance variations (p ˃ 0.05) were observed in the resistance patterns among the C. coli and C. jejuni isolates to the other investigated antimicrobials (Figure 1).Interestingly, it was found that the C. jejuni isolates were more resistant to 8 and 9 antimicrobial categories (18.5 and 35.8%, respectively) than the C. coli isolates (10 and 30%, respectively). Meanwhile, the resistances to 6, 7, and 10 antimicrobial classes were higher in C. coli isolates (10, 10, and 40%, respectively) than the C. jejuni isolates (1.2%, 9.9%, and 34.6%, respectively). In total, the resistance rates to 10 antimicrobial classes were more prevalent among the campylobacter isolates from the chicken liver samples (60%). Meanwhile, the resistance patterns to 7 and 9 antimicrobials categories were more prevalent among the campylobacter isolates from the chicken cecal parts (15.3 and 42.3%, respectively). Moreover, the resistance proportions to the 8 antimicrobial classes were the highest among the campylobacter isolates from the chicken breast meats (26.1%) and those to the 6 classes were the highest among the campylobacter isolates from the chicken cloacal swabs (11.1%) (Figure 2).Of note, our results reveal that 26 campylobacter isolates (25.7%) were XDR: 12 (48%), 6 (26.1%), 5 (19.2%), and 3 (11.1%) were obtained from chicken liver, breast meats, cecal parts, and cloacal swabs, respectively; meanwhile, 75 campylobacter isolates (74.3%) were recognized as MDR. Interestingly, our results present that all the tested campylobacter isolates showed an MAR index ≥0.39 (Table 6), which represent high-risk contamination sources, where the antimicrobials are frequently utilized. A total of 8 campylobacter isolates had MAR indices >0.9 (resistance to 17 antimicrobial agents); 2 (8%), 3 (13%), 1 (3.8%), and 2 (7.4%) were isolated from chicken liver, breast meats, cecal parts, and cloacal swabs, respectively (Table 6). Moreover, statistical significance variations were observed in the resistance profiles between the campylobacter isolates from various chicken sources to 16 antimicrobial agents (p = 0.006) and to 6 and 10 antimicrobial classes (p = 0.037 and 0.017, respectively) (Figure 2A). Additionally, statistical significance variations were observed in the resistance patterns among the C. jejuni and C. coli isolates to 8 and 15 antimicrobial agents (p = 0.038 and 0.024, respectively) (Figure 2B).3.3. Molecular Grouping of the Campylobacter Species from Various Chicken SourcesA total of 26 XDR campylobacter isolates (resistant to 16 and 17 antimicrobials) were submitted for conventional PCR assays. Those isolates were recovered from the chicken liver (12), breast meats (6), cecal parts (5), and cloacal swabs (3). All the 26 screened campylobacter isolates (100%) were identified as genus Campylobacter by the PCR detection of the 23S rRNA gene. Additionally, 19 isolates (73.1%) and 7 isolates (26.9%) were positive for mapA and ceuE genes, respectively, and confirmed to be C. jejuni and C. coli, respectively. Of the 19 C. jejuni isolates, 8 (42.1%), 4 (21.1%), 5 (26.3%), and 2 (10.5%) were recovered from chicken liver, breast meats, cecal parts, and cloacal swabs, respectively. Moreover, 7 C. coli isolates were recovered from 4 (57.1%) chicken liver, 2 (28.6%) breast meat, and 1 (14.3%) cloacal swab samples. These findings were consistent with those of the phenotypic identification techniques. Additionally, statistically significant variations (p = 0.008) were observed in the incidences of the C. coli and C. jejuni isolates between the samples of the chicken cecal parts. Meanwhile, no statistical significance differences were observed in the incidence of the C. coli and C. jejuni isolates between the cloacal swabs, breast meat, and chicken liver samples (p = 1, 0.567, and 0.22, respectively) (Figure 3).3.4. Molecular Investigation of the Virulence-Related Genes among the Investigated IsolatesAll 26 molecularly confirmed campylobacter isolates were examined for the presence of three significant virulence genes, which have fundamental roles in the pathogenesis of Campylobacter spp. (flaA, virB11, and wlaN). Of the 26 examined isolates, 26 (100%) were positive for the flaA gene, 13 (50%) were positive for the virB11 gene, and 9 (34.6%) were positive for the wlaN gene. Of note, four virulence gene profiles were detected among the tested campylobacter isolates (Figure 4). A total of 12 XDR campylobacter isolates (46.2%) revealed the most frequent virulence gene profile (flaA).Regarding the isolates’ sources, it was found that the virB11 gene was more prevalent among the campylobacter isolates from the chicken cloacal swabs (66.7%) and cecal parts (60%). Meanwhile, the wlaN gene was more prevalent among the campylobacter isolates from the chicken cecal parts (60%), followed by the cloacal swabs and breast meats (33.3% each). Moreover, 14 (53.8%) campylobacter isolates contained at least 2 virulence genes; 5 (41.7%), 4 (66.7%), 3 (60%), and 2 (66.7%) were recovered from the chicken liver, breast meats, cecal parts, and cloacal swabs, respectively. Additionally, 8 (30.8%) campylobacter isolates harbored 3 investigated virulence genes; 3 (25%), 1 (16.7%), 3 (60%), and 1 (33.3%) were recovered from the chicken liver, breast meats, cecal parts, and cloacal swabs, respectively. There was a statistically significant difference (p = 0.048) in the virulence profile I of the campylobacter isolates among the different chicken sources (Figure 4).According to the spp. level, it was found that the virB11 gene was more prevalent among the C. coli isolates (71.4%), while the wlaN gene was more prevalent among the C. jejuni isolates (36.8%). Moreover, 5 C. coli isolates (71.4%) contained at least 2 virulence genes. Meanwhile, 6 C. jejuni isolates (31.6%) harbored 3 investigated virulence genes. Additionally, no statistical significance variations (p ˃ 0.05) were observed in the virulence gene profiles between the C. coli and C. jejuni isolates (Figure 4).3.5. In Vivo Experimental StudyAn identified XDR and multi-virulent C. jejuni strain, which was resistant to 17 tested antimicrobials, sensitive only to cefoxitin antibiotic and exhibited the virulence profile I (flaA, virB11, and wlaN), was used as a challenge strain in our in vivo experiment to assess the growth promoting, immunostimulant, antibacterial, anti-virulence, and activities of the eugenol and trans-cinnamaldehyde mixture.3.5.1. Growth PerformanceThe growth performance variables throughout the experimental period are presented in Table 7. Significant differences were observed between the various experimental groups over the entire rearing period. The C. jejuni challenge significantly (p < 0.05) reduced the BW and BWG and increased the FI and FCR in the PC group, compared to the challenged groups treated with the mixture of eugenol and trans-cinnamaldehyde and cefoxitin. Interestingly, the FCR was most significantly improved (p < 0.05) in the NC and the mixture of eugenol and trans-cinnamaldehyde-treated groups, followed by the cefoxitin-treated group, with no significant differences between the NC and the mixture of eugenol and trans-cinnamaldehyde-treated groups. Meanwhile, chicks in the NC, cefoxitin, and the mixture of eugenol and trans-cinnamaldehyde-treated groups exhibited a significant (p < 0.05) increased BWG over the entire rearing period, unlike the PC group.3.5.2. Analysis of Cytokines Genes ExpressionThe data of the cytokines genes expression analysis are shown in Figure 5. At 7 and 14 dpi, the relative expression levels of the TNF-α gene were significantly decreased (p < 0.05) in the mixture of the eugenol and trans-cinnamaldehyde-treated group, followed by the cefoxitin-treated group, in comparison to the PC group, and there was no significant difference between the NC and mixture of the eugenol and trans-cinnamaldehyde-treated groups. Moreover, the most marked reduction in the IL-2 expression level was observed in the mixture of the eugenol and trans-cinnamaldehyde-treated group, in respect to the PC. Meanwhile, the relative expression levels of the IL-6 and IL-8 genes were decreased in the cefoxitin-treated group, followed by the mixture of the eugenol and trans-cinnamaldehyde-treated group. Interestingly, no statistical significance variations were observed among the expression levels of the IL-6 and IL-8 genes among the NC, cefoxitin, and eugenol and trans-cinnamaldehyde mixture-treated groups at 14 dpi. Additionally, the IL-10 gene expression level was significantly upregulated (p < 0.05) in the group treated with the mixture of eugenol and trans-cinnamaldehyde, compared to the PC group, at both time points.3.5.3. Expression Analysis of the C. jejuni Virulence Genes via the RT-qPCR AssayThe expression levels of the C. jejuni flaA, virB11, and wlaN virulence genes after treatment with the mixture of eugenol and trans-cinnamaldehyde and cefoxitin at 7 and 14 dpi are shown in Figure 6. The flaA, virB11, and wlaN mRNA expression levels were significantly (p < 0.05) downregulated in the cefoxitin, followed by the mixture of eugenol and trans-cinnamaldehyde-treated groups, compared to the PC group at both time points. Interestingly, no statistical significance variations were observed among the expression levels of C. jejuni flaA, virB11, the and wlaN genes between the cefoxitin- and mixture of eugenol and trans-cinnamaldehyde-treated groups at 14 dpi. Additionally, the flaA relative expression level was significantly (p < 0.05) decreased in the chicks treated with cefoxitin and the mixture of eugenol and trans-cinnamaldehyde at 7 dpi, with respect to the PC group, with no statistically significant differences between the latter groups.3.5.4. Quantification of the C. jejuni DNA CopiesThe quantification data of C. jejuni in the cecal contents of chicks are presented in Figure 7. The most significant (p < 0.05) reduction in C. jejuni loads was detected in the chicks treated with cefoxitin, followed by the mixture of eugenol and trans-cinnamaldehyde, compared to the PC group at 7 days post-challenge with XDR C. jejuni. Interestingly, no statistical significance variations were observed among the log10 copies of the C. jejuni populations in the cecal contents of the chicks treated with the mixture of eugenol and trans-cinnamaldehyde and cefoxitin at 14 dpi.4. DiscussionIt was stated that multiple worldwide crises were established as a result of the extensive spread of antimicrobial-resistant bacteria, including methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant Staphylococcus aureus (VRSA), Klebsiella spp., Mycoplasma spp., as well as zoonotic foodborne bacteria, such as Campylobacter spp., Salmonella Enteritidis, Salmonella Typhimurium, and E. coli [40,41,42,43,44,45,46,47,48]. Campylobacter spp., as commensal bacteria in the intestinal tract of domestic animals and chickens, are recognized as the primary sources of campylobacter infections in humans. Recently, Campylobacter spp. have become resistant to several antimicrobials, especially macrolides, FQ, and tetracyclines being important public health problems around the world. In the current work, we determined the high incidence of campylobacter isolates (67.3%) between different chicken specimens in Sharkia Governorate, Egypt. Our results are higher than those observed in previous studies carried out in Egypt, (7.6%) [49], (20.3%) [50], (26.9%) [51], and (32.8%) [7], meanwhile, these findings are consistent with the results of a recent study carried out in Kenya (69%) [52]. In the current work, Campylobacter spp. were more prevalent between the cloacal swabs (90%). These results are higher than those observed in India (71%) [53], Kenya (69%) [52], and Egypt (54.3%) [7]. In the present study, Campylobacter spp. were not detected in the chicken luncheon meats (0%), which is similar to the results of a previous study conducted in Egypt [54]. Additionally, C. jejuni was the most commonly recognized spp. (54%), which is consistent with the findings of previous studies carried out in Tunisia (68.9%) [55], and China; (20.3%) [56] and (12.3%) [12], respectively. Generally, the differences in the prevalence of Campylobacter spp. among the different studies might be correlated to the contamination and health conditions, climate factors, the geographical locations, the sources of tested samples, and the conventional identification methods [57].Unsurprisingly, there are differences in the antibiotic resistance profiles between different countries, as a result of the variations in the prescribed antibiotics. Of note, our examined isolates showed complete resistance to ampicillin and erythromycin (100% each). These results are higher than those observed in other studies conducted in South Korea (57.4 and 14.9%, respectively) [58] and Pakistan (15 and 10%, respectively) [59]. Herein, high resistance levels to doxycycline and ciprofloxacin were observed among our examined isolates (83.2 and 79.2%, respectively); these results were lower than those reported in a previous study conducted in Tunisia (100 and 99.2%, respectively) [55]. Alarmingly, all global warnings are related to the emergence of MDR strains, but, recently, the XDR spread is taking concrete steps. In the current study, 74.3 and 25.7% of the examined campylobacter isolates were identified as MDR and XDR, respectively. These results are consistent with those detected in a recent study carried out in Egypt, where 28.5 and 69% of the tested isolates were MDR and XDR [54]. Herein, our examined isolates had MAR indices ≥ 0.39. These results were higher than those observed in another study conducted in South Korea, where all the tested isolates had MAR indices ≤ 0.2 [58]. In developing countries, the high resistance rates of campylobacters could result from the excessive uncontrolled utilization of antimicrobial agents in animal and human treatments without any prescription and as growth promoters in veterinary medicine. Moreover, our findings show alarmingly high resistance prevalence of doxycycline, erythromycin, and ciprofloxacin because they are the antibiotics of choice for the treatment of human campylobacter infection resulting in significant problems, where antimicrobial treatment becomes limited. Thus, the utilization of antimicrobials should be controlled in humans and animals [4]. Thus, it is significant to utilize medicinal plants, such as phytochemicals, as an alternative to the antimicrobials [60,61,62].Out of the 26 molecularly investigated campylobacter isolates, 19 (73.1%) and 7 (26.9%) were positive for mapA and ceuE genes, respectively, being identified as C. jejuni and C. coli, respectively. Interestingly, there was a 100% correlation between the molecular and conventional identification results. These results were in agreement with a previous study conducted in Egypt, where 89.5% of the tested isolates were identified as C. jejuni, while the remaining isolates (10.5%) were identified as C. coli [63].Herein, 50% of the examined campylobacter isolates were positive for the virB11 gene. These results are in contrast with the results of another study conducted in South Korea, where all the examined isolates were negative to the virB11 gene [64]. These differences in the prevalence of the virB11 gene may be attributed to the isolates’ sources and samples’ types.Interestingly, our results reveal that all the examined campylobacter isolates (100%) are positive for the flaA gene. These findings are in complete agreement with a previous study conducted in Egypt [49], suggesting that the flaA gene has a significant virulence role in campylobacter isolates due to its significant part in the invasion, adhesion, and motility Campylobacter spp.In the current work, the dietary supplementation of a mixture of eugenol and trans-cinnamaldehyde was able to modulate the immune responses of broiler chickens to overcome experimental C. jejuni infection and enhance their growth performance. In the present study, the mixture of eugenol and trans-cinnamaldehyde enhanced the overall growth rate and restored the decreased body weight and impaired the FCR post-infection with XDR C. jejuni, in comparison to the PC group. In agreement with our findings, a recent study reported the growth-promoting effect of carvacrol in broiler chickens challenged with C. jejuni [65]; however, the efficacy of the mixture of eugenol and trans-cinnamaldehyde on the broiler chickens’ performance after the experimental infection with the XDR C. jejuni strain has not been studied until now. Additionally, another study reported that broiler chicks fed with various levels of EOs exhibited an improved FCR and BWG after experimental infection, when compared to the PC group [66]. Moreover, the dietary supplementation of plant-active principles had better growth performance variables and C. jejuni load in both intestinal content and excreta [67]. According to previous studies, the positive effects of EOs on broiler chickens’ growth performance might result from the capability of their bioactive compounds to enhance their immune functions [68], stimulate their appetite and antimicrobial characteristics [69]. However, the EO extracts’ mode of actions are not yet defined and could vary as a result of the different forms, sources, and structures of the active components being utilized in diets [66].Cytokines have significant regulatory roles in the inflammatory response of the intestinal tract. When the pathogens invade the gastrointestinal epithelial cells, the immune cells of the intestinal tract are triggered to produce cytokines that have fundamental roles in the host immune responses against microorganisms [70]. The TNF-α is a significant proinflammatory cytokine that regulates the host immune response against microorganisms via differentiating and proliferating the immune cells [71], but the excessive and prolonged secretion of proinflammatory cytokines might result in gastrointestinal damage [72]. Moreover, TNF-α, IL-2, IL-6, and IL-8 trigger an inflammatory response through recruiting the antimicrobial cells, including macrophages and neutrophils [71]. In the current study, the splenic proinflammatory cytokine (IL-2, IL-6, IL-8, and TNF-α) gene expression levels were upregulated in chicks that received a control basal diet and challenged experimentally with C. jejuni; however, the relative expression levels of IL-6 and IL-8 genes were significantly decreased in the cefoxitin-treated group, followed by the mixture of the eugenol and trans-cinnamaldehyde-supplemented group concerning the PC group. Additionally, supplementing our challenged birds with a mixture of eugenol and trans-cinnamaldehyde had regulatory effects on the proinflammatory cytokine genes expressions through a significant downregulation of TNF-α and IL-2 expression levels, which may counteract the inflammatory effects caused by C. jejuni, indicating their immunostimulant activities. This may result from the role of dietary EOs in improving the host non-specific immune response via the nonspecific killing of parasites, tumor cells, bacteria, and fungi, which lead to minimizing the microorganism loads [73]. Meanwhile, IL-10 has mainly opposed and complex effects on the immune and inflammatory responses via suppressing their actions [74]. In the current study, the highest expression level of the IL-10 gene was observed in the mixture of the eugenol and trans-cinnamaldehyde-supplemented group, which may minimize the excessive inflammatory response indicating their significant anti-inflammatory characteristics. In agreement with our results, recent studies showed the immunostimulant and anti-inflammatory effects of Eos via minimizing the expression levels of TNF-α [19], IL-2, and IL-6 [66] genes and decreasing that of the IL-10 gene [66] in broiler chicks challenged with Salmonella spp. due to their regulatory effects; however, the anti-inflammatory and immunostimulant effects of the mixture of eugenol and trans-cinnamaldehyde in broiler chickens challenged with the XDR C. jejuni strain has not yet been investigated.Targeting the virulence factors of microorganisms through utilizing novel anti-virulence treatments is a promising alternative method, which could be utilized to disarm the bacterial microorganisms [60]. Following the previous statement, our data show that the mixture of eugenol and trans-cinnamaldehyde downregulated the expression levels of C. jejuni flaA, wlaN and virB11 virulence genes in challenged broilers concerning the PC group. In agreement with our results, recent studies reported the in vivo anti-virulence effects of EOs via suppressing the virulence genes’ expressions of Salmonella spp. in challenged broiler chicks [19]. Previous studies were continuously reporting the modulatory effects of EOs on C. jejuni virulence genes’ expressions in vitro [7,17,18]; however, to the best of our knowledge, there are no studies reporting the in vivo efficacy of the mixture of eugenol and trans-cinnamaldehyde on C. jejuni virulence in challenged broiler chickens.Campylobacter spp. are commensal bacteria in the gastrointestinal tract of poultry and chickens, which considered the main reservoir for human campylobacter infections [1]. Herein, the quantitative analysis of cecal C. jejuni in challenged broilers showed that the dietary supplementation of a mixture of eugenol and trans-cinnamaldehyde significantly (p < 0.05) minimized C. jejuni populations 7 and 14 dpi, in comparison to the PC group. Our findings are in agreement with the results of previous studies, which demonstrate that the dietary supplementation of carvacrol significantly (p < 0.05) decreases C. jejuni counts in ceca [65], cloacal swabs, and the colon [75] of challenged broiler chickens, in comparison to the PC group. Additionally, a previous study reported a reduction in the cecal C. jejuni counts in challenged broilers after dietary supplementation with an EO mixture containing eugenol, indicating their inhibitory effects against C. jejuni [76]; however, the efficacy of the mixture of eugenol and trans-cinnamaldehyde on cecal C. jejuni loads in broiler chickens challenged with the XDR C. jejuni strain has not been explored to date.5. ConclusionsThe current work shows an alarming prevalence of avian XDR Campylobacter spp. in Egypt. The wlaN and virB11 genes were more prevalent among the examined campylobacter isolates from chicken cecal parts and cloacal swabs. Interestingly, our findings clearly prove the beneficial effects of a mixture of eugenol and trans-cinnamaldehyde on broiler chickens’ growth performance through decreasing the fecal load of C. jejuni and modulating the expression of proinflammatory and anti-inflammatory cytokines and decreasing the excessive inflammatory response. Furthermore, the administration of the mixture of eugenol and trans-cinnamaldehyde for broilers challenged with XDR C. jejuni had immunostimulant, anti-inflammatory, antimicrobial and anti-virulence effects. Therefore, our findings recommend the application of the mixture of eugenol and trans-cinnamaldehyde as an alternative to antimicrobials for the treatment and control of Campylobacter spp. infection in the veterinary fields, as well as for immunocompromised individuals.

animals : an open access journal from mdpi

10.3390/ani11072133

PMC8300223

Bovine respiratory disease (BRD) threatens cattle production and welfare globally. We sought to quantify the effects of vaccine treatments and animal temperament classification on feed intake, feeding behavior and weight responses following challenge with bovine viral diarrhea virus, one of the pathogens associated with bovine respiratory disease. Commercially available respiratory vaccines were utilized, and temperament classification was based on exit velocity. Although overt clinical signs of respiratory disease were not observed following challenge, feed intake, weight gain, feed efficiency, and feed bunk frequency and duration were negatively affected. Animals administered a modified-live vaccine had more desirable feed intake, feed bunk duration, longer meal events and slower eating rates compared with those administered a killed or no vaccine. Temperament affected feeding behavior patterns, where calm steers had a greater duration of feed bunk visits and meal events, and slower eating rates compared with excitable steers. There were greater differences due to vaccine treatments in most feeding behavior traits within calm vs. excitable steers. The modified live vaccine mitigated the negative effects of the viral challenge to a greater extent than the killed vaccine for feed intake and feeding behavior patterns, and corresponded with previously reported findings regarding the effects of these vaccine types on immune responses.

This study examined the effects of multivalent respiratory vaccine treatment (VT) and animal temperament classification on feeding behavior traits, feed intake and animal performance in response to a bovine viral diarrhea virus (BVDV) challenge. Nellore–Angus crossbred steers (n = 360; initial body weight (BW) 330 ± 48 kg) were assigned to one of three vaccine treatments: non-vaccinated (NON), modified live (MLV) and killed (KV) regarding respiratory viral pathogens, and inoculated intranasally with the same BVDV1b strain. Cattle temperament categories were based on exit velocity. Overt clinical signs of respiratory disease were not observed, yet the frequency and duration of bunk visit events as well as traditional performance traits decreased (p < 0.01) following BVDV challenge and then rebounded in compensatory fashion. The reduction in dry matter intake (DMI) was less (p < 0.05) for MLV-vaccinated steers, and MLV-vaccinated steers had longer (p < 0.01) durations of bunk visit and meal events and slower (p < 0.01) eating rates compared with KV- and non-vaccinated steers following BVDV challenge. Greater differences in most feeding behavior traits due to VT existed within calm vs. excitable steers. Respiratory vaccination can reduce the sub-clinical feeding behavior and performance effects of BVDV in cattle, and the same impacts may not occur across all temperament categories.

1. IntroductionAlthough multivalent vaccines for bovine respiratory disease (BRD) are widely used in US feedlots [1], BRD remains one of the most costly and prevalent diseases in the beef industry [1,2,3,4]. The efficacy of BRD vaccines is impacted by a number of management factors, including stressors associated with weaning, commingling and transportation [5,6], as well as animal temperament [7,8].Vaccine type also affects the degree of protection against BRD. In general, modified-live vaccines (MLV) have been shown to elicit more robust and longer-lasting immune responses [9,10] compared with killed vaccines (KV). Furthermore, modified live vaccination programs have been reported to reduce morbidity [11] and mortality rates [12], and lymphocytopenia [13] and pyrexia [14] compared with killed or no vaccine programs.Cattle with more excitable temperaments have been reported to have compromised immune functions compared with calm cattle, which may be due to stress-induced responses associated with elevated serum cortisol concentrations [15,16]. Steers with excitable temperament phenotypes had lower in vivo lymphocyte proliferation and lower in vivo vaccine-specific IgG concentrations [7], and bulls with excitable temperaments have been shown to have reduced innate immune responses [17]. These results would suggest that temperament may alter the magnitude of protection against BRD elicited by vaccines.There is limited research examining the effect of vaccine treatment (VT) and (or) the interaction of VT × temperament on feed intake, performance and feeding behavior responses in cattle following a disease challenge. Therefore, the objectives of this study were to examine the effects of multivalent VT for BRD and temperament classification on feed intake, performance and feeding behavior responses following a bovine viral diarrhea virus (BVDV) challenge in growing beef steers.2. Materials and Methods2.1. Animal and Experimental DesignAll animal procedures were reviewed and approved by the Texas A&M University Institutional Animal Care and Use Committee as well as the Texas A&M University Institutional Biosafety Committee (Animal Use Protocols #2010-08 and #2013-0069). The animals utilized in this study were half-blood (F2 and F3) Angus–Nellore steers (n = 360) from the Texas A&M University McGregor Genomics herd, which consists of a Bos taurus–Bos indicus crossbred population that was specifically developed to support genomic studies. Four trials were conducted during consecutive years from 2010 to 2013. The steers were born in the spring (mid-February to late April annually) and were not vaccinated against BRD pathogens as calves. Steers were weaned at approximately 7 mo of age and received 3 clostridial vaccinations with Closti Shield 7 (Novartis Animal Health US, Inc., Greensbro, NC, USA) at approximately 70 day of age, at 3 weeks prior to weaning and at weaning. Following weaning each year, calves were managed as a single group and remained on pasture or were fed a growing ration, depending on the year, until being transported 165 km from McGregor, TX, to College Station, TX, in January or February. Steers were confirmed to be BVDV-PI negative through evaluation of ear-notch samples by antigen capture ELISA and were seronegative for BVDV antibodies (Texas Veterinary Medical Diagnostic Laboratory; TVMDL, Amarillo, TX, USA). Throughout this study, low-stress cattle handling methods were emphasized during movement, processing and data collection.Animals were housed and managed as a single contemporary group each year following weaning until assignment to 1 of 4 pens at the Texas A&M University Beef Systems Research Unit (College Station, TX, USA). Each of the pens was equipped with 4 electronic feed bunks (GrowSafe System LTD., Airdrie, AB, Canada) with 20 to 26 steers per pen. A high-forage growing diet was used in this study that consisted of 31.5% corn, 36.5% chopped alfalfa, 24.5% dry distillers’ grains, 2.5% commercial premix and 5% molasses, which was formulated to meet the nutrient requirements of steers gaining 1 kg per day. The cattle were acclimated to the diet for 4 to 8 wk prior to the start of each year’s trial, and feed was delivered twice daily to ensure ab libitum access.2.2. Vaccination and Challenge ProtocolsAt approximately 12 mo of age, steers were stratified by sire and genomic cow families, and randomly assigned to 1 of 3 vaccine treatments that consisted of a killed virus (KV) vaccine (n = 119), a modified live virus (MLV) vaccine (n = 123, and no (NON) vaccine (n = 118). Both vaccines were labeled for protection against infectious bovine rhinotracheitis, parainfluenza-3, bovine respiratory syncytial virus and bovine viral diarrhea, and were administered according to the label directions. Steers assigned to the KV treatment received an initial vaccine dose (Vira-shield; Novartis Animal Health US, Inc., Greensbro, NC, USA) 56 or 49 d prior to BVDV challenge, with a second dose administered 21 d later. Steers assigned to the MLV treatment were vaccinated with a single dose of Arsenal 4.1 (Novartis Animal Health US, Inc., Greensbro, NC, USA) on the same day that the second KV dose was administered. The non-vaccinated steers did not receive a BRD vaccine or sham injection prior to BVDV challenge, but the non-vaccinated steers were handled similarly each time to the KV- and MLV-vaccinated steers. The MLV-vaccinated steers were isolated from the KV- and non-vaccinated steers for 7 to 10 d following vaccination, with an empty pen between to avoid animal nose-to-nose contact and prevent potential cross-contamination from virus shedding. Following this post-vaccination isolation, steers were comingled again prior to being assigned to their respective study pens. The vaccine treatments were balanced across study pens.All steers were challenged with the type 1b non-cytopathic BVDV strain CA0401186a that was obtained from the USDA-ARS National Animal Disease Center, Ames, IA [18]. This BVDV strain, originally isolated from a persistently infected BVDV Holstein calf, was selected for this study, as previous research demonstrated that it elicited typical immunological and clinical symptoms of morbidity, but with minimal risks of extreme illness or death [18], as is typical for most field strains of BVDV. Each steer was administered 5 mL of BVDV inoculum containing 1 × 105 TCID/mL intranasally (2.5 mL dose per nasal passage). Challenge dates (Day 0) were 11 May, 10 May, 15 May and 4 June for trial years 2010–2013, respectively.2.3. Data CollectionBody weight and rectal temperature were measured on Days –28, 0, 3, 7, 10, 14, 28 and 42 relative to the BVDV challenge (Day 0). Exit velocity was measured using infrared sensors on Days 0 and 14 as the time to transverse a fixed distance of 1.8 m upon exiting a squeeze chute (Farm Tec, Inc. North Wylie, TX, USA). Relative exit velocity (REV) was computed as (individual EV—mean EV) ÷ mean EV for each animal within year, and averaged for Days 0 and 14. Dry matter intake, ADG and feeding behavior traits were evaluated during each of the 4 14-d experimental periods (EP) relative to the BVDV challenge, which occurred on Day 0: Period 1 (Days –14 to −1), Period 2 (Days 0 to 13), Period 3 (Days 14 to 27) and Period 4 (Days 28 to 41). As BW was not measured on Day –14, ADG during the first 14-d period was computed from the BW measured on Days –28 and 0, with the assumption that growth was linear during the initial 28-d period. A timeline of the experimental procedures is provided in Figure 1.The steers were observed twice daily during the first 14 d following the BVDV challenge, and once per day thereafter to assess the clinical symptoms associated with BRD. Evaluations of cough, ocular and nasal secretion, depression, diarrhea and anorexia were recorded using a 0 to 5 clinical illness score (CIS; 0 = no symptoms; 1 to 5 were indicative of least severe to most severe). The criterion used to define BRD cases in this study were clinical scores of >3 for a single clinical symptom or combined scores of ≥3 for 2 or more clinical symptoms. Rectal temperatures were recorded on pre-determined days rather than as a final clinical threshold following the initial clinical assessment, as in field protocols for BRD diagnosis [19]. Animals that exhibited a rectal temperature of >40 °C were administered tulathromycin (Draxxin, Zoetis Animal Health), regardless of their CIS. This manuscript is part of a series that reports on various animal responses to a BVDV challenge. Investigations of clinical symptoms [19] and sire effects on DMI and ADG [20] following BVDV challenge in these cattle have been previously reported. This manuscript focused on the effects of animal temperament and VT on DMI and feeding behavior responses following a BVDV challenge.2.4. GrowSafe DataA GrowSafe system (DAQ 6000E) was used to measure feed intake and feeding behavior traits from 14 d prior to until 42 d following the BVDV challenge. The system consisted of feed bunks equipped with load bars to measure feed disappearance and RFID antennas within each feed bunk to record animal presence via detection of EID ear tags. Assigned feed disappearance (AFD) rates were computed daily for each feed bunk to assess data quality. Data for each pen were omitted from analysis due to system malfunctions, power outages or low (<95%) pen average AFD rates. During the 2010 and 2013 trials, data for 14 d and 2 d, respectively, were removed due to low AFD rates. The average AFD for the remaining days were 97.1% and 99.3%, respectively. No data were removed from the 2011 and 2012 trials, with average AFD rates exceeding 99%.The feeding behavior traits evaluated in this study were based on frequency and duration of bunk visit (BV) events, head-down (HD) duration, frequency and duration of meals events, and time to approach feed bunk (TTB) following feed-truck delivery (Table 1). A BV event commenced when the EID ear tag of an animal was first detected at the feed bunk and ended when the time between the last 2 consecutive EID recordings exceeded 100 s, the EID ear tag was detected at another feed bunk or the EID ear tag of another animal was detected at the same feed bunk [21]. Bunk visit frequency was defined as the number of independent events recorded, regardless of whether or not feed was consumed, and BV duration was defined as the sum lengths of all BV events recorded during a 24-h period [22]. Head-down duration was computed as the sum of the number of times an EID ear tag was detected each day multiplied by the scan rate of the GrowSafe system. R statistical software (R Core Team, 2014, Vienna, Austria) was used to compute TTB each day as the interval length between feed delivery for each pen and each animal’s first BV event following feed delivery [22]. Estimated values for missing feed intake data were derived from a linear regression of the feed intake on the day of the trial [23]. Bunk visit eating rate was computed as the ratio of daily DMI to daily BV duration.To compute meal data, a 2-pool Gaussian–Weibull distribution model was fitted to log-transformed non-feeding interval data. The intercept of the 2 distributions was used to define the meal criterion [24,25], which was the longest non-feeding interval considered to part of a meal event. The individual animal meal criterion was used to compute the frequency and duration of daily meal events. Meal eating rate was computed as the ratio of daily DMI and daily meal duration.2.5. Statistical AnalysisMixed model procedures of SAS 9.4 (SAS Inst. Inc., Cary, NC, USA) were used to analyze DMI, ADG and feeding behavior data. The model included VT and EP as fixed effects; REV as a covariate; the interactions of VT × EP, VT × REV, EP × REV and VT × EP × REV; and the random effects of year and pen within year. The 3-way and the EP × REV interactions were non-significant for all dependent variables and thus were removed from the final models. Previous analyses of these data have documented significant sire effects on DMI and ADG [20]. However, in the current study, sire was excluded from the statistical models in order to fully evaluate the effects of temperament on the response variables, as sire differences in animal temperament have been demonstrated to exist in this population [26].To examine the interactive effects between VT and REV, an unequal slope model was fitted for the dependent variables with significant (p < 0.05) VT × REV interactions. Subclass means for steers with calm and excitable temperaments were compared at mean REV minus 1 SD and mean REV plus 1 SD, respectively, using the PDIFF option in SAS. Differences between VT and EP least squares means were compared using the PDIFF option of SAS (SAS 9.4). Additionally, contrast statements were used to examine the nature of the dependent variable responses (linear, quadratic or cubic) across EP. Finally, to determine if temperament phenotype was equally distributed across the VT groups, steers were sorted by REV and classified into calm, moderate or excitable temperament groups based on them being ± 0.5 SD from the mean REV within the trial. The distribution of temperament classification within VT was examined with the PROC FREQ procedure of SAS using the CHISQ option (SAS 9.4).3. Results3.1. Vaccine Treatment and Experimental PeriodThe least squares means for DMI, performance and feeding behavior responses are presented in Table 2. Compared with Period 1, DMI, ADG and G:F were reduced by 15.9%, 27.7% and 20.0%, respectively, during the 14-d period following BVDV challenge (Period 2), and subsequently increased during Periods 3 and 4 in a cubic (p < 0.01) manner. Although the main effect of VT did not affect DMI, ADG or G:F, there was a VT × EP interaction (p < 0.05, Figure 2) for DMI. The reduction in DMI during Period 2 following BVDV challenge was less (p < 0.05) for MLV- (−10.6%) compared with KV- (−18.2%) and non-vaccinated steers (−18.9%); correspondingly, the subsequent increase in DMI during Period 3 was greater for KV- and non-vaccinated steers compared with MLV-vaccinated steers. While the VT × EP interaction was not significant (p = 0.11) for ADG, the effect of VT treatment on the reduction in ADG during Period 2 showed a similar trend to DMI (Figure 2).Compared with Period 1, the frequency and duration of BV events, the frequency and duration of meal events, HD duration and the frequency of BV events per meal all decreased by 15% to 25% during Period 2 and subsequently increased during Period 3 in a cubic (p < 0.01) manner. In contrast to these feeding behavior traits, meal eating rate actually increased by 6.4% during the 14 d after the BVDV challenge (Period 2) and continued to increase during Period 4 in a cubic (p < 0.01) manner. In contrast to meal eating rate, eating rate during BV events was not affected by the BVDV challenge. Compared with Period 1, TTB following feed delivery increased (p < 0.05) by 37.1% during the 14 d after the BVDV challenge, with TTB decreasing during Period 3 and 4 to values similar to those observed prior to BVDV challenge.In contrast to DMI, the VT × EP interactions were not significant for the feeding behavior traits. However, VT significantly altered feeding behavior traits throughout the study (Table 2), with MLV-vaccinated steers having distinctly different feeding behavior patterns compared with KV- and non-vaccinated steers. The MLV-vaccinated steers had 5% to 7% greater (p < 0.01) HD duration and durations of BV and meal events, and 4% to 5% slower (p < 0.01) BV and meal eating rates compared with KV- and non-vaccinated steers. Additionally, MLV steers had a 4% to 6% greater (p < 0.01) number of BV events per meal compared with KV- and non-vaccinated steers. The TTB following feed delivery and BV frequency were not affected by VT.3.2. TemperamentThe distribution of steers with divergent phenotypes for temperament across VT was evaluated based on being ± 0.5 SD from the mean REV of 0 ± 0.24. For steers with excitable temperaments, the distribution of steers was similar (p = 0.14) for NON, KV and MLV vaccine treatments at 28.2%, 33.3% and 38.5%, respectively. Likewise, steers with calm temperaments were equally distributed for NON, KV and MLV vaccine treatments (37.6%, 25.7% and 36.6%, respectively).Relative exit velocity was a significant covariate (p < 0.01) for DMI, such that steers with calm temperaments (mean REV–1 SD) consumed 5.0% more feed than steers with excitable temperaments (mean REV + 1 SD), irrespective of the vaccine treatment (Table 3). There was a tendency (p = 0.08) for REV to affect ADG, with calm steers having a 5.3% numerically greater ADG compared with excitable steers. However, REV did not affect (p = 0.69) G:F, and the VT × REV interactions were not detected (p ≥ 0.29) for DMI, ADG or G:F.With the exception of BV frequency and meal eating rate, REV was a significant covariate for all feeding behavior traits. In general, HD duration, and BV and meal duration all decreased (p < 0.01) as REV increased (Table 3). However, VT × REV interactions were detected (p < 0.05) for both HD and meal duration. In KV- and non-vaccinated steers, these traits were not affected by REV (the slopes did not differ from zero), but in MLV-vaccinated steers, both HD and meal duration decreased as initial REV increased (Figure 3). Within calm steers, MLV-vaccinated steers had greater (p < 0.01) HD and meal duration compared with KV- and non-vaccinated steers, whereas VT differences in HD and meal duration were not detected in steers with excitable temperaments. There were also significant VT × REV interactions for both BV and meal eating rates (Figure 3). Within steers with a calm temperament, MLV-vaccinated steers had lower BV and meal eating rates compared with KV- and non-vaccinated steers. However, BV and meal eating rates were not affected by VT in steers with excitable temperament phenotypes.Although the frequency of BV events was not affected by REV, there was a VT × REV interaction (p < 0.05) for the frequency of meal events. Meal frequency increased as REV increased in KV- and non-vaccinated steers; however, REV had no effect on meal frequency in MLV-vaccinated steers (Figure 4). The KV- and non-vaccinated steers with excitable temperaments consumed more meals per day than MLV-vaccinated steers, whereas VT did not affect meal frequency in calm steers. Reflecting the influence of temperament on meal frequency, the number of BV events per meal declined (p < 0.05) as REV increased. Irrespective of VT, TTB following feed delivery was affected by REV, with excitable steers taking almost 4 min longer to approach the feed bunk following feed delivery than steers with calm temperaments (Table 3).4. Discussion4.1. Responses to the BVDV ChallengeThis study demonstrated the effects of sub-clinical illness on performance traits in Bos indicus crossbred steers that Griffin [27] and others have reported as plaguing the beef industry. Detailed descriptions of health traits and antibody titer responses were previously reported by Downey-Slinker et al. [19]. In this study, 14% of the steers had a CIS of 1 or 2 (CIS = 0 to 5); however, none of the steers met the criteria for clinical BRD diagnosis following the BVDV challenge and none of the steers died. As reported by Downey-Slinker et al. [19], within the 14-d period following BVDV challenge, 40% of steers presented with pyrexia (1 SD over the Day 0 rectal temperature for two measurement days), 55% presented with lymphopenia (>40% reduction in lymphocyte counts) and 41% presented with thrombocytopenia (>40% reduction in platelet counts). Both lymphopenia and thrombocytopenia are well-established indicators of subclinical BVDV infection in beef cattle [12,13]. Other studies have also shown that animals challenged with BVDV Type 1b [10] or BVDV Type 2 [28] strains do not always manifest with observed clinical signs of BRD. Burciaga-Robles et al. [29] reported that calves challenged with BVDV Type 1b had minimal or no observed clinical signs of BRD.Despite the lack of clinically diagnosed BRD cases, there were substantial reductions in DMI, ADG, G:F and feeding behavior traits during the 14-d period following the BVDV challenge. Similar patterns have been documented in other studies with clinically ill cattle. During a spontaneous outbreak of BRD in growing bulls (8–9 mo of age), Jackson et al. [22] reported that DMI was reduced by 39% during the week prior to observed clinical BRD diagnosis. Likewise, the frequency and duration of BV events declined by 2.9 events/d and 4.4 min/d, respectively, during the week prior to an observed clinical diagnosis of BRD [22]. Carlos-Valdez et al. [30] reported that Angus crossbred steers challenged with Mannheimia haemolytica after exposure to a persistently infected BVDV Type 1 calf had reduced DMI, ADG and G:F during the first 4 d following challenge. Similarly, Theurer et al. [31] reported that calves challenged with M. haemolytica spent less time at both the feed bunk and hay feeder compared with calves that were not challenged. In addition, Wolfger et al. [32] reported that an increase in feed intake per meal event, along with increases in the frequency and duration of meal events, was associated with a lower risk of developing BRD. Hutcheson and Cole [33] reported that calves observed to be clinically ill had 11% lower intake and 29% lower ADG compared with calves observed to be healthy. Sowell et al. [34] found that morbid steers had fewer feeding bouts and spent less time at the feed bunk compared with healthy steers, and Sowell et al. [35] reported that clinically healthy steers had more rapid responses following feed delivery than steers identified as being clinically ill. Likewise, Daniels et al. [36] reported that clinically ill calves had a lower frequency of feeding bouts and spent less time at the feed bunk compared with clinically healthy calves. In dairy cows, reductions in DMI and feeding behavior responses prior to and following a diagnosis of clinical mastitis [37], metritis [38] and ketosis [39,40] have been observed.The effects of the BVDV challenge caused ADG, DMI and most of the feeding behavior traits to decline during the 14-d period immediately following the BVDV challenge (Period 2), which subsequently increased during Periods 3 and 4 in a cubic manner. These performance trait responses during Period 2 were associated with substantial reductions in lymphocyte and platelet counts on Day 14 following the BVDV challenge, as previously reported by Downey-Slinker et al. [19]. Compared with clinically healthy calves, Buhman et al. [41] reported that the frequency and duration of feeding events were reduced in morbid calves 11 to 27 d after feedlot arrival, and increased thereafter during the next 28-d period. These authors attributed this increase in feeding activity to a post-sickness compensation. Carlos-Valdez et al. [30] reported a post-sickness compensation in calves exposed to persistently infected BVDV calves for 72 h then challenged with M. haemolytica. Following reductions in DMI, ADG and G:F during Days 0 to 4 after M. haemolytica challenge compared with control calves, there was a subsequent increase in ADG and G:F during Days 5 to 17 following the challenge [30]. Calves challenged with M. haemolytica during Days 5 to 17 appeared to compensate for the loss in production and showed an increase in ADG and G:F compared with control calves [30]. Holland et al. [42] reported that crossbred heifers treated for BRD had a lower ADG compared with those not treated during the preconditioning phase; additionally, there was a greater compensation in ADG during the first 28 d following the preconditioning phase for cattle treated three times compared to cattle that had never been treated for BRD. A similar compensation was observed in this study, with DMI, ADG and G:F being substantially greater during Period 3 following the BVDV challenge. Likewise, the frequency and duration of BV events, and HD and meal durations were greater, and the TTB following feed delivery was faster during Period 3 than Period 2, demonstrating that the steers quickly compensated for the BVDV challenge in this study.4.2. Vaccine Treatment EffectsThe reduction in DMI following the BVDV challenge was less pronounced in MLV-vaccinated steers compared with the KV- and non-vaccinated steers. These results coincided with the previous findings of Downey-Slinker et al. [19], whereby MLV-vaccinated steers had a reduced (33.9%) incidence of lymphopenia compared with KV- (64.7%) and non-vaccinated steers (68.1%). Although VT did not affect the proportion of steers exhibiting pyrexia during the 14-d post-BVDV challenge period, the MLV-vaccinated steers had lower rectal temperatures compared with KV- and non-vaccinated steers on Days 3 and 7 following the challenge [19].Vaccine treatment clearly altered feeding behavior patterns, such that MLV-vaccinated steers had a greater duration of both BV and meal events, greater HD duration and more BV events per meal compared with KV- and non-vaccinated steers. Additionally, as VT did not alter DMI, eating rates during both BV and meal events were slower in MLV- compared with KV- and non-vaccinated steers. These results, in conjunction with those of Downey-Slinker et al. [19], suggest that the multivalent MLV provided a greater level of protection to the BVDV challenge compared with the KV. Although some studies have reported no difference in antibody response between KV and MLV vaccines [43], Downey-Slinker et al. [19] found that the MLV-vaccinated steers in the present study had greater BVDV Type 1b titer concentrations compared with KV-vaccinated steers prior to BVDV challenge, but KV-vaccinated steers had greater titers at 14 d following the challenge. Additionally, MLV vaccines have been shown to reduce susceptibility to lymphopenia and reduce the fever response to a greater extent as compared with KV vaccines [13,14]. Collectively, results from these studies suggest that the MLV vaccine was more effective at mitigating subclinical symptoms of BRD compared with the KV vaccine.4.3. Temperament EffectsThe effects of temperament on the DMI and performance of cattle in multiple breeds have been well documented, such that more excitable steers have decreased DMI and ADG compared with calm steers [44,45,46]. In agreement with previous research, the results from the current study found that calm steers had greater DMI and numerically greater ADG compared with excitable steers. However, there have been mixed results on the effect of temperament on feed efficiency. Bruno et al. [45] reported that temperament did not affect G:F, even though cattle with calm temperaments had increased DMI and ADG compared with excitable cattle. Bos indicus crossbred steers [47] and heifers [46] with excitable temperament phenotypes had lower ADG and less favorable G:F compared with cattle with calm temperaments. Likewise, Cafe et al. [48] reported that Angus steers with excitable temperaments upon feedlot arrival tended to have less favorable G:F than steers with excitable temperaments. In contrast to the latter studies, temperament phenotype did not affect G:F in the current study among these 50% Bos indicus steers.Reflecting the effect of temperament on DMI, steers with calm temperaments spent more time each day consuming feed, whether quantified as HD duration or the durations of BV or meal events. Furthermore, eating rate during BV events was slower in calm than excitable steers. Similarly, Cafe et al. [48] reported that cattle with a faster EV spent less time at the feed bunk compared with cattle with a slower EV. Olson et al. [46] reported that heifers with calm phenotypes had 9% greater meal duration, and consumed meals that were 22% longer and 17% larger compared with excitable heifers. These results, along with the results of Cafe et al. [48] and Olson et al. [46], suggest that cattle with calm temperaments have more favorable feeding behavior patterns compared with cattle with more excitable temperaments.As previously discussed, the reduction in DMI associated with an increase in REV was not affected by VT. In contrast, the effects of temperament on feeding behavior responses were impacted by VT, suggesting that feeding behavior responses were more sensitive in detecting differences due to VT than DMI responses. The MLV-vaccinated steers with calm temperaments had greater HD and meal durations, and slower meal eating rates than KV- and non-vaccinated steers, whereas VT differences in these feeding behavior traits were not detected in steers with excitable temperaments.It is interesting to note that the MLV-vaccinated steers exhibited favorable feeding behavior patterns, despite the fact that there was a numerically higher proportion of steers with excitable temperaments in the MLV- compared with the non- and KV-vaccinated treatments (38.5%, 28.2% and 33.3%, respectively). The results from this study would suggest that the benefits of the multivalent MLV vaccine were more evident in calm than in excitable steers, which may be related to the fact that steers with an excitable temperament have heightened physiological responses to stress. Multiple studies have reported that excitable steers have greater cortisol responses to stress than calm steers [16], which has been shown to negatively affect the immunocompetence of cattle [49]. In addition, Oliphint et al. [7] reported that cattle with excitable temperaments had reduced immune responses to vaccination compared with calm cattle. These results suggest the potential beneficial effects of the MLV vaccine may have been mitigated by the increased stress responsiveness exhibited by excitable temperaments. Furthermore, previous analyses of data from this study demonstrated that subjective temperament scores were moderately correlated in a negative manner with antibody titer responses following the BVDV challenge [50].5. ConclusionsThe objectives of this study were to quantify the effects of the multivalent VT and animal temperament on DMI, performance and feeding behavior responses following a BVDV1b challenge. The results demonstrated that the multivalent VT clearly altered DMI following BVDV challenge, such that MLV-vaccinated steers had less of a reduction in DMI compared with KV- and non-vaccinated steers. Furthermore, feeding behavior patterns were substantially affected by VT, with MLV-vaccinated steers having increased feeding duration and slower eating rates compared with KV- and non-vaccinated steers. These results, in conjunction with those of Downey-Slinker et al. [19], suggested that the MLV vaccine mitigated the impacts of the BVDV challenge to a greater extent compared with the KV and NON treatments. Additionally, temperament affected DMI and feeding behavior patterns, with calm steers having increased DMI and feeding durations, and slower eating rates during the BV events compared with excitable steers. Previous analyses in these cattle have demonstrated a substantial genetic influence for temperament at weaning [26], and for DMI and ADG following this BVDV challenge [20]. We observed that the same impacts of VT may not occur across all animal temperament categories The increased stress responsiveness of excitable steers in this study appeared to have mitigated the beneficial effects of the MLV vaccine. This study indicated that temperament classification may be at least a partial proxy for genetic background when pedigrees are unavailable.

animals : an open access journal from mdpi

10.3390/ani13091485

PMC10177341

Among the most important factors in the production of livestock are morphological traits, such as body weight, body length, chest depth, chest width, cannon circumference, and the body index. Platelet-derived growth factor D (PDGFD) is a candidate gene that has the potential to affect fat deposition and body size in sheep. In this study, we discovered two intronic InDels (13 bp deletion, 14 bp insertion) with the potential to significantly affect the morphological traits of three different indigenous Chinese sheep breeds. These InDels can be used as DNA markers for sheep marker-assisted selection (MAS) breeding.

Platelet-derived growth factor D (PDGFD) is a member of the PDGF gene family, and it plays an important role in the regulation of adipocyte development in mammals. Furthermore, genome-wide association studies (GWAS) have previously identified it as a candidate gene associated with fleece fiber variation, body size, and the fat-tail phenotype in domestic Chinese sheep. In this study, a total of 1919 indigenous Chinese sheep were genotyped to examine the association between nucleotide sequence variations in PDGFD and body morphology. Our results detected both a 14 bp insertion in intron 2 and a 13 bp deletion in intron 4 of PDGFD. Moreover, these two InDel loci had low to moderate polymorphism. Notably, the 13 bp deletion mutation of PDGFD was found to significantly affect sheep body size. Yearling rams in the Luxi black-headed sheep (LXBH) containing a heterozygous genotype (insertion/deletion, ID) were found to have larger body length, chest depth, and body weight than those with wild genotypes. Furthermore, adult ewes in the Guiqian semi-fine wool sheep (GSFW) containing a homozygous mutation (deletion/deletion, DD) were found to have smaller chest width than their peers. Moreover, yearling ewes in this group with the same homozygous mutation were found to have lower body weight, chest width, and cannon circumference compared to those of other individuals. This study demonstrates that PDGFD InDel polymorphisms have the potential to be effective molecular markers to improve morphological traits in domestic Chinese sheep.

1. IntroductionChina has many extensive and genetically diverse types of sheep that are well adapted to local agricultural and climate conditions, being resistant to common diseases [1]. Lanzhou fat-tailed sheep (LFT) have long tails and deposit the majority of their adipose tissue within them, thus allowing them to adapt well to harsh climates and environmental conditions [2,3]. Luxi black-headed sheep (LXBH) have short tails with plenty of fat, as well as a high fertility rate and excellent meat production capacity [4,5]. Furthermore, Guiqian semi-fine wool sheep (GSFW) are selected for their dual purpose of high mutton and wool production [6]. Increasing sheep productivity and improving meat quality are consistent priorities necessary for the continuous development of the Chinese sheep breeding industry. Livestock growth and carcass traits belong to the medium heritability range (0.20–0.35); thus, the accuracy of sheep selection and breeding can be improved by gene marker-assisted and genomic selection of candidate genes related to the above traits [7,8].Previous selective sweep analyses and genome-wide association studies (GWAS) identified several key candidate genes functionally associated with growth, body size, and tail morphology in sheep, such as platelet derived growth factor D (PDGFD), bone morphogenetic protein 2 (BMP2), thyroid-stimulating hormone receptor (TSHR), xylulokinase (XYLB), and fibroblast growth factor 7 (FGF7) [9,10]. In recent years, the PDGFD gene was discovered to belong to the PDGF family, which encodes for a mitogenic factor used by mesenchymal cells and is located in the blood serum [11]. Moreover, PDGFD contains a tissue-specific expression pattern. It is expressed more abundantly in the pancreas, pituitary gland, ovaries, and adipose tissue than in other tissues in humans [12]. Studies have shown that PDGFD is involved in various intracellular signaling pathways, (PI3K/Akt, MAPK, mTOR, and Notch signaling) and regulates IGF1R, VEGF, and Snail, among other proteins. By affecting these pathways, PDGFD has the potential to regulate multiple-organ fibrosis, atherosclerosis, tissue repair, and cancer in humans, as well as fleece fiber diameter in domestic sheep [13,14,15,16].Furthermore, PDGFD is highly conserved amongst different species [17]. In a recent study, one cattle nonreference sequence (NRS) variant overlapping with the intergenic region between the protein-coding genes PDGFD and DYNC2H1 (dynein cytoplasmic 2 heavy chain 1) was identified by multiple de novo assemblies, which may have biological significance [18]. PDGFD was also shown to affect intermuscular fat deposition in Dianzhong cattle [19]. Furthermore, PDGFD plays a critical role with regard to adipocyte proliferation and differentiation, and it is directly and indirectly involved in fat metabolism as it relates to the tails of sheep [10]. There are two single-nucleotide polymorphisms (SNPs) within the PDGFD gene that have been shown to significantly affect the size of sheep tails [20]. These studies suggest that PDGFD is closely related to sheep growth and development. However, it is notable that the functional role of PDGFD polymorphisms in relation to morphological traits of sheep is still poorly understood. Therefore, we aimed to characterize InDel variations of PDGFD in three different Chinese sheep breeds to validate the association between InDel polymorphisms and morphological traits, as well as provide fundamental reference data for use of PDGFD in domestic sheep breeding.2. Materials and Methods2.1. Animals, Phenotypic Data, and DNA ExtractionA total of 1919 sheep of three breeds were chosen for this study: Guiqian semi-fine wool sheep (GSFW, n = 1243, Bijie, China), Luxi black-headed sheep (LXBH, n = 618, Liaocheng, China), and Lanzhou fat-tailed sheep (LFT, n = 65, Lanzhou, China). Of the 618 LXBH sheep, 37.2% (n = 230) were lambs (≤3 months), 44.3% (n = 274) were yearlings (4–18 months old), and 18.5% (n = 114) were adults (>18 months). Among the GSFW group, 477 animals already had body size records (yearling rams, n = 143; yearling ewes, n = 155; adult ewes, n = 179). Within each sheep breed, healthy sheep individuals raised under similar feeding and management conditions were selected as tested sheep in this study. Morphological characteristics, i.e., body weight, body height, body length, chest circumference, chest depth, chest width, and cannon circumference, were recorded for each sheep in the study [21,22]. On the basis of prior established formulas, body size indices, such as body trunk index (the ratio of chest circumference to body length), body length index (the ratio of body length to body height), chest width index (the ratio of chest width to chest depth), chest circumference index (the ratio of chest circumference to body height), cannon circumference index (the ratio of cannon circumference to body height), and limb length index (the ratio of the difference between body height and chest depth to body height), were also calculated [23,24]. Furthermore, genomic DNA was extracted from ear tissues using standard phenol/chloroform extraction procedures as described previously and diluted to 20 ng/μL with ddH2O after determination of concentration [25]. To study genetic differences in the PDGFD gene, 30 DNA samples from each sheep breed were chosen at random and placed into three different DNA pools [26,27,28].2.2. InDel Detection and GenotypingThe Ensembl database was used to find six possible InDel sites within sheep PDGFD gene. NCBI Primer-Blast was subsequently used to design six pairs of primers (Table 1) based on the PDGFD reference sequence for sheep (NC_056068.1). These primers were synthesized via Sangon Biotech (Xi’an, Shaanxi, China). PCR was performed using a 13 μL reaction volume and Touchdown program described previously [25,26]. PCR products were then directly examined via electrophoresis with a 3% agarose gel for confirmation.2.3. Statistical AnalysesGenotype frequency, allele frequency, and population indices (heterozygosity, He; the number of effective alleles, Ne; polymorphism information content, PIC) were all calculated using Microsoft Excel (2019). Population indices were calculated following Nei’s methods [29]. In order to check whether the polymorphisms deviated from the Hardy–Weinberg equilibrium (HWE), the chi-square test (χ2) was carried out. Differences in allele frequencies, as well as the genotypic distribution, for each sheep population were assessed using the χ2 test or Fisher’s exact test. If all expected counts were greater than 5, the Pearson chi-square test was used. If there was a certain expected count <5, the Fisher exact test was used. The SHEsis online platform (http://analysis.bio-x.cn, accessed on 4 October 2022) was used for linkage disequilibrium (LD) analysis. The association between different genotypes of InDels and morphological traits of each sheep breed was evaluated using the Student’s t-test (only two genotypes) or one-way ANOVA (three genotypes) via SPSS 25 software. The statistical model used was as follows: Yijk = µ + Gi + Sj + Ak + eijk, in which Yijk represents the phenotypic value for morphological traits, μ represents the population mean, Gi represents the fixed PDGFD genotype effect for each group of sheep, Sj represents the fixed effect of gender, Ak represents the fixed effect of the age, and eijk represents the random error.3. Results3.1. InDel Genotyping and SequencingOn the basis of the combined results of gel electrophoresis, as well as sequencing profiles, two noncoding InDel mutation sites within PDGFD gene were detected: a 14 bp insertion (rs590816164) in intron 2, and a 13 bp deletion in intron 4 (rs1092650847) (Figure 1). Genotyping analysis showed that there were two genotypes at the P4-ins-14bp site (insertion/deletion, ID, 312 bp/298 bp; deletion/deletion, DD, 298 bp) (Figure 2), and three genotypes at P5-del-13bp site (insertion/insertion, II, 185 bp; ID, 185 bp/172 bp; DD, 172 bp) (Figure 3). The P4-ins-14bp locus was polymorphic for both LFT and GSFW sheep, whereas the P5-del-13bp showed genetic polymorphisms amongst all three sheep breeds.3.2. Genetic Parameters and Linkage Disequilibrium AnalysisGenotypes, allele frequencies, and population genetics analyses are shown in Table 2. Amongst all sheep breeds, the frequency of the wild genotype was greater than 0.6 at both InDel loci. Furthermore, wildtype frequency was found to be greater than the frequency for both homozygous mutants and heterozygotes. Population genetic analyses indicated that the P4-ins-14bp locus demonstrated low levels of polymorphism (PIC < 0.25) amongst all tested sheep breeds. Furthermore, the P5-del-13bp demonstrated low levels of polymorphism (PIC < 0.25) for Luxi black-headed sheep and Lanzhou fat-tailed sheep, and moderate levels of polymorphism (0.25 < PIC < 0.5) for Guiqian semi-fine wool sheep. Furthermore, in GSFW populations, the P4-ins-14bp locus was found to deviate from the Hardy–Weinberg equilibrium (p < 0.05). No significant deviations from HWE were identified for the P5-del-13bp locus in all tested sheep breeds (p > 0.05).Chi-square test results pertaining to the allelic frequency distribution of both InDel mutation sites amongst all three sheep breeds showed that the allele and genotype frequencies of the P4-ins-14bp locus were significantly different amongst all three breeds (p < 0.05) (Table 3). Furthermore, at the P5-del-13bp locus, the frequencies for both genotype and allele were considered statistically different amongst all three breeds (Table 4). According to linkage disequilibrium analysis (Figure 4), both P4-ins-14bp and P5-del-13bp loci showed a strong LD state within the LFT (D′ = 1, r2 = 0) population, and a weak linkage state in the GSFW (D′ = 0.060, r2 = 0.001) population.3.3. Association Analysis of PDGFD InDel and Body Morphometric TraitsThe association between genotype and phenotypic traits (seven body size traits and six body size indices) was analyzed. Significant deviations from HWE were identified for the P4-ins-14bp locus (p = 0.001142); thus, this site was not used for association analysis [30]. For the P5-del-13bp locus, significant differences were observed in chest depth, body length, and body weight amongst LXBH yearling rams (p < 0.05), and individuals with wild genotype had a smaller overall body size than individuals with a heterozygous genotype. Furthermore, a homozygous mutation (DD) at the P5-del-13bp locus showed a strong, negative influence on body weight, chest width, and cannon circumference in the GSFW sheep (p < 0.05). The body weight, chest width, and cannon circumference of yearling ewes with DD genotypes were significantly smaller than those of other individuals (p < 0.05). The chest width of adult ewes with a homozygous mutation was significantly smaller compared to other individuals (p < 0.05) (Table 5). Some traits did not differ significantly amongst individuals with different genotypes, and these results are not shown in the results table.4. DiscussionThe PDGF family includes proteins that are both proangiogenic and regulatory factors stimulating connective tissue growth [31]. PDGF comprises four subtypes: PDGFA, PDGFB, PDGFC, and PDGFD [32]. A prior study suggested that PDGFC is involved in adipose expansion [33]. The protein coded for by PDGFD mediates PDGF receptor chain tyrosine kinase (PDGFR-β) dimer formation and activation [34]. PDGFD is associated with angiogenesis required for tissue development and growth due to the inclusion of PDGF receptors on the surface of adipocytes and pre-adipocytes [35]. Furthermore, PDGFD has been shown to play a key role in the formation and development of pre-adipocytes, which can be seen in all species of mammals [27].Adipose tissue is the main means of caloric storage in the body and plays a crucial role in the regulation of metabolism and body shape [36]. The large amount of tail fat in fat-tailed sheep was shown to have a significant influence on fat deposition in other parts of their bodies [37]. Thus, elucidating genetic variations of genes related to fat metabolism and deposition, as well as analyzing how these variations associate with morphological traits may provide more information about useful molecular markers to genetically improve sheep morphology [38]. For example, the gene encoding for cyclic AMP response element-binding protein 1 (CREB1) regulates fat metabolism in sheep adipose tissue, with one mutation within its first intron region strongly associated with differences in body measurement traits [39,40]. PDGFD, which is located on chromosome 15, has also been identified as a candidate gene for the phenotypic traits of sheep (i.e., fat tail and body size) [5,10].Related studies have shown that genes affecting animal reproductive traits may also affect animal morphological and growth-related traits. Recently, Kang et al. discovered that a 7 bp InDel variation in intron 8 of lysine demethylase 3B (KDM3B) not only affected the litter size of Australian white sheep, but was also significantly associated with the body size traits of Lanzhou fat-tailed sheep and Luxi black-headed sheep [41,42]. Previously, Su et al. discovered that an 18 bp deletion in intron 2 of PDGFD was associated with litter size for Australian white sheep [25]. Data from the Ruminant Genome and NCBI databases showed that PDGFD is highly expressed in the reproductive system (i.e., ovary) [25]. These results imply that PDGFD may control multiple traits simultaneously (i.e., gene pleiotropism).With these studies in mind, PDGFD was considered as a potential candidate gene for our study. In this study, we detected two InDel mutations, P4-ins-14bp and P5-del-13bp, both of which were found to be in Hardy–Weinberg equilibrium (p > 0.05) for Lanzhou fat-tailed sheep and Luxi black-headed sheep. These results indicate that no large-scale migration or mutations occurred in the tested population. It is worth noting that the P4-ins-14bp InDel existed in both Lanzhou fat-tailed sheep (2.59%) and Guiqian semi-fine wool sheep (8.31%) populations with relatively low frequency; however, Luxi black-headed sheep populations did not have this InDel. This may be due to different breeding requirements for those sheep varieties or the relatively conservative nucleotide sequence of PDGFD, as its genetic variation has been shown to be species-specific and vary on the basis of population size with regard to analysis and testing [43]. The P5-del-13bp locus showed moderate polymorphism in Guiqian semi-fine wool sheep, indicating that this population has rich genetic diversity.Furthermore, both P4-ins-14bp and P5-del-13bp loci were located in the intronic regions of PDGFD. Studies have shown that intronic mutations may function to regulate transcriptional activity by preventing or increasing transcription factor binding, thus influencing gene expression via splicing, which leads to phenotypic variation [44,45,46,47]. Wang et al. found an intron mutation at the seventh intron of human transcription factor hepatocyte nuclear factor 1A (HNF-1A) that caused abnormal mRNA splicing and impaired its activity as a transcription factor [48]. Li et al. identified nucleotide mutations within a cis-regulatory element in the bone morphogenetic protein receptor type-1B (BMPR1B) intron 1, which could control pig prolificacy via the cis-regulation of BMPR1B expression [49]. Therefore, we further investigated the effect of these two intronic mutations on the phenotypic variation of tested sheep. A strong genotype–phenotype association between the P5-del-13bp polymorphism and morphological traits of local sheep was observed. For Guiqian semi-fine wool sheep, the P5-del-13bp locus was found to negatively influence body weight, which is associated with negative responses in fleece production [50]. In the future, further studies are needed to explore the mechanism of phenotypic change caused by noncoding genetic variants. Overall, our findings indicate that PDGFD is a critical gene related to morphological traits in sheep.5. ConclusionsIn this study, a 14 bp insertion in intron 2 and a 13 bp deletion in intron 4 within the PDGFD gene were detected in multiple breeds of sheep. Additionally, the 13 bp noncoding genetic variation had a significant effect on morphological traits (body weight, body length, chest depth, chest width, cannon circumference, etc.) of Luxi black-headed sheep and Guiqian semi-fine wool sheep, indicating the possibility of using these InDel mutation sites as DNA markers to increase the growth and development of indigenous Chinese sheep. However, the concrete mechanisms via which PDGFD gene and the 13 bp deletion affect morphological traits are still indistinct.

animals : an open access journal from mdpi

10.3390/ani11030814

PMC8001643

We explored the existence of Anaplasma phagocytophilum and related variant in samples of goats and sheep obtained from Antalya and Mersin provinces, representative of Mediterranean region of Turkey. Based on 16S rRNA and groEL genes of A. phagocytophilum and related variants, we examined blood samples by polymerase chain reaction (PCR) followed by sequencing. The results showed that the prevalence of A. phagocytophilum and A. phagocytophilum-like 1 infection was 1.4% and 26.5%, respectively. Sequencing confirmed molecular data and showed the presence of A. phagocytophilum and A. phagocytophilum-like-1 variant in the sampled animals.

Anaplasma phagocytophilum causes tick-borne fever in small ruminants. Recently, novel Anaplasma variants related to A. phagocytophilum have been reported in ruminants from Tunisia, Italy, South Korea, Japan, and China. Based on 16S rRNA and groEL genes and sequencing, we screened the frequency of A. phagocytophilum and related variants in 433 apparently healthy small ruminants in Turkey. Anaplasma spp. overall infection rates were 27.9% (121/433 analyzed samples). The frequency of A. phagocytophilum and A. phagocytophilum-like 1 infections was 1.4% and 26.5%, respectively. No A. phagocytophilum-like 2 was detected in the tested animals. The prevalence of Anaplasma spp. was comparable in species, and no significant difference was detected between sheep and goats, whereas the prevalence significantly increased with tick infestation. Sequencing confirmed PCR-RFLP data and showed the presence of A. phagocytophilum and A. phagocytophilum-like-1 variant in the sampled animals. Phylogeny-based on 16S rRNA gene revealed the A. phagocytophilum-like 1 in a separate clade together with the previous isolates detected in small ruminants and ticks. In this work, A. phagocytophilum-like 1 has been detected for the first time in sheep and goats from Turkey. This finding revealed that the variant should be considered in the diagnosis of caprine and ovine anaplasmosis.

1. IntroductionAnaplasma phagocytophilum is the agent of tick-borne fever (TBF) or pasture fever, a disease affecting some species of domestic ruminants (cattle, sheep, goats). The bacterium is a pathogenic species for livestock such as ruminants as well as humans in temperate and tropical countries [1,2,3,4]. Anaplasma phagocytophilum is transmitted by Ixodes spp. and infects host neutrophils and monocytes, where reproduction occurs [1,5]. Anaplasma phagocytophilum infection is known as pasture fever and characterized by fever, anorexia, lateral recumbency, dullness, and loss of milk yield in affected hosts [2,4,6].Increased attention to A. phagocytophilum reveals new information about the genetic diversity of the pathogen. Recently two Anaplasma variants related to A. phagocytophilum have been documented in cattle, sheep, goats, and ticks [7,8,9]. In Japan, A. phagocytophilum-like 1 has been detected in deer and Hemaphysalis longicornis [10], cattle [11], Ixodes spp. [12], and Haemaphysalis megaspinosa [13]. A. phagocytophilum-like 2 has been identified in Hyalomma asiaticum [14], sheep and goats from China [15]. Recently those Anaplasma variants have been documented in ruminants from Tunisia [7,8], South Korea [16], and Italy [17].Various Anaplasma species including A. phagocytophilum have been documented in ruminants and ticks in Turkey [5,6,18,19,20,21]. However, until now no data on A. phagocytophilum variants is available in Turkey. In the current study, 16S rRNA, groEL (heat shock protein) PCR and sequencing were performed to identity A. phagocytophilum and A. phagocytophilum-like variants in small ruminants from sampling sites in Antalya and Mersin provinces, where the representative Mediterranean area of Turkey.2. Materials and Methods2.1. Study Region and Sample CollectionThis survey was conducted in small ruminants farmed in three districts (Alanya, Akseki, Manavgat) from Antalya (latitude 36° 53′ N, longitude 30° 42′ E) and two districts (Anamur, Bozyazı) from Mersin (latitude 36° 47′ N, longitude 34° 37′ E) provinces of Turkey (Figure 1). This area has a Mediterranean climate, with hot humid summers and warm rainy winters. The goats and sheep are kept in closed areas in villages near to the coast during the winter months, and they are taken to the plateaus in the Taurus Mountains in the early spring and grazed in the pastures here until autumn.The sample size was calculated using the online tool Sample Size Calculator (www.calculator.net/sample-sizecalculator.html, accessed on 1 February 2019), for a confidence level (CL) of 95%, an error margin of 5%. According to this, during April–July 2019, a total of 433 apparently healthy small ruminants (296 goats, 137 sheep) were included in the survey. Blood samples were drawn from the punctured jugular vein into anticoagulated (K3-EDTA) vacutainer tubes and stored at −20 °C freezer until DNA extraction. The goats and sheep were also checked for tick infestations, and a total of 1475 ticks were removed. The collected ticks were preserved in 70% ethanol in Eppendorf tubes. They were identified using taxonomic keys [22]. The animals were grouped into categories according to species (goat and sheep) and the presence of ticks (yes/no). This study secured the approval of the Elazig Veterinary Control Institute (number: 2018/02).2.2. DNA Extraction and Amplification of 16S rRNA GeneDNA was isolated from 200 µL volumes of whole blood using a DNAeasy Blood Minikit according to the vendor’s recommendations. Genomic DNA from blood of clinically infected cattle with A. phagocytophilum [6] was used for positive control. Anaplasma phagocytophilum-like variants DNAs, received from Alberto Alberti (University of Sassari, Sassari, Italy) were used as positive controls.To detect A. phagocytophilum and A. phagocytophilum-like variants, a nested 16S rRNA PCR was carried out described by Kawahara et al. [10]. The PCR reaction conditions were made according to the previously described studies [10,21]. The nested amplicons were examined by 1.5% agarose gel electrophoresis and visualized using the gel Documentation System (Vilber Lourmat, Marne La Vallee Ceedex, France).2.3. Restriction Fragment Length Polymorphism (RFLP)XcmI and BsaI restriction enzymes allow the specific discrimination amongst A. phagocytophilum and related variants [8,17]. For differentiation of A. phagocytophilum and related variants, the nested amplicons obtained in this study were digested with the XcmI and BsaI restriction enzymes as previously described [8,17].2.4. GroEL PCRTo confirm the results of the RFLP assay, the positive samples were screened by a groEL nested PCR for the amplification of A. phagocytophilum [23]. The semi-nested PCR reported by Ybañez et al. [24] with the primers EEGro1F/AnaGroe712R and AnaGroe240F was utilized for amplifying of A. phagocytophilum-like 1 groEL gene. Oligonucleotide primers used in this study were presented in Table 1.2.5. Sequencing and Phylogenetic AnalysesAnaplasma phagocytophilum (n = 6) and A. phagocytophilum-like 1 (n = 10) positive PCR amplicons were purified using the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany). The purified amplicons were sent to BM Labosis (Ankara, Turkey) for Sanger sequencing to determine DNA sequences of the 16S rRNA gene. Multiple alignments were performed with the CLUSTAL Omega ver. 1.2.1 (https://www.ebi.ac.uk, accessed on 1 February 2019). The representative sequences have been submitted to the GenBank (MT881655 and MT881656 for 16S rRNA gene of A. phagocytophilum-like 1 and A. phagocytophilum, respectively). The sequence alignment was performed using MUSCLE in Geneious prime [25].Phylogenetic analyses of the 16S rRNA sequences obtained in this work and the other sequences submitted to GenBank were carried out. The maximum likelihood analysis (ML) carried out in Mega X [26] was utilized to determine the phylogenetic relationship of the Anaplasma spp. To sequence evolution, best-fit model was assessed as TN93+G+I by using the jModel test v.0.1.1 [27]. Reliability of internal branches of the tree was evaluated by the bootstrapping method with 1000 iterations [28].2.6. Statistical AnalysisAssociation of the presence of Anaplasma spp. with host species and presence of tick was performed with Epi Info 6.01 (CDC, Atlanta), using the χ2 test and Fisher’s exact test.3. Results3.1. Tick InfestationOf the 433 small ruminants examined, 190 (43.9%) were infested with at least one tick species. A total of 1475 adult ticks (449 females, 1026 males) were collected from goats (1409/1475, 95.5%) and sheep (66/1475, 4.5%). Six tick species were identified among all collected ticks. Rhipicephalus bursa (1269/1475, 86%) was the dominant tick species, followed by R. turanicus (98/1475, 6.6%), Dermacentor marginatus (94/1475, 6.4%), Hyalomma marginatum (8/1475, 0.5%), R. sanguineus s.l. (5/1475, 0.3%), and Ixodes ricinus (0.06%, only one specimen). The goats were infested with all the identified tick species, whereas sheep were infested with R. bursa and R. turanicus.3.2. Prevalence and Distribution of Anaplasma spp.The prevalence of A. phagocytophilum and related variants in sampled goats and sheep is presented in Table 2. Overall, 121/433 (27.9%) samples collected in studied regions tested positive for Anaplasma spp. by 16S rRNA PCR. The infection rate in goats and sheep was determined as 28% and 27.7%, respectively. RFLP revealed the prevalence of A. phagocytophilum and A. phagocytophilum-like 1 as 1.4% and 26.5%, respectively. No PCR amplicons derived from goats and sheep were digested by the BsaI enzyme, confirming the absence of Chinese variant (A. phagocytophilum-like 2). Of the 121 positive samples with 16S rRNA PCR, 110 (95.6%) were positive with groEL nested PCR. Six of them (6/110, 5.4%) were positive for A. phagocytophilum and 104 (94.5%) were positive for A. phagocytophilum-like 1 (Table 2).Association of the frequency of A. phagocytophilum and A. phagocytophilum-like 1 variant in small ruminants with species and tick infestation is documented in Table 3. The prevalence of Anaplasma spp. was comparable in species, and no difference was detected between infection rates in sheep and goats (p = 0.9603). However, the prevalence significantly increased with tick infestation in small ruminants (p = 0.0003) (Table 3).3.3. Molecular and Phylogenetic AnalysesTo validate the RFLP results and identify genetic variants of A. phagocytophilum-like 1, randomly selected 10 representative samples were sequenced. The sequences shared 100% identity to each other. Therefore, one representative sequence for A. phagocytophilum-like 1 was submitted to the NCBI GenBank database, and deposited with accession number MT881655. This finding indicated that one variant was identified, and named as Aplike1OvineCaprine in this work. BlastN analysis demonstrated that the Aplike1OvineCaprine variant indicated high similarity (99–100%) to those Anaplasma isolates deposited in the GenBank as uncultured Anaplasma sp. and A. phagocytophilum. Moreover, the Aplike1OvineCaprine variant was 100% identical to those of A. phagocytophilum-like 1 detected in sheep (Aplike1Ov1, KX702978) and goat (Aplike1GGo2, KM285227) from Tunisia, and cattle from Turkey (Aplike1Bv, MT338494) (Table 4). The A. phagocytophilum Akseki11 Sheep Turkey isolate obtained in this study shared 99.3–99.6% identity isolated from Niviventer confucianus (A. phagocytophilum ZJ-HGA strain, DQ458805) and human (A. phagocytophilum HZ strain, NR_074113), respectively.Phylogenetic analysis using the 16S rRNA gene showed that our variant (Aplike1OvineCaprine) clustered a distinct group with those of A. phagocytophilum-like 1 previously published sequences reported in sheep, goats, cattle, deer, and Haemaphysalis ginghaiensis (Figure 2).4. DiscussionAnaplasma phagocytophilum causes tick-borne fever in small ruminants and granulocytic anaplasmosis in horses and dogs [1,2]. It is an emerging tick-borne pathogen for humans as well [3]. The genetic diversity of A. phagocytophilum is much greater than expected. Indeed, recent studies have revealed the existence of two distinct Anaplasma species or variants related to A. phagocytophilum, one in Japan and the other in China [11,12,13,14,24]. Then, these pathogens were designated as A. phagocytophilum-like 1 and A. phagocytophilum-like 2 variants [7,8]. More recently, both genotypes have been documented in ruminants and R. turanicus in Tunisia [7,8,17,29], cattle in South Korea [16], and small ruminants in Italy [17]. In the present study, a survey was carried out to detect and identify A. phagocytophilum and A. phagocytophilum-like variants in small ruminants from the Mediterranean region of Turkey. Our findings provide molecular evidence for the presence of A. phagocytophilum and A. phagocytophilum-like 1 in sampled sheep and goats. In the previous studies carried out in Turkey, A. phagocytophilum has been reported in small ruminants [18,21,30]. However, this is the first time that A. phagocytophilum-like 1 variant in sheep and goats have been reported in the country.Contrary to A. phagocytophilum, it has been suggested that both Japanese and Chinese variants do not cause clinical infection in ruminants [8,17]. In this study, a high prevalence for A. phagocytophilum-like 1 variant was determined (26.5%), but no clinical infection for tick-borne fever was observed in sheep and goats during sample collection. This result is consistent with the previous suggestions that A. phagocytophilum-like variants are considered non-pathogenic for ruminants [8,16,17]. The prevalence of A. phagocytophilum-like 1 (26.5%) in small ruminants obtained in this study was higher than that observed in Tunisian sheep (7%) and goats (13.1%) [8], however, it was lower than that observed in other studies conducted in Mediterranean small ruminants (122/203, 60%) from Tunisia and Italy [17].It has been previously suggested that serological cross-reactions occur between A. phagocytophilum and other Anaplasma species [31,32]. The same situation may be true in some circumstances for molecular markers, for example a pair of primers (SSAP2f/SSAP2r) based on the 16S rRNA gene of A. phagocytophilum were designed for the specific amplification [10]. However, it has been shown that these primers also detect distinct Anaplasma variants related to A. phagocytophilum [7,8,24]. In this work, the frequency of pathogenic A. phagocytophilum was 1.4%, which is not consistent with the previous studies in Turkey that reported values of 66.7% in Central Anatolia [30] and 19.7% in Eastern Anatolia [21]. The high infection rates obtained in the previous studies may be due to the selected primers for the amplification of A. phagocytophilum. EE1/EE2 and SSAP2f/SSAP2r primers have been selected to detect A. phagocytophilum in the studies conducted in Central Anatolia [30] and Eastern Anatolia [21], respectively. However, the EE1/EE2 primers are universal for the detection of all Anaplasma spp. including A. phagocytophilum-like variants [33]. It has been also reported that the SSAP2f/SSAP2r can amplify not only A. phagocytophilum, but also A. phagocytophilum-like variants [7,8,24]. This study provides molecular data for the circulation of A. phagocytophilum and A. phagocytophilum-like 1 Turkish small ruminants. Therefore, cross-reactivity between A. phagocytophilum and related variant should be considered in interpreting the findings of surveys to be carried out in the area, where A. phagocytophilum and A. phagocytophilum-like variant co-exist.As several domestic and wild mammals are hosts or reservoirs for A. phagocytophilum [1,2], abundance and intensity of the tick vector, I. ricinus in Europe including Turkey are considered a major factor affecting the distribution of the pathogen in a specific area. It is well known that there is no I. ricinus in the Eastern and Central Anatolian regions of Turkey [34]. It has been reported that A. phagocytophilum is transmitted by Ixodes spp. (I. persulcatus, I. scapularis and I. ricinus) in some parts of the world including in Europe [1,35]. In Turkey, I. ricinus collected from humans were positive for A. phagocytophilum [5]. So far, data on the transmission of A. phagocytophilum-like variants by ticks are lacking. A recent study reported that R. turanicus was common in sampled sheep and goats in Tunisia, and one R. turanicus tick feeding on the goat was found to be infected with A. phagocytophilum-like 2 [28]. In the present study, potential vectors of A. phagocytophilum-like 1 was not studied, but we found that the sampled sheep and goats were commonly infested with R. bursa (86%), R. turanicus (6.6%), D. marginatus (6.4%), and very rarely I. ricinus (0.06%, only one specimen). Our finding also showed a correlation between Anaplasma positivity and the presence of ticks (p = 0.0003), compatible with the finding that the prevalence of A. phagocytophilum-like 1 was higher in goats infested by ticks than in not infested [7]. Based on the abundance of Rhipicepahlus and Dermacentor ticks and the very rarity of I. ricinus, we can assume that Rhipicephalus and Dermacentor may play an important role in the transmission of A. phagocytophilum-like 1 rather than I. ricinus. This assumption is supported by the previous findings that a high prevalence of A. phagocytophilum-like variants have been reported in ruminants in the higher semi-arid area of Tunisia, where I. ricinus is not present [8]. However, more detailed studies are needed to validate this assumption and to establish what tick species may play a role in the transmission of A. phagocytophilum-like 1 in Turkey.Our sequencing validated RFLP findings, and showed that the sampled small ruminants were found to be infected with A. phagocytophilum-like 1. Phylogenetic analysis indicated two main separate branches. The Aplike1OvineCaprine (MT881655) variant obtained in this study, as well as those previously reported from sheep, goats, cattle, and ticks, formed a monophylogenic clade distinct from A. phagocytophilum and A. phagocytophilum-like 2, and other members of Anaplasma spp. [7,8,24].5. ConclusionsThis work provides molecular data for the circulation of A. phagocytophilum-like 1 for the first time in Turkey. The novel strain is widespread in small ruminants in the Mediterranean area of Turkey with an overall prevalence of 26.5%. This finding revealed that the variant should be considered in the diagnosis of caprine and ovine anaplasmosis.

animals : an open access journal from mdpi

10.3390/ani11041144

PMC8073855

The purpose of this study was to detect polymorphism in thyroid hormone-inducible hepatic protein gene (THRSP) and analyze its influence on the fatty acid composition of milk in Jersey and Holstein-Friesian cattle. One single nucleotide polymorphism (SNP) was detected and determined in 224 cows. It was demonstrated that the analyzed variant had a significant influence on several fatty acids content in milk. Obtained results could be applied in breeding programs for improving the quality of milk.

Thyroid hormone-inducible hepatic protein is involved in the de novo synthesis of fatty acids in the lactating mammary gland. Different variants of the gene that encodes this protein may be associated with its different activity. The primary aim of this study was to find polymorphism in the THRSP gene and estimate the relationship between individual genotypes and fatty acid composition in milk. Investigations were carried out on 224 cows represented by two breeds—Jersey (n = 80) and Polish Holstein-Friesian (n = 144). Polymorphism in THRSP was detected by Sanger sequencing; however, genotypes were determined by the PCR-RFLP method. It was shown that the analyzed variant had a significant (p < 0.05) influence on palmitic and stearic fatty acids as well as on fatty acids with a chain length of 14, 16, and 6–16 in Jersey breed and on caproic, palmitic, myristoleic, and palmitoleic fatty acids in H-F. Obtained results indicated that analyzed SNP in bovine THRSP gene (rs42714482) may be considered as a potential marker for fatty acid composition in milk

1. IntroductionMilk fat is a source of fatty acids in the human diet, which can be beneficial for health or can be associated with the risk of some diseases [1]. The fatty acid profile can be modified by many factors; among them, environmental and genetic seems to be most important. Whereas environmental conditions can be managed by farmers if the genetic background is more complex to handle. First, we need to know which genes and their variants are correlated with fatty acid composition in milk. Therefore, it is important to analyze candidate genes for this trait, which can be typed based on the physiological role of the encoded protein, QTL mapping or GWAS.Thyroid hormone-inducible hepatic protein gene (THRSP) encodes Spot14 (S14) protein, which is associated with regulation of the de novo fatty acid synthesis in the liver, adipose tissue, and lactating mammary gland [2]. It was shown that overexpression of THRSP in bovine mammary epithelial cells increased triacylglycerol levels and enhanced the expression of following lipogenic genes: fatty acid synthase (FAS), peroxisome proliferator-activated receptor γ (PPARγ) and sterol regulatory element-binding protein 1 (SREBP1) [3]. THRSP gene polymorphisms were investigated in many domestic animals. Analysis indicated that the THRSP gene in chicken is duplicated and 2 forms are present—THRSPα and THRSPβ. In both paralogs, insertion-deletion (indel) polymorphism was found. It was shown that THRSPα variants were correlated with the deposition of abdominal fat in chickens [4]. In another study, one SNP was detected in chicken THRSPα which together with earlier analyzed indel were associated with growth and body composition traits [5]. Similarly, detected SNP in the goat THRSP gene was studied in relation to growth traits. It was shown that different genotypes were correlated with body weight and chest girth in the Boer goat breed [6]. In pigs, however, SNP found in 5′ proximal regulating region of THRSP was associated with average backfat thickness, average daily weight gain, and loin-eye area [7]. Also in cows, polymorphism in the coding region of this gene was analyzed in relation to meat and carcass traits, as well as fatty acid composition in meat. It was shown that different THRSP genotypes were correlated with water holding capacity and meat tenderness in Qinchuan cattle and with unsaturated, monounsaturated, and few individual fatty acids in the meat of Hanwoo cattle [8,9]. In dairy cattle—Italian H-F, THRSP variants were associated with some milk production traits [10]. Bovine THRSP gene is located on chromosome 29 (Bta29) and consists of 2 exons. The length of the transcript is 1398 bp; however, the protein is composed of 148 aa [11,12]. Although the role of THRSP in the synthesis of fatty acids during lactation, polymorphism in the gene that encodes this protein was not studied in relation to the fatty acid composition of milk. Therefore this study aimed to detect polymorphism in the THRSP gene of Jersey and Polish Holstein-Friesian cattle and to perform association analysis for milk fatty acids profile.2. Materials and Methods2.1. AnimalsThe experiment covered 224 cows that belong to Jersey (n = 80) and Polish Holstein-Friesian (n = 144) breeds. The first group was reared in a tie-stall barn in Greater Poland Voivodeship; however, the second was reared in a free-stall barn in West Pomeranian Voivodeship. Feeding and management of animals on both farms were very similar. Cows were fed by use of a total mixed ration (TMR) diet that contains corn silage, grass haylage, alfalfa silage, straw, solvent-extracted soybean meal, as well as minerals and vitamins. The nutritional composition of feeds applied in the experiment is presented in Table 1. Blood samples were collected into tubes containing K3EDTA during a routine veterinary check-up. Milk samples were collected in lactations 1–4 during trial milking performed by the Polish Federation of Cattle Breeders and Dairy Farmers. The mean day of lactation for Holstein-Friesian cows was 195; however, for Jersey, it was 151.2.2. Polymorphism AnalysisDNA was isolated from peripheral blood by use of MasterPure™ DNA Purification Kit for Blood Version II (Epicentre Biotechnologies, Madison, WI, USA). Following primers pair, covering bovine THRSP exon 1 was designed using Primer3 software [13]: F 5’-GCTGTGTTGACCTACTGGC-3’, R 5’-CGGCCACCATTACCTTTCCT3’. Primers were designed based on the ENSBTAG00000011666 sequence [10]. PCR cycling was as follows: initial denaturation at 94 °C/5 min, 35 cycles of 94 °C/30 s, 61 °C/45 s, 72 °C/30 s, and final extension at 72 °C/5 min. PCR was performed in a total volume of 15 µL that contains 50–80 ng of genomic DNA, 1.5 mM MgCl2, 0.3 mM of dNTP mix, 12 pmol of each primer, and 0.35 U of Taq polymerase (EURx, Gdansk, Poland). The presence of specific amplicons (598 bp) was confirmed in 1.5% agarose gel with Perfect™ 100–1000 bp DNA Ladder (EURx, Gdansk, Poland). Sequencing of amplicons was performed by an external service (Genomed, Warsaw, Poland). PCR-RFLP method was applied to determine detected THRSP gene variants (rs42714482) [14]. The same pair of primers and conditions as mentioned above were used in PCR. A total of 10 µL of obtained amplicons were digested by BstC8I enzyme (SibEnzyme, Novosibirsk, Russia) in 55 °C at least 3 h. Restriction fragments were separation in 4% agarose gels with a 50 bp DNA Ladder (Genoplast, Rokocin, Poland).2.3. Milk Composition AnalysisTo avoid a period of negative energy balance, milk samples were collected after the 90th day of lactation during morning milking, according to PN-EN ISO 5555:2002 standard. Next samples were transported to the laboratory and stored at −20 °C until further analysis. Lipids were extracted from milk by use of chloroform and methanol mixture in a 2:1 ratio. Next fatty acids were converted into methyl esters using boron trifluoride according to PN-EN ISO 12966-2:2011 standard. Analysis of fatty acids methyl esters was performed using gas chromatography mass-spectroscopy method (GC-MC) in agreement with PN-EN ISO 5508:1996 standard. Fatty acids were identified based on their relative retention time in relation to retention times of standard (SupelcoTM 37 Component FAME Mix, Sigma-Aldrich, Saint Louis, MI, USA). Following fatty acids were analyzed:saturated: C6:0 (caproic), C8:0 (caprylic), C10:0 (capric), C12:0 (lauric), C14:0 (myristic), C16:0 (palmitic), C18:0 (stearic);unsaturated: C14:1 (myristoleic), C16:1 (palmitoleic), C18:1n-9c (oleic), C18:1n-9t (elaidic), C18:2n-6c (linoleic), C18:3n3 (α-linoleic).Peaks were analyzed by use of TurboMassTM software (PerkinElmer, Waltham, MA, USA).Additionally, the following indexes were calculated: sum of fatty acids with a chain length of 14 (ΣC14), 16 (ΣC16), 6–16 (ΣC6–16), 18 (ΣC18), Δ9-desaturase index for fatty acid with 14 carbons (Δ9IC14), 16 (Δ9IC16), 18 (Δ9IC18), for monounsaturated fatty acids (Δ9MUFA), saturated fatty acids (SFA), unsaturated fatty acids (UFA), monounsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFA), UFA/SFA ratio, atherogenic index (AI) and thrombogenic index (TI). The Δ9 desaturase and atherogenic/thrombogenic indices were calculated using the formulas proposed by Lock and Garnsworthy [15] and Ulbricht and Southgate [16].2.4. Statistical AnalysisStatistical analysis was conducted using R packages [17]. Pedigree data were arranged in Pedigree Viewer, ver. 6.5 (University of New England, Armidale, Australia) [18]. An additive relationship matrix based on a three-generation pedigree was generated using the kinship2 R package (Rochester, MI, USA) [19]. To estimate associations between individual THRSP genotypes and analyzed traits following mixed linear model was estimated and applied using the lmekin function from the coxme R package (Rochester, MI, USA) [20]:Y = µ + G + LS + β1A + β2DLC + α + e where: Y—the value of the analyzed trait, µ—overall mean, G—fixed effect of THRSP genotype, LS—fixed effect of lactation number and season, β1A—regression coefficient for the age of cow, β2DLC—regression coefficient for the day of lactation when milk was collected, α—random polygenic effect taking into consideration pedigree relationships, e—the random error. Statistical significance of results was indicated below a p-value of 0.05 (p < 0.05).3. ResultsSequencing of bovine THRSP exon 1 allowed detection of one missense SNP located in position 193 of the transcript (152 in cds) (Figure 1). This C/T substitution led to alanine to valine exchange in position 51 of the protein.Restriction analysis showed that the BstC8I enzyme can differentiate detected variants in the THRSP gene. Individual genotypes were determined based on the following restriction fragments lengths: TT—205, 164, 124, 105 bp, CT—205, 164, 124, 116, 105, 89 bp, CC—164, 124, 116, 105, 89 bp (Figure 2).The distribution of THRSP genotypes and alleles with Hardy-Weinberg equilibrium (HWE) in both analyzed breeds of cows are presented in Table 2. It was shown that TT genotype was most frequent in Jersey (0.33) group; however, CC was most frequent in Polish H-F (0.46). In the case of alleles, reverse tendency was observed where T was the major allele in Jersey cows (0.58), while C in H-F (0.68). Analysis showed that the distribution of genotypes was in agreement with the HWE expectation. Frequencies of individual genotypes were significantly different between Jersey and Polish Holstein-Friesian breeds (χ2 = 29.442, p < 0.01)Fatty acid composition of milk with some indexes in relation to THRSP variants is shown in Table 3 and Table 4. In Jersey cows, statistically significant differences (p < 0.05) were found between individual genotypes and following milk fatty acids: palmitic, stearic as well as following indexes: fatty acids with a chain length of 14, 16, and 6–16. In Polish H-F cows, associations (p < 0.05) were found for caproic, palmitic, myristoleic, and palmitoleic fatty acids in milk. In both breeds, only one trait was common as correlated with THRSP genotypes—palmitic fatty acid. Jersey cows with TT genotypes were characterized by the highest value; however, Polish H-F with the same variant had the lowest.4. DiscussionTHRSP is a nuclear protein that can regulate lipogenesis [21]. It was found that THRSP may regulate milk fat synthesis by directly affecting the activity of some classical lipogenic enzymes [3]. Recently, Salcedo-Tacuma et al. [22] indicated THRSP as an inhibitor of lipid synthesis in adipose tissue of periparturient Holstein cows.THRSP gene is involved in de novo fatty acid synthesis. The fatty acids that are synthesized de novo belong to short-chain and medium-chain length acids, from C4 to C14 and also some C16. The C18 fatty acids and some C16, however, arise from the plasma lipids [23]. In our study, we analyzed fatty acids from C6 to C18. We decided to include C18 fatty acids because an earlier report showed a significant decrease in C18:0 upon THRSP overexpression in goat mammary epithelial cells [24].THRSP gene in Polish H-F and Jersey breeds in relation to milk fatty acids composition was analyzed for the first time. In a previous study, Oh et al. [9] selected 8 SNPs in the bovine THRSP gene from GenBank [13] for analysis in Korean cattle (Hanwoo). Only two of them, namely g.78 G > A and g.184 C > T were polymorphic in this breed. Comparison of sequences from Ensembl [11] and GenBank [14] showed that the g.184 C > T variant is the same as detected in our work (29:17994763 C/T, rs42714482). Analyzed polymorphism causes amino acid substitution in position 51 of the encoded transcription factor, which probably may modify its interaction with regulated genes and, thus, differentiate processes connected with fat milk synthesis, including the composition of individual fatty acids in milk. We found that the frequency of individual genotypes in Korean cattle (CC—0.17, CT—0.46, TT—0.37) was very similar to those observed in our study in Jersey breed. Evolutionary analysis of SNPs located in genes of Hanwoo cattle in relation to data from other cattle breeds, i.e., Jersey, Simmental, Angus, and Holstein showed that the Korean breed was distinctly separated from the other four breeds. Further SNPs analysis, however, showed that the THRSP gene is not classified as Hanwoo-specific, which may reflect similar genetic parameters in Korean and Jersey breeds [25]. When comparing the frequency of THRSP genotypes in Italian Holstein-Friesian cattle (CC—0.48, CT—0.42, TT—0.10), the same was observed in our work for Polish Holstein-Friesian breed [10].As mentioned earlier, the THRSP gene was not investigated in relation to milk fatty acids. It was studied, however in relation to carcass traits and fatty acid composition of muscle fat in Korean Cattle, as well as to health and milk production traits in Italian Holstein cattle. It was shown that g.184 C > T SNP is significantly correlated with myristic, palmitic, myristoleic, oleic, linoleic, linolenic fatty acids in m. longissimus dorsi, as well as with saturated fatty acids, monounsaturated fatty acids, and monounsaturated/saturated ratio [9]. In our study, we found an association between THRSP genotypes and palmitic fatty acid in milk of Jersey and Polish H-F cows and myristoleic fatty acid in HF cows. In the muscle of Hanwoo and the milk of Jersey cattle, the highest content of palmitic fatty acid was found for the TT genotype. In the Polish H-F breed, however, a reverse tendency was observed, because the TT genotype was correlated with its lowest value. Similarly, TT genotype was favorable for myristoleic fatty acid in the muscle of Hanwoo cows but in the milk of Polish Holstein-Friesian cows, it was disadvantageous. These differences may result from fatty acids content in milk of particular cattle breeds. A significant breed effect on the content, fatty acid profile and atherogenic or thrombogenic properties of milk fat was reported by Sobotka et al. [26]. The concentrations of long-chain saturated fatty acids were significantly higher in the milk fat of Jersey cows than in the H-F breed. The milk of H-F cows had lower fat content but provided more health benefits than the milk of Jersey. It was also confirmed by the analysis of Jersey, H-F, and three other breeds reared in the Netherlands. Breed differences were found for individual fatty acids, among them for palmitic and myristoleic [27]. We cannot also exclude that observed differences in association tendency may be linked with different rearing conditions for both breeds. We found also an association of TT genotypes with higher content of caproic and myristoleic fatty acid in Polish H-F and with lower stearic fatty acid in Jersey, but this relationship was not observed in Hanwoo cattle. Analysis of THRSP gene in Italian Holstein cows also showed a significant association for milk yield, milk fat yield, milk protein yield and productivity, functionality, and type index, with allele T being favorable for these traits [10]. In our study, we did not find any relationships between individual THRSP genotypes and milk yield, fat yield, and fat content both in Polish Holstein-Friesian and Jersey breeds. Similarly, THRSP genotypes were not found to influence the atherogenic and thrombogenic indexes. As mentioned earlier, we also noticed significant associations for stearic, caproic, palmitoleic fatty acids and fatty acids with a chain length of 14, 16, and 6–16. Stearic acid is common in nature, both in animal and vegetable organisms; however, its level is usually higher in animal than vegetable fat [28]. In milk, it arises from the plasma lipids, but interestingly, the highest expression of THRSP is correlated with its decrease in goat mammary epithelium [24]. In the Jersey breed, TT genotype was associated with the lowest level of stearic acid in milk. Caproic, caprylic, and capric fatty acids are known as the reason for the specific aroma of goat and sheep milk [29]. Analysis of caproic acid amount in the milk of different species showed its highest level in goat milk, lowest in sheep; however, in cow milk, it was slightly higher than in sheep [30]. In Polish Holstein-Friesian cows, the TT variant of THRSP related to the highest content of this fatty acid. Palmitoleic acid shows anti-inflammatory and antidiabetic activity and is produced by desaturation of palmitic acid [31]. In Polish H-F animals we found that TT genotype is correlated with the lowest amount of palmitoleic, as well as earlier mentioned palmitic fatty acid, which can reflect their metabolic relationships. In the case of confirmed associations of THRSP polymorphism and fatty acids with a chain length of 14, 16, and 6–16 in Jersey cattle, we obtained unclear results. The highest value of the first trait was found for the CC genotype; however, the second and third traits were found for the TT genotype.5. ConclusionsThe conducted experiment showed that THRSP polymorphism (rs42714482) is associated with milk fatty acid composition in Jersey and Polish Holstein-Friesian cattle. It covers palmitic, stearic caproic, myristoleic, and palmitoleic fatty acids, as well as fatty acids with a chain length of 14, 16, and 6–16 in a particular breed. Only palmitic acid content was the common trait for both breeds, however, with opposite tendency, which may reflect breed differences or/and feeding and housing conditions. Analyzed SNP in bovine THRSP gene could be taken into consideration as a potential marker for fatty acid composition in milk.

animals : an open access journal from mdpi

10.3390/ani11040995

PMC8066114

Limited information exists on the physiological changes that occur in the horses competing in endurance races. The objective was to provide the initial data describing changes in laboratory measurements of the horses competing in endurance races under temperate conditions and to compare the data between the Arabian horses, which are one of the most popular horse breeds in the world, and a Lithuanian horse breed—Žemaitukai. The study was carried out on 112 horses. Blood samples were collected before and after an endurance race. The Arabian horses were faster compared to the local breed (Žemaitukai). The study showed significant changes in horse blood gasometrical and biochemical indicators.

Fédération Equestre Internationale (FEI) has described equine endurance racing as the second largest discipline in the world, above which is only show jumping. The Žemaitukai is an ancient indigenous Lithuanian horse breed known since the 6th or 7th century. The Arabian horse breed is one of the oldest human-developed horse breeds in the world. Compared with other race horse breeds, the muscle tissue of Arabian horses is characterized by significant differences in structure—a predominance of oxidative fiber type I is observed in Arabians, making them the prevailing breed in endurance racing. The Arabian horses are recognized as the leading breed in endurance competitions. Speed, pace, and total time in the race strategy have been extensively studied in human sports, and in contrast, this strategy appears to have been virtually ignored in equestrian sport, despite the potential for contributing to performance optimization. In relation to speed and total time in the race, there are limited data on postrace physical, biochemical, and blood gas parameters of endurance horses. Thus, this study was carried out to investigate the effects of speed on the blood parameters of the Arabian and Žemaitukai horses during an endurance race. Blood samples were taken before and immediately after the exercise. Biochemical and blood gas indicators were analyzed. The study showed significant increases in mean blood gasometrical indicators, such as partial carbon dioxide pressure (8.09–15.18%, p < 0.001); base excess in the extracellular fluid (14.01%, p < 0.001 in the Arabian horses and 172.01% in the Žemaitukai breed, p = 0.006); decreases of the blood electrolyte ionized calcium (4.38–8.72%, p < 0.001) and the hematocrit and hemoglobin values (20.05–20.12%, p < 0.001 in the Arabian horses and 6.22–6.23% in the Žemaitukai breed, p = 0.003–0.004); and decreases in the base excess in the blood values (29.24–39.38%, p < 0.001) and lactate (13.45–31.97%, p < 0.001) in the blood of both breeds in the post-competition horses. Significant increases after competition were determined for the values of creatinine (21.34–30.82%, p = 0.001–0.004), total bilirubin (50.84–56.24%, PH < 0.001), and albumin (2.63–4.48%, p = 0.048–0.001) for both breeds. For the faster Arabian horse breed, recovering after racing took half the time that the local Žemaitukai breed did.

1. IntroductionFédération Equestre Internationale (FEI) has described equine endurance racing as the second largest discipline in the world, above which is only show jumping [1]. Endurance races are described as long-distance competitions (40–160 km) that are organized into loops, which are arranged over variable terrain, and as a rule, these competitions are completed in one day. In endurance competitions, horse and rider combinations are required to complete the racecourse in good condition, aiming to win [2]. Endurance can be described as an exercise performed at a moderate speed whilst covering a long distance [3]. A horse is an interesting physiological model in this context, because different breeds can be used in all types of physical exercise. For example, the Arabian breed is recognized as well adapted to endurance racing, because the Arabian horses are able to run at an average speed of 20 km/h or greater for up to 160 km (in bouts of 30–40 km) [4]. This level of performance is based on aerobic metabolism, adaptation of the cardiorespiratory system, effective body heat dissipation, and maintenance of homeostasis. The Arabian horse breed is one of the world’s oldest breeds. This breed is distinguished by its natural beauty, graceful movement, and athletic endurance. Since Arabian horses originated from the Middle East, they feature the unique ability to thrive in a hot, dry environment [5]. The standard height of the Arabian horse is from 145 to 155 cm (standing) [6]. The standard weight is between 360 and 450 kg; chest girth—113.2 cm; body length—145 cm; cannon bone girth—18.5 cm. Compared with other race horse breeds, the muscle tissue of Arabian horses is characterized by significant differences in structure—the predominance of oxidative fiber type I is observed in Arabians, [7,8]. In an Arabian horse’s muscle, the higher proportion of oxidative type I fibers (characterized by a low glycogen content and high triglyceride storage capability) results in a greater use of fat for energy [9]. Due to their exceptional qualities, Arabian horses have been extensively used for over 100 years by horse breeders for the improving of other horse breeds [10]. Arabian horses are often used not only for refinement of other horse breeds, but also for their endurance abilities [11]. The studies of the phenotypic traits of the Žemaitukai showed the average body measurements were: wither height—133.6 cm; chest girth—173.2 cm [12]; body length—142.2 cm; cannon bone girth—17.7 cm; weight, 360–420 kg [13]. The Žemaitukai is a pony by height and type [14]. However, the wide chest, high indices of extension, and massiveness indicate that these horses might have substantial draughting power. Due to thin, strong legs, a low bony index, round hips, a wide trot, good jumping technique, energetic temperament, and mobility, the Žemaitukai horses are considered as very suitable for quick trotting and jumping. Thanks to the aforementioned qualities, the Žemaitukai breed is highly valued and used as a versatile horse breed [13]. Based on the results obtained in our last study, it was stated that the horses of the Žemaitukai breed are suitable for endurance competing [15]. To maximize performance during a race, an athlete has to regulate speed over the entire course of a race. The distribution of energy expenditure during the race is defined as “pacing strategy.” Pacing strategy is considered to be a key factor determining overall performance in endurance racing. It is considered that the best strategy in a race is maintaining an even pace, and sometimes gradual declining in speed [16]. The aim of every athlete is to reach a certain distance in the shortest time possible [17]. In endurance athletes, an increased aerobic capacity allows skeletal muscle to metabolize more fat (use energy from fatty acids), while at the same time using carbohydrates as an energy source [18]. An aerobic conditioning program in endurance-type athletes induces an increase in the activity of oxidative metabolism and a decrease in anaerobic metabolism. The adaptive response to exercise is associated with changes in gene expression, metabolism, muscle cell cycle progression, and protein homeostasis. However, the exact mechanisms that occur in equine muscles during exercise related to skeletal muscle endurance in high-intensity training are not well understood [19]. Extensive studies of speed, pace, and total time in racing strategy have been carried out in human sports, whereas in equestrian sport, pacing strategy has been practically ignored, although it is one of the key factors to be analyzed when seeking to optimize horse performance [10]. As regards speed, data on postrace physical, biochemical, and blood gas parameters of endurance horses are limited. Therefore, the aim of this study was to investigate the impacts of breed on speed and the blood parameters of the Arabian and Žemaitukai horses during an endurance race.2. Materials and Methods2.1. Location, Animals, and Experimental DesignThe research was carried out in accordance within the provisions of the Law of the Republic of Lithuania—order number 8–500 on the protection, keeping and use of animals, of 6 November 1997 (the Official Gazette “Valstybės žinios” number 108–6595, dated 28 November 1997), order number 4–361 of 31 December 1998 of the State Veterinary Service of the Republic of Lithuania on breeding, care, transportation of laboratory animals, and order number 4 of 18 January 1999 of the State Veterinary Service of the Republic of Lithuania on the use of laboratory animals for scientific tests. The study approval number was PK012868. The study was conducted on 60 (27 female and 33 male) clinically healthy Arabian horses and 52 (34 female and 18 male) clinically healthy Žemaitukai horses at six endurance competitions (45 km races) in Lithuania and at the Lithuanian University of Health Sciences Veterinary Academy. All horses were 10 ± 4 years old, with an average body weight of 408 ± 41 kg. The horses were transported from different locations, having been delivered to the competitions at least two hours prior to their first veterinary examinations. Veterinary inspections were carried out according to FEI regulations. After each loop of the races, all competing horses underwent veterinary inspections, and the physical parameters of each horse were recorded. All horses successfully passed veterinary inspections, which were performed prior to the start of the race, during the race, and after the race. For a horse to be considered fit enough to continue the event, its HR must be below 65 bpm within 20 min of arrival. The criterion for inclusion of horses was successful completion of the race, and only the finishers were included in the further research. All horses were trained for endurance racing and had participated in similar competitions before; however, such factors as the degree of competition experience, the number of competition kilometers completed, and the number of successful races over the recent period were different for each animal. All researched horses were dewormed and vaccinated at a similar time and were not receiving any medications and/or suffering from any infections in the preceding three weeks. All animals were housed in a similar environment and fed a similar diet with adequate mineral and vitamin supplementation; the amounts of salt and water were not limited in the diet.2.2. Analytical ProceduresBlood samples were taken from each animal by applying the technique of jugular venipuncture with 1.6 mL heparinized vacutainer blood collection tubes for blood gas (Terumo Europe, Belgium) and a 5.0 mL tube without anticoagulant for serum biochemistry (Terumo Europe, Belgium), 30 min before the start and no later than 30 min after the finish. None of the horses showed stress during the blood sampling, which was carried out in less than 30 s for each sample. The samples were taken immediately after the first and last veterinary inspections; then they were identified and stored in an ice bath for a maximum period of 30 min until processing. Using EPOC blood gas analyzers (EPOC, Canada, Ottawa), the following indices were analyzed: hydrogen potential (pH), partial carbon dioxide pressure (pCO2, kPa), partial oxygen pressure (pO2,kPa), base excess in the blood (BE, mmol/L), base excess in the extracellular fluid (BE ecf mmol/L), bicarbonate (HCO3, mmol/L), oxygen saturation (sO2, %), total carbon dioxide in the blood (tCO2, mmol/L), hematocrit (HCT, L/L), hemoglobin concentration (Hgb, mmol/L), glucose (Glu, mmol/L), sodium (Na+, mmol/L), potassium (K+, mmol/L), calcium (Ca++,mmol/L), and lactate (Lac, mmol/L).Serum urea (Urea, mmol/L), glucose (Glu, mmol/L), aspartate aminotransferase (AST, U/L 1), iron (Fe, umol/L), creatinine (CREA, µmol/L), magnesium (Mg, mmol/L), total protein (TP, g/L), total bilirubin (TB, µmol/L 1), albumin (ALB, g/L), cuprum (Cu, umol/L), zinc (Zn, umol/L), and gamma-glutamyl transferase (GGT, U/L) were measured using an automated analyzer, Hitachi 705 (Hitachi, Japan).Horse speed and recovery time data were taken from the electronic time-keeping system (ECR v.7.01 Systems, Kaunas, Lithuania) designed for endurance events. Speed data were calculated by dividing the length of the course by the time taken for the horse to complete the course (subtracting hold times). Recovery time is the time that horse spends in the recovery area after crossing the end line of a loop until it crosses the line into the vetting area to be presented for its horse inspection.The races were held in Lithuania, in accordance with the FEI rules. The competitions were carried out on variable terrain with some muddy and some firm areas, and with slight elevation changes (±300 m). The environmental conditions during the competitions varied, with the mean temperature being 22.5 °C (within the range of 12.50–26.5 °C) and the mean relative humidity being 73.20% (within the range of 52–96%). For a horse to be considered fit enough to continue the event, its HR must be below 65 bpm within 20 min of arrival.2.3. Data Analysis and StatisticsThe data were analyzed using the IBM SPSS Statistics v20.5 for Windows. The distributions of the evaluated traits were used to carry out the assessment according to the Kolmogorov–Smirnov test. The mean (M) value and standard error of the mean (SE) were calculated. To compare blood parameters before and after competition, a paired t-test was used; to compare the differences between the breeds before and after competition, the Student’s t-test for independent samples was employed. The relationships between speed and recovery time and the blood parameters of a horse were assessed using the Spearman method. The results were considered to be significant at p < 0.05.3. Results3.1. The Indices of Biochemical Parameters of the Arabian and Žemaitukai Horses before and after CompetitionThe analysis of biochemical parameters (Table 1) showed significant increases in the values of CREA (21.34–30.82%, p = 0.001–0.004), TB (50.84–56.24%, p < 0.001), and ALB (2.63–4.48%, p = 0.048–0.001) after competition in both breeds.Before and after competition, significantly higher levels of Fe (p = 0.001 and p < 0.001), CREA (p < 0.001), TP and TB (p = 0.023 and p < 0.001), and ALB (p < 0.001) were observed in the Arabian horses compared to those in the Žemaitukai breed, for which higher AST (p < 0.001), Cu (p = 0.002 and p < 0.001), and GGT (p < 0.001) were found.3.2. The Indices of Acid–Base Balance in the Arabian and Žemaitukai Horses before and after CompetitionThe study showed significant decreases in mean blood gasometrical indicators, such as pCO2 (8.09–15.18%, p < 0.001) and BE (efc) (14.01%, p < 0.001 in the Arabian horses and 172.01% in the Žemaitukai breed, p = 0.006) and in the blood electrolyte Ca ++ (4.38–8.72%, p < 0.001); and there were increases in HCT and Hgb (20.05–20.12%, p < 0.001 in the Arabian horses and 6.22–6.23% in the Žemaitukai breed, p = 0.003–0.004), BE (b) (29.24–39.38%, p < 0.001), and Lac (13.45–31.97%, p < 0.001) in the blood of both breeds in the post-competition horses compared to those measured before competition (Table 2).3.3. Speed and Recovery Time by BreedThis section describes the speed and recovery time after competition for the Arabian and Žemaitukai horses, and the relationships between speed and blood indices. The study showed that based on electronic time-keeping system data, for the faster Arabian horse breed, recovery time was twice as fast, compared to the local Žemaitukai breed (p < 0.001) (Table 3).It was found that of all blood biochemical parameters of the Arabian breed, Fe before competition most positively correlated with the speed of horse (r = 0.675, p < 0.01) and GGT (r = 0.600, p < 0.01) after competition. Most negatively correlated were Ca, Mg, TB, and Cu (r = 0.600–0.900, p < 0.01) before competition; and Fe, CREA, TP, ALB, and GGT (r= 0.600–0.996, p < 0.01) after competition (Figure 1). The recovery time of Arabian horses was mainly positively related with blood urea and Zn (r = 0.600–0.800, p < 0.01) and negatively correlated with Mg, Cu, and GGT in the blood (r = 0.500–0.975, p < 0.01) before and after competition (Figure 1).Analysis of the correlations between the speed of the Arabian horses (Figure 1E) and the indicators of acid-base balance showed positive correlations with pH, pO2, cSO2, HCT, and Hgb (p < 0.05) and negative correlations with pCO, BE (efc), Na+, K+, Ca++, tCO2, Glu, and Lac (p < 0.05) before competition. Post-race blood tests for HCO3, Ca++, and tCO2 showed positive correlations of these parameters with the speed of horse, whereas pH, Na+, K+, HCT, Hgb, and BE (b) negatively correlated with the speed of horse (p < 0.05).Based on the correlation analysis of biochemical parameters of the Žemaitukai horses (Figure 2), it can be concluded that p before competition and TB after competition were mainly negatively related with the speed of horse (p < 0.01). It was also found that the recovery time of this breed (Figure 2) has a negative significant correlation with GGT before competition, and with Mg and p after competition (p < 0.01).The indicators of acid–base balance of the Žemaitukai breed showed the highest positive correlation of horse speed with blood pH before competition, and with Ca++ (r = 0.900, p < 0.01), HCO3, and tCO2 after competition (r = 0.800, p < 0.01). The highest negative correlations of horse speed were with pCO and Lac before competition (r = −0.800–0.802, p < 0.01) and with HCT and Hgb after competition (r = −0.670–0.672, p < 0.01) (Figure 2).4. DiscussionIn the study, significant differences between the pre- and postrace blood parameters of the Žemaitukai and Arabian endurance horses were found. Pre-race blood samples showed lower hematocrit values; however, they still fell within the reference range [23]. In postrace blood testing, an increase of hematocrit was observed. An insignificant increase in hemoglobin was found in postrace blood samples. These findings were described in previous studies [24,25]. Splenic contraction induced by adrenergic stimulus and sweating, causing extensive body fluid losses, has been observed in the conditions of more prolonged exercise [26]. The study revealed relationships between breed and HTC, Hb change, and TP concentration. According to Fan et al. [27], increases in HTC percentage and TP concentration can be indicative of dehydration status occurring in result of the action of xanthine oxidase in free radical production, which in turn is determined by the permeability of muscle cell membrane. Dehydration is normally observed in all horses participating in endurance races. However, dehydration levels may vary. They are directly related to the sweat rate, which is determined by the amount and rate of physical work performed and by the environmental temperature and humidity [28]. Training and heat acclimatization can increase the sweat rate by 10 to 20 percent [29]. Decreased blood volume, high expenditure of energy, and muscular damage can indicate changes in equine biochemical profiles during long endurance racing [30]. Decreased blood volume first of all indicates an extensive loss of body fluid and electrolytes and low intake of fluid. During this study, a negative correlation between the CREA values and horse speed was established for both breeds after competition. Due to muscular metabolism, the final catabolite creatinine is developed, and the final catabolite of endogenous protein breakdown is urea [31]. An elevated concentration of urea can be observed in the horses after prolonged exercise. This was revealed by a study of horses competing in 121 km and 164 km endurance races [32]. This is consistent with the increase in postrace concentration of urea observed in our study. The postrace increase in creatinine concentration was also observed in the horses competing in the 160 km endurance races [33]. Another study revealed that a higher degree of dehydration due to intensified physical effort can result in increased levels of urea and creatinine after the 80 km races [34], implying a significant positive correlation between the above-mentioned parameters. The study showed that CREA levels were increased in both breeds after exercise; however, for the Žemaitukai horses, the rise of creatinine was higher, leading to significant decreases in the urea–CREA ratio. The mechanism for changes in serum creatinine following aerobic training could be related to the theoretical and empirical reports insisting that creatinine concentration is positively associated with body mass index, body fat, and fat distribution [35]. Before and after competition, higher levels of CREA were observed in the Arabian horses, probably as a result of the larger muscle mass [36], although the weights of breeds of horses were similar. A low body fat percentage and a large amount of muscle are of benefit to horses being considered to be elite level endurance racers. Creatinine and urea elevation may also result from a higher metabolic rate [37].The serum iron concentration was significantly reduced in postrace Žemaitukai horses and increased in the Arabian horses. The ability to maintain prolonged submaximal exercise and the activity of iron-dependent oxidative enzymes is sufficiently closely related to tissue iron levels, which are a determining factor for endurance performance during prolonged submaximal exercise [38]. Lack of hemoglobin may significantly affect physical performance, as there will be a decrease in oxygen transport to exercising muscle. Decreases of hemoglobin levels and tissue iron content may have an adverse effect on performance [38]. Iron supplementation has been reported [39] to increase physical performance and motivation and improve efficient energy use in humans involved in various types of physical activities, whereas iron deficiency contributes to reduced aerobic capacity of muscle, and decreased tissue concentrations of nonheme iron can have detrimental effects because it functions as an enzyme cofactor [40]. Serum concentrations of the iron storage protein ferritin have been correlated with performance in nonanemic female marathon runners [41].Exercise did not affect mineral requirements greatly, despite the possible increase in the request for minerals associated with the need for more energy to the muscles (Ca, Mg, and p) and production of saliva and sweat (Na+) [42]. The blood K+ levels decreased after racing in both breeds. Similarly, another study found that the loss of potassium via sweat and renal fluid reabsorption through the kidneys were associated with potassium and hydrogen ion release [43]. Di Filippo et al. (2005) [42] also reported that the Arabian horses had a lower K+ concentration after 60 km of endurance racing [44]. Lowered K+ levels during exercise have a significant negative effect on a horse’s performance [45]. Loss of potassium may result in fatigue, weakness, reduced intestinal motility, and paralytic bowel, and sometimes even lead to altered electrocardiographic traces [46].In this study, a lactate increase after endurance exercise in the Arabian horses was observed. However, decreased blood lactate concentrations might occur due to poor glucose utilization by the metabolizing tissues [47]. Substitution in the metabolic preference for glucose over lipids was the mainstay of increase in the blood lactate levels. Besides, active exportation of lactate from muscles into the blood is possible [48]. Another study found a sudden upsurge of lactate concentration after 15 min of recovery [49]. The horses with high lactate concentrations showed better results than those with low lactate levels, indicating that other mechanisms may be involved in the regulation of blood lactate concentration [50]. Nevertheless, postrace blood lactate concentrations were higher than pre-race ones for all horses; however, they were elevated more in the Arabian horses than in the Žemaitukai breed, probably owing to the significantly higher anaerobic capacity in the Arabian horses. The postrace values of pH, HCO3−, BE, and Hct in both horse breeds were higher than the pre-race ones. Increases in pH values of the horses can be related to the loss of chloride ions via sweat. Due to the body’s need to restore the balance of negative charges, the loss of chlorine via sweat results in retention of the second most abundant ions in the organism, bicarbonate ions (HCO3−) [51]. In turn, the excess of HCO3− triggers hypochloremic metabolic alkalosis [45]. According to Johnson (1995), this alkalosis is an important clinical complication in exhaustion syndrome and in the cases of exertional rhabdomyolysis [52]. To maintain pH homeostasis, the body has three lines of response: chemical buffers, respiratory regulation, and renal regulation [43]. Due to changes in the blood pH, respiratory compensation occurs almost immediately, altering the pCO2 [52]. Over a long run, regulation of the acid–base balance requires excretion of H+ ions and retention of bicarbonate ions by the kidneys [53]. In line with these findings, the results show changes in postrace pCO2 and pO2. These findings indicated respiratory acidosis, which added to metabolic alteration, is called metabolic alkalosis with respiratory compensation. This respiratory modifications is common in sporting animals when faced with metabolic alkalosis [43], and is a reflection of the animals’ organic health. Increased extraction of oxygen from the blood determines the reduction of venous oxygen content [54]. Under resting conditions, oxygen extraction ranges between 20% and 40%. During exercise, approximately 70–80% of the oxygen delivered to the active muscles may be extracted. This demonstrates that there is a reserve of oxygen in the blood that can be utilized immediately to meet the needs of the contracting muscles at the onset of exercise. Increased extraction of oxygen from the blood is driven by the decreases in perivascular PO2, which in turn are driven by the reductions in cell PO2 [9]. The apparent breed difference in aerobic and anaerobic capacity may, in part, reflect breed variation in muscle fiber types and the muscle concentrations and activities of enzymes involved in glycolysis [55]. Excess body fat increases the energy requirements of weightbearing work, such as running, by increasing the energy requirements of exercise for any given intensity of work during a maximal oxygen consumption test [56]. This may be detrimental to running performance in that the running speed that can be sustained for a given duration is reduced [57], thereby increasing race time.5. ConclusionsThis study showed that for the faster Arabian horse breed, recovering after racing took half the time taken by the local Žemaitukai breed. Based on hematocrit changes during exercise, the Arabian horses were found to have a higher sweat rate. Before and after competition, higher levels of creatinine were observed in the Arabian horses, probably as a result of having more muscle mass than Zemaitukai horses. Postrace blood lactate concentrations were higher for all horses; however, they were elevated more in the Arabian horses because of the greater muscle mass and higher activity level of the latter breed. Lower amounts of electrolytes and smaller changes during races showed that the Žemaitukai horses had a lower capacity for heat tolerance, suggesting that the Žemaitukai horses were less trained for endurance competition. Those differences between the two breeds might be based on different amounts of fat free body mass. Further studies are required to determine the inter-breed differences in muscle architecture and body composition.

animals : an open access journal from mdpi

10.3390/ani13091462

PMC10177438

The heritability of a trait is the proportion of phenotypic variance explained via genetic variance. Prior to the advent of genomics, heritability was estimated using extensive pedigree analyses. With the availability of genome-wide genotyping arrays, an alternative method became available to estimate heritability using genomic relationship matrices derived from genotype data. We used approaches that consider patterns of linkage disequilibrium and relatedness to estimate heritability of osteochondrosis dissecans in Hanoverian Warmblood horses based on genotype data from SNP arrays and imputed genotype data. Taking into account linkage disequilibrium patterns and relatedness in the data, heritability estimates on the linear scale for fetlock-, hock- and stifle-OCD were 0.41–0.43, 0.62–0.63, and 0.23–0.25, respectively, with standard errors of 0.11–0.14. In summary, SNP-based approaches are able to capture a greater proportion of additive genetic variance than previous estimates based on pedigree data.

Before the genomics era, heritability estimates were performed using pedigree data. Data collection for pedigree analysis is time consuming and holds the risk of incorrect or incomplete data. With the availability of SNP-based arrays, heritability can now be estimated based on genotyping data. We used SNP array and 1.6 million imputed genotype data with different minor allele frequency restrictions to estimate heritabilities for osteochondrosis dissecans in the fetlock, hock and stifle joints of 446 Hanoverian warmblood horses. SNP-based heritabilities were estimated using a genomic restricted maximum likelihood (GREML) method and accounting for patterns of regional linkage disequilibrium in the equine genome. In addition, we employed GREML for family data to account for different degrees of relatedness in the study population. Our results indicate that we were able to capture a larger proportion of additive genetic variance compared to pedigree-based estimates in the same population of Hanoverian horses. Heritability estimates on the linear scale for fetlock-, hock- and stifle-osteochondrosis dissecans were 0.41–0.43, 0.62–0.63, and 0.23–0.25, respectively, with standard errors of 0.11–0.14. Accounting for linkage disequilibrium patterns had an upward effect on the imputed data and a downward impact on the SNP array genotype data. GREML for family data resulted in higher heritability estimates for fetlock-osteochondrosis dissecans and slightly higher estimates for hock-osteochondrosis dissecans, but had no effect on stifle-osteochondrosis dissecans. The largest and most consistent heritability estimates were obtained when we employed GREML for family data with genomic relationship matrices weighted through patterns of regional linkage disequilibrium. Estimation of SNP-based heritability should be recommended for traits that can only be phenotyped in smaller samples or are cost-effective.

1. IntroductionOsteochondrosis (OC) is one of the most important orthopaedic diseases of the juvenile horse [1]. Due to disturbances in enchondral ossification, damage to the subchondral bone is the reason for the formation of intraarticular osteochondral fragments and subchondral bone cysts. If osteochondral fragments occur, the disease is referred to as osteochondrosis dissecans (OCD). Osteochondrosis occurs at certain predilection sites. Joints frequently affected are the metacarpophalangeal/metatarsophalangeal (fetlock), tarsocrural (hock) and femoropatellar joints (stifle) [2]. Therefore, it is of utmost interest to evaluate genetic parameters for OCD as precisely as possible in order to take breeding measures.The aetiology of OCD appears to be multifactorial with a relevant genetic contribution [3,4]. There have been estimations regarding the heritability of OCD based on pedigree [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21] and genotyping data [22,23]. Those estimates are shown in a previous review [1] and supplemented with results of more recent studies shown in Supplementary Table S1.Before the genomics era, estimates of heritability were based on pedigree data. The introduction of SNP arrays enabled the estimation of heritability based on genotyping data. Genome-based heritability estimates offer many advantages through eliminating the need to collect extensive pedigree data. Apart from the time-consuming data collection, analysis of pedigree data poses the risk of biased results due to incorrect, incomplete, or varying depth of pedigrees. Heritability estimates between populations may vary because of population history, gene frequency, or environmental exposures [23].There are various approaches to estimating heritability based on genotyping data. The fraction of phenotypic variance that can be explained using variants that have been identified as causal variants through genome-wide association studies (GWAS) is named hGWAS2. hGWAS2 is limited because, for most complex diseases, only a small proportion of variants has already been identified [24]. hSNP2 is the proportion of phenotypic variance explained using all SNPs on a genotyping platform. hSNP2 is the upper limit for hGWAS2 and can be a measurement of the proportion of already identified causal variants in the actual genetic variance of a trait [24]. The difference between hGWAS2 and hSNP2 is often referred to as missing heritability [25,26]. We want to estimate hSNP2 and assess different methods using genomic REML algorithms (GREML). Heritability estimation methods based on genotyping data require certain assumptions regarding the population structure of the underlying population, indirect genetic effects, the presence of artificial or natural selection within the population, and linkage disequilibrium. These assumptions are specific for each population and trait and can severely bias the heritability estimates [27]. Different methods require certain assumptions [28]. The aim of heritability estimation using SNP array or Beadchip data is to approach h2, which is the actual narrow sense heritability of a trait [29].We estimated hSNP2 using GCTA GREML [30], which is a single-component model to estimate heritability based on a genomic relationship matrix (GRM) and unrelated individuals [31]. As this approach is very sensitive to patterns of linkage disequilibrium (LD) [32], we used a similar single-component approach that is implemented in the software LDAK and weights SNP effect sizes according to regional LD patterns to construct the GRM [32]. LD describes the non-random association of alleles at two or more loci. LD varies because of factors such as population history, natural or artificial selection, mutation, and other forces that cause changes in allele frequency [33]. It can cause upwards biased estimates of heritability due to repeated tagging of SNPs [34].As Beadchip arrays are usually based on common SNPs, we want to use imputed SNPs for heritability estimation to capture the effects of more causal variants [27,35,36,37]. However, it is recommendable to prune for minor allele frequency (MAF) because rare variants are imputed less accurately [36,38].In a previous study on osteochondrosis in horses, SNP-based heritability for osteochondrosis in the hock was estimated in a population of 479 North American Standardbred. Horses using REML analysis in GCTA and LDAK with the weighted GRM in an imputed data set with ~1.25 million SNPs. The OCD frequency in this study population was 0.27. The analyses were repeated using a smaller study population, with individuals pruned for relatedness at 0.25. SNP-based estimates seemed to be biased upwards via LD, which implies the need to account for LD in heritability estimations in horse populations [20].Zaitlen et al. [24] proposed a method to estimate heritability based on a population with different degrees of kinship that avoids the need to remove closely related individuals from the study population. This method is implemented in GCTA and is known as GREML analysis for family data. It provides estimations of SNP-based heritability in family data as well as narrow sense heritability hf2, which enables the quantification of genetic effects resulting from kinship and, thus, enables the detection of higher amounts of heritability [24]. This method has not yet been used in horse populations with very specific relatedness structures.The aim of this work is to estimate the heritability of the trait OCD in the fetlock, hock, and stifle joints based on SNP data. We conducted a GREML analysis, an LDAK analysis using an LD-weighted GRM based on unrelated individuals, and a GREML analysis for family data using two simultaneously constructed GRMs for individuals with different relatedness structures. Beside the effects of the different GRMs, we will observe the effects of imputation and the different MAF restrictions, comparing the heritability estimates with previous pedigree-based analyses using a similar study sample [5].2. Materials and Methods2.1. AnimalsThe horses included in this study were a subset of the study population previously analyzed by Hilla et al. [5]. For the present study, 446 four-year-old Hanoverian warmblood horses were included. The inclusion criteria were as follows: only one horse per sire and maternal sire was allowed, either as a control or a case. The controls and cases were randomly distributed among the sires and maternal sires. The control horses had to be free of all diseases found during the veterinary health examination for pre-selection at auctions, at licensing, or during the purchase examination. The cases were horses with OCD only and free of any other disease recorded in the veterinary health check. The veterinary health check included clinical and radiographic examination of all four limbs. Only osteochondral fragments at the specific predilection sites of the fetlock, hock, and stifle joints were classified as OCD [5,39]. Osteochondral fragments plantar in the fetlock joints and at the insertion sites of the short sesamoid ligaments at the proximal phalanx of the hindlimbs were classified as plantar and dorsodistal fragments of the fetlock joints; thus, there were not considered OCD. Distal and proximal interphalangeal joints, fetlock joints, hock joints, and stifle joints were evaluated for contour changes stemmed from periarticular osteophytes or exostoses and for a narrow or absent joint space. These changes were classified as osteoarthroses. Radiographic changes in the shape, symmetry, contour, and structure of the navicular bone and the shape, size, number, and location of the canales sesamoidales were scored on a scale of 1–4 [40]. Only horses with a score of 1 were considered free of radiographic changes to the navicular bone. Horses with the presence of a sidebone were also scored as not being free of radiographic changes. After removing all horses affected by diseases other than OCD, the data set was filtered for cases and controls. The strict inclusion criteria resulted in the final data set comprising 446 horses. Traits were encoded as 0/1 variates for each joint. We did not consider an overall score for OCD because genetic correlations of OCD between the different joints were moderately negative (fetlock-OCD with hock- and stifle-OCD: −0.12 and −0.18) or moderately positive (hock-OCD with stifle-OCD: 0.17) [5].The phenotypic and genotype data were provided by the Association of Hanoverian Warmblood breeders (Hannoveraner Verband e.V., Verden, Germany). The frequencies of OCD were 0.2489 (n = 111), 0.3139 (n = 140), and 0.0291 (n = 13) in the fetlock, hock, and stifle joints, respectively. Relationships expressed through the contingency coefficient between the frequencies of fetlock, hock, and stifle OCD were close to zero because 96, 125, and 11 horses represented the sole cases of fetlock, hock, and stifle joint OCD, respectively.2.2. MethodsGenome-wide genotyping data was obtained using the GGP Equine (71 589 SNPs) genotyping array. Descriptive statistics of the population were calculated with SAS, version 9.4 (Statistical Analysis System, Cary, NC, USA, 2023). The SNP data have been imputed to 1,617,270 SNPs with an information score of 0.95 using BEAGLE 5.4 [41] and publicly available whole genome sequencing data for horses (Supplementary Table S2). Subsequently, the imputed and non-imputed data sets were pruned at MAFs of 0.01, 0.025, or 0.05 using PLINK 1.9 [42,43], resulting in six different data sets. Using all six data sets, heritabilities for OCD of the fetlock, hock, and stifle joints were estimated using the GREML analysis implemented in GCTA (genome-wide complex trait analysis) 1.94.1 [30] with one GRM [44], resulting in SNP-based heritability (hSNP2). Subsequent estimations were performed using the LD-weighted genomic relatedness matrix as implemented in LDAK 5.2 [45] and the integrated REML analysis [32], resulting in estimations of hSNPw2. Using the GREML analysis for family data with two GRMs simultaneously, based on all pairs of individuals and related individuals [24] implemented in GCTA 1.94.1 [30], we estimated hf2. The GRM based on all pairs of individuals captured information on the sharing of causal variants tagged using SNPs. The second GRM considered only individuals who were identical-by-state above a certain threshold (0.05) and, consequently, only related individuals. Hence, it captured information on shared causal variants that could not be tagged using SNPs [24,29]. Both GRMs fitted into a mixed linear model and were supposed to provide estimates of hSNP−all−pairs2 from the first GRM and the missing heritability hSNP−related2 from the second GRM. Those values were summed up to hf2 [24].We obtained heritability estimates (hfw2 and hSNP−all−pairs−w2) by implementing the LD-weighted genomic relationship matrix estimated with LDAK [32] in the GREML analysis for family data [24]. Sex was included in all analyses as a covariate. As we used 0/1-data, all heritability estimates were transformed onto the liability scale using the prevalence option of GCTA. The study design for heritability estimations is illustrated in Supplementary Figure S1.3. ResultsThe results of our heritability estimates for osteochondrosis in the fetlock joint are given in Table 1. Additionally, estimates for hSNP−all−pairs2 and hSNP−all−pairs−w2 are given in Supplementary Table S3. The SNP-based heritabilities estimated with GREML revealed that the heritability estimates decreased in the imputed data set compared with the original data set. The SNP-based heritabilities estimated with GREML and LDAK for fetlock-OCD differed in several aspects. Accounting for regional LD patterns increased heritability estimates for the imputed genotype data but slightly decreased heritability estimates for the original data sets. Heritability estimates using GREML analysis for family data resulted in higher estimates for both data sets, the original and imputed genotype data, as well as when regional LD patterns were considered. The effects of using different MAFs had only small effects when we used LD-weighted genomic relatedness matrices with LDAK. Standard errors for heritability estimates using GREML analysis for family data slightly increased from 0.11–0.12 to 0.13–0.14 on the linear scale.After transformation onto the liability scale, we obtained fairly high estimates for heritability and their standard errors.When comparing the GREML analysis for the original and imputed data sets, the same trends were observed for the heritability estimates for hock- and fetlock-OCD (Table 2, Supplementary Table S4). However, the increase in heritability estimates was much smaller when GREML analysis was applied to family data than to fetlock-OCD. The consistency and magnitude of the heritability estimates were highest when we used GREML analysis for family data with LDAK.The most consistent heritability estimates were obtained for stifle-OCD for the analysis accounting for family data and LD patterns for both data sets (the original and imputed genotype data) (Table 3, Supplementary Table S5). The effect of family structure on heritability estimates was small, while LD patterns had slightly larger effects. In general, differences between the different approaches were relatively small. Transformation onto the liability scale gave meaningless estimates >1 due to the low frequency of cases.4. DiscussionAccording to the findings of previous studies, it seems to be recommendable to account for linkage disequilibrium when estimating heritability based on SNP data [20,29,31,32,34,46]. Horses have long-range linkage disequilibria, which is why SNPs can show effects of a risk variant as far away as 1 Mb [47]. Additionally, LD is higher within breeds than across breeds [48], which is important since population data are usually used for heritability estimations. In general, REML-based estimates, such as those obtained from GREML analysis in GCTA, are sensitive to patterns of LD [32]. The linkage disequilibrium between SNPs is used to create the GRM and the LD between SNPs; causal variants can cause bias in heritability estimation [32,36]. As the intensity of linkage disequilibrium varies regionally along the genome, LDAK weights the SNPs according to local patterns of LD [32]. While we cannot observe a large impact of LD using our original data set, we see slight differences between GREML analysis and LDAK analysis in the imputed data set. The difference between heritability estimates increases with increasing MAF restrictions, which is attributable to the fact that less genetic variation is captured with SNPs when lower frequencies are recorded. Additionally, allele frequency and linkage disequilibrium are dependent on each other [49], which explains why the estimations conducted with LDAK are able to compensate changes in MAF. We assume that linkage disequilibrium does not play a major role in our study population. One possible explanation could be that we included many individuals with diverse LD structures, meaning that they outweighed each other in our analysis.One cause of undetected heritability could be that rare variants, and eventually even variants with large effects, may not be mapped on the available genotyping arrays that mainly include common SNPs [36]. Therefore, it is recommended to perform heritability estimations on imputed data sets [29]. To capture as much variation as possible, we imputed our Beadchip data to 1,617,270 SNPs, which corresponds to the recommendations given by Evans et al. [29] for heritability estimations. When comparing the results of the original and imputed data sets, we observe for all traits analyzed the most consistent estimates when family data and LD patterns are accounted for. Even the differences between the original and imputed data shrink or are no longer present. In the present data set, imputation had no or very little effect on SNP-based heritabilities; thus, we were unable to detect variation due to rare alleles.The single-component analyses in GCTA and LDAK calculated GRM based on the available SNP data to subsequently estimate heritability. For those analyses, it was recommended to prune for relatedness to eradicate bias caused by common environmental or other non-additive genetic effects [29]. The resulting unrelated individuals are by definition distantly related individuals because they share distant ancestors [50]; however, they are assumed to provide random genetic variance [28]. The need for pruning for relatedness arises from the model assumption in the GREML analysis that all measured genetic effects are direct effects. If related individuals were included, the indirect genetic effects between those individuals would be counted as direct effects and, thus, inflate heritability estimates [28]. Indirect genetic effects may result from genetic maternal effects [28]. The idea of the GREML analysis for family data was to find a way to circumvent pruning for relatedness in a study population and, thus, ensure a larger study population, which in turn should lead to lower standard errors. Additionally, the GREML analysis for family data estimates hf2 and, thus, is able to capture higher heritability [24]. While hf2 provides an unbiased estimate of the heritability of the trait, the proportions of the single components do not always seem to be assessed correctly [24,29]. We only observed this phenomenon in the imputed data set for stifle-OCD when we employed GREML for family data without an LD-weighted genomic relationship matrix. In all other analyses, we could not observe imbalanced contributions to the heritability estimates resulting from the two different GRMs. The most likely reason for this issue is the very low frequency of cases for stifle-OCD.In our analyses for fetlock-OCD, we can confirm that we detected higher estimates of heritability with GREML for family data than with the single-component REML algorithms, whereas for hock- and stifle-OCD the increase in heritability estimates was rather small. We assume that in our population a significant amount of heritability for fetlock-OCD may be due to indirect genetic effects that are captured examining the genetic effects between individuals with varying degrees of relatedness. This is the first analysis with GREML for family data that has been performed in a horse population. The most consistent heritability estimates were obtained using GREML for family data and an LD-weighted genomic relationship matrix for the original and imputed genotype data. With frequencies of cases closer to 0.5 in the population under study, differences between the original and imputed data sets diminished. However, we have to note that GREML for family data is designed for human populations with their specific relatedness structure and significantly larger available data sets [24,31]. While in a human population, full siblings are common, in horse population, full-siblings are uncommon.Since OCD is defined as a binary trait, all results of the REML analyses have been transformed onto the liability scale as recommended [30,51,52,53]. In agreement with previous studies, upward bias may occur, particularly when estimates on the linear scale are high and more frequencies deviate from 0.5 [11,12].The present study used data from Hilla et al. [5]. We selected horses as representatively as possible and avoided including closely related animals, such as paternal half-siblings and maternal sire half-siblings. The results obtained from the present study allow us to assume that analyses using GREML for family data and an LD-weighted genomic relationship matrix result in higher heritability estimates compared to estimates based on pedigree data (hped2) in a Hanoverian Warmblood horse population. However, larger genotype data sets should be available to reach lower standard errors. Nevertheless, heritability estimations based on SNP-based methods may give reliable results even in much smaller data sets compared to pedigree-based estimates.Similar results were reported for hock-OCD in a population of North American Standardbred horses [20]. Compared to our results, this previous study showed larger increases in heritability estimates when taking into account LD patterns compared to the standard GRM. Thus, we assume stratification based on families and breed history may have contributed to this result. In addition, the LDAK version used by McCoy et al. [20,32] was a less improved version of the software, which could have had an effect on the results. Heritability estimates are specific for populations because of different familial structures and selective signatures in the genome [1,33], which may also contribute to the difference in our results. In the present study, standard errors on the linear scale were at 0.11–0.14, resulting in 95% confidence intervals from ±0.22 to ±0.27, while standard errors on the linear scale were at 0.12 and 0.16 in the previous study on US Standardbreds.In summary, we recommend the use of GREML for family data with an LD-weighted genomic relatedness matrix to estimate heritabilities, particularly for traits which are difficult or very costly to record. Due to restrictions in sampling and varyingly strong LD patterns in populations, the approach as provided by LDAK should be implemented in the estimation procedure. The pursuit of more precise heritability estimates is worthwhile means of achieving estimated breeding values with higher reliabilities and a higher selection response in health traits.5. ConclusionsEstimation of heritabilities based on SNP arrays is recommended because reasonably high accuracy of estimates can be achieved in smaller samples compared to pedigree-based studies with similar sample sizes. The use of genomic REML analysis for family data with LD-weighted genomic relationship matrices allows the capture of most of the additive genetic variance and provides the most consistent estimates at different MAFs. The present study yielded higher heritability estimates with reasonable standard errors than a previous study for the same population. Further studies with larger data sets should be performed to validate these results.

animals : an open access journal from mdpi

10.3390/ani11061614

PMC8227101

Foot-related lameness is one of the most significant welfare issues in farm animals. Contrary to dairy cows and meat sheep breeds, epizootiological data on foot-lesions and associated lameness in dairy sheep are scarce. In this study, data were collected from 30 representative intensive dairy sheep farms. Multivariate statistical analysis was used to produce a typology of intensive farming systems which resulted in the assignment of farms in two distinct clusters. Six hundred adult ewes were randomly selected from six flocks (three flocks per cluster) and a cross-sectional study was implemented to investigate the epizootiology and potential risk factors of foot-related lameness, foot-lesions and diseases. Ovine interdigital dermatitis and infectious footrot were the most common infectious foot diseases, while white line disease and hoof wall cracks were the most prevalent non-infectious lesions. Infectious footrot was the main cause of lameness and increased with age, whereas body condition score was associated with increased prevalence of ovine interdigital dermatitis. Comparisons between the clusters regarding foot-related lameness, foot-diseases and lesions at the animal, the limb, and the hoof level are presented, and relevant literature, mechanisms, hypotheses, and challenges of the field are discussed.

Foot-related lameness, foot-diseases and lesions are emerging issues in dairy sheep; however, relevant epizootiological studies are scarce, and risk factors have not been elucidated. The objectives of this cross-sectional study were (i) to address this dearth of knowledge by investigating the epizootiology of lameness-related foot-lesions and diseases, and (ii) to assess the impact of potential risk factors on foot health, in intensive dairy sheep farms. Thirty farms were assigned in two representative clusters using a multivariate statistical analysis. Three farms per cluster and 100 multiparous milking ewes per farm (total n = 600) were selected and enrolled in the study. Foot-related lameness, ovine interdigital dermatitis (OID), infectious footrot (IFR), white line disease, hoof wall cracks, as well as health and welfare traits were recorded. Overall prevalence of foot-related lameness was 9.0% and was primarily associated with IFR; however, additional infectious and non-infectious foot diseases and lesions also contributed. Among infectious foot diseases, OID was the most prevalent (21.3%) followed by IFR (8.0%); WLD and hoof wall cracks were the most prevalent non-infectious foot-lesions (37.7% and 15.3%, respectively). IFR and OID prevalence increased with age (p < 0.05) and BCS (p < 0.01), respectively, suggesting that host-related factors and husbandry practices are important determinants of its occurrence.

1. IntroductionIn recent years, the growing demand for sheep milk and products thereof has resulted in a remarkable increase of the global dairy sheep population (ca. 20.0%), and total sheep milk production (ca. 28.0%) [1]. To address this trend, farming systems have been evolving and adapting to more intensive management schemes, exploiting improved genotypes and modern husbandry practices and technologies. This is mainly evident in developed Mediterranean countries with a long-lasting tradition on dairy sheep farming, and a well-organized sheep milk processing industry (e.g., Greece, Spain, France, Italy, etc.).Intensification of production has reshaped the labor conditions and the socioeconomic status of farmers, converting sheep farming into an attractive career opportunity for young people in rural and peri-urban areas. It has also led to significant benefits for the dairy sheep sector including: (i) increased productivity; (ii) efficient utilization of available resources (such as high-yielding breeding stocks, modernized infrastructures, specialized labor, alternative feedstuff, and optimized land use); (iii) precision farming through adoption of advanced monitoring systems and husbandry practices; and (iv) the establishment of evidence-based biosecurity and hygiene measures for the control of infectious and parasitic diseases [2]. Nevertheless, a growing public skepticism occurs regarding the ethical treatment of animals and their well-being in intensive farming systems [3]. Indeed, several studies have stressed the emergence of health and welfare issues that further influence public perception challenging the sustainability of farms [4,5,6,7]. In addition, heterogeneity of intensive farms in terms of their structure, management, and production methods, minimizes the possibility of universally applicable herd health and disease control protocols [8]. Therefore, an updated, objective, and representative typology of these farms based on their characteristics and animals’ health and welfare status is useful for the assessment of flock health management status and the proposal of targeted modifications to address critical challenges.Lameness is a condition with considerable impact on dairy sheep productivity, health, and welfare. It is defined as the deviation from normal gait caused by a wide range of factors, usually followed by signs of pain or discomfort and described as a clinical sign rather than a disease itself. The most common causes of lameness include various foot and hoof infections and lesions affecting foot tissues (including joints and bones) that lead to a condition collectively described as foot-related lameness [9]. The major infectious causes of foot-related lameness are either bacterial, namely, infectious footrot (IFR) (Dichelobacter nodosus and Fusobacterium necrophorum), contagious ovine digital dermatitis (CODD) (Treponema spp.), ovine interdigital dermatitis (OID) (F. necrophorum), and pedal joint abscess (PJA) (F. necrophorum and Actinomyces pyogenes), or viral, namely, orf (Parapoxvirus), foot and mouth disease (Apthovirus), and bluetongue (Orbivirus). Non-infectious causes of lameness include white line disease (WLD), laminitis, and other foot lesions and injuries (e.g., toe granulomas, hoof wall cracks, overgrown hooves, foreign bodies, etc.) [10].Foot-related lameness ranks highly on the list of the most important health issues with the potential to significantly undermine animals’ performance and farms’ sustainability. The reduction on milk yield (ca. 20.0%, ref. [11]), body weight (11.6%, ref. [12]), and wool production (8.0%, ref. [12]) has been documented in lame sheep, explaining the considerable monetary losses derived from foot-related lameness. For example, the annual economic losses in the UK and New Zealand due to lameness have been estimated to reach GBP 24–80 M and NZD 11 M, respectively [13,14].It is known that both genetic and environmental factors predispose to foot diseases, lesions, and lameness thereof in sheep [9,14,15,16]. For example, polymorphisms of the DQA-2 loci in the ovine Major Histocompatibility Complex (MHC-ovar) genes have been associated with the immune response of sheep to IFR, demonstrating a pivotal role in the susceptibility or resistance against the disease [17]. Additionally, hoof conformation and hoof keratin quality are fundamental to maintain the foot health status [16], however, selective breeding for these traits has not been extensively applied. Environmental factors predisposing to foot-related lameness include animal management, nutritional status, housing conditions (bedding moisture, ventilation, temperature, etc.), season and climate, farmers’ knowledge/skills on foot care, and overall hygiene status [18,19]. Other animal factors potentially affecting foot health in sheep are age, productive stage, and milk yield [9]. Nevertheless, little attention has been paid to foot-related lameness in dairy sheep and the available literature is scarce; thereby, much uncertainty still exists regarding the epizootiology and risk factors of lameness-related foot diseases and lesions [9].A decade ago, in 40% of the studied intensive and semi-intensive Chios dairy sheep farms in Greece, prevalence of foot-related lameness was greater than 5.0% [18], with the prevalent diseases being IFR, OID, PJA, and WLD; since then, farming systems have evolved and foot-related lameness epizootiology is likely to have changed. In meat and wool sheep farms, updated information on foot-related lameness etiology and epizootiology is currently available, underpinning its significance. For example, in the UK, prevalence of specific diseases ranged from <1.0% to >25.0% [20,21], with more than 80.0% of meat sheep flocks, reporting an increase on the occurrence of foot-related lameness [21]. Similarly, in dairy cows, various studies in the UK and the USA place foot-related lameness among the most significant health and welfare issues, with the prevalence ranging from 21.0 to 35.0% [15,22,23,24]. However, extrapolation of epizootiological data from dairy cows or/and meat and wool sheep and applicability of published research in dairy sheep are problematic (for example, in dairy cattle, foot-related lameness is equally caused by infectious and non-infectious lesions unlike dairy sheep where infectious lesions are more frequently observed in lameness cases). This is due to (i) different species and dissimilar productive orientation, husbandry, health management, and breeding practices at the farm level, and (ii) the regional variety of soil-climatic conditions associated with different production methods. Hence, foot-related lameness cannot be evaluated and prioritized on ad hoc basis and subsequently, the suggestion of lameness mitigation strategies is not possible in intensive dairy sheep farms, although an urgent need to cope with the problem is evident.To address the aforementioned research gap, the objectives of this cross-sectional study were (i) to investigate and describe the epizootiology of lameness-related foot diseases and lesions, and (ii) to assess the impact of potential risk factors on them, in intensive dairy sheep farms in Greece.2. Materials and Methods2.1. Area of the StudyThirty intensive dairy sheep farms (high-input farms with considerable capital investment on breeding stocks, labor and infrastructures) were initially included in the study (n = 10,630 ewes). The farms were distributed in thirteen counties across Greece (Achaea, Aetolia-Acarnania, Attiki, Drama, Karditsa, Kilkis, Korinthos, Kozani, Larissa, Magnisia, Serres, Thessaloniki, Trikala) as presented in Figure 1, and located mainly in plain areas with the topography ranging from coastal areas to inland plateaus and the climate from typical Mediterranean to continental.Farms were surveyed on-site between May and July 2020, using a structured on-purpose built questionnaire. Data regarding farm structure, flock characteristics and management, labor, infrastructures, feeding and nutrition, reproduction, biosecurity and hygiene measures, disease control protocols, as well as animal overall and foot health status were collected (prevalence, severity, and control measures of lameness-related foot diseases and lesions). A score from 0 to 12 for preventive flock management was calculated assigning one degree for each of the following preventive measures implemented on a regular basis: vaccination against (i) clostridial diseases, (ii) contagious agalactia, (iii) enzootic abortion, (iv) pasteurellosis, (v) gangrenous mastitis, (vi) IFR, (vii) dry-off intramammary antibiotic treatment, (viii) anthelmintic treatment of lambs, (ix) anthelmintic treatment of ewes, (x) vitamin and mineral supplementation, (xi) preventive antibiotic treatment in lambs, and (xii) foot-trimming more than once per year. This score was used as a rough indicator of the preventive veterinary and hygiene status at the flock level.2.2. Farm SelectionA multivariate statistical analysis was used to produce a typology of intensive farming systems by grouping the farms into representative clusters as described in the statistics section below. Based on the results, farms were assigned into two clusters with 22 (Cluster 1) and 8 (Cluster 2) farms, respectively. Three farms per cluster were randomly selected and enrolled in the main study.2.3. Animal Selection and RecordingFrom each farm, 100 multiparous milking ewes, at the beginning of lactation (20–50 days post-lambing), were randomly selected (total n = 600 ewes). The farms were visited during scheduled routine foot-trimming, which was performed by trained personnel under the supervision of two veterinarians. Before and during foot-trimming, one of the veterinarians clinically assessed and recorded the hoof wall overgrowth and cracks, as well as the occurrence, topography, and severity of foot diseases and lesions including OID, IFR, and WLD at the hoof level. The other veterinarian performed physical examination of the animals and recorded clinical findings and welfare indicators at the animal level. The recorded traits included (i) body condition score (BCS, 1–5, 1 = emaciated, 5 = obese with 0.25 increments) [25], (ii) occurrence of foot-related lameness, arthritis, respiratory disease, ocular and nasal discharge, body abscesses, mastitis, udder lesions and deformities (i.e., skin lesion, abscess, and asymmetry) (0 = absence, 1 = presence), and (iii) wool quality (0 = good quality, 1 = poor quality) and fleece cleanliness (0 = clean, 1 = dirty). Ear tag, breed, and age were also recorded.2.4. Statistical AnalysesSPSS v23 software (IBM Corp., Armonk, NY, USA) was used for the statistical analyses, and statistical significance was set at the 0.05 level. To classify the 30 farms into representative clusters, a two-steps approach was followed. In brief, at the first step, principal component analysis (PCA) was used for data mining purposes, resulting in six principal components (PCs) with eigenvalues greater than 0.8 (Figure 2) to be retained. The six extracted PCs explained about 86.5% of the total variation described by the nine initially considered variables. However, the Kaiser–Meyer–Olkin measure of sampling adequacy was low (0.338) and Bartlett’s test of sphericity was not significant (p = 0.186), indicating a poor performance of the PCA; thus, the extracted PCs were not used for the subsequent clustering of the farms. Instead, k-means cluster analysis was used to allocate the farms into two clusters based on factors potentially predisposing to foot-related lameness prevalence (years of farmer’s experience, ewes’ replacement rate, annual milk yield per ewe, prolificacy, stocking density, and annual incidence of foot-related lameness).Descriptive statistics were estimated (mean ± standard error), whereas comparisons between the two clusters regarding farm characteristics were performed using analysis of variance (ANOVA). Chi-square test (χ2 test) was used to compare the prevalence of foot-related lameness and specific non-infectious and infectious foot lesions between (i) the two clusters at the animal-, the limb-, and the hoof level, (ii) front/rear limbs at the limb level, and (iii) front/rear and inner/outer hooves at the hoof level. χ2 test was also used for comparisons between the two clusters considering other recorded health and welfare traits.For the cross-sectional study, all data were integrated into three databases, corresponding to records at the animal-, the limb-, and the hoof- level (600, 2400, and 4800 records, respectively). Prevalence of (i) lameness, and (ii) lameness-related foot diseases and lesions were estimated. To test the contribution of potential risk factors in predicting the occurrence of foot-related lameness, foot diseases and lesions at the animal level, a set of multilevel binary logistic regression models was used, where random effects of farm j and animal i and fixed effects of age, wool quality, and BCS were considered. Logit [Pr (Yij = 1)] = β0j + β1*AGEij + β2*WOOLij + β3*BCSij + εij(1) where Yij = dependent variable (occurrence of foot-related lameness, OID, IFR, WLD, and hoof wall cracks, on at least one limb), β0j = intercept, β1 = coefficient of age (AGE) (4 levels), β2 = coefficient of wool quality (WOOL) (2 levels), β3 = coefficient of body condition score (BCS), εij = random residual error.3. Results3.1. Descriptives and Comparisons between the ClustersDescriptive statistics and comparisons between the two clusters regarding farm and animal characteristics are presented in Table 1. In general, farms in the two clusters had similar structure and characteristics except for (i) milk yield per ewe per lactation which was significantly higher in Cluster 1 (ca. 383 kg) compared to Cluster 2 (ca. 257 kg) (p < 0.001), and (ii) number of empty ewes and abortion rate which were significantly higher (p < 0.01 and p < 0.05, respectively) in Cluster 2 (ca. 10.0% and 4.0%, respectively) compared to Cluster 1 (ca. 4.0% and 1.0%, respectively) (Table 1).3.2. Prevalence of Foot-Related Lameness, Foot Diseases and Lesions at the Animal LevelA total of 210, 98, 230, and 62 ewes of two, three, four, and >four years old, respectively were used for the study. Average BCS of the studied ewes was 2.8 ± 0.01. The overall prevalence of foot-related lameness was 9.0% (54/600), and was significantly higher in Cluster 1 (11.3%, 34/300) compared to Cluster 2 (6.7%, 20/300) [χ2 (1, n = 600) = 3.99, p < 0.05, Table 2]. The major cause of lameness was IFR (5.5%, 33/600) and its combination with (i) OID (1.2%, 7/600), (ii) WLD (1.0%, 6/600), and (iii) OID and WLD (0.2%, 1/600); other causes of lameness were WLD (0.8%, 5/600) and excessive hoof wall overgrowth (0.3%, 2/600).In the case of infectious foot diseases, OID was the most prevalent (21.3%, 128/600) followed by IFR (8.0%, 48/600). Regarding non-infectious foot lesions, almost all the studied animals had at least one overgrown hoof (99.3%, 596/600), whereas the prevalence of WLD and hoof wall cracks were 37.7% (226/600) and 15.3% (92/600), respectively. Prevalence of foot diseases and lesions per age group are presented in Figure 3. All foot lesions were observed in each of the studied farms, with OID, IFR, WLD, and hoof wall cracks prevalence ranging from 7.0 to 48.0%, 2.0 to 14.0%, 30.0 to 51.0%, and 13.0% to 18.0%, respectively.Prevalence of foot lesions, and other health and welfare traits recorded at the animal level, in the two clusters, and comparisons between them are summarized in Table 2; OID prevalence was significantly higher in Cluster 2 (p < 0.01); on the contrary, IFR prevalence tended to be higher in Cluster 1 (p = 0.071).3.3. Prevalence of Foot Diseases and Lesions at the Limb LevelPrevalence of foot diseases and lesions at the limb level and comparisons between (i) the two clusters, and (ii) front/rear limbs in the studied sheep population, are presented in Table 3. IFR prevalence was significantly higher in Cluster 1 compared to Cluster 2 (p = 0.05). Additionally, both OID and IFR prevalence were significantly higher in rear compared to front limbs (p < 0.001 and p < 0.01, respectively).Prevalence of foot diseases and hoof lesions according to the number of the affected (i) limbs (OID), and (ii) hooves (IFR, WLD, and hoof wall cracks) are presented in Figure 4.3.4. Prevalence of Foot Diseases and Lesions at the Hoof LevelPrevalence of IFR, WLD, hoof wall overgrowth, and hoof wall cracks at the hoof level and comparisons between (i) the two clusters, (ii) front/rear hooves, and (iii) inner/outer hooves are presented in Table 4. IFR prevalence at the hoof level was significantly higher in Cluster 1 compared to Cluster 2 farms (p < 0.05), and in rear compared to front hooves (p < 0.01).3.5. Risk Factors for the Occurrence of Lameness and Lameness-Related Foot LesionsThe effects of potential risk factors on the occurrence of lameness and lameness-related foot diseases and lesions are presented in Table 5. A one-degree increase on BCS (i) tended to be associated with ca. 2.9 times higher likelihood of foot-related lameness (p = 0.064), and (ii) was associated with ca. 3.7 times higher likelihood to develop OID lesions (p < 0.01). Additionally, ewes with good wool quality were ca. 2.3 times more likely to develop OID lesions (p < 0.05) compared to ewes with poor wool quality. Greater than four-year-old ewes were ca. 7.7 and 12.5 times more likely to develop IFR lesions compared to three (p = 0.064) and four-year-old (p < 0.05) ewes, respectively.4. DiscussionTo the best of our knowledge, this is the first study assessing the epizootiology and distribution of foot diseases and lesions in intensively reared dairy sheep in Greece, addressing a significant dearth of knowledge and facilitating better understanding of foot-related lameness; OID and IFR were the most observed infectious foot diseases, whereas WLD and hoof wall cracks were the most common non-infectious hoof lesions. In addition, IFR was the major cause of lameness in the studied flocks.In our study, a multivariate approach was utilized, integrating potential lameness risk factors as classifying variables to achieve an a posteriori typology of intensive farming systems. Similar approaches have been exploited in the past, to classify dairy small ruminant farms in Greece, based on their structure and characteristics, aiming to describe and understand the changes, challenges, and future perspectives of the sector [26,27]. Apart from milk yield and specific reproduction efficiency traits (empty ewes and abortion rate), no other remarkable differences were observed between the two clusters. Therefore, it can be assumed that a common evolutionary pattern exists in intensive dairy sheep farms in Greece; nevertheless, peculiarities of management and husbandry practices are likely to modify productivity and foot-health status. Cluster 1 farms had significantly increased milk yield per ewe per lactation and a higher prevalence of foot-related lameness and IFR compared to Cluster 2; hence, it could be hypothesized that sheep with higher milk yield are more susceptible to foot-related lameness. This is a hypothesis supported by studies in dairy cows showing that high yielding animals are more likely to develop foot lesions and lameness [23,28,29]; however, this is not sufficiently evidenced in dairy sheep [11], and under the current study-design, it is not possible neither to conclude the pathophysiological mechanisms nor estimate the effect of milk yield on foot-related lameness. Cohort studies comparing groups of high- and low-yielding ewes with regard the occurrence of foot-related lameness across lactation are necessary to address this hypothesis. On the contrary, the effect of foot-related lameness on milk yield has been documented in dairy sheep [11,30]. Similarly, in dairy cattle, foot-related lameness was associated with remarkable reduction on daily milk yield (from 1.6 to 2.7 kg/day; [31,32]) and milk yield per lactation period (from 270 to 574 kg [33]).Overall prevalence of foot-related lameness in the studied ewes population (9.0%) was increased compared to the mean prevalence reported by the farmers during the survey (2.4 and 5.3%, for Cluster 1 and Cluster 2, respectively), and the estimated prevalence in intensive and semi-intensive Chios sheep farms in Greece about a decade ago (6.8%) [30]. In the UK, results regarding prevalence of foot-related lameness are controversial, demonstrating either a reduction [13,34,35] or an increase [21] compared to the average prevalence (8.0–10.0%) reported 15 to 20 years ago [36,37]. The results of our study and the data from the UK indicate that lameness mitigation strategies are not universally efficient; also, we need to be cautious when interpreting the progression of foot-related lameness prevalence, as it is frequently underestimated by the farmers, due to different attitudes and beliefs and the underdiagnosis and underreporting of the problem [38,39]. Interestingly, in our study, foot-related lameness was considered by all farmers as a significant challenge for the health and welfare status of their flock. Nevertheless, most of them could recognize only severely lame animals and major foot diseases and lesions thereof (i.e., IFR, OID but not WLD); this is not consistent with sheep farmers in the UK, who could recognize even mildly lame sheep, although their attribute towards lameness varied [37].OID was the most frequently observed infectious foot disease with almost one out of five animals in our study being diagnosed with the disease and the prevalence among the flocks varying from 7.0 to 48.0%. Mean prevalence of OID in the UK has been estimated at 6.9% [13], however, prevalence up to 45.0% has been reported [40]. Frequently, OID is considered as the early stage of IFR and defined by infection and exudative inflammation signs at the interdigital space, caused by F. necrophorum before the invasion of D. nodosus and the underrunning of the hoof [41,42,43]. In our study, OID was separately assessed and defined by the occurrence of mild superficial lesions and loss of hair, limited at the interdigital space, without the occurrence of underrunning of the hoof wall. In general, early stages of OID either are not followed by lameness, or result in unnoticed lameness cases [43]; this is consistent with our findings where mild to moderate OID cases were observed and not found to be associated with lameness unless IFR co-existed. Absence of lameness indicates that OID is possible to remain underdiagnosed or undiagnosed in intensive sheep farms until its complication by D. nodosus and the development of IFR. Furthermore, OID was more commonly observed at the rear limbs compared to the front limbs. According to the available literature, this is the first time that differences between front and rear limbs regarding OID infections are observed in sheep. However, this is in agreement with the distribution of lesions due to infectious foot diseases in both dairy cows and beef cattle [23,44,45,46]. In the case of dairy cows, it has been suggested that the closer contact of rear limbs with manure (F. necrophorum is normally found on the digestive tract), and potentially body weight distribution, may explain the susceptibility of rear limbs to the disease. Nevertheless, in sheep farms, manure and bedding material are drier than in cow farms, whereas research on weight distribution and posture/gait biomechanics is scarce; hence, support or rejection of these hypotheses are not feasible.IFR is a globally spread and exceptionally contagious and painful disease of the foot. It is considered the most significant lameness-related bacterial foot disease in sheep causing extended foot lesions [47]. These lesions include the underrunning of the hoof, and in advanced cases, the total separation of the horn from the underlying hoof matrix [43,48,49]. Its prevalence in meat sheep varies from negligible (0.4%, [50]) or moderate (3.7% [37]; 8.5% [51]) to high (>40.0% [52]) or extremely high (>95.0% [43]) and its transmission occurs mainly in damp conditions, rather than in dry, hot, and cold conditions [53]. Under the UK temperate climate, 80.0 to 90.0% of sheep flocks are affected by IFR [21,54], with the annual cost of the disease estimated between GBP 20 to 80 million [55,56]. In our study, IFR was observed in all the studied flocks, and 8.0% of the studied ewes were affected by IFR in at least one hoof; this is within the expected prevalence range, considering that recordings took place in autumn, when warm weather and high moisture favor survival, proliferation, and transmission of F. necrophorum and D. nodosus [9,42,49]. IFR was present in 87.0% (47/54) of the ewes with lameness alone or in combination with other foot lesions. This finding underlines its significance in the epizootiology of foot-related lameness in intensive dairy sheep farms, which is in agreement with Gelasakis et al. [30] and Kaler and Green [37] who documented that more than 60.0 and 90.0% of lameness cases were due to IFR in dairy and meat sheep, respectively. IFR had a significantly higher prevalence in cluster 1 ewes, when examined at the limb and the hoof level, implying an association between high milk yield and susceptibility to IFR, which is consistent with findings in dairy cows [28]. IFR was more commonly observed at rear limbs and hooves compared to front limbs and hooves, possibly explained by the mechanism detailed for OID. As expected, and in compliance with the available literature, no significant differences were found between inner and outer hooves as regards to the occurrence of the disease.The most prevalent foot lesion in the studied sheep population was WLD (ca. 38.0%). High prevalence of WLD is not unusual. Previous studies have reported prevalence as high as 75.0% in some flocks [10,57]. WLD occurs when the hoof wall is detached from the laminar corium, a condition that is usually underdiagnosed or under-reported by the farmers who are unable to recognize it. This was also the case in the surveyed farms. Thereby, high prevalence of WLD could be associated with misdiagnosis and the absence of efficient treatment. In most cases, WLD is not directly associated to lameness, but predisposes to it due to debris accumulation inside the defect, leading to proliferation of bacteria, foot infection, abscess formation (WLA), pain and lameness. Although, the etiology and risk factors of WLD remain unknown in sheep [10], type of ground [58] and nutritional deficiencies (biotin and sulfur containing amino acids) are considered as causative or predisposing factors in dairy cattle [59]. Moreover, genetic background has recently been evidenced for WLD in meat sheep breeds using data from 9169 sheep and 22 farms with increased WLD prevalence (47.0% and 24.0%, for Scottish Blackface and Texel sheep, respectively) in the UK [57]. In dairy sheep, data regarding the epizootiology of WLD is limited. Recently, Gelasakis et al. [30] reported white line lesions (WLD and WLA) as the second most significant cause of foot-related lameness in intensively reared Chios sheep (with an incidence risk of 1.1% and 0.8%, respectively); however, only lame animals were included in that study. Regarding the distribution of WLD lesions at the limb and the hoof level, no significant differences were found, which agrees with findings from dairy goats in the UK [60]; hence, it could be assumed that WLD is not likely to be associated with (i) body weight distribution and biomechanics during standing and walking, (ii) anatomical peculiarities of hooves, and (iii) different environmental exposures, at the limb or the hoof level (e.g., associated with their placement, front/rear or inner/outer). In the present study, none of the factors assessed was found to affect the prevalence of WLD. Breed has been found to predispose in WLD in both meat sheep and dairy cows, whereas WLD prevalence in cows increased with age and milk yield [58,61]. Although we used only purebred animals of different breeds, the breed effect could not be estimated as it was confounded by the farm effect. When assessed at the hoof level, WLD tended to be most prevalent in Cluster 2 ewes (p = 0.06), indicating that differences between intensive farming systems regarding WLD prevalence are likely to occur. In every case, it can be concluded that a different study-design (e.g., cohort study) should be exploited to investigate the etiopathogenesis and epizootiology of WLD.This is the first time hoof wall cracks in dairy sheep are studied, whereas relevant studies in meat and wool sheep are also not available. Hoof wall cracks were observed in ca. 15.0% of the ewes. Based on horse and cattle studies, inappropriate housing conditions, poor bedding, increased moisture, manure exposition, vitamin and trace mineral deficiencies, poor-quality hoof keratin and hoof conformation, as well as increased age and BCS lead to the disruption of hoof wall integrity and may predispose to hoof wall cracks and injuries [62,63,64,65,66]. Poor wool quality could be linked to a higher prevalence of hoof wall cracks, due to the fact that keratin of both wool and hoof horn have similar low- and high- sulfur protein fractions [67,68,69]; however, this was not confirmed by the model applied in our study. In dairy cows, a much lower prevalence (1.0%) of hoof wall cracks has been found, compared to beef cows (64.0%), mainly in the outer, front hooves associated with the gait pattern and the charging of the front limbs [66,70]. In our study, no significant differences were found between the clusters, front/rear limbs and hooves, and inner/outer hooves, whereas none of the traits assessed as risk factors had a significant effect.Inappropriate foot care (i.e., foot-trimming and foot-bathing), also predisposes to foot-related lameness. Indeed, severe hoof overgrowth may cause foot deformities and is likely to predispose to infectious foot diseases, foot lesions, and lameness [71]. Τhe majority of farms in the present study exercised routine foot-trimming once per year. Foot baths are rarely used, if at all. As expected, overgrown hooves were observed in almost all the studied animals regardless of cluster or farm. For this reason, the observed differences regarding hoof overgrowth at the limb and the hoof level are biologically meaningless. In dairy cows, differences between front and rear hooves regarding prevalence and severity of hoof overgrowth have been documented, with front hooves being more frequently severely overgrown [72]. In permanently housed dairy sheep, hooves are not worn-down. Animals usually walk on soft straw bedding material which does not favor natural wearing of the hoof. In addition, increased intake of concentrates in high yielding sheep further deteriorates the situation by increasing the growth rate of the hoof horn [73]. In the studied farms, excessive hoof overgrowth implies that routine foot-trimming once per year was not adequate, hence, twice per year may be necessary (i) to maintain normal shape and conformation of the hooves and the uninterrupted movement of the animals, (ii) to avoid injuries and infections, and (iii) for the early diagnosis and allocation of foot-lesions. In every case, it is crucial that foot-trimming is performed by experienced foot-trimmers to avoid excessive trimming leading to injuries and bleeding. According to Winter et al. [13], the remarkable decrease in the prevalence of sheep lameness in England from 2004 to 2013 in meat breeds was linked to (i) the control of excessive foot-trimming, (ii) the prompt treatment of lameness, and (iii) the overall training and knowledge-transfer to farmers for the management of lameness. Based on this observation, they concluded that foot-trimming should be excluded from routine foot care practices, as its benefits were not evidenced. However, extrapolating observations from meat sheep, reared under grazing systems where natural worn-down of the hooves occurs, is not appropriate.Season is among the environmental factors affecting foot-related lameness in both meat [74] and dairy sheep [9]. Our study was performed between mid to late autumn (from October to December), where increased moisture and relatively warm weather in combination with accumulating manure facilitate the proliferation and transmission of pathogens causing infectious foot diseases, thus justifying the observed high prevalence of OID and IFR [42,74]. Under the current study-design, it was not possible to assess the seasonal variation of infectious foot diseases prevalence, and a longitudinal study is warranted to address this objective. A similar study was recently implemented in meat sheep, demonstrating increased prevalence of IFR during late summer/early autumn and spring [74].Farming system influences the etiology and epizootiology of foot-related lameness in dairy sheep [9], with intensive farms being more challenged [39]. In intensive farms, increased stocking density [13] and inappropriate hygiene conditions [75] predispose to foot diseases and lameness thereof. Hence, increased stocking density in Cluster 2 could explain the significantly higher prevalence of OID, ocular and/or nasal discharge, mastitis, and udder skin lesions, as well as the dirtier fleece of the ewes. Noteworthy differences regarding basic hygiene and biosecurity protocols were not recorded among the farms; routine cleaning and disinfection of the premises, with commercial disinfectants, at least twice per year, during early autumn and spring were practiced in all cases following standard procedures.Average BCS of the studied ewes was acceptable (2.8) for the production stage (early lactation). Ideally, BCS during early lactation should be 3.0, however, negative energy balance at this stage in high-yielding dairy ewes is common resulting in a rather expected reduction on BCS during the first weeks of lactation [76]. Ewes with higher BCS were more likely to be diagnosed with OID. It is the first time such a relationship is observed in dairy sheep, although similar observations have been previously reported in dairy cows where heavier animals had increased probability of developing non-infectious hoof disorders [77]. Similarly, ewes with better wool quality in the present study had a higher risk of being infected by OID. Most OID cases were mild and not complicated by IFR; it can be hypothesized that ewes with better wool quality and/or higher BCS had a better quality of hoof horn preventing the transition from OID to IFR and lameness.In the studied flocks, distribution of animals within the four age groups was not equal, although animals were randomly selected. Nationwide drop of sheep milk price by 20.0 to 25.0% in 2017–2018 resulted in a wide reduction of the replacement rate that year, to decrease replacement cost. This was also the case in the studied flocks, leading to a reduced three-year-old age group. Random selection of animals and analyses performed were adequate to handle and compare data between unequally sized age groups. Greater than four-year-old ewes had significantly increased probabilities of developing IFR compared to three- and four-year-old ewes. This is consistent with several studies in meat sheep [16,43,74], and is probably the result of increased susceptibility to footrot and hoof deterioration with age [16]. This is also supported by several studies in dairy cows [78,79,80,81,82,83], where increased prevalence of infectious foot-diseases with age is attributed to (i) hoof deformities associated with metabolic and other stressors [78,84], (ii) changes in the loading capacity of the sole and other soft tissues of the feet [85], (iii) longer exposure to housing environment and pathogens [78], and (iv) inadequate treatment and relapsing foot lesions [81].5. ConclusionsA remarkable prevalence of foot-related lameness, foot diseases and lesions in intensive dairy sheep farms in Greece has been evidenced, indicating an increasing trend over the last decade. This underpins the urgent demand for targeted mitigation strategies against foot-health challenging conditions and particularly infectious foot diseases (i.e., IFR and OID). These strategies need to adjust on the peculiarities of intensive livestock farming and should be mainly based on (i) hygiene and biosecurity measures, (ii) preventive veterinary measures and control of foot diseases and lesions, and (iii) training of farmers on foot care practices and holistic management of foot health and lameness. Among the studied risk factors, age, and BCS were associated with infectious foot diseases; however, further investigation of potential risk factors in cohort and large-scale studies are crucial to elucidate their effects. Building on current knowledge and covering of the gaps in the epizootiology and potential risk factors of foot-related lameness, diseases, and lesions will further enhance health and welfare of dairy sheep and the sustainability of farms.

animals : an open access journal from mdpi

10.3390/ani11030884

PMC8003826

Increased growth rates of ewe lambs between three and seven months of age can potentially have negative impacts on mammary development and milk production, affecting their capacity to wean a lamb as yearling ewes. This experiment was designed to examine the impacts of an increased growth rate of ewes between weaning and their first breeding at seven months of age on mammary development using ultrasonography and to establish if mammary ultrasound measures could be indicators of growth of yearling ewe progeny. Mammary measures were taken in late pregnancy, early lactation and weaning in 59 single-bearing yearling ewes either preferentially fed and achieving 47.9 kg at breeding at seven months of age, or fed to achieve 44.9 kg at breeding. Mammary measures did not differ between live-weight gain treatments, indicating no evidence of negative effects on mammary development of yearling ewes. Some mammary measures, however, were positively associated with the growth of the progeny to weaning suggesting that ultrasonography has the potential to identify yearling ewes that would wean heavier lambs.

The experiment aimed to examine the impacts of an increased growth rate of ewes between three and seven months of age on udder development using ultrasound and to establish whether ultrasonography could be used to identify ewe mammary structures that may be indirect indicators of singleton growth to weaning. Udder dimensions, depths of gland cistern (GC), parenchyma (PAR) and fat pad (FP) were measured in late pregnancy (P107), early lactation (L29), and at weaning (L100) in 59 single-bearing yearling ewes selected from two treatments. The ‘heavy’ group (n = 31) was preferentially fed prior to breeding achieving an average breeding live-weight of 47.9 ± 0.38 kg at seven months of age. The ‘control’ group (n = 28) had an average breeding live-weight of 44.9 ± 0.49 kg. Udder dimensions, GC, PAR and FP did not differ between treatments. Lamb growth to L100 was positively associated (p < 0.05) with PAR at P107 and GC at L29. There was no evidence of negative effects of the live-weight gain treatments on udder development of yearling ewes as measured by ultrasonography. The results suggest that this ultrasound method has the potential to identify pregnant yearling ewes which would wean heavier singletons.

1. IntroductionA major determinant for achieving puberty and successful breeding of yearling ewes at seven months of age is the attainment of 40–70% of mature live-weight [1,2]. Yearling ewes that weighed 40 to 45 kg at breeding had greater performance than those bred at 35 kg or below, therefore Kenyon et al. [2] recommended a minimum live-weight of yearling ewes of 40 kg at breeding. Further, heavier live-weights at breeding have been shown to improve the reproductive performance of yearling ewes resulting in a greater number of yearling ewes mated during the breeding period, increased fertility rate, litter size, and lambing percentage [2,3]. Using this combined knowledge, farmers aim to feed their Romney-type yearling ewes to achieve suitable growth rates post-weaning to ensure they reach live-weights greater than 40 to 45 kg at breeding at seven months of age. Increased growth rates prior to puberty, however, have been reported to have negative impacts on mammary gland development and milk production in yearling ewes [4,5]. McCann et al. [6] reported that yearling ewes grown at higher rates between weaning and breeding at seven months of age produced less milk than yearling ewes with lower growth rates, indicating an impairment of the mammary gland development and function. Farmers, therefore, need to balance the desire for heavier live-weights at breeding to improve yearling ewe reproductive performance while limiting any potential negative impacts on lactation performance and growth of the progeny to weaning.Yearling ewes have a period of accelerated growth of parenchymal mammary tissue, called allometric phase, between two and five months of age [7,8]. During this period, the ductal network of the mammary gland expands extensively into the mammary fat pad [9]. The development of the ductal network during this period will determine future alveolar development, and therefore future milk production [5,10]. A high plane of nutrition prior to puberty has been reported to reduce the development of parenchyma [5,8] and increase fat accumulation in the fat pad [8,9], which combined may explain the reported subsequent lower milk production [5,11].Ewe mammary internal structures can be visualized using ultrasonography [12]. Specifically, ultrasound has been used to investigate the mammary parenchyma [13,14] and the mammary gland cistern (sinus lactiferous) [15,16]. In dairy ewes, mammary morphology and milk production were reported as traits of interest in genetic selection, leading to an increase in mammary size and milk yield compared to non-dairy breeds [17,18]. In addition, currently, most studies that have examined mammary structures using ultrasonography have utilised dairy breed ewes [12] with only a small number of studies examining dual-purpose meat and wool breeds [15,19]. Ruberte et al. [19] examined the relationship between mammary ultrasound images and mammary anatomy of mature ewes, whereas Caja et al. [15] focused on measures of cistern size using ultrasound and its correlation with milk yield of mature ewes. Currently, no studies have used ultrasound to examine the mammary gland of non-dairy yearling ewes. Over the last 20 years, ultrasound technology has improved, allowing for more detailed examination of the mammary structures through greater image resolution which allows the assessment of the development of the parenchyma and identification of abnormalities in the parenchyma [12]. The present experiment was the first, to these authors’ knowledge, to utilise ultrasound on non-dairy yearling ewes during their first pregnancy and lactation and to investigate the relationship between ultrasound measurements of yearling ewe udders and the growth of their progeny.The primary objective of this experiment was to investigate the effects of an increased growth rate between weaning (three months of age) and breeding (seven months of age) on mammary gland dimensions and structures of single-bearing Romney yearling ewes during their first pregnancy and lactation using ultrasonography. It was hypothesised that yearling ewes with an increased growth rate between weaning and breeding (heavy) would have a smaller parenchymal area and a greater fat pad area than yearling ewes with a lower growth rate between weaning and breeding (control). The second objective was to develop an ultrasound technique to identify mammary internal structures that could be used as indirect indicators of the growth to weaning of the progeny of yearling ewes. It was hypothesised that the ultrasound measurements of the young dam’s mammary gland in lactation would be correlated with the early growth rate of their progeny.2. Materials and MethodsThe experiment was conducted at Massey University’s Riverside Farm (latitude: 40°50′35″ S, longitude: 175°37′55″ E), 10 km north of Masterton, New Zealand. All animal handling procedures were approved by the Massey University Animal Ethics Committee (MUAEC-17/16).2.1. Experimental Design2.1.1. BackgroundAs previously described by Haslin et al. [20], at weaning, at approximately 86 days of age (127 days prior to breeding; P-127), 270 twin-born Romney ewe lambs (hereafter called yearling ewes) were allocated to one of the two treatment groups using a stratified random sampling procedure to ensure that the average live-weight of the groups did not differ (28.6 ± 0.16 kg). The intent of this experiment was to bring yearling ewes to different live-weight targets at their first breeding (P0) at seven months of age. The ‘heavy’ group (n = 135) was preferentially fed until breeding (10/05/2018; P0) achieving an average live-weight of 47.9 ± 0.38 kg. The ‘control’ group (n = 135) had an average live-weight of 44.9 ± 0.49 kg at P0. Both groups grazed lucerne sward (Medicago sativa L.; heavy for 105 days (P-127 to P-22) and control for 100 days (P-127 to P-27)), then ryegrass (Lolium perenne L.) and white clover-based sward (Trifolium repens L.) for 22 days heavy (P-22 to P0) and 27 days control (P-27 to P0), and both were offered a cereal-based concentrate feed (CP 10.5%, NDF 17.6%, ADF 7.1%, ME 12.8 MJ/kg DM) pre-breeding (Figure 1). Cereal-based concentrate feed mass offered to the heavy group at a rate of 200 g/yearling ewe/day for 68 days (P-119 to P-51) and 300 g/yearling ewe/day for 51 days (P-51 to P0; Figure 1). Yearling ewes in the control group were offered 200 g/yearling ewe/day for 43 days (P-94 to P-51; Figure 1). Individual animal feed intake was not measured. Pasture allowances were controlled using a rotational grazing system and by differing pre- and post-grazing herbage sward heights. All yearling ewes were managed as a single mob from P0 and bred for two 17-day periods (P0 to P34) to crayon-harnessed Romney rams at a ratio of 1:40 [3]. Yearling ewes were identified as mated in the first 17-day period by recording the presence of a crayon colour mark on their rump [3]. Pregnancy diagnosis was determined at 84 days of pregnancy (P84) using transabdominal ultrasound (Figure 1).2.1.2. Present StudyRomney yearling ewes from each treatment group were randomly selected at pregnancy diagnosis (approximately 10 months of age) to select only yearling ewes mated during the first 17 days of breeding that were identified as single-bearing (P84; heavy, n = 31, 52.3 ± 0.85 kg and control, n = 28, 51.4 ± 0.85 kg; Figure 1). Only single-bearing yearling ewes were selected as yearling ewes carrying singles are more frequent than those carrying twins [21]. At 138 days of pregnancy (P138), yearling ewes from both treatment groups were randomly assigned to one of four lambing paddocks (average stocking rate 8.02 yearling ewe/ha; heavy n = 8, 6, 9, 8 and control n = 9, 5, 9, 5 per lambing paddock) to ensure yearling ewes from each treatment group in each paddock. Cross-suckling was not controlled, as ewes and lambs are developing an exclusive bond [22] making cross-suckling not frequent. Two yearling ewes in the heavy group died during the lambing period. All yearling ewes lambed within 15 days (1/10/2018 to 16/10/2018). The lactation period was deemed to have begun after the first lamb had been born from all yearling ewes (1/10/2018; L1) and all lambs were weaned at approximately 100 ± 4 days of age (17/01/2019; L100; heavy n = 24 and control n = 24).From P0 to L100, both groups were managed and grazed together using a rotational grazing system on ryegrass and white clover pasture under commercial New Zealand grazing conditions. The pre-grazing pasture mass during pregnancy and lactation was on average 868 ± 39 and 1265 ± 58 kg DM/ha, respectively. Due to low pasture availability in pregnancy, all yearling ewes were offered lucerne bailage (CP 13.3%, NDF 48.8%, ADF 36.1%, ME 9.6 MJ/kg DM) at a rate of approximately 1.0 kg/yearling ewe/day from P34 to P138.2.2. Animal MeasurementsUnfasted live-weights of yearling ewes were recorded at P-127, P0, P84, P107, L29 and L100. Body condition score (BCS) of yearling ewes were recorded at P0, P84, P138, L29 and L100. Lambs were tagged within 18 h of birth, during twice-daily lambing rounds, at which time their date of birth, paddock, sex, dam number and birth weights were recorded. Lambs were reweighed at L29 and L100.2.2.1. Udder Score and MorphologyYearling ewe udder scoring, and morphological trait measurements were performed at P107, L29 and L100 by a single trained operator. The scoring system, adapted from Griffiths et al. [23], assessed udder health and included the palpation of both udder halves and teats (Table 1). Yearling ewes were placed in a sitting position to allow access to the udder for palpations.Morphological traits were measured while yearling ewes were standing and included udder circumference (UC, cm) measured above the teats [24], using a tape (Scrotal Measuring Tape, Shoof international LTD, New Zealand), and the height of each udder half (cm), using a ruler to measure the distance between the rear udder attachment along the outside edge of the udder, and the udder floor [25] (Figure 2). Udder volume (UV, cm3) was calculated using UC and an average of udder height (UH) according to Ayadi et al. [25] (1) and (2). R = UC/2π(1) UV = π × R2 × UH,(2) where UV = udder volume (cm3); π = 3.14159; R = radius (cm); UH = udder height (cm); UC = udder circumference (cm).2.2.2. Ultrasound ScanningUltrasound scans were performed by a single operator, at P107, L29 and L100 (Figure 1). At L29 and L100, ultrasound scans were not conducted for yearling ewes whose lambs had died (heavy n = 5 and control n = 4). At L29 and L100, yearling ewes were separated from their lambs four hours prior to the ultrasound scanning to allow the udder to accumulate milk according to Ruberte et al. [19] and Caja et al. [15]. Yearling ewes were placed in a sitting position (i.e., shearing position; Figure 3b,c) to allow easy access to the udder. Ultrasound scans were performed with an ultrasound scanner fitted with a linear transducer with 5.0–10.0 MHz imaging frequency (Sonosite M-Turbo Ultrasound with L38xi, Bothell, Washington, DC, USA). Vegetable oil was used as a coupling gel. The transducer was applied on the external base of each teat at an approximate angle of 30° from the caudal-cranial axis (Figure 3a) with an inclination of approximately 45° in relation to the teat [26] (Figure 3b,c). A light and consistent pressure was applied to the udder through the transducer to minimise variations related to pressure on the images. There was variability in the position of yearling ewes but, the effects of these variations were minimised by indicating to the handler on which position (on right of left leg) the ewe had to be sited for the ultrasound scan (Figure 3b,c), and by identifying the most representative and consistent mammary structures on the images during the scan prior to capturing the images.A minimum of three images were taken from each udder half. Images included the gland cistern, mammary parenchyma, putative fat pad and the boundary between the mammary gland and the abdominal wall. One image of suitable resolution per udder half, where all structures were identifiable and present was selected for image processing [27]. Udder halves with an udder palpation score of 4 or 5 (Table 1) at a specific time point (P107, L29 or L100) were considered “abnormal” [23] and were not included in the image selection (heavy: 1 ewe with 1 half and control: 2 ewes with 1 half each).The image processing was undertaken using ImageJ software [28] as used by Abràmoff et al. [29]. The scales between pixels and millimetres were calculated based on the number of pixels, the scanning depth (mm), and the transducer width (mm) (Figure 4). This method relies on the ability of the operator to interpret and identify lines on the images. To standardize the assessment compartment depth, drawing templates were created for each time point as used by Molenaar et al. [30] and included four representative images from four different yearling ewes with and without the lines drawn for each compartment (Appendix A). The total depth of mammary gland conservative (MTc) was the smallest likely demarcation (abdominal wall) of the mammary gland (Figure 5a), and total depth of the mammary gland generous (MTg) was the largest likely demarcation of the mammary gland visible on the image [30] (Figure 5a). The MTc, MTg, fat pad (FP), parenchyma (PAR), and gland cistern (GC) depths were estimated at the deepest point for each sub-compartment, excluding the skin layers, using the straight tracer (Figure 5a) and were expressed in millimetres.To assess the development of the parenchyma at P107, L29, and L100, three regions of interest (ROI; [26]) were randomly drawn in the parenchyma area, each square measured 6.7 mm2 (Figure 5b). The brightness of each pixel corresponded to echogenicity and was numerically represented on a scale of 256 levels of grey [31]. Echogenicity is defined as the capacity of tissues to interact and reflect the sound waves of the transducer [32]. This capacity varies with tissues, i.e., liquids have very low echogenicity [32] and fat has greater echogenicity but attenuates as the depth increases [30].2.3. Statistical AnalysisAll statistical analyses were conducted using SAS v9.4 (SAS Institute Inc., Cary, NC, USA). Yearling ewes that died (heavy n = 2) or whose lambs died prior to L100 (control n = 4 and heavy n = 5) were excluded from the experiment. The final dataset included 24 yearling ewes and their singletons in each treatment group and a total of 284 images.Growth of yearling ewes from P-127 to P0 was analysed using a linear mixed model including treatment group (control vs. heavy) as a fixed effect and age at P-127 as a covariate. Live-weight of yearling ewes at P107, L29 and L100 was analysed using a linear mixed model allowing for repeated measures. The model included treatment group, day of measurement (P107, L29 and L100) and their two-way interaction as fixed effects. Lambing date was fitted as a covariate in the model as used by Pettigrew et al. [33]. The BCS of yearling ewes was analysed using a generalized linear model allowing for repeated measures with a Poisson distribution and a log transformation. Treatment group, day of measurement (P0, P84, P138, L29 and L100) and their two-way interaction were included as fixed effects. Growth of the progeny from birth to L29 and from L29 to L100 was analysed using a linear mixed model allowing for repeated measures, and including treatment group, time (birth to L29 and L29 to L100) as fixed effects, date of birth as a covariate and lambing paddock as a random effect.The ROI grey-scale values, GC, FP, PAR, MTc and MTg of the right and left udder halves were analysed using general linear mixed models allowing for repeated measures. These models included udder half (right vs. left), day of measurement, treatment group and two-way interactions of udder half and day of measurement and treatment group and day of measurement as fixed effects, with a Tukey–Kramer adjustment, lambing date as a covariate and yearling ewe as a random effect. The grey-scale, GC, PAR, MTc, MTg did not differ (p > 0.05) between udder halves, therefore, an average of the two halves was calculated for each day of measurement and used in further analyses. For FP, udder halves were significantly (p < 0.05) different at L100 and thus the FP measures of the right and left halves at L100, remained separated in the analyses.To determine the effect of individual yearling ewe live-weight at P0 on the grey-scale values, GC, PAR, FP, MTc and MTg, the linear mixed models were re-run without treatment as a fixed effect and including udder half, day of measurement and their two-way interaction as a fixed effect, lambing date and live-weight at P0 as a covariate, a two-way interaction between live-weight at P0 and day of measurement, and yearling ewe as a random effect.The residuals were generated using general mixed models. Ewe live-weight, ewe BCS, UV, UH, UC and MTg were adjusted for treatment group and lambing date. In the model, PAR, GC and MTc were adjusted for the treatment group, MTg and lambing date. Lamb growth from birth to L29, L29 to L100 and birth to L100 were adjusted for the treatment group, lambing date, and sex of lamb. Pearson correlations were used to test for linear associations between the residuals of ewe live-weight, UC, UH, UV, GC, PAR, FP, MTc and MTg at each time point (P107, L29, L100) and lamb growth from birth to L29, from L29 to L100 and from birth to L100.Multiple regression analyses of lamb growth from birth to L29 and from birth to L100 were carried out using general linear models. General linear models were used to examine whether each predictive variable was individually correlated with lamb growth. Predictive variables correlated with lamb growth with p ≤ 0.20 were included in the model [34]. Correlations between selected predictive variables were examined to identify high collinearity, resulting in Equations (3) and (4) respectively. Lamb growth from birth to L29 = GC at L29 + MTc at L29 + BCS at P0(3) Lamb growth from birth to L100 = PAR at P107 + FP at P107 + Ewe LW at L29 + GC at L29 + MTc at L29(4)Backward manual variable eliminations were used to select the model that best explained the variation in lamb growth from birth to L29 and to L100 by removing predictive variables with p > 0.10. Confounding effects were evaluated after each variable removal and were examined by checking the changes in predictive variable coefficients. Any non-significant predictive variable causing greater than a 20% change in the model coefficients was considered a confounding variable and included in the model [34].3. Results3.1. Growth and Live-WeightYearling ewes from the heavy group had greater growth rates between P-127 and P0 than control yearling ewes (p < 0.05; 147 ± 4.4 vs. 133 ± 4.4 g/d, respectively) resulting in a tendency for different live-weight at breeding (p = 0.09, 47.5 ± 0.71 vs. 45.8 ± 0.71 kg, respectively). Live-weight of yearling ewes, however, did not differ (p > 0.05) between treatments at P107, L29 or L100 (Table 2).Yearling ewe BCS did not differ (p > 0.05) between treatment groups at any time point (Table 2). Ewe BCS did not differ (p > 0.05) between P0, P84 and P138, which in turn were greater (p < 0.05) than BCS at L29 and L100, which did not differ (p > 0.05; Table 2).Lamb live-weights at birth, L29 and L100 (Table 2) and lamb growth from birth to L29, from L29 to L100 (average 340.8 ± 13.5 g/d and 201.5 ± 8.19 g/d, respectively) and lamb growth from birth to L100 [35] did not differ (p > 0.05) between treatments.3.2. Udder Scores and MorphologyThe udder scores, UH, UC and UV at P107, L29 and L100 were presented in Haslin et al. [35]. Briefly, teat palpation score, udder depth score, the proportion of asymmetric udder and dimensions (UH, UC and UV) did not differ (p > 0.05) between treatment groups at any time point. The control group had greater (p < 0.01) udder palpation scores at P107 than the heavy group [35].3.3. Ultrasound MeasurementsThe depth of the gland cistern (GC), parenchyma (PAR), total mammary conservative (MTc), total mammary generous (MTg) and the grey-scale value did not differ between udder halves (p > 0.05; data not shown). The depth of the fat pad (FP) did not differ at P107 between udder halves (p > 0.05; data not shown), but at L100, the left udder half had a deeper FP than the right udder half (p < 0.05; 19.2 ± 0.91 left vs. 15.4 ± 0.99 right).The depths of GC, PAR, FP, MTc, MTg and the ROI grey-scale values did not differ (p > 0.05) between treatment groups at any time point (Table 3). Live-weight at P0, irrespective of treatment group, had no effect (p > 0.05) on GC, FP, MTc, MTg and ROI grey-scale values but negatively impacted PAR at P107 (p < 0.05; estimate −0.20 mm).The depth of GC, PAR, MTg and MTc were greater at L29 (p < 0.001) than L100 which was, in turn, greater (p < 0.001) than P107, irrespective of treatment groups. The depth of FP was greater at L100 than at P107, irrespective of treatment groups (p < 0.001; Table 3).3.4. Correlations between Udder Measurements, Ewe Live-Weight, Ewe BCS and Lamb GrowthAt P107, UC was positively correlated with UV, UH and PAR (p < 0.05), and GC and PAR were negatively associated with FP (p < 0.01; Table 4). BCS of yearling ewes at P138 was positively associated (p < 0.01) with FP at P107 and live-weight at P107, and negatively associated (p < 0.05) with PAR at P107 (Table 4). At L29, UV was positively correlated with UH and UC (p < 0.05), and PAR was negatively correlated with GC (p < 0.05). Live-weight of yearling ewes at L29, irrespective of treatments, was positively associated with BCS at L29 (p < 0.01) and UC (p < 0.05; Table 5). At L100, UV was positively correlated with UH, UC, FP of left half (p < 0.05), UH was positively associated with UC and FP of the left half (p < 0.05). Live-weight of yearling ewes was positively associated (p < 0.001) with BCS of yearling ewes at L100 (Table 6) At L100, FP of the left half was negatively correlated with PAR (p < 0.01; Table 6).Lamb growth from birth to L29, irrespective of treatments, was positively associated with GC at L29 (p < 0.05; Table 5) and FP at L100 on the right half (p < 0.05; Table 6). Lamb growth from birth to L29 to L100 was positively associated with PAR at P107 (p < 0.05; Table 4) and FP at L100 on the left half (p < 0.05; Table 6) but negatively associated with FP at P107 (p < 0.05; Table 4). Lamb growth from birth to L100 was positively associated with PAR at P107 (p < 0.01; Table 4), GC at L29 (p < 0.05) but negatively associated with PAR at L29 (p < 0.05; Table 5).3.5. Multiple Regression of Lamb GrowthThe best model explained 12.2% of the variation in lamb growth from birth to L29 included only the effect of GC at L29 (Lamb growth from birth to L29 = 261.4 (±35.1) + 4.9 (±1.9) GC at L29; estimate (±SE)). The difference between a yearling ewe with an average GC at L29 and a GC in the 90th percentile was 6.6 mm (Table 7), resulting in a 32.3 g/d difference in lamb growth from birth to L29.For the period birth to L100, the best model explained 37.6% of the variation in lamb growth and included the effect of PAR at P107, yearling ewe live-weight (LW) at L29 and GC at L29 (Lamb growth from birth to L100 = 37.4 (±48.5) + 7.2 (±1.9) PAR at P107 + 2.0 (±0.65) LW at L29 + 1.2 (±0.63) GC at L29). The difference between a yearling ewe with an average PAR at P107 and a PAR in the 90th percentile was 2.6 mm (Table 7), resulting in 18.7 g/d in lamb growth from birth to L100. The difference between a yearling ewe with an average GC at L29 and a GC in the 90th percentile was 6.6 mm (Table 7) resulting in 7.9 g/d in lamb growth from birth to L100.4. Discussion4.1. Treatment EffectsThe first objective of this experiment was to investigate the effects of increasing growth rates of yearling ewes between weaning and breeding on mammary gland dimensions and internal structures. It was hypothesised that yearling ewes with an increased growth rate would have a reduced mammary parenchymal and a greater fat pad area compared with yearling ewes with a lower growth rate. Despite differences in yearling ewe growth rate between weaning and breeding, depth of the gland cistern, parenchyma, fat pad and total depth of the mammary gland conservative and generous did not differ between treatments, which contrast with previous studies [5,8,11]. The differences in yearling ewe growth rates between treatments from weaning to breeding, however, were small (14 g/d) and the magnitude of the growth (less than 150 g/d on average) was lower than those achieved in previous studies (i.e., from 173 to 305 g/d; [8,11]). The small differences in the growth rate of yearling ewes between treatments in the present experiment (14 g/d) may have impacted the ability of the treatments to alter mammary gland development. There are also differences in breeds between studies that may also explain the difference between the results of the present experiment and previous studies (Hampshire Down [8], Suffolk crossed Dorset and Suffolk [11], Dorset [5]). In the present experiment, there was no evidence of negative impacts on mammary gland development of Romney-type yearling ewes achieving live-weight gains of approximately 147 g/d between their weaning at three months of age and their first breeding at seven months of age and reared under New Zealand conditions.No differences were found between the right and the left udder half for total depths of the mammary gland, depths of the gland cistern and parenchyma. This finding is consistent with previous studies [13,14]. At weaning, however, the fat pad differed between udder halves. This difference could be explained by a preference of the lamb for one udder half, which may have been over-stimulated [36], in particular for ewes rearing a singleton. Overstimulation of one udder half may lead to an early partial udder involution of the non-preferred udder half resulting in changes in the mammary tissues [37].4.2. Associations between Udder Measurements and Lamb GrowthThe second objective of this experiment was to develop an ultrasound technique to identify mammary internal structures that could be used as indirect indicators of lamb growth to weaning. Mammary parenchymal depth in late pregnancy measured by ultrasound was moderately and positively associated with lamb growth to weaning. Singletons born to yearling ewes with a large parenchyma in late pregnancy being 1832 g heavier at weaning at 100 days of age than singletons born to yearling ewes with an average parenchyma in late pregnancy. The mammary parenchyma includes the secretory tissue involved in the production and secretion of milk [38,39] with the number of secretory cells determining milk production [40,41]. Thus, a deeper parenchyma in pregnancy could indicate a greater number of secretory cells and potentially greater milk production. Strzetelski et al. [32] reported in primiparous dairy heifers that the percentage of secretory tissue measured by ultrasound was highly positively correlated with milk production. While the impact of the parenchymal depth in late pregnancy on singleton growth to weaning was moderate, the results suggested that the ultrasound method could potentially be used as a technique to identify pregnant yearling ewes that will wean heavier singletons.The gland cistern depth in early lactation was moderately and positively correlated to lamb growth from birth to early lactation and to weaning. Singletons born to yearling ewes with a large gland cistern in early lactation being 938 g and 790 g heavier at 29 and 100 days of age respectively, than singletons born to yearling ewes with an average gland cistern in early lactation. The mammary gland cistern is the cavity that milk drains into between milkings or suckling events [38]. Ewes with larger cisterns were reported to produce more milk than ewes with smaller cisterns [15,16,42]. It is likely, therefore, that yearling ewes with larger gland cisterns in early lactation would have greater milk production than those with smaller cisterns. The associations between ultrasound measurements and milk production in non-dairy yearling ewes, however, are still unknown. These results again suggested that ultrasound has the potential to identify yearling ewes that would wean heavier singletons. Further research is warranted to confirm these findings and investigate the use of ultrasound to examine the association between mammary gland, milk production and lamb growth in non-dairy yearling ewes.The ultrasound scan in late pregnancy was easier to perform, as lactating yearling ewes and lambs had to be separated and were off feed for four hours pre-scanning, whereas late-pregnant yearling ewes did not require a waiting period prior to scanning. The measurement of the parenchyma in late pregnancy relies on the ability of the operator to identify the demarcation between tissues which can be difficult due to the image resolution and the attenuation of signal as the scanning depth increases [30]. The measure of the gland cistern did not so much rely on the operator ability, as the gland cistern appears clearly as black on the image [15,30]. The precise measure of the parenchyma and gland cistern with this ultrasound technique could not be performed at scanning as the correspondence between millimetres and pixels varies depending on the scanning depth. The ultrasound technique used in this experiment would therefore be challenging to apply on larger flocks of yearling ewes. More research is required to improve the ultrasound technique for its potential application on larger flocks.5. ConclusionsThis experiment was the first to use ultrasound to investigate the relationship between udder measurements and the growth of progeny of non-dairy yearling ewes during pregnancy and lactation. The depth of the mammary parenchyma in late pregnancy and of the gland cistern in early lactation were indicators of growth from birth to weaning of singletons. Under the conditions of this experiment, there was no evidence of negative effects of the differing live-weight gain treatments between three and seven months of age on mammary gland development of Romney-type yearling ewes during their first pregnancy and lactation as measured by ultrasonography. The results of the association between lamb growth and mammary ultrasound measures suggest that the ultrasound technique used in this experiment has the potential to identify pregnant yearling ewes which would wean heavier singletons. More research is needed to investigate the use of ultrasound to examine the associations between mammary ultrasound measurements, milk production and lamb growth in non-dairy yearling ewes.

animals : an open access journal from mdpi

10.3390/ani11030890

PMC8003938

Owner-based reports of dogs presumed to come from commercial breeding kennels (CBKs) suggest high levels of fear in this population. Fear in kenneled dogs is a serious behavioral welfare concern as it may lead to both acute and chronic stress. Novel social and non-social stimuli have been shown to elicit behaviors associated with fear in animals. New knowledge on the levels of fear in dogs from CBKs could be used to further refine protocols intended for assessment of welfare in CBKs and to improve breeders’ management practices. The aim of this study, therefore, was to evaluate how dogs from CBKs reacted to social (i.e., a person approaching) and non-social (i.e., a traffic cone and a dog statue) stimuli, and to perform a preliminary evaluation of how these responses might be used as indicators of dogs’ overall socialization levels in kennels. Results revealed that dogs had primarily mildly fearful responses to the stimuli presented. These findings are encouraging as extreme fearful reactions were rarely recorded. Nevertheless, there is a clear margin for commercial breeders to improve the socialization protocols in their kennels to better incorporate both social and non-social stimuli.

Understanding the behavioral welfare of dogs in commercial breeding kennels (CBKs) is important for improving breeders’ management practices as well as dog welfare. In the current study, breeding dogs from CBKs were exposed to novel stimuli to evaluate their behavioral responses, with emphasis on indicators of fear. Subjects were presented with a standard stranger-approach test, a traffic cone, and a realistic dog statue. Sixty dogs were exposed to the three stimuli and behavioral responses were scored using an ethogram developed for this study. Dogs spent significantly more time investigating the environment, staying further away from the stimulus, and they took longer to approach and investigate when presented with the cone than with the dog statue or stranger (p < 0.01). These findings suggest that the cone elicited more fear-related behaviors than the dog statue and stranger. Given these results, in addition to socializing their dogs to unfamiliar people and other dogs within their kennels, commercial breeders should be encouraged to increase the exposure of their dogs to more diverse novel stimuli to reduce non-social fear and support the welfare of dogs while they reside in the kennel and when they transition to new homes.

1. IntroductionUntil recently, studies focusing on the behavior and welfare of dogs from commercial breeding kennels (CBKs) were scant [1]. While research on the behavior of these dogs is still fairly limited, basic knowledge from investigations of similarly confined dog populations, such as those kept in shelters or laboratories, may have implications on the lives of dogs from CBKs [2,3,4]. For example, studies have illustrated the importance of housing quality, predictability of the social environment and frequency and quality of human–animal interactions for animal welfare and longevity [5,6,7,8]. Previous research on the welfare of dogs presumably from CBKs conducted using data originating from dog owner reports [9] has lacked empirical evidence from direct observations. However, with advances in the development of dog welfare assessment tools and increased direct access to commercial dog breeding premises [10,11,12,13], researchers are beginning to better understand the behavior and health status of this population and how to assess their overall welfare in a more holistic way. It is important for researchers, breeding facility operators, and inspectors to have access to highly functional, validated, and easy to use tools designed to measure canine welfare in the field. For example, the development of the Field Instantaneous Dog Observation (FIDO) tool, which offers a means to assess the physical and behavioral welfare states of dogs in CBKs, including their physical and behavioral health, has played an important role in the move toward the collection of direct observational data from this population [10].An animal’s welfare can be greatly impacted by high levels of fear or a predominantly fearful emotional state, which may lead to both acute and chronic stress [14,15]. Dogs experiencing acute stress are more likely to exhibit submissive and/or fear-related behaviors, such as paw-lifting and lowered postures, whereas prolonged chronic stress has been shown to induce behavioral stereotypes [5]. Understanding the levels of fear in CBK populations, and how fear may impact dogs’ quality of life after transitioning out of the kennel, is a necessary line of inquiry to ensure their welfare. At the end of their breeding careers, eligible dogs from CBKs may be rehomed as pets. For inadequately socialized dogs, experiencing high levels of social and non-social fear could seriously impact their ability to transition smoothly to their new homes. This is critical because dogs that present major behavioral problems once rehomed may be at high risk of abandonment, surrender to a shelter or euthanasia [6]. Stella and colleagues [11] suggest that the assumption that all retired dogs from CBK are equally good candidates for rehoming might significantly compromise the safety and welfare of those more fearful individuals. The authors advised that, for these dogs, there should be a greater degree of socialization in place to help prepare them for such transitions.Both social and non-social stimuli have been shown to elicit behaviors associated with the flight or fight response in dogs, with social stimuli provoking a higher frequency of fear-related behaviors [15]. However, to date, there are no published studies that quantify the levels of fear in dogs from CBKs based on their behavioral reactions to both social and non-social stimuli. Such information could potentially be used to further refine behavioral assessment protocols intended for assessment of welfare in CBKs and to maximize rehoming success after retirement. Additionally, better understanding of social and non-social fear responses in this population of dogs could help to inform standards of care and management practices for dogs raised in CBKs. The aim of this study, therefore, was to evaluate how dogs from CBKs reacted to social (i.e., an unfamiliar person approaching) and non-social (i.e., a plastic traffic cone and a dog statue) stimuli, and to gauge the degree to which behavioral responses to exposure to these stimuli might be used as indicators of the adequacy of socialization practices used in CBKs.2. Materials and Methods2.1. Ethics StatementThe procedures described were reviewed and approved by the Purdue University Institutional Animal Care and Use Committee (PACUC 1809001796), and permission to visit the kennels and record the videos was granted by the owners prior to the commencement of the study.2.2. SubjectsThe subjects for the current study (n = 60) were randomly selected from a pool of 447 dogs from 26 CBKs, located across the Midwestern United States. These subjects were part of a larger data collection effort for another study (ongoing) which aimed to investigate management and welfare risk factors affecting rehoming outcomes in retiring dogs from CBKs. Within that larger study, dogs were assessed using a refined version of the FIDO tool [10] and categorized as “red”, “yellow” or “green” (RYG). As reported by Bauer et al. [10], the assessment was based on an unfamiliar person approaching the front of the pen, maintaining a sideways orientation, and scoring the immediate behavioral reaction of the dog. A dog was scored “red” if it showed signs of fear and/or stereotypic behavior; “green” if it showed affiliative approach, solicited attention or was undisturbed by the presence of the approaching tester; or “yellow” if it showed an ambivalent approach/avoidance behavior or could not clearly be scored “red” or “green”. Tests were videotaped for later analysis. The subsample of 60 dogs was selected from the main data spreadsheet (Excel) using a random number generator (google.com). The subsample included 36 dogs scored as “green”, 11 scored as “yellow”, and 13 scored as “red”. There were 15 males and 45 females. Furthermore, this population consisted of a variety of 26 purebreds and designer crossbreeds (Table 1).The average age of the dogs was 3.4 years (range = 1 to 7, SD = ±1.42). Physical health data collected using the FIDO tool [10] showed overall good health conditions for all dogs: no coughing, sneezing, lameness, nasal discharge or wounds were observed and body condition was normal to stout. Tear staining or ocular discharge was observed in half of the dogs. Dogs had clean coats: only 10/60 dogs had mild spotting (i.e., less than 25% of the body wet or with debris).The sampled dogs were from 20 different commercial breeding facilities, representing different housing systems, varied flooring types/materials (e.g., concrete, tenderfoot), indoor pen sizes (ranging from 0.7 to 4.5 m2, mean = 2.3 m2, SD = 1.2 m2), varied access to the outdoors (e.g., indoor only or with free indoor/outdoor access), and access to separate exercise areas (e.g., additional daily or weekly access to separate outdoor exercise yards). The breeders enrolled in the original study volunteered their participation, and their kennels all exceeded minimum space and exercise requirements for dogs mandated by U.S. federal law [16].2.3. ProcedureThe three stimulus–response tests used in this study were conducted by two female researchers. During the test, dogs were confined into the indoor portions of their home pens (i.e., they had no access to the outdoor) to ensure availability for scoring. As the indoor pen size varied in dimension as previously described, and because dogs were free to move within the pen space (i.e., the tested dog was not positioned at a specific starting location to avoid additional handling stress), the distance between the stimuli and the dog at the start of the test varied. The first stimulus–response test conducted was a three-step stranger approach in which the tester: (1) opened the pen door with a sideways orientation and without making direct eye contact with the focal dog, (2) offered a treat to the dog directly from her hand, always maintaining a sideways orientation, and (3) offered a second treat from one hand while reaching to gently touch the dog with the other hand. Finally, the tester stepped back from the pen and closed the gate (adapted from [11]). The second response test consisted of placing a plastic orange traffic cone into the pen with the dog and locking the pen gate. The cone was left in the pen and the dog was allowed to explore it for 30 sec before the tester removed it. The final stimulus–response test included the placement of a realistic dog statue (Boston terrier figurine, 40 cm height) in the pen with the dog. Again, the dog was allowed 30 sec to investigate the object before it was removed by the researcher. The objects were thoroughly disinfected between kennels. If the tested dog licked, chewed or eliminated on the objects, then these were cleaned and let to air out before moving to the next dog. It is important to note that this set of tests was selected in order to elicit a variety of responses, from interest to mild fear, without provoking extreme reactions of avoidance and aggression that could have harmed the animals.Each of these tests was video recorded using a digital video camera (Sony Handycam HDR-CX405) mounted on a tripod. Videos (3/dog = 180 video clips) were subsequently analyzed using the behavioral scoring software BORIS (Version 7.8.2). The dogs’ behaviors during each test were analyzed using an ethogram based on the available literature on fear and stress in dogs [5,14,15,17]. The behavioral variables were then grouped into eight main categories: fear, stress, aggression, stereotypic behaviors, activity, vocalization, and non-fearful investigation. The full ethogram used to code the videos is provided in Table 2. The ethogram was pilot tested with a small additional subset of videos (n = 9) not included in the analysis.2.4. AnalysisAll statistical analysis was performed using SPSS (IBM, Version 26). To assess intra-rater reliability, nine subjects were randomly selected (15% of the sample), and all three reaction tests for each subject were re-analyzed from video by the same observer two months after initial scoring. Levels of agreement were determined using Intraclass Correlation Coefficients and interpreted as follows: ICC < 0.50, poor agreement; 0.50–0.75, moderate agreement; 0.75–0.90, good agreement; >0.90, excellent agreement [18].For the following analysis, each dog acted as its own control. Wilcoxon sign-ranked tests were conducted to compare time spent at the front of the pen versus the back of the pen. This was used to gauge the general willingness of each dog to engage with or avoid the stimulus (i.e., stranger, traffic cone, or dog statue) positioned at the front of the pen. Kruskal–Wallis analysis of variance tests were conducted to compare the behavior of dogs across the three stimuli. For significant results, Wilcoxon tests were used for post hoc pair comparisons, applying Bonferroni correction (p < 0.016).Additional analyses focused on sex, age and pen size to examine differences between behavioral variables of interest. For these analyses, behaviors were considered independently of the stimuli (i.e., same behaviors were summed across stimuli). The breeds of dogs were also recorded; however, there were not enough subjects in each breed to have the requisite power to perform breed comparisons. A Wilcoxon sign-ranked test was used to compare behavioral differences of male versus female dogs. A simple logistic regression was calculated to indicate any possible associations between the behavioral variables and the dogs’ ages and pen sizes.Descriptive analysis was used to characterize behaviors that were only seen in the stranger approach test: specifically, the “affiliative approach to tester”, “contact/no contact”, and “takes treat” data. Finally, cross tabulations were created to explore whether showing higher levels of sociability towards unfamiliar people was associated with fewer signs of fear towards non-social stimuli. This was performed by comparing each dog’s “red”, “yellow”, or “green” classification from the behavioral portion of the previously conducted FIDO testing with their response to the cone and dog statue.3. ResultsDue to their very low occurrence, some behaviors were excluded from the analysis, including “growl”, “whine”, “body shake”, “yawning”, and “shivering”. Similarly, “biting” and “teeth baring” were never seen, and “pacing” and “circling” were only noted two and three times, respectively, out of the 180 observations. Behaviors associated with more intense fear reactions (i.e., “escaping” and “retreat”) were observed in two dogs (less than 2% of the sample) during the stranger approach test, in three dogs (less than 5% of the sample) during the response to the dog statue test and in four dogs (less than 7% of the sample) during the response to cone test. Thus, these behaviors were also excluded from the analysis. Behaviors not included in the analysis are indicated in Table 2. Other behaviors (i.e., “bark”, “paw-lifting”, and “lip-licking”) were combined into a larger “stress-related” category for analysis. Additionally, postures (i.e., “lowered posture”, “neutral posture”, and “rigid posture”), which were included as modifiers (e.g., attributes of behaviors), were collapsed into single independent variables (i.e., independently of the behavior they were associated with) for analysis.Intra-rater reliability analysis confirmed a good agreement (0.76), on average, across all behavioral variables. An excellent agreement was observed for 58.8% of the variables analyzed (10/17), while the remaining variables showed a moderate to good agreement (0.56–0.80, Table 3).When analyzing the time spent in the front or back portions of the pens during each stimulus–response test, a Wilcoxon test revealed that dogs spent significantly more time in the back of their pens, compared to the front, when introduced to the cone (Z = −2.113; p = 0.035). No significant difference was found for time spent in each portion of the pen when dogs were introduced to the other two stimuli. However, on average, dogs spent more time at the back of the pen (i.e., further away from the stimulus) with the dog statue (mean: back = 5.44 s; front = 3.16 s) and more time at the front of the pen (i.e., closer to the stimulus) during the stranger interaction (mean: back = 5.45 s; front = 7.14 s).Statistically significant differences in the dogs’ responses to the three stimuli for the following behaviors were demonstrated by the Kruskal–Wallis test: “investigating the environment” (χ2 = 59.8, p < 0.0001), “sniffing the stimulus” (χ2 = 75.8, p < 0.0001), “latency to approach and interact with the stimulus” (χ2 = 53.8, p < 0.0001), “low” posture (χ2 = 9.84, p = 0.007), “rigid” postures (χ2 = 15.3, p < 0.0001), and “walking” behaviors (χ2 = 28.5, p < 0.0001). Post hoc comparisons revealed that the subjects spent significantly more time “investigating the environment” when presented with the cone compared to the dog statue or the stranger (p = 0.0001, Figure 1a). Additionally, dogs spent significantly more time engaged in “sniffing” behavior and had a shorter latency to approach and interact when presented with the dog statue compared to the cone or the stranger (p = 0.0001 for both behaviors, Figure 1b,c). Pairwise comparisons also indicated that dogs spent more time displaying rigid postures when presented with the dog statue than when presented with the cone and the stranger (p = 0.010 and p = 0.016, respectively, Figure 1d). They also spent more time in lowered postures when presented with the dog statue compared to the stranger (p = 0.001, Figure 1e) but not with the cone. Finally, dogs spent less time performing “walking” behavior during the stranger approach test than they did when presented with either the cone or dog statue (p = 0.0001, Figure 1f).Further analysis revealed no significant differences in behavioral responses for all stimuli between male and female subjects. Pen size was found to influence a dog’s position within the pen: dogs in larger pens (i.e., more square meters of surface) spent less time in the back of pen compared to dogs in smaller pens (t = −2.08, p = 0.042). All other associations were non-significant.For the behaviors observed only during the stranger approach test, dogs spent on average 0.68 sec in affiliative behavior (2.27% of total observation time, range: 0 to 9.5 sec; SD = ±1.68). During the stranger approach test, the tester attempted to give the subject a treat from their hand as well as to touch the subject with their hand. In this interaction, the tester was able to give the dog a treat directly 30 times (50% of subjects) and to make contact 25 times (41.67% of subjects).Cross tabulations revealed how dogs’ reactions to the cone (Table 4) and the dog statue (Table 5) were associated with their levels of sociability towards people, based on the Red–Yellow–Green (RYG) assessment.4. DiscussionThe goal of this study was to evaluate the effects of different stimuli in eliciting social and non-social fear in a population of dogs from CBKs and to perform a preliminary evaluation of how these responses might be used as indicators of dogs’ overall socialization levels in kennels. In this study, we recorded intense fear responses, such as freeze, escape attempt, or aggression, elicited by the three selected stimuli in very few animals (maximum 4/60 dogs). Instead, milder signs of fearful reactions were observed, including preferences for the back portions of their pens away from test objects, actively avoiding and increasing the distance from the stimuli, taking longer to approach and interact with, or complete failure to come into contact with the stimuli. The absence of extreme or even relatively strong fear responses may indicate that these dogs were somewhat able to cope with the challenges presented and, in turn, suggests that even though there is ample space for improvement, some level of effective socialization may have already taken place at these kennels. It is important to note that these results may also have been mitigated by the test design, which was not intended to provoke intense fear responses. For completeness, it should be mentioned that the response recorded toward these stimuli may be exclusive to this study population. Due to the lack of an external control group, it is not possible to determine if the type of response these stimuli would elicit in other populations (such as pet and shelter dogs) differs from that of dogs from CBKs. This goes beyond the scope of this paper as our aim was not to compare the prevalence of fear behavior in CBKs to an expected outcome; however, it is something to keep in mind in future applications of this test. Finally, the dogs used in this study came from breeders who volunteered to participate, so results may not be reflective of all U.S. commercial breeding kennels.From our analysis, the orange traffic cone, a non-social stimulus, triggered significantly higher rates of fear-related behaviors compared to social (or presumed social) stimuli, such as the stranger and the dog statue. The dogs’ fear when introduced to the cone was expressed through more time spent in the back portion of the pen away from the cone (which was placed in the front portion of the pen). This position in the pen was not observed as frequently when the dogs were presented with the other two stimuli. It should be considered that not only was the back portion of the pen further away from the stimulus, but it typically gave dogs access to outdoor runs. It is therefore likely that dogs showing a preference for the backs of their pens may have been looking to escape even though they were not actively scratching at the doors. A second indicator of the uneasiness provoked by the cone was that dogs spent more time investigating the environment than the cone. While dogs may spend a considerable amount of time investigating new environments, the only new environmental element presented during this test was the cone itself. This extended activity may, therefore, have been a type of redirected investigative or coping behavior [19], suggesting that the cone was creating an emotional conflict and leading dogs to perform alternative investigative behaviors. As the cone represented a novel, motionless object to which this population of dogs had never been exposed, these findings may suggest that the dogs in our sample may not have been adequately exposed to enough novel objects during their lives in the kennels to have been unafraid of the traffic cone. This finding has welfare implications as dogs that struggle to cope when presented with a motionless object, such as a plastic cone, are also likely to struggle when exposed to an unfamiliar household once re-homed, thus potentially resulting in a stressful transition [11]. It could be argued that the presentation of a novel object is, by design, fear-inducing and that any dog may react fearfully to such an exposure, even if well-socialized, so conclusions about lack of socialization should be drawn with caution. The individual variability recorded (as shown by the boxplots in Figure 1) is an indicator that some dogs were not intimidated by the cone or were able to overcome an initial fear and engage with the object. Whether this variability was due to genetic, environmental, or experiential factors is not possible to say at this point. Therefore, it is important that future intervention studies investigate if the implementation of specific socialization protocols does indeed reduce the overall level of fear towards novel objects in CBK populations or if other factors play a major role.The dog statue elicited a different set of responses from the cone and stranger. Dogs approached the statue significantly faster than they approached the other stimuli. Given that the statue was a realistic reproduction of a life-sized conspecific, it is likely that the dogs’ first reactions were similar to those experienced during a social encounter with a conspecific, i.e., social investigation was the immediate response triggered. Subjects spent more time “sniffing” the dog statue when compared to the other two stimuli. Furthermore, dogs that explored the statue were more often observed in either a “lowered” or a “rigid” posture than those that explored the other two stimuli. Such rigid and lowered postures are typical in agonistic interactions, suggesting that dogs may have initially perceived the statue as an unfamiliar conspecific. On the other hand, a “lowered” posture could indicate either a submissive affiliative behavior, an attempt to solicit social interaction in a non-threatening manner, or maybe a state of fear. Unfortunately, the nature of our study did not enable classification of the emotional states of the dogs during such interactions. However, it should be considered that, while some dogs reacted in an affiliative manner during the first half of the test, they may have switched to fear behavior upon failure to elicit an appropriate social response from the statue or after realizing that this stimulus was actually an inanimate object, like the cone. This is in line with previous studies using fake dogs as proxies for unfamiliar dogs in dog–dog interactions. Barnard et al. [20] found that their subjects responded with the same general behaviors when presented with a fake dog reproduction as with a living dog. Given that the subjects seemed to react (at least initially) to the statue as if it was another dog, it is possible that this stimulus could reflect the response of the tested dog to another unfamiliar dog after being rehomed. The greater amount of exploration and lower initial fear toward the dog statue suggests that the dogs in this sub-sample were generally socialized to other dogs. Indeed, common management practices used by the participating breeders included allowing the dogs into outdoor exercise yards in groups and flexible group composition when socially housing them. Overall, it is interesting that the dogs seemed to respond to the dog statue more as a conspecific than as a non-social object. The use of artificial conspecifics has many advantages, allowing performance of social behavior and cognition experiments in a more controlled and repeatable manner compared to living animals [21,22,23]. The use of artificial dogs during behavioral assessments in homes and shelters has gained attention as it permits testing of dogs safely and in a standardized way, although its validity is still controversial [20,24,25,26]. Future studies should further investigate how dogs perceive social and non-social stimuli and what elements of these stimuli might elicit fear responses. This could potentially impact the design and conduct of future behavioral and cognitive tests, as well as socialization practices in breeding kennels.The stranger approach test elicited the lowest level of “walking” behavior compared to the other stimuli. In addition, when the stranger was present, dogs spent more time at the front of the pen (i.e., in relatively close proximity to her) and they spent less time moving away from the stranger compared to the cone stimulus. This suggests that the dogs were less fearful of the stranger than the cone. However, it is interesting to note that in the stranger test, the “latency to approach and interact with the stimulus” was similar to that of the cone and was significantly higher than that of the dog statue. This could indicate that the stranger elicited a greater level of fear or a greater level of conflict behavior compared to the dog statue. This finding is further supported by the reduced amount of time dogs spent sniffing the stranger compared to the dog statue. Whether the level of fear induced by the stranger was due to insufficient socialization of these dogs to unfamiliar people or whether it was due to the unfamiliarity of the test situation and the movement of the tester while performing the test is unclear. During the stranger approach tests, the tester was only able to hand-feed a treat to 50% of the dogs, which indicates that many of the subjects were not comfortable enough in the presence of a novel person to approach and take a treat. Additionally, the tester was only able to touch 25 of the 60 dogs, suggesting that the majority were not comfortable with a stranger slowly reaching out and attempting to touch them. Although these dogs were handled daily by their caretakers, it is possible that many were not sufficiently exposed to or handled by unfamiliar people often enough for them to generalize their social responses to new people. In a pilot study carried out by this group (unpublished), it was observed that some dogs reacted fearfully to strangers, even though they were highly social toward their own caretakers. For successful rehoming outcomes, it is critical that dogs are able to generalize positive perceptions of their interactions with their caretakers to new people to whom they are exposed to avoid being chronically distressed. The implementation of effective socialization protocols in CBKs may, therefore, help to achieve this goal. Currently, our research team is investigating if brief daily caretaker interaction may have an effect on the dogs’ behavior, not only toward the caretaker but also, by generalization, toward unfamiliar people. If successful, this could be an easy-to-implement protocol with great beneficial impact.The cross-tabulation analysis allowed for comparison of a subject’s behavior during the cone and dog statue reaction test and their stranger approach RYG classification. When presented with the cone, dogs categorized as “red” showed greater average durations for fear-related behaviors, such as gaze avoidance and lowered postures than dogs scored as “yellow” or “green”, as well as longer latencies to approach and interact with the stimulus. In contrast, dogs in the “green” category spent a greater average duration performing behaviors such as play, sniffing, and neutral posture than did dogs in other categories. Collectively, these findings support the idea that “green” dogs had lower levels of fear when exposed to the cone stimulus than the dogs in the “red” category. Similarly, when exposed to the dog statue, dogs in the “red” category, on average, spent more time performing gaze avoidance and maintaining a low posture, and demonstrated a higher increased latency to approach and interact with the stimulus than dogs categorized as “green” or “yellow”. When presented with the dog statue, dogs in the “green” category exhibited play behavior, which was completely absent in the dogs scored as “red” or “yellow”. This again suggests that “red” and “yellow” dogs had greater levels of fear in the presence of the dog statue than dogs in the “green” category. It should also be noted that all categories had a high average duration for “sniffing” behavior when presented with the dog statue. This suggests that most dogs were highly interested in the stimulus, regardless of their RYG classification. This observation may have practical applications: a high prevalence of “red” dogs in the kennel may indicate higher likelihood of generalized social and non-social fear in a population. If future experiments confirm this association, this very brief stranger approach test may be used as an easy initial screening tool before performing more comprehensive behavioral assessments.Although high levels of fear may be a consequence of poor or absent socialization practices in the kennels, it is important to note that a range of different factors may play a role in shaping fear responses. These include genetics, physical health, handling and breeding practices, exercise, conspecific interactions, parity, breed and more [6,27,28,29,30]. In this study, no significant differences were found between the behavioral reactions of males and females to any of the three stimuli. The effect of breed could not be discerned, because breed could not be included in the analysis. However, there is evidence from previous work that breed may have a significant effect on fear levels. More than 50 breeds have been determined to show potentially heritable fear/anxiety [29]. Additionally, Morrow et al. [30] found significant differences in the percentage of subjects demonstrating fear-related avoidance behaviors and the age of onset of these behaviors between three various breeds. Future work building upon the current study should, therefore, include breed as a potential risk factor for fearful temperament. Finally, there was a significant association indicating that dogs in larger pens spent less time in the back portion of the pen than dogs housed in progressively smaller pens. It is likely that dogs in larger spaces could be closer to the front of the pen and still maintain a comfortable distance from the object (placed at the front of the pen). In contrast, dogs in small pens had to retreat to the far end of their pens to create distance from the stimulus presented. There may also have been a confound in that small breeds typically are housed in smaller pens. Hence, future studies should account for both space allocations and breed factors to better understand the effects and impacts of larger pen spaces on the levels of fear expressed and dogs’ coping strategies when faced with a new, challenging situation. This line of research could have an impact on space recommendations for other populations of kenneled dogs such as those maintained in working, shelter and laboratory environments.A systematic data collection and risk analysis that combines animal, resource, and management metrics could help move toward a reliable quantitative assessment of fear and identify key areas for improvement for dogs in CBKs.5. ConclusionsFindings from this study revealed that the cone stimulus elicited the greatest levels of behaviors indicative of fear than either the dog statue or stranger. It is important to reiterate, however, that overall, dogs had primarily mildly fearful reactions to the cone and that behaviors suggesting extreme fear or aggression were rarely recorded. Both of the other stimuli (the stranger and the dog statue), although less apparently intimidating, still provoked a certain amount of fear in a moderate portion of the study population. These findings, overall, are encouraging, although they also highlight a continued need for improvement. What constitutes a minimum to optimal amount of socialization in dogs has yet to be determined [31,32]. However, introducing stimuli in a positive and gradual way, especially during the early stages of life, has been reported to help minimize the development of fear-related behaviors [33,34]. Hence, it is advisable that breeders have socialization protocols in place that incorporate controlled exposure to both social and non-social stimuli. Such protocols should ultimately aim to improve the dogs’ welfare and to maximize chances of success after retirement and rehoming. Indeed, further research is needed to fully understand which management practices and/or protocols may be more effective for evidence-based welfare improvements in this population of dogs. Simple behavioral tests like the one presented here can be used to explore the effects of short, targeted interventions where, for example, adult dogs are re-tested after simple socialization practices have been implemented.

animals : an open access journal from mdpi

10.3390/ani11061793

PMC8234069

Although agonistic interactions between cats are often regarded clinically as a source of stress, there is currently limited research evidence regarding the welfare impact of keeping multiple cats as pets. The aim of this study was to compare welfare indicators between cats living in single and multi-cat households, as well as between cats living in multi-cat households where agonistic behaviour was/was not reported by owners. Indicators included a spatial judgment bias task (JBT) and the cat stress score (CSS). CSSs were higher in cats from single compared with multi-cat households. CSSs were lower for cats that showed a more ‘pessimistic’ response in the JBT, suggesting these cats appeared to be less stressed. JBT results did not vary depending on the presence of, or reports of agonistic behaviours between, cohabiting cats. These data suggest that mood states (as measured by the JBT) were not impacted by the social groupings investigated, and that cats from single-cat households showed more signs of stress (as measured by CSS) than those in multi-cat households. Alternative explanations cannot be discounted, particularly due to the narrow sample population and broad scope of husbandry conditions that were unaccounted for. Further research is warranted to explore the extent to which variables that could not be controlled may have confounded findings.

Although agonistic interactions between cats are often regarded clinically as a source of stress, there is currently limited research evidence regarding the welfare impact of keeping multiple cats as pets. The aim of this study was to compare welfare indicators between cats living in domestic single and multi-cat households, as well as between multi-cat households where agonistic behaviour was/was not reported by owners. Indicators included a spatial judgment bias task (JBT), where longer latencies to ambiguous probes are interpreted as being related to a more ‘pessimistic’ mood state, and the cat stress score (CSS), where high scores are indicative of high stress levels. Of 128 focal cats between the ages of 9–22 months, 94 were from multi-cat households, 126 had useable CSS data and 42 had JBT results suitable for analysis. CSSs were significantly lower for cats showing a more ‘pessimistic’ response in the JBT. It is possible that the cats that appeared to be the most relaxed may have been showing inactivity relating to negative affective states and/or were the least active/food motivated, and therefore slower in the JBT. CSSs were significantly higher in cats from single compared with multi-cat households, and did not vary with reports of agonistic interactions in multi-cat households. JBT results did not vary depending on the presence of, or reports of agonistic behaviours between, cohabiting cats. These data suggest that cats from single-cat households may be more likely to show signs of acute stress than those in multi-cat households. Alternative explanations are possible. For example, lower CSSs in the multi-cat group may reflect ‘relief’ effects resulting from separating cats for the test period, or inactivity relating to negative affective states. Due to the narrow sample population and broad scope of husbandry conditions, the potential for confounding variables limits the degree by which results can be used to inform causation of the relationships identified. Further research is warranted to replicate this work and explore potential confounders.

1. IntroductionDomestic cats are a popular companion animal, with approximately 10 million cats estimated to be kept in the UK as pets [1]. Although cats may appear to be well adapted to fitting in alongside human lifestyles, there are some aspects of modern living which may be a welfare concern. In particular, those working in clinical behaviour commonly report cases where cats adapt poorly to living in close proximity with other cats, especially those who are unrelated and/or unfamiliar [2]. With an estimated 43% of owned cats in the UK being reported as living with at least one other cat, and approximately half of these cats being reported to endure conflicts with cohabiting cats, this is an area of cat behaviour that may have a considerable impact on the welfare of owned cats [1,3].The ancestral species of the domestic cat, the African Wild cat (Felis silvestris lybica) [4], was likely to have been largely asocial [5]. Despite this, domestic cats living in feral colonies can form cohesive social groups, typically consisting of a core group of related females which may be associated with several roaming males, where resource availability and distribution allows [6]. However, commonly occurring social groupings in domestic environments, such as the co-habitation of cats introduced as adults, and behaviours observed between co-habiting cats, are considerably different from those in naturally occurring groups [7,8]. Close contact between incompatible cats has been suggested to be an important cause of undesirable stress-related behaviours [2,9]. Behaviours that owners find problematic can have further impact on welfare through consequences, such as relinquishment to shelters [10] or euthanasia. Given the apparent variation in social compatibility between cats, evaluating the degree to which contact with conspecifics may be perceived as positive or negative to individuals is an important area of research.Measuring the welfare of domestic cats is a growing area of research that has thus far predominately been applied to controlled environments such as shelters or research facilities [11]. Methods of welfare assessment for cats in different social environments have mainly involved changes in physiological and behavioural parameters during or after a perceived challenge. Examples include activation of the hypothalamic pituitary–adrenal stress response pathway [12,13,14], as well as visual behavioural responses to stress, measured using an integrated behavioural observation measure for stress assessment, the ‘cat stress score’ (CSS) [15,16,17,18,19,20]. Prospective longitudinal data have also been used to examine the differences between multi/single-cat households [21]. The results of previous studies have yielded conflicting results which, despite highlighting a number of potentially important factors, have yet to provide clear evidence for the welfare implications of cat sociality [22].One limitation of the current welfare assessment tools developed for use in cats is the difficulty in identifying the valence (positive or negative) of the animal’s emotional state [11]. Judgment biases have been used to evaluate welfare state based on the valence of emotions, since animals experiencing putatively poor welfare, such as being housed in unpredictable environments or frequently being exposed to ‘unpleasant’ experiences, have a more ‘pessimistic’ reaction to ambiguous stimuli in judgment bias tests (JBT) [23,24,25]. This welfare measure appears to be sensitive to the effects of social stress [26]. JBTs have been used in a range of species, and we have adapted protocols in order to develop a spatial JBT suitable for application to pet cats in the home environment [27,28,29].CSSs [30] were used in this study as an additional welfare indicator and for comparison with JBT results. CSS scales were adapted with the addition of further categories [31] to increase sensitivity.The objectives were to: Investigate whether measures of relative judgment bias co-vary with CSSs. The CSS is believed to measure discrete behavioural responses to recent stress exposure, whereas judgment bias varies with underlying (and hence potentially longer term) affective states [32]. Hence, we would not expect to find a direct correlation between these measures.Assess the welfare of pet cats based on the presence of, and relationships with, other cats in the household, using the JBT and CSS. Cats in multi-cat households were hypothesized to show a longer latency to one or more ambiguous probes in a JBT (i.e., show a more ‘pessimistic’ judgment bias) and have higher CSSs compared to those living in single-cat households.Cats from multi-cat households where the focal cat was reported to show and/or be the recipient of agonistic behaviour were hypothesized to show a longer latency to one or more ambiguous probes in a JBT and higher CSS compared to those who did not exhibit/were not the recipients of agonistic behaviour.2. Materials and Methods2.1. Study PopulationThe study population comprised 241 domestic pet cats from 105 different households. Of these, 128 cats were ‘focal’ cats, registered with two existing UK longitudinal questionnaire-based studies with the University of Bristol—the ‘Bristol Cats’ project and ‘Cats Longitudinal Analysis of Welfare Study’ (C.L.A.W.S). Focal cats were between 9–22 months of age at the time of data collection. This age range was selected due to restrictions on the availability of cats in the existing cohorts. The remaining cats in the study population were all of those co-habiting with the focal cats. The sample for this project was recruited from Southern England, the Midlands and South Wales. Owners were recruited via email, or telephone contact if they did not have a registered email address/had indicated a preference. Visits were carried out between 21 October 2013 and 26 May 2014. All owners gave fully informed, written consent to take part in the study. The cats were pets, kept primarily for the purpose of owner companionship, living in the owner’s homes. The majority of focal cats had both indoor and outdoor access (81.3%), a proportion of which remained stable across the multi-cat groups (81.3%) and agonistic behaviour groups (78.7%). Other husbandry conditions varied between cats and households, although several aspects of husbandry were recorded, these variables were not factored into statistical analysis due to the restricted sample size and broad scope of variation that existed. Potentially confounding variation that existed between the selected groups should therefore be considered a limitation of the study. Due to the nature of the study population (pet cats owned by participants who were volunteering their time to take part), the only experimental control placed on the cats outside of testing was to ask owners to avoid feeding focal cats for at least three hours prior to the arranged visit times in order to attempt to control for the effects of satiety. Based on the results of piloting [29], it was anticipated that roughly half of focal cats would complete the JBT, so attempts were made to recruit twice the desired number. Focal cats were excluded from the study if they were pregnant or lactating, on any pharmacological therapy likely to affect behaviour, were currently undergoing behaviour therapy, had any current medical illness, had a history of infectious disease with potential carrier status or were known by their owners to be fearful of strangers or handling.The average age of focal cats was 14.46 months; 61.7% were male, of which two cats were not neutered. The remaining cats (38.3%) were neutered females. The proportion of males was slightly lower but stable across the multi-cat groups (47.7%) and agonistic behaviour groups (48.9%) examined. A total of 94 (73.4%) focal cats lived in multi-cat households, and the total number of cats in these households ranged from 2–8 (see Table 1 for frequencies). Information regarding the focal cat’s agonistic interactions with other cats in the household was collected via a verbal questionnaire, with the exception of one eight-cat household where the cats were kept separated and so no data regarding agonistic interactions were collected. CSS and JBT data were only collected for focal cats. Where households contained more than one cat registered to the ‘Bristol Cats’ or ‘C.L.A.W.S.’ studies, aged between 9–22 months, multiple focal cats from the same household were tested. No more than two focal cats were used from any one household, and two cats were selected at random from eligible cats using a random number generator [33]. The number and percentages of focal cats used for analysis of CSS and JBT data are shown in Table 1.2.2. Experimental ProtocolCat owners were visited in their homes for data collection. All data collection was conducted in one room of the house—ideally a quiet room free from distractions, with adequate space for the JBT. At least one owner would usually be present for the entire process, in addition to the lead researcher (researcher A). An additional researcher (researcher B) was present for the initial visit and, in some cases, subsequent visits. Researcher B was not always the same person.The chronological stages of the experimental protocol are described below. The average time from arrival to the subject’s home to the time of completion for one cat on the initial visit was 45 min, and 30 min for each subsequent visit. All equipment was wiped down with Dettol® antibacterial spray between visits. Each focal cat was visited on up to five separate occasions to complete training and testing for the JBT.2.2.1. Habituation Habituation times were based on those described by Kessler and Turner [30] and tested during the piloting stage of this project [29]. This was 10 min on the first visit and five minutes for each subsequent visit. Habituation began once the researchers, owners and cat were in the room of the house selected for data collection, and the JBT equipment (excluding the food bowl) was placed out for the cat to investigate. 2.2.2. Verbal Questionnaire (Multi-Cat Households Only) At the initial visit, owners of multiple cats read a list of social behaviours that had previously been described as agonistic/indicative of cats that do not perceive one another as part of the same social group [2,34]. They were asked to report whether they had witnessed the focal cat show, or were a recipient of, any of these behaviours over the last month, with any other household cat. The behaviours, as described to owners, and their abbreviated definitions, are as follows: Fluffing: ‘fluffing up’ at one another (when the hairs go on end making the cat look bigger).Freezing: freezing and staring at one another—for example, in corridors or doorways, for five seconds or more at a time.Vocalising: hissing/spitting/yowling/growling at one another.Blocking: blocking or inhibiting one another’s movements or being reluctant to pass one another in tight spaces (e.g., corridors and doorways).Aggression: fighting (aggression not occurring out of play), e.g., scratching/biting, chasing and attacking.2.2.3. CSSThe adapted 9-point CSS [31] was used. This scoring system added half points to the original scale [30]. In the current study, the same scale points were used but were named as complete integers; hence, a score of 2.5 became a score of 3, and score 3.5 became a score of 4 [31]. This resulted in a scale ranging from 1 = fully relaxed to 9 = terrorized. Scoring was conducted by selecting the score that best fitted with the majority of behavioural signals shown by each cat. Where behavioural signs were ambiguous between points on the scale, the lowest possible score was given.The scores for each cat were taken immediately after the habituation period at every visit, meaning that multiple scores were collected for cats undergoing JBT training and testing. Three scores were taken at each visit, with a two-minute interval between each. Video footage was taken for the duration of scoring in order to assess reliability by comparison, with scores given by an independent observer. The mean CSS across all visits for each cat was used in data analysis.2.2.4. Feline Temperament ProfilingAt the initial visit only, following assignment of CSSs, an adapted version of the Feline Temperament Profiling (FTP) scoring system [35] was used to assess the cats’ behavioural responses to the researcher and screen out those unsuitable for further testing. The protocol for FTP was adapted to enable rapid assessment of whether cats were showing an ‘unacceptable’ level of fear, which would suggest they would not respond well to the level of handling required for JBT (and therefore ensured their welfare was protected). An ‘unacceptable’ level of fear was reached if the cat showed three or more ‘potential’ indicators (e.g., dilated pupils, avoiding eye contact) or any one clearly fearful reaction (e.g., aggression or hiding). If this was the case, the assessment was stopped, and the judgment bias training was not attempted. Steps 7–10 of the FTP protocol were omitted, as they were deemed unnecessary for the purposes of this study. A copy of the adapted protocol used can be found in the supplementary information.2.2.5. JBTA spatial JBT was used, which consisted of a training phase followed by a test phase. During the training phase, cats were presented with a bowl which was either baited with a treat on one side of the experimental set up, or a bowl which was empty on the opposite side of the experimental set up. Training was completed once the cat had reached criteria to demonstrate that they were more likely to anticipate a food reward in the baited bowl location compared with the empty bowl location. During the following test phase, ‘probe’ bowls were presented at intermediate locations, and the cat’s latency to approach each bowl was recorded as an indication of judgment bias. The detailed protocol is described in full below, based on that of Tami et al. [28] and a similar protocol developed for dogs [27]. This protocol has been piloted for use in pet cats in the home environment [29], and the results were used to inform the methodology of this study. Adaptations resulting from piloting included: the use of a door that opens from the centre to reduce the risk of side biases developing during training; the use of a spatial measure in order to determine whether cats had ‘checked’ bowls instead of relying on observer perception; the use of a novel food bowl instead of the cat’s own bowl to avoid variation in bowl height/size and prior associations; and recruiting the owner’s help to restrain the cat instead of a secondary researcher, where possible, to reduce potential stress caused by handling by an unfamiliar person.Judgment bias training started at the initial visit, following FTP (excluding those cats scored as having an ‘unacceptable’ level of fear). The JBT could take up to five sessions for the cat to learn and complete, and it was always carried out after the habituation period and CSS observations.A minimum of two people were required to be present for each training session and testing. Researcher A was always the same person throughout data collection, and was responsible for placing the bowl in position and measuring latencies. The owner would usually fulfil the role of the second person, handling and releasing the cat. A second researcher (B) was present on the first day of training to help the owner. If the owner was physically unable to handle and release the cat themselves, researcher B would attend subsequent visits to fulfil this role (n = 7).The study received human and animal participant ethical approval from the local University ethics committee.Training PhaseThe training phase was made up of a maximum of 60 trials, with 15 trials per session. Cats had one training session per day, and training sessions were conducted on consecutive days. In some cases, this schedule had to be adapted to fit with owner availability. One cat had two training sessions on the same day, and 13 cats did not have training on consecutive days.The apparatus was set up as shown in Figure 1, using masking tape to mark bowl locations on the floor. A clear laminate plastic semicircle measuring 60 cm across was placed underneath the bowl every time it was presented, at all locations.The positions of the rewarded (R) and unrewarded (U) bowl locations remained constant for each cat, but were randomized between cats using an online random number generator [33].The cat was held behind the screen by the owner (or researcher B) while the bowl (which was novel to the cat) was baited/not-baited with a food reward (one beef Whiskas® Temptations biscuit; this was kept standard across all cats to prevent variation in responses due to food type).The bowl was always baited in the same way; the bowl was picked up by researcher A, taken to the baiting location (halfway between the two bowls) and either baited or not baited. Either way, the box with the food in it was opened and closed, and a treat was placed in the bowl (then removed if it was a non-baited trial). The bowl was then tapped twice with a pen to earn the cat’s attention and signal that a trial was about to begin. The bowl was placed in either the ‘R’ or ‘U’ position, on top of the plastic semicircle. The order of ‘R’ and ‘U’ trials was pseudo-random, with no more than two in the same location continuously to decrease the likelihood of side biases developing. The barrier (two solid panels that opened outwards, creating a ‘doorway’ in the centre for the cat to proceed through) was then opened by the owner or researcher B, ensuring minimal distraction and reducing the risk of side bias. The cat was released as soon as the barrier was opened; facing straight forward, the time taken for the cat to move from the starting point and place a foot onto the plastic semicircle was recorded.The cat was given up to 45 s to reach the semicircle; if the cat took longer than this, the maximum time of 45 s was recorded. Once the cat had reached the semicircle, or 45 s had elapsed, the cat was recalled/moved back behind the barrier by the owner or researcher B. If the bowl was baited, the cat ate the treat before being recalled. At the first training session, the cat was presented with three ‘R’ and three ‘U’ trials in a pseudo-random order (as described above). During these trials, researcher A gave two additional taps on the edge of the bowl with a pen, up to three times per trial, to earn the cat’s attention and aid learning. In all subsequent training sessions, ‘R’ and ‘U’ trials were presented in a pseudorandom order and no additional taps were provided. The cat was said to have learned the task when it was consistently visiting the ‘R’ location faster than the ‘U’ location. Consistently faster was defined as: for any six consecutive trials in one discrete training session, the longest latency to reach the rewarded side must be at least half a second less than the shortest latency to reach the unrewarded side. Each cat was given a maximum of 60 learning trials to reach the criterion. Once the cat had learned the task it was presented with the test phase. Test PhaseThe test phase consisted of 14 trials. This was ideally run the day after the cat reached the criterion; however, in 18 cases, this was not possible due to the need to fit in with the time schedules of owners (range = 1–9 days). Although this variability was unavoidable, it may have impacted on learning the discrimination as highlighted in the discussion. During the test phase, the cat was presented with bowls in intermediate locations (‘probes’), as shown in Figure 1. The bowl nearest the rewarded location was termed ‘R-near’ (nearer rewarded), the bowl nearest the unrewarded location was termed ‘U-near’ (nearer unrewarded) and the central bowl was ‘M’ (middle). These positions were equidistant from the ‘R’ and ‘U’ locations and were not reinforced with a food reward. The probes were presented in a pseudo-random order, interspersed with the reference ‘U’ and ‘R’ locations. For the final trial the bowl was placed in the ‘U’ location and baited to ensure the cat was learning spatial associations and not following olfactory cues (the false negative, ‘U = R’). Each probe location was only presented once in order to prevent potential confounding effects of a reduced latency to these positions through repeated presentation due to cats learning that these locations were unrewarded [27]. Baiting/non-baiting of bowls was performed in exactly the same way as the training trials. The cat’s latency to reach each probe was measured when the cat placed a paw on the plastic semi-circle surrounding the bowl. The precise bowl locations and order of occurrence for the test phase are presented in the Supplementary Materials. 2.3. Statistical AnalysisIn order to account for potential bias caused by testing multiple focal cats from the same household (n = 23), all statistical analysis presented in this section has been replicated with a reduced sample size (n = 105) to eliminate repeated testing of cohabiting cats. An online random number generator was used to select a single focal cat from each household to include in this analysis [33]. This resulted in a reduced population of 103 focal cats with useable CSS data, and 35 with useable JBT data. In this reduced population, statistical significance and the direction of the relationships identified were unaffected, demonstrating a low likelihood of bias caused by testing multiple focal cats from the same household. Due to the already limited sample size, results from the full population of 128 focal cats are presented and discussed below. Dependent variables were skewed and could not be normalised using log or square root transformations; hence, non-parametric methods were used. Data were analysed using IBM SPSS version 21 unless otherwise stated. 2.3.1. CSS Preliminary Analysis To ensure consistency in scoring CSS, the lead researcher and an independent scorer assessed sets of 19 videos repeatedly in order to ensure their methods were reliable; this was before they scored the study population. A ‘strong’ level of agreement between scores after two repetitions was achieved (weighted Kappa = 0.768). Inter-rater reliability was ‘moderate’, and was assessed by comparing 20 scores taken during visits with scores given using the video footage of these visits by the independent scorer (weighted Kappa = 0.531). Weighted Kappa results were assessed, and interpretation was determined using the ‘StatsToDo’ online resource [36]. Although the CSS is often described as a scale, differences between integers are not necessarily equivalent; hence, they were analysed as categorical variables. Categories were combined into three groups (CS1–2; CSS 3; CSS 4–6) because of the low number of cats showing very low or high CSS values.2.3.2. JBT AnalysisWilcoxon signed ranks tests (T) were used to compare latencies to each bowl location presented during the JBT. Kruskal–Wallis tests were used to compare categories of CSS with probe latencies in the JBT.2.3.3. Comparing Welfare Measures Based on Social ConditionsMann–Whitney U tests (U) were used to compare the distribution of probe latencies for cats that lived in multi-cat households with those from single-cat households, as well as for cats that lived in multi-cat households where there were agonistic inter-cat interactions and those where there were not. Chi-squared tests (χ2) were used to compare CSS groups between these household types.3. Results3.1. CSSCSS data were not collected for two focal cats, as they remained hidden during visits. For the remaining cats (n = 126), the average CSS for each cat over all visits was used in the analysis (Figure 2). The median score was three, with an interquartile range of one.3.2. JBT A total of 42 (32.81%) focal cats had JBT results that were used in analyses. A further 12 cats completed the JBT but had either travelled on average faster to the U compared with the R position during the test phase or had become uninterested/distracted in the test phase. The remaining 76 cats did not reach the JBT learning criterion; of these, 30 cats showed behavioural signs indicative of fear or frustration; 17 were not interested in the food reward; 13 did not pass FTP; 11 lost interest in the task; two showed a combination of the above reasons; and one became injured between training sessions (due to an unrelated incident).The number of trials taken to reach the learning criterion for the JBT was positively skewed. The median number of trials was 14.5, ranging from the minimum number possible (6) to 51 trials. The interquartile range was 13. Initial exploration of the latencies to ambiguous probes for those cats that completed the JBT showed similar results for latencies to the R, R-near, M and U-near bowl locations (Figure 3). Notably, there was a high degree of variation between individuals in the population, with a proportion of cats having much longer latencies to all locations.Median latencies to the U location were significantly longer than to all other locations on test day: R (T = 902.0, p = <0.001), Nr-R (T = 3.0, p = <0.001), M (T = 23.0, p = <0.001), Nr-U (T = 165.0, p = 0.003 and the ‘false negative’ U = R (T = 486.0, p = 0.005). These results confirmed that the cats who had reached the learning criterion could distinguish between the reference locations, and perceived this location as least likely to be rewarding. This also demonstrates that the cats were unlikely to respond to olfactory (or other) cues to locate the treats, and that their response was based on prior learning about the location. Using a Bonferroni corrected p value (0.008), median latencies to the R location on test day were not significantly different from latencies to the Nr-R (T = 453.0, p = 0.563) or M (T = 521.5, p = 0.029) probes. This suggests that cats did not perceive these intermediate probes as less likely to be rewarding. Median latencies to the R location were significantly shorter than latencies to the Nr-U probe (T = 776.5, p = <0.001), suggesting that cats were less likely to expect the bowl to contain a reward in the Nr-U position. When comparing intermediate probes with one another, latencies to the M probe were not significantly different from latencies to the Nr-R probe (T = 282.0, p = 0.085), suggesting that cats did not differentiate between the two in terms of reward probability. Latencies to the Nr-U probe were significantly longer compared with latencies to both the Nr-R (T = 765.0, p = < 0.001) and M (T = 698.5, p = 0.002) probes, supporting the conclusion that cats perceived this location as less likely to be rewarding. 3.3. Comparing CSS and JBT DataCategories of CSSs were significantly associated with latency to the Nr-U probe (χ2 = 6.07, df = 2.0, p = 0.048; Figure 4), and showed a trend towards association with the Nr-R probe (χ2 = 5.90, df = 2.0, p = 0.052; Figure 5). No significant association was identified between CSSs and latency to the M probe (χ2 = 0.30, df = 2.0, p = 0.861). Post hoc analyses suggested a significant difference between scores 1–2 and a score of 3 (U = 101, p = 0.012) for Nr-U probe latency, where cats with the lowest stress scores were slower to reach the Nr-U probe. For the Nr-R probe, cats with CSSs of three appeared slower to reach the bowl compared with those cats that had the highest CSSs of 4–6 (U = 7, p = 0.031). This trend was the only result that was not replicated when statistical analysis was repeated with a smaller sample size of 105 cats from separate households. This was due to cats scoring 4–6 with useable JBT data being randomly removed from the study population. 3.4. Comparing Welfare Measures between Multi-Cat and Single-Cat HouseholdsOf the 42 cats that fulfilled the inclusion criteria for analysis of JBT data, 33 (78.5%) were from multi-cat households. Probe latencies did not differ significantly between single and multi-cat households (Nr-R: U = 134.0, p = 0.673; M: U = 101.0, p = 0.152; Nr-U: U = 122.5, p = 0.432). Effect sizes were calculated to further examine differences between groups because of the small sample of cats in single-cat households. In all cases, the estimated effect size was small (<0.3).A significant difference was identified between CSSs in multi-cat households compared to single-cat households (χ2 = 9.465, p = 0.009, n =126 cats with CSS data). Post hoc analysis using a Bonferroni corrected p value (0.017) indicated that cats from multi-cat households were more likely than solitary cats to have CSSs of 1–2 (χ2 = 8.977, p = 0.003, n = 126), and solitary-cat households were more likely than multi-cats to have CSSs of 3 (χ2 = 8.977, p = 0.003, n = 126). No significant relationship was identified between household type and CSSs of 4–6. Figure 6.3.5. Comparing Welfare Measures Based on the Occurrence of Agonistic Interactions in Multi-Cat HouseholdsOne of the 94 focal cats living in multi-cat households was kept separately from the other cats in the home, and therefore excluded from analysis. A total of 70 of the 93 remaining cats displayed and/or were recipients of agonistic behaviour (fluffing, blocking, freezing, aggression and/or vocalising) from other cats in the home (75.3%). CSS data was collected for 68 of these cats. CSS categories did not vary significantly between cats who displayed/received agonistic behaviour and those who did not (χ2 = 4.33, p = 0.115, n = 92).Of the 33 cats from multi-cat households that completed the JBT, 26 (78.8%) displayed and/or were recipients of agonistic behaviour from other cats in the home. Probe latencies did not differ significantly between these cats and those that did not show/were not recipients of agonistic behaviour (Nr-R: U = 87.0, p = 0.880; M: U = 60.0, p = 0.183; Nr-U: U = 86.0, p = 0.846). Effect sizes were calculated to further examine differences between groups because of the small sample of cats in non-agonistic households. In all cases, the estimated effect size was small (<0.3).4. DiscussionThe measures used to investigate the welfare implications of cat housing conditions in this study included the CSS and the JBT. These measures were first compared with one another to investigate the extent to which they co-varied, and then compared between groups of cats in order to assess welfare in relation to different social housing conditions.The broad scope of variation that may have existed between the social groups examined in this study could have potentially confounded results (for example, the degree of relatedness between cats, the age at which they were introduced, relationships with neighbouring cats or with other household pets were not recorded or taken into account). A potentially confounding variation that existed between the selected groups should therefore be considered a limitation of the study, and one which limits the extent to which results can be used to inform causation.4.1. Usability of CSS and JBT Welfare Measures in Home EnvironmentsThe CSS is an observational measure of stress which is most likely to reflect a cat’s response to its immediate environment at the time of testing. The CSS was a quick, easy and non-invasive measure to take in the homes of pet cats. A moderate-high level of within and between-scorer reliability was achieved for CSS data. This can be considered a limitation of the study, and one which may have been improved by increasing the number of training sessions scorers completed. CSSs within the population had a relatively narrow distribution, with the vast majority of cats scoring either 2 or 3, despite the use of an ‘adapted’ scale with additional points [31]. This is likely to be because the majority of cats were relaxed in their home environment and were given time to habituate to the researchers and equipment. Further increases in the sensitivity of the CSS to pick up on differences between cats with low or intermediate scores may be a beneficial area of future research when assessing cats in their home environment. An important consideration related to the CSS is the difficulty in attributing scores to emotional valance, especially ‘mid-range’ scores. Low scores of 1–2 reflect behavioural signs of relaxed cats, and scores near the mid-range (3–5) indicate more alert cats; however, the behavioural ‘stress’ signals relate more to general activity (and are not specific to negative emotions). Scores of six and upwards indicate behavioural responses to stress that are more likely to be associated with negative emotional states. No cats in this study had scores of seven or over. The higher scores identified in this study may therefore simply reflect cats that were showing signs of being more alert/active, and not in distress. It is unlikely that cats with very low CSSs in this study were showing acute stress-related inactivity [37], as these cats were screened out as unsuitable for testing. It is, however, possible that inactivity associated with low CSSs may have been associated with a ‘depression’-like state thought to be associated with negative affective states [38].Unlike the CSS, the JBT aimed to gain a measure of longer-term affective states. Under one-third of the original sample of focal cats had JBT data suitable for analysis, a similar proportion reported in previous studies applying the JBT to cat populations [25,28,29]. The most common reasons for incompletion of the task were behavioural signs indicative of fear, or disinterest in the food reward. This is likely to have biased the sample of cats with useable JBT data to those that were less fearful and more food motivated. Developing ways to reduce the number of cats being excluded from JBT due to fearful responses will be important for future work. This may involve reducing the potentially fear-inducing elements of the JBT protocol (e.g., handling/restraint) as much as possible, and altering the habituation protocol. It should also be considered that, although food is likely to be the easiest reward to manipulate for the purposes of the JBT, the preference that cats may have for social interaction with people could have influenced results and reduced the ‘success rate’ of training [39].Conducting the JBT in the owners’ homes placed potentially confounding factors out of experimental control. It was necessary to fit testing around owners’ schedules, meaning that, in some cases, testing did not occur the day following a cat reaching training criterion. The high variability in environmental conditions and pre-existing associations cats had about the environment may have influenced their performance of the task. Household distractions are also likely to have affected results. Probe latencies in the JBT were especially sensitive to distractions, and to the fact that each intermediate was only presented once to prevent learnt reduction in responses to repeated presentations [40]. Although six cats were excluded from analysis as a result of distraction, it is possible that others also became distracted but that this was not observed during the trials. Relying on owners to restrain the cats during testing resulted in differences in handling ability, which could have affected the likelihood of cats completing testing and may have influenced their affective state. The presence of unfamiliar researchers may have had a similar affect. Food motivation varied between cats and would be influenced by husbandry systems outside the control of the study, which is likely to have affected performance in the JBT, despite attempts to control for satiety by instructing owners not to feed their cats for 3 h before visits. 4.2. Comparing CSS and JBT ResultsCats with CSSs of 1–2 (relaxed or very relaxed) had longer latencies to reach the near unrewarded probe in the JBT compared to those with CSSs of three (alert). Similarly, cats with CSSs of three showed a trend for longer latencies to the Nr-R probe compared with those cats with highest CSSs of 4–6. These were unexpected results that indicated low CSSs were associated with a more ‘pessimistic’ judgment bias. It is possible that the low stress scores identified in more ‘pessimistic’ cats may be a reflection of ‘relief’ resulting from separation from other cats or other longer-term stress causing factors that had not been recorded in this study, but which may have resulted in a negative judgment bias. The latency to each bowl is influenced by the relative motivation of cats to achieve the food reward, and cats that are not alert and/or interested in the task may be both more relaxed in appearance and slower or less interested in investigating novel locations if there is a chance of non-reward. It is possible that this relative inactivity associated with low CSSs was caused by a ‘depression’-like state associated with negative affective states [38].4.3. Comparing Groups of Cats Using JBT and CSS Welfare Measures Cats from multi-cat households were more likely to have ‘relaxed’ or ‘very relaxed’ CSSs of 1–2 compared with those from single-cat households; the latter were more likely to have an ‘alert’ score of 3. There was no difference between groups for the highest CSSs (4–6). CSSs did not vary between multi-cat households depending on the occurrence of reported inter-cat agonistic behaviours. This result was contrary to our hypothesis and indicates that cats from single-cat households were more ‘alert’ due to stress. Young cats living in groups from domestic settings have previously been identified as less stressed than their single-housed counterparts, and so this finding may be a valid representation of stress, particularly due to the young age of the focal cats [14]. Again, there are other possible explanations. For example, cats from multi-cat households were typically separated from other cats for at least five minutes prior to the recording of stress scores, and they may have been showing enhanced relaxed behaviours as positive emotional responses to this separation and the opportunity for more one-to-one time with their owners or researchers. It is also possible that the low CSSs identified for multi-cats in this study were reflective of negative affective states [38].No significant differences were identified for JBT results when comparing single-cat households with multi-cat households, or when comparing multi-cat households based on the occurrence of agonistic interactions. These results may indicate that the presence of other household cats or the occurrence of agonistic behaviours between household cats do not have a significant impact on affective state. This may be because positive effects of living with other cats counteract the negative impact of agonistic encounters. It is also possible that the JBT may not be sensitive enough to detect changes in affective states caused by chronic social stress due to a ‘masking’ effect that other, more transient, changes in the environment may have [41].It is possible that the lack of significant JBT results in respect to the social conditions of cats measured in this study can be attributed to the test used being inaccurate in recording the underlying affective state of participating cats. This explanation is deemed likely because JBT latencies recorded were similar between the R, Nr-R, and M locations. When significant differences were found, they were detected at the Nr-U probe, which was least likely to be perceived as rewarded. The lack of distinction between probes could be a result of the U position not being punishing enough, due to the fact it is a lack of reward rather than an overt risk [42]. This explanation may be further supported by previous research comparing groups of cats using a similar JBT set up, which did not identify differences between the groups tested.It is possible that the method for categorising groups of household cats into ‘agnostic’ or ‘non-agonistic’ did not successfully capture differences in social groupings that have the most significant effect on cat welfare. Agonistic behaviours were owner reported due to restraints on time, controllability of the cats’ movements and the level of disturbance caused to owners. This would have resulted in a higher risk of inaccuracies caused by interpretation or awareness. For example, owners may have struggled to differentiate between social play and agonistic aggression. Responses were reduced to a binary (occurring/not occurring) variable for analysis. Cats showing/receiving rare or transient signs of aggression would therefore be categorised in the ‘agonistic’ group alongside those with frequent and severe aggression. This method of categorisation may have masked differences between groups and resulted in limitations related to owner reports, including inaccuracies or misinterpretations. It is also possible that other social behaviours, such as the occurrence of positive interactions or avoidance behaviours, have a greater impact on the welfare of groups of cats than the presence of agonistic behaviours.Lastly, the lack of significant results identified between the agonistic and non-agonistic groups may be interpreted as evidence that the JBT and CSS are not valid indicators of welfare in the home environment, as the occurrence of agonistic behaviours between cats can be interpreted as an additional indicator of poor welfare [2,21,43].5. ConclusionsDespite limitations in methodology and demographics of the study population which limit interpretation of causation and extrapolation of results to the general population of pet cats, this study has demonstrated the potential use of a spatial JBT in the home environment. The results can be used to improve the efficacy of such testing protocols, should they be repeated in future research, including ensuring the environment used for testing is as ‘quiet’ as possible in order to prevent distraction from the task and reduce potentially confounding transient emotional states. CSSs were lower for cats that showed a more ‘pessimistic’ response to the Nr-U probe in the JBT, suggesting that ‘pessimistic’ cats appeared to be less stressed. It is possible that the cats who appeared to be the most relaxed were also the least active/food motivated, and therefore slower in the JBT than the more ‘alert’ cats. This relative inactivity may be reflective of negative affective states [38].CSSs were higher in cats from single-cat households compared to multi-cat households. These results may indicate that young cats from multi-cat households are less stressed or, alternatively, that in the context of this study, low scores of 1–2 were potentially indicative of ‘relief’ responses or inactivity that may be related to negative affective states.JBT results did not vary depending on the presence of, or reports of agonistic behaviours between, cohabiting cats. This suggests that aspects of social housing we measured may not have influenced the affective states of the cats involved. Other explanations, including questioning the validity of the JBT (and CSS) as an accurate measure of welfare in this study, have also been highlighted.Factors that may have resulted in the high level of variation in JBT results include aspects of husbandry and social interactions, such as those with neighbouring cats, that were not under experimental control. The majority of the cats in this study had unsupervised access outdoors, introducing a high degree of variability and potentially stressful experiences.Care should be taken when extrapolating the results of this study to the general population due to potential bias in the sample (cats of a specific age group belonging to owners who were already enrolled in existing cat epidemiology studies). Although prior screening of cats using FTP was important to ensure the welfare of cats included in the study, the removal of cats showing signs of fear prior to testing will have further biased the sample. The small sample size should also be considered a limitation when interpreting and attempting to generalise the results of this study. Despite this limitation, sample size calculations indicated that it was unlikely that significant relationships were missed.

animals : an open access journal from mdpi

10.3390/ani13061090

PMC10044235

With the development of the breeding industry, the production of fishmeal will not be able to meet the needs of feed production in the future. Clostridium autoethanogenum protein meal (CAP) is a by-product of people using carbon monoxide exhaust from the steel industry to produce ethanol, with a potential to become a substitute ingredient for fishmeal to alleviate the shortage of fishmeal. This study was conducted on pearl gentian grouper (Epinephelus fuscoguttatus ♀ × Epinephelus lanceolatus ♂), an economic fish species widely cultured in Southeast Asia. After using different concentrations of CAP to replace fishmeal in the feed for eight weeks, respectively, we found that it could replace 45% of fishmeal in the pearl gentian grouper feed through growth performance, various physiological and biochemical indexes analysis experiments, but a 60% replacement level would significantly affect health and growth. These findings provide a reference for the promotion and further research of CAP as a new environmentally friendly ingredient, which is positive for the sustainable development of the farming industry and even for the utilization of waste gas from the steel industry.

In this study, we present data from an eight-week growth trial with pearl gentian grouper fed either a reference diet (FM) with a fishmeal level of 50%, or test diet wherein 15% (CAP15), 30% (CAP30), 45% (CAP45), and 60% (CAP60) fishmeal was replaced by Clostridium autoethanogenum protein meal (CAP). Results showed that the weight gain and daily feed intake ratio of CAP60 were significantly lower than the FM group. In the serum, compared to the FM group, the content of malondialdehyde (MDA), the activities of alanine aminotransferase in CAP60 and CAP45 groups, and acid phosphatase in the CAP60 group were significantly higher, while the content of total cholesterol in CAP60 and CAP45 groups was significantly lower. In the liver, compared to the control group, the content of MDA in the CAP60 group was significantly higher. 3-hydroxy-3-methylglutaryl coenzyme A reductase in CAP30 to CAP60 groups and farnesoid X receptor in CAP60 were significantly upregulated. In distal intestines, the activities of trypsin and superoxide dismutase of CAP30 to CAP60 groups were significantly lower than the FM group. In conclusion, for pearl gentian grouper, CAP could replace up to 45% of the fishmeal in the feed, while a 60% replacement level will affect cholesterol bile acid metabolism and health.

1. IntroductionPearl gentian grouper (Epinephelus fuscoguttatus ♀ × Epinephelus lanceolatus ♂), with the heterosis of fast growth and strong disease resistance, is farmed in large quantities in China [1]. It is a predatory fish which requires about 45% fishmeal in the diet [2]. Fishmeal is an expensive ingredient with limited supply. There have been many studies to assess the feasibility of replacing fishmeal with other ingredients [3].Single-cell protein generally refer to the biological proteins processed by microalgae, fungi, and bacteria, which high contents of crude protein and are rich in amino acids, various vitamins, trace elements, and bioactive substances [4]. Single-cell protein can be produced from agricultural by-products, oil industry waste gas [5], or through further processing of fermentation residues from the brewing and monosodium glutamate manufacturing industries [6]. The full utilization of single-cell protein can lead to resource conservation, and in recent years, it has been widely studied as a promising alternative to fishmeal substitute in aquafeed [7].Clostridium autoethanogenum is a strictly anaerobic, gram-positive, rod-like, motile bacterium first reported in 1994 by Abrini et al. [8]. It is capable of producing ethanol using CO as a carbon source and is therefore used to treat exhaust gases from the steel industry and to produce ethanol [9]. The mash produced during this fermentation process is extracted and then processed to produce a protein ingredient, called Clostridium autoethanogenum protein (CAP). CAP crude protein content is higher than 80%, amino acid composition is similar to fishmeal, and no anti-nutritional factors have been detected [5]. Moreover, the results of genomic sequences showed that Clostridium autoethanogenum had no virulence genes [10], so CAP is considered a viable alternative to fishmeal. It is worth noting that CAP lacks crude fat compared to fishmeal [11] and also lacks phosphorus when used in shrimp feed [12], which needs to be compensated by additional supplementation. In previous studies, the proportion of CAP substituted fishmeal can reach 63% [13], 50% [14,15], or 42.8% [16] in Micropterus salmoides, 58.0% in Acanthopagrus schlegelii [5], 30% in Larimichthys crocea [17], and 30% in Litopenaeus vannamei [18], while higher levels of substitution would significantly reduce the growth and had varying impacts on the liver or intestinal health of aquatic animals. In one of the studies on M. salmoides, it was shown that 75% CAP substitution level causes swelling of hepatocytes [13], while according to another study, more than 57.14% CAP substitution levels caused morphological atrophy of the intestine in M. salmoides. CAP substitution levels above 60% also cause upregulation of tumor necrosis factor α (tnf-α) and downregulation of interleukin 10 (il-10) in the intestine of L. crocea [17].Up to now, the application of CAP as a kind of ingredient for grouper has not been reported. This study aimed to investigate the feasibility and suitable substitution amount of fishmeal in the feed of pearl gentian grouper by replacing it with CAP. In addition to growth performance, physiological and biochemical indices, gene expression, and morphology of the tissues of grouper were monitored and evaluated comprehensively in this study.2. Materials and Methods2.1. Diets and Experimental DesignThe control group feed was made with gluten (wheat, crude protein: 73.75%; crude fat: 0.11%), soybean meal (crude protein: 68.21%; crude fat: 9.00%), brown fishmeal (crude protein: 68.21%, crude fat: 9.00%) as protein ingredients and wheat flour (crude protein: 9.19%, crude fat: 0.38%) as the main binder, which was named FM. Four experimental diets were formulated to be isonitrogenous and isolipidic with 15%, 30%, 45%, and 60% fishmeal replaced by CAP (crude protein: 84.14%, crude fat: 0.19%, produced by Shoulang Biotechnology Co., Ltd. Beijing, China), named CAP15, CAP30, CAP45 and CAP60, respectively. Methionine and arginine were added in the experimental diets to the same level as the control group. The nutritional composition of CAP and fishmeal is shown in supplementary material Table S1.Except for choline chloride, all ingredients were thoroughly mixed after squeezing through a 60-mesh sieve and then mechanically mixed with choline chloride and water. The 2.5 mm diameter pellets were extruded using a twin-screw extruder (F-26, South China University of Technology, Guangzhou, China) then air-dried for 48 h, and finally, stored at −20 °C before use. The formulation and proximate composition of the experimental diets are shown in Table 1.2.2. Experimental Animal and Feeding ManagementThe experiment was carried out in the Zhanjiang Hi-Tech Park of Ocean (Zhanjiang, China). The juvenile pearl gentian grouper for the experiment were obtained from a local breeding factory (Zhanjiang, China). A total of 500 fish (18.01 ± 0.82 g) were randomly distributed into 20 tanks (0.3 m3) at a density of 25 fish per tank, and each diet was assigned to four tanks. Fish were fed to satiation at 8:00 and 17:00 each day. Continuous aeration was provided to each tank. The water was changed once a day, the temperature was 26 to 30 °C, and the salinity was 24 to 32%. Dead fish were weighed and the mortalities were recorded. The feeding trial lasted for 8 weeks.2.3. Sample CollectionFish from each tank were starved for 24 h, then counted and weighed. Three fish per tank were randomly selected to record body weight, visceral weight, and hepatic weight to calculate the VSI and HSI. Six fish per tank were randomly selected for blood sampling from the caudal vein. The blood samples were kept in the ice box. The front positions of liver and part of the distal intestines taken from two fish per tank were cut off and stored in 4% formaldehyde solution for hematoxylin-eosin (H&E). For FM, CAP30, and CAP60 groups, the other part of the distal intestines was cut to 1 mm in size, and stored in 2.5% glutaraldehyde solution for transmission electron microscopy (TEM) samples preparation. The liver, dorsal muscle, and distal intestine samples from four fish per tank were collected, immediately frozen in liquid nitrogen, and then stored at −80 °C for analysis of the biochemical indices, enzyme activities, and gene expression. Three fish were taken from each tank and used for proximate composition assay.2.4. Proximate Composition AssayProximate composition of experimental diets and whole fish samples were measured. Moisture was analyzed by drying the samples to constant weight at 105 °C. Crude lipid was measured by petroleum ether using the Soxhlet method. Crude protein was determined using the Dumas Nitrogen method by Primacs 100 analyzer (Skalar, Dutch) and estimated by multiplying nitrogen by 6.25.2.5. Biochemical Indices and Enzyme Activities AssayAfter 12 h of sampling, the blood samples were centrifuged (3500 rpm, 15 min, 4 °C), then serum was obtained. About 200 mg of tissue was weighed and then homogenized in 9:1 (saline solution: tissue) saline solution with ice bath. After centrifugation (2500 rpm, 10 min, 4 °C), supernatants were collected.For serum samples, the contents of total protein (TP), triglyceride (TG), uric acid (UA), total cholesterol (T-CHO), malondialdehyde (MDA), glutathione (GSH), the activities of acid phosphatase (ACP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), as well as total antioxidant capacity (T-AOC) were measured. For the treated liver samples, the contents of MDA, GSH, the activities of ACP, superoxide dismutase (SOD) and lysozyme, as well as total antioxidant capacity (T-AOC) were measured. For the treated distal intestine samples, the activities of trypsin, SOD, and lysozyme were measured.Among the above indices, except for the trypsin, SOD and lysozyme were determined by enzyme-linked immunosorbent assay (ELISA) detection kits (Shanghai Enzyme-linked Biotechnology, Shanghai, China), following the instructions. All the other indices were determined by detection kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China), following the instructions.2.6. Tissue Morphological HistologyAccording to the method of Chen et al. [19], the samples for H&E stain were fixed with paraffin and sectioned, dewaxed, and stained with Hematoxylin solution, then were dehydrated and sealed with neutral gum. The histology of samples was photographed using an optical microscope (Nikon Ni-U, Japan). For the hepatic sections, nuclear excursion, inflammatory, and ballooning degenerated hepatocytes were assessed. For the distal intestines sections, ten mucosal folds were randomly selected, the height and width were measured, and the number of goblet cells within 200 µm from the top was counted. Ten measurement points were selected at equal intervals to measure the thickness of the muscularis.The electron microscopy samples were prepared and observed with reference to the method of Huang et al. [20]. Briefly, tissue samples were fixed in 2% osmium tetroxide phosphate buffer, then dehydrated in ethanol, an embedded in epoxy resin before being cut into ultrathin sections by an ultramicrotome (Leica EM UC7, Japan). Ultrathin sections were dyed by uranyl acetate and lead citrate, and photographed using a transmission electron microscope (HITACHI HT7600, Japan) at an accelerating voltage of 80 kV. Thirty intestinal microvillus were randomly selected in each sample to measure the length, and the number of microvillus within 10 μm was counted to calculate the density.2.7. Real-Time Quantitative PCR AssayTotal RNA was extracted from tissue using an RNA extraction kit (Transgen Biotech, Beijing, China), and the concentration was detected with a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA) and adjusted to 600 ng μL−1 by adding ultrapure water. Evo M-MLV RT kit (Accurate Biology, Changsha, China) was used to synthesize cDNA. The methods and reactions followed the instructions.The amplification of real-time quantitative PCR was performed in a total volume of 10 μL, comprising 0.1 μL (5μM) of each primer, 1 μL of cDNA, 3.8 μL of sterilized double-distilled water, and 5 μL of 2X SYBR® Green Pro Taq HS PremixII (Accurate Biotechnology, China), using a quantitative thermal cycler (Light Cycler 480, Roche Diagnostics, Switzerland). The target gene expression was calculated by the 2−∆∆Ct method [21] using β-actin as the reference gene [22] and the FM group as the reference group. The information of primers was presented in Table 2. Some of these primers were derived from the studies of Xu et al. [23] and Yin et al. [24], which were designed according to the full-length sequences from the sequenced transcriptome of the pearl gentian grouper (not published).For the aspect of cholesterol bile acid metabolism, four genes, 3-hydroxy-3-methylglutaryl coenzyme A reductase (hmgcr), farnesoid X (fxr) and cholesterol 7α-hydroxylase (cyp7a1), cholesterol 27α-hydroxylase (cyp27a1) were selected. For the IGF-1/PI3K/AKT/mTOR signaling pathway, six genes, insulin-like growth factor (igf-1), serine/threonine-protein kinase (akt), mammalian target of rapamycin (mtor), eukaryotic translation initiation factor 4E-binding protein 1 (4e-bp1), and ribosomal protein S6 kinase 1 (s6k1) were selected. For the Nrf2-Keap1 signaling pathway, nuclear erythroid 2-related factor 2 (nrf2) and kelch-like ECH associating protein 1 (keap1) were selected.2.8. Calculations and statistical analysisThe calculation formulas were as follows:Weight gain (WG,%) = 100 × (Wt − Wi)/Wi;Specific growth rate (SGR,%) = 100 × (lnWt − lnWi)/t;Survival rate (SR,%) = 100 × Nt/Ni;Feed conversion ratio (FCR) = Wfi/(Wt − Wi)Daily Feed intake ratio (DFI,%/day) = 100 × Wfi/((Wg − Wi)/2 × t)Condition factor (CF, kg/cm3 %) = 100 × Body weight (g)/Body length (cm3)Hepatosomatic index (HSI,%) = 100 × Hepatosomatic weight (g)/Body weight (g)Viscerasomatic index (VSI,%) = 100 × Viscerasomatic weight (g)/Body weight (g)where Wt is final body weight (g); Wi, initial body weight (g); Wfi, Total feed intake (g); Nt, the final fish number; Ni, the initial fish number; t, the duration of experiment days.All data were subjected to statistical verification using one-way analysis of variance (ANOVA) with SPSS 25.0 (International Business Machines Corp, Armonk, NY, USA). When there were significant differences between groups, Tukey’s multiple comparison test was performed with differences significant at p < 0.05. Further, a follow-up trend analysis was performed using orthogonal polynomial comparison, to determine whether the effects were linear or quadratic [17].3. Results3.1. Growth PerformanceGrowth performance is presented in Table 3. CAP substitution of 15–45% fishmeal in the diet had no significant effects on FBW, WG, and SGR of grouper, but for fish fed a CAP60 diet, all these indicators were significantly lower than fish fed the control and CAP15 diets (p < 0.05). With the increase of dietary CAP levels, the FBW, WG, SGR, and DFI showed both significantly negative linear and quadratic patterns (p < 0.05). In addition, the CAP30 group had the highest DFI, which was significantly higher than the CAP60 group (p < 0.05). There were no significant differences in the FCR and SR between each group.3.2. Morphometric Parameters and Whole Fish CompositionAs shown in Table 4, in morphometric parameters, with the increase of dietary CAP levels, HSI and VSI showed significantly positive linear and quadratic patterns (p < 0.05). HSI of fish fed CAP60 diet was significantly higher than other groups; VSI of fish fed CAP60 diet was significantly higher than the control group (p < 0.05). For body compositions, moisture, crude protein, and crude lipid of whole fish did not show any significant difference among each group (p > 0.05).3.3. Serum Biochemical Indices and Enzyme ActivitiesThe results of serum biochemical indices and enzyme activities are presented in Table 5. With the increase of dietary CAP levels, the contents of T-CHO and GSH showed both significantly negative linear and quadratic patterns (p < 0.05). Among them, the content of T-CHO of fish fed CAP15, CAP45, CAP60 diets were significantly lower than the in the control group, and in addition, the content of T-CHO of fish fed CAP45, CAP60 diets also had significantly lower than in the CAP30 group (p < 0.05). The content of GSH of fish fed CAP60 diet were significantly lower than the CAP15 group (p < 0.05). The content of MDA and the activity of ALT of fish fed CAP30, CAP45, CAP60 diets, the activity of ACP of fish fed CAP60 diet were significantly higher than the control group, also the content of MDA of fish fed CAP45 and CAP60 diets were significantly higher than the CAP15 group, showed both significantly positive linear and quadratic pattern (p < 0.05). As the level of dietary CAP increased, the activity of AST showed a significantly positive quadratic pattern, with the fish fed CAP45 were significantly higher than those fed the CAP60 diet (p < 0.05). There were no significant differences in the contents of TP, TG, UA, and T-AOC among each group (p > 0.05).3.4. Biochemical Indices and Enzyme Activities in the Liver and Distal IntestinesTable 6 shows the biochemical indices and enzyme activities in the liver and distal intestines. In liver, the content of MDA of fish fed a CAP60 diet and the activity of ACP of fish fed a CAP45 diet were significantly higher than the FM and CAP15 groups, showed both significantly positive linear and quadratic patterns (p < 0.05). The content of GSH of fish fed CAP60 diet was significantly lower than the FM and CAP15 groups; theT-AOC of fish fed CAP60 diet was significantly lower than the control group, and showed both significantly negative linear and quadratic patterns (p < 0.05). In addition, the activity of SOD and lysozyme was not significantly different between each group (p > 0.05).In distal intestines, with the increase of dietary CAP levels, the activities of trypsin and SOD of fish were significantly lower than the control group when 30% or more fishmeal was replaced in the diet, and showed both significantly negative linear and quadratic patterns (p < 0.05). However, there were no significant differences in the activity of lysozyme among each group (p > 0.05).3.5. Hepatic and Distal Intestinal Morphological ObservationFigure 1 shows the images of HE-stained sections of distal intestines, and Table 7 shows the morphological histology indices. With the increase of dietary CAP, the height of mucosal fold, the thickness of muscularis, and the relative number of goblet cells showed both significantly negative linear and quadratic patterns (p < 0.05). Among them, the height of mucosal fold and the relative number of goblet cells of fish fed CAP60 diet were significantly lower than the FM and CAP15 groups, the thickness of muscularis of fish fed CAP60 diet were significantly lower than the fish fed a CAP15 diet (p < 0.05).Figure 2 shows the TEM of distal intestines, and Table 8 shows the morphological histology indices. The microvillus height of fish fed the CAP60 diet was significantly lower than those in the control group, and the microvillus density of fish fed the CAP60 diet was significantly lower than the FM and CAP30 groups, showing a significant positive linear and quadratic trend with increasing dietary CAP levels (p < 0.05), and appearing disorganized and sparse.HE-stained sections of the liver under 400 times magnification are shown in Figure 3. Hepatocytes were polyhedral in shape, the nucleus was positioned in the middle, and no obvious inflammatory cell infiltration was observed in fish fed the control and CAP15 diets. Comparatively, more hepatocytes with nuclear excursion were observed in last three groups, and more ground-glass hepatocytes were observed in fish fed the CAP45 diet, and in the CAP60 group, hepatocytes with ballooning degeneration were observed.3.6. Real-Time Quantitative PCR AssayThe relative levels of gene expression in the liver are presented in Figure 4. hmgcr, fxr showed both significantly positive linear and quadratic patterns with the increase of dietary CAP levels, while cyp7a1, mtor, and nrf2 showed both a significantly negative linear and quadratic pattern (p < 0.05). Compared to the control group, hmgcr in CAP30, CAP45, CAP60 groups and fxr in the CAP60 group were significantly upregulated, while cyp7a1 in CAP45and CAP60 groups were significantly downregulated. mtor in CAP45, CAP60 groups and nrf2 in the CAP60 group were significantly downregulated, compared to the FM and CAP15 groups. As the dietary CAP levels increased, cyp27a1 and keap1 showed a significantly negative linear pattern (p < 0.05). There were no significant differences in the relative expression levels of s6k1 among each group (p > 0.05).Figure 5 showed the results of gene expression in the distal intestines. With the increase of dietary CAP levels, the relative expression level of epinecidin showed both significantly positive linear and quadratic patterns, while mtor showed both significantly negative linear and quadratic patterns, while s6k1 showed only significantly negative quadratic patterns (p < 0.05). The relative expression of epinecidin in the CAP45 group was significantly upregulated relative to the control group, and further, it also exhibited significant upregulation in the CAP60 group relative to the CAP45 group (p < 0.05). Other than that, there were no significant differences in the relative expression levels of akt, toll-like receptor (tlr22), tnf-α, il-10, and interleukin 1β (il-1β) among each group (p > 0.05).As shown in Figure 6, with the levels of dietary CAP increased, a negative linear and a quadratic trend was found including the mRNA expression level of mtor and myosin heavy chain (myhc), while s6k1 showed only a negative linear trend, and 4ebp1 showed a positive linearly and quadratic trend in the dorsal muscle, relatively (p < 0.05). Furthermore, compared to the control group and fish fed CAP15 diet the mRNA expression level of 4ebp1 in CAP60 was significantly upregulated, while myhc was significantly downregulated (p < 0.05).4. DiscussionThe results of this study indicate that CAP is a feasible alternative to fishmeal in the feed of pearl gentian grouper. Even a 45% replacement level does not significantly affect the growth of grouper. Similar results have been shown in other studies of marine fish. In a study on A. schlegelii, Chen et al. [5] found that CAP could replace 58% of fishmeal without significantly affecting growth indices, while in a study on L. crocea, Wu et al. [17] found that CAP could replace 30% of fishmeal in the feed. The studies on A. schlegelii have found that replacing fishmeal in feed with high levels of CAP reduced the palatability of the feed [5]. In this trial, the mean value of DFI in the CAP30 group was even higher than that in the control group, which may indicate that feeds with appropriate CAP substitution levels have better feeding attraction. However, when 45% or more fishmeal was replaced in the diet, the DFI of grouper was reduced compared with the control group, but the difference was not significant, indicating that the palatability of CAP is still relatively suitable for grouper. Therefore, the significant reduction of growth performance in the CAP60 group may be influenced by other factors.ACP is a hydrolase that destroys the structure of pathogens by hydrolyzing phosphate esters to enhance immunity [25]. In this experiment, the activity of ACP in serum and liver increased with the increasing dietary CAP levels, which was contrary to the findings of Wu et al. on L. crocea [17], probably due to the difference of species. Another study indicated that ACP activity within the serum and liver of hybrid grouper was increased by negative stress [26], suggesting that the effects of excessive levels in the diet of CAP on grouper may negatively affect the immunity of grouper.The serum activities of AST and ALT are low in normal conditions, but when liver damage occurs, these two transaminases escape from the hepatocytes, causing a high level of their activities in the serum [27]. GSH is an important antioxidant substance in cells and is essential for maintaining the stability of the intracellular environment [28]. Therefore, when lipid peroxidation increases, the GSH content decreases, in addition to an increase in MDA content [29]. In this experiment, the change of activity of two transaminases, the contents of MDA, GSH in the serum and the changes of MDA, GSH, T-AOC in the liver illustrated the oxidative damage to the organism by excessive substitution of fishmeal by CAP. In a study on M. salmoides, Yang et al. [16] also found that fishmeal in CAP replacement feed significantly increased serum MDA levels, similar to this experiment. However, in the study of A. schlegelii, 58% CAP substitution levels in diet did not significantly affect the antioxidant capacity of the liver [5]. It is possible that these results are caused by differences in the experimental species.The Nrf2-Keap1 signaling pathway is an important endogenous signaling pathway affecting the resistance to oxidative stress [30]. Among them, nrf2 is a transcription factor that can upregulate the transcription of antioxidant-related genes to maintain cellular defense [31]. When the intracellular oxidative damage situation normalizes, keap1 expression is upregulated and Keap1 enters the nucleus to bind with Nrf2 and exit the nucleus, returning to its steady-state [32]. In studies on Sillago sihama [33] and Pseudosciaena crocea [34], keap1 expression level in the liver was positively correlated with nrf2. Similar results were presented in this study, and the changes in the expression of these two genes also suggested that the level of CAP substitution to fishmeal in the feed affected the antioxidant capacity of the grouper liver.In addition to changes in biochemical indicators, the morphological structure of liver cells can be significantly influenced by diet [35]. In histopathology, the position of nuclei can reflect the health of the hepatocytes [36]. The ground-glass microstructure of the hepatocytes indicates hypertrophy of the smooth endoplasmic reticulum, a sign of inflammation [37], while ballooning degeneration of the hepatocytes is often considered a form of apoptosis [38]. This result revealed that substitution of fishmeal in feed with high level CAP induced increased hepatocytes inflammation.Cholesterol is an important component of cell membranes and a precursor of bile acids and steroid hormone. Since fish can synthesize cholesterol, it is usually considered a non-essential nutrient [39]. Liver is the principal site of cholesterol synthesis, as well as the place of producing bile acids by cholesterol, where HMGCR and CYP7A1, CP27A1 are the key enzymes involved in cholesterol synthesis and bile acid production, respectively [40]. The farnesol X receptor (FXR) is the bile acid receptor that inhibits the expression of cyp7a1 to negatively regulate the production of bile acids [23]. Although we designed the feed formulation with fish oil to equalize the crude fat in each group, the results showed that the CAP high level replacement group showed differences in cholesterol metabolism compared to the FM group. According to the study by Wu et al. [41], the addition of cholesterol in the diet of giant grouper (Epinephelus lanceolatus) resulted in higher blood cholesterol content than the control group, but the expression of hmgcr was downregulated as a result. Similarly, in this study with increasing dietary CAP replacement level, the expression level of hmgcr was significantly upregulated, while the serum T-CHO content in CAP45 and CAP60 groups was significantly lower than the FM group. Furthermore, the change in the relative expression of fxr, cyp7a1, and cyp27a1 indicates that bile acid synthesis is inhibited when high level of fishmeal was replaced by CAP, which may be due to the deficiency of cholesterol. However, more in-depth verification is needed to further investigated.The mTOR is the signaling pathway involved in regulating the metabolism of nutrients such as proteins, lipids, and nucleic acids in the organism to preserve normal cell growth and proliferation [24]. In the IGF-1/PI3K/AKT/mTOR signaling pathway, IGF-1 activates the PI3K/AKT pathway, which in turn activates the formation of mTOR, phosphorylates S6K1, and inhibits 4EBP1 activity [42]. The phosphorylation of S6K1 promotes the initiation of translation, while 4E-BP1 is a negative regulator of the translation process [43]. igf-1 is significantly highly expressed in fish liver [44]. Although the expression of igf-1 and s6k1 in liver was not significantly affected in this experiment, the difference in mtor still indicated that the substitution of CAP for fishmeal would still affect the IGF-1/PI3K/AKT/mTOR signaling pathway to some extent. It has been reported that replacing fishmeal with blended alternatives leads to down-regulation of the expression of mtor, s6k1 in grouper [45], which is similar to the present study. In addition, the expression of these two genes in grouper is also affected by certain amino acid deficiencies [46]. Subject to IGF-1 mediated effects, changes in the expression level of myhc result in changes in the content of myosin [44]. In a previous study on pearl gentian grouper, Yang et al. [47] used its expression as an index to evaluate muscle protein synthesis capacity. In this study, the expression level of mtor, 4ebp1, s6k1, and myhc in dorsal muscles all showed to varying degrees that CAP has a negative effect on muscle protein synthesis at higher levels of pair substitution. After all, CAP has not yet been found to contain active small peptides or unknown growth factors, which are nutrients that are abundant in fishmeal [2].The height of the intestinal mucosal folds, as well as the length and density of the microvilli determine the absorption surface area of the intestine together [20,48]. In this study, microvilli in the CAP60 group were disorganized, sparse, and short; intestinal folds and muscle layer thickness were also decreased with the increasing CAP levels, similar to the results of studies in M. salmoides [16], reflects the damage of the intestinal. In addition, this result of decreased trypsin activity further suggests that the absorption capacity of the intestine would be affected. Based on the results of akt, mtor, and s6k1 gene expression assays, it can be further inferred that high levels of CAP substitution have a negative impact on intestinal development. Deficiencies in intestinal digestive and absorptive capacity were important in affecting growth performance in this experiment.The mucus layer of the intestine has a barrier role in protecting the intestine, and its main component, mucoglycoprotein, is secreted by the goblet cells [49]. Goblet cells are highly differentiated columnar epithelial cells [50] that appear vacuolated after HE staining treatment. In this study, the high number of goblet cells in the CAP15 group versus the control group suggests that there is some improvement in intestinal barrier function by CAP, but further, the high levels of substitution can diminish intestinal barrier function. It has also been reported that exposure to oxidative damage leads to a reduction of intestinal goblet cells in pearl gentian grouper [51]. In this study, the activities of SOD were decreased in the high-level CAP substitution group. All these results suggest that a diet with high levels of CAP substitution for fishmeal also causes some degree of oxidative damage to the distal intestines of pearl gentian grouper. Epinecidin has the ability to kill microorganisms directly or inhibit their growth [52]. In previous studies on grouper, it has been found that the expression level of epinecidin-related genes in the distal intestine can be affected by the vitamin and probiotic preparations added to the feed [53,54]. Combined with the result that the relative expression of epinecidin mRNA in distal intestine was upregulated with the increase in feed CAP substitution in this experiment, it is hypothesized that CAP, as a single-cell protein, contains unknown factors that stimulate the distal intestine leading to an increase in epinecidin expression, the mechanism of which needs to be further investigated.5. ConclusionsIn this study, CAP replacing 30% of the fishmeal in the diet had no significant negative effect on pearl grouper; when the replacement level was up to 45%, it caused some liver and intestinal health problems but did not significantly affect growth. However, higher levels of substitution (60%) caused a decrease in some non-specific immune indicators in the blood, also caused oxidative damage to the liver and distal intestine, which has a significant negative impact on health, and affected protein synthesis, ultimately reducing the growth of grouper. It was also worth noting that high levels of CAP significantly affected cholesterol and bile acid metabolism in the liver. In conclusion, CAP is a new environmentally friendly protein source suitable for use in pearl gentian grouper feeds. The direction of future research can be new feed formulations by combining CAP with other protein ingredients or adding certain attractants to improve the acceptance of CAP by grouper, in order to increase the application prospects.

animals : an open access journal from mdpi

10.3390/ani12010043

PMC8749634

Boar semen can contain many bacterial species, some of which can have a negative impact upon the quality of the semen, as well as on the sows’ reproductive capacity. Semen contamination may occur at time of collection or during semen processing. The aim of this study was to identify gram-negative bacteria that appear in boar semen and to establish models of antimicrobial resistance of isolated gram-negative bacteria. Semen doses examined contained bacterial species with a known negative effect on sows’ reproductive tracts (Pseudomonas, Enterobacter, Klebsiella, E. coli), and more than half of these isolates were resistance to gentamycin (56.52%) and penicillin (58.69%) antimicrobials commonly used in boar semen extenders. This work proved the presence of pathogenic multiple resistant bacteria in semen, and therefore, we highly recommend periodic microbiological screening of bacteriospermia in boars to avoid the use of low-quality semen in the pig industry.

Bacterial contamination of boar semen occurs with some frequency in artificial insemination centers and may have a negative effect on the quality of the semen as well as on the sows’ reproductive capacity. Normally, the source of bacterial contamination in pig seminal doses is the own boar. However, distilled water or laboratory equipment used to elaborate the seminal doses can be an important source of bacterial contamination. This study focused on the identification of gram-negative bacteria in boar semen, and impact on the quality of ejaculates obtained from boar, as well as on the establishment of antimicrobial resistance patterns of isolated gram-negative bacteria. Semen samples were collected from 96 boars, ranging in age from 12–36 month, from three artificial insemination centers from the North-West of Romania. Bacterial species were identified by two methods: matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry and API 20 E (BioMerieux, France). The main bacteria isolated from the doses diluted semen were gram-negative bacteria (47.91%), with a majority of the contaminant bacteria belonging to the family Enterobacteriaceae: Seratia marcescens 19.56%, Proteus mirabilis 15.21% and Escherichia coli 10.86% and to the family Pseudomonaceae: Ralstonia picketii 17.39%, Burkholderia cepacia 10.86%, Pseudomonas aeruginosa 8.69%, and Pseudomonas fluorescens 4.34%, respectively. More than half of these isolates (56.52%) were resistant to gentamycin and 58.69% were resistant to penicillin. These antibiotics are very frequently added in sperm diluent in the centers for the processing of sperm from boars in Romania. Regular monitoring for bacterial contamination is an important aspect of a control program.

1. IntroductionIn the pig breeding industry, diluting and preserving the longevity of semen greatly improves and streamlines artificial insemination techniques. Through the techniques of dilution and preservation of sperm life, other microorganisms such as bacteria, fungi can adversely affect the quality of diluted semen. Bacterial contamination of semen or bacteriospermia is a fairly widespread problem in semen collection and processing centers [1,2]. For this reason, antimicrobial substances are introduced in the diluted doses of semen, substances that prevent the development and multiplication of these bacteria. These antimicrobial substances are chosen on the basis of efficacy on the main bacteria isolated and identified in the doses of diluted semen. These bacteria are usually gram-negative bacteria.The main sources of bacterial contamination that influence the initial number of bacteria in the semen are boar feces, prepuce, preputial fluids, prespermatic fraction, hair, boar skin, boar’s respiratory secretions, people (laboratory workers, visitors), drinking water, distilled water used in the preparation of diluted doses of semen, feed, shelter, air, ventilation system, testicular, urethral, bladder infections, laboratory materials (peristaltic pump, boxes, BPS station, etc.) [3,4,5]. Overgrowth by contaminant bacteria of certain genera has a deleterious effect on semen quality and longevity [5]. Agglutination of sperm occurs, decreases sperm motility and viability and as a result sows will present regularly reproductive disorders translated by estrus returns, postinsemination vulvar discharges, abortions, mummifications, and the low reproductive performance of the herds were reported [4,5].Strict and rigorous attention to hygiene during semen collection and processing may reduce bacterial contamination [5]. The normal flora of the skin, hair and respiratory tract of boar cannot be reduced. However, personnel can minimize the bacterial load by monitoring as strictly as possible the rules of hygiene and sterility of the collection and processing equipment of the semen. It is particularly important to know the bacterial microbiota in boar semen and the profile of their antimicrobial resistance [6,7,8].The objectives of present study focused on identification and microbiological examination of the boar sperm doses and analysis of the quality of these samples, using a rapid and precise working technique, the matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry and API 20 E (BioMerieux, Marcy l᾽Etoile, France) for strains identification and further the establishment of the main antimicrobial resistance pathotypes of isolates gram-negative bacteria.2. Materials and MethodsA 1-year (April 2020–April 2021) retrospective study was performed by sending 96 extended swine semen for routine quality control bacteriological screening at the Banat University of Agricultural Sciences and Veterinary Medicine “King Michael I of Romania”, Timisoara, Romania, Department of Infectious Diseases.2.1. Origin of SamplesSemen samples were collected from 96 boars, ranging in age from 12 to 36 months, from 3 artificial insemination centers located in northwestern Romania. The boars included in this study belonged to different breeds (Large White, Landrace, Duroc, Pietrain, and PIC).Semen doses from these boar centers were distributed in 46 pig farms.Eighty boars from the first center belonged to Large White, Landrace breed, and the hybrid PIC. Form this center there were collected 44 raw semen samples. Sixty-four boars from the second center belonged to Large White, Duroc, and Pietrain breed, and there were collected 35 raw semen samples. Thirty boars from the third center belonged to Duroc and Large White breed, and there were collected 17 raw semen samples. The number of samples was chosen so as to be representative for each boar center. All centers used the same extender (gentamicin sulfate, 12.5 g) which contain gentamicin as antibiotic.All the boars were routinely used for artificial insemination (AI), received a commercial feed pellet and were housed in individual boxes equipped with nipple drinkers, according to the European Commission Directive for Pigs Welfare.2.2. Semen EvaluationAll boars were clinically healthy, but on farms there were registered reproductive failures manifested by genital and urinary tract infections, abortions, etc. Collection of ejaculate was performed by manual pressure, after the sanitization of the prepuce zona. The prespermatic fraction of the raw ejaculations was discarded to maintain only the sperm-rich fractions.After collecting the semen samples, the volume, concentration, color, and motility were evaluated at their own board stud where the samples came from. Color was assessed visually by observing the semen in transparent microtubes, examining the degree of turbidity, presence of the blood or an unusual color.The sperm volume was assessed immediately after the ejaculate sampling, by direct observation in the graduate sampling container.The sperm concentration was also determined immediately after the ejaculate was collected, using the Sperm Sue spectrophotometer.The semen analysis (concentration and mobility) was performed using the CASA IVOS version 12 system produced by Hamilton–Thorne Bioscience, using Animal Motility Software, Viadent option. For the sample analysis, Leja blades of 30 µL, with 4 chambers (Cryo BioSystem, France) were used, in which 10 µL semen was placed, with the help of an automatic pipette and subsequently the samples were analyzed. The system scanned automatically 10 different microscopic fields through the chamber.After raw semen samples examination, all 96 samples were diluted 1:9 with a standard commercially extender which contains gentamycin sulfate and glucose. An aliquot of 200 µL of each diluted semen sample was added to sterile microtubes containing Stuart transport medium and transported at Faculty of Veterinary Medicine for bacteriological exams. The raw semen and the extender were bacteriologically tested in the boar centers laboratories.2.3. Bacterial Isolation and IdentificationIn this study, culture was performed by inoculating aliquots of diluted semen (DS) on the surface of Columbia blood agar (Oxoid, Hampsire, UK) with a sterile glass L shape hockey stick spreader. The Petri dishes were incubated under aerobic conditions at 37 °C and after 24–48 h each morphologically different colony was used for subcultivation on Columbia blood agar, Mac Conkey agar (Oxoid, Hampsire, UK) and EMBL (Eosin Methylene Blue) agar (Oxoid, Hampsire, UK).After cultivation on specific media the cultures were examined every 6 h. Isolated bacteria were identified using standard microbiological procedures: growth and colonial characteristics, gram staining, cellular morphology, catalase, and oxidase reaction, hemolysin production and coagulase test.The final pure culture represented a basic material for bacterial identification by API 20 E system (bioMerieux, France) and MALDI-TOF, with MALDI Biotyper (Bruker Daltonic, Karlsruhe, Germany).2.4. Antimicrobial Susceptibility TestSusceptibility tests were performed using disc diffusion method (Kirby–Bauer), according to the Clinical and Laboratory Standard Institute (CLSI) protocol [9,10]. As there are no CLSI susceptibility breakpoints available for Ralstonia pickettii or Burkholderia cepacia, the antibiotic susceptibility results were interpreted using the CLSI criteria for Pseudomonas spp. For this purpose, the antibiotics most often used in reproductive diseases in sows were used. The antimicrobial agents tested included: ceftiofur (30 µg), lincomycin (2 µg), enrofloxacin (5 µg), gentamycin (10 µg), neomycin (30 µg), flumequine (30 µg), apramycin (30 µg), penicillin (10 IU), and ampicillin (10 µg). As the control strain was used Escherichia coli ATCC 25922.2.5. Statistical AnalysesThe data were processed using the nonparametric test Mann–Whitney U, to assess the difference in semen concentration and motility in regard to bacterial isolation.A coefficient of variation (CV%) was calculated to analyze the variability between the boar centers.The cluster examination for antimicrobial resistance was performed with BIONUMERICS 8 (Applied Maths, a bioMerieux company). The results were obtained by a temporary evaluation license, and we have received the permission to publish these results.3. ResultsNo alterations regarding sperm color and concentration were observed, but the presence in semen samples of E. coli, Burkholderia cepacia, Serratia marcescens, and Proteus mirabilis was negatively associated with sperm motility (p < 0.05).The boar ejaculate is milky white in all raw semen samples with no other shade or color, and the presence of blood was not noticed. Normal color of boar ejaculate is white, with bluish shadows [1,2].Regarding the concentrations of the harvested semen, following the application of the nonparametric Mann–Whitney U test, there were no distinctly significant differences (p ˃ 0.05) between boar breeds and boar centers.The mean concentration and motility of sperm cells were 389 ± 127.5 (×106 mL−1) and 91.9 ± 7.3% respectively.There were 21 positive samples (47.72%) for bacterial contamination from the first center, 16 positive samples (45.71%) from the second center, and 9 positive samples from the third center (52.94%).Statistical analysis of variability between the artificial insemination centers was 39.31%, meaning a low degree of dispersion of values and no difference between boar centers.Out of the 96 tested doses of diluted semen (DS), only 46 (47.91%) were positive at bacteriological exams, the other 50 samples were microbiologically negative. From the bacteriologically positive semen samples, only 6 samples (13.04%) presented mixed contamination, with more than one bacterial species being isolated.Aerobic cultivation performed on 96 doses of diluted semen (DS) led to the isolation of 9 different species of bacteria identified through MALDI-TOF and API 20 E (Bio Merieux, Marcy l᾽Etoile, France) as pathogen bacteria, as well as skin and mucosal commensals and environmental bacteria. The specific species were Serratia marcescens, Ralstonia pickettii, Proteus vulgaris, Pseudomonas fluorescens, Burkholderia cepacia, Klebsiella oxytoca, Pseudomonas aeruginosa, Enterobacter spp. and Escherichia coli (Table 1, Figure 1). The most frequently occurring microorganisms were represented by Serratia marcescens, Ralstonia pickettii and Proteus mirabilis.In addition to these gram-negative bacteria with known pathogenic action, gram-positive bacteria were also identified as Streptococcus porcinus, Staphylococcus equorum, Staphylococcus succinus, and Aerococcus viridans. These isolates were not further tested because gram-negative bacteria have the most harmful effect on the semen [3,4].The antimicrobial susceptibility characterization was performed in all 46 isolates obtained (Table 2).The resistance profiles cluster analysis resulted in three groups. The first group included 13 isolates (from a total of 21 isolates) which belong to the Pseudomonaceae family. The second group included 23 isolates with heterogeneous resistance profile, including 18 isolates from the Enterobacteriaceae family. The third group included two isolates of Proteus mirabilis and E. coli with different behaviors to antimicrobials (Figure 2).4. DiscussionThe sources of bacterial contamination of semen doses are many and very diverse. Althouse and Lu (2005), in the studies performed, described the bacterial strains belonging to 25 different genera that have already been detected as semen contaminants. The presence of bacterial contamination in pig semen doses may have a negative effect on their quality and durability [3].Of the 96 doses of diluted semen (DS) studied, only in 46 doses (47.91%), the presence of bacteria was demonstrated. This relatively high percentage of positivity demonstrates that bacteriospermia or bacterial contamination is frequent in laboratories and semen processing centers. Beneman et al. (2018), and other researchers reported even higher bacterial contamination percentage (86%) [7,11]. Of the total samples tested (96 samples), 50 samples were microbiologically negative.According to the data presented in Table 1 and Figure 1, several genera and gram-negative bacterial species were isolated from the diluted seminal material. Bacterial species were present in 46 samples of the 96 doses of doses of diluted semen studied (47.91%). From 31 samples (67.39%), a single gram-negative bacterial genus was isolated. In the other 15 samples (32.60%), two gram-negative bacterial genera were identified.Serratia marcescens was the most frequent isolated (19.56%), followed by Ralstonia pickettii (17.39%), Proteus vulgaris (15.21%), E. coli (10.86%), Burkholderia cepacia (10.86%), Klebsiella oxytoca (8.69%), Pseudomonas aeruginosa (8.69%), Enterobacter spp. (4.34%), and Pseudomonas fluorescens (4.34%). Martin et al. (2010) identified gram-negative bacteria, the most common isolated microorganism being E. coli (79%), followed by Proteus spp. (36%), and Pseudomonas spp. (8%), and but also gram-positive bacteria—Staphylococcus spp. (12%), and Streptococcus spp. (9%) [12].In a study conducted by Tvrda et col. [13] in Slovakia, 12 bacterial genera and 16 bacterial species were isolated and identified in boar ejaculates immediately following semen dilution, using MALDI-TOF mass spectrometry, as follows: Proteus vulgaris, E. coli, Pseudomonas aeruginosa, Pseudomonas putida, Klebsiella pneumoniae, Aerococcus viridans, Staphylococcus aureus, Staphylococcus chromogenes, Staphylococcus simulans, Clostridium difficile, Enterococcus hirae, Bacillus cereus, Bacillus licheniformis, Bacillus subtilis, Acinetobacter iwoffii, Rothia nasimurium, and Corynebacterium spp.In our research, Serratia marcescens was isolated in a proportion of 19.56%. In other studies, the proportion in which S. marcescens was isolated is variable. Althouse et al. [2] isolated S. marcescens in proportion of 10.3%, Ubeda et al. [1] in proportion of 12.5%. Schultze et al. [14] isolated S. marcescens in proportion of 2.3% and Ralstonia pickettii in proportion of 11.4%.Most of the bacteria isolated and identified in this study are opportunistic bacteria, but which can form, alone or in combination with other bacteria, biofilm on the surfaces in the semen processing laboratory [15,16]. For this reason, we consider that it is particularly important to have a high degree of hygiene of the staff, equipment, and laboratory where the semen is taken and processed.Tvrda et al. [13] reported that 76% of semen samples had been contaminated with relatively high variety of bacterial genera, predominantly well-known uropathogens.Serratia marcescens is a bacterial contaminant that can be spermicidal when is present in extended boar semen [15,16,17]. This particular contaminant appears to originate from carrier boars, where it resides in the preputial cavity, but has also been shown to easily contaminate the semen-processing laboratory. Regarding the Serratia marcescens it was observed its high capacity to create biofilm on wet surfaces and deteriorate the sperm cell to the point of causing sperm death quickly.Ralstonia pickettii, another gram-negative, nonfermentative bacteria, was isolated in this study in a proportion of 17.39%. These bacteria can be isolated from the system of producing distilled water; that is, water used for diluting semen [16]. From a medical point of view, these strains are of particular importance, because they are responsible for the appearance of the pyometra in sows, after insemination. Similar results were obtained by other researchers [4,7,18,19,20,21,22]. Ralstonia picketii and Achromobacter xylosoxidans can be found in the water distillation system of a boar stud facility that uses this water to expand raw semen. Clark et al. [16] and other research showed that the presence of A. xylosoxidans and R. pickettii in water for semen extension of porcine semen does not detrimentally affect sperm motility or pH of the final solution regardless the choice of semen diluent [12,16,18,23,24].Escherichia coli was isolated in proportion of 10.86%. Our results differ from the findings of Maroto Martín et al. [12]. They reported the presence of E. coli in a proportion of 79%, followed by Proteus spp. 36% and Pseudomonas spp. 8%. In Italy, other researchers isolated E. coli 53% [23]. The results obtained in this study are similar to the results obtained by Althouse et al. [3], research where E. coli was isolated in proportion of 6.4%.Enteric bacteria, especially E. coli, negatively affect fertility by decreasing sperm motility, affecting the acrosome, and causing sperm agglutination [12,14,25,26,27,28,29]. In research conducted by Ubeda et al. (2013), Klebsiella oxytoca was isolated in proportion of 11.79%, Serratia marcescens 12.55%, and Escherichia coli 1.52% [1].Maroto Martín [12] and Martins [27] confirm the general opinion that boar ejaculates are more predisposed to gram-negative bacteria contamination.Semen contamination becomes relevant when it is associated with reduction of male fertility or with decreased semen boar quality. Bacterial contamination of the sow᾽s reproductive tract by artificial insemination can cause metritis, endometritis, vulvar discarches, return to the estrus, reduction of litter size, and an increased number of stillbirths and mummies [14,17,18,21,30,31,32,33]. Although the boars did not have clinical symptoms, we cannot say that they did not have bacteriospermia, since many infections of the male reproductive tract do not imply clinical diseases and are only visible when a poor seminal quality is perceived. As such, reproductive indices decrease or females increasingly present metritis, uterine infection, etc. For this reason, we consider that it is particularly important to have a high degree of hygiene among the staff, equipment, and laboratory where the semen is processed. The production of semen doses with low bacterial contamination and high viability of sperm will be possible only with a strict hygienic control in the processing of semen and having the antibiotic resistance profile of bacteria that can contaminate semen [12,34,35,36].The antimicrobial usage in semen extenders aims to reduce the bacterial contamination. However, it is known that 80–90% of the bacteria isolated from the doses of semen present various levels of antibiotic resistance [12,21,37,38]. In this study, all gram-negative bacterial strains presented resistance phenotype against at least one of the tested antimicrobial groups.The extender used in the artificial insemination center contains gentamycin. In our study 26/46 isolates (56.52%) showed resistance to gentamycin, 11 were intermediate and 10 isolated were sensitive to this antibiotic. This shows that the effectiveness of the antibiotic in doses of diluted semen is quite low. The behavior of bacterial strains for gentamycin can be different; other researchers [7,28] obtained 80% of isolates sensitive to gentamycin [7].In their study, Tvrda et al. [13] reported that gentamycin was effective enough to eradicate gram-negative bacteria, E. coli, Klebsiella pneumoniae, Proteus vulgaris strains showed 100% sensitivity to ampicillin. We can according to Maroto Martín et al. [12] Bresciani et al. [25] and Gączarzewicz et al. [28], who pointed out that a significant proportion of the bacteria commonly found in boars ejaculate in Europe may be resistant to gentamicin.Most of the gram-negative bacteria isolated in our study had a high-level resistance rate of antimicrobials when tested against neomycin, penicillin, lincomycin, and ceftiofur (Table 1). These results are correlated with previous studies [13,31,34] reporting that several bacterial genera and species exhibit a certain degree of resistance to gentamycin and aminoglycosides (the most common antibiotics used in semen extenders) leading to a concerning assumption that none of these antibiotics was able to effectively eradicate Gram negative bacteria present in diluted semen samples.5. ConclusionsIn this study, following the bacteriological examinations of the doses of diluted semen, 47.91% of them were positive, demonstrating high bacterial contamination. A predominance of gram-negative opportunistic bacteria were observed in the contaminated samples, which may be involved in uterine infections in sows or in reducing the number of fetuses.We consider it essential to make an accurate bacterial-type diagnostic and quantification method, as well as a proper antibiotic selection to use for the dose’s diluent.This work proved the presence of pathogenic gram-negative bacteria with multiple resistance to antibiotics in semen, and therefore, we highly recommend periodic microbiological screening of bacterial contamination in boars to avoid the use of low-quality semen in the pig industry.Hygienic semen collection and processing techniques and stringent laboratory procedures must be the first and primary lines of defense in successfully managing contamination. Controlling bacterial growth in extended semen with antibiotics must be a secondary method of bacterial contamination of the doses of diluted semen.Semen processing is not yet standardized among artificial insemination centers from Romania, and the critical points during production need to be identification.

animals : an open access journal from mdpi

10.3390/ani13060967

PMC10044550

Weaning is a stressful period for pigs that causes gastrointestinal disruption and low growth rates. For a long time, zinc oxide at pharmacological doses, along with different antibiotics, has been used prophylactically during this phase to reduce the incidence of these gastrointestinal problems. Nowadays, the increasing concern about the environment, along with the global and constant growth of bacterial resistance to antibiotics, has led to the prohibition of the prophylactic use of zinc oxide in diets for piglets in the EU, along with the implementation of new regulations on the use of antibiotics. Consequently, the pig sector faces the important challenge that supposes developing alternatives to the classical system based on the use of these antimicrobial compounds. This study is the first step to achieving this goal by minimizing the use of various antibiotics and zinc oxide in weanling pigs by supplementing citrus flavonoids and only one antibiotic (amoxicillin). Accordingly, the influence of zinc oxide plus antibiotics and citrus flavonoids plus amoxicillin in weaned pigs has been investigated and its impact on growth performance, gut microbiology profile, gut signaling, intestinal architecture, and serum biomarkers indicative of stress and inflammatory responses have been evaluated. Citrus flavonoids plus amoxicillin improved growth performance and gut health, evidencing a positive microbial modulation, stress status reduction, and a positive effect on the gastrointestinal barrier, and other digestive functions. Additionally, the expression of some bitter taste receptors in the intestine has been increased when supplementing both dietary strategies, the one based on zinc oxide or the one based on citrus flavonoids supplementation. Consequently, the present study shows that in weanling piglets, the supplement of citrus flavonoids with amoxicillin might be a promising alternative to the dietary use of pharmacological doses of zinc oxide with more than two antibiotics, therefore minimizing the use of antimicrobial compounds without detrimental effects on performance.

Since citrus flavonoids have antioxidant and anti-inflammatory properties, it was hypothesized that these compounds would become a suitable alternative to the use of therapeutic doses of zinc oxide at weaning. A total of 252 weaned pigs ([LargeWhite × Landrace] × Pietrain) were distributed according to BW (5.7 kg ± 0.76) into 18 pens (6 pens per diet, 14 pigs/pen). Three experimental diets for the prestarter (0–14 d postweaning) and starter (15–35 d postweaning) period were prepared: (i) a nonmedicated (CON) diet, (ii) a CON diet supplemented with zinc oxide at 2500 mg/kg, amoxicillin at 0.3 mg/kg and apramycin at 0.1 mg/kg (ZnO), and (iii) CON diet with the addition of a commercial citrus flavonoid extract at 0.3 mg/kg and amoxicillin at 0.3 mg/kg (FLAV). Pig BW, ADG, ADFI, and FCR were assessed on d7, d14, and d35. Samples of intestinal tissue, cecal content, and serum were collected on day seven (18 piglets). FLAV treatment achieved greater BW and ADG during the starter and for the entire experimental period compared with the CON diet (p < 0.05), whereas ZnO pigs evidenced intermediate results. Jejunum tissue analysis showed that pigs fed the FLAV diet overexpressed genes related to barrier function, digestive enzymes, and nutrient transport compared to those pigs fed the CON diet (p < 0.05). An increase in the abundance of bacterial genera such as Succinivibrio, Turicibacter, and Mitsuokella (p < 0.05) was observed in the FLAV compared with the CON and ZnO piglets. ZnO and FLAV increased the expression of TAS2R39, while ZnO pigs also expressed greater TAS2R16 than CON (p < 0.05) in the intestine. FLAV treatment improved the gut function, possibly explaining a higher performance at the end of the nursery period. Consequently, citrus flavonoids supplementation, together with amoxicillin, is a promising alternative to the use of zinc oxide plus amoxicillin and apramycin in weanling pigs, minimizing the use of antibiotics.

1. IntroductionAmong livestock production, the swine industry provides one of the most important protein sources for human nutrition [1]. Even though in recent decades a high degree of improvement has been achieved at the genetic level (prolificity, efficiency, and performance), paradoxically, the weaning phase continues to be a critical period within this system. Indeed, the improvement in a sows’ prolificity has led to a reduction in the birthweight of neonates, also increasing litter heterogeneity with serious consequences during postnatal life [2]. For decades, several antimicrobial compounds, such as zinc oxide at pharmacological doses and antibiotics as growth promoters, have been widely used as a practical strategy to reduce and control postweaning syndrome in pigs. For instance, zinc oxide possesses a multifactorial mode of action, improving digestion and nutrient digestibility, acting as an antioxidant and immunomodulatory molecule with antibacterial effects and, consequently, impacting positively on intestinal morphology and health [3]. However, antibiotics and zinc oxide are involved in the continuous increase of antibiotic-resistant bacteria and, even more, zinc oxide is the direct cause of negative effects polluting the environment [4]. Therefore, accordingly to the worldwide cross-sectorial strategy for “One Health” fixed by the FAO, the OIE, and the WHO, the use of antibiotics should be controlled and restricted. Meanwhile, zinc oxide at pharmacological doses has been completely banned in the European Union since June 2022 [5].Different feed additives have been studied and proposed as an alternative to zinc oxide to improve postweaning pigs’ health and reduce the incidence of diarrhea [3]. Citrus flavonoids possess antioxidant and anti-inflammatory properties, and different studies have evidenced their effects as antimicrobial and immunomodulatory molecules [6,7,8]. Additionally, citrus flavonoids and their metabolites have shown beneficial effects on the intestinal barrier through different mechanisms of action [8]. Citrus flavonoids, as bitter compounds, modify the gene expression of bitter taste receptors (TAS2R), similar to other compounds with antimicrobial, anti-inflammatory, and antioxidant capacity [9], and some bitter compounds have shown promising effects to substitute zinc oxide at pharmacological doses [10]. Therefore, it was hypothesized that citrus flavonoids would become a suitable alternative to the use of therapeutic doses of zinc oxide, also reducing the use of antibiotics in weanling pigs. The objective of this study was to evaluate the effect of supplementing citrus flavonoids and amoxicillin in replacing the use of pharmacological doses of zinc oxide together with several antibiotics in weanling piglets as a first step in the demedicalization process of this productive phase. Accordingly, gene expression, microbiota, and histomorphology in the guts of weanling pigs, along with performance parameters and stress markers, were studied to elucidate the potential application of citrus flavonoids in weanling pigs.The experiment was planned, designed, and conducted at the end of the year 2018, aiming to deal with the reduction of antimicrobials in swine production (antibiotics and zinc oxide) and facing a new scenario and anticipating the ban on zinc oxide in 2022.2. Materials and Methods2.1. Ethics StatementExperimental procedures were approved by the Ethics Committee of the Universitat Autònoma de Barcelona (approval code CEEAH2788M2) based on the European Union guidelines for the use of animals in research [11].2.2. Animals and HousingAt weaning (21 ± 1.6 d), 252 pigs ((Large White × Landrace) × Pietrain) with an initial average BW of 5.7 ± 0.76 kg were used in a 35-d study. Animals were identified through an ear tag, individually weighted, blocked according to the initial BW, and distributed into three experimental diets in 18 pens (6 pens per treatment, 14 pigs per pen) within the same weanling room. Each pen (3.12 m2) allotted entire males and females. The commercial weaned facility owned by the Universitat Autònoma de Barcelona had a fully slatted floor, unlidded hoppers (TR5, Rotecna, Spain), and nipple drinkers. Temperature and ventilation rates were monitored using thermostatic heaters and exhaust fans adjusted depending on the age of the pigs (28 to 22 °C).2.3. Experimental Design and Dietary TreatmentsTwo-phase diets (Table 1) were formulated to meet or exceed nutrient requirements [12]: the prestarter (PS) phase from days 1 to 14 and the starter (ST) from days 14 to 35. Three experimental diets were prepared according to the supplementation or not of antimicrobial and experimental products. For the PS feed phase, a nonmedicated (CON; no antibiotic or pharmacological levels of zinc oxide) was used as the basal diet for all experimental diets. The positive control diet (ZnO) consisted of the addition of zinc oxide at 2500 mg/kg, amoxicillin at 0.3 mg/kg, and apramycin at 0.1 mg/kg products to the CON diet. The third diet consisted of the CON diet plus the addition of a commercial citrus flavonoid product (Bioflavex®, HTBA, Barcelona, Spain) at 0.3 mg/kg and amoxicillin at 0.3 mg/kg (FLAV). During the ST phase, diets remained the same except for the ZnO diet. It consisted of the addition of zinc oxide at 1500 mg/kg plus the addition of amoxicillin at 0.3 mg/kg, neomycin at 0.19 mg/kg, and tiamulin at 0.10 mg/kg. The antimicrobial plan used in the positive control was applied based on the in-farm therapeutic strategy already conducted in the commercial conditions according to the health status and inherent challenges (common clinical signs of postweaning diarrhea associated with ETEC E. coli, Streptococcal meningitis, and swine dysentery at the second phase of the nursery period). In the FLAV treatment, amoxicillin was used throughout the study due to the broad spectrum of this antibiotic and the high prevalence of Streptococcal meningitis on the farm. For each dietary treatment, the feed was offered in mash and pellet form during the PS and ST phases, respectively. Both feed and water were provided ad libitum.2.4. Experimental Procedures and SamplingIn order to calculate the average daily gain (ADG), average daily feed intake (ADFI), and feed conversion ratio (FCR), the individual BW of the pigs, as well as feed disappearance from each feeder, was recorded on days 7, 14, and 35. Pigs were daily monitored for mortality and illness incidents.2.4.1. Feed AnalysisProximate analytical determination of diets was performed following the Association of Official Agricultural Chemists International (2005) methods: dry matter (AOAC 934.01), crude protein (AOAC 968.06), ether extract (AOAC 920.39), and ash (AOAC 942.05). Neutral detergent fiber content was determined using the Ankom nylon bag technique (Ankom 200 fiber analyzer, Ankom Technology, Macedon, NY, USA).2.4.2. Proinflammatory and Stress MarkersFrom each pen, three pigs (average BW) were selected to take blood samples on days 0, 7, and 14 of the trial. The same animal was sampled during the experimental period. Samples were taken by jugular puncture and collected in an individual vacutainer tube without anticoagulant (10 mL). Tubes were subsequently centrifuged at 1500× g for 10 min to obtain serum. Aliquots of serum were stored at −80 °C until chemical analyses. A PigMap analysis was performed on the Olympus AU 400 analyzer [13]. Two commercial ELISA kits were utilized to analyze cortisol (DRG, Marburg, Germany) and porcine TNF-alpha (R&D Systems, Abingdon, UK) content.2.4.3. Histomorphological AnalysisOn day 7, samples of the jejunum and ileum tissue (about 5 cm) were collected from one pig per pen (n = 6). Selected animals exemplified the average pig BW within the pen. Pigs were sedated using zolazepam and xylazine and subsequently euthanized by an overdose of pentobarbital. Samples were fixed in 4% paraformaldehyde and then embedded in paraffin. Then, sections were stained with a hematoxylin-eosin solution. Villus height, crypt depth, villus height to relative crypt depth ratio (V:C), and the number of goblet cells and lymphocytes were measured by using a light microscope. Only full and vertical-oriented villi were considered within the analysis.2.4.4. Gene Expression AnalysisFrom the same jejunum and ileum sections, samples of about 1 cm were collected. Intestinal samples were rinsed in phosphate-buffered saline and immediately snap frozen in 1 mL of RNAlater (Deltalab, Rubí, Spain). Samples were stored at −80 °C until analysis. An Open Array Real-Time PRC Platform (Applied Biosystems, Waltham, MA, USA) was self-designed as described by González-Solé [14]. Briefly, extensive literature research was accomplished to finally select 56 genes. The main criteria for selection were the involvement of the gene on physiological functions such as immune response, barrier function, digestion processes, nutrient transport, and stress responses. Sample procedures, data collection software, and sample quality analysis were executed as previously defined by Villagómez-Estrada [15].In addition, four genes related to bitter taste receptors were individually analyzed (TAS2R7, TAS2R16, TAS2R38, and TAS2R39) from jejunum and ileum samples. RNA was extracted by homogenizing tissues in TRIzol (Invitrogen, Waltham, MA, USA) using a Polytron Instrument (IKA, Königswinter, Germany). PrimeScript RT Reagent Kit (Takara, Frankfurt, Germany) was used to transcribe RNA to cDNA, following the manufacturer’s instructions. The quality of RNA was assessed by a NanoDrop instrument (ThermoFisher, Madrid, Spain) at 260, 280, and 230 nm. Quantification of gene expression was performed as described in Paniagua et al. [9].2.4.5. Microbial Molecular AnalysisOn day 7, from the same selected animals, cecal content (500 mg) was collected for 16-S microbial sequencing analysis. Samples were immediately frozen and stored at −80 °C until analysis. Bacterial DNA was obtained by following the manufacturer’s instructions for the commercial QIAamp DNA Stool Min Kit (Qiagen, West Sussex, UK). After the proper verification of DNA concentration and purity (NanoDrop 1000 Spectrophotometer; Termo Fisher, Wilmington, DE, USA) it was eluted in 200 μL of Qiagen buffer AE ad stored at −80 °C until analysis. Samples were amplified (500 cycles) using the MiSeq® Reagent Kit v2 (MiSeq, Illumina, San Diego, CA, USA) for the V3–V4 regions of the bacteria. Sequence reads were processed on the QIIME v.1.9.1 pipeline [16] and clustered to operational taxonomic units (OTUs) at a 97% sequence similarity. Reads were selected by the subsampling open reference approach at 10% of sequences subsampled [17] and assigned to a taxonomy using 16S GreenGenes v.13.8 reference database [18] at a 90% confidence threshold. OTUs with a relative abundance across all samples lower than 0.005% were eliminated [19].2.4.6. Statistical AnalysisPrior to ANOVA analysis, the normality and homogeneity of the data were assessed using the Shapiro–Wilk test. Data were analyzed through the MIXED procedure of SAS (version 9.4, SAS Institute, Cary, NC, USA) considering a randomized complete block design. The effect of the experimental diet was considered a fixed effect, whereas the BW block counted as a random effect. The pen was the experimental unit. In the case of acute phase proteins, histomorphological measurements, gene expression, and microbiota community data corresponded to the individual pig selected that represented the average BW in each pen, therefore the animal represented the pen (the experimental unit). For acute phase proteins and stress markers, the initial measurements (d 0) were used as covariables. Digesta bacteria diversity was analyzed at the OTU level using a vegan package [20]. Shannon and Inverse Simpson estimators were used to assess the alpha diversity on the basis of raw counts. While beta diversity dissimilarities were evaluated by multivariate ANOVA using the Adonis function. An ANOVA test was performed to test the experimental group’s differences in microbial richness and diversity. The zero-inflated log-normal mixture model was considered by the differential abundance analysis.Biostatistical analysis of gene expression and microbiota community data was performed in R Studio v.3.5.1 software. The means and SEM values for bitter taste gene expression data presented in the figures correspond to nontransformed data and p-values to those obtained by an ANOVA of transformed data. They were analyzed using a mixed-effects model (version 9.2, SAS Inst.Inc., Cary, NC, USA), where the model included treatment, tissue, and the interaction between treatment and tissue as fixed effects. For both, gene expression and microbiota analysis, the Benjamini and Hochberg false discovery rate (FDR) multiple testing correction was performed [21]. Significance was declared at a probability of p ≤ 0.05 and tendencies were considered when the p-value was between >0.05 and <0.10.3. Results3.1. Growth PerformancePig performance response for the experimental diets is shown in Table 2. Feeding the FLAV diet tended to increase the pigs’ BW at day seven when compared to CON animals (p < 0.10). Interestingly, pigs from the FLAV treatment achieved greater BW and ADG during the ST and for the entire experimental period compared to the CON diet (p < 0.05). However, BW and ADG from the ZnO animals showed no differences when compared to CON and FLAV treatments at the end of the study. Although both treatments, FLAV and ZnO, improved ADFI during the ST and for the entire experimental period, no influence of experimental diets was observed on FCR (p > 0.10).Mortality was 1.59% and was not related to any dietary treatment (p = 0.752).3.2. Gene Expression AnalysisFor a complimentary evaluation of the experimental diets’ effects on pig physiological status, a gene expression analysis of jejunum and ileum tissue was performed on day seven postweaning (Figure 1 and Figure 2). A total of 46 genes were successfully amplified, though only up to 12 genes significantly differed among the experimental diets (p < 0.05; Table 3 and Table 4). Gene expression differed between intestinal tissues. The analysis of jejunum tissue showed dissimilarities in gene expression compared to ileum tissue. Pigs fed the FLAV diet overexpressed genes related to barrier function (MUC13), digestive enzyme (DAO1, GCG, HNMT, and SI), and nutrient transport (SLC13A1 and SLC15A1) compared to those pigs fed the CON diet (p < 0.05; Table 1). Interestingly, pigs the fed ZnO diet showed downregulation of the Zn transporter gene (SLC39A4) compared to pigs fed the CON and FLAV diets (p < 0.0001). Ileum analysis showed that pigs fed the CON diet had an upregulation of genes involved in barrier function (MUC2 and TFF3), inflammatory response (IFNGR1 and IL8), and antioxidant enzymes (GPX2 and SOD2m) but a downregulation of the PPARGC1α gene (immune response) compared to the FLAV diet (p < 0.05; Table 3). However, the CON pigs had downregulation of some nutrient transport genes (SLC11A2 and SLC39A4) than the ZnO pigs (p < 0.05; Table 4).3.3. Bitter Taste ReceptorsRegarding the TAS2R gene, no interaction was observed between experimental diets and tissues for the different receptors studied. The gene expression of TAS2R7 and TAS2R38 was not affected either by diet or by tissue. Nonetheless, the effect of diet and tissue was observed on TAS2R16 and TAS2R39 gene expression (Figure 3 and Figure 4).Pigs supplemented with zinc oxide had a significant increase in the gene expression of TAS2R16 (p < 0.05; Figure 3) compared to the CON animals, and the expression of this receptor was greater in the ileum than in jejunum samples (p < 0.05; Figure 4), regardless of the experimental diet. Additionally, FLAV and ZnO diets increased the gene expression of TAS2R39 compared with the CON pigs (p < 0.05; Figure 3), whereas TAS2R39 was less expressed in the ileum than in the jejunum (p < 0.05; Figure 4).3.4. Microbial Molecular AnalysisNo effect of the experimental diet was observed for alpha estimators (p > 0.10; Table 5). Beta diversity analysis revealed distances between clustered samples of the FLAV and CON groups (PADONIS = 0.023; Figure 5).At the genus level, when pig diets were supplemented with zinc oxide, there was a decrease in the relative abundance of several genera such as Lactobacillus, Desulfovibrio, Anaerovibrio, Lachnospira, Lachnobacterium, and Actinobacillus compared to the CON diet (p < 0.05; Figure 6). Whereas, when the FLAV diet was offered to pigs the relative abundance of genera such as Dorea, Desulfovibrio, and Actinobacillus, among others, decreased compared to the CON diet (p < 0.05; Figure 7). Interestingly, the abundance of several genera of bacteria such as Lactobacillus, Roseburia, and Clostridium increased compared to the CON diet (p < 0.05; Figure 7). Comparing the FLAV diet to the ZnO diet, an increased abundance of Mitsuokella, Lactobacillus, Megasphaera, Succinivibrio, Veillonella, Streptococcus, and Fibrobacter (p < 0.10; Figure 8) was observed.3.5. Intestine Histomorphometry MeasurementsIleum histomorphometry parameters are shown in Table 6. No effect of experimental diets on ileum histomorphometry parameters was observed except for a tendency on villus height (p = 0.094) and villus height to crypt depth V:C ratio (p = 0.059). Pigs fed the ZnO diet tended to have longer ileum villi and a higher V:C ratio than those fed the CON diet.3.6. ProInflammatory and Stress MarkersAs shown in Table 7, serum cortisol levels measured at day seven postweaning were lower in pigs fed the FLAV diet compared to those fed the ZnO diet (11.2 vs. 25.1 ng/mL; p = 0.007). No influence of experimental diets was observed on serum TNF-alpha and PigMap levels (p > 0.10).4. DiscussionThe main objective of this study was to explore a commercial approach to minimize the use of multiple antimicrobial substances in weanling diets, such as zinc oxide and antibiotics, through supplementation with citrus flavonoids. As the trial was performed under commercial farm conditions, the FLAV treatment also included amoxicillin in addition to citrus flavonoids due to the historical sanitary status of the farm and ZnO treatment included more antibiotic molecules following their conventional strategy to reduce postweaning health disorders. Although the major function of antibiotics is to act as antimicrobial substances against bacteria modifying the normal microbial community of the intestine, their anti-inflammatory effects have been also described [22]. Consequently, the effects of the ZnO and FLAV groups are discussed. However, the potential effects of all substances included in each strategy should be considered and the effects observed cannot be attributed only to zinc oxide or citrus flavonoid supplementation even if some parts of the discussion are focused on the potential effects of the supplementation of these substances.Weaning, in addition to activating stress signaling pathways, can also alter the normal expression of intestinal genes in pigs [23,24]. After seven days of feeding, the FLAV treatment increased the jejunum mRNA levels of the genes related to barrier function (MUC13), digestive enzymes (DAO1, GCG, HNMT, and SI), and nutrient transport (SLC13A1 and SLC15A1) compared to those fed the CON diet. Citrus flavonoids, such as naringenin, possess well-demonstrated properties as anti-inflammatory and antioxidant molecules [8,25,26] that would explain the reduction in the inflammatory and antioxidant response in the intestines of these pigs. Interestingly, de Groot et al. [27] performed a study to establish inflammation patterns in weanling pigs to better understand, in future research, which interventions can positively affect the intestine in this phase. Thus, their results showed that the gene expression of proinflammatory cytokines was increased in the jejunum, ileum, and colon after weaning and during the first 15 days postweaning, for example, the expression of the IL-8 gene. Therefore, the downregulation of the IL-8 gene found in FLAV pigs, but not in the CON group, can be associated with a lower inflammatory environment in the jejunum at day seven. Indeed, citrus flavonoids’ properties improving barrier function have been largely described in humans and animals, mainly attributed to the hesperetin and naringenin flavonoids, which can regulate the expression and secretion of mucins [8]. For instance, MUC13, a transmembrane mucin glycoprotein, is an important component of intestinal structure and whose unsuitable expression may predispose to infectious and inflammatory diseases [28]. In the same way, a greater mRNA expression of nutrient transporters and digestive enzymes such as SI may indicate an enhanced intestinal maturity and absorptive function that likely contributed to improved digestion and growth. Interestingly, a downregulation of the Zn transporter gene SLC39A4 was found in ZnO pigs indicating negative feedback to restore zinc homeostasis [29].One of the hypotheses of this work was that citrus flavonoids may be a suitable alternative to pharmacological doses of antibiotics and zinc oxide probably led by the effect of bitter compounds on bitter taste receptors (TAS2R). These taste receptors have been described along the gastrointestinal tract constituting the chemosensory system, modulating the proper digestive response (secretion, motility, absorption of nutrients, or aversion), food intake, and behavior [30,31,32,33]. In the present study, FLAV and ZnO pigs had greater expressed TAS2R39 than CON animals in the intestine at day seven, whereas ZnO increased the gene expression of TAS2R16 as well. Recently, it has been demonstrated that different TAS2Rs differentially modulate food intake in rats through different enteroendocrine secretions [34]. In accordance with these results, FLAV and ZnO pigs showed a higher ADFI at the end of ST and for the whole study (days 0 to 35). Additionally, activation of TAS2R in the stomachs of rats has been proven to stimulate ghrelin secretion, an orexigenic hormone that enhances gastrointestinal motility [35] that could explain the increase in the ADFI of pigs when zinc oxide or citrus flavonoids were used in our study. Unfortunately, the gene expression of the TAS2Rs was not analyzed in the stomachs of the pigs, so further research will be needed to elucidate if these receptors are involved in the orexigenic effect observed in FLAV and ZnO animals. Moreover, ZnO treatment also increased gene expression of TAS2R16, whereas FLAV pigs showed numerically higher gene expression than CON, though not statistically different. Paniagua et al. [9,36], observed a modulation in the gene expression of TAS2R16 in the rumen epithelium when bulls were supplemented with citrus flavonoids. Therefore, in agreement with these results, bitter substances increased the gene expression of TAS2R16 in the intestine of pigs. Although some antimicrobial molecules, such as chloramphenicol and erythromycin, have been described as bitter molecules activating TAS2R [37], as far as we know, the antibiotics used in our study have not been confirmed as bitter molecules able to activate TAS2Rs. Consequently, zinc oxide and citrus flavonoid supplementation might explain the activation of TAS2Rs observed in the intestines of the pigs.No statistical interaction was found for treatment and tissue (jejunum and ileum) when TAS2Rs’ gene expression was analyzed, however, the gene expressions of TAS2R16 and TAS2R39 also showed different expressions depending on the tissue, jejunum, or ileum, whereas TAS2R7 and TAS2R38 were not affected. Colombo et al. [38], studied the expression of some TAS2Rs in the gastrointestinal tract of five young pigs, and they found differences in different gastrointestinal points (stomach, jejunum, and colon), and concluded that the expression for all TAS2Rs studied was very low. Genes such as TAS2R7 were predominantly found in the stomach, TAS2R16 in the jejunum and TAS2R38 was not frequently expressed. Thus, the results of the present study might be considered in agreement with Colombo et al. [38], as TAS2R7 evinced low gene expression in the intestines of all the animals; TAS2R16′s expression was found in the jejunum of the animals studied by Colombo et al. [38], and our results evinced that this receptor responded to the treatment in the tissues studied; probably TAS2R38 is not important in the gastrointestinal tract of the pig, based on the results of both studies; finally, Colombo et al. [38] did not study the TAS2R39, whereas our results have evinced that this receptor showed a higher degree of gene expression and response to the treatments. As far as we know, this is the first study in pigs where the gene expression of TAS2R39 and its response to bitter molecules have been described. Although much more research is needed to properly elucidate the functions and responses of TAS2R39 in pigs, this receptor might be involved in the multifactorial mechanism of action of zinc oxide and citrus flavonoids observed in pigs at weaning. Furthermore, TAS2R activation is related to the secretion of some enteroendocrine molecules, such as PYY and CKK [9,30,34,36]. Interestingly, CCK is an intestinal peptide that increases the release of digestive enzymes from the pancreas in the duodenum, facilitating digestion [30]. Thus, citrus flavonoids might increase digestive enzymes in the jejunum through the activation of TAS2Rs. Additionally, citrus flavonoids exert different positive effects on intestinal health through different biological mechanisms (anti-inflammatory, antioxidant, and barrier function improvement) [8].Among all parameters that can influence animal growth performance, intestinal integrity, along with a stable microbiota, is particularly important to enhance feed efficiency and maintain overall health. Curiously, when intestinal morphology was evaluated, no differences were observed between experimental diets. However, taking a deep look into gut health parameters, diets offered to pigs modulate the microbiome community composition. Although alpha diversity indices showed a similar richness within experimental diets, the beta diversity analysis showed dissimilarities among those microbial communities. This finding suggests that the difference is not due to the presence of specific bacteria but to the larger abundance of certain bacteria [39]. Indeed, pigs fed with the ZnO diet decreased the relative abundance of genera such as Lactobacillus and Lachnospira compared to the CON diet. Interestingly, FLAV diets increased the abundance of these same bacteria genera together with Roseburia compared to the control diet. The importance of the intestinal microbiota for gastrointestinal function and their association with animal health and growth has been previously shown [39,40,41,42]. Indeed, several bacteria such as species of the Lactobacillus, Bifidobacterium, Lachnospira, or Roseburia genera are known as beneficial functional microbes which have great potential to contribute to the reduction of the growth of pathogenic bacteria, stimulation of the immune system, feed intake, and feed efficiency [43]. Furthermore, many members of these bacteria can produce, directly or indirectly, certain metabolites known as postbiotics, such as short-chain fatty acids (SCFA). The SCFAs (i.e., butyrate, propionate, and acetate) are produced by fermenting dietary nondigestible carbohydrates [44]. It is widely accepted that the role of SCFA as a colonocyte energy supply which instead can exert beneficial effects on intestinal barrier function and reduction of gut inflammation [45], as well as a crucial physiological effect on several organs, including the brain [46]. Therefore, these findings support the idea that FLAV treatment might improve pig growth by maintaining a stable and healthy intestinal microbiota able to increase its absorptive capacity.Moreover, it is worth noting the lower plasma concentration of cortisol found at day seven postweaning in pigs fed the FLAV diet. Cortisol, in addition to being one of the widest biomarkers of stress in pigs, is considered a powerful suppressor of the immune response (production of cytokines and immunoglobulins) and whose secretion predominates under the prolonged exposition to a stressful stimulus [47]. As mentioned before, the stress in pigs, as a consequence of weaning, can severely impact the body‘s homeostasis and consequently, performance. Therefore, there is a primary interest in minimizing stress levels by improving management activities or through dietary interventions. The present results suggest that the FLAV diet can reduce the stress levels in pigs by approximately twofold compared to the ZnO diet, at least after the first seven days postweaning. Previous studies conducted in beef cattle have shown that supplementing citrus flavonoids can reduce aggressiveness by modulating animal behavior [9,37,48,49], so citrus flavonoids supplemented in the FLAV treatment might explain this reduction of cortisol in plasma. Nonetheless, further research is needed to corroborate the possible positive effects of citrus flavonoids on animal behavior and stress reduction in pigs, as well as the mechanisms of action associated.In the present study, pigs fed the FLAV diet showed improved performance during the first 35 days postweaning compared to the CON diet, whereas the ZnO pigs were intermediate. Interestingly, ZnO negatively affected the abundance of some beneficial genera, such as Lactobacillus and Lachnospira, whilst only a few nutrient transporter genes were better expressed in the ilium of these animals compared with the CON piglets. On the contrary, as previously discussed, the FLAV piglets evinced better intestinal maturity together with a greater microbiome gut community, increasing the relative abundance of some genera known as beneficial functional bacteria (Lactobacillus or Roseburia), which might explain the greater performance observed in the FLAV animals at the end of the study. Consequently, although the ZnO group evinced greater ADFI, only numerical improvement was observed for BW and ADG at 35 days postweaning compared with the CON animals, whereas the FLAV piglets achieved greater BW, ADFI, and ADG.5. ConclusionsTaking these results together, it must be stressed that feed and its associated nutritional components influence directly the major components of gut health such as gene expression, microbiota environment, and intestinal integrity. Consequently, supplementing feed with functional additives such as citrus flavonoids together with antibiotics like amoxicillin can positively impact intestine digestive and immune function and, thereby, on pig growth along with lesser inflammatory and pro-oxidative status, representing thus a first step to minimize the use of antibiotics and avoiding pharmacological doses of zinc oxide. Additionally, TAS2R seems to be involved in the multifactorial mechanism of action of these substances (zinc oxide and citrus flavonoids) supplemented with antibiotics. The next steps should be to evaluate the effect of the supplementation of these flavonoids without any dietary antibiotics added to confirm them as a promising alternative to dietary antibiotic use as well, improving health and performance in weanling pigs.

animals : an open access journal from mdpi

10.3390/ani11071971

PMC8300318

Environmental enrichment is a combination of techniques that aim to improve the quality of life of zoo animals. However, institutions might be reluctant to add certain enrichment items due to the belief that their presence could negatively affect the visitor experience in the zoo. To explore the veracity of this belief, we assessed visitor attitudes towards two types of enrichment items (naturalistic vs. artificial looking) in an outdoor walk-through enclosure for ring-tailed lemurs in Zoo Planckendael (Belgium). We developed a questionnaire that was answered by 371 visitors. We also took into consideration the behaviour of the animals and their visibility. We found that the visitor attitudes were more influenced by the behaviours displayed by the lemurs than the appearance of the enrichment items. We suggest that more emphasis should be placed on designing enrichment items that provide the animals with opportunities to display more active and appropriate behaviours, regardless of the appearance of the objects, in order to improve animal welfare while simultaneously improving the visitor experience.

Decisions on environmental enrichment programmes are sometimes based on the assumption that non-natural or artificial looking items negatively affect visitor experiences. In this study, we developed a questionnaire to assess zoo visitor attitudes towards enrichment appearance in an outdoor walk-through enclosure for ring-tailed lemurs (Lemur catta). Naturalistic and artificial looking enrichment items were alternately provided in the enclosure. A total of 371 visitors filled out the questionnaire: 174 in the naturalistic and 197 in the artificial conditions. Both researchers and visitors conducted behavioural observations of the lemurs. Our results suggest that the appearance of the items did not have an effect on visitor attitudes and that visitors recognised both naturalistic and artificial items as enriching for the animals. Moreover, the behaviour and visibility of the lemurs had a greater effect on the visitors’ attitudes. We suggest that during the design of enrichment items, less concern should be placed on the appearance of the items and more on their effect on animal behaviour. Ultimately, this would improve both animal welfare in captivity and the visitor experience.

1. IntroductionIt is generally accepted that modern zoos have four roles, including conservation, education, research, and entertainment of the visitors [1]. The role of animal–visitor interactions (AVI) is often more focused on the impact such AVIs have on the welfare of animals, and less on zoo visitors. Yet, previous studies suggest that AVIs influence zoo visitor experiences. Spooner et al. examined the effects on the visitor experience of both animal shows [2] and educational theatre [3] at zoos. They found that live animal shows but also family theatre without live animals were effective in increasing visitors’ knowledge on basic animal features as well as awareness of the zoos’ conservation efforts. Additionally, public animal-training sessions produced more positive zoo experiences for visitors [4].The effects of zoo animal behaviour on the perceptions and behaviours of visitors have also been studied on several occasions. For instance, Altman [5] observed that the more ‘animated activities’ displayed by polar bears (Ursus arctos maritimus), the more the visitor attention was directed to the animals and their behaviours, as showed by an increase in the behaviour content in the conversations among the visitors. Godinez et al. [6] and Miller [7] examined visitor responses to the behavioural displays of zoo animals, with particular attention to stereotypies. Negative visitor perceptions were related to the animals being out of sight [6] or displaying stereotypic behaviours [6,7]. Tofield et al. [8] examined both naturalistic exhibits and behavioural displays on zoo visitor reports taken from interviews and observed that visitors found enriched exhibits more attractive and spent more time at more naturalistic enclosures.Environmental enrichment is a combination of techniques with the aim to enhance the quality of life of zoo-housed animals by providing environmental stimuli necessary for optimal psychological and physiological wellbeing [9,10,11,12]. Good environmental enrichment programmes facilitate the expression of species-specific behaviours by increasing behavioural diversity [9,10,13,14], including extending foraging time and problem-solving [15], reducing extended periods of inactivity and undesired behaviours [15,16,17,18] and lowering the frequency of aggressive behaviours [13,19]. Environmental enrichment programmes also aim to increase the animal’s sense of control and ability to choose [12], which is an important contributor to animal welfare since the perception of unpredictable situations, being unable to choose and not having a sense of control over the environment, have been related with stress [11].Environmental enrichment, as noted by Mellen and MacPhee [9], is a holistic approach that involves both the enclosures where animals are housed and the stimuli and events that occur within such enclosures. Over the last few decades, animal enclosures have transitioned to naturalistic environments to enhance visitor attitudes and improve animal welfare [1,9,20,21,22,23]. The environment surrounding an animal influences the visitor perception of that animal [24,25] and visitors show a preference for more naturalistic enclosures [24]. Naturalistic enclosures tend to be more complex and, therefore, improve animal welfare. In addition to the permanent elements of the enclosure design, often more temporary environmental enrichment objects are added. These objects can either be naturalistic or artificial looking, mostly depending on the materials used to build them. The design and provision of environmental enrichment relies heavily on the enthusiasm and dedication of motivated animal caretakers. However, in order to succeed, enrichment programmes need support from other entities within the zoo, such as directors, curators, managers, veterinary and other zoo staff [9,26]. Still, institutions may be reluctant to add certain enrichment devices due to the (untested) assumption that their presence can negatively affect the visitor experience, especially in a more naturalistic enclosure [21,27]. One way to overcome this problem, is to provide naturalistic looking items. Indeed, it has been shown that naturalistic exhibit designs are more attractive and informative to the visitors [1,20,21,24,25] and it could be argued that natural looking enrichment would add to the cohesion of a naturalistic enclosure [28,29]. However, artificial materials are often considered cheaper, more durable, easier to be cleaned and disinfected, as well as easier to modify without compromising the structure of the material, compared to more natural looking materials [30]. This could be a reason why environmental enrichment (especially artificial looking items) is sometimes provided off-exhibit, in areas not accessible to public view [21]. Yet, it has been suggested that animals in an enriched environment are more likely to exhibit active and species-appropriate behaviour and this is probably more informative and interesting to zoo visitors [9,22,23,26]. Visitor attitudes are of importance to the zoo management and a positive influence on visitor experience would thus be an additional benefit of enrichment. Davey et al. [14] assessed visitor behavioural responses to environmental enrichment improvements in a mandrill (Mandrillus sphinx) enclosure at Beijing zoo (China). They observed that the improved enclosure attracted more visitors, who stayed longer, suggesting that the enriched design was more attractive than the previous exhibit. However, they did not compare artificial with naturalistic looking enrichment, but a barren enclosure with a remodelled naturalistic exhibit with enrichment devices.Most studies on the impact of environmental enrichment on zoo visitors’ experience suggested that a mismatch between enrichment appearance and exhibit design has little effect on the attitudes of the visitor [21,27,29,31]. Jacobson et al. [31] explored the effect of artificial looking enrichment on visitor perceptions in a naturalistic chimpanzee (Pan troglodytes) enclosure in Lincoln Park Zoo (USA). They concluded that the visitors were not affected by the aesthetics of the enrichment and showed that the largest proportion of visitors thought the naturalism of the enclosure was important. Similar results are reported in a study investigating the effects of enrichment in a polar bear enclosure on visitor perception: enrichment type, either artificial or naturalistic looking, did not alter visitor perceptions about the zoo, the enclosure, the species on display, or the individual animals [21]. McPhee et al. [27] surveyed zoo visitors about their perceptions of naturalism in polar bear, tiger (Panthera tigris altaica), lynx (Lynx canadensis), and fishing cat (Prionailurus viverrinus) exhibits in Brookfield Zoo (USA), but their results did not report an influence of the type of enrichment on visitor impressions.In another study conducted at Brookfield Zoo [28], visitors were asked to score videos and photographs of different exhibits, with or without animals, and with different types of enrichment, naturalistic or artificial looking. Contrary to previous research, Razal and Miller’s results suggested that different types of enrichment, as well as the naturalistic appearance of the enclosure, might have an impact on how visitors perceive animals and their exhibits. They found that visitors rated the pictures with naturalistic looking enrichment higher for questions related with good animal welfare and the enclosure suitability and naturalness. Visitors also ranked the naturalistic looking enrichment items higher.According to Luebke and Matiasek [32], the most common factors that impact visitors’ experience in zoos involve not only exhibit features, but also animal visibility and behavioural activity. Most of the studies on the impact of environmental enrichment in visitors, though, did not control for animal visibility and behaviour. To better comprehend visitor perception of enrichment, it is necessary to have more studies that control for animal behaviour and visibility [28].Since cultural background has been shown to have an effect on attitudes towards animals and animal welfare [33], visitor experience regarding the presence of environmental enrichment might also differ between human societies. To our knowledge, all the studies that assessed visitor attitudes towards enrichment or how enrichment influences visitors’ experiences have been conducted in the United Kingdom [29], the United States of America [21,27,28,31] or China [14].The aim of this study was to assess visitor attitudes towards enrichment appearance on a walk-through enclosure for ring-tailed lemurs (Lemur catta) in an European zoo. Our goal was also to determine if the visibility and behaviour of the lemurs made a difference on visitor attitudes towards the enrichment devices. We expected that visitor attitudes did not differ between naturalistic and artificial looking enrichment [21,27,31], but that lemur behaviour had an influence instead [32].2. Materials and Methods2.1. Study SubjectsThis study was conducted in the ring-tailed lemur enclosure in Zoo Planckendael (Belgium) for 10 days in April 2019. The enclosure consisted of a walk-through outdoor enclosure, a visible indoor enclosure (but inaccessible to the animals at the time of the study) and an off-exhibit indoor enclosure, which was not visible from the visitor side. The outdoor enclosure had a natural looking design that included vegetation, grass, climbing structures (including trees), and a pond. Visitors were allowed to walk on a designated path through the outdoor enclosure, where six ring-tailed lemurs roamed (Table 1). It was forbidden for visitors to touch or feed the lemurs. The outdoor enclosure was 545 m2 and the indoor enclosure was 48 m2.In total, 371 zoo visitors participated in the study; 174 visitors participated on days with naturalistic looking enrichment and 197 visitors participated on days with artificial looking enrichment. The information related with the visitors that were interviewed can be found in Table 2.2.2. Visitor Attitudes QuestionnaireTo collect visitor attitudes, a questionnaire was developed, consisting of two parts. The first part included 14 attitude statements and one multiple-choice question on the behaviour of the lemurs. The participant had to indicate for each of the 14 statements on a 7-point Likert scale to what degree they agreed with the statement, ranging from ‘Completely disagree’ (score = 1) to ‘Completely agree’ (score = 7), with a neutral midpoint at score 4. In the multiple-choice question on lemur behaviour, visitors were asked to report whether they had seen any social behaviours, feeding behaviours, inactivity and/or activity. This question was included to collect data on the visitor perception on the lemur behaviour. The second part of the questionnaire focused on environmental enrichment and included a short text explaining the purpose of enrichment and five supporting images of examples of enrichment items in other species. After reading this text, participants were asked to score three enrichment-related statements using the 7-point Likert scale. In this section, the open-ended question ‘Do you see any enrichment items in this enclosure?’ was also included to ensure that researchers and visitors were considering the same objects as enrichment items.To avoid order-effects [34], three versions of the questionnaire were made by varying the order of the statements. For each participant, the starting and ending times were collected to determine the questionnaire interval. The questionnaire was originally in Dutch and it was randomly offered to adult visitors on a clipboard while they were walking through the enclosure. They were asked by the researcher to read and complete on their own a questionnaire on housing of the ring-tailed lemurs. Visitors were in no way informed about the appearance of the enrichment items, nor was their attention directed to the enrichment items by the researchers. Evaluation of the enrichment items was only requested halfway into the questionnaire, when the researcher had already left the visitors to complete the questionnaire on their own. To avoid interviewer and response biases, visitors were assured in the introductory text that their answers would be anonymous [34].2.3. Enrichment ItemsFive enrichment items were designed, taking into account the ring-tailed lemurs’ needs and the behaviours that we wanted the animals to display when using them. The designs were approved for safety by the zoo veterinarian, the curator and the caretakers’ coordinator. We built two versions of each item: one naturalistic and one artificial looking (Figure 1). Among the materials the zoo directive considered natural looking were wood, rope, burlap, and beige fire hose. Plastic, PVC, bright colours, and steal were considered artificial-looking material by the zoo directive and they were what we used for the design. A description of the items can be found in Table 3.Any time a food-related item was provided to the lemurs, two or more of the same devices were present to avoid feeding competition [10,17,19]. In these items, a third of the lemurs’ daily diet was included to avoid overfeeding. All devices used in this study had been previously offered to the lemurs during another research conducted by our group [35], and were therefore not novel to the animals.2.4. Experimental ConditionsIn order to determine the effect of enrichment appearance, we designed two experimental conditions that consisted of the presence of either naturalistic looking or artificial looking enrichment items in the ring-tailed lemur outdoor enclosure. To avoid habituation to the enrichment items [17,22], three out of five enrichment items were presented each day, varying between combinations of three items, over the 10 days of data collection. The enrichment items on display were always either all naturalistic looking or all artificial looking, never were the two types of enrichment displayed simultaneously. The specific combinations of enrichment items were determined pseudo-randomly beforehand, assuring that all items got an equal amount of time to be displayed.2.5. Ring-Tailed Lemur Behavioural ObservationsWhile the visitors were completing the questionnaire, we observed the behaviour of the ring-tailed lemurs. We used a focal continuous sampling method [5] and a simple ethogram (in Table 4). As we had the time when each visitor started and ended filling the questionnaire, the activity data could afterwards be linked to the individual visitor attitudes. Observational data were collected using ZooMonitor [36].2.6. Statistical AnalysisFirst, to assess if visitor attitudes differed between conditions (having seen naturalistic or artificial looking items), a Mann–Whitney U test was performed on the 17 attitude statements. Shapiro–Wilk tests (using the build-in R function shapiro.test) showed the Likert-scale data to be non-normally distributed (p-values < 0.05). As the normality assumption for performing t-test was not met, alternatively nonparametric Mann–Whitney U tests were run to compare scores of agreement on each of the statements between the two experimental conditions (naturalistic vs. artificial looking enrichment). As a means for correction for multiple testing, we recalculated p-values by the Holm–Bonferroni method (using the build-in R function p.adjust).Second, in order to investigate the relationship between visitor attitudes and lemur behaviour, we linked the questionnaire answers of each visitor to the behaviours of the lemurs during the same questionnaire interval. Two ordinal logistic regression models were performed to assess whether lemur activity level and behaviour influenced visitor attitudes. The first one was based on the data reported by the multiple-choice question by the visitors, and the second one was based on the behavioural observations carried out by the researchers. The multiple-choice question asked the visitors what the lemurs were doing during the time of their visit. They could choose one or more of the following options: (a) Feeding, drinking, foraging; (b) Inactive (sitting, resting, sleeping); (c) Active (walking, running, jumping, climbing; (d) Social behaviour (playing, grooming); and/or (e) Other. For this question, we did not want to draw the visitor’s attention towards the enrichment items, so we did not include ‘Interaction with enrichment’ as a behaviour in the list. A researcher observed the lemur behaviour during the time window where each participant was taking the survey (questionnaire interval). For each lemur, it was observed if they were active, inactive, displaying social behaviours, or out of view (if they were indoors), and if they were using an enrichment device or not. The total number of seconds that the lemurs exhibited behaviour of a certain category—sum of all lemurs—was divided by the total sum of all lemur behaviours (in seconds) exhibited during the questionnaire interval to get the proportion of each behavioural category for each participant. Then, we analysed the relation between the proportions of lemur behaviour of each category and the scores of the participants for each of the Likert-type questions.Finally, for the open question ‘Do you see any enrichment items in this enclosure?’, percentages were calculated for four pre-defined response categories. These categories included: (1) Enrichment items designed for this study and swinging rings permanently present on the enclosure; (2) Other items or features of the enclosure; (3) Sounds and musical background; (4) People (including visitors and caretakers).Mann–Whitney U tests, explorative analyses, linking researcher observed lemur activity to participant questionnaires and ordinal logistic regressions (for what the rms package [37] was used) were performed in R studio (R Development Core Team, 2016).2.7. Ethical StatementThe Royal Zoological Society of Antwerp waived the requirement for formal ethical approval of this study for the following reasons, regarding animal welfare and visitor participation. Concerning animal welfare, this study was conducted in compliance with relevant Belgian and European legislation, and in agreement with international and scientific standards and guidelines. Due to the non-invasive character of the study, and absence of any potential discomfort, our study does not meet the definition of an animal experiment as mentioned in Chapter I, Article 16 of the Belgian ‘Act on the protection and wellbeing of animals’ (Wet van 14 augustus 1986 betreffende de bescherming en het welzijn der dieren gesynchroniseerd met de wet van 27 December 2012). Regarding visitor participation, trained zoo staff and research interns approached adult visitors in the outdoor walk-through exhibit for the ring-tailed lemurs and asked them if they were willing to participate voluntarily in the survey. All subjects who agreed to participate gave their verbal informed consent for inclusion in this study before completing the printed questionnaire. No personal information was collected, and all participants were assured in the introductory text that participation in the questionnaire would be completely anonymous. This study did not require ethical approval since it did not involve any interventions or handling of the animals. The enrichment items used in this study were evaluated and approved by the veterinarian, curator, and caregiver coordinator of Zoo Planckendael.3. Results3.1. Visitor Attitudes towards Naturalistic and Artificial Looking EnrichmentVisitor attitude scores were compared between the naturalistic and the artificial condition for all 17 questionnaire statements (Figure 2). No statistically significant differences were found in any of the statements.3.2. Relationship between Visitor Attitudes and the Lemur Behaviours as Observed by the ResearchersOrdinal logistic regression analyses were conducted to investigate the relation between lemur behaviour—the proportions of time the lemurs were active, inactive, inside (out of view for the participants) and engaging with the enrichment items during the time window where the participant was completing the survey—and participants’ attitude scores on each of the survey questions.When the lemurs showed more active behaviour, visitors tended to agree more with the following statements: ‘I think the enrichment in this enclosure is fit for the lemurs’ and ‘The lemurs look happy’. The time that the lemurs spent interacting with the enrichment was significantly related to higher Likert-scores on the following statement: ‘I think the enrichment in this enclosure is fit for the lemurs’ (Table 5).3.3. Relationship between Visitor Attitudes and the Lemur Behaviours as Perceived by the VisitorsConcerning the results on the behaviours the visitors reported in the multiple-choice question (Table 6), we found that when visitors saw social behaviour (binary: yes or no), they provided significantly lower Likert-scores on the statements ‘The lemurs look stressed’ and ‘I think the enrichment disturbs the view of the enclosure’, and higher scores on the statements: ‘I think the lemurs can behave as they would do in the wild’, ‘I think this enclosure replicates the natural habitat of the ring-tailed lemurs well’, ‘The lemurs look relaxed’ and ‘The lemurs look happy’.When visitors observed the lemurs during feeding, they agreed significantly more with the statement ‘I think the enrichment in this enclosure is fit for the lemurs’ and less with the statement ‘I think the enrichment disturbs the view of the enclosure’.If visitors reported seeing inactivity of the lemurs during their visit, they agreed significantly more with ‘The lemurs look bored’.3.4. Results to the Open QuestionThe analysis of the answers to the open question ‘Do you see any enrichment items in this enclosure?’ showed that out of the 174 visitors that saw naturalistic-looking items, 89% mentioned one of our enrichment items, and 58% mentioned other features of the enclosure. Out of the 197 visitors who saw artificial looking items, 77% mentioned one of our items, and 52% mentioned other features of the enclosure. In the naturalistic-looking condition, none of the visitors considered sounds enriching, and 0.6% (1 visitor) mentioned people as enriching. In the artificial looking condition, 1% (2 visitors) considered sounds and the presence of people enriching for the lemurs.4. DiscussionThe aims of our study were to determine if visitor attitudes were influenced by the appearance of enrichment items and to explore whether the perceived activity levels and the observed activity levels of ring-tailed lemurs had an effect on visitor attitudes towards these items. The results obtained from the open-ended question ensured that researchers and visitors considered the same objects as enrichment items. Above three quarters of the visitors in each condition (89% in the naturalistic looking condition and 77% in the artificial) mentioned at least one of our items as enriching.The analysis of the attitude statements of the 371 completed questionnaires showed that statements regarding the behaviour of the animals and their natural habitat elicited more neutral responses from the visitors, indicating that the visitors did not have a strong opinion about these aspects. As for the other statements, visitors responded generally in favour of the zoo, its management and care: lemurs appeared well taken care of, relaxed and happy, not stressed, and their exhibit and the enrichment in it were rated good-looking.Our results showed that enrichment appearance did not have an influence on visitor attitudes. Visitors generally reacted positively to all enrichment items, naturalistic and artificial alike, and recognised the purpose of both naturalistic and artificial items as being enriching for the animals. The fact that the lemur enclosure in Zoo Planckendael is an immersive experience, where visitors walk through the enclosure following a pathway, might make the visitors more flexible to accept artificial enrichment devices, compared with naturalistic enclosures that lack such constructed elements. Nevertheless, these results are in line with those already published on visitor attitudes towards enrichment performed in other, non-European, zoos [21,27,31].Providing appropriate enrichment items can increase the activity level and natural behaviour of the animals by offering them opportunities to engage in species-appropriate behaviours [19,35,38]. It has been suggested that observing active and engaged animals could promote positive visitor perceptions towards them [28]. Indeed, in our study, we observed that lemur activity levels and behaviour influenced visitor attitudes, more than the appearance of the enrichment items. Visitors scored higher in positive attitudes when they saw the lemurs displaying more active behaviours and interacting with the enrichment devices. In general, when lemurs were active, visitors increasingly reported they looked happy and that the enrichment in the enclosure was fit for the lemurs. When lemurs were inactive, meaning they were asleep or resting, visitors increasingly reported they looked bored. Moreover, when the visitors reported the lemurs displaying social behaviours like grooming or playing together, visitors considered that the animals were behaving as they would in the wild, that their enclosure mimicked their wild environment well, and that the enrichment items did not disturb the view of the enclosure. Visitors further believed that, when they saw social behaviours, the lemurs looked relaxed, happy, and not stressed both with naturalistic and artificial looking enrichment.Importantly, when lemurs were interacting with the enrichment items, visitors provided higher scores for enclosure suitability. Moreover, when seeing the lemurs feeding on a food item from a food-based enrichment device, visitors found the enrichment items to disturb the view of the enclosure less and agreed more on that the enrichment was suited for lemurs.The more active behaviours were displayed by the lemurs as observed by the researchers, the more positive visitor attitudes about the enclosure, the enrichment, and the lemurs’ emotional state. We argue that the functionality, rather than the appearance, of the enrichment items is important, considering its efficiency to elicit the display of more active and species-specific natural behaviours [31], which in turn enhances visitors’ zoo experience. The study of enrichment functionality was not directly addressed by the current study, but preference assessments have been used to test preferred food items in enrichment devices in four species of lemurs [39]. The effects of visitors on the behaviours of ring-tailed [40] and crowned (Eulemur coronatus) [41] lemurs in walk-through exhibits have also been examined. Further research could include the study of the relationship between type of enrichment, lemur behaviour, and the perceptions, attitudes and behaviours of the visitors. Moreover, what similar exhibits or factors related to enrichment within such exhibits have on visitors themselves.Several tools can be used to ensure that zoo visitors understand the importance of environmental enrichment for the animals, including educational signs, animal training and presentations, or having staff or volunteers draw the attention of visitors to the enrichment devices in an enclosure and clarify how the animals use them [1,21], especially in a walk-through enclosure [42] like the one in this study.5. ConclusionsOur results suggest that visitors of the ring-tailed lemur enclosure were not affected by the appearance of the environmental enrichment items. Instead, visitor attitudes were affected to a greater extent by the activity and the behaviours that the animals displayed. Behavioural data on the activity of the ring-tailed lemurs, continuously recorded by the researchers during the experiment, as well as the visitors’ own reports of activity were shown to influence the visitor attitudes. Visitors found that the lemurs were well taken care of, scored the welfare of the lemurs high, and reported that both the enclosure and enrichment items were good-looking and appropriate for the species. We consider that our study supports the idea that zoo management decisions on environmental enrichment programmes should not only be based on the appearance of the enrichment devices, but on how they can stimulate certain species-specific behaviours in the animals. This would ensure better welfare for the animals in zoo settings, since they would benefit from good enrichment items regardless of their appearance, as well as enhance the experience of visitors.

animals : an open access journal from mdpi

10.3390/ani11123338

PMC8698122

The use of crop, fruit, and vegetable processing co-products for animal feeding has been of increasing interest worldwide to minimize feed waste. Additionally, it has a positive impact from an environmental standpoint being a more efficient use of feed resources. However, the use of many co-products has limitations related to poor palatability for animals and the logistical aspects of product delivery. Citrus pulp is a by-product of the citrus industry presenting a nutritional composition that makes it attractive for use as an ingredient in animal feeding. Previous research has shown that it is feasible to utilize citrus pulp in beef cattle rations. The objective of the present study was to evaluate the effect of the inclusion of fresh citrus pulp in the diet of feedlot steers on animal performance and carcass and meat nutritional properties and quality. The findings of the study showed that fresh citrus pulp may be used as an energy source in rations for feedlot steers as it does not affect animal performance or carcass and meat quality, but, rather, has a positive effect on dry matter intake and a better feed to gain ratio.

The use of fruit by-products such as citrus pulp represents a feeding ingredient that deserves to be evaluated as an energy source in animal rations. Thirty-six British breed steers were allotted to one of the three feeding treatments (12 steers/treatment): 0%, 15% and 30% of fresh citrus pulp inclusion in the ration in a randomized complete block design to evaluate animal performance and carcass and meat quality traits. In the present study, the inclusion of fresh citrus pulp up to 30% of the diet did not affect the animal average daily gain (p > 0.05) but steers that were fed the pulp consumed less feed (p < 0.05) and presented a lower feed conversion ratio (p < 0.05) than their counterparts without citrus pulp in their diet. No effect of fresh citrus pulp was observed on carcass and meat quality (p > 0.05). A greater lipophilic antioxidant capacity (p < 0.05) in meat was observed when fresh citrus pulp was offered at 15% of the diet. Fresh citrus pulp used up to 30% as a feed ingredient in feedlot rations does not negatively affect animal performance or meat quality but, rather, has a positive effect on dry matter intake and a better feed conversion ratio.

1. IntroductionIn recent decades, there has been an increasing interest for the reutilization of fruit and vegetable processing co-products in farm animal nutrition due to the social and environmental pressures faced by modern society [1,2,3,4]. In addition, the use of fruit waste for animal feeding would reduce feeding costs incurred by farmers [5]. Citrus pulp represents an important by-product for the Uruguayan citrus juice industry that can be used as a high energy feed in ruminant rations [6]. Residues of citrus juice production are principally composed of water, soluble sugars, fiber, organic acids, amino acids, proteins, minerals, and lipids, as well as flavonoids and vitamins [7]. The performance of steers fed citrus pulp in their ration has been similar to those fed with corn diets [8] being suitable for its inclusion in a balanced diet replacing other energy feeds [9]. The inclusion of citrus pulp into a corn-silage diet has shown no significant changes in the acetic to propionic acid ratio in the rumen and although not considered a roughage, the pulp might contain roughage-like properties that tend to promote higher ruminal pH values [6].On the other hand, using plant by-products containing phytochemicals, such as citrus pulp, can enhance the deposition of bioactive compounds in muscle tissues that delay the oxidative deterioration of color and flavor, and extend the shelf-life of meat [10]. Limited data exist on the effect of citrus pulp on beef meat quality although several studies have been performed in lambs [11,12,13]. Feeding whole citrus pulp to lambs enhances the antioxidant status of muscle more through an increase in the deposition of α-tocopherol than through the effect of flavonoids [14].The hypothesis of the present study was that the inclusion of fresh citrus pulp up to 30% of the ration (dry-matter basis) of steers fed high-concentrate diets could be used as a feed ingredient lowering feeding costs. In lambs, greater levels of citrus pulp of 30% have shown negative impacts such as rumen parakeratosis [6].The objective of the investigation was to evaluate the effect of increasing levels of fresh citrus pulp on steers fed high-concentrate diets, on voluntary dry matter intake, animal performance, ruminal pH, meat quality traits and its antioxidant capacity.2. Materials and MethodsThe experiment was carried out at the Intensive Beef Fattening Unit of Marfrig Group “El Impulso” (33°12′ S and 58°05′ W), Rio Negro, Uruguay. The feeding period was of 104 d starting on 6 August and ending on 18 November 2020. All methods and conditions employed in this study were approved by the Committee for the Ethical Use of Animals of the National Institute Agricultural Research, Uruguay (INIA, Protocol number 2020-12).2.1. Experimental Design and Dietary TreatmentsThirty-six Angus, Hereford, and Angus–Hereford cross steers were blocked by weight and breed and were assigned randomly to one of the three dietary treatments. At the start of the experiment, steers weighed 384 ± 26.4 kg. The steers of the first treatment were offered a control diet without fresh citrus pulp (FCP0), the steers of the second treatment were offered the same diet as those in treatment 1, but replacing corn silage and corn grain with 15% of fresh citrus pulp (FCP15) on a dry-matter basis and the steers of the third treatment were offered the same diet as those in treatment 2, but with 30% of fresh citrus pulp (FCP30) on dry-matter basis (Table 1). The three levels of FCP were defined to study the response curve to FCP inclusion in the diet. The greater proportion of FCP (FCP30) was defined based on previous research [6] where they reported that high levels of citrus pulp in the diet could have negative impacts on animal performance. Fresh citrus pulp was mainly composed of lemon with an energy concentration of 2.54 Mcal/kg DM of ME and 6.94% of CP (DM basis).The experiment was a randomized complete block design with three treatments and 12 replications in which each animal was placed in an individual pen. The experimental pens had a mean size of 50 m2 (2.5 m × 20 m) divided with two strands of electric wire, fed by a solar-panel electrifier. Each pen had its own drinking and feeding trough which were half plastic tanks with a capacity of 100 L. Pens were ballast soil surfaced. Ballast soil consists of compacted broken stone and is used to give stability to the surface to avoid mud.The feed was offered twice per day, and the same diet was offered throughout the experimental period. The first meal at 08:00 h provided 60% of the diet, and the second meal at 14:00 h provided the remaining 40% of the diet. The adaptation period was of 14 d, starting on August 6 and ending on August 20. The different components of the diets were weighed separately with a digital scale and then mixed and offered to each animal.Diet adjustments were based on slick-bunk management [15,16] and weighed rejects. Diet adjustment was performed every day based on observation of the bunk before the first meal was offered in the morning, with a scale from 1 to 5 (1 = slick bunk, 5 = full bunk). One animal of the FCP30 treatment was removed by day 50 of the experimental period because it refused to consume the diet although no health problems were detected.2.2. Animal DeterminationsAt the beginning of the trial, the steers were drenched for fluke (Fasimac®, Elanco, Bogotá, Colombia) and vaccinated for respiratory diseases (HBV1, Pi3, Pasteurella multocida, Histophilus somni and Arcanobacterium pyogenes) with Alliance Respiratoria® (Boehringer Ingelheim, Buenos Aires, Argentina), and for clostridial diseases with Sintoxan® (Merial, Buenos Aires, Argentina).Daily dry matter intake (DMI, dry-matter basis) was calculated from the difference between the quantity offered and refused each day multiplied by its DM percentage. Steers were weighed without fasting at 08:00 h every 14 days before the first meal was provided. Average daily gain (ADG) was calculated as a linear regression of each weight of each individual animal [17]. Feed to gain ratio (F:G) was calculated by dividing the average daily DMI by the average daily gain. Water intake (WI) was calculated as the difference between the remaining water level and the volume needed to reach its original level in the tank. Rumen pH was measured hourly in 12 animals (four animals per treatment in the same block) using ruminal boluses (SmaXtec Premium Bolus, SmaXtec Animal Care GmbH, Graz, Austria) throughout the experimental period. The boluses were administered by a veterinarian from the company country representative, using an oral applicator. Boluses were powered by a lithium battery, which communicated through an internal antenna with an external receiver.2.3. Feed DeterminationsDiets were sampled weekly and monthly composites were sent to the Nutrition Laboratory of INIA La Estanzuela to perform the chemical composition (Table 1). Dry matter content of the diets was determined by drying them in a forced-air oven at 60 °C for 48 h according to Harris [18]. The DM content (%) was calculated using (dry weight/wet weight) × 100. The neutral detergent fiber (NDF) and the acid detergent fiber (ADF) concentrations corrected by ash were determined using an Ankom fiber analyzer (ANKOM-2000I; ANKOM Technology, Macedon, NY, USA) [19]. Ash was determined by the method 942.05 of AOAC [20]. Ether extract was determined using ANKOM-XT15 (ANKOM Technology, Macedon, NY, USA) according to the AOAC method 954.02 [20], and acid detergent lignin (ADL) was determined as described by Goering and Van Soest [21]. Nitrogen (N) was measured using a N analyzer (Kjeltec 8200; FOSS Analytical, Hillerod, Denmark), and crude protein (CP) was calculated as N × 6.25.2.4. Energy Calculation of the DietSamples of each individual component of the diet were chemically analyzed and total digestible nutrients (TDN) were calculated according to Weiss et al. [22]. Conversion from TDN to net energy of maintenance (NEm) or net energy of gain (NEg) was performed using calculation formulas of the NASEM [23].2.5. Carcass and Meat Quality DeterminationsSteers were slaughtered in a commercial meat processing plant where hot carcass weight (HCW) was recorded after slaughter. After 48 h chilling, carcasses were ribbed between the 10th and 11th rib and one hour later marbling score (MARB), ribeye area (REA), and fat thickness (FAT) were measured. Marbling scores were determined using the USDA degrees of marbling [24] and subcutaneous fat thickness was measured over the ribeye between the 10th and 11th ribs in the fourth quartile from the chine along the dorsal side of the Longissimus thoracis (LT) muscle. Ribeye area was traced onto acetate tracing paper and later the area was measured with a software (Foxit Software, Fremont, CA, USA). Instrumental meat color (CIE L*: lightness, a*: redness and b*: yellowness) was also measured on each left half-carcass side in triplicate with a Minolta colorimeter CR-400 (Konica Minolta Sensing Inc., Osaka, Japan) using a C illuminant, a 2° standard observer angle, and an 8 mm aperture size, and calibrated with a white tile before use.At the deboning room, a 5 cm sample was removed from the LT muscle (from the 11th rib following caudal direction) of each left half-carcass. The samples were vacuum-packaged and transported to the Meat Laboratory of INIA Tacuarembó. Each meat sample was divided into two steaks of 2.5 cm thickness. In one steak, subcutaneous fat was removed and cut into small pieces to be subsequently frozen at −80 °C. It was later pulverized using a Robot Coupe R2 (Robot Coupe®, Montceau-les-Mines, France). Immediately following homogenization, each sample was packed in individual sterile whirl-pack bags (Nasco, Fort Atkinson, WI, USA) and placed into a −80 °C freezer until micronutrients and antioxidant capacity determinations were performed.The other steak was aged at 0–2 °C for 5 days. After aging, instrumental lean color (CIE L*: lightness, a*: redness and b*: yellowness) was measured on each steak in triplicate after 40 min blooming with a colorimeter (as previously described). Subsequently, Warner-Bratzler shear force (WBSF; model D2000- WB, G-R Electric Manufacturing Co, Manhattan, KS, USA) was assessed according to the American Meat Science Association guidelines [25]. Steaks were weighed before and after cooking and cooking losses were calculated as ((weight of raw steak − weight of cooked steak)/weight of raw steak) × 100. Steaks were cooked in a preheated clam-shell-style grill (GRP100 The Next Grilleration, Spectrum Brands, Inc., Miami, FL, USA) until the internal temperature reached 71 °C. After cooking, six cores of 1.27 cm diameter were removed from each steak parallel to the longitudinal orientation of muscle fibers. Individual shear force values were averaged to assign a mean peak WBSF value to each sample. Steaks were weighed before and after cooking and the cooking losses were calculated as a percentage of the weight of the raw meat.2.6. Micronutrient AnalysisThe extraction procedures for micronutrients and antioxidant capacity determinations were performed in raw LT lyophilized samples. Tocopherol, retinoid, and carotenoid derivatives were analyzed following the procedures described by Xu [26] and Bertolín et al. [27], with modifications. Briefly, 0.1 g of dry samples was vortexed, sonicated (10 min, 20 °C) and extracted overnight at room temperature with 1 mL 10% potassium hydroxide (ethanol:water, 50:50). After the addition of 1 mL of 10 μg/mL butylated hydroxytoluene in hexane:ethyl acetate (9:1), the tubes were vortexed for 15 min and centrifuged at 12,500 rpm for 5 min. The extraction procedure was repeated twice and the upper layers were collected and evaporated under vacuum in a CentriVap™ concentrator (Labconco Co., Kansas City, MO, USA) at 30 °C for 30 min. The dry residue was dissolved in 0.5 mL of methanol, vortexed for 1 min and filtered (0.22 μm) into an amber vial. Finally, 50 μL were injected on a HPLC Prominence Modular System (Shimadzu Co., Kyoto, Japan) equipped with a Nucleosil C18 column (5 μm; 250 × 4.6 mm), a diode array detector (PDA, SPD-M20A), and a fluorescence (RF-10AXL, Shimadzu, Torrance, CA, USA) detectors, controlled by the LabSolution Software® (Shimadzu, Torrance, CA, USA). The mobile phase, with a flow rate of 1.5 mL.min−1, consisted of acetonitrile (solvent A), methanol (solvent B), and water (solvent C). The HPLC flow gradient started with isocratic 60:35:5 (A:B:C) up to 5 min, 2 min to 60:40:0, 3 min at 60:40:0, 10 min to 22:78:0, 5 min at 22:78:0, and then 10 min to initial condition. The temperature of the autosampler and the column were adjusted at 15 °C and 35 °C, respectively. Alpha tocopherol was detected by fluorescence set at 295 nm for excitation wavelength and 330 nm for the emission wavelength. Retinol and derivatives were determined by UV-vis absorbance at 325 nm and lutein, β-carotene, and carotenoids at 450 nm. The analytes were identified by comparison of the retention times and spectral analysis with those of the pure standards and data reported in literature. Data analysis was performed by applying LabSolution Shimadzu Software® (Shimadzu, Torrance, CA, USA). Alpha tocopherol, retinoid, and carotenoid concentrations were expressed as μg α-tocopherol/g, μg retinol/g, and μg β-carotene/g of muscle (wet basis), respectively.2.7. Antioxidant Capacity DeterminationsFor 1,1-diphenyl-2-picrylhydrazyl free radical (DPPH) and hydrophilic oxygen radical absorbance capacity (ORAC hydrophilic) assays, 0.1 g of lyophilized samples was mixed with 1 mL of methanol:water (80:20) solution, homogenized, and placed in an ultrasonic bath for 10 min at 20 °C, and then stored for 4 h at 20 °C avoiding light exposure. After that, samples were centrifuged at 12,500 rpm for 5 min to obtain a methanolic extract. All spectrophotometric data were obtained using a multi-mode microplate reader Synergy H1®, with two automatic reagent dispensers (BioTek Instruments Inc., Winooski, VT, USA). The quantification of free-radical-scavenging of DPPH (2,2-diphenyl-1-picrylhydrazyl) was performed mixing 15 μL of appropriated diluted extract with 235 μL DPPH solution (125 mM). The mixture was shaken and incubated for 24 h at 20 °C in dark conditions. The reduction of light absorption was measured at 517 nm and quantified using a Trolox standard curve. The results were expressed as μmol Trolox equivalents (TE) per 100 g of muscle (wet basis).The ORAC assay was performed based on the methodology described by Wu et al. [28] with some modifications. For the hydrophilic ORAC assay, an aliquot (25 μL) of the diluted methanolic extracts or control (gallic acid) or standard solutions of Trolox (0 to 60 mM) was transferred to 96-well microplates. In the lipophilic ORAC assay 0.1 g of dry samples was mixed with 2.0 mL of hexane, vortexed, and sonicated for 10 min at 20 °C. Then, the hexane layer was removed and evaporated for 20 min at 30 °C in the vacuum CentriVap concentrator. The residue was dissolved in 50 μL of acetone and then diluted with 150 μL of 7% MCD solution (methyl-β-cyclodextrin, in acetone:water 50:50). The MCD solution (7%) was used for sample dilution, as for Trolox standards (0, 10, 20, 30, 40, 50, and 60 mM). Both ORAC analyses followed the same kinetic procedure. Each plate was placed in the microplate reader and 150 μL of fluorescein solution (0.008 μM) prepared with 75 mM phosphate buffer was added using the automatic dispenser. After shaking and incubation at 37 °C for 30 min, 25 μL AAPH (153 mM) freshly prepared with 75 mM phosphate buffer was added to each well. The plate was shaken, and the fluorescence was monitored every minute for 60 min. The fluorescence wavelengths were set at 485 nm for excitation and 528 nm for emission. The results were estimated based on the standard curve of Trolox concentrations and the areas under the fluorescence decay curves using Biotek Gen 5™ version 3.09.07 (BioTek Instruments, Inc., Winooski, VT, USA) coupled to the microplate reader. The ORAC activity was expressed as μmol Trolox equivalents (TE) per 100 g of muscle (wet basis).2.8. Statistical AnalysisThe experiment was analyzed as a randomized complete block design (RCBD) with feeding treatment (FCP0, FCP15, FCP30) as a fixed effect and the block was included as random using the PROC MIXED procedure of the SAS (SAS Institute, Cary, NC, USA, version 9.4). Ruminal pH was analyzed as repeated measures over time and the autoregressive (AR (1)) covariance structure was used based on the smallest value of the Akaike’s information criterion when compared to compound symmetry and unstructured covariances structures. Characteristics measured on the carcass were analyzed using covariates. Slaughter weight (SW) was used as a covariate to analyze HCW, and HCW was used as a covariate to analyze MARB, REA, and FAT. Homogeneity of variance and normality for all data was evaluated using the studentized residuals plots. Kenward–Roger approximation was used to calculate denominator degrees of freedom for different covariance structures for adjustment of the F-statistic. After ANOVA, least squares means were calculated for treatment comparisons with a significance level of α = 0.05, using the PDIFF option of LSMEANS, when F-tests were significant (p < 0.05). Orthogonal polynomial contrasts were used to determine the linear and quadratic effects of increasing citrus pulp levels on the response variables with a significance level of α = 0.05. Contrast of fresh citrus pulp (FCP15 + FCP30) treatments vs. FCP0 was also performed.3. Results3.1. Animal PerformanceSteers did not differ (p > 0.05) among treatments on initial and final weight or on ADG (Table 2). Although animal performance was not affected by the inclusion of FCP in the diet, steers from FCP30 consumed less feed (p < 0.05) and were more efficient (p < 0.05) at converting feed nutrients into increased body mass than those animals from FCP0 and FCP15. Increasing levels of FCP in the diet linearly decreased (p < 0.05) dry matter intake and F:G. Water intake among treatments was similar, however, steers’ water intake per kg of DM consumed increased linearly (p < 0.05) as the level of FCP in the diet increased.3.2. Ruminal pHRuminal pH values increased linearly (p < 0.05) as the level of FCP increased in the diet from 6:00 to 8:00 h. The first meal was offered to the steers at 8:00 h while the second was offered at 14:00 h. Subsequently, 3 h after the first meal (i.e., from 11:00 h) there was a quadratic effect (p < 0.05) of increased FCP levels for 12 consecutive hours (Figure 1). Furthermore, from 11:00 to 13:000 and from 15:00 to 17:00 h the ruminal pH of the steers from FCP15 and FCP30 treatments was greater (p < 0.05) than those from FCP0.3.3. Carcass and Meat QualityTraits measured on the carcass such as HCW, MARB, REA, or FAT were not affected (p > 0.05) by dietary treatments (Table 3). There were no significant differences (p > 0.05) among treatments on WBSF, cooking losses, or instrumental lean color.When we analyzed the relationship between the response variables and the dietary treatments through orthogonal polynomial contrasts, linear and quadratic trends were not significant (p > 0.05) and neither were the contrast between FCP15 or FCP30 and FCP0 (without citrus pulp) for any of the variables.3.4. Micronutrients and Meat Antioxidant CapacityMeat from steers fed with FCP30 presented greater (p < 0.05) content of α-tocopherol compared to the other two treatments. In addition, meat from FCP15 had a greater (p < 0.05) α-tocopherol content than FCP0 (Table 4). No differences (p > 0.05) were observed in retinol content among dietary treatments although FCP0 showed a lower (p < 0.05) retinoid content than FCP30. It is important to note that α-tocopherol and retinoid contents increased linearly (p < 0.05) as the inclusion level of FCP in the diet increased. A greater (p < 0.05) content of β-carotene was observed on FCP15 compared to FCP30 although no differences (p > 0.05) were found on total carotenoid content among treatments.Regarding antioxidant activity determinations, no differences (p > 0.05) were found in DPPH or hydrophilic ORAC values. Nevertheless, a greater (p < 0.05) lipophilic ORAC value was observed on meat from FCP15 than the other two treatments. It is worth noting that lipophilic ORAC values showed a quadratic response (p < 0.05) as the levels of FCP into the diet increased. Furthermore, we observed a greater (p < 0.05) lipophilic ORAC value in meat from treatments that included FCP (FCP15 or FCP30) in the diet compared to FCP0.4. DiscussionCitrus pulp was included in the ration to formulate similar diets which would explain why no differences were observed in ADG among treatments. It is important to note that in our experiment steam flake was kept at the same level through the feeding treatments and corn silage and ground corn were the feed components replaced by FCP. Lenehan et al. [29] evaluated the effects of replacing rolled barley with citrus pulp and no differences were observed on ADG between dietary treatments of young growing cattle offered grass silage ad libitum. However, Gouvea et al. [30] reported an increase in ADG in Nellore bulls when they replaced ground corn—that has less energetic concentration than steam flake—with pelleted citrus pulp. Therefore, if the inclusion of FCP in the diet does not change the energetic concentration of the diet it would be expected to have no effect on ADG.Regarding lower DMI and the better F:G of the steers with increasing levels FCP in the diet, we speculate that the rate of passage might slow down and thus feed digestibility could be increased to some extent compensating for lower FI. In our study, no differences were observed on WI among treatments despite that the inclusion of FCP that could meet part of water requirements of ruminants [6]. Nevertheless, the ratio WI:DMI was greater with increasing FCP levels which could be explained by the high bulk density of citrus pulp decreasing the DMI due to rumen fill [6].Even though large amounts of pectin are present in citrus pulp that are rapidly and extensively degraded in the rumen, they yield little lactate causing less decline of ruminal pH [6]. The diets in our experiment were not particularly high in energy concentration and pH was never below 6.6 although differences were observed among treatments. Indeed, there was a quadratic effect of increased FCP levels after 3 h of the first meal and for 12 consecutive hours. It has been reported that there is no effect on ruminal pH when substituting levels of ground corn with pelleted citrus pulp [30]. On the other hand, previous research has found a reduction of ruminal pH in Jersey steers when dried citrus pulp in the ration reached to 67 and 100% replacement of corn silage [31]. In addition, Lenehan et al. [29] reported a decrease of ruminal pH when substituting barley (pH = 6.79) with citrus pulp (pH = 6.64). In our experiment, the reduction of corn grain and corn silage levels also induced a lower pH in the FCP30 treatment. However, pH values recorded hourly were far above the limit where subacute acidosis would occur (pH = 5.6) [32,33].As with ADG, carcass traits such as HCW, REA, FAT, and MARB did not differ among treatments. These findings agree with previous research where no differences were observed in carcass traits of Nellore bulls when they compared four different dietary energy sources including citrus pulp [34]. In another study, researchers did not observe differences in carcass characteristics of heifers when evaluating two levels of ruminal undegradable protein and two levels of citrus pulp in the ration [35].Oxidation of meat lipid and protein fractions has been reported as the main non-microbial cause of its quality deterioration during aging [36]. Tenderness is the most important trait affecting beef palatability and hence consumer acceptance [37,38,39]. In our study, WBSF values of meat aged for 5 d did not differ among feeding treatments which agrees with previous research [34,35].Meat color is the most important characteristic that determines a consumer’s purchase decision [40]. Lipid and myoglobin oxidation processes in meat seem to be linked and the latter and are associated with lean discoloration due to the conversion of oxymyoglobin to metmyoglobin [41]. The concentration and activity of many endogenous skeletal muscle antioxidants can be influenced by diet [42,43]. An effect of dietary citrus pulp on the delay of lipid oxidation in lamb meat has been reported, but its effect on color stability has been unclear [12]. In the present study, instrumental color values (L*, a*, b*) of meat aged for 0 d and 5 d were not affected by the inclusion of FCP in the diet probably because the inclusion levels were lower than those required to affect meat color. Previous studies evaluating instrumental meat color (L*, a*, b*) in beef cattle observed greater L* values [44] or no differences [45] due to the inclusion of citrus pulp in the diet.Salami et al. [45] studied the dietary effect of dried citrus pulp on α-tocopherol content on meat and they reported no differences among feeding treatments in which barley was replaced by 0%, 40%, and 80% of dried citrus pulp. On the other hand, it was observed that feeding lambs with a diet containing 35% of dried citrus pulp resulted in a threefold greater concentration of α-tocopherol in muscle [14]. In addition, it has been reported that meat from Angus steers that were fed with a dried citrus pulp (150 g/kg DM) diet for 90 days presented a threefold greater concentration of α-tocopherol than the control diet [44]. Our results showed that meat from FCP30 treatment had three times more α-tocopherol content than FCP0 achieving the suggested threshold concentration (3 to 3.5 µg/g muscle) to delay lipid and pigment oxidation [46,47,48]. Profile analyses for identification of other tocopherols and tocotrienols in further investigations could also contribute to corroborating its antioxidant bioactivity. Retinol content observed agrees with the levels previously reported [49], but its concentration did not differ among treatments. According to the extraction and analysis conditions, this retinoid would correspond to retinyl palmitate, although it is not possible to validate the identification. Previous studies in muscle tissue demonstrated that after absorption, retinoids are stored mainly as retinyl palmitate [50]. Other pro-vitamin-A compounds such as carotenoids and β-carotenes follow the same metabolic pathways. Research conducted in beef cattle showed that β-carotene is essentially the only carotenoid absorbed from the intestine [51]. It has been reported that β-carotene plays a role as an antioxidant under low accumulation of reactive substances, but its function may change to pro-oxidant, or it may be degraded, under higher accumulation [52]. Research has shown that β-carotene levels found in beef muscle are very dependable on the diet composition [51,53,54]. It has been reported that there is a wide range of β-carotene concentrations from 0.06 µg/g in steers under grain-based diets [54] to 2.00 µg/g in bulls fed with corn silage supplemented with enriched n-3 concentrate [52]. In our experiment, β-carotene concentrations in all treatments were within the mentioned range but closer to the highest concentration. A possible explanation might be that corn silage, which contains greater levels of β-carotene than forage hay, was used as a feed ingredient [55]. In addition, the FCP used was of lemon which has a greater concentration of β-carotene than orange pulp [56]. We hypothesize that a better combination of corn silage and FCP would explain the greater levels of β-carotene in FCP15 than in FCP30 treatment although no differences were observed in the carotenoid contents among treatments. Unfortunately, we were unable to know the concentrations of β-carotene and carotenoids in the diet. More research would be needed to understand the effect of the citrus pulp as a feed ingredient including its carotenoid profile and the mechanisms involved in its deposition in muscle tissue.Values of the three antioxidant activity analyses were in the range of those reported in a previous study, considering that, in our experiment, we presented the values per 100 g of muscle [28]. A greater value of hydrophilic ORAC was found in FCP0 than in FCP15 or FCP30 treatments. Phenolic compounds and amino acids, particularly sulfur-containing amino acids, dipeptides such as carnosine and anserine, and antioxidant enzyme activity, among others, are responsible for the hydrophilic antioxidant activity [57,58]. On the other hand, antioxidants such as tocopherols, retinoids, and carotenoids, influence the lipophilic ORAC antioxidant activity and the FCP15 treatment showed a greater deposition of lipophilic antioxidant compounds into animal muscle. Beef is a complex matrix and there does not exist a single method that can fully detect its antioxidant status [28].The inclusion of FCP into feeding rations would provide feasible opportunities to enhance the antioxidant activity and nutritional quality of beef although considerable variation in the chemical composition of the diet and citrus pulp can affect its effect.5. ConclusionsWe conclude that FCP may be used as a feed ingredient in rations for feedlot steers, not affecting animal performance but with a positive effect on DMI and better F:G. Carcass and meat quality were not affected by dietary treatments although the content of some micronutrients and the antioxidant capacity was increased when FCP was added in the diet. However, it is necessary to resolve some logistical aspects associated to storage and delivery of FCP to include this by-product as an animal feed ingredient.

animals : an open access journal from mdpi

10.3390/ani11061634

PMC8227967

Porcine circovirus 3 (PCV-3) was first identified in pigs in the USA and was subsequently detected in several other countries, including Brazil. PCV-3 can be associated with diseases in pigs. To date, there are only a few reports of PCV-3 in wild boars worldwide. This study aimed to investigate the presence of PCV-3 in wild boars in Paraná state, Brazil. The results revealed that PCV-3 was present in the serum and lungs of the sampled boars. The complete genome of the PCV-3a strain was determined and compared with other PCV-3 strains around the world. Phylogenetic analysis has shown a close relationship to the strains already described in domestic and wild pigs. At this moment, there is no evidence that PCV-3 causes disease in wild boars. However, the monitoring of circulation of PCV-3 in wild boars is important for pig industry biosecurity because these animals share pathogens with domestic pigs.

Porcine circovirus 3 (PCV-3) was identified in domestic pigs worldwide. Although PCV-3 has also been detected in wild boars, information regarding its circulation in this free-living animal species is scarce. To investigate PCV-3 occurrence in free-living wild boars in Brazil, 70 serum samples collected between January 2017 and June 2019 in Paraná state, Brazil were analyzed by PCR assay. Amplicons measuring 330 bp in length were amplified in seven (10.0%) of the serum samples and confirmed to be PCV3-specific by nucleotide (nt) sequencing. As the amplified products from the serum samples yielded only intermediate levels of viral DNA, lung samples from the seven PCR-positive wild boars were also evaluated by PCR. Of these samples, five lung samples were positive and provided high levels of viral DNA. The three lung samples that presented the highest levels of viral DNA were selected for amplification and sequencing of the whole PCV-3 genome. The three full-length sequences obtained were grouped in PCV-3 clade “a”, and the sequences exhibited 100% nucleotide similarity among them. The PCV-3 field strains of this study showed nucleotide and amino acid similarities of 98.5–99.8% and 98.8–100%, respectively, with whole-genome PCV-3 sequences from around the world.

1. IntroductionPorcine circovirus 3 (PCV-3) belongs to the genus Circovirus and was recently identified in the USA through metagenomic analysis [1,2]. Subsequently, PCV-3 has been reported in several countries of South America, Europe, and Asia [3,4,5,6,7,8,9,10]. Retrospective studies revealed PCV-3 circulation since 1993 in Sweden [11], 1996 in China [12] and Spain [5], and 2006 in Brazil [13].The PCV-3 genome consists of 1999–2001 nucleotides (nt) of circular, single-stranded DNA featuring two major open reading frames (ORFs). ORF1 encodes the replicase protein (Rep), which is composed of 296–297 amino acids (aa); this ORF is the most conserved region of the genome and shares 55% aa identity with the Rep of porcine circovirus 2 (PCV-2) [1]. ORF2 is located on the negative strand and encodes the capsid protein (Cap), the only constituent of the viral capsid; it is composed of 214 aa, sharing approximately 26–37% identity with the PCV-2 Cap protein [1,2]. ORF3 encodes a putative 231-aa protein, and its function has not been elucidated [1].The PCV-3 sequences available in GenBank have high nucleotide identity between strains [14]. The evolutionary analysis in phylogenetic studies indicate the presence of a common ancestor dated approximately 1967 [15]. Considering a maximum genetic distance of 3% within the complete genome and a bootstrap support higher than 90%, Franzo and colleagues [16] suggest only two clades that can be defined as genotypes. Specifically, PCV-3a forms clade 1, while PCV-3b form clade 2. To date, only two strains are included in clade 2 (GenBank access numbers MG372488 and MG372490).PCV-3 has been detected in symptomatic [1,2] and asymptomatic pigs [14] and in other animal species, including dogs [17], cattle [18] and wild hosts [19]. The PCV-3 prevalence in domestic pigs varies from 6.5 to 68.6% [4,5,6,10,12,14], while in wild boars, it varies from 9.1 to 57.1% and which, due to their habits, may contribute to the spread of the virus [13,20,21,22]. PCV-3 infection in wild boars has been reported in Germany [20], Italy [21], Spain [22], and recently, in Brazil [13].Brazil is the fourth-largest pork producer and exporter worldwide, and the state of Paraná is the second-largest pork producer in Brazil, accounting for 19.8% of total pork meat production in 2019 [23]. The total population of free-living wild boars in Brazil is unknown, but sightings are common in crop fields and near livestock farms in various Brazilian regions, including Campos Gerais, Paraná state [24]. This study attempted to investigate PCV-3 occurrence in free-living wild boars in Campos Gerais, Paraná state, and to genetically characterize the PCV-3 strains detected in this animal species.2. Materials and Methods2.1. Sample CollectionFrom January 2017 to January 2019, 70 free-living wild boars were harvested in the Campos Gerais region of the state of Paraná; specifically, 14 juvenile females, 14 juvenile males, 31 adult females, and 11 adult males were obtained. The classification of the animals into juvenile and adult animals was performed according to Hebeisen et al. (2008) [25].Hunting was performed by exotic wildlife controller agents who were authorized by the Brazilian Institute of Environment and Renewable Natural Resources (IBAMA) according to IN 03/2013 [26], registered in the Federal Technical Register of Potentially Pollutive Activity (CTF/APP), and closely monitored by the Brazilian Army. Paired serum and lung samples were collected from all 70 free-living wild boars and stored at −80 °C. Sera were used to assess the frequency of PCV-3 infection in the studied population, as is commonly carried out in pioneering studies in wild boars [21,22].After screening of serum samples, the lungs of the animals that had positive serum samples were evaluated for the presence of PCV-3 following the same methodology. This tissue was selected because it is a replication site of the virus and a high viral load can be found [22].2.2. DNA Extraction and PCRViral DNA was extracted from serum samples using a DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. PCR assays for PCV-3 diagnosis were performed using a pair of primers (5′-CCA CAG AAG GCG CTA TGT C-3′ and 5′-CCG CAT AAG GGT CGT CTT G-3′) that amplify a 330-bp fragment of the capsid gene [1] in 25-μL final reaction volume. The amplified products were analyzed by electrophoresis on a 1% agarose-TBE gel and stained with ethidium bromide.2.3. Genome SequencingComplete genome sequencing of three PCV-3 PCR-positive lung samples was performed using primers 5′-CAC CGT GTG AGT GGA TAT AC C-3′, 5′-GTC GTC TTG GAG CCA AGT G-3′, 5′-TGT TGT ACC GGA GGA GTG-3′, and 5′-GAA GTT GCG GAG AAG ATG-3′, described by Palinski et al. (2017) [1] and a primer with degenerate 3′ end GCCGAC-TAATGCGTAGTCNNNNNNNNN described by Franzo et al. (2018) [6]. The selection of the three PCV-3 samples for sequencing was based on viral load and on wild boar geographic location.The amplicons were purified using the PureLink® Quick Gel Extraction Kit (Invitrogen, Carlsbad, CA, USA), quantified with a Qubit™ Fluorometer (Invitrogen™ Life Technologies, Eugene, OR, USA), and analyzed by electrophoresis on a 2% agarose gel. The ABI3500 Genetic Analyzer and BigDye™ Terminator v3.1 A Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) was used for sequencing, which was performed in both directions with the forward and reverse primers employed in the PCR assay. Quality assessment and sequence analyses were determined by Phred [27] and Lucy [28] software. The sequences were assembled using Cap3 [27] to generate consensus sequences.2.4. Genome AnalysesComplete PCV-3 genome sequences were accessed on GenBank. The misaligned strings were excluded from the alignment. Therefore, only one sequence was selected as representative of all identical sequences. The construction of the tree comprised complete PCV-3 sequences from wild boars and domestic pigs from different continents and recent and contemporary strains. As reported by Franzo and colleagues [16], two representative sequences of the PCV-3b clade were also included.Phylogenetic analysis of the complete genome sequences of PCV-3 was performed using the neighbor-joining (NJ) method in MEGA 6.0 software [29]. Bootstrap values were determined with 1000 replicates, and the evolutionary distances were computed using the Tamura 3 parameter model [29]. The genome sequences were compared with other PCV-3 sequences that are available in GenBank.The complete PCV-3 sequences of this study and PCV-3 sequences from wild boars and domestic pigs from different continents and recent and contemporary strains were analyzed using BioEdit version 7.2.6.1 [30].3. ResultsPCV-3 was detected in 7 (10%) out of the 70 serum samples. PCV-3-positive samples were obtained from adult female boars from Castro and Ponta Grossa Counties. Additionally, five (71%) of the seven lung samples from wild boars with PCV-3-positive serum samples were positive. The complete genomes of the three sequenced PCV-3 Brazilian strains were deposited in GenBank (accession numbers: MT075517, MT075518, and MT075519). The sequences presented 100% nt similarity among themselves and were classified as PCV-3a (Figure 1).The Brazilian PCV-3 wild boar strains showed 99.6% nt similarity with Chilean (MN907812) and Chinese strains (MG897494) and 99.5% similarity with American (KX966193) and Taiwanese strains (MK343155) identified from domestic pigs.The PCV-3 sequences described in this study (MT075517, MT075518, and MT075519) revealed 98.7–99.8% nt and 99.0–100% aa similarity with the Brazilian strains from domestic pigs (MF079254, MF079253, MK645715, MK645718, MK645719, MK645716, and MK645717) (Table 1), 98.5–99.8% nt and 98.8–100% amino acid (aa) sequence similarity with PCV-3 sequences from domestic pigs worldwide, and 98.6–99.2% nt and 98.9–99.5% aa sequence similarity with PCV-3 sequences from free-living wild boars (MH579736; MH579747; MG820624; MH699985).4. DiscussionThe PCV-3 phylogenetic tree (Figure 1) showed that the PCV-3 sequences from this study were more similar to Brazilian, American, Taiwanese, Chinese and Chilean strains obtained from domestic pigs than to other PCV-3 nt sequences obtained from wild boar and available in GenBank. This result suggests that PCV-3 may circulate between Brazilian domestic and feral pigs. In addition, the notable similarity between strains from Brazilian wild boars and those obtained in different countries and years suggests the high genetic stability of PCV-3 field strains.In this study, PCV-3 DNA was detected in 7 (10%) out of the 70 serum samples, suggesting a systemic infection. Previous studies have shown that wild boars are a potential reservoir for PCV-3 infection in wild and domestic pigs [22]. Additionally, according to Klaumann et al. (2018) [31], PCV-3 could be detected over a long time, suggesting that wild boars may exhibit long-lasting infections. Other viruses belonging to the Circovirus genus, such as PCV-2, also produce persistent viremia in pigs [32]. PCV-3 was reported in serum from wild boars in some countries. Italy was the first country to report PCV-3 infection in wild boars, where 33% of serum samples collected from 2014 to 2015 were positive [21]. In Spain, 42.6% of serum samples collected from wild boars from 2004 to 2018 were positive for PCV-3, demonstrating that the virus has been circulating since 2004 [22]. A longitudinal study in Spain, which involved the capture and recapture of wild boars for over a year, detected PCV-3 in serum samples of 52.6% of the evaluated animals [22].The prevalence of PCV-3 DNA in serum samples of free-living wild boars in Brazil was lower (10%; 7/70) than that observed in European studies [21,22]. PCV-3 studies in Brazil are scarce but showed that PCV-3 has been present in domestic pigs since 2006 [4], and in wild boars since 2013 [13]. In the country, PCV-3 was previously detected in serum of sows presenting stillbirths [3] and the prevalence of PCV-3 in serum samples of domestic pigs was higher (26.7%; 41/154) [9] than observed in wild boars in our research. In a retrospective study, PCV-3 was detected in 47.8% (32/67) of different domestic pig samples such as lung, lymph nodes, and spleen; however, the prevalence for each organ was not reported [4]. A recent study by Varela et al. (2020) [13] detected PCV-3 in 36.3% (29/80) of retropharyngeal lymph nodes from wild boars captured from 2013 to 2015 in Rio Grande do Sul state; other samples types and animals age and sex were not evaluated.Our study was the first to investigate the occurrence of PCV-3 in free-living wild boars in a specific Brazilian region (Paraná state, South Brazil), and to describe PCV-3 recovered from the serum and lungs of these species in Brazil. Selection of sample types for PCV-3 detection was based on previous studies that indicated spleen [20,22], tonsil [13,22], liver [22] and lungs [22] to be the most useful tissues for PCV-3 detection. Furthermore, lungs had the highest prevalence of positivity (57.1%) compared to other tissues [22], and they are considered as a target for PCV-3 replication [31]. Although the viral load in the submandibular lymph node was higher, the percentual of positivity in this tissue was lower compared to other tissues [22]. Serum is considered the most appropriate sample for epidemiological studies in wild boars, despite PCV-3 viral load being lower when compared with viral load found in other tissues [22]. So far, there are few published studies of PCV-3 detection in wild boars [13,20,21,22], and 66.6% of them use serum as biological material for initial PCV-3 investigation [21,22].Studies published to date observed a high prevalence of PCV-3 in adult wild boars (47.5%) and a low prevalence in juveniles (8.6%) [22]. Surprisingly, in our study, PCV-3 was only detected in adult free-living wild boars, suggesting less virus circulation in juvenile free-living wild boars. Usually, adult males live alone while adult females are frequently seen in flocks with or without offspring. The lower PCV-3 detection observed in juvenile wild boars suggests a low viremia at the time of sample collection [10], possibly due to the presence of PCV-3 antibodies from maternal colostrum.5. ConclusionsThis report is the first to describe PCV-3a in free-living wild boars in Paraná state, South Brazil. The identity matrix demonstrated a high nt similarity among the three PCV-3 strains from this study and other strain sequences that are available in GenBank. PCV-3 prevalence in wild boars should be evaluated to determine the dynamics of virus evolution in this pig population.

animals : an open access journal from mdpi

10.3390/ani11123407

PMC8697944

Freeze-drying (or lyophilization) is a method to preserve cells and tissues in which frozen material is dried by sublimation of ice. One of the main advantages is that nitrogen and dry ice are no longer required for the storage and shipment of biological material, which can be kept at room temperature or 4 °C, resulting in enormous reductions in costs. Although widely used to preserve biomolecules and macromolecular assemblies, freeze-drying of cells and tissues is currently experimental. Here, we lyophilized sheep ovarian tissue with a novel device named Darya and assessed effects on tissue integrity and gene expression. We show that ovarian tissue survives lyophilization procedures, maintaining its general structure and reacting to the different experimental steps by regulation of specific genes. Our results contribute to the optimization of protocols to freeze-dry ovarian tissues and may find application in programs of animal and human reproductive tissue preservation.

Cryopreservation is routinely used to preserve cells and tissues; however, long time storage brings many inconveniences including the use of liquid nitrogen. Freeze-drying could enable higher shelf-life stability at ambient temperatures and facilitate transport and storage. Currently, the possibility to freeze-dry reproductive tissues maintaining vitality and functions is still under optimization. Here, we lyophilized sheep ovarian tissue with a novel device named Darya and a new vitrification and drying protocol and assessed effects on tissue integrity and gene expression. The evaluation was performed immediately after lyophilization (Lio), after rehydration (LR0h) or after two hours of in vitro culture (IVC; LR2h). The tissue survived lyophilization procedures and maintained its general structure, including intact follicles at different stages of development, however morphological and cytoplasmic modifications were noticed. Lyophilization, rehydration and further IVC increasingly affected RNA integrity and caused progressive morphological alterations. Nevertheless, analysis of a panel of eight genes showed tissue survival and reaction to the different procedures by regulation of specific gene expression. Results show that sheep ovarian tissue can tolerate the applied vitrification and drying protocol and constitute a valid basis for further improvements of the procedures, with the ultimate goal of optimizing tissue viability after rehydration.

1. IntroductionLong-term storage of biological samples in medical applications may be achieved by two different approaches: cryopreservation or lyophilization [1]. Cryopreservation allows preservation of cells, tissues, or any other biological construct by cooling samples to very low temperatures [2]. It has become the most widely used and reliable approach to maintain cell or tissue structure and functional integrity during transportation and storage. Despite being currently considered the gold standard for the storage of active cells, it has some important limitations related to costs, as it requires a continuous and guaranteed supply of liquid nitrogen, regular maintenance, dedicated space, and trained staff; in addition, it greatly complicates sample transport and has a heavy carbon footprint due to the use of liquid nitrogen. Such limitations could be overcome by the use of lyophilization. By drying tissues, samples could be stored and transported at ambient temperatures or at 4 °C. Lyophilization (or freeze-drying or vit-drying) is the process of drying a frozen sample via sublimation of ice (freeze-drying), by transiting ice directly to the gas phase without passing through the intermediate liquid one [3] or by desorption of glass material (vitrified) into gas (Vit-Drying [4]). Lyophilization has a long history of successful use in different areas of medicine and the food industry, however, most applications of freeze-drying or drying of mammalian cells could only preserve the integrity and/or functionality of certain subcellular components [5,6,7,8,9,10,11,12]. For example, freeze-drying of EBV-transformed B-lymphoblastoid cells could stabilize the total RNA in the cells for routine diagnostics [13]. Surface-labeled, freeze-dried lymphocytes can be used as the reference material for counting CD4+ T cells [14,15]. The sperm chromatin structure in freeze-dried spermatozoa is largely intact and can be used to fertilize oocytes by intracytoplasmic sperm injection in a range of species including mouse, rat, cat, dog, rabbit, bull, pig, horse, chimpanzee, giraffe, jaguar, weasel, long-haired rat, and human, with offspring reported in some species [16,17,18,19,20,21,22,23,24,25,26,27,28,29]. Despite being developed many years ago [30], the possibility to freeze-dry cells or tissues maintaining vitality and functions is still under optimization, and this also applies for gonadic tissues. Currently, ovarian tissue is most commonly preserved by cryopreservation. Despite various degrees of success reported in different species, which included morphological, cytological, and molecular alterations observed after cryopreservation with different approaches [31,32,33,34,35,36,37,38], ovarian tissue banking was recently accepted as a fertility-preservation technique by the American Society for Reproductive Medicine and is no longer considered experimental [39]. Conversely, procedures to lyophilize ovarian tissue are still in progress and very limited data were reported to date; to our knowledge, only Lee et al. described successful dry-preservation of cat ovarian tissue, with at least partial follicle survival after rehydration (2019) [40].The aim of this study was to investigate the histological and molecular structure of sheep ovarian tissue after lyophilization with a new freeze-drying device named Darya (FertileSafe, Nes-Ziona, Israel; [41]) and by using a new method of drying following vitrification (Vit-Drying). These preliminary results will contribute to the improvement of the experimental procedures for the lyophilization of reproductive tissues and will pave the way for application in fertility preservation programs, both in the animal and human fields.2. Materials and MethodsAll chemicals in this study were purchased from Sigma-Aldrich S.r.l. (Milan, Italy) unless stated otherwise.2.1. Ethics ApprovalThe ovaries used for in vitro experiments were collected at a local slaughterhouse in Sardinia, Italy, which does not require ethics approval.2.2. Sample CollectionOvaries were collected from regularly slaughtered ewes aged between 2 and 4 years. Immediately after collection, they were placed in a balanced saline solution (Dulbecco saline phosphate buffer, PBS) added with penicillin (50 mg/mL) and streptomycin (50 mg/mL) and maintained at the constant temperature of 27–30 °C. Ovaries were brought to the laboratory within 2 h of recovery and placed in a glass plate to be processed. They were washed twice in PBS and then placed in Dissection Medium (DM) [TCM199 with 25 mM N2 HydroxyethylpiperazineN2 ethan sulfonic Acid (HEPES) to stabilize pH, 50 IU/mL streptomycin, 50 IU/mL penicillin, and 0.005 M NaHCO3 and polyvinyl alcohol (PVA) 0.1% (w/v)] at pH 7.22 ± 0.1 and 4 °C. Ovaries were sectioned with a sterile microblade and tissue sections of 1 mm3 were collected.2.3. Experimental DesignThe sections of ovarian tissue (1 mm3) were immediately subjected to vitrification and drying. After rehydration, tissue sections were cultured in vitro for two hours. Sections of fresh ovarian tissue were included as controls (CTR; Figure 1). Eight animals were used for this study; each experimental group included sections derived from the ovaries of each ewe (eight biological replicates per group), as follows:CTR (n = 8): Sections of fresh ovarian tissue stored for further analysis immediately after dissection.Lio (n = 8): Sections of ovarian tissue stored for further analysis immediately after lyophilization by vitrification procedure (without warming procedure).LR0h (n = 8): Sections of ovarian tissue stored for further analysis after lyophilization and rehydration procedures.LR2h (n = 8): Sections of ovarian tissue stored for further analysis after lyophilization and rehydration procedures followed by 2 h of in vitro culture.Tissue integrity was assessed by histological analysis after lyophilization procedures (LR0h and LR2h) and in fresh controls (CTR). All experimental groups were subjected to analysis of expression of a panel of eight genes.2.4. Tissue Vitrification and DryingTissue sections were exposed to two different solutions (equilibration and vitrification solutions). First, samples were immersed for 25 min in 1 mL of equilibration solution (ES) consisting of TCM-199, 25 mM HEPES, 7.5% ethylene glycol (EG), 7.5% dimethyl sulfoxide (DMSO), and 20% fetal calf serum (FCS). Subsequently, samples were transferred into 1 mL of vitrification solution [VS: TCM-199 with HEPES 25 mM, EG 18%, DMSO 18%, bovine serum albumin (BSA) 0.6%, and trehalose 0.5 M] for 15 min [42]. Samples were immediately subjected to Vit-dry procedures using Darya lyophilizer device (FertileSafe, Nes-Ziona, Israel), which allows to control the condensation temperature and reaches a vacuum pressure of 80 mTorr in less than 10 s. Before use, the Darya device was sterilized in an autoclave.Tissue sections were lyophilized at −50 °C for 20 h, then at −35 °C for two hours, at −25 °C for a further two hours, and, finally, they were directly plunged into liquid nitrogen.2.5. RehydrationRehydration was achieved simultaneously with temperature recovery by sequential exposure of the samples to TCM-199 solutions with 20% FCS and decreasing sucrose concentrations (1 M, 0.5 M and 0.25 M) at 38.6 °C. Samples were kept inside each solution for 5 min. Tissue samples for gene expression analysis were immediately stored in RNALater™ (Qiagen, Hilden, Germany) or transferred to a Petri dish containing culture medium (IVC: TCM-199 with 100 μM cysteamine, 10% FCS, 2.1 g/L sodium bicarbonate, 0.36 mM pyruvate) and stored in RNALater™ after 2 h of in vitro culture. Samples were stored at −80 °C until further processing. Tissue samples for histological analysis were immediately stored in Bouin’s solution.2.6. Gene Expression AnalysisGene expression analysis by real-time PCR was performed and is described according to MIQE guidelines [43] and in line with recent recommendations [44].2.7. Total RNA Isolation and Reverse TranscriptionTissue samples for molecular analysis were plunged into RNALater (ThermoFisher Scientific) immediately after each treatment and stored at −80 °C until RNA isolation. Total RNA was isolated using 1 mL TRIzol reagent (Invitrogen Corporation, Carlsbad, CA, USA) and treated with DNase I (Invitrogen Corporation) according to manufacturer’s protocols. The resulting RNA quantity and purity were spectroscopically checked with NanoDropLite (Fisher Scientific S.A.S., Illkirch Cedex, France), while RNA integrity was evaluated by electrophoresis in a 1% agarose gel in Tris Borate EDTA Buffer (1 µg RNA per sample). Five hundred ng total RNA from each sample were reverse transcribed in a 20 μL reaction with 50 mM Tris HCl (pH 8.3), 75 mM KCl, 3 mM MgCl2, 5 mM DTT, 1 mM dNTPs, 2.5 μM Random Hexamer primers, 20 U of RNaseOUT™, and 100 U of SuperScript™ III RT (all provided by Invitrogen Corporation). Negative control reactions (without the enzyme) were carried out to confirm the absence of genomic DNA contamination. The reaction tubes were incubated at 25 °C for 10 min, at 42 °C for 1 h, and, finally, at 70 °C for 15 min to inactivate the reaction.2.8. Real Time Polymerase Chain ReactionRelative quantification of transcripts was performed by real-time polymerase chain reaction (PCR) in a Rotor-Gene Q 5 plex HRM (Qiagen). The PCR was performed in a 15 μL reaction volume containing 7.5 μL 2× Qiagen PCR Master Mix (Qiagen), 200 nM of each primer (Table 1) and cDNA equivalent to ~20 ng RNA. The PCR protocol consisted of two incubation steps (50 °C for 5 min and 95 °C for 2 min), followed by 40 cycles of amplification program (95 °C for 15 s, gene-specific annealing temperature for 30 s, Table 1), a melting curve program (65–95 °C, starting fluorescence acquisition at 65 °C and taking measurements at 10 s intervals until the temperature reached 95 °C) and finally a cooling step to 4 °C. Fluorescence data were acquired during the elongation step. To minimize handling variation, all samples were analyzed within the same run using a PCR master mix containing all reaction components apart from the sample. The PCR products were analyzed by generating a melting curve to check the specificity and identity of the amplification product. For each primer pair, the efficiency of the PCR reaction was determined by building a standard curve with serial dilutions of a known amount of template, covering at least three orders of magnitude, so that the calibration curve’s linear interval included the interval above and below the abundance of the targets. Only primers achieving an efficiency of reaction between 90 and 110% (3.6 > slope > 3.1) and a coefficient of determination r2 > 0.99 were used for the analysis.Target gene expression was normalized against the geometrical mean of three housekeeping gene expressions: ribosomal protein L19 (RPL19), actin B (ACTB) and succinate dehydrogenase complex flavoprotein, subunit A (SDHA).2.9. HistologySamples of LR0h, LR2h, and CTR experimental groups were fixed in Bouin’s solution for 12 h, paraffin-embedded, 3 µm sectioned, stained in Hematoxylin Eosin (HE) and Masson’s Trichrome, and visualized under an optical microscope. Cortical and medullary tissues were histologically evaluated to identify potential morphological changes affecting follicular structures and the surrounding parenchyma. Digital computer images were recorded with a Nikon (Tokyo, Japan) Ds-fi1 camera.2.10. Statistical AnalysisData were analyzed with GraphPad Prism version 8.0.0 for Windows, GraphPad Software, San Diego, CA, USA. After testing for normality using a Kolmogorov–Smirnov test, gene expression data were analyzed with the General Linear Model analysis of variance (ANOVA), followed by Tukey’s post-hoc comparison (if ANOVA was significant). Differences were considered significant when p < 0.05.3. Results3.1. RNA IntegrityAgarose gel electrophoresis highlighted differences in the integrity of the total RNA isolated from tissues subjected to different treatments (Figure 2A).Total RNA includes different types of RNA, and ribosomal RNA (rRNA) is quantitatively preponderant (about 80%). For this reason, rRNA subunits (28 S, 18 S, and 5.8 S) may be visualized after electrophoretic separation in agar gel. Visualization of two sharp and clear bands, representing the 28 S and 18 S rRNAs, indicate good RNA integrity. Here, electrophoresis showed excellent integrity for CTR samples; the tissues subjected to the lyophilization process only (Lio) maintained fair integrity (it is possible to distinguish the rRNA bands, but there is a slight smear representing partially fragmented RNA). Tissues subjected to lyophilization and rehydration (LR0h) showed increasing degrees of fragmentation and variability between samples, which further increased after two-hour IVC (LR2h; Figure 2A).3.2. Gene ExpressionACTB, RPL19, and SDHA are endogenous reference genes suitable for expression studies in different tissues [42,45,46]. In the present experiment, their expression is not stable across experimental groups but reflects the degree of RNA integrity (Figure 2B).As RNA fragmentation may affect the quantification of gene expression, we compared the expression of target genes before and after normalization against reference gene expression to dissect the variation due to RNA integrity from the variation due to specific gene regulation.In absence of normalization against the reference genes, five of the eight analyzed genes (BAX, CIRBP, OCT4, FSHR, and STAR) show patterns of expression similar to the reference genes and in accordance with RNA integrity (Figure 3): the relative quantification of these transcripts show significantly lower levels in tissues subjected to lyophilization (Lio) and rehydration (LR0h and LR2h). The expression of NLRP5 and HSP90b in the lyophilized samples (Lio) is similar to controls, while a decrease is observed after rehydration and IVC (LR0h and LR2h). Finally, SOD1 transcript abundance shows no variation across the four experimental groups.Normalization of the target gene expression against the reference gene abundance identified which gene was specifically regulated by the treatment in the live/surviving cells. The abundance of BAX, CIRPB, NLRP5, OCT4, and STAR mRNAs did not differ following treatments. On the contrary, HSP90b expression significantly increased after lyophilization (Lio) and remained stable after rehydration and two-hour IVC (LR0h and LR2h). The expression of SOD1 increased after lyophilization (Lio) and after dehydration (LR0h) and maintained the high level after IVC (LR2h). Conversely, FSHR expression showed a decrease approaching significance (p = 0.056; Figure 4) after lyophilization, which was maintained after rehydration and IVC.3.3. HistologyThe histological analysis of CTR, LR0h, and LR2h samples showed progressive morphological alterations both in the follicular and in the interstitial and stromal components.Approximately 80% of primordial and primary and 90 to 100% of secondary follicles were characterized by a multifocal detachment of the granulosa cells from the basement membrane (Figure 5B,B1). Multifocal to coalescing scant eosinophilic irregular cytoplasmic vacuolization and severe fragmentation and condensation of the nuclear chromatin (Figure 5C1), with multifocal intranuclear slightly basophilic vacuolization, were noticed. Diffusely, the medullary and cortical stroma were expanded by weakly basophilic edema.4. DiscussionLyophilization of reproductive tissues is currently experimental. In the context of reproduction, the use of freeze-drying has been successfully applied to mouse spermatozoa, which maintain chromosomal stability [11] and the ability to fertilize oocytes and support normal embryonic development [9]. Conversely, the lyophilization of gonadal tissue, both in male and female, is still being tested. Nevertheless, the many advantages of such cost-effective preservation methods encourage further efforts to optimize protocols and improve the quality of the tissues subjected to such procedure.The present work showed that sheep ovarian tissue subjected to vitrification/lyophilization procedures survives and reacts by regulating the expression of specific genes.Moreover, histological analysis indicated structural alterations due to lyophilization, such as the detachment of the granulosa cells from the basal membrane, cytoplasmic modifications, and disaggregation of the nuclear chromatin. However, histologically, the tissue retained its general microscopic features allowing the identification of the main cell types including intact primordial follicles (Figure 5). Nevertheless, additional studies on a higher number of samples at different time points of lyophilization are ongoing to fully characterize and quantify the modification observed in the different stages of follicle development and on the ovarian tissue.Moreover, histological analysis showed that the tissue maintains its general structure, including intact follicles at different stage of development (Figure 5), with only limited morphological and cytoplasmic modifications.To achieve ovarian tissue lyophilization, we employed a new benchtop device called “Darya”, recently designed to lyophilize cells and tissues, and previously applied to freeze-dry ram sperm [41]. Freeze drying requires the optimization of three main procedures: freezing, drying, and rehydration. Excessive dehydration can cause macromolecular denaturation and reduction in cell volume, with possible irreversible collapse of the cell membrane [47]. A further deleterious effect is the mechanical stress caused by ice formation around the cells, which forces the cells into the extremely limited space of the unfrozen solution [48]. The Darya instrument allows rapid sublimation at low temperatures below zero which, combined with the action of sugar-based cryoprotective solutions such as Trehalose and DMSO, causes reduced osmotic effects and facilitates sample dehydration.At present, tissue degeneration after freeze-dry procedures is non-negligible. A deeper understanding of how cells react to the process, together with the specific molecular mechanisms, is beneficial to improve freeze-dry experimental protocols. In our experimental design, we have attempted to understand the variation in gene expression after lyophilization procedure, after lyophilization and rehydration, and during a 2 h post-rehydration period of in vitro culture, to dissect the molecular responses specific to each experimental step. The selected gene panel comprises germ-cell specific markers (NLRP5, POU5F1 [OCT4]), genes specifically expressed in supporting somatic cells (FSHR and STAR), and genes involved in cell stress response (BAX, SOD1, CIRBP, and HSP90b). OCT4 (POU5F1) encodes a transcription factor specific of germ cells [49], while NLRP5 is a maternal effect gene exclusively expressed in oocytes and early embryos [50]; the receptor of the follicle stimulating hormone (FSHR) and the steroidogenic acute regulatory protein (STAR) are expressed only in granulosa cells; finally, BAX, SOD1, CIRBP, and HSP90b encode molecules involved in the cellular response to different stress conditions [51,52,53,54].The first observation derived from the molecular analysis is that lyophilization procedures cause a partial RNA degradation; dissection of the different experimental steps showed that samples subjected only to lyophilization (Lio) maintained RNA integrity, while after rehydration and subsequent in vitro culture RNA showed significant signs of degradation (Figure 2). These observations suggest that part of the cells did not survive the process and highlight the need of an optimization of these experimental phases to better preserve tissue integrity. In accordance, the analysis of the single genes confirmed a decrease in transcript abundance during the entire lyophilization process, which reflects the decrease in total RNA content (Figure 3).RNA degradation can compromise a proper gene expression analysis. The presence of damaged RNA molecules can prevent adequate quantification, altering primer binding, enzyme activity (reverse transcriptase and DNA polymerase) and molecule elongation during cDNA synthesis or during PCR. These phenomena can cause an imprecise quantification of both target and reference genes. In the present work, the degradation of RNA clearly affected the transcript levels of all genes expressed in the ovarian tissue. To overcome this criticism and carry out an appropriate evaluation, we examined the expression of the target genes with and without normalization against reference genes (Figure 3 and Figure 4). This allowed us to distinguish the transcript variations due to the overall RNA degradation from the variations due to activation or suppression of specific genes.Comparison of the two sets of expression indeed identified a specific response to lyophilization in terms of gene regulation, which indicates that surviving cells precisely up- or down-regulated certain genes following lyophilization, rehydration, or subsequent in vitro culture.The expression of all genes was detected in all experimental groups, indicating the survival of both germ and somatic cells to the entire experimental process. The relative quantification of the reference genes (ACTB, RPL19, and SDHA) reflects the levels of RNA degradation, with a significant decrease between controls and lyophilized samples, and a further decrease following rehydration (Figure 3); such patterns indicate a negative effect of lyophilization and rehydration, which possibly cause death of part of the cells, whose RNA undergoes degradation.The analysis of the target genes in absence of normalization showed patterns in accordance with RNA degradation in four cases: BAX, CIRBP, FSHR, and STAR. OCT4 showed a significant decrease in all treated groups, while NLRP5 and HSP90b showed similar levels of expression between controls and lyophilization (Lio) and a significant decrease following rehydration. On the contrary, SOD1 did not show variations in expression (Figure 3).To identify the genes that responded specifically to lyophilization in terms of up- or down-regulation, we then compared the expression levels after normalization against the reference genes. Five genes showed similar transcript levels among the four experimental groups: BAX, CIRBP, NLRP5, OCT4, and STAR, while FSHR showed a difference close to significance between controls and the three treated groups (Figure 4). These results indicate that the surviving tissue maintained proper function, both in somatic cells (FSHR and STAR), and germlines (NLRP5 and OCT4). The expression patterns of HSP90b and SOD1 clearly show that the tissue that survived the conservation process activated the expression of genes involved in the cellular response to stress conditions (Figure 4). HSP90b mRNA levels increased after lyophilization (Lio) and maintained similar levels following rehydration and subsequent IVC (L0R and L2R). HSP90b is involved in the correct folding of proteins that have become structurally unstable due to exposure to various types of cell stress (heat or cold shock, hyperosmotic stress, or heavy metal toxicity) [51,52,53,54]. Its steady upregulation indicates that exposure to vit-dry procedures exerts negative effects on protein stability. Conversely, SOD1 showed two significant increases following lyophilization (Lio) and rehydration (L0R and L2R; Figure 5). The enzyme superoxide dismutase 1, encoded by SOD1, is responsible for the elimination of free radicals (Reacting Oxygen Species [ROS]) achieved by converting superoxide ions into molecular oxygen and hydrogen peroxide; the increasing SOD1 expression we observed following lyophilization, rehydration and IVC is, therefore, an index of growing oxidative stress (Figure 4).The results of histological analysis indicate structural alterations due to lyophilization, such as the detachment of the granulosa cells from the basal membrane, cytoplasmic modifications and disaggregation of the nuclear chromatin. However, the tissue retained its general structure, and it is possible to distinguish morphologically intact primordial follicles (Figure 5). These histological observations, together with the results of the molecular analysis, suggest that at least part of the tissue was preserved and able to tolerate the lyophilization protocol applied in the present study.Our observations are in partial agreement with the study performed on cat ovarian tissue by Lee et al. [40], who characterized the influence of microwave-assisted dehydration after exposure to trehalose on morphology and viability of living ovarian tissues. While their preservation procedure did not involve cryopreservation, the two protocols share the dehydration and rehydration steps, which may irreparably damage the vital functions of the tissue. Whereas we observed structural alterations, they report proper morphology and DNA integrity of follicles and stromal cells, together with a partially maintained transcriptional activity following 10 min of drying.Despite the reported molecular and structural alterations, these studies on dry-preservation techniques are encouraging and provide foundation for protocol optimization. The abundant literature on ovarian tissue cryopreservation (both slow freezing and vitrification) also describes alteration of tissue integrity in terms of follicular and stromal morphology, cell viability, DNA integrity, and gene expression [31,32,33,34,35,36,37,38]. Nevertheless, extensive research in different species has led to optimization of the techniques and significant improvement of tissue integrity following cryopreservation, with its recent inclusion in clinical practice of fertility preservation [39].5. ConclusionsIn conclusion, the present study provided a preliminary microscopic and molecular evaluation of the effects of lyophilization on ovine ovarian tissue. The analysis of the different steps dissected the specific effects of lyophilization or rehydration of the sample. The results indicate that at least part of the ovarian tissue tolerated the vit-dry preservation protocol and may be useful for further improvements of the procedures, with the ultimate goal of optimizing tissue viability after rehydration.

animals : an open access journal from mdpi

10.3390/ani11071908

PMC8300119

The purpose of the study was to identify SNPs in genes encoding TLR7 and TLR8 in goats of Carpathian breed and analyze their association with the SRLVs provirus concentration. A total of 14 SNPs were detected, 6 SNPs in the TLR7 gene locus and 8 SNPs in the TLR8 gene. These SNPs were located in intron, 3′UTR and 5′UTR regions and within the coding sequences leading to the synonymous mutations. Our results revealed that 9 out 14 identified polymorphisms were associated with the SRLVs proviral concentration. This finding supports a role for genetic variations of TLR7 and TLR8 in SRLVs infection.

Toll-like receptors (TLRs) 7 and 8 are important in single-stranded viral RNA recognition, so genetic variation of these genes may play a role in SRLVs infection and disease progression. Present study aimed to identify SNPs in genes encoding TLR7 and TLR8 in goats of Carpathian breed and analyze their association with the SRLVs provirus concentration as index of disease progression. A total of 14 SNPs were detected, 6 SNPs in the TLR7 gene locus and 8 SNPs in the TLR8 gene. Nine of the 14 identified polymorphisms, 4 in the TLR7 gene and 5 in TLR8 gene, were significantly associated with the SRLVs proviral concentration. These SNPs were located in 3′UTR, 5′UTR and intron sequences as well as in the coding sequences, but they led to silent changes. Homozygous genotypes of three TLR7 SNPs (synonymous variant 1:50703293, 3′UTR variant 1:50701297 and 5′UTR variant 1:50718645) were observed in goats with lower provirus copy number as well as in seronegative animals. The results obtained in this study suggest that SNPs of TLR7/TLR8 genes may induce differential innate immune response towards SRLVs affecting proviral concentration and thereby disease pathogenesis and progression. These findings support a role for genetic variations of TLR7 and TLR8 in SRLVs infection and warrants further studies on the effect of TLR7/TLR8 polymorphisms on SRLVs infection in different populations.

1. IntroductionMaedi visna virus (MVV) and caprine arthritis encephalitis virus (CAEV) also referred as small ruminant lentiviruses (SRLVs) are two related retroviruses which infect sheep and goats. These viruses infect monocytes, macrophages and dendritic cells and despite immune response cause a lifelong infection which can persist for months in latent or subclinical form. The most prevalent clinical signs as an outcome of SRLVs are associated with arthritis, neurological disorders, mastitis, emaciation and pneumonia [1]. Transmission occurs from infected dams to offspring by colostrum/milk consumption and between adults mainly through respiratory secretion [2]. Since no effective vaccines are available, the current control strategies against SRLVs infections are mostly based on the detection and culling of infected animals [3]. What is more, SRLVs infections have become a worldwide problem bringing considerable financial losses in the small ruminant industry [3,4,5]. In Poland, any SRLVs control programs have been never implemented, and thus, infections with SRLVs are quite common. The overall true prevalence at the flock level reached 33.3% and 71.9% in sheep and goats, respectively [6,7].In the course of SRLVs infection, their genome is integrated into the host genome in the form of proviral DNA which load is a factor determining disease prediction. It was shown that high proviral load corresponded to higher lesion score [8,9]. On the other hand, the presence of animals with low SRLVs proviral load may suggest the implication of host factors that may restrict and control viral replication. Such animals which are also referred as long term non-progressor show competent humoral immune response in the absence of virus replication leading to lower SRLVs transmission [10,11,12]. This different reactivity to SRLVs infection may suggest different host immune response ability to control this infection. In SRLVs-infected animals, both innate and adaptive immunity are induced. Moreover, several genes associated with resistance/susceptibility to SRLVs infection and disease outcome have been identified [13,14]. Toll-like receptors (TLRs) are host pattern recognition receptors (PRRs) that play a pivotal role in the innate immune system. PRRs activate the innate immune system through different signaling pathways upon recognition of variety of structurally conserved molecules derived from pathogens known as PAMPs (pathogen-associated molecular patterns). TLRs are type I transmembrane proteins which contain three domains, an ectodomain (ECD) containing leucine-rich repeat (LRR) motifs that mediate PAMPs recognition and a cytoplasmic Toll/interleukin-1 receptor (TIR) domain linked by a single transmembrane (TM) domain [15]. TLR involved in response to viral infection are TLR1-TLR4 and TLR6-TLR9 [16]. In retroviral infections, TLR7 and TLR8, which share a high degree of structural similarity and recognize viral ssRNA (single-stranded RNA) within endosomal compartments, are of particular interest [17]. Activation of these TLRs initiates the MyD88-dependent pathway, mainly in NF-κB activation for inducing the expression of pro-inflammatory cytokines, chemokines, and type I and type III interferons [18,19]. It is well known that cytokines impairing expression is a key point in SRLVs-dependent immunity, pathogenies, and appearance of lesions [20,21]. However, so far, the role of lentivirus-induced TLR signaling has not been widely studied in small ruminants. It was showed that in the course of SRLVs infection, TLR7 and TLR8 become activated, inducing IFN-α, IL-6, TNF-α production and expression of antiviral proteins [21]. Upregulation of TLR7 and TLR8 was noted in naturally SRLVs-infected sheep showing lung lesions [13]. Single nucleotide polymorphisms (SNPs) of genes encoding proteins involved in innate response have aroused much attention, and a number of studies have been conducted to identify SNPs in these genes in different species [22,23,24]. It has been found that genetic variants of TLRs can be associated with different outcome of several diseases [15,25]. Polymorphisms within TLR7 and TLR8 have been linked to susceptibility and progression of different viral diseases including these caused by Human Immunodeficiency virus (HIV-1), Hepatitis C virus (HCV), Chikungunya virus (CHIKV) and Crimean-Congo hemorrhagic fever (CCHF) virus [26,27,28,29]. Mikula et al. [30,31] suggested that mutations in TLR7, TLR8 and TLR9 may play an important role as host factor predisposing sheep for infection with SRLVs. Despite these observations, there is currently a lack of any studies reporting polymorphisms of caprine genes encoding TLR7 and TLR8 and their possible role in SRLVs infection. Herein, this study was conducted to identify SNPs in genes encoding TLR7 and TLR8 in goats of Carpathian breed and analyze their association with the SRLVs infection and provirus concentration. We focused on goats of Carpathian breed because represent a remnant population of an ancient breed; thus, analysis in this population could be important for its conservation.2. Materials and Methods2.1. Animals and Sample PreparationThe study was performed in one flock, counting 32 adult goats representing Carpathian breed. All 32 goats were examined in this study. The goats of Carpathian breed were widely found in Carpathian Mountain in Poland in the 19th and 20th century and then became an extinct breed. In 2005, all of these goats were moved to the National Research Institute of Animal Production in Cracow where this flock has been recreated and currently is covered by the genetic resources protection program, supported by the Ministry of Agriculture and Rural Development; however, its risk of extinction is high. All goats were clinically healthy and maintained at the same environmental and feeding conditions. Age of goats ranged from 2 to 10 years (average ~6 years). Blood samples were taken by jugular venipuncture in EDTA tubes and in second tubes for serum collection, and the blood was collected from all animals at the same day. DNA extraction was performed using Sherlock AX isolation kit (A&A Biotechnology, Gdynia, Poland) and quality of preparation was checked using NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA). Status of goats for SRLVs infection was confirmed by serological testing using ELISA (ID Screen MVV/CAEV Indirect Screening test, IDVet, France), according to the manufacturer’s recommendations. All procedures associated with animal handling and treatments were approved by the Local Ethical Committee on Animal Testing at the University of Life Sciences in Lublin (Poland).2.2. SNP Identification–Variant CallingAs a first approach, the screening for SNPs identification was performed using already available RNA sequencing (RNA-seq) data obtained from 12 goats from tested flock (data available under GEO GSE168160 accession number). For SNPs identification, the quality of raw data was measured with the use of FastQC software [32] which was followed by trimming procedure that focused on removing adapter content, reads of low quality (phred quality < 20) and low read length (minimal read length set to 35) with the use of Trimmomatic software [33]. Then, the quality of trimmed reads was checked again to ensure the effectiveness of trimming. The next step was mapping procedure which was maintained with the use of Tophat software [34] to Capra Hircus ARS1 genome. Then, the duplicates were marked with the use of MarkDuplicates function of Picard Tools, and finally, the variant calling procedure was utilized with the use of Freebayes software [35] with a minimum coverage threshold set to 5. The filtration of the obtained variants was done with the use of VCFtools software [36] with the following parameters: minimum coverage set to 10; minimum quality set to 30; minimum combined coverage set to 120 and minimum combined quality set to 360. The annotation of the TLR8 and TLR7 gene variants was done with the use of Ensembl Variant Effect Predictor online software [37].2.3. SNP GenotypingBased on information on RNA-seq SNPs, for both TLR8 and TLR7 loci, the primers span gene’s regions containing polymorphisms were designed and PCR amplification, Sanger sequencing and SNPs analysis were performed on samples of all 32 goats. The primers span gene’s region containing SNPs was designed using Primer3 (v. 0.4.0) based on the reference sequence shown in Table 1 and Supplementary Figures S1 and S2. The PCR amplification products were obtained using AmpliTaq Gold™ 360 Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) and purified using enzyme mixture-EPPiC (A&A Biotechnology, Gdynia, Poland). The Sanger sequencing was performed using BigDye™ Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific) and BigDye XTerminator™ Purification Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. The amplicons were capillary sequenced on 3500xL Genetic Analyzer (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). Data were analyzed using Data Collection Software (Applied Biosystems). All identified variants were submitted to European Variation Archive (EVA) and received the Project ID PRJEB42246.2.4. Poviral Load QuantificationDNA extracted from peripheral blood leukocytes (PBLs) [38] was quantified by the real time PCRs using Rotor-Gene Q Series ver. 2.0.3 (Qiagen) with primers and probe specifically designed for SRLV A5 subtype, which circulation in this flock was confirmed [39]. Sequence of forward and reverse primers and probe were CA5F (5′ TGGGAGTAGGACAAACAAATCA 3′), CA5R (5′ TGACATAT GCCTTACTGCTCTC 3′) and CA5P (5′ 6-FAM-TCACCCATTGTAGGCATAGCTGCC-BHQ-1 3′), respectively. A reference plasmid encompassing the target gag region was generated by the cloning of a 625 bp fragment into pDrive plasmid used to generate a standard curve based on 10-fold serial dilutions of plasmid DNA from 108 to 10. Amplification was performed in a total volume of 20 μL, according to the following cycling conditions: initial incubation and polymerase activation at 95 °C for 15 min and followed by 45 cycles of 94 °C for 60 s and 60 °C for 60 s. The reaction mixture for each PCR test contained 10 μL 2× QuantiTect Multiplex NoROX PCR buffer (Qiagen, Hilden, Germany), 400 nM of each of the primers, 200 nM of the specific probe, 5 μL of extracted genomic DNA. A non-template control (DEPC H2O) was included in each run. All samples were tested in duplicate, and the results were expressed as a mean copy number of provirus per 500 ng of genomic DNA of each goat. 2.5. Statistical AnalysisTo identify association between genotypes and proviral load, a simple linear model approach was performed using t-test. In this analysis SNP was used as a classification variable and proviral load values were used as analysis variables. Quantitative results given as means ± SD presented specific significant differences between animals within given genotype groups. The association between SRLVs DNA proviral load and age of tested goats were estimated using the Kruskal–Wallis H test. For this purpose, the animals were divided into 4 groups: 1 (2–3 years old), 2 (4–6 years old), 3 (7–8 years old) and 4 (9–10 years old). p-value of <0.05 was considered statistically significant. All statistical analyses were performed using SAS Enterprise (Statistical Analysis System, Version 8.02, 2001, Wadowice, Poland).The genotype distribution of the SNPs of TLR7 and TLR8 were tested for deviation from the Hardy-Weinberg equilibrium (HWE) by means of χ2 analysis using Court lab- HW calculator. p-value of <0.05 were considered statistically significant. Phase program was used to assess frequency of haplotype distribution, and Haploview program [40] was used (default mode) to generate haplotype block structures and calculate LD values between SNPs.3. Results 3.1. Serology and qPCROut of 32 goats used in this study, 29 were found seropositive by ELISA and positive to quantitative polymerase chain reaction (qPCR) confirming the infection with SRLVs. Three goats were negative in both ELISA and qPCR. The average number of proviral copies in positive samples varied from 1 to 263 per 500 ng of genomic DNA. The correlation between proviral load and age of goats was not statistically significant (H = 6.893613; p < 0.0754; Kruskal-Wallis test).3.2. TLR7 Gene-Detected SNPs, Allele and Haplotype FrequencyThe variant calling method allowed to identify four polymorphisms in TLR7 gene locus. Two of them were 3′UTR (untranslated regions) variants (C/T 1:50701297 and T/C 1:50702074), one 5′UTR variant (C/T 1:50718645) and one synonymous polymorphism (T/C 1: 50703293) (Table 2). Additionally, Sanger sequencing performed on specimen collected from all goats revealed two additional SNPs, one in promoter and second in intron 1 regions (G/A 1: 50718760 and C/T 1: 50718466), which was not found in the first approach when whole blood transcriptomes were sequenced (Table 3).The allele frequency analysis showed that two of all analyzed SNPs (promoter G/A 1:50718760 and 3′UTR T/C 1:50702074) were monomorphic but showed the mutant allele compared to the reference sequence. For the rest four polymorphisms, three genotypes were identified. The genotype and allele distributions of these four SNPs were shown in Table 4.The Synonymous Variant (SV) T/C; 3′UTR and 5′UTR polymorphisms showed the similar allele distribution. These SNPs were most abundantly present in heterozygous goats (Table 4). Distribution of CC and TT genotype was from 16% to 31% and from 23% to 31%, respectively. For 3′UTR (C/T 1:50701297) and 5′UTR SNPs (C/T 1:50718645), only five animals, including 3 seronegative goats, carried CC genotype. For SV T/C 1:50703293 polymorphism, these goats carried TT genotype. The frequency of intron 1 (C/T 1:50718466) allele showed the predominance of CC genotype (91%), while opposite homozygotes TT and heterozygotes CT were 3% and 6%, respectively. One out 3 seronegative goats carried TT genotype while 2 other goats carried CC genotype. Moreover, the analyzed population was not in Hardy-Weinberg equilibrium according to intron C/T 1:50718466 (Table 4). Haplotype analysis allowed to detect 8 haplotype regions. Three haplotypes represented frequency from 10% to 51% while two haplotypes had frequency lower that 1% (Table 5). The linkage disequilibrium (LD) analysis showed the presence of one LD block and strong LD between four SNPs (Supplementary Materials: Table S1 and Figure S3).3.3. TLR8 Gene-Detected SNPs, Allele and Haplotype FrequencyIn TLR8 gene, using RNA-seq, six polymorphisms were detected from which two were the 3′UTR variants and the rest were the synonymous SNPs (Table 2). Additionally, Sanger sequencing allowed detecting two extra SNPs in the 3′UTR region: (1:50659346 C/A, 1:50659136 A/C) (Table 3).In TLR8 locus, 3′UTR T/C 1:50659202 SNP was monomorphic but showed the mutant allele compared to the reference sequence. For 1:50664682 A/G SNP, two genotypes were identified. The frequencies of identified AA and GT genotypes were 95% and 5%, respectively. For the rest six polymorphisms, three genotypes were identified. One type of homozygote (AA or CC) was found in most goats, including also seronegative animals while heterozygotes goats accounted for 32–42% (Table 6).The haplotype analysis revealed the presence of 10 haplotypes from which frequency ranged from 29.7% to 13.30% (Table 5). The linkage disequilibrium (LD) analysis showed the one LD block involving 6 polymorphisms (Supplementary Materials: Table S1 and Figure S3).3.4. Association between SNPs and Provirus Copy NumberBecause only three goats were uninfected with SRLVs, it was impossible to create control group with an equivalent number of animals to those of serologically positive and compare allele and genotype frequencies between infected and uninfected goats. Therefore, only associations between identified SNPs and provirus copy number of SRLVs were estimated for both TLR genes. It was showed that nine of the identified polymorphisms showed significant association with provirus copy number. For TLR7 gene, four polymorphisms were significantly associated with SRLVs provirus copy number (Figure 1). For TLR7 intron variant 1:50718466, goats with CC genotype had significantly higher proviral load that CT genotype goats (p value < 0.05). For three other SNPs (synonymous variant 1:50703293, 3′UTR variant 1:50701297 and 5′UTR variant 1:50718645), the heterozygote goats CT were characterized by higher provirus copy number (p < 0.05).For TLR8 gene, five polymorphisms were significantly associated with provirus copy number (Figure 2). For TLR8 1:50666071 synonymous variant, homozygotes GG goats had lower proviral load than heterozygotes AG goats (p < 0.05). For two synonymous TLR8 polymorphisms (1:50664064 and 1:50664208), the goats with the CC genotype showed the lower copy number of SRLVs compared to CT animals. For 3′UTR variant 1:50659346 AC genotype, goats had significantly higher proviral load than CC genotype goats (p value < 0.05). The exact opposite was the case with 1:50659136 SNP, for which homozygotes AA goats had significantly higher SRLVs proviral load than both other genotypes (p < 0.01).4. DiscussionThe present study is the first attempt showing the genetic variability in TLR7 and TLR8 genes in SRLVs infected goats as well as association of SNPs with SRLVs proviral concentration.We identified 6 polymorphisms in the TLR7 gene locus and 8 SNPs in the TLR8 gene. These SNPs were located in intron, 3′UTR and 5′UTR regions and within the coding sequences leading to the synonymous mutations which do not result in any amino acid change in encoded protein, due to genetic code redundancy. Synonymous substitutions and mutations, affecting noncoding DNA regions, are often considered silent mutations; however, it is not always the case that the mutation is silent. This kind of polymorphisms can produce different effects on gene expression leading to functional differences of various significance. Several recent reports consider such mutations, particularly concerning human diseases [41,42]. Silent mutations may affect translational efficiency and protein folding by changing codons read by tRNAs [43,44] and alter mRNA stability structure or splicing leading to variation in protein expression [45]. SNPs in the promoter regions could affect their activity and regulation producing changes in gene expression levels. Moreover, through different mechanisms, silent mutations may influence gene regulation, differences in mRNA and protein abundance, and proteins’ structure and functionality [45]. Untranslated regions (UTRs), which are localized at both ends of transcript (mRNA), make numerous conformational structures, tridimensional loops and hairpins that interact with numerous proteins and other functional and regulatory compounds like ribosomes or microRNA. UTRs are known to play crucial roles in the post-transcriptional regulation of gene expression. The main role of 5′UTR is controlling of translation efficiency as well as transcript stabilization while 3′UTR is mostly implicated in regulation of transcript stabilization, including modulation of the transport of mRNAs out of the nucleus [46,47]. Furthermore 3′UTR can be microRNA target sites, where microRNAs bind and regulate genetic expression. Therefore, SNPs or mutations in these regions might alter the existing target sites for microRNAs [48]. The importance of UTRs in regulating gene expression is underlined by the finding that mutations that alter the UTR can lead to serious pathology [49].The role of TLR7 and TLR8 SNPs in SRLVs infection surely requires a comparison of infected and uninfected animals. Unfortunately, only 3 out of 32 tested goats were uninfected, so statistical analysis was not possible. However, it is worth pointing out that goats from tested flock were serologically examined several times over the past few years, and only these 3 goats remained always negative. Serological studies revealed that very young goats presented antibodies against SRLVs while these 3 goats, which have 3, 7 and 10 years old, were uninfected. This clearly indicates an implication of host factors that may restrict and control SRLVs replication. Because it was impossible to create control group with an equivalent number of animals to those of serologically positive and compare allele and genotype frequencies between infected and uninfected goats, we focused on investigating the association between TLR7 and TLR8 mutations and SRLVs proviral concentration, as important index of disease progression. It has been demonstrated that host control of infection with SRLVs, including provirus level, may have a genetic basis [10,50,51,52]. Genetic studies in humans have pointed to a role for TLR SNPs, especially TLR7, in HIV infection and their association with viral loads. The presence of the most frequent TLR7 polymorphisms, TLR7 Gln11Leu (rs179008), was associated with increased viral load and altered CD4 T cell counts during HIV infection [53,54] while polymorphisms in TLR7 (re179010) and TLR8 (rs3764880) may reduce the risk of disease by participating in inhibition of viral load leading to the slower progression of infection [28,55,56]. Our results revealed that 9 out 14 identified polymorphisms were significantly associated with the SRLVs proviral concentration. In particular, homozygous genotypes of three TLR7 SNPs (synonymous variant 1:50703293, 3′UTR variant 1:50701297 and 5′UTR variant 1:50718645) were observed in goats with lower provirus copy number as well as in seronegative animals. However, additional analysis is needed to confirm this association. In this analysis, the effort should be directed to goat’s samples collected at different time points to check the evolution of proviral load over the time.The direct influence of the SNPs detected in this study on phenotypic features is unknown. Similar to what was showed for SRLVs infection in sheep [57], it can be assumed that during SRLVs infection in goats, TLR7 and TLR8 became activated, inducing production of antiviral cytokine, including IFN5-α, IL-6 and TNF-α production, which are critical in the development of antiviral immune response [21]. Activation of IFN type I pathway leads to expression of genes with IFN-sensitive response elements (ISREs) which fight viruses through different mechanisms. It has been confirmed that SRLVs contain ISRE and that IFN may modulate viral transcription promoting an antiviral state [14,58]. Therefore, we can speculate that SNPs detected in this study may affect the level of TLR7/8 expression, causing the differences in the production of downstream cytokines, like IFN-α, which finally can lead to the differences in SRLVs proviral concentration.Studies on the association between TLR variants and HIV infection/disease progression suggested that they may be specific for different human populations [59]. Diversity reported between breed’s susceptibility/resistance to SRLVs [50,60] may suggest that SNPs detected in this study may be typical for goats of Carpathian breed. We especially focused on goats of this breed because the Carpathian goat is an ancient breed which in the 19th/20th century was present in Carpathian Mountain in Poland and then became an extinct breed. It emphasizes the need to select animals for higher resistance to infections. Considering the fact that animals belonging to this flock showed close to 100% seropositivity against SRLVs, the identification of genetic markers associated with SRLVs proviral load may be important for further breeding program. Removal of infected animals, especially those with high proviral load can limit the spread of the virus since such animals are highly efficient in shedding the virus [10].In conclusion, this is the first report showing single nucleotide polymorphisms in genes encoding TLR7/8 in goats and the association between TLR7/8 polymorphisms and SRLVs proviral concentration in goats of Carpathian breed. Limited number of animals tested in this study and lack of possibility to compare infected and non-infected animals are undoubtedly limitations of this work. However, the obtained results suggest that SNPs of TLR7/TLR8 genes may induce differential innate immune response towards SRLVs affecting proviral concentration and thereby disease progression. These findings support a role of genetic variations in TLR7 and TLR8 in the course of infection with SRLVs and warrants further studies on TLR7/TLR8 polymorphisms.

animals : an open access journal from mdpi

10.3390/ani11041055

PMC8068398

In the last few years, the fourth industrial revolution has found its way into agriculture. Under the heading of smart farming, various so-called smart products are offered that can positively influence the daily work of farmers as well as animal welfare. These smart products can record data from the farming operation, extract essential information, and in some cases even make decisions autonomously. Particularly in Germany, where social criticism of intensive livestock farming has been raised, such smart products could make a significant contribution to improving animal welfare. However, a key prerequisite is the acceptance of the users, who are typically the livestock farmers themselves. There is so far hardly any knowledge about farmers’ attitudes towards smart products in livestock farming. In this study, the attitudes of German livestock farmers towards smart products are evaluated by categorizing them into groups by a factor analysis and a cluster analysis. Based on the analysis of an online questionnaire, in which German livestock farmers (n = 422) participated, four clusters could be derived. The main distinguishing characteristics of the clusters are the influence of the social environment, the expected effort for implementation, the general trust in smart products, and the technology readiness of the farms. As a result, this study provides valuable insights, for technology providers of smart products for livestock farming, as well as for policymakers.

In recent years, the fourth industrial revolution has found its way into agriculture. Under the term smart farming, various so-called smart products are offered that may positively influence both the daily work of farmers and animal welfare. These smart products can collect data from the farm, extract important information, and in some cases even make decisions independently. Particularly in Germany, where intensive livestock farming is criticized by society, such smart products could make a significant contribution to improving animal welfare. However, an important prerequisite is the acceptance of the users, who are usually the livestock farmers themselves. So far, there is little knowledge about farmers’ attitudes towards smart products in livestock production. In this study, a factor analysis and a cluster analysis are conducted to evaluate the attitudes of German livestock farmers towards smart products. Based on the analysis of an online questionnaire in which German livestock farmers (n = 422) participated, four clusters could be derived. The main distinguishing characteristics of the clusters are the influence of the social environment, the expected effort for implementation, the general trust in smart products, and the technology readiness of the farms. As a result, this study provides valuable insights for technology providers of smart products for livestock farming as well as for policy makers.

1. IntroductionIntensive, conventional livestock farming has been the subject of vehement public criticism for several years. In addition to reasons related to the way in which the animals are kept themselves, such as the space provided per animal, animal welfare issues, animal–human interaction, and overarching scandals in food supply chains are further responsible for the fact that the professional image of farmers and livestock farmers is deteriorating [1]. In addition to animal welfare, livestock farming and agriculture further play a key role within the international debate on climate change and sustainability, especially with regard to the sustainable use of natural resources [2]. As a consequence, a strong aversion to current systems of intensive livestock farming prevails, specifically for pigs and poultry, but also in dairy farming [3]. Furthermore, the sensitivity towards agriculture along with requirements aimed at improving farm animal welfare standards is increased [4,5,6,7]. Common choices for livestock producers to counter criticism include open farm tours for everyone or live streaming from livestock facilities. Another way for farmers to respond to societal and policy demands and provide a higher standard of farm animal welfare is to participate in specialized animal welfare programs [8,9]. For instance, one such animal welfare program was established in Germany in 2015 under the name of the Initiative Animal Welfare (IAW). Its objective is to improve farm animal welfare in German poultry and pig production. Based on current consumer surveys, it can be inferred that the IAW program is very well accepted overall by consumers in Germany [10]. Since 2020, a total of 4200 pig farmers (24.7% of total pig farmers in Germany), and 2000 chicken farmers (60% of total meat chicken farmers in Germany) have participated in the IAW [11]. However, recent studies demonstrate that farmers perceive a great effort with participation in animal welfare programs, which significantly diminishes their intention to participate [12,13,14]. Since Germany represents one of the largest meat producers worldwide, it is considered to act as one of the main players on the international market [15]. For this reason, it can be assumed that examples and results from Germany can also be relevant on an international level. Thus, the question arises how to facilitate farmers’ work and especially the effort associated with livestock farming in a profitable way.In the last few years, the fourth industrial revolution has found its way into agriculture, driven by the continuously increasing diffusion of information and communication technologies. Smart farming is expected to enable the development of a more sustainable agriculture and livestock farming through its technologies [16,17]. In the field of livestock farming, smart ear tags for dairy cows were developed, which enable monitoring of individual animal behavior by tracking the movements of cows within the stable environment in real time [18]. Furthermore, lameness detectors are used that can automatically detect lameness in cows in real time and inform the farmer [19]. Smart ear tags equipped with accelerometers are further applied in sow farming for early lameness detection [20]. Common features of these smart livestock technologies include inherent supportive reconfiguration capabilities to perform agile actions in real time, especially in the event of sudden changes in farm operating conditions [21]. In addition, smart products are often referred to as intelligent devices or products that are interconnected and interact via local and global, often wireless, network infrastructures, thus serving as a link between the virtual and physical world by responding in real time and autonomously making decisions [22]. Smart products distinguish themselves from classical cyber-physical systems, e.g., by haptic user interfaces based on gesture identification or acceleration sensors. They are characterized by using specific technologies to capture and communicate information about themselves, their condition, and the surrounding environment [23]. The benefits of smart livestock farming are considered highly important, for example, by facilitating work time or simplifying animal monitoring. Regarding climate change and sustainability, it is assumed that smart farming reduces the ecological footprint of agriculture and mitigates greenhouse gas emissions in livestock farming as well as livestock monitoring and disease detection [24]. Nevertheless, in addition to technological and organizational challenges, hurdles remain in the adoption of these technologies. These include, for example, data security and data sovereignty, high costs of implementation, or limited farmer skills and knowledge [25,26].A fundamental condition for benefiting from the advantages of smart products in terms of farm animal welfare is farmer acceptance and use [27]. Previous studies have concentrated incrementally on technologies in the context of precision farming, while adoption of smart livestock farming has rather been neglected. For example, the relevance of the role of farm attributes and socio-demographic characteristics of farmers, such as education or age have been investigated [28,29]. With regard to smart products, only the use and acceptance of smartphones among farmers has been studied in Germany to date [30]. Further empirical studies explicitly investigating the acceptance of smart products in German livestock farming are non-existent. However, it is known that livestock farmers cannot be considered as a homogeneous group in terms of their attitudes and that a positive attitude also has an effect on acceptance as well as actual behavior [31]. For this reason, and to address the research gap, livestock farmers in Germany were surveyed about their attitudes and (potential) usage behavior with regard to smart products. A factor analysis and subsequently a cluster analysis was approached for this article in order to identify potential target groups for the use of smart products and to characterize them in more detail so that target group-specific managerial implications can be derived. The results bear important implications for manufacturers and policymakers at national and international levels, resulting in opportunities to tailor smart products to the needs and wishes of livestock farmers.2. Materials and Methods2.1. Study DesignA questionnaire among livestock farmers in Germany from June to August 2020 was analyzed for this article. It was based on an anonymous and standardized online questionnaire. The questionnaire was pre-tested for one week by different experts from agro-economic research and farmers. To reach as many livestock farmers as possible, and to achieve as much homogeneity, especially with regard to personal characteristics, compared to the basic population of livestock farmers in Germany, it was then distributed via a chair’s own mailing list and via the Lower Saxony Chamber of Agriculture. Private networks and social media, such as Facebook, Xing, and Instagram, were used to further distribute the link to the questionnaire. Further magazines in agricultural practice were asked to distribute the questionnaire (topagrar, profi, farm food future, Oldenburger intenational, der Hoftierazt). The questionnaire was preceded by a brief description of the background and objectives of the study, namely the investigation of livestock farmers’ attitudes towards the use of smart products. Furthermore, a definition of smart products was prefixed to the questionnaire (“More and more devices are connected to the internet. Smart products collect data via corresponding sensors, analyze them and forward them via the internet or receive data from other smart products. The “intelligence” of these products means that they perform tasks independently and communicate with other products”). Three examples were given: smart phone, smart ear tag, smart pig counter. Based on the extended “unified theory of acceptance and use of technology” model, the questionnaire was constructed [32]. With its key constructs being performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, price value, and habit, it provides a conceptual framework for analyzing decision makers’ intention to use or not to use a certain technology or system. Further, it has been applied to investigate acceptance issues for technologies or systems within agriculture [33].The questionnaire mainly consisted of closed questions. It was also divided into three parts. Part A asked farmers about socio-demographic and farm characteristics, such as age, gender, education, work experience, farm size, and individual farm operations. Part B consisted of 55 statements designed to determine farmers’ attitudes toward smart products. The respondents were asked to rate predefined statements on the basis of five-point Likert scales from 1, meaning, “I totally disagree” to 5, meaning, “I totally agree”. Part C consisted of 10 statements designed to determine farmers’ perceptions of national adoption of smart products. 2.2. Statistical AnalysisIBM SPSS Statistics 26 software for Microsoft was utilized to perform the factor and cluster analyzes as well as the descriptive analysis of the questionnaire. First, the data were cleaned regarding incompleteness and non-livestock farmers as well as outliers. In order to obtain an overview of the sample, socio-demographic and farm characteristics were analyzed using univariate methods. Subsequently, an exploratory factor analysis was conducted to combine a large number of items into a smaller number of factors, allowing for a simplified interpretation of farmers’ attitudes towards the use of smart products. A principal component analysis was applied for this article using the varimax rotation method and Kaiser Normalization. The varimax rotation was exerted because the variance of the squared loadings reaches the maximum value with this rotation and, thus, facilitates the interpretation, as the assignment of statements to the respective factors is simplified [34]. Despite the variables of the Unified Theory of Acceptance and Use of Technology, Likert-scaled items on farmers’ trust and technology readiness with respect to behavioral intention and use of smart products on their farms were included in the factor analysis. The before mentioned number of items was tested for suitability for factor analysis. The Kaiser–Meyer–Olkin (KMO) criteria and the “measure of sampling adequacy” (MSA) were applied for this purpose. If KMO and MSA are below 0.5, the result is considered “unacceptable” and factor analysis should not be performed [35]. In this study, a value of 0.930 was determined, which is classified as “marvelous” [36]. Another quality criterion is Bartlett’s test of sphericity, which assesses the null hypothesis so that the input variables from the survey population are in uncorrelated form [36]. In this study, Bartlett’s significance is at a value of p = 0.000, which is highly significant because it means that the null hypothesis can be rejected and thus there is a correlation between the variables. Moreover, as in principal component analysis, only those factors that have an eigenvalue greater than 1 are extracted. Here, the factor analysis explained 59.03 percent of the total variance. Finally, seven factors including 34 items were identified (see Section 3.2). A reliability analysis provided information about the internal consistency of the factors. Cronbach’s Alpha value was chosen to measure the consistency. All seven factors show a valid internal consistency [37,38].Afterwards a hierarchical cluster analysis was performed to classify farmers into different segments in terms of their attitudes towards smart products. For this purpose, the previously formed factors were applied to characterize the clusters. This type of cluster analysis is utilized to generate homogeneous groups from a heterogeneous population [36,39]. The process of conducting the cluster analysis was organized into three steps. Initially, the single-linkage method was performed to remove eventual outliers. The objects with the smallest linkages were combined with another. Then, the optimal number of clusters was determined based on Ward’s method. The analysis was successful in identifying four clusters. The dendrogram, a screen plot, and the elbow criterion were adopted to support the decision on the optimal number of clusters [36]. Eventually, a k-means clustering was performed to optimize the Ward’s results. Discriminant analysis was applied to verify the results. In this study, 100.00% of the original cases were correctly classified [36,38].Additional results of the discriminant analysis (eigenvalues and Wilks–Lambda) further prove that the results of the cluster analysis are of high quality. Wilks–Lambda is a significance test that confirms a significant difference between the groups identified in this study. In addition, the eigenvalue was greater than 1, and significant differences (p < 0.000) were found between the clustering factors [37]. No significant difference was found between the clusters with regard to the factor Performance, which was therefore not included in the cluster analysis. For a detailed characterization of the generated clusters, the Tamhane T2 post-hoc multiple comparison test was conducted, which demonstrated significant differences between the clusters. The Tamhane T2 test is based on the assumption that the groups may differ in variance. For a more detailed description and to test significant differences between the clusters, a univariate ANOVA, and a cross-tabulation were applied [38]. In addition, the clusters were analyzed with respect to socio-demographic and farm characteristics.3. Results3.1. Sample DescriptionA total of 422 complete data sets were available for analysis. Moreover, 24.4% of the respondents were female. Since the proportion of female employees in German agriculture is about 36%, women are slightly underrepresented in this study [40]. The majority of respondents live in Lower Saxony (22.3%), North Rhine-Westphalia (20.1%), and Bavaria (14.7%). The same federal states have the highest share of professional farms in total farms in Germany [41]. Moreover, 20.9% of the respondents are older than 55, slightly underrepresenting the average age of employees, as one-third of all agricultural employees in Germany are older than 55 [41]. The average farm size is 375.15 hectares, which is significantly higher than the average number of hectares per farm (61.9 hectares) in Germany [42]. Furthermore, 37.9% of the respondents hold an agricultural apprenticeship, have attended a special agricultural college, or have a Meister (master farmer) degree. In this case, the average of the educational distribution of German farmers, where 68% have the above types of educational qualifications, is fallen rather short [40]. Moreover, 42.9% of the respondents in the sample hold a university degree. Overall, the proportion of farmers with a university degree in Germany amounts to 12%, which leads to an overrepresentation of academics [40]. Furthermore, 65.8% of the respondents stated that they already use smart products (including smart phones) on their farm. Due to the small sample size and the sampling procedure (network sampling), this sample cannot be considered representative for all German livestock farmers with regard to the participants’ farm characteristics. Nevertheless, due to the highly significant results, these may provide interesting insights, especially for large full-time livestock farms.3.2. Results of the Factor AnalysisThe final factor solution included seven factors with 34 statements (see Table 1). The means (μ) and standard deviations (σ) are also plotted in Table 1. The first factor, “performance”, describes the farmers’ assessment of the operational benefits of using smart products and contains ten statements (Cronbach’s alpha: 0.891). Overall, the factor “performance” is assessed positively. The second factor, “animal benefits”, covers farmers’ attitudes with respect to livestock, and contains four statements (Cronbach’s alpha: 0.863). It is rated positively in general. The third factor, “social environment”, describes the influence that the social surrounding exerts on farmers in connection with smart products (Cronbach’s alpha: 0.779). The factor “social environment” varies and displays an almost indifferent picture on average. The fourth factor, “effort”, summarizes four statements about the effort incurred by using smart products (Cronbach’s alpha: 0.770). On the overall average, the effort is rated low. The fifth factor, “trust”, summarizes five statements about farmers’ attitudes toward trusting smart products (Cronbach’s alpha: 0.744). The overall average presents an indifferent picture. The sixth factor, “technology readiness”, combines four statements about the systemic conditions that should be fulfilled for the adoption of smart products (Cronbach’s alpha: 0.652). Although the overall average is positive, high standard deviations are evident. The seventh factor, “facilitating conditions”, describes the basic framework conditions that simplify the use of smart products on the farm (Cronbach’s Alpha: 0.728). Similar to technology readiness, high standard deviations are also observed for the factor “facilitating conditions”. The performed tests to check the quality of the factor analysis indicated that all factors met the common requirements. The factor analysis explained 59.03 percent of the total variance among the 34 statements. Since the KMO value is relatively high at 0.930, these statements are well suited for cluster analysis. The identified factors were adopted as clustering variables to form groups for smart products attitudes.3.3. Results of the Cluster AnalysisBased on the factors identified, a cluster analysis was conducted. The aim of the cluster analysis was to divide the farmers into groups according to their attitudes toward smart products. The factor performance had to be eliminated from the cluster analysis because no significant differences between the groups were identified. Table 2 presents the results of the cluster analysis, the mean values of the cluster-forming factors and the underlying variables. The first cluster (cluster A) contains 96 livestock farmers. It is characterized by an overall positive attitude towards the use of smart products and acts independently of its social environment, therefore entitled “independent proponents”. On the one hand, they are of the opinion that animal pathologies could be detected faster with the help of smart products (μ = 4.19; σ = 0.812) and, on the other hand, that they would let smart products assist them in animal control (μ = 4.32; σ = 0.900). Moreover, cluster A is independent of its social environment, especially in terms of society’s expectations (μ = 2.25; σ = 0.795) and colleagues’ opinions (μ = 2.63; σ = 0.861). Farmers tended to reject the statements that safe handling (μ = 2.77; σ = 1.192) and learning to handle (μ = −2.36; σ = 1.027) smart products was difficult. Cluster A farmers show little concern regarding statements related to trust. They would follow a recommendation that a smart product provides them (μ = 3.72; σ = 0.706) and demonstrate a basic trust in technologies and machines (μ = 3.92; σ = 0.816). In addition, they exhibit a high level of technology readiness, both because they use various sensors on their farm (μ = 4.84; σ = 0.443) and because computers are deployed for specific tasks, such as herd management (μ = 3.96; σ = 1.123). The statement indicating that internet access or mobile connection is available on the entire farm tends to be agreed with (μ = 3.51; σ = 1.384). Compared to clusters B, C, and D, farmers in cluster A show the least influence by their social environment. At the same time, they show the highest trust and the highest technology readiness given not entirely optimal internet conditions on their farm. The results of the cluster analysis reveal relatively high standard deviations within certain statements. This indicates the existence of heterogeneous attitudes within the individual groups, especially with regard to their technological readiness and the facilitating conditions.The second cluster (cluster B) comprises 65 livestock farmers. It is characterized by a high perception of the benefits of smart products, while the technology readiness and facilitating conditions remain insufficient. Farmers in this cluster are declared as the “hindered adopters”. Their opinions of possible benefits in livestock management are highly positive when it comes to obtaining real-time information about the animals’ condition (μ = 4.32; σ = 0.900) or that animal pathologies can be detected more rapidly (μ = 4.31; σ = 0.789). In addition, cluster B assumes that its social environment exerts little influence on its attitude toward smart products, especially regarding colleagues (μ = 3.20; σ = 1.003) and society’s expectations (μ = 2.62; σ = 1.041). In terms of effort related to the use of smart products, farmers in cluster B show no expected difficulties. This mainly concerns the learning effort (μ = 2.02; σ = 0.992) as well as the potential stress caused by the use (μ = 1.97; σ = 0.918). In addition, farmers in cluster B trust smart products in general. They are rather indifferent whether their data are secure when they work with products from reputable manufacturers (μ = 3.38; σ = 1.071), but assume that they receive sufficient utilization-related information from the manufacturer (μ = 4.00; σ = 0.866) in case of purchase. In terms of technology readiness, cluster B does not perform particularly well. They tend to negate statements that computers are involved in various tasks on the farm like herd management (μ = 2.03; σ = 1.104) and that information systems on the farm are capable of autonomous decision processes (μ = 1.47; σ = 0.730). In general, farmers from cluster B continue to rather lack facilitating conditions as the availability of a stationary or mobile internet access on the entire farm (μ = 2.68; σ = 1.147). Compared to clusters A, C, and D, cluster B considers smart products to provide the highest benefit toward livestock and the lowest effort in learning costs and handling, while they tend to barely meet the technically necessary prerequisites for adoption.The 157 farmers in the third cluster (cluster C) qualify as “unrestrained promoters” since they intend to adopt smart products in principle and fulfill the technical and operational requirements at the same time. Farmers in cluster C indicate that the use of smart products offers great advantages, as abnormalities in the animals’ natural behavior can be detected more rapidly (μ = 4.38; σ = 0.583). They would further unquestionably let a smart product assist them in animal control (μ = 4.47; σ = 0.656). Regarding the statements about the role of the social environment, it becomes apparent that especially the family (μ = 4.20; σ = 0.755), as well as society (μ = 3.86; σ = 0.763) markedly influence the farmers’ attitude. They do not perceive a great effort in learning how to use smart products (μ = 2.38; σ = 0.1.034) and do not consider the use as a stressor (μ = 1.97; σ = 1.032). Regarding trust, farmers of cluster C appear indifferent up to positive. They tend to agree with the statement that they trust smart products and their decisions (μ = 3.56; σ = 0.737) and would tend to follow a recommendation received from a smart product (μ = 3.66; σ = 0.694). Overall, they show very favorable farm conditions, which is confirmed by statements concerning the presence of an internet connection on the entire farm (μ = 4.15; σ = 1.051). Compared to clusters A, B, and D, they are most affected by the social environment, whereas at the same time they feature the most favorable digital environment in terms of Internet access.A total of 104 farmers were assigned to cluster D. Since the farmers in this cluster are highly skeptical, they are referred to as the “trustless critics”. Cluster D is mostly indifferent to the benefits a smart product could confer on livestock. This is reflected in the statements about whether animal diseases can be detected more rapidly (μ = 3.07; σ = 1.026) and whether farmers would let a smart product assist them in animal control (μ = 2.92; σ = 1.129). With regard to the social environment, cluster D is indifferent or independent. This is evident both in statements of a potential positive impression in society (μ = 3.12; σ = 0.938) and in the context of endorsement within the family (μ = 3.09; σ = 0.946). In addition, cluster D expects a slightly increased effort in the adoption of smart products, on the one hand with regard to the operation (μ = 3.28; σ = 1.038), on the other hand with regard to the stress caused by utilization (μ = 3.10; σ = 1.084). Trust in smart products tends to reveal a critical image. Questions about trust in smart products and their decisions (μ = 2.63; σ = 0.926) and about data security when working with smart products from reputable manufacturers (μ = 2.48; σ = 0.985) are answered doubtfully. Although various sensors are applied on farms from cluster D (μ = 4.11; σ = 0.944), autonomous decisions are not performed by information technologies (μ = 1.71; σ = 0.889). Hence, technology readiness tends to be limited. Furthermore, internet access for farmers from cluster D is limited on the farm (μ = 2.76; σ = 1.235), which constrains the conditions for smart product adoption. Compared to clusters A, B, and C, farmers in cluster D assess the animal benefits from smart products the lowest. At the same time, they evaluate the effort the highest and express the greatest concerns about trust. However, as with cluster A, high standard deviations occur, which indicate partially heterogeneous attitudes within the cluster.Table 3 shows some characteristics within the analyzed clusters. Most of these differences are at significant levels. Compared to the others, cluster B contains the highest proportion of women. In terms of age, it is also apparent that farmers from cluster B are on average about eight years younger. At the same time, farmers from cluster B hold on average the least amount of hectares of land, significantly less than cluster A and C. Furthermore, it becomes apparent that cluster B includes most of the farmers in a secondary occupation. Furthermore, farmers from cluster B average nine years less work experience. Few significant differences between clusters were identified with respect to the type of livestock farming. These were evident in sow keeping, dairy cattle, and horses. Thus, the fewest sow keepers are located in cluster B. Cluster D contains the smallest number of dairy farms, while the proportion in clusters A and C is rather large. Furthermore, the share of horse keepers is highest in cluster B. Beyond socio-demographic and farm characteristics, significant differences were discovered in behavioral intentions regarding the use of smart products. On the question of whether the intention to use smart products in the near future is present, clusters A and C show significant affirmation, whereas cluster D tends to negate this intention. Eventually, significant differences were revealed in the actual use of smart products. In particular, clusters A and C show a high proportion of farmers who already use smart products on their farms. In contrast, the proportion in clusters B and D is significantly lower. Usage was further assessed in terms of frequency of use. Initially, it emerged that clusters A and C most frequently use their smartphones for non-communicative purposes compared with clusters B and D, for example to monitor or control machinery or livestock facilities. Small machines or devices that perform tasks independently without human intervention are used less frequently, though least in clusters B and D.4. DiscussionThis study analyzed data from 422 livestock farmers who participated in an online questionnaire about their attitudes toward smart products. By applying a cluster analysis with a factor analysis in advance, four clusters of farmers could be formed. Cluster A (“independent proponents”) is characterized by its support for smart products. It shows the least influence of the social environment and at the same time the highest level of trust and technology readiness. Cluster B (“hindered adopters”) demonstrates a strong intention to use smart products. It evaluates the benefits in terms of livestock the highest and rates the effort as very low. However, it is the weakest in terms of technology readiness and facilitating conditions. Cluster C (“unrestrained promoters”) likewise shows a distinctly positive attitude toward smart products. It is most strongly influenced by the social environment. At the same time, it is most likely to feature facilitating conditions and similarly exhibits a high level of technology readiness. Cluster D (“trustless critics”) displays the strongest distrust of smart products. The farmers in this cluster perceive only a minor benefit for the livestock, consider an increased effort to learn operation and handling, and lack trust in smart technologies.Performance is defined as the extent to which a person believes that using the system or a technology, in this case smart products, provides a benefit. Based on the statements of the factor analysis, performance in relation to the use of smart products consists of a higher profit as well as the possibility to reduce working hours and to facilitate and accelerate work processes. In addition, performance includes the ability of farmers to produce more sustainably with the help of digitization. Although no significant differences were found among the four clusters, performance expectancy plays a role in the adoption of new technologies and systems. Economic aspects are considered as one of the main reasons for farmers to adopt smart farming technologies [21]. Expected performance in terms of profit further favors farmer adoption of a technology due to the prospect of more stabilized revenues [43]. In addition, expected financial performance represents an important positive influence on the adoption of sustainable practices in agriculture [44]. Furthermore, the opportunity itself to contribute to a more sustainable agriculture is perceived as an advantage of smart products in the overarching context of smart farming [16]. However, the ability to facilitate the control and management of livestock is another perceived benefit [45,46]. Even though performance was identified as a factor, it could not be included in the cluster analysis because no significant differences between the groups could be observed. Therefore, all clusters perceive certain benefits from the use of smart products.Animal benefits are defined as the degree of expectation an individual believes that the use of a system or technology, in this case smart products, will be beneficial and provide support and relative benefits to the livestock. Many farmers consider the constant monitoring and detection of changes in natural behavior as well as the early identification of diseases as a benefit of smart livestock farming [47]. Regarding the improvement of farm animal welfare, personal motivation, and joy in improving farm animal welfare is considered an important motivator, although it is often suppressed by financial incentives [48]. However, the care and responsibility for the livestock as well as the maintenance of animal health also contribute to the personal satisfaction [49]. Therefore, it can be assumed that these factors may also affect the farmers’ use behavior of smart products. The individual clusters differ significantly in their attitudes toward animal benefits. While clusters A, B, and C anticipate distinct benefits for livestock from the use of smart products, cluster D is considerably more skeptical, and would tend not to let a smart product assist them in monitoring livestock.Social influence is determined by important people, such as friends, colleagues, and relatives, who influence the individual to use a system or technology. Further, the role of society is also considered. Social pressure from society in particular exerts discomfort on livestock farmers, which was revealed in a study on the willingness to participate in animal welfare programs [50]. Research that examined the social influence of colleagues, friends, and family on strategic decisions on the farm identified a contribution of social influence related to issues such as sustainable agriculture and environmental conservation [51]. In addition, the experience of a farmer’s colleagues with new technologies, such as smart products was determined to significantly influence the farmer’s future use [52]. Moreover, family members greatly affect strategic decisions regarding the development of the farm [53]. Adopting smart products also represents a strategic decision for a farm. For farmers, familiar social contacts and interaction with colleagues are particularly important for successful technology implementation, as their learning processes are primarily socially oriented [54]. The effect of social influence is most evident in cluster C. Farmers from cluster A in particular appear to be independent of social influence. The influence of the social environment can also be observed in clusters B and D, albeit to a lesser extent.Effort in this article refers to the expected and actual effort required to implement and use smart products. Previous studies on animal welfare programs suggest that the additional workload associated with participation is criticized by many farmers. This particularly applies to the temporal burden due to documentation duties [55]. A different study found that willingness to participate in animal welfare programs decreases as implementation efforts increase [56]. Studies on precision agriculture report that farmers face adaptations in their farm by shifting from experience-based decisions to data-driven processes, leading to uncertainties about the potential costs of the technology [26,57]. At the same time, the demands on farmers’ knowledge and skills increased due to the growing complexity of technologies [47]. Farmers’ experiences with previous technology implementations, where production increases and labor savings were overestimated, may also contribute to unmet expectations and critical attitudes toward innovative technologies [57]. Therefore, here, the above-mentioned thesis that the willingness to use smart products decreases with increasing effort can be agreed to. Farmers in clusters A, B, and C do not appear to perceive learning how to use smart products as an actual effort at all, compared to cluster D. The same is evident in the questions about the perception of handling simplicity and a possible stressor that could emanate from smart products.Trust is not only an essential factor for a successful business relationship in the agricultural sector, but also substantial for the adoption of new technologies. Regarding the implementation of farm animal welfare-oriented measures, the extent to which farmers consider the implementation of a certain measure as important and reasonable in order to ensure sustainable livestock farming appears to play an overarching role [58]. Further, trust appears to be an impactful factor in the adoption of technologies, such as smart products. Empirical research for example suggests that low rates of technology adoption are more often caused by lack of confidence than by cost. Even if farmers discover the potential benefits of the technology, some lack confidence that the new technology will work as purported [59]. Simultaneously, increasing confidence by believing that a technology will deliver the expected results, as well as comprehending the process to achieve an objective with a technology, is assumed to lead to higher adoption rates of a technology [60]. Similar results emerged in the study of farmer adoption of innovative green technologies [61]. Particularly, at this point, the present results highlight that the population of farmers cannot be understood as a homogeneous group. Four clusters of farmers were found that differ in their attitudes and willingness to use smart products (Table 2). These results indicate that livestock farmers in clusters A, B, and C would tend to follow a recommendation that a smart product provides them, while cluster D rejects the statement. When asked about data security, clusters B and C show no trust-related concerns. Cluster A is overall indifferent regarding this statement whereas cluster D is characterized by massive distrust.Technology readiness refers to the availability of sufficient resources and organizational knowledge to successfully implement smart products [62,63]. Technology readiness thus encompasses the extent to which an organization is prepared for a technology implementation. It includes general information technology systems such as internal networks and specific information technology systems required to support the system [64]. Another factor is the extent to which organizational operational processes are aligned with the implementation challenges and can be adapted to the new system requirements [65]. In addition, it includes the extent to which an organization possesses or is able to acquire the necessary technical skills, knowledge, and workforce to implement a technology [66]. One of the few studies that investigated technology readiness for smart farming technologies on farms also emphasized the importance of technology readiness for adoption [67]. The clusters distinctly vary in their technology readiness. On the farms of clusters A and C, for example, computers and sensors are regularly utilized. Farmers from cluster D may use sensors, but computers are seldom utilized. The question of whether information technology systems independently make decisions on the farm was rather negated overall. This is least the case for clusters B and D. The construct facilitating conditions describes to which degree the respondents believe that a favorable infrastructure exists on the farm that facilitates the use of the system. In this study, facilitating conditions refer primarily to broadband coverage and the availability of an Internet connection. If farmers believe that a system or technology matches their needs and is compatible with their environment, it is considered likely that the technology will be adopted because they consider it a positive investment [68]. Furthermore, the availability of the necessary infrastructure for a technology such as smart products also represents facilitating conditions [21]. A previous study also indicated that a favorable farm environment increases the likelihood that a farmer will adopt a technology [69]. Cluster C exhibits the best prerequisites in terms of facilitating conditions with regard to Internet accessibility. Cluster A displays a similar tendency. However, it is noticeable that cluster D possesses limited Internet conditions, while cluster B ranks the weakest.From the above-mentioned factors, several opportunities emerge to enhance the attractiveness of smart products use by livestock farmers. These recommended approaches address manufacturers that develop and offer smart technologies, and policymakers in particular. Farmers from cluster A (“independent proponents”) are unconditionally positive towards smart products and independent from their social environment. This is also reflected in their behavioral intention and actual use behavior. Around four out of five farmers from cluster A already use smart products in their livestock and intend to continue in the future. Recommendations for action in cluster A address in particular policymakers, who should continue to improve broadband coverage in Germany. Farmers from cluster B (“hindered adopters”) tend to express highly positive attitudes towards the benefits of smart products and perceive their usage as effortless, while they rather lack technology readiness and facilitating conditions. It is assumed that for these reasons uncertainty about the intention of use prevails, as it complicates the use. Furthermore, it is remarkable that the livestock farmers in cluster B are significantly younger and consequently exhibit less work experience in terms of time than in the other clusters. In addition, cluster B contains more women and the amount of farmland is smaller than it is in the other clusters. Recommendations for action, similar to cluster A, are primarily directed at policymakers. This includes the promotion of nationwide broadband coverage, since the majority of agriculture is practiced in rural areas. Similar to animal welfare programs, policy makers could consider establishing financial incentives to promote smart livestock farming technologies that are proven to improve animal welfare in livestock farming. This idea is supported by the fact that livestock farmers perceive a distinct benefit for their livestock, but are stymied by the technical infrastructure required. Livestock farmers from cluster C (“unrestrained promoters”) strongly support smart products, underlie the influence of their social environment, and fulfill all prerequisites regarding technology readiness and facilitating conditions for the adoption of smart products. The use behavior clearly reflects this. With regard to socio-demographic and farm characteristics, it is evident that livestock farmers from cluster C operate the largest farms regarding farm size. The only recommendation for action for manufacturers and policymakers at this point is to continue to support these livestock farmers. Livestock farmers from cluster D (“trustless critics”) represent the most critical of smart products by far. They perceive no specific benefit to livestock, find increased effort to learn and use smart products, and lack trust in them. This is further reflected in the behavioral intention and use behavior. However, the only notable finding in the farm characteristics is the significantly lower number of dairy farmers than in clusters A and C. At first, it may be stated that compared with cluster A, B, and C, a substantially higher effort is required to motivate farmers of cluster D to use smart products. Recommended approaches focus especially on manufacturers and policymakers. Given the lack of obvious benefits to livestock farmers regarding their livestock, manufacturers should communicate the benefits and limitations of smart products simply and clearly. For instance, as additional documentation requirements are perceived as a heavy burden by farmers when participating in animal welfare programs, manufacturers should highlight benefits of smart products in facilitating documentation requirements or livestock monitoring. Since the expected effort also exerts a major influence on attitudes toward the use of smart products, usage should be associated with a positive experience, which is a reason for designing products that are simple and user-friendly in terms of haptics and user interface in order to facilitate uncomplicated use. This likewise includes ensuring that learning how to use and handle smart products becomes as easy as possible in order to reduce uncertainty and complexity. Hence, trust strongly influences the behavioral intentions of livestock farmers from cluster D, the exchange of monitoring data is an important aspect that can be achieved, for example, by establishing internationally binding legal standards. Manufacturers should further focus on service and, in the case of livestock farmers, especially on availability. Proper installation, necessary maintenance, and support for operational and problem issues form part of this. Another way to achieve greater trust can be to directly involve farmers in the development of smart products. This involves open communication with farmers and actively finding compromises. As facilitating conditions are one of the most important determinants for the use of smart products by farmers, the government should provide the necessary support for adoption, including the development of a more widespread broadband coverage. Further, the government could consider providing monetary incentives for the implementation of smart products, insofar as they contribute to enhancing farm animal welfare standards. To increase farmers’ trust, policymakers should remove ambiguities around data security and data sovereignty. Like most non-experimental studies, this study has some limitations that should be considered when interpreting the results. First, the study provides non-representative results due to the modest sample size of 422 farmers and the different composition of the sample compared to the population of German livestock farmers. The farmers in this study operate farms with an average of 375.15 hectares, which is markedly higher than the average for German farmers [42]. Nevertheless, due to the highly significant results and Germany’s role at the world market, these may provide interesting insights, especially for large full-time livestock farms. In addition, conducting an online survey certainly entails certain limitations. While it is easy to implement and realize, it may limit representativeness due to differences in internet use in certain segments of the population. Women, elderly persons, and persons with low education use the internet less frequently than men, younger persons, and persons with higher education [70]. Conducting the survey online enables some segments of the population to be reached more easily than others. Individuals who describe themselves as technology-averse might choose not to participate in an online survey in the first place. In addition, no response rate can be determined from the online survey, as no information is available on how many people received the link via the various mailing lists. An additional point of discussion is the occurrence of occasional high standard deviations within the results, which indicates that the attitudes of livestock farmers still differ within a cluster. This applies, in particular, to cluster D. Finally, it is worth mentioning the subjective nomenclature of the clusters, which originated from the authors, and therefore should not be considered as generally binding.5. ConclusionsThe aim of this article was to identify groups of livestock farmers that differ in their attitudes toward smart products and to analyze these groups. This objective was achieved by grouping livestock farmers into four clusters that differ significantly. These clusters further demonstrate differences between their socio-demographic and farm characteristics at a significant level. Based on the results, it is possible to deduce what is relevant for the farmers in each cluster. Important findings of this article include that technology readiness, trust, and the social environment exert a substantial influence on farmers’ attitudes towards smart products. At the same time, it becomes evident that certain performance expectancy is prevalent among all farmers surveyed, which is the reason why performance expectancy is not suitable as a delimiting criterion between the clusters.Future research could focus more precisely on the individual animal types and livestock housing systems. In addition, it could be scientifically investigated how the use of smart products actually affects the working time spent. One possible research method to draw comparisons regarding the temporal and financial situations of the livestock farmers could involve work diaries. These results could provide more detailed information than those obtained in the present study and would therefore deliver very interesting managerial results. Furthermore, the economic effects of using smart products in livestock farming could be analyzed by comparing efficiency aspects such as profitability, liquidity, and stability between farms using smart products and those not using smart products with the help of matching algorithms. In addition, farmers’ opinions should be taken into account for the future and compared separately between users and non-users of smart products. For this purpose, choice-based experiments could serve to explore under which conditions farmers would implement specific products. In addition, it would be interesting to investigate how the use of smart products actually affects animal welfare.

animals : an open access journal from mdpi

10.3390/ani13050787

PMC10000186

Animal diseases have increasingly become an important external impact that restricts the healthy development of the pig industry. African swine fever has seriously threatened the stable development of the pig industry in China and caused huge economic losses for farmers. The control behaviour by farmers is particularly important in the front line of epidemic prevention. To explore the factors that affect epidemic prevention and control behaviour by farmers, field survey data was empirically analysed. The results show that gender, education level, technical training, breeding years, breeding scale, specialisation, and risk awareness have significant effects on the biosecurity measures taken by farmers.

Animal diseases are a serious threat to animal husbandry production and diet health, and effective prevention and control measures need to be explored. This study investigates the factors influencing the adoption of biosecurity prevention and the control behaviours of hog farmers towards African swine fever and provides appropriate recommendations. Using research data from Sichuan, Hubei, Jiangsu, Tianjin, Liaoning, Jilin, and Hebei, we employed a binary logistic model to empirically analyse these factors. Regarding individual farmer characteristics, male farmers emphasised biosecurity prevention and control in farms, with higher education actively influencing the adoption of prevention and control measures. Farmers who received technical training were actively inclined to adopt such behaviours. Furthermore, the longer the duration of farming, the more probable the farmers were to neglect biosecurity prevention and control. However, the bigger and more specialised the farm, the more inclined they were to adopt prevention and control behaviours. With respect to disease prevention and control awareness, the more risk-averse the farmers were, the more they actively adopted epidemic prevention behaviours. As the awareness of epidemic risk increased, the farmers tended to adopt active epidemic prevention behaviours by reporting suspected outbreaks. The following policy recommendations were made: learning about epidemic prevention and improving professional skills; large-scale farming, specialised farming; and timely dissemination of information to raise risk awareness.

1. IntroductionThe high incidence of and difficulty in controlling major animal diseases has led to a focus on animal health and its relationship with biosecurity [1]. The Food and Agriculture Organization (FAO) considers biosecurity to be directly related to agricultural sustainability, food safety, and environmental protection [2]. The application of biosecurity is the most important measure for avoiding the spread of disease [3]. Disease prevention through biosecurity measures is considered an important factor in improving animal production [4]. In August 2018, China experienced an outbreak of African swine fever (ASF), which spread expeditiously across several provinces, resulting in a steep decline in hog production and dramatic market fluctuations [5]. The outbreak significantly impacted the hog industry, mainly in terms of markets and hog farming entities. The infection led to abnormal fluctuations in the prices of meat, especially pork, disrupting market development and resulting in considerable disparities in regional pork prices [6]. Most farm owners lacked the experience and expertise to prevent and control the virus. Therefore, the measures taken were inadequate, culminating in low-level biosecurity in farming environments. Consequently, the epidemic brought huge economic losses to hog-farming entities [7]. Differences in production size, biosecurity standards, production inputs, and sales practices between hog farms can affect the potential risk of disease transmission and lead to a wide variation in the ability to take preventive measures between farms [8,9]. The current biosecurity situation in China is not promising; the state of some farms is alarming in the wake of the African swine fever, limiting their productivity, and causing heavy economic losses. Therefore, it is necessary to understand the current situation of biosecurity in Chinese hog farms and investigate the factors that influence the prevention and control of epidemics, to improve the level of biosecurity.Existing studies on the epidemic prevention behaviours of farmers have mainly focused on specific aspects. The epidemic prevention behaviour of farmers is the first step in animal epidemic prevention and control. Farmers usually perform four kinds of epidemic prevention behaviours, which include, the improvement of feeding conditions, the use of vaccines and veterinary drugs, the reporting of animal epidemics, and the treatment of sick and dead livestock [10]. Strengthening biosecurity management measures is the most important means to prevent and control African swine fever in China’s pig farms. The construction of a biosecurity system requires attention to the safety of inputs, optimisation of the feeding environment, management of personnel and vehicles, disinfection, and waste disposal [11]. Zhou et al. [12] examined the disease prevention and control behaviours of farmers and found that small-scale hog farms focused mainly on disinfection, feed and water management, disease prevention and control, and incoming and outgoing vehicle management. Meanwhile, retail farmers emphasised the regulation of the internal and external environment of hog farms, personnel management, and the use of vaccines and drugs. In a study on biocontainment measures in avian influenza prevention and control, Huang et al. [13] focused on preventing rodents and birds, as well as measures to manage incoming vehicles and disinfection of personnel. It has also been suggested that four factors: vaccines, veterinary drugs, disinfection and cleaning, and quarantine, play a crucial role in the prevention and control of hog diseases [14]. The immunisation file, also known as the animal household register, is also an important link in the animal epidemic prevention chain. It records in detail the ages of the animals, entry bar, immunisation date, immunisation identification number, vaccine type, injection, etc [15]. The establishment of immunisation files by farmers is vital to ensure the implementation of compulsory immunisation programs, and the conducting of risk assessments for the epidemic control of animal diseases to accurately grasp the immunisation status and monitor animals, both overall and individually, to avoid missed and late immunisations and to ensure the overall quality of the immunisation [16]. Moreover, ensuring drinking water quality is an important component of livestock rearing and management [17]. Water quality can be evaluated by testing the farm’s drinking water source to check whether the bacterial count exceeds the limit [18]. Since all types of vehicles are in direct contact with the outside world, vehicle biosecurity and management are also important aspects of hog farm biosecurity controls [19].Scholars have long examined the factors influencing disease prevention and control in farms. He Zhongwei et al. [20] studied the influencing factors of prevention and control behaviours in poultry farmers, taking the individual characteristics, breeding characteristics, epidemic awareness, external environment cognition, and policy implementation as explanatory variables, with which to study the impact of whether farmers adopt epidemic prevention behaviour. Li Yanling et al. [21] analysed the factors affecting epidemic prevention behaviours from four aspects: farmer characteristics, production and operation characteristics, cognitive characteristics, and environmental characteristics. Studying the relationship between epidemic prevention behaviours and the characteristics of poultry farmers, Li Jie et al. [10] used further four prevention behaviours: regular disinfection, whether to believe the official warning, whether to clean the chicken cage, and whether to adhere to the “all-in and all-out” system. Song et al. [22] conducted an empirical analysis of hog farms in the Hubei province and concluded that factors such as farmer literacy, whether they had participated in the training, their knowledge of epidemics, the scale of farming, and whether they had joined a cooperative organisation significantly influenced whether the farmers adopted the disease prevention and control measures. Through a study in the Xinjiang Autonomous Region, Chen et al. [23]. analysed disease prevention and the control behaviours of cattle and sheep breeding cooperatives and found that the education level of the breeders, whether they had received training, their annual income, and the provision of timely assistance were the main influencing factors. Luo et al. [24]. studied the disease prevention and control behaviours in beef cattle and meat sheep farms in the Qinghai province and determined that the education level, whether they had received disease prevention and control training, if the farm maintained records of disease prevention and control, and previous certifications were the main influencing factors determining the adoption of the disease prevention and control behaviours by the farmers. Through literature reviews, it can be summarised that there are four main categories of factors that influence the disease prevention and control behaviours of farmers: the characteristics of the farmer, including their gender, age, and education level; the production and operation characteristics, such as the proportion of the farming income in the total household income; the cognitive characteristics of the farmer, such as their cognitive level, the willingness for disease prevention and control and safety awareness; and environmental characteristics, including socio-economic environment and government policies, and what social services are provided [21,25,26].While many scholars have adequately researched the disease prevention and control behaviours of farmers alongside any influencing factors, there remain some shortcomings. Most of this relevant research merely focuses on a specific region or province and the representativeness of the samples and the universality of the pattern are, thus, inadequate. For this reason, we used the logistic model to empirically analyse the main factors affecting the relevant animal disease prevention and control behaviours, from the perspective of the farmers. Moreover, we analysed the microscopic research data on hog farmers in Sichuan, Hubei, Jiangsu, Tianjin, Liaoning, Jilin and Hebei, and proposed more targeted countermeasures to improve the prevention and control capacity of the farmers, by providing policy recommendations to improve China’s animal disease prevention and control at the grassroots level.2. Materials and Methods2.1. Data SourcesThe data used in this study were obtained from a survey of hog farming and disease prevention and control conducted by the research team in the third and fourth quarters of 2020. The survey area includes seven provinces (municipalities directly under the Central Government) in Sichuan, Hubei, Jiangsu, Tianjin, Liaoning, Jilin, and Hebei. The selection of the provinces and regions for the research mainly used the following approach: each province was ranked from the highest to lowest by the number of slaughtered fattened hogs in 2018, the top 18 provinces whose total production reached 90.78% of the national total were categorised into six grades, with each grade representing a different output level, and one province was selected in each grade as a representative of the corresponding production level. Moreover, considering the geographical variability of hog farming, we selected the provinces to research according to six major geographical sub-regions of China. The north-western region, where a very low number of pigs are produced, was not included in the 18 provinces. The remaining five regions were considered in an integrated/comprehensive manner. When sampling representative provinces in different grades, we selected provinces in different geographical subdivisions, as far as possible, to indicate the variability of the geographical areas used for farming. The specific selections are listed in Table 1.In addition, some of the provinces selected included those with African swine fever outbreaks, to consider the research content in terms of disease prevention and control. The samples were selected by adopting the principle of stratified stochastic sampling. To gain a comprehensive picture of the occurrence and non-occurrence of the African swine fever outbreak, we selected one infected and one non-infected county per province. Based on the above, we selected the specific locations for research, as shown in Table 2. The survey, which mainly targeted hog farmers, included a combination of field interviews and questionnaires. A total of 267 questionnaires were returned, and 262 valid samples were obtained, excluding those with missing or abnormal data. The items in the questionnaire mainly included basic characteristics of farming entities (age, gender, education level, and years of farming, etc.), the production and operation information of the farm (scale of farming, type of operation, and mode of farming, etc.) alongside the biosecurity prevention and control measures undertaken at the farm, including quarantine, disinfection, as well as vehicle and personnel management.2.2. Descriptive Analysis of the Survey SampleThe survey sample was selected in line with the basic characteristics of hog farming entities, and primarily included males, who accounted for 81.30% of the total sample. Individuals aged between 45–54 accounted for half of the overall group, and those aged 45 or above accounted for 73.66%, in line with the characteristic that the farming entities mainly consisted of middle-aged and elderly male groups. The majority had 6–15 years of education, and their education levels were dominated by junior secondary and senior secondary school, which accounted for 90.08%. The number of years of farming was evenly distributed, with the highest percentage (26.72%) having been engaged in hog farming for 6–10 years. Meanwhile, 79.39% of the farmers had received technical training, whereas approximately 20% had not. The percentage of entities whose farming income accounted for over 50% of their total household income was 92.75%. The proportion of farmers who adopted the farming mode of self-breeding and self-rearing was 69.47%, followed by commercial hog fattening at 26.34%. Additionally, 82.44% of the farms had not participated in industrial organisations, whereas the remaining 17.56% of farms had joined industrial organisations. The basic characteristics of the survey sample distributions are listed in Table 3. The detailed data can be found in the Supplementary Materials File (S1_survey data of pig farms).Epidemic prevention and control in the farm. According to the code of construction of ecological scale pig farms, the distance of the farm site from the main traffic routes and residential areas should meet certain requirements. The farm boundary should be no less than 300-metres from the main traffic routes, 300-metres from residential areas and other livestock farms, and 1000-metres from areas such as slaughterhouses, livestock and poultry processing plants, livestock and poultry trading markets, waste disposal sites and scenic tourist areas. With respect to the location of farms, the proportion of hog farms located over 300-metres from other farms was 68.70%, those located over 300-metres from residential areas were 69.47%, while 17.94% of the farms were located less than 300-metres from main traffic routes. With the above three indices taken together, the location of more than 60% of the survey sample farms was in accordance with construction regulations. The proportion of farms with an “all-in, all-out” management of feeder hogs is a mere 33.59%, indicating that most farmers buy and sell feeder hogs in separate batches. Furthermore, 90.08% of the farms have separate pens for sows, feeder hogs, and piglets, while 9.92% of farms had not conducted separate zoning management; and 76.72% of the farms had dedicated means of transport. With respect to immunisation records, 87.79% of the farms maintained them. Meanwhile, 83.59% of the farms had separate clean and dirty paths, but 40% of the farms did not conduct effective control of birds, as shown in Table 4. See a Supplementary Materials File (S1_survey data of pig farms) for complete information.2.3. Variable SelectionBy reviewing the literature, we found that the disease prevention and control behaviour of hog farmers is mainly influenced by the characteristics of farmers, variables such as farm size and degree of specialisation, and risk awareness of farmers. The variables selected for this study included the following components. The characteristics of the farmers mainly included gender of the farmer, educational level, and whether the farmer had received training in hog farming techniques. The farm characteristics mainly included the number of years of farming, the scale of farming, and the degree of specialisation. The farmers’ risk awareness included their investment risk appetite, whether the hog farms can effectively block the introduction of African swine fever, whether they will report any suspected outbreaks in hog farms, and whether they will report any suspected outbreaks in neighbouring hog farms. The farmers’ disease prevention and control behaviours included, whether the farm has immunisation records, whether water sources were tested, and whether the farm has a dedicated means of transport. The descriptive statistical analysis of the variables is detailed in Table 5. The observed original data of the variables can be found in a Supplementary Materials File (S1_survey data of pig farms).2.4. Model ConstructionThis study constructed a logistic model to test the factors influencing the adoption of preventive and control behaviours by farmers, which was estimated using the maximum likelihood estimation method. The dependent variables are divided into two categories: y=1 if the hog farmers adopt disease prevention and control behaviours, while y= 0 if not. X=(x1,x2,⋯,xk)T denotes the independent variable. The relationship between the probability P(y=1|X) of occurrence of the event and the independent variable was studied, with the logistic regression equation as follows:(1)P(y=1|X)=F(β0+β1x1+⋯+βkxk)=eβ0+β1x1+⋯+βkxk/(1+eβ0+β1x1+⋯+βkxk)+ε3. ResultsIn this study, Stata software was used to regress a binary logistic model for each of the three factors influencing hog epidemic prevention behaviours. Prior to model regression, variance inflation factor tests were conducted separately to verify the existence of multicollinearity problems. The results showed that the highest value of the inflation factor of the variables did not exceed 2, implying that there was no multicollinearity problem.The estimation results show that the three models fit well and the LR statistical significance levels are all highly significant at the 1% significance level. As shown in Table 6, in terms of individual characteristics, the gender of the farmer has a significant effect, a 10% significance level, on the adoption of disease prevention and control behaviours, whereas having a dedicated means of transport in place, and its coefficient is significantly negative, indicating that male hog farmers are more inclined to adopt dedicated means of transport. This is related to the fact that male farmers, as the main breadwinners of the family, pay greater attention to taking precautionary measures to ensure that household assets are protected from epidemics. At the 1% level of significance, the number of years of education has a positive effect on the adoption of disease prevention and control behaviours by hog farmers. The higher the educational level, the more active a farmer is in adopting the behaviour of testing the water source. This indicates that people with higher educational levels have greater knowledge and access to information about epidemics and take biological prevention and control at the farm more seriously. Technical training had a positive effect on the building of immunisation records and the disease prevention and control behaviours of dedicated means of transport, at significance levels of 10% and 1%, respectively. This indicates that the farmers’ participation in technical training helped to augment their knowledge of disease prevention and control, improved their relevant awareness, and promoted the relevant behaviours required to prevent epidemics.With respect to farming characteristics, the number of years of farming has a significantly negative effect on the adoption of disease prevention and control behaviours, with a 1% significance level. This indicates that the more years of farming they possess, the less of a priority the farmers attached to disease prevention and control. When they have adequate farming years, they tend to accumulate extensive farming experience, which, nevertheless, will prompt them to rely on their experience in animal disease prevention and control; thus, they are found to lack in science, which results in gaps in their disease prevention and control behaviours. The size of the farm had a significantly positive effect on the adoption of immunisation records and water testing behaviours at the 5% and 1% levels. This indicates that as the scale of farming increases, farmers pay additional attention to the adoption of disease prevention and control measures to avoid huge economic losses from the onset of diseases. At the 5% significance level, the degree of specialisation contributes positively to both the building of immunisation records and the adoption of dedicated means of transport, suggesting that the higher the proportion of total household income from farming, the greater the focus on the biosecurity situations in the farms, with more human, material, and financial resources invested in farming, which is conducive to the adoption of biosecurity prevention and control behaviours.With respect to disease prevention and control awareness, the farmers’ appetite for risk had a significant negative effect on building immunisation records and water testing at the 10% and 1% levels, respectively. This suggests that the more risk-averse the farmers were, the more active they were in adopting disease prevention and control behaviours. Conversely, the more risk-seeking the behaviour, the more passive they were in adopting disease prevention and control behaviours. This can be attributed to both types of farmers adopting different attitudes towards risks: active and passive. At the 10% significance level, the more the farmers believed that the farm can effectively intercept the introduction of African swine fever, the more likely they were to adopt active disease prevention and control behaviours and test water sources. In contrast, farmers were more likely to build immunisation records when reporting suspected outbreaks at their farms. At the 1% significant level, farmers were more inclined to have their water sources tested and their means of transport dedicated to the farm when reporting suspected outbreaks found in neighbouring hog farms. These data show that the farmers’ active awareness of disease prevention and control can also generate active disease prevention behaviours and exhibit a positive effect.4. DiscussionBiological vectors and trade are the main methods for African swine fever transmission. Biological vectors generally refer to organisms that infect and carry the African swine fever virus (ASFV). For example, ticks, wild suids, and domestic pigs. Susceptible pigs have direct or indirect contact with these infection sources, which then transmit the African swine fever. In trade, infected pork products, swill, alongside equipment and vehicles are the media of ASFV [27,28]. Epidemiological studies revealed three main routes of ASFV transmission: 46% was spread by unsterilised vehicles and workers, 34% by swill-feeding pigs, and 19% by trans-regional transportation of pigs and products [29]. Unsterilised vehicles and people are the main vectors in the spread of African swine fever, and strict control of vehicles and people entering farms is particularly necessary. The questionnaire survey revealed that most farms have strict control over people and vehicles, whereby foreign vehicles and people are not allowed to enter at will, and farms have special transport as well as vehicle decontamination areas, although this does not cover all farms, and a small number of farms are still loosely managed. In terms of feed and water, only 2.63% of farmers feed their pigs with slop, which is excellent. However, for drinking water, more than half of the farmers have not tested their water sources, which greatly increases the risk of contamination in the farms and poses a threat to the health of the hogs.Regarding the prevention and control measures of ASF, there is no commercialised vaccine at present, and the focus on prevention and control of ASF is “prevention”. The only control measures are to find infected pigs as soon as possible, to either cull or treat them harmlessly, or block and isolate them [30]. Chai Weihua et al. [31] discussed the measures to strengthen pig breeding management and disease prevention and control based on the three elements of infectious disease control: controlling the source of infection, cutting off the transmission routes, and protecting susceptible populations. Starting with “prevention”, represents cutting off all the paths through which the virus may spread. It is strictly forbidden for people outside the pig farm to enter, and the staff should be disinfected and protected when entering and leaving. Disinfect the materials and vehicles entering the pig farm, and when introducing breeding pigs from outside, it is necessary to set up an isolation and observation period before they become gregarious and attempt to adopt a population management mode of “all-in and all-out” to improve the biosecurity barrier. An Dong et al. [32] concluded that ASF should be prevented and controlled by isolation measures, standardised disinfection procedures for farms, training of farmers on biosecurity knowledge, and international border epidemic prevention. The biosecurity measures taken on the surveyed farms focus on cutting off the means of transmission. Biological control is carried out to prevent birds, rodents, and insects from entering the farm, and to disinfect and control the people and vehicles. In terms of pig management, 90.08% of the farms are zoned, with sows, feeders, and piglets kept separately. However, only 33.59% of the farms adopt “all-in, all-out” management practices for their pigs, which undoubtedly increases the risk for the farms not currently adopting such management practices, as those newly introduced pigs may carry viruses into the farm, increasing the risk of infection for pigs throughout the farm.In terms of the farmers’ characteristics, men are more aware of epidemic prevention and actively take epidemic prevention actions [33]. As the head of the family, men bear the main business responsibilities, thus, taking active epidemic prevention actions to avoid economic losses. The more educated the farmers are, the more inclined they are to take epidemic prevention measures [34]. The higher the educational background, the more knowledge you will receive, and the higher your ability to acquire information and learn new knowledge. When an epidemic occurs, you will take active defensive measures. Farmers with more years of farming possess rich farming experience, yet this also means that they will follow the old rules. Compared with farmers with fewer farming years, they pay attention to different disease prevention and control measures [35]. The larger the size of the farm when faced with an epidemic, the greater the losses suffered will be, having the intrinsic motivation to adopt epidemic prevention and control behaviours as well as the financial ability to support them [22].The relationship between risk perception and the adoption of conservation behaviours by farmers has been examined by scholars. The results of Valeeva’s study revealed the direct importance of the intrinsic risk characteristics of farmers for their adoption of risk management strategies [36]. Self-protective (risk-averse) behaviours, in general, directly contribute to the farmers’ decisions to exhibit more specific farm-protective behaviours. This is the exact opposite of the derived empirical results. Conversely, the results of the study indicated that farmers with a higher perception of animal disease risk tend to adopt bird-proofing measures, and those who prefer risk tend to disinfect their vehicles. This is consistent with the results of this paper. The reason for these different results is the different meanings of the measurement indicators. In the questionnaire, “high risk, high return”, “medium risk, medium return” and “low risk, low return” were used as options to measure the farmers’ risk preferences. The empirical results show that farmers with a ‘high risk, high return’ bias are more likely to adopt epidemic prevention behaviours. The outbreak of ASF in China led directly to a shortage of pig stocks and a sharp rise in pork prices [37]. On the one hand, restocking brings huge economic benefits to farmers, but at the same time, they are exposed to the risk of further contracting the ASFV. These farmers, who prefer “high risk, high reward”, are driven to replenish their farms by the huge economic gains, but at the same time take active measures to protect themselves against the risk of the ASFV.5. Conclusions and Recommendations5.1. ConclusionsMost of the farmers in the survey area are middle-aged or older males, mostly above 45 years of age, who have a primary school and junior secondary school education and have received technical training. Most of the farms can implement zoning management, and dedicated means of transport, and have built immunisation records. Nevertheless, there are still approximately 30% of the farms whose layout does not meet the ecological codes of construction, and most of them have not adopted an “all-in, all-out” management style.With respect to individual farmers’ characteristics, gender, education, and technical training had a significant impact on the adoption of biosecurity measures. Specifically, male farmers paid significant attention to biosecurity control in their farms. The more educated the farmers, the more active they were in adopting biosecurity measures; and farmers who had received technical training were more inclined to adopt biosecurity behaviour than those who had not.With respect to farming characteristics, the number of years of farming, the scale of farming, and the degree of specialisation had a significant effect on the adoption of disease prevention and control behaviours in the farms. Specifically, the more years of farming, the more likely the farm was to neglect biosecurity control. However, the larger the farm, the greater importance it placed on adopting disease control measures; and the more specialised the farm is, the more inclined it was to adopt disease prevention and control behaviours.With respect to risk awareness, the more risk-averse a farmer was, the more active they were in adopting disease prevention and control behaviours. Conversely, the more risk-seeking a farmer was, the more passive they were in adopting disease prevention and control behaviours. When farmers have an active awareness of disease prevention and control, they will adopt active disease prevention and control behaviours while reporting suspected epidemics found in neighbouring farms.5.2. Policy Recommendations5.2.1. Learning about Epidemic Prevention and Improving Professional SkillsFarmers’ knowledge reserves of farming techniques and animal epidemic prevention, their mastery of epidemic prevention measures, and their own scientific literacy, all have a direct impact on the epidemic prevention behaviours adopted by farmers. Therefore, it is important to provide farmers with training in epidemic prevention and control from multiple angles, subjects, and in a variety of ways. Knowledge of epidemic prevention and control should be promoted through the internet, APP, TV, and veterinary medicine dealers. Subjects such as animal husbandry and veterinary departments, scientific research institutes, breeding associations, and government departments should increase technical support for farmers and organise regular epidemic prevention training for farmers to improve their understanding of the principles of various epidemic prevention and control measures and their effectiveness, and to master the know-how of using various prevention and control measures to build the first barrier for epidemic prevention.5.2.2. Large-Scale Farming, Specialised FarmingPromoting large-scale farming models requires strengthening the promotion and guidance of large-scale farming. Regularly organised visits to large-scale farming demonstration farms for retail and small-scale farmers to raise their awareness of the importance of developing large-scale farming and carrying out biocontainment measures. Improve the degree of specialisation in farming. Promote semi-modern and fully enclosed management. Farms should strictly control the entry and exit of personnel and staff should strictly implement the bathing and disinfection entry system to avoid the risk of virus infection. Environmental disinfection is carried out regularly and quantitatively in pig houses. Establish a sound farm biosecurity prevention and control system to provide a healthier growing environment for pig breeding.5.2.3. Timely Dissemination of Information, Raising Risk AwarenessAnimal husbandry and veterinary departments should release epidemic information in a timely manner. By using new media technology to build a smart platform for animal epidemic prevention and control, timely epidemic information can be released, making it possible for farmers to learn about epidemic information through their mobile phones. Making farmers aware of the serious dangers of epidemics and raising their risk awareness will help increase their willingness to implement epidemic prevention and control measures. The government should vigorously publicise relevant laws and regulations such as the Animal Epidemic Prevention Law and the Code of Practice for the Disposal of Major Animal Diseases to raise the awareness of farming subjects to understand the law and their responsibilities and to know the rules and regulations.

animals : an open access journal from mdpi

10.3390/ani12030256

PMC8833700

Eastern grey kangaroos, like most wildlife, are facing an increasingly uncertain future under rapid climate change. How individuals and populations cope with extreme climatic events will influence their capacity to adapt and persist. Here, we analyzed how drought impacted eastern grey kangaroo populations by focusing on their body condition, demography, activity rates at water points, and the likelihood of parasitic infections. We found that body condition was lower as environmental conditions became more extreme and that fewer males in the population were observed. The proportion of juveniles within the population increased as more favorable conditions returned. Kangaroos with poor body conditions were more likely to become hosts to ticks, while higher parasite egg burdens in scats occurred in autumn. Our study has shown that the impacts eastern grey kangaroos face during climatic events such as drought can be severe and may have long-term consequences.

Extreme climatic events such as droughts and floods are expected to become more intense and severe under climate change, especially in the southern and eastern parts of Australia. We aimed to quantify the relationship between body condition scores (BCS), demography, activity rate, and parasitic infections of eastern grey kangaroos on a large conservation property under different climate extremes by employing camera traps established at artificial water points (AWPs). The survey period included a severe drought, broken by a significant flooding event. Climatic and environmental conditions were documented using remotely sensed indices of moisture availability and vegetation productivity. These conditions were found to affect all health and population parameters measured. BCS, juvenile proportions, and sex ratios were most correlated with 6-month lags in climatic conditions, while the activity rate of kangaroos at AWPs was most correlated with vegetation productivity. Ticks were mostly found on individuals with a poorer BCS, while the concentration of parasitic eggs in feces was higher in autumn than in spring. Our study offers a glimpse into some of the environmental drivers of eastern grey kangaroo populations and their health, information that may become increasingly important in today’s climate. It further emphasizes the importance of this knowledge for wildlife conservation efforts appropriate to managing the impact of climate change alongside other threats.

1. IntroductionThe global climate is changing at an unprecedented rate, leading to an increase in the frequency and intensity of extreme climatic events and posing significant threat to many wildlife species [1]. Natural phenomena such as droughts and floods are expected to become more intense and frequent as the world warms above 1.5 °C pre-industrial levels [2,3], due to an intensification of El Niño and La Niña events in many parts of the world [4]. On top of this, climate change is occurring in a world where wildlife are already facing mounting pressure from a range of anthropogenic activities, including habitat destruction and fragmentation, pollution, pathogens, and overexploitation [5]. The persistence of individuals and populations may hinge on their ability to adapt to extreme weather events caused by a changing climate [6], particularly as environments increase in aridity. The capacity for adaptation may disfavor species of temperate or mesic regions rather than species that are already arid-adapted. This may be especially noticeable for species in temperate regions with narrow climatic niches [7,8], but extreme events may also have an impact on the ranges of such species that currently extend into semi-arid regions [9]. Furthermore, the global climate is now changing so rapidly that many species simply do not have time to evolve resistance to it [10]. Understanding how wildlife are affected by events such as droughts and floods is therefore essential for addressing future issues in conservation [11].While population abundance and ecological dynamics are often impacted by environmental change and climate, so are the health and welfare of wildlife [10]. Direct impacts during heatwaves can result in mass death events [12,13], but the long-term and gradual impacts of drought can lead to immunosuppression, disease susceptibility, and reproductive and developmental decline [6]. Reduced health and welfare can lead to a decline in the reproduction rate, skewing demography towards older populations as individuals become more likely to reach advanced life stages without successfully reproducing [14]. While extreme conditions are more likely to affect a population, conditions with lower intensity repeated over longer periods can also affect the body condition of individuals [15]. Body condition is known to be closely linked to the survival and reproduction rate of many species [16,17]. Poor body condition can impair the immune system, putting individuals at risk of pathogenic and parasitic infections [10]. Individuals with lower body conditions can also become more vulnerable to predation as they become too weak to flee from predators, or choose to trade costly behaviors such as vigilance for more vital ones such as feeding [16]. While parasite infections are often the cause of poorer body conditions, poor body condition can also make individuals more susceptible to parasite infection [18]. Furthermore, as body condition declines, individuals can suffer from defective immune systems as well as lower levels of maintenance and overall function, further increasing the likelihood of parasitic infection [18,19]. When food and water becomes scarce during drought, wildlife will often congregate at resource points in larger numbers than usual, increasing the risk of parasite transmission and outbreaks [20,21].In Australia, drought is a pseudo-cyclic phenomenon that has occurred for millennia [22]. Under climate change, droughts in Australia are becoming more intense and lasting longer in southern and eastern parts of the country [3]. These temperate environments are likely to experience more frequent droughts and increasing aridity. While most anthropogenic structures exert pressure on wildlife, artificial water points (AWPs) may offer much-needed water [23,24]. Similar effects have been observed for water points dug by wildlife themselves [25]. Large herbivores, such as kangaroos, have been observed to utilize AWPs, particularly during hot periods when the vegetation is drier [26]. However, even though kangaroos may utilize AWPs to drink, there is no evidence that AWPs influence densities or assist with population growth [26,27]. While dingoes can suppress kangaroo populations through top-down regulation [28], in the absence of dingoes, kangaroo abundance and population dynamics are driven by primary productivity and the availability of food sources [29,30,31]. Despite this, access to water from AWPs and natural water sources is likely important for individuals in managing their health during drought [31].Here, we aimed to detect changes in populations and health parameters of eastern grey kangaroos (Macropus giganteus) during two extreme weather events, namely, drought followed by flooding. To do so, we monitored body condition, demography, presence of parasites, and activity rates of eastern grey kangaroos at AWPs on a wildlife reserve in semi-arid south-western Queensland over a period of 18 months. Long-term changes in weather conditions were captured using the standardized precipitation evapotranspiration index (SPEI), which measures the driving effects of temperature on water demand [32], while shorter-term changes in the amount of live green vegetation were captured using the normalized difference vegetation index (NDVI) [33]. NDVI has previously been used to analyze the relationship between vegetation and wildlife performance [34].We expected that the body condition of individuals would be lower during drier months (i.e., drought), and the use of AWPs would increase due to the lack of moisture in forage. We also predicted that higher body conditions would be observed in times of higher SPEI and NDVI, and that demographic shifts would occur as juvenile kangaroos become more frequent under favorable conditions due to increases in survival rates. The proportion of individuals with external parasites was expected to increase with the decreasing body condition and health of individuals.2. Materials and Methods2.1. Study SiteThe study was conducted on the 480 km2 Mourachan Conservation Property (MCP) in south-western Queensland, near the township of St George. This private semi-arid rangeland reserve, managed by Australia Zoo, is surrounded by cattle and sheep farms. Although a small number of cattle are run on one section of the property under wildlife friendly principles [35], the remainder is maintained as a conservation reserve where kangaroos and other wildlife are protected from persecution [36]. The MCP includes four macropod species: eastern grey kangaroos, red kangaroos (Macropus rufus), black wallabies (Wallabia bicolor), and red-necked wallabies (Macropus rufogriseus). Eastern grey kangaroos are distinguished from the only other large macropod, the red kangaroo, by body color and facial markings. Southern Queensland and most of Australia was in drought for the majority of 2019 due to El Niño, but weather conditions changed at the beginning of 2020, and the MCP was inundated with flood waters in late January 2020. Furthermore, 2019 was the driest year on record for the MCP (Figure 1).2.2. Camera TrapsCamera traps (Strike Force HD Pro X, Browning, Morgan, UT, USA) were setup in November 2019 at 15 AWPs distributed across the property (Figure 2). At the time of installation, they were the only available sources of water on the reserve. A total of 40 cameras were used to capture motion-sensed photos from November 2019 to April 2021, leading to a total of 10,736 camera trap days. Some cameras were lost due to being completely submerged during flooding, while some temporal breaks in data capture occurred because of COVID-19 border restrictions that limited access to the site.2.3. Body Condition Score and DemographyEastern grey kangaroos were given a score between 1 and 5, depending on their visual body condition, using a similar method described by Johnston et al. [38]. A subjective body condition scoring system was developed, and example photos were collated for each score. Examples can be found in Appendix A Table A1. Scores were described as: 1 (emaciated), 2 (very thin), 3 (thin), 4 (average/healthy), and 5 (optimal). Images used to score BC were selected based on their quality, mainly relying on the full or most of the body being clearly visible. If more than one individual was present on an image, all kangaroos were scored. Only one image per event (pictures taken within 5 min of each other) was scored. A total of 3993 camera trap images were used. The demographic class of each scored kangaroo was also recorded using the classification references described by Austin and Ramp [39], using the following categories: pouch young, young-at-foot, sub-adult, small adult, medium adult, and large adult. The sex of each individual was determined using the description in Jarman, et al. [40].2.4. Activity RateThe activity rate of kangaroos was measured to analyze how often they use artificial water points, both during drought and under more favorable conditions. We recorded the total number of kangaroo events per month, defining events as a series of camera trap images captured within 5 min of the previous image, where one or more kangaroos were visible. The number of kangaroos visible on an image did not influence the number of events (i.e., images within 5 min of each other with more than one kangaroo were still counted as a single event). The activity rate here, therefore, does not represent the number of individuals within the population, only how often kangaroos were observed using the AWPs.2.5. ParasitologyThe presence/absence of parasitic ticks was recorded from camera trap images. The images used to analyze tick presence were the same images used to measure BCS due to their higher quality. Rather than analyzing the relationship between tick-infected individuals and environmental conditions, we focused on the relationship between the presence of ticks and the BCS of the kangaroos.Kangaroo scat samples were used to estimate the average Fecal Egg Count (FEC) of the population during drought (2019) and post-flood (2021). Where possible, fresh scat samples were collected directly after observing foraging eastern grey kangaroos. When no kangaroos were present, samples were collected based on the size and shape of the scat [41]. We validated that the scats collected in November 2019 were from eastern grey kangaroos through genetic analysis [36]. Samples were collected from each AWP in November 2019 (spring) and stored at −20 °C. It should be noted that storage at −20 °C can lead to some loss of eggs in the scats due to biological degradation [42]. Samples were also collected in April 2021 (autumn). FECs were performed by mixing 3 g of a scat sample in 60 mL of saturated salt solution. Eggs were counted using a Whitlock Universal 4 chamber worm egg counting slide (J.A. Whitlock and Co, Eastwood, Australia), following the methods described by Gordon and Whitlock [43].2.6. Statistical AnalysisGeneral linear mixed models (GLMM) were used to analyze the effects of climatic conditions and primary productivity on the body condition, demography, and activity rate of the population. Models for body condition and activity rate used a Poisson distribution, while demography used a binomial distribution. SPEI values were used to represent long-term trends and were, therefore, used with lags of 3, 6, and 12 months to represent a quarter of a year, half of a year, and a full year. NDVI was used to represent more immediate and short-term changes in health and population parameters related to primary productivity and was, therefore, used with a monthly lag. SPEI time series was downloaded from the SPEI Global Drought Monitor using the MCP as a single grid cell [44]. We used a polygon of the MCP as a mask to select tiles and obtain monthly NDVI values from the MODIS Terra NDVI composites [45]. GLMMs (generalized linear mixed model) with binomial distribution were also used to determine whether the body condition of kangaroos influenced the likelihood of tick infection. A t-test was performed to compare the Fecal Egg Count from scats between November 2019 and April 2021.All GLMMs were performed in R v4.1.1 (R Foundation for Statistical Computing, Vienna, Austria) [46] using the “glmer” function of the “lme4” package (Version 1.1-27.1) [47].3. Results3.1. BCS and DemographyThere was a strong positive relationship between SPEI and BCS, with the 3-month lag (SE = 0.011, z = 13.93, p = <0.001, and r = 0.6) and 12-month lag (SE = 0.009, z = 14.86, p = <0.001, and r = 0.57) showing similar patterns (Figure 3). However, a stronger relationship was observed between BCS and SPEI with a 6-month lag (SE = 0.013, z = 15.35, p = <0.001, and r = 0.87). Similarly, NDVI was also shown as being significantly correlated with BCS (SE = 0.156, z = 12.046, p = <0.001, r = 0.8) (Figure 3).SPEI with a 3-month lag (SE = 0.047, z = 8.612, p = <0.001, and r = 0.54) and a 12-month lag (SE = 0.037, z = 10.11, p = <0.001, and r = 0.62) were found to have the strongest effect on the sex ratio of the population, with fewer males observed using AWPs during drier conditions. SPEI values with a 6-month lag (SE = 0.051, z = 7.527, p = <0.001, and r = 0.14), and NDVI values (SE = 0.634, z = 5.643, p = <0.001, and r = 0.21) were found to have the weakest relationships with sex ratio. The ratio balanced out with the return of more favorable conditions (Figure 4A,B).The monthly adult to juvenile ratio of the MCP population visiting AWPs had a relatively weak relationship with SPEI at 3-month lags (SE = 0.065, z = 7.896, p = <0.001, and r = 0.18), and NDVI (SE = 0.863, z = 5.48, p = <0.001, and r = 0.14). Stronger effects came from SPEI with longer term lags, as shown by values with a 6-month lag (SE = 0.079, z = 11.824, p = <0.001, and r = 0.23) and a 12-month lag (SE = 0.081, z = 14.767, p = <0.001, r = 0.31).3.2. Activity RateActivity rates (average number of events per month) of eastern grey kangaroos significantly declined at AWPs as weather and vegetation indices increased after the drought broke (SPEI with a 3-month lag SE = 0.013, z = −72.00, p = <0.001, r = −0.72, a 6-month lag SE = 0.013, z = −71.286, p = <0.001, r = −0.75, a 12-month lag SE = 0.01, z = −63.667, p = <0.001, r = −0.59 and NDVI SE = 0.194, z = −69.01, p = <0.001, and r = −0.77). This shows that drier conditions increased the likelihood of kangaroos accessing AWPs (Figure 5A,B).3.3. ParasitologyThe number of kangaroos infected by ticks increased as their body condition declined (SE = 0.181, z = −3.712, p < 0.001, r = −0.4). Ticks were more likely to be observed on individuals with poorer body conditions. Fecal egg counts for April 2021 (post-flood) were found to be significantly higher than November 2019 (drought) (df; degrees of freedom = 41.971, t = −3.644, p = 0.001) (Figure 6A,B). All eggs observed were identified as Strongyle eggs.4. DiscussionWe found that the body condition of eastern grey kangaroos was negatively correlated with environmental factors explained by climatic conditions, as measured by long-term trends in moisture availability (SPEI) and primary productivity (NDVI). As both SPEI and NDVI increased, leading to more favorable conditions, the body condition of kangaroos increased. We also found that the SPEI data with a 6-month lag had the strongest relationship with BCS. However, this does not suggest that NDVI values cannot be used to predict a relationship between BCS and primary productivity since a correlation of r = 0.8 was found between NDVI and BCS. In other species, such as the roe deer (Capreolus capreolus) and red deer (Cervus elaphus), NDVI was found to be a strong predictor of body mass [34,48]. NDVI has also been shown to influence the body condition and reproductive timing of the African buffalo (Syncerus caffer) and African elephant (Loxodonta africana) [49]. Poor body condition of animals during drought is thought to be one of the major causes of deaths [50]. For example, Knight [50] found that 90% of fresh carcasses (of four species of ungulates) found during droughts in the southern Kalahari had poor body conditions. In Australia, the body condition of the southern hairy-nosed wombat (Lasiorhinus latifrons) is closely related to food availability and quality [51]. Kangaroos often choose greener, more nutritious vegetation, resulting in high mortality rates during drought when such resources are not available [52]. Poor body condition often leads to adverse health effects such as lower immune responses due to impaired immune systems, increasing infection likelihood while lowering resistance [10]. This can also impair nutrient uptake and lead to poor nutritional status [53]. Nutritional and hydrological stresses, such as drought, are known to affect other processes such as growth and reproduction, over and above increased mortality during these events [54].The sex ratio of eastern grey kangaroos visiting AWPs at Mourachan shifted during drier months, with fewer males being observed when environmental conditions were poorer (Figure 4A,B). However, the increase in males observed post flood did not result from juvenile male births but from the adult individuals. This suggests that some of the males were alive during the drought but simply did not use the water points as much as the females did. Previous studies on the mortality of kangaroos during droughts have reported that males had a higher mortality rate than females [55,56], which manifested itself through all age classes and is thought to have been caused by the differences in energetic costs, body size, and mobility between the sexes [55]. As the sex ratio of eastern grey kangaroos in our study returned quickly to parity and was not driven by recruitment, it is likely that behavioral decisions may explain this trend.The proportion of juvenile kangaroos within the population increased as environmental conditions improved post drought and exhibited a stronger relationship with longer-term data (SPEI 6 and 12) than short-term (Figure 4D). Reproduction in kangaroos is driven by biological and ecological constraints. Gestation period in this species is approximately 36 days [57] and is influenced by mother health and climatic conditions. Reproduction is more successful when females have a high BCS [52]. Juvenile kangaroos represented below 2% of observations at the height of the drought and were absent until July 2020, four months after the floods (Figure 1). For many mammal species, the energy requirements of reproduction can lead to a reduction in female body mass [58], requiring females to feed more to compensate for the energy allocated towards reproducing [59]. Further, eastern grey kangaroos are seasonal breeders and embryonic diapause is rare [60], meaning that response to favorable conditions may be delayed as individuals take time to improve body condition and synchronize with typical reproductive cycles. Under favorable conditions at Mourachan that began in autumn 2020, the proportion of juveniles gradually increased and peaked at the beginning of summer 2020/2021, where they comprised around 25% of observations. However, despite generally good conditions in the summer of 2020/2021, observations of juveniles declined to around 5% as summer progressed, possibly due to juvenile mortality as they are more vulnerable to higher temperatures and a lack of moisture in vegetation than adults [61,62]. In previous studies, juveniles and older kangaroos have been shown to be the first to die during droughts [56].For arid-adapted macropod species, such as the red kangaroo (Macropus rufus), rearing juveniles can cost the mother as much as 50% of her own daily energy requirement for maintenance near the end of the juvenile’s pouch stage [52]. If environmental conditions are favorable and the condition of mothers is high, up to 85% of pouch young can make it to the weaning stage [52,63]. During drought, the costs of caring for young can often surpass the daily energy intake of the mother, leading to higher rates of juvenile mortality [64]. Red kangaroo mothers have the adaptive advantage of suspending embryonic development (embryonic diapause) while environmental conditions are poor [52], a trait that has evolved to promote survival under arid conditions [60]. Without this advantage, grey kangaroos are less resilient to drought conditions. It remains to be seen whether eastern grey kangaroos under increasing aridification can utilize diapause more commonly than current evidence suggests. What is apparent, however, is that red kangaroos may be physiologically better suited to survive droughts than eastern grey kangaroos [65].We also found that environmental conditions affect the activity rate of kangaroos at artificial water points, with more events occurring during drier months when less water and live green vegetation was available across Mourachan (Figure 5). Eastern grey kangaroos access water points to drink before returning to a more favorable grazing site [26], while food availability is a greater driver of dispersion and density. Kangaroos living in rangelands often drink at AWPs, especially during summer when vegetation availability and moisture content are low. As growing vegetation became available after the flood, we found that kangaroos spread out and were less reliant on AWPs. Wildlife activity in semi-arid and arid ecosystems is often focused around sources of water, but while AWPs were originally thought to increase the abundance and density of kangaroos [66], it has now been accepted that AWPs do not influence their density [24]. In accordance with this, we suggest that changes in activity rates at AWPs at Mourachan were likely a response of changing physiological requirements due to the lack of moisture in vegetation, a circumstance that may become more challenging as droughts are predicted to become more frequent and intense.Tick infections were increasingly prevalent on eastern grey kangaroos with lower BCS (Figure 6A). The presence of ticks may contribute to declines in health, depending on the type of tick and the location on the body, but poor health may also make kangaroos more susceptible to ticks. One explanation for the latter relationship is that the grooming rate of the unhealthy kangaroos can decline. Grooming costs can include loss of saliva, leading to a loss of water, as well as a reduction in vigilance rates [67]. When in poor physical condition, various species will often reduce behaviors such as grooming to focus their remaining energy on more vital activities such as foraging [67]. This gives parasites, such as ticks, the opportunity to attach themselves to hosts without the risk of being removed, increasing the likelihood of infection [18]. Eastern grey kangaroo mothers often groom their young [68], and we did not detect any ticks on any juveniles we observed, which could be due to the mother removing the ticks while grooming the young. Although we could not identify ticks to species level in our study, three species of ticks in Australia have been confirmed as causing paralysis in species such as kangaroos [69]. The prevalence of ticks on kangaroos, and their impact on the condition, health, and resilience to a changing climate, remains unknown.The fecal egg count data revealed that the number of eggs per gram of scat was significantly higher in autumn than in spring, suggesting cooler, wetter conditions could be more favorable to parasites such as strongyles (Figure 6B). Kangaroos naturally carry a variety of gastrointestinal parasites [70]; however, strongylid nematodes are the most abundant in eastern grey kangaroos [70,71]. Juvenile kangaroos are the most at-risk, especially from trichostrongylid nematodes such as Globocephaloides trifidospicularis. Parasite infections in eastern grey kangaroos are more prevalent in winter during wetter conditions [72]. While the post-flood samples were not collected in winter, the flood would have created conditions more suited to parasites than the drought [73]. Floods can also bring in more parasites such as Fasciola hepatica, which use freshwater snails as intermediate hosts [73]. While we cannot attribute differences in parasite load to drought/flood cycles because we were unable to replicate sampling to accommodate seasonal or annual trends, we nevertheless note the potential for changes in patterns of parasite prevalence to impact on the disease burden of kangaroos, which may be heightened by increasing extreme conditions.5. ConclusionsThe impacts on the health and demography of eastern grey kangaroos identified in this study are only some of the effects caused by environmental events such as droughts. While droughts have been a part of Australia’s climate for thousands of years [22], predictions of increasing prevalence are concerning. Droughts in the southern and eastern regions of the country are expected to last longer, while reaching more extreme intensities due to the intensification of El Niño and La Niña events [4]. Much of the wildlife in Australia has evolved with droughts; however, rapid changes in global climate may reduce the resistance of many species to droughts [10]. Species such as kangaroos may suffer from poorer body conditions, leading to an increased risk of infections by parasites [18]. Lower body condition could also lead to lower birth rates due to the poor fitness of females, while juveniles may have higher mortality rates [52]. These challenges, along with current threats from human activities [61,74], may contribute to further declines in kangaroo populations. From a global perspective, climate change causes new challenges for wildlife conservation. It is therefore essential to understand how wildlife is affected by climate change to address this issue [11]. To improve our knowledge of the impacts of climate change on eastern grey kangaroos, longer term studies located in different ecosystems would be required to cover more populations. While our study showed a contrast between drought and flood, it only covered one drought/flood cycle; a longer project would, therefore, allow for the comparison of multiple cycles and could further clarify patterns we have found. Studies covering a longer period would also allow for the collection of data on parameters such as the growth rate and survival of juvenile kangaroos, while also enabling yearly comparisons of the reproduction rate, BCS, and demography of populations.

animals : an open access journal from mdpi

10.3390/ani11082338

PMC8388702

The relationship between animal welfare and housing conditions is still a matter of debate. The present study aimed to evaluate animal welfare of undocked heavy pigs from the same farm, raised in buildings with different ventilation systems, i.e., mechanical and natural, throughout the fattening period (90–160 kg average weight). Ventilation efficiency was evaluated using computational fluid dynamics. Results showed that overall pigs raised in the mechanical ventilated building were in a more positive affective state. Despite that, with hot temperatures, the higher occurrence of pig soiling indicated heat stress and consequent welfare impairment. The higher frequencies of dog sitting behavior also indicated worsening of welfare conditions in the middle–late phases of fattening, likely imputable to the lack of stimuli and boredom in the pigs.

The present study aimed to evaluate animal welfare of pigs from the same farm, raised with two ventilation systems. The study involved 60 pens of fattening pigs, raised in two buildings: one naturally ventilated (NV) and the other mechanically ventilated (MV). Pigs were assessed on three observation days: at 40 kg (T1), 100 kg (T2), and 160 kg (T3) of live weight. Animal-based measures were used such as qualitative behavioral analysis (QBA), behavioral measures (BMs), and lesion and health measures (LHMs). Housing conditions (HCs) measured at each observation day were the number of pigs per pen, space allowance, temperature, light, and CO2. The association study was performed using a general linear model and analysis of variance. Ventilation effect was analyzed by performing computational fluid dynamics. Results showed that overall pigs raised in the MV were in a more positive affective state. Despite that, with hot temperatures, the higher occurrence of pig soiling indicated heat stress in pigs and consequent welfare impairment. The higher frequency of pigs showing dog sitting behavior at T2 and T3 suggest welfare worsening in the last phases of fattening. The study concludes that ventilation system influences animal behavior and overall animal welfare, especially during the warmer season.

1. IntroductionAnimal-based measures (ABMs) are considered the most reliable indicators to assess the welfare status of an animal and to identify the risk factors in the management and environmental conditions [1]. They include a large variety of indicators such as behavior, clinical signs (e.g., skin lesions, pathologies), and physiological and productive parameters. ABMS allow measuring how a single animal (or a group of animals) reacts to environmental and management stressors. One of the main stressors in intensive pig herds is the fluctuation of the (indoor) environmental temperature, particularly when it overcomes the thermoneutrality threshold of the animal [2]. Further causes of environmental stress can be attributed to excessive relative humidity (rH) and harmful gas concentrations (e.g., CO2 and NH3) [3,4]. Concerning temperature control and gas removal, ventilation efficiency plays the main role, and it is mainly dependent on the ventilation system in the barn [5]. At present, pig barns in the Mediterranean area are mainly wind-driven and, thus, naturally ventilated, while mechanical ventilation systems are less common [6]. In naturally ventilated buildings, the external wind influences the indoor velocity magnitude and distribution [7]. In these systems, the accurate control of the indoor conditions is not always feasible, especially on warm days, when the windows and vents of the barns are fully open [8]. In mechanically ventilated barns, the ventilation efficiency can vary depending on the operational conditions of the ventilation system [9]. In both natural and mechanical ventilation systems, the ventilation efficiency can vary depending on the geometry of the piggery structure; therefore, it should be carefully evaluated.Inadequate ventilation, leading to subsequent changes in temperature, humidity, and presence of gas and dust [10,11,12], has been found linked to pig behavior, health, and physiology [9]. For some ABMs, the effects of poor ventilation on animal health are well known; for example, an insufficient air exchange can increase the occurrence of respiratory disease and thermal stress [7,13], while contrasting results have been observed with regard to animal behavior. In the case of aggressive behaviors and outcomes such as lesions, specific studies are lacking. Some studies reported that heat stress can lead to the development of aggressive behavior and consequent skin lesions [14]; other works showed that high CO2 concentration might induce higher inactivity rates of pigs and could increase the risk of overloading with subsequent occurrence of lesions in the middle area of the body and/or the prevalence of bursitis due to the prolonged contact of bone prominence with the floor [10].Furthermore, tail biting and relative tail lesions have been hypothesized to be influenced by the ventilation conditions (i.e., magnitude air velocity, air direction, and air exchange) and typologies (i.e., natural or mechanical). Hunter et al. [15] observed a higher presence of lesions in the case of mechanical ventilation than in the case of natural ventilation. Therefore, it is important to consider that, according to Hunter et al. [15], the natural ventilated building considered in the study was provided with straw litter, which is considered the “gold standard” to prevent tail biting behavior, and this aspect might have biased the results. On the contrary, the study by Scollo et al. [16] reported higher frequencies of tail lesions in the farms adopting natural ventilation.Lastly, no studies have been found on the effect of ventilation on the emotional state of fattening pigs. Today, it is worth noting that animals need a positive emotional state to be in a positive welfare condition [17]; therefore, ABMs to assess pig emotional state have been proposed, such as qualitative behavioral assessment (QBA), tail posture, and tear staining [18,19]. It has been questioned if the impact of ventilation, dust, and air quality could be a confounding factor for some emotional state indicators [20,21]. It has been hypothesized that temperatures out of the acceptable range can influence tail posture, which is also an indicator of tail biting behavior in a group of pigs, and pig emotions, leading to misinterpretation of the measure itself. Similarly, high dust and gas concentrations can mislead the interpretation of tear staining as an indicator of emotion in pigs, due to inflammation of the eye and conjunctiva by poor air quality [20].Lastly, previous research found evidence that the age and weight of the pigs might also influence their behavioral response toward housing conditions, temperature, and air quality [22], stressing the need to consider this variable when assessing the effect of ventilation on ABMs in a productive cycle.Being able to identify the relationships between ABMs and ventilation conditions might be helpful to prevent the occurrence of harmful social behaviors and, therefore, increase pig welfare conditions.This study hypothesized that there would be significant associations between housing condition management and behavior, lesions, and emotional state in undocked pigs during the fattening phase. Therefore, the first aim of the study was to quantify and qualify the main welfare issues of the pigs. The second aim was to define and compare the ventilation conditions in two different buildings of a case study farm, using computational fluid dynamics (CFD) simulations, and to investigate possible relationships between welfare indicators and housing and ventilation conditions, to evaluate how these can impact pig welfare.2. Materials and MethodsIn the study, a list of animal-based indicators of both negative and positive welfare status was tested in two groups of finishing pigs reared in a barn located in northern Italy. The two groups were housed in two different buildings, which were selected for the study because they have analogous management conditions but are characterized by different ventilation conditions (natural vs. mechanical ventilation).2.1. Livestock, Building, and AnimalsThis study was conducted in a commercial pig farm located in northern Italy, in the heart of the so-called Food Valley. At the beginning of the observations, a face-to-face interview with the farmer was conducted by two authors, P.T. (Paolo Trevisi) and M.V., using a questionnaire (Supplementary File S1). The purpose of the questionnaire was used to determine the overall rearing condition of the farm and main management practices. Briefly, the farm was a fattening farm belonging to a specific supply chain. Therefore, all the pigs came from the same group, genetic line, and overall housing conditions. All pigs were left undocked. Pigs at the beginning of the fattening period weighted 40 kg and were raised for 4 months until 160 kg of live weight. The farm had three employees caring for the animals, and there were written procedures about the prevention of tail biting and emergency culling.2.2. Description of the BuildingThe farm was originally designed to host dairy cows; however, in the 1990s, it was restored and converted to house finisher pigs. An aerial view of the particular shape of the building, similar to a “star” with six buildings labeled B1–B6, is shown in Figure 1 together with some representative pictures, both internal and external. The central core of the farm is a heptagon with an edge dimension of about 20 m; the inner height is 7.50 m at the eaves and 9.40 m at the top of the central dome. The buildings are 72.30 m long. Buildings B1, B2, B3, and B6 are 17.70 m wide, whereas B4 and B5 are 19.80 m wide.The present study focused on B3 and B5. The two different buildings were selected since they can be considered representative of the two different ventilation conditions of the farm. B3 is naturally ventilated and is (north–south) oriented. Natural ventilation (NV) in B3 is obtained through wall windows and ridge vents running all along the building length. B5 is SE–NW oriented and is equipped with a mechanical ventilation system. The mechanical ventilation (MV) is realized by means of six longitudinally equally spaced chimney fans (Fancom, The Netherlands). Furthermore, the wall windows on the lateral longer side allow for the entering of the fresh air.Each building hosts around 700 pigs (for a total of 4200 pigs in the farm). The buildings hosting the pigs are characterized by two lines of pens (19 + 19) with a central service corridor of about 0.90 m running along the whole length of the building. Pens have about 25% of the surface characterized by a slatted floor and 75% characterized by a full slab. Each pen is provided with a trough for liquid feed (provided twice a day) and two nipple drinkers. Environmental devices are constituted by a chain and a chain with wood placed in the middle of each pen.2.3. Sampling Procedure and Investigated ScenariosThe present study involved a total of 60 pens, 30 in the NV building (B3) and 30 in the MV building (B5). The average number of pigs/pen during the study was 28.23 ± 2.77 SD. The animals in each building were assessed on three observation days (T), for a total of 1694 pigs observed:T1: 1 week after their allocation in the building, at about 40 kg of body weight;T2: 1 month after the allocation, at about 90 kg of body weight;T3: the day before loading to the abattoir, at about 160 kg of body weight.At each observation point (i.e., observation day), 10 pens per building were randomly chosen, according to the Welfare Quality protocol [23].The three observation days (T1, T2, and T3) for the two buildings (B3 and B5) led to six different scenarios in which data on housing conditions and data on animal-based measures were collected. Moreover, the same six scenarios were numerically investigated by means of CFD simulations.2.4. Experimental Measures2.4.1. Animal-Based MeasuresThe ABMs were recorded on each observation day. Full references and explanations of the measures are reported in Table 1.Measures were divided into qualitative behavior assessment (QBA), behavioral measures (BMs), and lesions and health measures (LHMs). The same observer (i.e., the co-author M.V., with 5 years of experience with ABMs and trained to use the Welfare Quality protocol) recorded all the data. QBA was observed between 9:00 and 10:00 am and consisted of four observations (5 min each) for a total of 20 min. Then, the data were reported on a 125 mm scale and multiplied by the coefficients indicated in the protocol [23] to calculate the QBA score. BMs were evaluated between 10:00 am and 11:00 am with the direct observation of all pigs in each pen, three times per pen, for 5 min per observation. Then, the average of the three observations was calculated. Behavioral measures consisted of two types of observation: categories of behavior as described in the Welfare Quality [23], and abnormal behaviors or stereotypies.Categories of behavior included “inactive behavior”, “social behavior”, “exploratory behavior”, and “other active behavior”, as detailed in Table 1. The frequency of “social”, “exploratory”, and “other active behaviors” was determined by the total active behavior in each pen. The frequency of “inactive behavior” was calculated as a function of the total behavior observed, as explained in the Welfare Quality protocol for pigs [23]. Observed stereotypies and abnormal behaviors were recorded (i.e., tail biting, ear biting, dog sitting, bar biting), and they were calculated as the percentage of the mean of animals exhibiting the behavior (Ab) over the total number of animals (Atot) in the pen [24]. Ab/Atot × 100 (%).(1)Then, the sum of all stereotypies and abnormal behavior was calculated for each pen.LHMs were assessed in the afternoon on a sample of 15 pigs/pen. The assessment was carried out inside the pen at a distance of 0.5 m from the pig, using a headlight when necessary. Only the left side of the pig was observed. Skin lesions were scored in each pig using a score [23] ranging from 0 to 2 (i.e., 0: up to 4 lesions, 1: from 5 to 10 lesions; 2: more than 11 lesions); then, the most frequent (i.e., prevalence) score was calculated and considered for each pen. The lesion score, ranging from 0 to 200 [24], was calculated for each monitored area as follows:prevalence of lesion with a score of 1 + (2 × prevalence of lesion with score of 2).The same formula was used to calculate the dirtiness and tear staining index.Other LHMs were recorded using a Y/N score (where Y denotes presence, and N denotes absence), and the prevalence of pigs having a Y score was calculated in each pen [24].2.4.2. Housing Condition MeasuresAt each observation day, the most relevant parameters characterizing the housing conditions (HCs) were measured. These were light intensity, temperature, CO2 concentration, stocking density, and dustiness. Light intensity was measured using a Mini Light Meter (UNI-T UT383, Dongguan City, China). The temperature was recorded with a Datalogger (UNI-T UT330C USB, Dongguan City, China). CO2 concentration was measured with an IR sensor using a XAM8000 Multigas Detector (Dräger, Lübeck, Germany). The area of the pen was calculated using a Laser Distance Meter (Extech DT40M, Nashua, NH, USA), excluding the feeding area, and then divided by the number of pigs to obtain the stocking density. Light intensity, temperature, and CO2 were recorded at the pigs’ eye level as the average of three points in the pen: the corner closest to the center of the building, in the middle of the pen, and the opposite corner closer to the external wall.2.4.3. Statistical Analysis of HCs and ABMsAll statistical analyses were performed using R software [29]. The statistical unit was a pen. All observation days were considered separately. Descriptive analyses of ABMs were performed using the psyc.ir package [30]. Frequencies of behavior or LHM prevalence (considering the sum of scores of 1 and 2) showing a prevalence below 5% were not submitted to further statistical analyses. A general linear model (GLM) was carried out for the two buildings on the factors of the HCs, BMs, and LHMs intended as dependent variables using the building as a factor (independent variable). The GLM procedure was performed using the lme4 package [31], and the chi-squared test was used to evaluate the differences between the two buildings (lsmeans package, [32]). QBA descriptors were subjected to principal component analysis (PCA) using the FactoMineR package [33]. Statistical significance was set at p ≤ 0.05.2.5. Computational Fluid Dynamics SimulationsThe three-dimensional distribution of the ventilation conditions of buildings B3 and B5 was assessed using CFD simulations. The CFD simulations considered the model of the whole geometry of the pig farm, including the surrounding buildings, to take into account the interactions between the different structures. The geometrical model was developed in Autodesk Inventor [34], and the CFD analyses were carried out in VENTO AEC 2020 [35]. The geometrical model of the buildings is depicted in Supplementary Figure S1.CFD analysis is based on the governing fluid dynamics equations (continuity, momentum, and energy). The general Navier–Stokes equation, with Boussinesq approximation that relates the Reynolds stresses and velocity gradients through the eddy viscosity, has the following form:(2)ρUj(∂Ui∂xj)=−∂P∂xj+∂∂xj[(μ+μt)∂Ui∂xj].These models are also called eddy viscosity models and are classified on the basis of the number of transport equations. The model, chosen for the simulations, was the two-equation k–ε standard model, where the equations for k, kinetic energy per unit mass of the turbulent fluctuations, and ε, dissipation rate, are as follows:(3)∂k∂t+Uj∂k∂xj=μtρS2−ϵ+∂∂xj[1ρ(μ+ μtσk)∂k∂xj], (4)∂ϵ∂t+Uj∂ϵ∂xj=ϵk(C1ϵμtρS2−C2ϵϵ)+∂∂xj[1ρ(μ+μtσϵ)∂ϵ∂xj], where σk, σϵ,C2ϵ, and C1ϵ are experimental constants available from the literature [36].The boundary conditions (i.e., outdoor temperature, relative humidity, wind magnitude, and wind direction) for each one of the six scenarios considered were defined according to the data recorded by the weather station of ARPAE, placed in Rolo (RE), located only 5 km from the farm. They are summarized in Table 2. In the simulations, the reference wind velocity profile was defined by the following logarithmic profile:(5)u(z)=uKlog(z−dz0), where u(z) is the average wind speed at height z above the ground, u is the friction velocity, K is the von Karman’s constant (assumed equal to 0.40), d is the displacement length, and z0 is the aerodynamic roughness (in m).Moreover, for the building with mechanical ventilation, each chimney was defined in the model as a pressure–volume source with a pressure gradient of 16.8 Pa/m obtained from the datasheet of the fans (Fancom, The Netherlands). The mesh was selected after a grid independency study based on four different grids in terms of cell number. The final grid adopted for the analyses was characterized by 9 × 106 million cells.A preliminary experimental campaign was conducted on the farm to collect the air velocity magnitude using a hotwire anemometer (Delta Ohm, Italy) with an uncertainty of 0.01 m/s, to validate the numerical model for both buildings. The results of the validation process are shown in Supplementary Figure S2. The relative mean square error (RMSE) results were equal to 0.003 m/s for the natural ventilated case and 0.048 m/s for the mechanically ventilated building, confirming the limited difference between experimental and numerical results.3. ResultsDuring the study, clinical observations were carried out by the farm veterinary and the coauthors P.T. (Paolo Trevisi) and M.V.; no infective disease occurred, there was no need for antibiotic or other veterinary treatment, and the animals were in overall good health status.3.1. Animal-Based Measures3.1.1. Qualitative Behavior AssessmentConsidering the three observation days, the QBA score (average ± SD) resulted equal to 16.8 ± 2.2 and 27.8 ± 20.6, respectively, for building B3 with NV and building B5 with MV. The main difference was obtained in the first assessment (T1), where NV had a score of 17.8 and MV had a score of 51.07. The PCA analysis showed that the first and second dimensions (Dim) together explained 71.9% of data variance. Dim1 accounted for 47.6% and Dim2 accounted for 24.3% (see Figure 2).Figure 2 shows that Dim1 accounted for the observation day, while Dim2 accounted for the building. The output of the PCA is reported in Table 3. In general, considering the two buildings (Dim2), NV showed overall higher arousal and negative emotional states (i.e., tense, irritable, agitated) than MV, where the animals showed more positive state and lower arousal signs. On the other hand, considering the effects of the observation time (Dim1), pigs were perceived as being in a more positive emotional state (i.e., active, relaxed, enjoying, playful, positively occupied, lively, content, happy) at T1 and a negative emotional state (i.e., bored, aimless, distressed and listless) at T3.3.1.2. Behavioral MeasuresDifferent pig stereotypies were observed during the trial. They are summarized in Figure 3. The observed stereotypies show that belly nosing and tail biting were more frequent at T1 and T2, while dog sitting and licking were more frequently observed at T2 and T3, in both buildings. Dog sitting was the most evidenced stereotypy overall (see Figure 4).Because the prevalence of single stereotypies was substantially low in many cases, they were summed and considered together for a statistical comparison between the two buildings. Statistical results from the behavioral analysis are reported in Table 4.At time T1, piglets in the MV building had a higher prevalence of tail position, as compared with the NV case (p < 0.0001). Pigs in MV also showed higher stereotypies as compared to NV (p < 0.0001) and negative behavior toward pen mates (p = 0.02). Inactive behaviors were mostly observed in the NV building (p = 0.0002). At time T2, stereotypies still had higher frequencies in the MV building compared to NV (p = 0.046), while the other behaviors did not show substantial differences. At time T3, pigs in MV showed a higher prevalence of a hanging down tail position (p = 0.01), negative social behaviors (p = 0.04), and other active behavior (p = 0.03) compared to NV. Moreover, pigs in the MV were also more inactive than those from the NV (p = 0.004).3.1.3. Lesions and Health MeasuresOnly LHMs showing LSI with a score higher than 10 or with prevalence above 5% were considered in the analysis (see Table 5). The results showed that, at T1, pigs in the MV building had higher scores for tear staining and dirtiness compared to pigs in the NV building (p < 0.0001). Pigs in NV instead showed more tail lesions than those in MV (p = 0.01) and had a trend of higher front lesions (p = 0.07). At T2 and T3, no significant differences were observed in terms of LHMs. Only a slight trend (p = 0.09) was observed in front lesions, where NV pigs had a higher score than MV.3.2. Housing Condition MeasurementsThe results concerning the housing condition measurements in the two buildings, at the time of observation, are reported in Table 6. The records of CO2 at T1 and temperature at T2 are missing due to technical issues during the assessment. The results show that light and temperature did not differ between buildings, except for T3. In fact, at T3, the temperature was rather high in both buildings but significantly higher in NV building pens compared to the MV case (p = 0.04), whereas the light was significantly lower in the MV case as compared to NV (p = 0.04), even if the value was higher than the minimum (i.e., 40 lux) reported in the Dir 120/2008 EC. The CO2 concentration was always below 3000 ppm (the level indicated by EFSA as dangerous for pigs [10]), but showed statistical differences at T2 and T3, with opposite trends. Specifically, at T2, the CO2 concentration was higher in the MV case compared to NV (p < 0.0001), while, at T3, the NV pens showed the highest CO2 concentrations (p < 0.0001). This outcome seems in line with the number of pigs per pen, which, in the first two observations, was higher in MV vs. NV (p < 0.0001 at T1 and p = 0.0003 at T2), while, at T3, it was higher in NV compared to MV (p = 0.02). Space allowance was higher in MV as compared to NV (p < 0.0001 at T1 and T2), except for T3 where no significant differences were observed between buildings. The space allowance accomplished the minimum standards required by the legislation in each observation day.3.3. Computational Fluid Dynamics SimulationsSix simulations were performed in order to evaluate the indoor ventilation conditions (numerical scenario) on the six observation days in which housing conditions and animal-based parameters were measured: three simulations were set to analyze the air velocity magnitude in the NV building and three were solved for the study of the ventilation scenarios in the MV building. The simulations considered the different external conditions (i.e., wind velocity and wind direction, air temperature, and air relative humidity rH) of the relevant observation day (see Table 2).A qualitative comparison of the results is shown in Figure 5. It is possible to observe that indoor airflow distribution was substantially different between the two buildings, in terms of both airflow pattern and air velocity magnitude.At T1, the two buildings showed very different indoor ventilation conditions. The mechanically ventilated B5 building (see Figure 5(a.1)) presented an air velocity magnitude highly variable with the length, with very low air velocity close to the central body and progressively increasing toward the opposite extremity, with a velocity peak of 0.6 m/s. On the contrary, in the NV building (see Figure 5(a.2)), results show that, in the central portion, the indoor air velocity ranged between 0.1 m/s and 0.2 m/s, while air velocity decreased in the two lateral portions, close to the extremities, of the building.At T2 the outdoor configurations had similar air velocity magnitude and similar blowing wind direction in the two buildings. It is clear that the presence of the mechanical ventilation system in B5 (see Figure 5(b.1)), as expected, increased the indoor air velocity magnitude, while, in the natural ventilation case, the wind velocity was in general very low (see Figure 5(b.2)).Similar conditions also characterized T3 of building B3, naturally ventilated (see Figure 5(c.2)). Instead, in building B5, the ventilation system resulted in the airflow distribution and magnitude being very inhomogeneous along the building length (see Figure 5(c.1)) compared to T1 and T2. This confirms the remarkable inhomogeneity of the internal ventilation condition between the different areas of the B5 building.Further details of the air velocity magnitude are shown in Table 7.As the table shows, during the monitored period, the indoor velocity magnitude in the MV building was, overall, more homogeneous than the air velocity in the NV building. Moreover, the average value was 0.10–0.11 m/s for MV, while it was just 0.06–0.07 m/s for NV.4. DiscussionThis study quantified and qualified the main welfare issues of pigs raised in two different buildings of the same farm. The ventilation strategy, as assessed by the CFD simulations, showed remarkable variability in the ventilation conditions of each building across the three observation days.Overall, the QBA assessment showed that animals in the mechanically ventilated pens were in a more positive affective state, in accordance with the higher ventilation performance of the MV building, characterized by higher indoor air velocity. The QBA also evidenced a worsening in the affective states increasing with the age of the pigs. This last effect might also depend on the reduction in space allowance and the increase in temperature during summer, as well as changes in pig physiology, as previously reported [37,38,39]. Therefore, the comparison between the two buildings was performed separately for each observation time (i.e., T1, T2, and T3).At T1, pigs in the MV group showed lower tail LSI compared to NV (Table 5) and a higher proportion of pigs with tail position up (Table 4). Tail lesions are the outcome of tail biting behavior. Tail biting is currently considered an iceberg indicator of poor welfare, having a negative effect on the emotive state of pigs [40]. Tail biting is an abnormal behavior, and its occurrence has been found to be strongly dependent by many managerial and environmental factors [41]. In accordance with the result of this study, Lahrmann et al. [27] proposed that assessing tail position would allow quickly identifying tail-bitten pigs since these pigs would keep a low tucked tail, while pigs which show few or no tail lesions would keep the tail curled and “up”.In contrast, the behavioral analysis showed a higher frequency of negative social behavior in the group in the MV building, as well as higher stereotypy frequency. Despite that, lesion outcomes were not significantly different between the two animal groups in the different buildings. A discrepancy between these two indicators (behavior and lesion) was previously observed in other studies [24]. A possible explanation is that the lesions are the consequence of negative social behavior that occurred in a range of time (days or weeks), while the behavioral analysis in this study was a picture of the exact moment of the assessment since they were recorded by direct observation. Moreover, a limit of the present study was that behavioral analysis was carried out using direct observation; thus, although the behaviors were recorded in the same range of time, the observations were not conducted simultaneously. Pigs in the MV building showed an indeed higher score in tear staining and dirtiness, as compared to the NV building. Tear staining is the presence of a red stain in the left eye of a pig, as a consequence of the production of a red pigment by the eye pituitary gland. In pigs, it has been proposed as an indicator of negative emotional state because of a correlation with processing negative emotions [20,42]. Other studies have hypothesized that tear staining might also be stimulated by excessive gas concentration, dustiness, pen soiling, or other environmental conditions [24,43]. On this observation day, the MV building group showed a higher proportion of dirty pigs. Pig soiling has been frequently linked to higher gases in manure [44], and it might explain tear staining. The indoor air velocity was similar in the two buildings (Table 7); however, the higher number of pigs/pen with lower space allowance in the MV building compared to NV at T1 might also have enhanced this mechanism.At time T2, behavior and lesions did not show any differences, except for overall higher stereotypies in the pigs in the MV building, mainly due to the percentage of pigs showing ear biting behavior. Similar to tail biting, ear biting has been considered an indicator of poor welfare so far [45]. Among the predisposing factors for ear biting, air quality has been reported to influence its occurrence [46]. In MV building, the results showed a higher concentration of CO2 as compared to NV building. CO2 is a product of respiration, which is heavier than oxygen; therefore, it has been found to fluctuate at the pig level. It is likely to presume that, on this observation day, the inhomogeneity of the airflow and speed was not efficient to remove CO2. Moreover, CO2 was found to be highly related to the number of animals. The MV building had one more pig per pen and showed a lower space allowance, contributing to an increase in the CO2 indoor concentration. This result might explain the higher presence of ear biting in this group.Behavioral analysis evidenced also a high proportion of pigs showing dog sitting behavior on T2 and T3 observation days, in both buildings. Dog sitting has been considered a non-aggressive stereotypic behavior and an indicator of suboptimal welfare in pigs [47,48]. According to the study by Scollo et al. [49], pigs reared in intensive conditions increased the frequency of sitting behavior when space allowance decreased, e.g., in the fattening phase. This has been interpreted as the lack of space to lie down [49] or the consequence of boredom, leading to severe cognitive deprivation due to the barren environment [50,51]. A combination of the two factors might explain the results of the present study. Heavy pigs have a very restricted area available at the end of the cycle (because the current legislation states that pigs above 110 kg require min 1.00 m2, and, in heavy pig production, pigs can reach up to 180 kg at the end of the rearing period). Moreover, the behavioral analysis showed that the enrichment devices available to the pigs (metal chain and a metal chain with wood) were of marginal interest since pigs spent most of their time exploring the pen and very little time on the enrichment devices. Exploring the pen (over-exploring) has been considered another sign of boredom and poor welfare in intensive pig farms since the pens are usually in barren environments that do not provide cognitive stimuli to the pigs [52].When considering T3, behavioral analysis evidenced a higher frequency of low tail position and negative social behavior in MV compared to NV. A low tail position has been previously associated with tail lesions; however, at this assessment, no differences were observed for tail LSI. It is important to consider that, at T3, the two buildings raised the maximum score in dirtiness, corresponding to almost all pigs in each pen having manure on >50% of the body surface; therefore, this condition might have biased the results from the lesion assessment. Pig soiling is considered the outcome of abnormal eliminative behavior in pigs [44]. Normally, pigs on a partially slatted floor tend to release urine and/or feces on the slatted floor and rest on the full floor. When certain predisposing factors occur (see later), pigs can develop abnormal behavior, which leads to pen and pig soiling. One of the main identified factors is thermal discomfort [53]; in fact, with high temperature, pigs raised indoor tend to rest on the slatted floor and release urine and/or feces on the full floor [54]. In very severe heat stress conditions, pigs tend to release urine and/or feces, as well as rest, on the full floor with the purpose of heat loss [37]. This latter condition has been considered an indicator of poor welfare since, in normal conditions, pigs prefer to avoid contact with their excreta [44]. The optimal temperature range for heavy pigs (140–180 kg of live weight) is estimated to be 18–20 °C. Therefore, at T3, the temperatures in the two buildings were very challenging for the pigs (29–30 °C on average), and neither type of ventilation was able to significantly reduce this temperature. The indoor ventilation was consistently different at T3. The MV building showed high air velocity in the extremities, compared to the central zone. On the other hand, the NV building showed homogeneous low air velocity throughout the building length. This difference could have affected CO2 concentration measured, which was significantly higher in the NV building as compared to MV. Accordingly, in the NV pens, higher frequencies of polydipsia were observed. Polydipsia is a stereotypy that can occur when pigs are submitted to heat stress, in an attempt to cope with hot temperatures [55]. Moreover, behavioral analysis observed also significantly lower inactive behavior in NV pens compared to MV ones. Housing conditions also revealed that temperature, light, CO2, and number of pigs per pen were higher in the NV case compared to MV. Those factors might have influenced the pigs’ behavior. Some studies have observed an increase in activity and aggression in the presence of high temperatures, due to heat stress and difficulty in finding a comfortable place to rest for pigs kept under intensive conditions [14,44]. In the present study, negative social behavior did not differ at T3, while a trend of more front LSI in NV pens was observed. Other studies, in contrast, observed an increase in lying behavior at high temperatures [37]. The difficulty in finding a lying place could be exacerbated when the number of pigs per pen increases, as in the NV pen group. Similarly, some studies reported that increasing illumination in the pig farms can lead to an increase in activity, which does not impair pig welfare [56]. In accord with the results, CO2 concentration was found to be directly proportional to pig activity by Zong et al. [57].When the temperatures in the two building buildings were challenging, the higher air velocity in MV pens, even if not able to decrease the indoor temperature, could have contributed to a reduction in the heat perception at the pig level, as well as to a reduction in CO2 concentration, thereby influencing pig behavior and contributing to improving their welfare [10]. One limitation of the study was that the measurements could not be performed on the same day in both buildings, due to the farm flow chart, according to commercial agreements between farmers and buyers. However, this is the first study aimed at integrating ABM assessments and environmental measures provided by CFD simulations in heavy pigs. These preliminary results pose new questions regarding the effect of the interplay between outdoor and indoor conditions and ventilation systems on pig welfare, which will be further investigated.5. ConclusionsThis study pointed out that the indoor environment might influence animal behavior and overall animal welfare, and a detailed dynamic analysis of the indoor ventilation and outdoor wind and exposition is very important to improve the conditions in which the animals live and to identify the main risk factors that might impact animal health and welfare. Especially in the presence of hot temperatures, the high occurrence of pig soiling indicates severe heat stress in pigs and consequent welfare impairment. The high number of pigs showing dog sitting behavior also suggested welfare deterioration for the pigs, especially in the later phases of fattening, probably due to the combination of an absence of stimuli and heat stress. According to the results reported in this study, in hot climates, mechanical ventilation systems may not be sufficient to mitigate heat stress in pigs, and other solutions (e.g., cooling systems or water sprinklers) should be proposed to avoid welfare consequences for pigs.

animals : an open access journal from mdpi

10.3390/ani13081372

PMC10135314

Locomotion scoring requires skilled, trained observers to accurately detect lameness. Many studies have thus evaluated infrared thermography as an alternative lameness detection method as it does not require a skilled observer. However, there are few reports of the use of infrared thermography in cattle in tropical environments like Tanzania. This study, therefore, aimed to assess whether using an infrared camera to measure the foot skin temperature of hind limbs could potentially be used as an alternative on Tanzanian dairy farms. Three study farms were visited twice each during the afternoon milking on consecutive days. Locomotion scoring using a 4-point scale (0–3) was conducted on the first day as the cows exited the milking parlour after being milked. On the following day, the hind limbs of the cows were thermally imaged while they were standing in the milking parlour, using a forward-looking infrared camera. Mean foot skin temperature increase was associated with an increase in locomotion score; for example, the mean temperature was higher for cows with a locomotion score of 3 than those with a score of 2. Therefore, the present study confirmed that measuring foot skin temperature using an infrared camera has the potential to be employed for detecting lameness on Tanzanian dairy farms. However, improvements in accuracy and reductions in infrared camera costs are needed.

Lameness detection is a significant challenge. Locomotion scoring (LS), the most widely used system for detecting lameness, has several limitations, including its subjective nature and the existence of multiple systems, each with its own advantages and disadvantages. Therefore, this study aimed to evaluate whether the foot skin temperature (FST) of hind limbs, as measured using infrared thermography (IRT), could potentially be used as an alternative on Tanzanian dairy farms. Each of the three study farms were visited twice during the afternoon milking on consecutive days, with a total of 170 cows assessed. DairyNZ LS (4-point scale (0–3)) was undertaken on the first day as the cows exited the milking parlour after being milked, while on the following day, the plantar aspect of the hind limbs of the cows was thermally imaged while they were standing in the milking parlour, using a handheld T650sc forward-looking infrared camera. Mean FST was higher for cows with a locomotion score of 1 than those with a score of 0; higher for cows with a locomotion score of 2 than those with a score of 1; and higher for cows with a locomotion score of 3 than those with a score of 2, with each one-unit locomotion score increase being associated with a 0.57 °C increase in mean temperature across all zones. The optimal cut-off point of 38.0 °C for mean temperature across all zones was identified using a receiver operator characteristic curve. This cut-off point had a sensitivity of 73.2% and a specificity of 86.0% for distinguishing cows with a locomotion score ≥ 2 (clinical lameness). The prevalence of clinical lameness across all three farms was 33%, which meant that only 72% of cows with a mean FST across all zones ≥ 38.0 °C had been identified as clinically lame using LS. This study confirmed that IRT has the potential to be used to detect lameness on Tanzanian dairy farms. However, before it can be widely used, improvements in accuracy, especially specificity, are needed, as are reductions in equipment (IR camera) costs.

1. IntroductionEarly lameness detection accompanied by effective treatment is crucial to minimise the pain and discomfort associated with lameness [1,2,3], as well as to decrease the risk of irreversible claw damage [4].Visual locomotion scoring (LS) is the most commonly used active lameness detection method on dairy farms [5]. A wide range of different systems have been used; Schlageter-Tello et al. [5] identified 25 different LS systems that had been published in the peer-reviewed literature. Across all systems, the most critical challenge of LS systems is their subjective nature, with both within- and between-observer variation being high, especially when training is limited [6,7,8,9]. Thus, training is particularly important when farmers and farm staff are undertaking lameness detection using LS.An objective technique for identifying lame cows would thus be useful for on-farm detection of lameness, especially where farm staff training is difficult to achieve (e.g., because of lack of infrastructure or due to limited farmer knowledge). This is the situation in Tanzania, where active lameness detection using LS is extremely rare. In Tanzania, farm size is generally small and infrastructure limited, making training at the farm level expensive and challenging.One potential alternative to visual LS is infrared thermography (IRT), a non-invasive technique that records body surface temperature and produces a pictographic representation of the scanned anatomical area [10]. Body extremities and surface temperature mainly depend on blood perfusion and tissue metabolism [11]. Thus, changes in blood flow can influence the amount of radiated heat and can thus be detected by IRT [10]. One reason for changes in tissue blood flow is the inflammatory response. Thus, IRT has the potential to detect systemic infectious lameness such as foot-and-mouth disease [12], and localised infectious lameness such as bovine digital dermatitis [13,14,15], as well as non-infectious claw horn disease [16,17,18].However, hoof temperature can be affected by factors other than foot diseases, such as physiological status [19,20], environmental factors [21,22], and activity level [23]. Thus, the production system and management of the cows in a herd can significantly affect hoof temperature, potentially altering the accuracy of IRT as a means of lameness detection. Thus to determine the usefulness of IRT in a system, IRT needs to be tested in that system [24]. Unfortunately, no data are available for tropical pastoral systems such as those which are common in many regions of Tanzania. Therefore, the aim of this study was to investigate the association between the foot skin temperature (FST) of hind limbs at afternoon milking and locomotion scoring on three Tanzanian dairy farms where cows grazed tropical pasture between morning and afternoon milking.2. Materials and Methods2.1. Study Area and AnimalsThis study was undertaken in the Morogoro region of eastern Tanzania. This region has a sub-humid tropical climate with two wet seasons per year (long rains from March to June and short rains from October to December). Mean temperature ranges from 27 to 33.7 °C and 14.2 to 21.7 °C during the dry and wet seasons, respectively.A convenient selection of three dairy farms was made for this study. All three herds had >50 cows, were milked twice daily, grazed natural pasture off-farm between morning and afternoon milking, and had a flat concrete surface outside the milking parlour which was suitable for locomotion scoring. Calving was all year round on the three study farms, and the herds were composed of mostly multiparous milking cows, as no regular replacement of cows was done on any of the farms. All three farms had only dairy breed cattle—a mixture of breeds and crossbreeds of European dairy cattle (Friesian, Ayrshire, Jersey, and their crossbreeds) and grazed their cattle on natural pastures (principally Hyparrhenia spp., Megathyrsus maximus, Cenchrus ciliaris, and Brachiaria spp.) for approximately 8 h after morning milking.The distance between the grazing area and milking parlour on all farms was approximately 5 to 10 km with no constructed trackways. After the afternoon milking session, cows were allowed to graze around the farm before being taken to the free stall barns in the evening (6 pm), where they were given hay (principally, from Chloris gayana and Pennisetum purpureum) and homemade concentrates. All three farms had herringbone milking parlours. Farms 1 and 2 used milking machines (reverting to hand milking when there was no power), while farm 3 employed hand-milking only. During the study period, the milking herds had 52, 58, and 60 cows on farms 1, 2, and 3, respectively.Routine hoof trimming and systematic locomotion scoring were not undertaken on any of the farms, and none of the farms had accurate lameness treatment records.2.2. Study VisitsIn March 2020, farms were visited during the afternoon milking on two consecutive days. Locomotion scoring was undertaken on the first day, and on the second day the plantar aspects of both hind feet of all cows were imaged using an infrared camera.2.3. Locomotion ScoringAll locomotion scoring was undertaken by the first author (CWW) who was trained in the DairyNZ lameness score using a combination of video and supervised scoring of live animals (see Werema et al. [24] for further details). Prior to the study commencing in March 2020, CWW rewatched the training videos as a refresher [25,26].On each farm, all milking cows were locomotion scored and their ear tag number recorded as they exited the milking parlour. The site chosen for scoring observations on each farm was a flat well-maintained concrete surface that was at least 25 m long, and which was cleaned after every milking session. This area allowed the observer to view at a distance so as not to disturb cow flow and locomotion.2.4. Infrared ThermographyDuring the afternoon milking following the LS, IRT imaging was performed using a handheld T650sc Forward-looking Infrared camera (FLIR Systems, Oregon, United States) that at foot skin temperatures had an accuracy of ±0.4 °C. Emissivity was set at 0.95. Pictures were taken while the observer was standing at a distance of approximately 1 m from the cow, with the plantar aspects of both hind feet being imaged. Prior to imaging, no foot preparation was undertaken, except that the cows walked through a footbath containing only water on entering the collecting yard for the milking parlour approximately 30 min before milking.Foot images were analysed using FLIR Tools software (FLIR Systems Oregon, United States) with estimates of the surface temperature obtained from seven zones on each hind foot as described by Werema et al. [24] (see Figure 1). The maximum temperature for each zone was used for analysis in line with previous infrared studies aimed at lameness detection in the cow [24,27,28]. Mean ambient temperature during this study was 29.4 °C.2.5. Statistical Data AnalysesSPSS version 27 (IBM Corporation, Armonk, NY, USA) was used for all data analysis unless otherwise reported. Descriptive statistics were created for each zone temperature measure. The normality of foot temperature was visually assessed using Q-Q plots and histograms, followed by checking the skewness and kurtosis statistics. A generalised linear marginal repeated measures model was then used to evaluate the effect of the farm, foot, and zone within the foot on skin temperature. Farm and foot (right or left hind) were the independent variables, zone within foot the repeated variable and skin temperature the outcome variable. Interaction between farm and foot was tested and removed from the final model as it was non-significant (p > 0.05). Covariance structure was identified using Quasi-likelihood under the independence model criterion (QIC). Residuals were checked for normality using Q-Q plots and histograms. Post-hoc pairwise comparisons between zones were undertaken using the Šidák correction for multiple comparisons [29].The association between locomotion score and foot temperature was tested using six temperature measures (see Table 1 for temperature definitions). Univariable analyses were first performed to assess the association between the outcome (foot temperature) and predictor (locomotion score) variables. This analysis identified significant heteroscedasticity when we compared foot temperature across locomotion scores. Therefore, we employed a generalised linear model with an identity link function and robust estimators [30] to analyse the association between foot skin temperature and locomotion score. In model fitting, each temperature definition was used as the outcome variable, with farm and locomotion scores as the predictor variables. Two-way interactions between farm and locomotion scores were added, and backwards selection was then used to remove any interactions where p > 0.05. Both main effects were kept in the final model irrespective of p-value.Six receiver operator characteristic (ROC) curves were created, one for each temperature definition, with categorised locomotion score (lame (locomotion score ≥ 2) vs. not lame (locomotion score < 2)) to establish the sensitivity and specificity of IRT to predict locomotion score ≥ 2. Area under the curve (AUC) and coordinates of the curve (CC) were then used to evaluate a model’s predictive accuracy. Optimal temperature cut-off values were identified by maximising sensitivity plus specificity. The statistical package MedCalc Version 19.5.1 (MedCalc Software, Ostend, Belgium) was then employed to calculate positive and negative predictive values for those optimal cut-off points.3. ResultsAll milking cows were locomotion scored on all three farms. The distribution of locomotion scores is summarised in Table 2. Across the three farms, 33% of cows were identified as being lame (locomotion score ≥ 2).3.1. Effect of Foot and Foot Zone on Skin TemperatureOn average the FST of the left foot was higher than the right foot (37.510 °C and 37.411 °C, respectively). However, it is unlikely that this is biologically significant as mean difference was only 0.099 °C (95% CI: 0.003–0.194). However, biologically significant differences between zones were identified. For example, the difference between the zone with the lowest mean temperature (zone 6) and the zone with the highest mean temperature (zone 4) was 1.57 °C (95% CI: 1.392–1.748). Results for all zones and their comparisons are summarised in Table 3 and Table 4, respectively. Mean temperatures were higher for zones on the lateral claw than their equivalent zones on the medial claw (see Table 3); these differences were 0.243 °C, 0.284 °C, and 0.163 °C for zones 1 vs. 5, 2 vs. 6, and 3 vs. 7, respectively.3.1.1. Infrared Thermography versus Locomotion ScoringThe mean FST for different locomotion scores is summarised in Table 5. For all six FST measures (see Table 1 for definitions), the temperature was higher for cows with a locomotion score of 1 than those with a score of 0; higher for cows with a locomotion score of 2 than those with a score of 1; and higher for cows with a locomotion score of 3 than those with a score of 2. The comparisons of all six-foot skin temperature measures versus different locomotion scores are presented in Table 6. As an example, for mean temperature, the mean difference between cows with scores 0 and 1 was 0.91 °C (95% CI: 0.498–1.313); between scores 1 and 2 cows, it was 0.83 °C (95% CI: 0.431–1.232); and between scores 2 and 3 cows, it was 0.65 °C (95% CI: 0.026–1.271).3.1.2. Association of Foot Temperatures and Locomotion ScoresA linear association was demonstrated between individual cow locomotion scores and foot skin temperature for all six temperature measures. Detailed data are presented in Table 7. As an example, for mean temperature, each one-unit locomotion score increase was associated with a 0.57 °C (95% CI: 0.46–0.70) rise in mean temperature.3.1.3. A Receiver Operating Characteristic (ROC) AnalysisA receiver operating curve is illustrated in Figure 2. Optimal threshold values, area under the curve and calculated parameters for temperature measures are summarised in Table 8.4. DiscussionThe objective of the current study was to assess the usefulness of infrared thermography (IRT) as a tool for detecting lameness in dairy cattle partly grazed and housed in a tropical region (Tanzania) versus visual locomotion scoring (LS).4.1. Suitability of IRT for Measuring Foot Skin Temperature during MilkingOn all three farms, unlike in New Zealand [24], there was sufficient time during milking to collect thermal images from all cows. Hand milking increased the time available for image collection as it was slower than when the machine was used; but during hand milking, the milker obstructed the cow’s feet, and the cows were more restless, leading to frequent changes in foot posture. Thus, during hand milking, images were collected prior to a cow being milked.One limitation of this study is that only hind feet were imaged, as cows were photographed during milking. Not all lameness is hind-feet related; although data from housed cows suggest that hind limb lameness accounts for >90% of cases [31], in pasture-based dairy cattle in New Zealand, only 67% of lameness was in the hind limb [32]. No such data are available from Tanzanian dairy cattle. However, it is likely that only recording the FST of hind limbs lowered the sensitivity of IRT for lameness detection in these cattle. Further research is required to identify how much the sensitivity was lowered.Another potential issue with IRT under Tanzanian conditions is the imaging of dirty feet. Prior to milking on all three farms, the cows were walked through a footbath containing water. This was routine practice on all three farms and was undertaken with the intention of reducing dirt on feet, although it is unlikely that there was any impact on foot cleanliness [33]. Although this study was undertaken in the dry season, the feet of all cows were generally clean and were not washed prior to IRT. In the rainy season, the accumulation of mud on the feet during grazing might influence the accuracy of IRT, especially if the feet are not cleaned before imaging. However, Stokes et al. [27] reported that in the UK, maximum sensitivity plus specificity of IRT was achieved in feet that were not cleaned. Further longitudinal research is required to investigate whether this is also the case in Tanzania.4.2. Effect of Claw and Zone on Skin Foot TemperatureIn these cattle, the lateral claw had a higher mean temperature than the medial claw, with the highest difference of 0.28 °C being between zones 2 and 6 (see Table 3). This difference is somewhat larger than the difference of 0.1 °C we found under New Zealand conditions using the same camera and protocol for IRT [24], but it is smaller than the differences of 1.0 to 1.7 °C that have been reported in studies on housed cows [17,19]. The reason for this difference is unclear, but may be related to different protocols and different equipment.The effect of the claw was consistent across zones, with the mean temperature for zones on the lateral claw being higher than their equivalents on the medial claw (see Table 3). Additionally, the order was consistent across claws, such as if on the lateral claw, zone 1 had a higher mean temperature than zone 2, the same applied to zones 5 and 6 on the medial claw. However, the order was not the same as that reported by Werema et al. [24]. In the current study, zones 1 and 5 (coronary band) had a higher mean temperature than zones 3 and 7 (below accessory digit), whereas Werema et al. [24] reported that zones 3 and 7 had a higher temperature than zones 1 and 5. The reason for this difference is unclear. It may be related to differences in hindlimb disease (e.g., increased infectious lameness in Tanzanian cattle). However, as we did not record hoof or limb lesions in either study, this has to remain a suggestion.4.3. Infrared Thermography as a Predictor of Locomotion ScoreIn this study, mean foot skin temperature increased as locomotion scores increased (Table 5), consistent with previous studies of IRT and locomotion scores that used different protocols and devices [24,34,35]. Consistent with Werema et al. [24], the current study found that each of the locomotion scores (0, 1 and ≥2) had significantly different mean FST. The key difference between our two studies in regard to FST and locomotion score was that in Tanzanian cattle, the optimal cut-off point was 38.0 °C, whereas in New Zealand, our optimal cut-off was 34.5 °C [24]. However, both these cut-offs are much higher than the cut-offs used in studies on housed cattle in the Northern Hemisphere, e.g., 23.3 °C [35], 25.25 °C [21], 25.5 °C, [34], and 27.0 °C [27]. The reason for the large difference in optimal cut-off between our studies and those previous studies is unclear. However, it may be related to protocol differences (e.g., measuring different sites, using different equipment or cleaning feet before imaging), differences in the environment such as ambient temperature, or differences in the cause of lameness (e.g., infectious vs. non-infectious lameness). Further data on the factors driving these differences in the optimal cut-off points are required. However, our data suggest that optimal cut-offs are likely to be protocol and production system specific. Thus, cut-offs should not be transferred from one study in one system to another, even if similar protocols are used.The specificity and sensitivity of IRT at our 38.0 cut-off point for identifying cows with a locomotion score of ≥2 were both moderate (86.0 and 73.2%, respectively), and lower than the specificity and sensitivity we reported in New Zealand of 92.4 and 80.0%, respectively [24]. However, as with the difference between the two studies in relation to optimal cut-off, the difference between our two studies is much less than the differences between previous studies, which have reported the specificity and sensitivity of IRT. For example, Lin et al. [35] reported a sensitivity of 78.5% and specificity of 39.2%, while the equivalent figures for Rodríguez et al. [34] were 46.7% and 89.7%, respectively; and for Stokes et al. [27], they were 80% and 73%, respectively. As with optimal cut-off, we need more data on the factors which are driving this variability in specificity and sensitivity, but the much smaller differences between our two studies suggest that differences in protocols may be driving much of the difference between studies.For alternative methods of lameness detection, it is the specificity that is the most important. If a technique is easy to apply, then moderate sensitivity can be overcome by repeated measurement. However, even relatively high specificity (>90%) can result in farmer fatigue if a high proportion of identified cows are not actually lame [36]. This fatigue is determined by the positive predictive value (PPV), the proportion of the positive tests which are true positive. The PPV is determined by specificity and prevalence; thus, although the specificity was lower in our Tanzanian study than in New Zealand, the much higher lameness prevalence (33 vs. 14%) meant that PPV was higher in the Tanzanian study.Nevertheless, a PPV of 72% at the optimal cut-off of 38.0 °C is not ideal as ~1/4 of the cattle identified will not have changes in gait and posture associated with clinical lameness. However, such issues may be tolerated in the small dairy herds in Tanzania, where the number of false positive cattle identified at any one time may be relatively small. Further research is required to better understand Tanzanian dairy farmers’ approach to lameness and whether this technological approach is likely to be acceptable on Tanzanian dairy farms.One fundamental limitation of the current study is equipment cost. The FLIR camera is not likely to be affordable for most farmers, especially in lower-middle-income countries like Tanzania. However, the costs of infrared cameras are decreasing, and the development of smartphone apps which utilise IRT technology may make the technology affordable, even for Tanzanian farmers. Thus, our data suggest that if these changes continue, IRT may, in the future, be a useful objective method of detecting lameness in Tanzanian dairy cows.5. ConclusionsThe current study showed that, with a trained observer, locomotion scoring after milking can be a useful method of lameness detection under Tanzania conditions. However, the study farms were chosen based on having a suitable area for observation, so may not be typical of Tanzanian dairy farms. Further evaluation of LS on Tanzanian dairy farms is warranted, but the requirement for a trained observer may remain a key issue. The present study results demonstrated that FST measured by IRT was able to differentiate between cows with different locomotion scores. However, the sensitivity, and especially the specificity of IRT need to be improved before it can be recommended for lameness detection under Tanzanian conditions. Additionally, the development of cheap, accurate smartphone IRT apps is needed for such technology to become affordable even on large Tanzanian dairy farms.

animals : an open access journal from mdpi

10.3390/ani13111798

PMC10251871

Natural infections caused by Aeromonas veronii in intensive farming can lead to economic losses in tilapia farming. Overusing antibiotics and chemical antimicrobial agents in fish farming leads to antibiotic resistance, pollution, and consumer reluctance. The utilization of mangosteen (Garcinia mangostana) peel extract loaded in nanoemulsion (MSNE)-supplemented diets in Nile tilapia (Oreochromis niloticus) could improve growth performance, immune response, and disease resistance. Nevertheless, the effect of incorporating MSNE into Nile tilapia diets has not yet been studied. In this study, we assessed the efficacy of MSNE-supplemented diets on growth performance, immune response, and resistance to A. veronii infection in Nile tilapia. The particle size, polydispersity index, and particle surface charge of MSNE were 151.9 ± 1.4 nm, >0.3, and −30 mV, respectively. Furthermore, MSNE improved the in vitro inhibition against A. veronii, and MSNE-supplemented diets had a beneficial effect on growth performance, enhanced immune response, and disease resistance against A. veronii infection. In conclusion, mangosteen peel extract loaded in nanoemulsion has the potential to be used as a supplement in Nile tilapia culture.

Nanotechnology can enhance nutrient delivery and bioavailability; hence, it has recently been considered the most practical alternative technology for nutritional supplements and disease control in fish farming. The present study was designed to evaluate the effects of mangosteen peel extract loaded in nanoemulsion (MSNE) on the inhibition of A. veronii (in vitro) and in vivo growth performance, serum biochemical parameters, the immune response, and the disease resistance of Nile tilapia (Oreochromis niloticus) against A. veronii challenge. The particle size, polydispersity index, and particle surface charge of MSNE were 151.9 ± 1.4 nm, >0.3, and −30 mV, respectively. Furthermore, MSNE, mangosteen peel extract (MPE), and nanoemulsion (NE) improved the antimicrobial activity against A. veronii. Fish fed MSNE, MPE, and NE-supplemented diets had a significantly lower (p < 0.05) feed conversion ratio (FCR) and higher specific growth rate (SGR) than fish fed the control diet. Furthermore, the MSNE had significantly higher serum glucose and protein levels than the control group in Nile tilapia. Total immunoglobulin, serum lysozyme, alternative complement activity, and survival of Nile tilapia fed with MSNE were significantly higher (p < 0.05) than the control diet. Therefore, MSNE has the potential to be employed as a supplement in sustainable Nile tilapia farming.

1. IntroductionNile tilapia (Oreochromis niloticus) is a popular farmed fish in Thailand due to its faster growth, good flavor, and disease resistance [1,2,3]. The global production of Nile tilapia in 2020 was 4407.2 thousand tons [4], with Thailand contributing 210,419 tons, with a 2.2% increase in 2021 [5]. Nevertheless, pathogen infections in intensive tilapia farming have been a significant cause of economic losses in fish farming. Natural pathogen infections in tilapia have been found, such as those caused by Aeromonas spp., Flavobacterium columnare, Edwardsiella spp., Francisella spp., and Streptococcus agalactiae. Among bacterial diseases, A. veronii has recently resulted in high mortality rates on Thai tilapia farms [6,7]. The clinical signs of A. veronii infection in tilapia included ulceration, hemorrhagic septicemia, and enteritis, which are more common in juvenile tilapia because of their weaker immunity than adult tilapia [6,7,8].In tilapia farms, antibiotics and chemotherapeutics have been used to treat the disease. On the other hand, the overuse of antibiotics and chemical antimicrobial agents leads to antibiotic resistance, fish farming pollution, and consumer reluctance [9,10]. In recent years, plant extracts [11,12,13], probiotics, and prebiotics [14,15] have been considered as alternatives for treating diseases in Nile tilapia caused by A. veronii.Mangosteen (G. mangostana) is known as the “queen of fruits” and one of the best-tasting tropical fruits, which are consumed fresh or processed into jam, preserves, and wine [16]. However, the processing of mangosteen food products turns the mangosteen peel into waste. The mangosteen peel has been reported to contain bioactive compounds such as xanthones, tannins, phenolic acids, and other bioactive compounds [17]. Importantly, the mangosteen peel contains 40% xanthones (which are α-mangostins, β-mangostins, γ-mangostins, garcinone E, 8-deoxygartanin, and gartanin), which possess antimicrobial, antioxidant, anti-inflammatory, and anti-proliferative activities [18]. Moreover, mangosteen peel extract has been used to improve the survival of fish. Previous studies showed that mangosteen peel extract inhibits F. columnare in channel catfish (Ictalurus punctatus) [19] and increases the survival of black tilapia (O. niloticus Bleeker) against A. hydrophila [20]. Furthermore, a rind powder and shoot extract of mangosteen supplemented diet improved the hematological profile of clown anemonefish (Amphiprion percula) [21] and African catfish (Clarias gariepinus) fingerlings [22].In aquaculture, nano-dietary delivery is a novel method to improve the bioavailability and efficacy of nutraceuticals. This is because nano-dietary delivery ensures that the targeted elements reach the bloodstream more effectively [23,24]. However, mangosteen peel extract provides low biological activity owing to its low water solubility and ease of decomposition. To overcome this challenge, nanoemulsions have been developed and used to deliver hydrophobic mangosteen peel extract. Nanoemulsion is one of the most important types of emulsion, which consists of ultra-fine particles in the range of 20–200 nm [25]. Oil-in-water (o/w) nanoemulsions are a specific type of nanoemulsion that has been utilized to deliver hydrophobic compounds. Furthermore, the hydrophobic drug, which was loaded into a nanoemulsion, could improve biological activity, promote active compound stability, and increase drug absorptivity in their target organ. Nanoemulsions are typically synthesized from two immiscible solutions: oil and an aqueous solution. An appropriate amount of surfactant and energy is added to the oil–aqueous mixture to form a nanoemulsion, which can be synthesized using various techniques, such as high- and low-energy homogenization, based on its chemical composition and is easy to scale up [26]. Importantly, the effect of incorporating mangosteen peel extract loaded in nanoemulsion into Nile tilapia diets has not been studied yet. Therefore, the aim of the present study was to evaluate the antimicrobial efficacy of mangosteen (G. mangostana) peel extract loaded in nanoemulsion and their supplemented diets on the growth performance, serum biochemical parameters, immune response, and disease resistance of Nile tilapia against A. veronii infection.2. Materials and Methods2.1. Preparation of Mangosteen Peel Extract Loaded NanoemulsionMangosteen peels were obtained from the orchard in Lan Saka district, Nakhon Si Thammarat province, Thailand. The samples were cleaned, sliced into thin pieces, dried at 60 °C for 72 h, and ground into powder using a hammer grinder. The mangosteen peel extract (MPE) was extracted with 95% ethanol (w/v) at room temperature for 48 h, followed by filtering and evaporation using a rotary evaporator (Buchi, Switzerland). The ethanol extract of mangosteen peels is composed of xanthones [27,28], which possess antimicrobial, antioxidant, anti-inflammatory, and anti-proliferative activities [18].Mangosteen peel extract loaded into a nanoemulsion was prepared for the supplemented diets. Briefly, mangosteen peel extract loaded into nanoemulsion (MSNE) was fabricated using hot and high homogenization energy. The oil phase was prepared by mixing mangosteen peel extract (200 mg) with medium-chain triglyceride (MCT) as a liquid lipid (15 g) and cetyl palmitate (5 g). The oil mixture was dissolved with Span 80 (3 g) and Montanov 82 (1 g) over a hotplate stirrer at 500 rpm and 60–70 °C. Over a hotplate stirrer, purified water, Tween 20 (3 g), glycerol (2.5 g), and synperonic PE/F68 (2 g) were mixed to make an aqueous phase mixture. Additionally, this mixture was poured into the lipid phase and sonicated for 5 min in the sonicator unit (Qsonica sonicator, Newtown, CT, USA) using a 40-amp pulse on for 30 s and off for 5 s intervals. For the supplemented diets, a nanoemulsion solution (NE) was prepared using chemicals similar to those used in MSNE but without mangosteen peel extract.2.2. Characterizations of MSNEThe hydrodynamic diameter, polydispersity index (PDI), and particle surface charge of MSNE were characterized with dynamic light scattering (DLS) using a zetasizer (Nano ZS, Malvern Instrument, Malvern, Worcs, UK). DLS measurements were carried out using a He-Ne laser (λ0 = 633 nm, θ = 173°). The samples were diluted 20 times with purified water. The measurement conditions were set and performed in triplicate at 25 °C.Particle size and particle morphology were also observed using a transmission electron microscope (TEM) (JEOL-2100 Plus, JEOL, Akishima, TYO, Japan). The samples were also diluted 1/50 in purified water and dropped onto a carbon grid. The prepared samples were dried in a dry cabinet overnight before being characterized. The samples were observed under 80 kV with a magnification of ×25 k and ×100 k.2.3. Antibacterial Activity2.3.1. Broth Microdilution AssayThe minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) were evaluated using the broth microdilution assay [29]. Briefly, A. veronii was isolated from Nile tilapia (O. niloticus) farming in Nong Khai province, northeastern Thailand [6]. Furthermore, A. veronii was cultured overnight at 30 °C on trypticase soy broth (TSB). Two-fold serial dilutions from a stock solution (100 mg of peel extract mL−1) of MPE, MSNE, and enrofloxacin were prepared (each with three replicates) in Mueller–Hinton broth (MHB). Furthermore, 108 cells mL−1 (adjusted using 0.5 McFarland standard) of A. veronii were added to the solution, incubated at 30 °C for 24 h, and the growth was measured using a spectrophotometer at a wavelength of 625 nm. The MIC values were determined as the lowest concentration of compounds whose absorbance was comparable with the negative control tubes (MHB without inoculums). The minimum bactericidal concentration (MBC) was measured by culturing all of the tubes without turbidity. The MBC value is the lowest concentration of compounds and does not reflect bacterial growth.2.3.2. Disc Diffusion AssayAntibacterial activity was determined by using a disc diffusion assay [30]. Briefly, A. veronii was cultured overnight at 30 °C on trypticase soy broth (TSB). The bacterial density (108 cells mL−1) was inoculated on Mueller–Hinton agar (MHA) [31]. The MPE (62.5 µL mL−1), MSNE (62.5 µL mL−1), and NE (62.5 µL mL−1) were filtered (0.22 µm pore size), and 50 µL supernatants were placed on sterile paper discs (diameter 6 mm) on MHA. The plates were incubated at 30 °C for 24 h. The antibacterial activity was assessed by measuring the inhibition zone. Enrofloxacin 5 µg was used as a positive control, and dimethyl sulfoxide (DMSO, 0.5%) was used as a negative control [32,33].2.4. Experimental FishThree hundred sixty monosex (male) tilapia (15.35 ± 0.91 g) were obtained from the Napho Phanpla Limited Partnership tilapia farm in Nakhon Si Thammarat, Thailand. The fish were divided into twelve 500-L tanks (30 each) and allowed to acclimatize for two weeks. Water temperature, dissolved oxygen, and pH levels were maintained between 25–28 °C, 5.24–5.98 mg L−1, and 7.48–8.16, respectively. Ammonia nitrogen was managed by exchanging 50% of the water every two days and measuring the ammonia nitrogen using test kits (V-Unique, Bangkok, Thailand), which indicated less than 0.02 mg L−1. The fish were hand-fed approximately 5% of their body weight twice a day. All of the protocols were approved by the ethics committee of Rajamangala University of Technology Srivijaya (Approval No. U1-03662-2559).The in vivo concentration was selected based on the results of the MIC and inhibition zone. The complete randomized design with four treatments was carried out in triplicate. The stock solution of extract, i.e., 100 mg mL−1 distilled water, was used to prepare the experimental diets. The treatment diets were as follows: a control diet (Control), MPE (62.5 µL mL−1, i.e., 6.25 mg g−1 of feed), MSNE (62.5 µL mL−1, i.e., 6.25 mg g−1 of feed), and NE (62.5 µL mL−1). The control and the experimental diets were prepared by thoroughly mixing 1 mL of the MPE, MSNE, or NE, and 1 mL of distilled water (control) with 1 g of feed in the commercial diet (Charoen Pokphand Foods Public Company Limited, Samut Sakhon, Thailand), air-dried, and 4 °C used to store until feeding. The proximate composition of the commercial feed was as follows: lipid (3%), protein (30%), moisture (12%), and ash (8%).2.5. Growth PerformanceAfter 30 days of feeding, the standard formulas were used to calculate the growth performance and feed utilization performance of Nile tilapia [34]. Weight gain (WG) (g fish−1) = (final body weight (FW) − initial body weight (IW))(1) Specific growth rate (SGR) = [(ln (FW) − ln (IW)/days] × 100(2) Feed conversion ratio (FCR) = feed intake (g)/WG(3) Average daily gain (ADG) = (% gain)/(number of days)(4)2.6. Serum Biochemical AnalysisAfter 30 days of feeding, blood samples (6 fish per group) were collected from the caudal vein using a hypodermic syringe. The blood samples were allowed to clot for 3 h at 4 °C, and the serum was collected after centrifugation at 2600× g for 10 min at room temperature. The serum samples were used to measure blood urea nitrogen (BUN), total protein, glucose, albumin, direct bilirubin (D-bilirubin), total bilirubin (T-bilirubin), serum aspartate aminotransferase (AST), serum alanine transaminase (ALT), and total cholesterol using an automated chemistry analyzer (Pokleritalia 125, PKL, Italy).2.7. Immunological AssayThe total immunoglobulin (Ig) was estimated by using the method of Siwicki et al. [35]. The total plasma protein concentration was determined with bovine serum albumin (standard protein). The plasma protein was precipitated with 12% polyethylene glycol, incubated at room temperature for 30 min, and centrifuged at 12,500 rpm for 10 min. The supernatant (10 µL) was mixed with 500 µL of biuret reagent. It was incubated for 5 min at room temperature, and the absorbance was measured using a spectrophotometer at a wavelength of 550 nm. The precipitation of plasma protein concentration was determined with bovine serum albumin (standard protein). The total immunoglobulin was calculated by subtracting the precipitation of plasma protein concentration from the total plasma protein concentration.The alternative complement hemolytic 50 (ACH50) activity was analyzed [36]. Briefly, the serum was diluted in GVB-EGTA (gelatin Veronal buffer; 10 mM barbital, 145 mM NaCl, 0.1% gelatin, 0.5 mM MgCl2, 10 mM EGTA, pH 7.3–7.4) to a final volume of 250 µL. Then, 50 µL of goat red blood cells was added to the test serum for the preparation of a 2-fold serial dilution and incubated at room temperature for 90 min. The relative hemoglobin content of the supernatant was assessed using a spectrophotometer at a wavelength of 415 nm. The ACH50 activity was determined by assessing the amount of serum that induces 50% lysis of goat red blood cells.The activity of lysozyme in serum was evaluated by indicating the level of lysis of the Gram-positive bacterium Micrococcus luteus. Briefly, the lysozyme standard was diluted in 0.06 M phosphate citrate buffer (pH 6.0) and 0.09% NaCl to concentrations of 0, 2.5, 5, 10, 15, and 20 µg mL−1. The 100 µL of the lysozyme standard and the serum were added to 96 microplates with the addition of M. luteus. The absorbance was measured using a spectrophotometer at a wavelength of 450 nm [37].2.8. Challenged StudyAt the end of the experimental period (30 days), thirty fish from each group were challenged with intraperitoneal injection with A. veronii at 107 CFU fish−1, based on the previous study [6]. Afterward, mortalities or any clinical signs were observed for 15 days. The survival rate (SR) was calculated as follows [2]:SR (%) = (Total no. of survivors after challenge/total number of fish challenge) × 100(5)In addition, the relative percentage of survival (RPS) was calculated as follows [15] RPS = [1 − [(treatment mortality/control mortality) × 100](6)2.9. Statistical AnalysisStatistical analysis was conducted by using SPSS version 26 software for Windows (SPSS Inc., Chicago, IL, USA). The results were analyzed using a one-way analysis of variance (ANOVA), and significant differences between the groups were determined through the use of Duncan′s multiple range tests. The cumulative survival percentages of the experimental groups were analyzed using the Kaplan–Meier method and the Log Rank (Mantel-Cox) test. A difference of p < 0.05 was considered significant.3. Results3.1. Characterizations of MSNEThe particle sizes of MSNE and NE were 151.9 ± 1.4 nm and 146.4 ± 3.1 nm, respectively (Table 1). The polydispersity index and particle surface charge of MSNE and NE were lower than 0.3 and −30 mV, respectively. These results suggested that the protocol was successful in preparing nanoparticles.TEM results reported that the morphology of MSNE was spherical, and the particle size was smaller than 200 nm (Figure 1).3.2. Antibacterial Activity of MSNEBoth the minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) of MPE and MSNE were 62.5 µL mL−1. The results of the disc diffusion assay showed that the inhibition zone for A. veronii in MSNE (16.00 ± 3.00 mm) was significantly higher (p < 0.05) than the MPE (11.67 ± 1.15 mm) and NE (10.67 ± 1.15 mm) (Table 2 and Figure 2). MSNE, on the other hand, had a significantly lower (p < 0.05) inhibition zone than enrofloxacin (24.67 ± 1.15 mm).3.3. Growth Performance of MSNE-Supplemented DietFish fed with the MSNE diet had a significantly lower FCR (p < 0.05) than fish fed with the control, MPE, and NE diets, respectively (Table 3). In addition, the SGR, WG, and ADG of the MSNE diet were significantly (p < 0.05) higher than the control diet. Furthermore, MPE and NE diets also found higher WG and ADG compared to the control. The results indicate that the MSNE diet did not have any detrimental effect on Nile tilapia.3.4. Serum Biochemical Analysis of MSNE-Supplemented DietThe MSNE diet had significantly higher glucose and total protein levels than the control group (Table 4). Moreover, the results revealed that there were no significant differences in BUN, ALT, AST, total bilirubin, direct bilirubin, total cholesterol, or albumin among the groups.3.5. Immune Parameters Analysis of MSNE-Supplemented DietThe fish fed with MSNE had a significant increase (p < 0.05) in total Ig in comparison with the NE and control groups, respectively (Figure 3). The lysozyme activity and ACH50 activity of MSNE, MPE, and NE were significantly different in fish fed with the control diets.3.6. Survival Rates of MSNE-Supplemented DietThe mortality in the control group (66.7%), MPE (56.7%), MSNE (40.0%), and NE (66.7%) occurred on day 2 post-challenge. Importantly, fish mortality stopped at day 7, day 5, day 4, and day 7 in the control, MPE, MSNE, and NE, respectively (Figure 4A). A log-rank test showed that the survival percentages of the four experimental groups were significantly different (X2(3) = 6.893, p < 0.05) (Figure 4A). All of the fish deaths were caused by A. veronii, as determined through bacterial isolation from the spleen and liver. All of the dead fish showed clinical signs of a pale body surface, hemorrhage in the liver, and a swollen intestine with an accumulation of yellow liquid (Figure 4B,C).The cumulative mortality was significantly lower in the MSNE (76.7%) and MPE (60%) diets than in the control and NE groups (Table 5). The relative percent of survival (RPS) of the Nile tilapia fed MSNE (39.1%) was significantly higher than those of MPE (21.7%), and NE (4.3%).4. DiscussionNanotechnology improves nutrient delivery by increasing solubility and protecting against the harsh conditions of the gut, resulting in increased fish potential for nutrient absorption [38,39]. Importantly, nanotechnology has the potential to become a common practical nutritional supplement and fish disease control technology in fish farming [39]. Therefore, the current study evaluated the efficacy of supplementing Nile tilapia diets with mangosteen peel extract loaded in nanoemulsion (MSNE).In the present study, the particle sizes and zeta potential of MSNE were 151.9 ± 1.4 nm and −30 mV, confirming the successful formation of nanoemulsions that have a droplet size ranging from 10 to 500 nm [40]. The mean particle size and the negatively charged surface result are in accordance with the previous study on the nanoemulsion of mangosteen extract in virgin coconut oil [41]. Furthermore, nanoemulsions have the ability to fuse with and lyse bacteria, resulting in broad-spectrum antimicrobial activity [42]. However, the antibacterial activity of MSNE against A. veronii was significantly higher (p < 0.05) than that of MPE. Similarly, nano-mangosteen peel extract inhibited the growth of Staphylococcus aureus, Bacillus cereus, and Shigella flexineri more than the mangosteen peel extract (841 and 420 µm) [43]. Indicating a xanthone loaded in the NE or a rapid contact of the negatively charged nanodroplets with the bacterial cell wall, causing adhesion to the cell surface, membrane damage, and ultimately death [44].The results showed that the MSNE-supplemented diet had significantly better SGR and FCR in Nile tilapia than in the control group. Similarly, ginger nanoparticles and Aloe vera nanoparticles significantly improved growth performance in common carp (Cyprinus carpio) [45] and Siberian sturgeon (Acipenser baerii) [46]. This can be attributed to mangosteen peel antioxidants, such as phenolic compounds, that could enhance growth [47,48]. Additionally, less than 200 nm-sized nanoemulsions had the greatest bioavailability after ingestion [49]. Therefore, supplementing mangosteen-peel-extract-loaded nanoemulsion may be another possibility for enhancing its potential role in promoting fish growth. Nile tilapia growth and health have been shown to improve following nutritional supplements [15,50], and this was also observed in the current study, despite the relatively short feeding time (30 days). However, lowering the feeding period of the MSNE diet could reduce the active compound amount and synthesized cost of MSNE. Thus, an 8-week or longer feeding trial is necessary to evaluate more substantial changes in fish growth and health.Fish diets supplemented with nutrients may have altered immune responses due to metabolic, endocrine, or neurological pathways [51]. The mangosteen peel extract exhibited potential as an immune stimulant for fish [21]. The MSNE diet fed to Nile tilapia was observed to significantly increase serum lysozyme and ACH50 compared to the control group. The ability of fish to compete with the bacterial infection is determined by evaluating their levels of lysozyme, a key non-specific defense molecule of the immune system [52]. Lysozyme activity can be activated by an immunostimulant, which causes phagocytic cells to synthesize more lysozyme [53]. The alternative complement pathway protects fish from a wide range of possibly invading organisms [54]. In accordance with our findings, enhanced lysozyme and ACH50 were found in diets supplemented with ginger and cinnamaldehyde nanoparticles [45,51].Nile tilapia diets fed with MSNE and MPE improved their adaptive immune response (total Ig), which is associated with the immunomodulatory activities of mangosteen peel extract [55,56]. The antigen-presenting cells of each type can uptake different sizes of particles. Macrophages are involved in the uptake of particles whose sizes range from 50 to 500 nm. Whereas dendritic-like cells are associated with the uptake, the particle size ranges from 20 to 200 nm [57,58,59]. The current study showed that the size of MSNE was 151.9 ± 1.4 nm, indicating that it can be taken up by macrophages and dendritic-like cells, which may trigger helper T cells. Furthermore, helper T cells activate B cells to differentiate into plasma cells and produce immunoglobulins [60,61]. Moreover, mangosteen peel extract has antioxidant properties that enhance the immune system and reduce oxidative stress, resulting in cell protection from oxidative stress and disease infection [16,62]. The small droplets of nanoemulsions improved the stability and absorption of mangosteen peel extract, as well as its immunostimulant and antimicrobial properties [63].Blood biochemical analysis is one of the tools used to assess the nutritional and health status of fish [64]. In case of liver damage, ALT, AST, T-bilirubin, and D-bilirubin are released into the blood [65]. In addition, the increased levels of hepatic enzymes such as ALT and AST indicate that the fish have a high toxicity to nanoparticles. The experimental diets revealed that ALT, AST, T-bilirubin, D-bilirubin, and BUN values were significantly not different compared to the control diet, indicating that the experimental diets did not have a detrimental effect on liver and kidney function. Similarly, Nile tilapia fed Moringa oleifera leaf nanoparticle-supplemented diets did not increase the ALT and AST values [66].Moreover, the fish fed MSNE showed significantly higher survival than the other groups. This suggests that the MSNE efficiency to inhibit A. veronii both in vitro and in vivo is due to the smaller size of the nanoparticle, which makes it easier to approach the bacterial cell wall and inhibit bacterial activity [43]. Additionally, another possibility suggests that nanoparticles can increase the solubility and absorption of herbal drugs [38]. Similar findings were reported for ginger nanoparticles and chitosan-polymer-based nanovaccines, which prevented the infection of A. septicaemia in C. carpio [45] and A. veronii in Oreochromis spp. [67]. Furthermore, we postulate that the higher survival of MSNE is also due to the several antimicrobial compounds of the mangosteen peel, such as xanthone, tannin, saponin, flavonoid, and polyphenol, which can disrupt bacterial membranes, resulting in cell hemolysis [44,68,69,70].5. ConclusionsThe analysis of the zeta-potential and TEM images suggested the successful preparation of nanoemulsions from mangosteen peel extract. The MSNE exhibited potent antibacterial activity. Importantly, MSNE supplementation in fish diets increased growth performance, immune parameters, and the survival rate of Nile tilapia against A. veronii infection. Therefore, MSNE has the potential to be employed as a supplement in sustainable Nile tilapia farming via oral administration.

animals : an open access journal from mdpi

10.3390/ani11113146

PMC8614452

“Tree bats” are North American bats that day-roost in trees year-round and undertake seasonal migration in lieu of hibernation. These bats have been shown to be highly susceptible to collisions with wind energy turbines and are known to fly offshore during migration. Therefore, as offshore wind energy expands off the eastern U.S. coast, there is some concern about potential impacts. We monitored bats in coastal Virginia, USA, using acoustic monitors—devices that collect the unique echolocation call signatures of bat species. We found that nightly tree bat visitation offshore or on barrier islands was associated with wind speed, temperature, visibility, and seasonality. Using statistical modeling, we developed a predictive tool to assess occurrence probabilities at varying levels of wind speed, temperature, and seasonality. Probability of occurrence and therefore assumed risk to collision is highest on high temperature and visibility nights, low wind speed nights, and during the spring and fall seasons. We suggest a similar modeling regime could be used to predict the occurrence of bats at offshore wind sites to inform potential mitigation efforts.

In eastern North America, “tree bats” (Genera: Lasiurus and Lasionycteris) are highly susceptible to collisions with wind energy turbines and are known to fly offshore during migration. This raises concern about ongoing expansion of offshore wind-energy development off the Atlantic Coast. Season, atmospheric conditions, and site-level characteristics such as local habitat (e.g., forest coverage) have been shown to influence wind turbine collision rates by bats onshore, and therefore may be related to risk offshore. Therefore, to assess the factors affecting coastal presence of bats, we continuously gathered tree bat occurrence data using stationary acoustic recorders on five structures (four lighthouses on barrier islands and one light tower offshore) off the coast of Virginia, USA, across all seasons, 2012–2019. We used generalized additive models to describe tree bat occurrence on a nightly basis. We found that sites either indicated maternity or migratory seasonal occurrence patterns associated with local roosting resources, i.e., presence of trees. Across all sites, nightly occurrence was negatively related to wind speed and positively related to temperature and visibility. Using predictive performance metrics, we concluded that our model was highly predictive for the Virginia coast. Our findings were consistent with other studies—tree bat occurrence probability and presumed mortality risk to offshore wind-energy collisions is highest on low wind speed nights, high temperature and visibility nights, and during spring and fall. The high predictive model performance we observed provides a basis for which managers, using a similar monitoring and modeling regime, could develop an effective curtailment-based mitigation strategy.

1. IntroductionCollisions with wind turbines are an expanding conservation concern for bats [1,2,3]. In North America, non-hibernating, migratory “tree bats” (Genera: Lasiurus and Lasionycteris) are particularly susceptible to collisions and are often the majority bat group in post-construction carcass surveys at wind energy facilities [4,5,6,7,8]. The tree bat mortality rate at wind turbines appears to be highly correlated with the seasonal movements of these species [9,10,11,12,13,14,15] whereby collisions are generally elevated in spring and maximized in fall migration periods [5,7,16]. Increased mortality counts during migration may be attributable to space-use increase due to fall mating and migration, erratic juvenile dispersal behavior, and general attraction to turbines [13,14,17,18].North American tree bats are known to fly offshore with some regularity. This was first documented in anecdotal historical sightings from ocean vessels large distances off mainland coasts [19,20,21] and observations of tree bats on the island of Bermuda [22]. In the eastern North America, recent research has discovered high-flying tree bats 8.4–44 km from the main shoreline [23,24]. The occurrence of tree bats offshore and along shorelines follows a similar seasonal activity pattern to wind turbine collisions—a general peak during spring and fall migration [23,25,26,27,28]. The reasons for this behavior remains unknown but some speculate that the coastline serves as a topographic reference for navigation [29] or that favorable wind conditions over open ocean may aid in long distance migration [30]. It is posited that the eastern shoreline acts as a topographic barrier, concentrating southward migrating tree bats along the coast during fall [10].Wind energy in the eastern United States is expanding at an accelerating rate, particularly in the offshore sector [31,32,33]. To date, two offshore wind turbine operations exist in the eastern United States that account for <50 MW capacity [33]. However, an increasing number of offshore projects are now leased and in the beginning construction phases. It is projected that these projects will account for more than 20 GW of rated capacity [33], a 400-fold increase. Although projections indicate offshore wind facilities will likely be concentrated in the wind resource rich Northeast, some development is proposed off the mid-Atlantic coast along Virginia, Delaware, Maryland, and New Jersey [33]. The impact this rapid development will have on bats is unknown, however, risk is certainly non-zero particularly for tree bats as they are the most susceptible bat group to collision (particularly during migration) and are the only bat group consistently seen at offshore localities (again, particularly during migration).Onshore, extensive monitoring at wind facility sites post-construction have offered successful data driven conservation strategies to minimize bat mortality at turbines including, but not limited to, acoustic deterrents [34,35,36,37,38,39,40] and curtailment [41,42,43,44,45]. Curtailment is based on the knowledge that most bats generally avoid flying in overtly windy conditions, i.e., avoiding speeds generally above 5 m/s [44,45]. Therefore, at low wind speeds below this (or other) threshold(s), turbine managers feather turbine blades, bringing rotor movement to a minimum, and thereby minimizing bat fatalities. There has been some success in the use of curtailment to reduce bat mortality while also minimizing financial loss [42,44,46] through the development of “smart curtailment” algorithms. These are typically model based, multivariate approaches whereby curtailment is triggered by the expectation of high bat activity or probability of presence [42,47,48]. Although these strategies may hold promise for offshore wind energy impacts, unlike terrestrial systems, the factors that influence occurrence (and therefore the parameter values necessary to predict risk metrics) of bats offshore are poorly known.Monitoring bat activity offshore is challenging. Passive acoustic monitoring over ocean waters requires some type of infrastructure (platforms, buoys, lighthouses, etc.) to support acoustic detectors. Barrier islands and offshore structures offer an alternative approach to collecting acoustic data in the near “offshore” environment if located a considerable distance from the mainland shoreline. Yet, these sites are accessible and feasible as detector deployment infrastructure. A few studies have approached the problem in this way, deploying acoustic detectors on islands, structures at sea, and on the coastline [26,27,28]. However, most of this research has been concentrated in the Northeast. Further south in the mid-Atlantic, some degree of seasonally fluctuating barrier island use by bats has been observed [25] as has fall offshore flight [23,24]. However, temporally, and geographically limited sample sizes somewhat constrain generalizability and the development of predictive models to describe the factors influencing bat use and to test the feasibility of a predictive curtailment algorithms in the near-offshore environment.Our study sought to address these data gaps with a large sample of acoustic bat occurrence data off the Eastern Shore of Virginia (ESVA). From 2012 to 2019, the Virginia Department of Wildlife Resources deployed acoustic monitors at four barrier island sites and one offshore site. We used this large acoustic dataset to develop a model to describe migratory tree bat nightly occurrence relationships to season, atmospheric conditions, and site-specific characteristics. Our modeling served two purposes—description and prediction [49]. We describe the parameters that reveal the effects of various potential drivers of nightly occurrence of tree bats. Then, we use the model as a predictive tool of bat occurrence and hence potential risk for regional wind turbine collisions once deployed.We hypothesized that tree bat occurrence in mid-Atlantic coastal environments is closely related to season due to the seasonal fluctuations in which tree bats use coastal landscapes and oceanic space. We predicted strong positive effects in spring and fall, moderate effects in summer, and negative effects in winter. We also developed competing hypotheses that the seasonal effect is explained by either (1) unique sites, or (2) the availability of local day-roosting habitat and potentially important foraging habitat (e.g., trees/forests, fresh water). We predicted that if the seasonal pattern is best explained by site specifics that unique sites would have noticeably different occurrence relationships to season. If the seasonal pattern is best explained by roosting habitat, sites with limited roosting habitat would have similar occurrence relationships to season (e.g., peaks only during migration). Lastly, we hypothesized that tree bat occurrence is closely related to multiple atmospheric conditions. We predicted that occurrence would be negatively related to wind speed, positively related to nightly temperature.2. Materials and Methods2.1. Study AreaWe conducted acoustic monitoring on four barrier island sites and one offshore site off the ESVA (Figure 1 and Figure 2) 2012–2019. The ESVA is the southern portion of the Delmarva Peninsula, surrounded by the Chesapeake Bay to the west and the Atlantic Ocean to the east. Locally, the vegetation is mid-Atlantic Coastal Plain deciduous and evergreen (pine) mixed upland and bottomland forest in its interior and intertidal saltmarsh habitat along the coasts. On the eastern Atlantic boundary, a chain of barrier islands occur that are characterized by little physical relief above sea level with upland shrub thickets, scattered patches of forest and salt marsh [50]. On the eastern side of ESVA, we monitored on Assateague Island on the Assateague Lighthouse, Cedar Island on an inactive United States Coast Guard (USCG) station, Hog Island on an inactive USCG station, and Smith Island on the Cape Charles Lighthouse. Cedar Island, Smith Island, and Hog Island are similar in that they are primarily composed of saltmarsh and upland shrub thickets. Some overstory evergreen vegetation exists on Hog Island, however, it is extremely limited in extent. In contrast to other ESVA study sites, Assateague Island has considerable deciduous and evergreen forest habitat. Additionally, Assateague Island contains fresh water sources. On the western boundary, we conducted research near Silver Beach on a navigation light structure approximately 0.7 km off the western shore of the ESVA in the Chesapeake Bay.2.2. Acoustic DataFrom 2012 to 2019, we collected acoustic data at the five ESVA sites named Assateague Island, Cedar Island, Hog Island, Smith Island, and Silver Beach (Figure 1 and Figure 2). We used frequency division/zero-crossing acoustic detectors (Anabat SD1 and SD2, Titley Scientific, New Ballina, NSW (any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government) that record high frequency (15–150 kHz) echolocation pulses of bats. We placed the detectors on existing structures (lighthouses or similar) at heights of approximately 10–40 m. We collected data annually typically across three seasons—beginning in early spring and through late fall and recorded during the winter season at least once per site (Figure 3). We considered acoustic recordings on a nightly basis from sunset to sunrise.The post processing data structure was composed of timestamped individual echolocation sequences of bats (hereafter “bat passes” or “passes”). Bat passes is defined as a distinct series of echolocation pulses, or “clicks”, which is identified to one bat as they pass within range of the detector [51]. We used Kaleidoscope 4.5.0 Bats of North America—4.2.0 classifier (Wildlife Acoustic, Inc., Maynard, MA, USA) to identify passes to species, unidentified bat passes (“no ID”), or noise. We tallied nightly pass counts by individual species, no ID, and noise. To minimize false positives, we manually inspected subsets of passes identified to species to confirm identification. Due to the context of the problem and realization that >85% of identified passes were tree bats, we placed particular emphasis on correct identification of eastern red bats (Lasiurus borealis), silver-haired bats (Lasionycteris noctivigans), and hoary bats (Lasiurus cinereus), and only used tree bat pass data in our analysis. In wind turbine collision risk studies, recent evidence suggests that the hourly or nightly passage rates of bats pre-construction are poor predictors of fatality rates post-construction [52] so to account for this, instead of using hourly or nightly tallies of bat passes as our response variable, we restructured the data to consider only the binary occurrence (or non-occurrence) of tree bat(s) on a nightly basis.2.3. Atmospheric Conditions and Other VariablesWe compiled weather conditions from nearby National Oceanic and Atmospheric Administration (NOAA) weather stations on the ESVA (Climate data online; https://www.ncdc.noaa.gov/cdo-web/; accessed on 9 November 2019; Figure 1). We used the nearest available weather station to each site to approximate hourly weather conditions. We extracted hourly data on wind speed (m/s), wind gust speed (m/s), temperature (deg C), visibility (0–16.2 km), pressure (mmHg), precipitation duration (hours), precipitation (cm), and absolute humidity (mg/cm3). We filtered these data to reflect dates and hours in which the detector stations were active, i.e., reducing to a nightly basis between sunset and sunrise on active detector nights. We summarized each weather variable to reflect nightly conditions taking the nightly mean of wind speed, temperature, visibility, pressure, and relative humidity, the maximum wind gust speed, nightly cumulative sum of precipitation, and nightly cumulative number of hours precipitating. We also created a change in pressure variable calculated as the mean at the current night minus the mean of the previous night. Lastly, because bats may be more likely to be present at individual sites during different times of the year if they contain viable day-roosting habitat, we created a binary roosting habitat variable as has viable roost availability (forests) or, none or limited roost availability for each detector station. We also noted additional potentially relevant variables including ordinal date (day of year), site name, and year. We did not include the potentially relevant variable of detector height because we were limited by the number of unique heights (n = 5).2.4. Presentation of Data and Exploratory Data AnalysisWe performed an exploratory data analysis (EDA) to visualize the effect of wind speed, temperature, and seasonality. At each site, we noted tree bat occurrence or non-occurrence and calculated a 20-day moving average on the ordinal date (day of the year (1–366)). We also fit a smoothing line (generalized additive model (GAM) spline; [53]) to aid in visualization. We calculated 90%, 95%, and 99% quantile values in which 90%, etc. of all nights with tree bat occurrence were less than a wind speed threshold, greater than a temperature threshold, or between spring or fall date ranges. We calculated these quantiles for wind speed only, temperature only, and a mix of wind speed and temperature or a spring and fall date range.2.5. ModelingWe used generalized additive models (GAMs; [54,55]) with a binomial distribution (logit link function) to model the relationship between binary nightly occurrence of tree bats and the variables. GAMs are an extension of generalized linear models (GLMs; [56]) such that the expected response is a link transformed summation of an intercept and the product of slope coefficients and variables, however, some or all variable effects may be specified as semi- or non-parametric real functions denoted as splines (hereafter “smooths” or throughout, “f(x)”; [54,55,57,58,59,60]). Smooths are created by a series of coefficient scaled basis functions “tied” together at knots—evenly spaced segments along the variable range. These smooths then, can take on complex, non-parametric shapes in the relationship between the variable effect and the variable as opposed to the generally linear, or parametric relationships in GLMs that may not reflect actual biological patterns. We performed all analysis in program R [61] and fit GAMs using the R package mgcv [53].To test our first hypothesis and the subsequent competing hypotheses, we used a model selection process that tested three a priori models to select the appropriate model that best accounted for a potentially nonlinear seasonal effect. These three models contained smooth non-parametric function(s) of the ordinal date that took on factor level-specific shapes depending on the factor provided in the model. We supplied one model with no factor variable, one with roosting habitat availability that varied its intercept with site, and one with site only (Table 1).We used the minimum change in Bayesian information criterion (∆BIC; [62]) as the basis for model selection because BIC outperforms Akaike’s information criterion (AIC) when n is large (n > 3000; [63]) and tends to select more parsimonious models because the penalty for complexity is larger than AIC (for n > e2). We calculated each a priori model BIC using Schwarz’s method of BIC = –2 log(ℒ) + K log(n) implemented in the model.sel function of the R package MuMIn [64].We used the top a priori model structure in all further models as a baseline (i.e., this model structure was nested within any other further competing model). To include atmospheric conditions into the model, we first reduced variables by omitting those of high correlation (Pearson correlation coefficient > 0.7). We then created a global model that included all variables as smooth functions. We performed a dredge (i.e., fit all possible additive model combinations). We compared these models by BIC. We considered models <2 BIC points competing models in which we selected the top model by biological feasibility and interpretability [63].To test predictive performance, we performed a series of diagnostic tests on the final model. First, to assess general performance, we conducted a Monte-Carlo cross validation (MCCV; [65]) on the area under the receiver operating characteristic (AUC; [66]). The AUC is a threshold dependent, sensitivity and specificity dictated metric of predictive performance for binary data such that an AUC of 0.5 is no better than random and an AUC of 1 is perfect prediction [66]. To perform the MCCV, we (1) randomly selected 85% of the data for training and 15% of the data for testing the model, (2) fit the model and predicted on the withheld data, and (3) measured and saved the AUC using the R package pROC [67,68]. We repeated those steps for 1000 iterations. We calculated the mean of the AUCs, and a 95% confidence interval by taking 2.5% and 97.5% quantiles of those 1000 iterations.As a second metric of predictive performance, we again divided the data into 85% training and 15% testing groups. Using the training group, we fit the final model, then predicted occurrence probabilities from 0 to 1 on the testing group. We selected an optimal “cut-off” threshold using the Youdin index [69] to categorize occurrence. We used these categorizations to compare to their true occurrence values. Therein, we calculated a confusion matrix and values for sensitivity (true positive rate; true positives/true positives + false negatives) and specificity (true negative rate; true negatives/true negatives + false positives; [70,71,72]). We used various other R packages for data manipulation, cleaning, processing [73,74,75], visualizations, and mapping [76,77,78,79,80,81].3. ResultsWe recorded acoustic data over eight years (2012–2019) resulting in a total of 5735 nights of recording across all detectors. We recorded at Silver Beach across four years (762 nights), Assateague Island and Smith Island across five years (791 and 1328 nights, respectively), Hog Island across six years (1268 nights), and Cedar Island across all eight years (1586 nights). Per year, effort was primarily centered on warmer months of spring to autumn, however, winter effort existed 1–2 years per site resulting in nearly entire year effort across all sites (Figure 3). We detected tree bats on a total of 39.26% of recorded nights, which varied by site (min = 29.89% at Hog Island, max = 71.30% at Assateague Island).With respect to the exploratory data analysis, tree bats appeared to occur at sites with strong relationships to season (Figure 4). At Assateague Island, we detected nightly occurrence with a unimodal shape—low occurrence in winter, increase in spring, a peak in summer, and a decrease in fall. For all other localities, we detected bimodal effect shapes with respect to season—low occurrence in winter, a small peak in spring, a slight decrease in summer, then an increase and larger peak in fall. Sites showing a bimodal shape contained limited roosting habitat and probably limited foraging habitat.The proportion of nights with tree bat occurrences appear related to wind speed and temperature as 90% of occurrences were on nights where wind speed averaged below 4.06 m/s and average temperatures were above 12.66 °C (Table 2). As an additive effect to these atmospheric conditions, the spring and fall months appeared to carry a large proportion of positive occurrence nights as 90% of occurrences occur during either wind speeds below 4.5 m/s, above 12 °C, or were between the dates of 28 April–14 May or 16 August–1 September.Through model selection, the top approximating a priori model via minimum ∆BIC was model 2—the ordinal date shaped by roosting habitat model (Table 3). In post-hoc, our atmospheric variables reduced from seven to six potentially relevant variables by omitting nightly maximum wind gust (m/s) due to multicollinearity with nightly mean wind speed (m/s). We argue that nightly mean wind speed more closely relates to the wind conditions throughout an entire sampling night. We encountered missing values from weather stations that forced us to reduce the total number of nights from 5735 to 4864 so that each model used the same data in calculating model selection metrics.Our global model included all remaining terms as additive smooths. The model selection dredge resulted in 128 models from which we selected the top model via BIC (Table 4). This top model included an intercept, site as a factor, a smooth effect of ordinal date based on day-roosting habitat, and smooth effects of nightly mean temperature, wind speed, and visibility.The smooth effect of temperature and visibility was generally positive along the range of variable values, however, plateaus at higher values of each were evident (Figure 5). The smooth effect of wind speed was linear and negative along the range of variable values. The smooth effect of ordinal date was different for each roost availability type. For sites with limited roost availability the ordinal date effect was generally low in winter, locally maximized in spring at around ordinal date 125 (~May 5), lower in summer, and maximized in fall at around ordinal date 235 (~August 22). For sites with viable roosting habitat, the ordinal date effect generally increased from winter to spring, peaked in summer at around ordinal date 200 (~July 17), and decreased in fall (Figure 5). The intercept of the model was modified based on site. The greatest positive effect was Assateague Island (β0 + β1), the only site with viable roosting habitat (Table 5). The lowest effect was at Hog Island (β0 + β2; Table 5), the most distant barrier island from the ESVA mainland.Our final model was highly predictive. It contained a mean MCCV AUC value of 0.852–95% CI (0.828, 0.877). The optimal cutoff for predicting occurrence or non-occurrence was 0.393, which we used as a threshold to predict on withheld data. The model appeared to correctly predict occurrences as indicated by the confusion matrix (Figure 6). Therein, the number of false positives and false negatives were generally low (117 and 65 out of 730 data points). Sensitivity (true positive rate) and specificity (true negative rate) values were 0.826 and 0.671, respectively.4. DiscussionOur hypotheses were generally supported by our analysis. First, tree bats do occur at offshore barrier island sites, but occurrence is most related to season. This became apparent as peaks in the occurrence rate over ordinal date contained local maximums in spring and fall. This seasonal effect is demonstrated in the EDA (Figure 4), the smoothed ordinal effects of the model (Figure 5), and in that 90 and 95% of occurrence nights fell within either nights of certain wind speed and temperature conditions or somewhat narrow spring or fall date ranges (Table 2). These seasonal effects undoubtedly are related to the migratory behavior of tree bat species [9,10,11,15,82,83]. Why tree bats traverse large bodies of water seasonally remains speculative, however, it could be explained by a simple increase in space use during migration or favorable conditions for long-distance flight occurring offshore [30]. Curiously, while both fall and spring seasons contain local peaks in occurrence, fall occurrence rates are higher than spring. This could be explained by the fact that fall is mating season and tree bats are more active in searching for mates and thereby more likely to explore more space [14,84]. This appears consistent as female eastern red bats are known to have multiple mates in a single season [85]. These effects are compounded, too, by additional volant juveniles navigating long distances for the first time. Moreover, these effects occur at a time when the species’ population should be at a level higher following summer parturition and juvenile volancy than winter and spring which could incidentally cause a higher rate of occurrence in fall as compared to spring [86].Another obvious effect on occurrence was the presence or absence of viable roosting habitat (forests), which seemed to influence the shape of the seasonal pattern. The unimodal seasonal activity pattern observed at Assateague Island, which contained forests available for roosting habitat, was more typical of onshore sites—bats arrive in spring, activity peaks in mid-summer which corresponds to maternity activity, and bats settle into reduced activity states (cave hibernation (cave bats) or intermittent torpor (tree bats)) in fall and then winter [11]. The other survey sites that contained little or no forest patches seemed to be visited consistently in just spring and fall—an indication of vagrant, rather than maternity, use. These sites contained lower activity in general, suggesting that without quality roosting habitat, bat occurrence and residency time was low, aside from the spring and fall season. Therefore, our results support the latter of our competing hypothesis—the pattern of seasonal use is best explained by the availability of local day-roosting habitat. This point also supports that siting for offshore wind turbines should consider increasing distance to viable roosting habitat to reduce curtailment needs during the summer. A similar study also observed this [27], that bat activity decreases with increasing distance from mainland and decreasing forest coverage.Next, including nightly atmospheric conditions greatly improved the model. It was not surprising that wind speed had negative effects on occurrence and conversely temperature and visibility had positive effects on occurrence. For example, we found that ~95% of nights that contained positive tree bat occurrence were <~5 m/s (~11 mi/hr) and >~10 °C (50 °F). High wind speeds and low temperatures greatly increase the energy costs associated with flying [87] which may be particularly true at distant barrier islands where we speculate that the nightly origin of these bats was most likely non-local, i.e., from the ESVA mainland. We understand that a multitude of atmospheric conditions relate to the activity states of bats [28] and the migratory behavior of birds [30]. Indeed, many observations of over-ocean flying bats have been during calm conditions [88]. We were initially surprised that visibility was selected as a relevant variable considering that bats rely on audible cues to navigate during flight via echolocation. However, it is intuitive to assume that bats use visual cues when flying above the ocean and/or when traveling to the islands and structures that we detected them nearby. Bats are known to echolocate while traveling over the ocean, particularly when close enough to detect them with acoustics (e.g., [23,89]), however, hoary bats (Lasiurus cinereus) sometimes forgo echolocation when traveling, and therefore rely solely on visual clues intermittently [90]. It is not beyond the realm of possibility that over-ocean flying bats use vision when there are no reflective surfaces for echolocation (e.g., at high altitudes) and therefore are unlikely to engage in over-ocean flights when visibility is low. The negative relationship of occurrence to visibility could also be explained by poor conditions for flying in general (rain, wind, low temperatures) as poor visibility is generally associated with those poor weather conditions, which, require more energy to fly in (e.g., rain, [91]).Our modeling effort increased our understanding of the pattern of occurrence of migratory tree bats at barrier island sites in the mid-Atlantic. Importantly, this dataset revealed the conditions whereby occurrence along the coast is more or less likely. Whether inland or coastal, it is established that site characteristics, seasonality, and atmospheric conditions influence the activity rates of bats [25,26,27,28]. These effects are reinforced with our findings at the more southerly latitude of the ESVA. The occurrence of bats offshore was highly predictable when using the model. Our large AUC values from the MCCV indicated that, on average, given site specifics, day of the year, and atmospheric conditions, the occurrence probability of migratory tree bats is very accurate for the ESVA sites. We also argue that our study continues a trend of consistency across studies. Tree bats appear to use offshore areas on the east coast during a certain set of conditions—calm and warm weather, during fall (and to some extent spring), and nearer to shorelines or forest coverage than far [27,88]. Therefore, we believe our results are fairly generalizable to the surrounding region of the mid-Atlantic coastline.Nonetheless, our study is not without limitations. First, observing bats via acoustics contain potential biases in that the physics of ultrasonic sound (bat echolocation pulses) change with atmospheric conditions [92]. This, plus the fact that non-occurrence does not necessarily equate to absence (i.e., detection probability is not reliably 1; [93]), may over or underestimate probabilities of occurrence depending on the conditions, time of year, among other factors. Additionally, acoustic activity of bats and wind turbine collision risk are not always analogous [52]. Regardless, these issues largely concern the correct detection of absence rather than presence of bats. In our research, nights of known occurrence follow patterns that are consistent (i.e., prediction accuracy is high on withheld data when trained on multiple years of data). Lastly, our study was limited in the number of sites to support our results. As we were limited in detector deployment infrastructure and accessibility, we were restricted to only five sites which could restrict generalizability and could contain bias. As just mentioned however, our study does not differ in major ways from other studies. Even with just five sites, patterns of occurrence follow associations with atmospheric conditions, site specifics, and seasonality in a largely nonunique manner which, as a standalone study may suffer with site limitations, but in the greater literature is in support of what has previously been known [25,27,89].The development and deployment of predictive smart curtailment algorithms is currently underway onshore and may be a viable method to reducing bat collisions at offshore wind farms. While additional research is warranted to assess collision risk at project-level localities, these data and this analysis helps identify a starting-point in assessing the temporal and climatic conditions when tree bats may be most susceptible to impacts from wind turbines offshore in the Mid-Atlantic region. If nightly occurrence does indeed generally correlate to offshore wind strike risk, a similar algorithm or model could be used as a to predict when risk is more likely. Even more simply, if managers were to implement simple standards, such as curtailing on nights with average wind speeds <5 m/s, temperatures >10 °C, and/or during the spring and (especially) the fall, most bat occurrence (and potential risk) could be avoided. It appears that curtailment using a combination of variables as these could be a relatively inexpensive [94] and effective [44,95] way to reduce bat fatalities at offshore wind facilities.5. ConclusionsAlthough we do not suggest using our specific model as a smart curtailment tool per se, this framework provides a viable starting point for creating curtailment regimens in the Mid-Atlantic. Our model was highly predictive and parsimonious which may suggest generalizability. Our results suggest that tree bat occurrence, and therefore a potential for risk is most likely under general and definable conditions—during the spring and fall seasons and on nights with low wind speeds, high temperatures, and high visibility. As such, it would be feasible for wind energy managers to collect acoustic data pre- and post-construction, assess the frequency of visitation at their specific sites, use site specific effects, atmospheric conditions, and seasonality in a modeling framework, and test the predictive ability of the model for specific locations. Using this approach, managers could have some basis for understanding which conditions influence nightly occurrence and when and where bat collision risk is non-zero or high as a guide to curtailment or other mitigation practices to minimize bat mortality.

animals : an open access journal from mdpi

10.3390/ani11051465

PMC8161327

Currently, functional foods are gaining widespread attention. Polyunsaturated fatty acids (PUFA) and antioxidant compounds have beneficial effects on health. It is possible to increase the concentration of these compounds in the milk obtained from dairy cows by manipulating their diets, thereby improving milk quality and consequently the health of animals and humans who consume this milk. Annatto seed (Bixa orellana L.) is a source of antioxidants, whereas linseed oil is rich in omega 3 fatty acid. We evaluated the inclusion of annatto seeds and linseed oil in the diets of dairy cows and their effects on dry matter intake (DMI), nutrient digestibility, milk yield, milk composition and antioxidant capacity in milk and blood. There was no effect of treatment on nutrient digestibility and antioxidant capacity, but the addition of annatto seeds decreased DMI and milk production and linseed oil supplementation reduced milk fat content.

This study aimed to evaluate the effects of annatto seeds, linseed oil and their combination on DMI, apparent total tract digestibility, antioxidant capacity and milk composition of dairy cows. Four lactating Holstein cows (120 ± 43 days in milk; 15.98 ± 2.02 kg of milk/day, mean ± SD) were allocated in a 4 × 4 Latin square with a 2 × 2 factorial arrangement (with or without annatto seeds at 15 g/kg of dry matter (DM); with or without linseed oil at 30 g/kg of DM) and provided four different diets: control (no annatto seeds or linseed oil); annatto seeds (15 g/kg of DM); linseed oil (30 g/kg of DM); and a combination of both annatto seeds and linseed oil. Annatto seeds reduced DM intake, and milk yield, protein and lactose, but increased content of fat, total solids and short chain fatty acid, with no effect on total antioxidant capacity of milk. Linseed oil supplementation decreased medium chain fatty acid proportion and n-6/n-3 ratio, conversely it increased long chain fatty acids and n-3 fatty acid content of milk, ether extract intake and total-tract digestibility. Thus, linseed oil supplementation in dairy cow diets improved the milk FA profile but decreased milk fat concentration, whereas annatto seeds did not influence antioxidant capacity and depressed feed intake and milk yield.

1. IntroductionFoods containing a high concentration of n-3 fatty acids in the human diet can reduce the risk of cardiovascular disease and prostate, colon and breast cancer [1,2]. Hence, interest in the consumption of polyunsaturated fatty acids (PUFA)-rich dairy products has increased [3,4]. Cows receiving typical diets produce milk that contains approximately 70% saturated fatty acids (SFA), whereas mono- and polyunsaturated fatty acids represent approximately 25% and 5% of milk fat, respectively [5].Linseed oil, which contains approximately 70% PUFA, of which 50% is α-linolenic acid (ALA), is a rich vegetable source of fatty acid (FA) n-3, and supplementation with linseed oil in cow diets leads to an increase in n-3 FA in their milk. [6]. The increase in the PUFA content of milk fat is desirable for human health; however, this makes the milk more susceptible to oxidation [7]. Because PUFA has double bonds, it is susceptible to the loss of electrons because of the action of free radicals, luminosity and other agents, leading to lipid peroxidation [8]. Lipid peroxidation promotes a rancid flavor and reduces the shelf life of dairy products [8] and can predispose humans to metabolic diseases [9,10]. In the animal body, there is a balance between the formation of free radicals and endogenous antioxidant capacity. Endogenous antioxidant capacity is regulated by the enzymes catalase (CAT), superoxide dismutase (SOD) and glutathione peroxidase (GPX), which can delay lipid peroxidation [11]. The body can also receive antioxidant compounds, such as vitamin E, selenium, phenolic compounds, from diet (exogenous antioxidants) [12,13].The supply of these sources of antioxidants in dairy cow diets increases the antioxidant activity in milk [13,14]. The addition of carotenoids and vitamin E in the feeding of ruminants, can improve the organoleptic quality of the final product [15], both in meat and milk. This protects the product from the lipoperoxidation effects by improving the oxidative status of meat and milk [16]. Annatto (Bixa orellana L.) is a Bixaceae family plant, and its seeds contain carotenoids with antioxidants properties (Bixin and Norbixin). Bixin is a carotenoid that belongs to the apocarotenoid family [17], and whose antioxidant power is conferred by an extensive chain of double bonds, which allows it to combat singlet oxygens [18]. Annatto seeds have tocotrienols that, combined with bixin, could synergistically protect PUFAs from oxidation [19]. The objectives of this study were to evaluate the effects of dietary of annatto seeds, linseed oil, and their combination on dry matter intake (DMI) and nutrient total tract digestibility, as well as changes in the antioxidant capacity and milk FA composition of lactating cows. We hypothesized that linseed oil supplementation would increase n-3 concentration, whereas annatto seeds would increase antioxidant capacity and, consequently, decrease milk oxidation, and the association between the two treatments would increase n-3 concentration in the milk in conjunction with lower oxidation. 2. Materials and Methods 2.1. Cows, Diets, and Experimental ProceduresThe experiment was conducted at the State University of Maringa (UEM), Brazil. The experimental protocol was approved by the UEM Ethics Committee (number 6450240117). Four multiparous Holstein cows (120 ± 43 days in milk, 15.98 ± 2.02 kg of milk/day, 566 ± 64 kg of body weight (BW), mean ± SD) were housed in a 4 × 4 Latin square design with a 2 × 2 factorial arrangement (with or without annatto seeds; with or without linseed oil), for a total of 84 days with four experimental periods of 21 days, with 16 days for adaptation and 5 days for collection). Diets were formulated for dairy cows averaging 580 kg of BW, 25 kg/day of milk with 35 g/kg of fat [20]. Treatments were formulated as follows (Table 1): (1) control (no added annatto seeds and no linseed oil; (2) annatto seeds (annatto seeds at 15 g/kg of dry matter (DM) added); (3) linseed oil (linseed oil at 30 g/kg of DM added); and (4) a combination of both annatto seeds and linseed oil added. The forage to concentrate ratio of the experimental diets was 600:400, on a DM basis. To ensure the intake of annatto seeds and linseed oil, before providing the total mixed ration (TMR), these ingredients were weighed and mixed in the concentrate daily. The annatto seeds used contained 904 g/kg DM, 113 g/kg crude protein (CP), 51 g/kg ether extract (EE), 254 g/kg neutral detergent fiber (NDF) and have a total polyphenol content of 8.35 mg expressed as gallic acid equivalents (GAE)/g, total antioxidant capacity of 358 mM Trolox/mL, reducing power of 1.25 mg GAE/g and total antioxidant activity by capturing free radicals (DPPH) of 32.6 mg/mL. Animals were housed in a tie-stall and milking was performed twice daily. TMRs (Table 2) were offered after milking, with 60% of the total DM at 8:00 and the remaining (40%) at 16:00. Animals were weighed at the beginning and at the end of each period before the morning feed. The amount of feed supplied was adjusted to obtain 10% leftovers.2.2. Dry Matter and Nutrients Intake and DigestibilityIntake was measured daily by weighing the feed provided and the refusals; feed intake was calculated for statistical analysis during the collection period (from day 17 to day 21) in which daily samples were collected and stored for further analysis. To estimate the total apparent digestibility of DM and food, fecal samples (100 g) were collected directly from the rectum in the following schedule: 17th day at 08:00 and 17:00; 18th day at 02:00, 11:00, and 20:00; 19th day at 05:00, 14:00, and 23:00, as described by Morris et al. [21]. Fecal samples were dried at 55 °C for 72 h and ground to pass through a 2-mm sieve to determine indigestible neutral detergent fiber (iNDF) and then through a 1-mm sieve for bromatological analysis.Feed, refusals and fecal samples were pooled for each cow to obtain a composite sample per treatment and period. Samples of feed, refusals and feces were analyzed according to AOAC [22] for DM (DM, method 934.01), EE (method 920.85), ash (method 938.08), CP (method 981.10) and NDF according to Van Soest et al. [23]. The organic matter (OM) was calculated by the difference between the ash content and total DM. Non-fibrous carbohydrates (NFC) were obtained using the equation described by Sniffen et al. [24]: NFC = 100 − (CP + NDF + EE + MM), wherein MM is the mineral material. Indigestible NDF (iNDF) was used as an internal indicator to estimate daily fecal excretion. The iNDF was determined in the feed samples, refusals and feces, through its in situ incubation for 288 h into the rumen as described by Huhtanen et al. [25]. 2.3. Lipid Profile and Blood Total Antioxidant CapacityBlood samples were collected from the coccygeal vein on the 19th day of each period in tubes containing anticoagulant (BD Vacutainer®, K2EDTA 7.2 mg, São Paulo, SP, Brazil) 4 h after the morning feeding. After the sampling, samples were subjected to centrifugation at 1080× g for 15 min to obtain the plasma. Plasma was transferred to 2-mL vials and immediately frozen for later determination of triglycerides, total cholesterol, high density lipoprotein cholesterol (HDL-cholesterol), and blood total antioxidant capacity.The blood concentrations of triglycerides, total cholesterol and HDL-cholesterol were determined using commercial kits (Gold Analisa Diagnóstica, Belo Horizonte, MG, Brazil), and the analyses were performed on a spectrophotometer (Bio-2000IL; Bioplus®, São Paulo, SP, Brazil). The determination of blood total antioxidant capacity was conducted according to Erel [26]. The samples (5 μL) were incubated with 200 μL of 0.4 M acetate buffer at pH 5.8. Subsequently, 20 μL of ABTS+-(2,2-azinobis-(3-ethyl-benzothiazolin-6-sulfonic acid)) •+ solution was added in 30 mM acetate buffer at pH 3.6. The samples were incubated and absorbance was measured at 660 nm with a UV–vis spectrophotometer (Spectrum SP2000, Thermo Fisher, Waltham, MA, USA). Blood total antioxidant capacity was expressed in Trolox equivalent (μM Trolox/mL).2.4. Milk Yield and CompositionMilk production was recorded with the use of meters coupled to milking equipment. For statistical analysis, only the data referring to the collection periods were used. On days 20 and 21, milk samples were collected and proportionally composited based on milk yield during the morning and afternoon milking. Then milk samples were divided into two aliquots. The first aliquot, approximately 50 mL of the milk, was kept at room temperature and stored with 2-bromo-2-nitropropane-1,3-diol (Bronopol, San Ramon, CA, USA) for the determination of fat, protein, lactose, defatted dry extract and milk density. The second aliquot, approximately 100 mL of milk, without the addition of preservative, was frozen at −80 °C for the subsequent determination of reducing power, total antioxidant activity total, conjugated diene hydroperoxides (CD), thiobarbituric acid reactive substances (TBARS), and composition of fatty acids.The concentrations of nitrogen-urea, fat, protein, and lactose in milk were determined using a spectrophotometer (Bentley 2000; Bentley Instrument, Inc., Chaska, MN, USA) in the laboratory of the Dairy Analysis Program of the “Associação Paranaense dos Criadores de Bovinos da Raça Holandesa”, Curitiba, Pr, Brazil. The somatic cell count was obtained using an electronic counter (Somacount 500; Bentley Instrument, Inc., Chaska, MN, USA). Additionally, fat-corrected milk (4% FCM) was calculated using the Gaines [27] equation as follows: FCM = 0.4 × milk yield + 15 × fat yield.2.5. Reducing Power of MilkThe reducing power was analyzed from the milk sample extracts obtained by adding 9 mL of methanol to 1 mL of milk. Thereafter, the mixture was vortexed for 5 min and centrifuged at 1080× g for 10 min. The total reducing power of milk was determined by a method described by Zhu et al. [28] with some modifications. Milk proteins were precipitated by adding 1 mL of a trichloroacetic acid solution (200:800; v/v) to 1 mL of milk. The mixture was vortex-mixed for 10 min and centrifuged at 1058× g for 10 min at 4 °C. Absorbance was measured at 700 nm on a UV–vis spectrophotometer (Spectrum SP2000, Thermo Fisher, Waltham, MA, USA), and reducing power was reported as GAE (mg/L).2.6. Total Antioxidant Activity in MilkTotal antioxidant activity was analyzed from the milk sample extracts obtained by adding 9 mL of methanol to 1 mL of milk. Thereafter, the mixture was vortexed for 5 min and centrifuged at 1080× g for 10 min. Total antioxidant activity of the milk samples was determined by a method described by Rufino et al. [29], with the addition of radical ABTS+-(2,2-azinobis-(3-ethyl-benzothiazolin-6-sulfonic acid)) to the extract. Absorbance was measured at 734 nm on a UV–vis spectrophotometer (Spectrum SP2000, Thermo Fisher, Waltham, MA, USA) after 6 min of reaction. Total antioxidant activity was expressed in Trolox equivalent (μM Trolox/mL).2.7. Conjugated Diene Hydriperoxides (CD) in MilkThe CD in milk were evaluated using the methodology described by Kiokias et al. [30], which reflects the lipid oxidation of milk. A 50 μL sample of milk was added to 2.5 mL of an isooctane/2-propanol solution (2:1, v/v) and vortexed for 1 min. Samples were filtered on a 0.22 mm PTFE membrane filter, and absorbance was measured at 232 nm on a UV–vis spectrophotometer (Spectrum SP2000, Thermo Fisher, Waltham, MA, USA). The CD were expressed in mmol/kg of fat.2.8. Tiobarbituric Acid Reactive Substances (TBARS) in MilkAnalysis of TBARS was performed as described by Vyncke [31], with few adaptations. A 500 µL aliquot of milk was transferred into a 15-mL falcon tube (Cellstar, Greiner Bio-One, Americana, SP, Brazil) containing 2.0 mL of thiobarbituric acid solution (TBA 1%, TCA 15%, and 562.5 mM HCl). The samples were heated in a boiling water bath (100 °C) for 15 min, cooled in ice water for 5 min and then centrifuged at 1080× g for 10 min. The supernatant was transferred to a cuvette for later reading at 532nm on a UV–vis spectrophotometer (Spectrum SP2000, Thermo Fisher, Waltham, MA, USA). The results were expressed in concentration (mmol/kg fat).2.9. Composition of Fatty Acids in MilkTo determine the milk fatty acid profile, fat was extracted by centrifugation, according to the methodology described by Murphy et al. [32] and fatty acids were esterified according to the ISO 5509 method [33] using KOH/methanol and n-heptane. Fatty acid methyl esters were quantified using gas chromatography (Trace GC 52 Ultra, Thermo Scientific, West Palm Beach, FL, USA) with self-sampling, equipped with a flame ionization detector at 240 °C and a fused silica capillary column (100 m in length, 0.25 mm internal diameter and 0.20 μm; Restek 2560, Thermo Scientific). The gas flow was 45 mL/min for H2 (carrier gas), 45 mL/min for N2 (auxiliary gas), and 45 a 400 m/min of synthetic air (flame gases). The initial temperature of the column was set at 50 °C, maintained for 4 min, remained from at 10 °C to then increased to 200 °C, and was maintained for 15 min. Then, it was maintained at 20 °C attaining reaching 240 °C and maintained for 8 min at the final temperature. The quantification of fatty acids in the sample was conducted by comparison with the retention time of fatty acid methyl esters from standard samples (18919-1 Sigma Aldrich, São Paulo, SP, Brazil).2.10. Statistical AnalysisStatistical analysis was performed using the MIXED procedure of SAS version 9.3 (SAS Institute, Inc., Cary, NC, USA) according to the following model:Yijkl = μ + Ai + Oj + A × Oij + pk + al + eijkl. With pk ≈ N (0, σp2), al ≈ N (0, σa2) e eijkl ≈ N (0, σe2), where Yijkl is the observed value; μ is the general mean; Ai is the fixed effect of annatto (i = 1 or 2); Oj is the fixed effect of linseed oil (j = 1 or 2); A × Oij is the fixed effect of the interaction between annatto and linseed oil; pk is the random period effect; al is the random effect of the animal; eijkl is the residual error; N indicates the normal distribution; and σp2, σa2, and σe2 are the variances associated with the random effects associated with the period and animal, and residual variance, respectively. Fisher’s least significant difference test (LSD) was applied when there was an interaction between the oil and annatto factors. For all analysis, the significance level was p ≤ 0.05, and trends were declared at 0.05 < p ≤ 0.10.3. ResultsThere was no interaction between annatto seeds and linseed oil on DM and nutrient intake except on EE intake (p = 0.03) (Table 2). Indeed, DM, OM, CP, NDF and NFC intake was reduced by annatto seeds addition (p < 0.001), and linseed oil reduced the NFC intake by 8% (p = 0.03). There was no interaction between evaluated factors on DM and nutrients digestibility. In addition, annatto seeds did not affect digestibility. However, supplementation of linseed oil increased EE digestibility (p = 0.005), tended to decrease NDF digestibility (p = 0.07) and to increase CP digestibility (p = 0.09).There was no interaction between annatto seeds and linseed oil on milk yield, composition and antioxidant activity (Table 3). Diets containing annatto seeds reduced the production of milk (p = 0.01), protein (p = 0.01), and lactose (p = 0.01). Annatto seed addition increased the fat content and total solids content (p < 0.01) and supplementation with linseed oil decreased milk fat content (p = 0.03).An interaction effect between annatto seeds and linseed oil was observed in milk C6:0 (p = 0.007), C8:0 (p = 0.01), and C18:2 n6t (p = 0.04) (Table 4). The addition of annatto seeds increased the concentrations of C6:0 and C8:0 fatty acids in the absence of linseed oil (p ≤ 0.05), while in the presence of oil the concentrations of these fatty acids was not changed. Supplementation of linseed oil increased the concentrations of C18:2 n6t FA (p ≤ 0.05), while in combination with annatto seeds, the concentration of this FA was not affected. It was observed that linseed oil supplementation reduced the concentration of C10:0, C12:0, C13:0, C14:1, C15:0, and C16:0 in milk fat (p < 0.05), and increased the C18:0, C18:1 n9t, C18:2 n6t, C18:3 n3, C18:2 c9 t11 -CLA, C20:3 n6, C20:4 n6, and C21:0 (p < 0.05).An interaction effect between annatto seeds and oil was observed on short chain fatty acids (SCFA) (p < 0.01), addition of annatto seeds alone increased milk SCFA in the absence of linseed oil (p ≤ 0.05), while in the presence of oil it did not change the concentrations of these fatty acids (Table 5). Linseed oil decreased medium chain fatty acids (MCFA) (p = 0.01) and increased long chain fatty acids (LCFA) (p = 0.02) in milk fat. Additionally, linseed oil supplementation tended to reduce mono-unsaturated fatty acids (MUFA) (p = 0.07) and increase PUFA (P= 0.08) and saturated fatty acids (SFA) (p = 0.06). There was a tendency for interaction between annatto seeds and linseed oil on fatty acids n-6 (p < 0.10), wherein linseed oil supplementation increased the fatty acids n-6 in the absence of annatto but in the presence of annatto seeds this fatty acid was reduced (Table 5). There was also an approximately 176% increase of n-3 concentration in milk fat (p < 0.01) and the n-6/n-3 ratio was improved by linseed oil supplementation once the linseed oil was reduced by 58% (p < 0.01).The different treatments, as well as the addition of annatto seeds did not have any effect on the antioxidant concentrations in milk (Table 3) or blood (Table 6). The concentration of CD in milk has a tendency to increase with addition of linseed oil (p = 0.07) to the diet. There was no effect of annatto seeds on blood parameters, linseed oil supplementation has a tendency to increase HDL-cholesterol (p = 0.08).4. DiscussionThe addition of annatto seeds to the cow diet reduced feed intake. It was observed that cows rejected diets containing annatto seeds. It was associated with undesirable palatability, texture, and odor. Additionally, annatto seeds contain carotenoids, terpenoids, and terpenes [34,35,36], which affect the DMI [37,38]. Terpenoids and terpenes can be stored in the form of essential oils and released through structures, such as secretory glands and trichomes; consequently, these compounds are concentrated in the taste buds [39]. The main essential oil components of annatto seeds are two monoterpenes called α-pinene and β-pinene [40]. According to Estell et al. [41], α-pinene in the alfalfa pellets rendered them unpalatable for consumption by sheep. It is possible that the monoterpene α-pinene may be associated with animal rejection of annatto seeds.The linseed oil supplementation, regardless of annatto seed supply, increased EE intake and reduced the NFC intake. According to the NRC [20], increased dietary energy density by fat supplementation leads to a replacement of carbohydrates by lipids, which might result in a higher EE content in the diet and a lower proportion of carbohydrates. Lipid supplementation increased the digestibility of EE and CP. Similar results were found by Santos et al. [42] when dairy cow diets were supplemented with 2.5% linseed oil. Linseed oil had a higher proportion of PUFA and a low melting point facilitating micelle formation and increasing absorption rate in the intestine [20]. Thus, the addition of linseed oil to diets increased the digestibility of EE.Exceeding 5% of the dietary lipid DM may cause some adverse effects, such as inhibiting the growth of microorganisms that degrade the fibrous fraction in the roughage and affecting fiber and OM digestibility [43]. The percentage of EE in diets with oil was 5.6%, which may explain the tendency to reduce NDF digestibility. However, exceeding the 7% lipid limit can have more evident effects on reducing DMI, and digestibility of DM, OM, and NDF [44].In the present study, 15 g/kg of annatto seeds in DM did not alter nutrient digestibility. These results were consistent with the findings of Barbosa et al. [45] who used higher amounts of annatto seeds (100 g/kg, 230 g/kg, and 350 g/kg of DM) in diets for sheep and found that annatto seeds did not influence nutrient digestibility. De Lima Júnior et al. [46] reported increasing levels (0, 100, 200, and 300 g/kg total DM) of the annatto by-product in the sheep feed did not affect the digestibility of the layers.In this study, the treatments did not affect the lipid profile (cholesterol, triglycerides, and HDL-cholesterol). These parameters followed the standard values according to Kaneko et al. [47]. Although some studies have highlighted the potential effects of antioxidants on these parameters, this was not observed in the present study [48,49]. Similar results were reported by Lima et al. [50], where supplementation of the diet with annatto colorum (paprika) in 0.08, 0.12, and 0.16 g/kg of DM for lactating cows did not affect plasma cholesterol concentration but altered the plasma fatty acid profile. In the present study, linseed oil supplementation increased the HDL-cholesterol, which could be attributed to lipid intake.Milk and lactose production were reduced with diets containing annatto seeds, which also decreased DM and nutrient intake. According to Dado and Allen [51], milk yield is positively correlated with DMI, resulting in lower net energy intake available for milk synthesis in comparison with the control diet, which had higher production because of higher DMI. Additionally, the lower DMI decreased glucose precursors. As synthesis of lactose is dependent on glucose, lactose is comprised in annatto seed-fed animals [52]. Milk fat and total solid content increased after consumption of annatto seeds diets. These effects occurred probably because of to a concentration effect because these diets showed a 296 reduction in milk yield [53].The addition of annatto seeds did not affect the proportions and ratios of fatty acids in milk, especially the MCFA that are responsible for increasing the fat content, suggesting that these amounts of annatto seeds were not the main cause of the increased content of fat. Additionally, fat-corrected milk production and milk fat production did not show significant effects related to annatto seeds and lipid supplementation.The addition of annatto seeds did not affect the oxidation products and antioxidant activity, as the seeds were included with the objective of transferring its polyphenolic compounds to milk, to decrease oxidation of milk fat that was enriched with PUFA. Treatments with annatto seeds had approximately 125.25 mg GAE/g DM of total polyphenols, which was not satisfactory to affect concentrations, increase antioxidant activity, or decrease the oxidative profile. Cows supplemented with 24.35 mg GAE/kg DM of total polyphenols from propolis extract did not improve the oxidative stability of milk [20]. Conversely, Santos et al. [54] found no significant difference for the oxidative profile but found increased antioxidant activity with the supplementation of 1950 mg GAE/g DM of total polyphenols from grape residue in the presence of soybean oil.Regarding the antioxidant activity in blood and milk, lipid supplementation increased CD concentration because of increased PUFA intake, which has double bonds that predispose it to lipoperoxidation, thus causing electron loss [55] and increased CD levels in milk. Supplementation with annatto seeds did not affect the animal or milk oxidative status, which may have been caused by the low absorption range of carotenoids. Bixin is a carotenoid that belongs to the group of carotenes, whose absorptive process is very similar to that of lipids, because carotenes are fat-soluble pigments [56]. Carotenes are less polar molecules found in the center of emulsions, whose absorptive efficiency depends on the transfer of the emulsion to the micelles, which is not as effective for the carotene group compared to the xanthophyll group [56].Supplementation with linseed oil reduced the fat content in milk. The PUFA content in the linseed oil is responsible for this reduction [57,58] by altering ruminal biohydrogenation resulting in intermediate products, such as fatty acids trans-9, cis-11-CLA, and trans-10, cis-12-CLA, which also inhibit the synthesis of milk fat in the mammary glands [59,60]. However, consumption of material and EE was lower in diets supplemented with annatto seeds, according to Woolpert et al. [61], and diets with lower EE content led to an increase in de novo fatty acid synthesis. This is the main source of the formation of SCFA in the mammary glands.Linseed oil diet supplementation reduced the MCFA, owing to the higher concentration of PUFAs in flax oil [62]. Linseed oil PUFAs can be affected by incomplete biohydrogenation generating the trans-9, cis-11-CLA, and trans-10, cis-12-CLA [59,60]. These fatty acids are inhibitors of genes that are involved in the de novo synthesis [59,60], such that these CLAs reduced MCFA (C10:0, C12:0, C13:0, C14:1, C15:0, C16:0) and led to a tendency to reduce C14:0. It is noteworthy that these fatty acids are of mixed origin, originating from the de novo synthesis that occurs in the mammary gland and from food.In general, supplementation with linseed oil increased the LCFA (C18:0, C18:1 n9t, C18:2 n6t, C18:3 n3, C20:3 n6, C20:4 n6, and C21:0). This can be explained by the composition of linseed oil, which is a source of LCFA and PUFA, mainly C18:3 n-3. When the PUFA reaches the rumen, it can be affected by biohydrogenation to protect ruminal bacteria generating intermediates and saturated fatty acids as a final product. The fatty acids that leave the rumen are absorbed in the intestine and incorporated into milk fat [63]. The C18:2 n6t is produced during the biohydrogenation process, and its reduction may be associated with the addition of carotenoids. Hino et al. [64] performed an in vitro assay, in which they added 5 or 10 mg of β-carotene, 5 mg α-tocopherol, and 1 g of glucose for each liter of rumen liquid, and observed the growth of ruminal microorganisms. The association of β-carotene and α-tocopherol with sunflower oil decreased the inhibition of the growth of these microorganisms caused by oil; thus, increasing the use of LFCA by ruminal bacteria. Possibly the annatto seed carotenoids promoted the use of the C18:2 n6t fatty acid in the presence of oil, resulting in a lower concentration of n-6 compared to the diet with flaxseed oil.Supplementation with linseed oil decreased SFA and increased MUFA and PUFA. The increase in PUFA results in a better fatty acid profile for human consumption. The results of the present study are consistent with the findings of Santos et al. [42] and Suksombat et al. [65], who reported similar changes in milk fatty acids when linseed oil was used, and those of Caroprese et al. [66] and Petit and Côrtes [67], who used linseed to feed dairy cows.Linseed oil is a rich source of C18:3 n-3, which is partially hydrogenated, and a large part of it becomes overpassed, which promotes an increase in the concentration of n-3 in milk fat. There was an increase of approximately 176% in the concentration of n-3 in milk obtained from cows fed linseed oil supplemented diets as compared to those fed normal diets. Previous studies demonstrated an increase of n-3 concentration in milk from dairy cows receiving linseed oil supplementation [42,58,68]. Fatty acids n-3 has a beneficial effect on human health owing to its potential to reduce cardiovascular disease [1], prostate, colon, and breast cancer [2]. Oil supplementation in the cow diet improves the n6:n3 ratio by 58%, which makes their milk healthier, according to the World Health Organization [69].5. ConclusionsThe addition of 15 g/kg DM of annatto seeds did not affect milk antioxidant capacity but reduced feed intake and milk yield. The addition of 30 g/kg DM of linseed oil decreased milk fat content and the ratio between fatty acids n-6:n-3.

animals : an open access journal from mdpi

10.3390/ani11123411

PMC8697970

At the end of lactation, antibiotics (DCT) or internal teat sealants (ITS) can be used to treat or prevent mastitis in dairy cows. Recommendations on how to perform such treatments are available, but little is known about how well these are followed by farmers and veterinarians. To increase this knowledge, questionnaires about farmer routines and veterinary advice were sent to 2472 farmers and 517 veterinarians in Sweden. Fourteen percent of the farmers and 25% of the veterinarians responded. Among the farmers, 81% used DCT to some cows, 3% used DCT to all cows, and 16% did not use DCT at all. Almost all veterinarians prescribed DCT, most only to some cows in a herd while 8% sometimes recommended DCT to all cows in a herd. Most of the farmers did not use ITS and half of the veterinarians never prescribed ITS. Milking system and milk production, and post-graduate training and number of mastitis cases per month were associated with several of the answers by the farmers and veterinarians, respectively. Overall, many farmers and veterinarians followed the recommendations, but it was also clear that more communication is needed as well as an up-date of the recommendations.

Dry-cow therapy with antibiotics (DCT) and treatment with internal teat sealants (ITS) are often used to control mastitis in dairy cows. However, the knowledge on farmer and veterinary compliance with recommendations for DCT and ITS is scarce. Thus, the main aim was to collect information on farmer routines and veterinary advice for such treatments. Associations with herd and veterinary variables were also studied. Web-based questionnaires including questions on demographics and the use of DCT and ITS were sent to 2472 farmers and 517 veterinarians in Sweden. The answers were summarized descriptively, and associations with demographics were evaluated using univariable regression models. The response rate was 14% for farmers and 25% for veterinarians. Among the farmers, 81% used selective DCT (SDCT), 3% used blanket DCT (BDCT), and 16% did not use DCT. Almost all (93%) veterinarians prescribed DCT and among those most recommended SDCT while 8% recommended BDCT. Eighty-two percent of the farmers did not use ITS and 45% of the veterinarians never prescribed ITS. Milking system and milk production, and post-graduate training and number of mastitis cases per month were associated with the largest numbers of farmer and veterinary answers, respectively. In conclusion, many farmer routines and veterinary advice complied with the recommendations available at the time, but a clear need for more education was also identified. The results also indicated that an up-date of the national recommendations was warranted.

1. IntroductionDry-cow therapy with antibiotics (DCT) has been a part of control programmes for mastitis in dairy cows in many countries of the world for more than 50 years, as reviewed by [1]. In many countries blanket DCT (BDCT), i.e., treatment of all cows at drying-off has been recommended. In the Nordic countries, however, selective DCT (SDCT), i.e., only treating cows with infectious subclinical mastitis, has always been the norm as outlined recently by Rajala-Schultz et al. [2]. When using SDCT it is important to have good routines for selection of cows to ensure that only cows with a good prognosis for treatment success are treated. Recommendations on DCT have been available from the main Swedish advisory organization (Växa Sverige (former Swedish Dairy Organization), Stockholm, Sweden (www.vxa.se), accessed on 29 November 2021) for dairy farmers for many years. Moreover, such recommendations are also included in the Swedish national guidelines for veterinary use of antibiotics [3], and some aspects of the use of DCT are regulated in the national legislation for veterinarians [4]. In addition, in the Nordic countries, drugs for DCT can only be prescribed by a veterinarian. Recommendations on DCT are also available in other countries [5,6,7,8], but there is no consensus on the best routine to perform DCT. Recently, however, the interest for SDCT and for finding the best tools to use when selecting cows suitable for such treatment has increased e.g., [9,10,11].In many countries, the use of internal teat sealants (ITS) to all, or a selection of cows, is also recommended at drying-off to reduce the risk of new intramammary infections (IMI) during the dry period reviewed by [12]. ITS has been available in Sweden for some time, but if and how often these products are used is not known. Moreover, as for DCT, ITS can only be prescribed by a veterinarian, and the use of ITS have so far not been included in the recommendations to farmers and guidelines for veterinarians.To our knowledge, detailed investigations of routines for DCT and ITS used by commercial farmers are scarce [13], and none has been performed in Sweden. Thus, it is not known if Swedish farmers follow the recommendations from the Swedish advisory organization or not. Moreover, it is not known if field veterinarians working with dairy cattle use those recommendations when advising farmers. In addition, the attitudes of farmers and veterinarians to the importance of DCT and ITS to animal health and production are not known.Thus, the main aims of this study were to collect information on farmer routines and attitudes as well as on veterinary advice on DCT and ITS using web-based questionnaires. In addition, we wanted to investigate if routines and advice were associated with herd and veterinary variables. The long-term goal was to evaluate the need for more education and an update of recommendations.2. Materials and MethodsTwo web-based anonymous questionnaires, one for farmers and one for veterinarians, were produced (Questback Essentials, Stockholm, Sweden). The farmer questionnaire was modified from the questionnaire used by Vilar et al. [13]. The questionnaires (in Swedish) are provided in the Supplementary Documents S1 and S2. The contents and quality of the questionnaires were tested by a small group of farmers and veterinarians before performing the study.The questionnaires included several sections, the first contained demographic questions about the herd (number of cows/year, county, conventional or organic production, average annual milk production per cow, average estimated bulk milk somatic cell count (SCC), milking system), or the veterinarian (year of veterinary degree, country of veterinary degree, county, gender, post-graduate training in bovine diseases, number of years in cattle practice, number of mastitis cases/month). The other parts contained questions on drying-off, dry cow therapy using antibiotics, treatment with internal teat sealants, and the dry period. The questions on drying-off and the dry period will be presented in a separate publication. Some questions were mandatory and some questions resulted in additional questions depending on the answers. In both questionnaires, DCT was defined as infusing antibiotics (mostly long-acting) via the teat canal into all udder quarters after the last milking during drying-off, i.e., just before the start of the dry period, and treatment with ITS was defined as infusing bismuth subnitrate (not antibiotics) via the teat canal into all udder quarters after the last milking during drying-off, i.e., just before the start of the dry period.The questionnaires were sent to the target groups in the end of 2019/beginning of 2020 (late autumn/early winter). Information about the questionnaire and a link to the website was distributed to the farmers via an email sent to all Swedish dairy producers having an email address and being affiliated to one of the dairy cattle farmers organisations Växa Sverige, Skånesemin or Rådgivarna i Sjuhärad (n = 2472 representing 75% of all Swedish dairy farmers (Jordbruksverket 2019)). Information about the questionnaire and a link to the website was distributed to the veterinarians via an e-mail sent to all veterinarians born in 1950 or later that had treated at least one case of bovine mastitis during 2018 and were registered with an email address at the National Board of Agriculture (n = 530). The email addresses of 43 veterinarians were not valid so the questionnaire reached 487 veterinarians. The same questionnaire was also sent via email to all veterinarians (n = 30) employed at the three dairy cattle farmers organisations mentioned above. Thus, the questionnaire was sent to 517 veterinarians. Both questionnaires were open for 4 weeks and reminders were sent via e-mail approximately once a week.The national recommendations for DCT available at the time of the study can be summarized as follows: Cows for DCT are selected based on udder health class (UHC). The UHC is provided by the Swedish official milk recording scheme (Kokontrollen, Växa Sverige, Stockholm) and is calculated from the cow SCC (CSCC) at 2–3 consecutive milk recordings using regression analysis [14]. The UHC at the last milk recording before drying-off is used for selection of cows as follows; UHC 0–2 (average CSCC < 130,000 cells/mL) = no DCT (healthy), UHC 3–8 (average CSCC 130,000 to 600,000 cells/mL) = DCT is decided based on SCC and growth of bacteria, UHC 9 (average CSCC > 600,000 cells/mL) = no treatment sue to poor prognosis. Only cows with subclinical mastitis due to IMI sensitive to penicillin are selected. All four udder quarters are treated with long-acting antibiotics after the last milking during drying-off after thorough cleaning of the teat end (with alcohol). At the time of the study two long-acting DCT products were available in Sweden, Benestermycin® (Boehringer Ingelheim Animal Health Nordics, Malmö, Sweden) containing benetamine penicillin, penetamate hydroiodide and framycetin sulphate, and Siccalactin® (Boehringer Ingelheim Animal Health Nordics, Malmö, Sweden) containing benzyl penicillin benzatine and dihydrostreptomycin. Teat dip/spray immediately after treatment and check the udder and teat dip/spray morning and evening for 24 h after treatment. If the herd has no access to UHC, CMT results can be used to control udder health. Given the long withdrawal time of the products the time to estimated calving should be at least 6 weeks if long-acting antibiotics are used. All DCT cows should be checked with CMT after calving and control of CSCC at first milk recording, and milk samples for bacteriology should be taken if CMT > 2 (scale 1–5) or CSCC is above 150,000 cells/mL. At the time, there were no Swedish recommendations for ITS available.To ensure sufficient numbers to perform valid statistics farmers or veterinarians in each geographical area the counties (n = 21) were compiled into regions according to Nomenclature of Territorial Units for Statistics (NUTS) level 2 with two modifications: Middle and Upper Norrland were combined into Norrland, and Stockholm and East Middle Sweden were combined into East Sweden. This resulted in six geographical regions, i.e., Norrland, North Middle Sweden, East Sweden, Småland and the islands, West Sweden, and South Sweden. The cut-offs for the continuous variables: number of cows/years, year of veterinary degree, and number of years in cattle practice were set to give approximately equal number of observations per category. The categories for the other continuous variables: milk production, bulk milk SCC, and number of mastitis treatments per month were pre-set in the questionnaire. Three categories for the bulk milk SCC were used (<200,000 cells/mL, 200,000 to 300,000 cells/mL, and >300,000 cells/mL) but the number of herds having >300,000 cells/mL were very few. Thus, only two categories were used in the analyses, i.e., <200,000 cells/mL and ≥200,000 cells/mL.The answers to the questionnaires were summarized descriptively. Statistical differences within each target group based on information on herds and veterinarians were evaluated using univariable logistic or multinomial logistic regression models when questions had enough observations per outcome and answer category. In the models, the answers in the questionnaires were treated as outcomes and the demographic variables as explanatory variables. Due to the vast number of univariable analyses, no multivariable analyses were performed. The herd variables (categories) used were region (six regions as specified above), production type (conventional, organic), milking system (automatic milking system (AMS), tied-up, parlour, rotary, combinations), number of cows/year (<53, 53–77, 78–137, ≥138 cows), average yearly milk production per cow (<9000 kg energy corrected milk (ECM), 9000–11,000 kg ECM, >11,000 kg ECM), and estimated bulk milk SCC (<200,000 cells/mL, ≥200,000 cells/mL). Housing of lactating cows was omitted from the analyses due to the categories being very similar to milking system categories, where the housing system was included. The veterinary variables (categories) were year of veterinary degree (1977–1991, 1992–2001, 2002–2008, 2009–2014, 2015–2020), country of veterinary degree (Sweden, Denmark/Norway/Finland, other countries in Europe), region (six regions, as specified above), gender (female, male), post-graduate training in bovine diseases (yes, no), type of post-graduate training (Hälsopaket mjölk, ViLA, other), number of years in cattle practice (<5, 5–9, 10–14, 15–19, 20–24, ≥25 years) and number of mastitis treatments per month (<1, 1–3, 4–8, 9–15, >15 cases).3. ResultsThe questionnaire response rate was 14% (340 of 2472) for farmers and 25% (130 of 517) for veterinarians. The results are presented descriptively below followed by results from the statistical analyses when relevant.3.1. Herd and Veterinary VariablesDescriptive information on the respondents is given in Table 1 and Table 2. In short, most dairy herds were situated in West Sweden or Småland and the islands, had conventional production and housed their dairy cows in insulated free-stall buildings. Automatic milking system (AMS) was the most common milking system and the average number of cows per herd and year was 116 (median 78). Most herds produced 9000 to 11,000 kg ECM/year and had an annual average bulk SCC below 200,000 cells/mL.Among the veterinarians, the year of veterinary degree varied from 1977 to 2020, with a median of 2005. Most of them obtained their degree in Sweden, and worked in East Sweden, Norrland, or West Sweden. The majority was female with some type of post-graduate training in cattle diseases. The number of years in cattle practice varied markedly (0–39 years; median 10–14 years) as did the number of mastitis cases per month (<1 to >15 cases; median 4–8 cases/month).3.2. Questions about DCT to FarmersDescriptive statistics on DCT routines used by the farmers are given in Table 3. Below a short description of the results is given.3.2.1. Use of DCT and Selection of CowsEighty-one percent of the herds used SDCT, mainly treating the occasional cow or fewer than 25% of the cows, and 3% used BDCT. Farmers who did not use DCT (16% of the herds) stated good udder health and/or concern for antimicrobial resistance as the main reasons for not doing so.Most of the SDCT farmers selected cows using UHC, the CSCC at the last milk recording before drying-off, and/or the occurrence of clinical mastitis during the previous lactation. Milk samples for bacteriological examination were taken before deciding on DCT from some of the cows in half of the herds.3.2.2. Choice of AntimicrobialsMost herds (55%) only used Siccalactin® while 22% only used Benestermycin®. Remaining herds used both of those two products, with no preference to one or the other, and a few herds used Carepen® (Boehringer Ingelheim Animal Health Nordics, Malmö, Sweden) an intramammary product only containing short-acting benzyl penicillin procaine. Most (84%) farmers always treated all four udder quarters, while the rest stated that they only treat lactating udder quarters or those with high CMT.3.2.3. Preparing for and Administering TreatmentsAround half of the farmers stated that they clean their hands before treatment and 40% used clean gloves. Almost all farmers used the same routine at the actual treatment, i.e., wiping with serviette provided and full insertion of the tube tip into the teat canal. Most (71%) did not think there were any risks or difficulties with the treatment but 11% stated that kicking of the cow is a risk and 11% that there is a risk to introduce dirt/bacteria into the teat.3.2.4. Assessment of Treatment EffectsThe effects of the DCT were always/almost always evaluated by 58% of the farmers and 19% stated that they rather often examined the CSCC at the first milk recording after calving. CMT was always/almost always and rather often performed in 27% and 18% of the herds, respectively. Milk sampling for bacteriology after calving was rare. Almost all farmers thought that DCT improves cow udder health in the beginning of the next lactation while 64% and 63% thought that the milk production and cow longevity, respectively, improves. Approximately one-fourth of the farmers were uncertain about the effects of DCT on those parameters. Just over half of the farmers did not know if calf health is affected by DCT while 35% did not think it had any effect.3.3. Associations between Demographic Variables and Answers Given by FarmersAs can be seen in Table 3, all six herd variables, but especially milking system and milk production, were associated with the responses given by the farmers. Detailed information on the results is given in Supplementary Table S1. Here, only a short summary with some examples is given.3.3.1. Use of DCT and Selection of CowsIt was less common to use DCT in herds producing <9000 kg ECM or having <53 cows than in herds with higher milk production and ≥138 cows, respectively. In herds with a BMSCC ≥200,000 cells/mL it was more common to state poor udder health as the reason for using DCT while herds producing ≥9000 kg ECM more often stated recommendation from an advisor as the reason than in herds producing less milk.In herds using SDCT and producing ≤11,000 kg ECM, it was more common to treat the occasional cow, compared to in herds with higher production where it was more common to treat at least 25% of the cows. In SDCT herds, it was also more common to treat at least half of the cows, compared to the occasional cow, in herds with a BMSCC ≥200,000 cells/mL than in herds with lower BMSCC. A larger proportion of cows was treated in herds with AMS than in herds with tie-stall milking, and in herds with ≥53 cows than in herds with fewer cows.3.3.2. Choice of AntimicrobialsThe selection of DCT product differed between regions, for example it was more common to use Benestermycin® in West Sweden than in East Sweden and more common to use Siccalactin in Norrland than in Småland and the islands. Benestermycin® was also more commonly used in the highest producing herds while Siccalactin® was more common in herds with lower production.3.3.3. Preparing for and Administering TreatmentsTo wash hands before DCT was more common in herds with AMS or tie-stall milking than in herds with parlour milking. To use clean gloves at DCT was more common in herds producing ≥9000 kg ECM and in herds with milking parlour than in herds with lower production and herds with tie-stall milking or AMS, respectively. Wiping the teats with paper was more common in organic than in conventional herds while wiping with a moist single-use towel was more common in herds with tie-stall milking or parlour than in AMS herds. The use of full insertion of the tip of the intramammary tube was more common in some regions than in others. More herds with organic production stated that there are risks or difficulties with the treatment.3.3.4. Assessment of Treatment EffectsThe proportion of herds that evaluated the effects of DCT by examining the CSCC at the first milk recording varied with the number of cows at the herd. Overall, it was more common to always do so in smaller herds. To evaluate the effects of DCT by CMT after calving was more common in herds with a BMSCC < 200,000 cells/mL than in herds with higher BMSCC and more common in herds with parlour milking or tie-stall milking than in herds with AMS.3.4. Questions about DCT to VeterinariansDescriptive statistics on advice given by veterinarians to dairy farmers and/or their personnel on DCT are given in Table 4. Below, a short description of the results is given.3.4.1. Prescriptions of DCT and Selection of CowsAlmost all veterinarians prescribed DCT, most commonly a few times per month or more. It was not common to regularly use bacteriological investigation of milk samples before prescribing DCT. However, when this was done, it was most common to use an accredited laboratory. Almost all veterinarians stated that they only prescribe one type of DCT product and Siccalactin® was most common. Veterinarians prescribing DCT mainly did so to some cows in the herds, only 8% stated that they sometimes recommend BDCT. The reason for prescribing BDCT was mostly that the herd had problems with Streptococcus agalactiae.The UHC at the last milking before drying-off was the most common factor used when selecting cows for DCT. Other factors influencing the selection were bacterial growth in milk samples, clinical mastitis during lactation, CSCC at the last milk recording, and positive CMT reaction at drying-off. Almost all veterinarians recommended treatment of all four udder quarters. Those who did not, recommended treatment only of quarters with high CMT or lactating quarters.3.4.2. Giving AdviceAlmost two-thirds of the veterinarians did sometimes or often give advice about how to perform the actual treatment. Among those who did not give such advice most stated that they did not perceive an interest for this or did not have enough knowledge as reasons.3.4.3. Preparing for and Administrating TreatmentsWhen asked which of the presented routines they thought should be a part of a good treatment routine use clean gloves and to wipe the teat end with provided serviettes were the most common answers followed by washing hands before treatment and massage of teat/quarter after infusion of the product. Close to half of the veterinarians recommended full insertion of the tip of the tube while just over one-quarter recommended partial insertion. Most veterinarians thought that there are risks or difficulties with the treatment; poor hygiene/risk to introduce bacteria and risk for antimicrobial resistance being the most common risks considered.3.4.4. LegislationOne-third of the veterinarians stated that they did not know about the national legislation on DCT. Among those who prescribed DCT half stated that they always/almost always, and one-third that they often, follow the legislation.3.4.5. Assessment of Treatment EffectsTo evaluate the effects of DCT by CMT after calving and by examining CSCC at the first milk recording were common recommendations. Almost all veterinarians thought that DCT improves cow udder health in the beginning of the coming lactation while around 60% thought it improves milk production and cow longevity. Almost half of the veterinarians stated that they did not know if DCT affects calf health and around one-fourth did not know if DCT affects milk production and cow longevity.3.5. Associations between Demographic Variables and Answers Given by VeterinariansAs can be seen in Table 4, all veterinary variables, but especially post-graduate training and number of mastitis cases per month, were associated with the responses given by the veterinarians. Detailed information on the results is given in Supplementary Table S2. Here, a summary of the results with some examples is given.3.5.1. Prescriptions of DCT and Selection of CowsThe frequency of prescribing DCT varied between regions and was, for example, higher in South and West Sweden than in North Middle Sweden. Veterinarians with post-graduate training in cattle diseases prescribed DCT more often than those without training. Moreover, veterinarians treating more mastitis cases per month prescribed DCT more often than those with fewer cases. Veterinarians working in Norrland performed bacteriological culturing themselves more often than veterinarians in Northern Middle Sweden and South Sweden. It was more common that veterinarians treating relatively few mastitis cases per month sent milk samples to an accredited laboratory than veterinarians treating many cases. The preferred DCT product varied between regions. For example, veterinarians in West Sweden more often prescribed Benestermycin® than veterinarians in Norrland. More veterinarians with post-graduate training prescribed Benestermycin® than those without training. It was also more common to prescribe Benestermycin® among veterinarians treating more than eight mastitis cases/month than among those treating fewer cases.Veterinarians with post-graduate training in cattle diseases more often recommended DCT to the occasional cow than veterinarians without training. Veterinarians without post-graduate training more often stated that a CMT reaction in an udder quarter affected the selection of cows for DCT than veterinarians with training3.5.2. Giving AdviceIt was more common among veterinarians with post-graduate training or treating relatively more mastitis cases to give advice on how to perform the intramammary treatment than among veterinarians without training and treating fewer cases, respectively.3.5.3. Preparing for and Administrating TreatmentsMore female veterinarians than male veterinarians thought that hand washing should be a part of a good DCT routine. Veterinarians treating 1–3 mastitis cases per month more often recommended hand washing than those treating more than eight cases per month. More veterinarians that had graduated in other European countries thought teat wiping with paper should be a part of the DCT routine than veterinarians that had graduated in the Nordic countries. More veterinarians without post-graduate training stated that using full insertion of the tip of the tube should be a part of the treatment routine than those with training.3.5.4. LegislationFewer veterinarians without post-graduate training knew about the legislation on DCT than among veterinarians with such training. The stated compliance with the legislation was associated with year of degree, region, and gender. For example, it was more common among veterinarians graduating before 2001 compared to those graduating after 2009, among veterinarians working in South Sweden compared to those working in Småland and the islands, and among female compared to male veterinarians, to state that they always/almost always complied with the legislation. The compliance with legislation also varied with number of years in cattle practice and number of mastitis cases per month. For example, the compliance was lower among veterinarians with <5 years in cattle practice than among those with 15–19 or ≥25 years in practice, and lower among veterinarians treating 9–15 cases per month than among those treating fewer cases.3.5.5. Assessment of Treatment EffectsThe year of degree was associated with how often the veterinarian recommended follow-up of the effects of DCT by checking CSCC at the first milk recording after calving. For example, this was more often recommended by veterinarians graduating before 2002 than by those graduating 2015–2020. Likewise, more veterinarians with ≥25 years in practice often recommended this practice than those with <5 years in practice. The same two variables were also associated with how often the veterinarian recommended follow-up of the DCT effects by CMT after calving. For example, this recommendation was more common among veterinarians graduating before 2015 than among those graduating 2015–2020, and more common among those with ≥5 years than among those with fewer years in practice. Those two variables, along with post-graduate training, were also associated with how often the veterinarians recommended follow-up using bacteriological examination of milk samples after calving. Such recommendation was, for example, more common among veterinarians graduating before 2002 than among those graduating after 2008, and among veterinarians with ≥25 years in practice than among those with less than 10 years. This recommendation was also more common among veterinarians with post-graduate training in cattle diseases than among those without such training.3.6. Questions about ITS to Farmers and Associations between Herd Variables and Answers Given by FarmersDescriptive statistics on routines for ITS used by the farmers are given in Table 5. In short, most farmers did not use ITS. The most common reason given for this was that they did not think it was necessary due to good udder health (37%). Almost 30% stated that they did not know about ITS. Among those using ITS, the most common reasons given were recommendation from advisor (55%) or udder health problems in the herd (47%). In those herds, 56% treated all cows, mostly based on recommendation from advisor, while remaining herds treated the occasional cow (42% treated every second cow). Most of the herds using ITS did not take milk samples for bacteriology before decision on treatment. Approximately half of the herds using ITS did not think the treatment involved any risks or difficulties while 37% mentioned the risk to introduce dirt/bacteria. Among herds using ITS 46% answered that they sometimes combine ITS with DCT while 25% said that they always do so. To evaluate the effects of ITS treatment by checking the CSCC at first milk recording was common in 54% of the herds and rather common in 14% of the herds while using CMT after calving was common or rather common in 33% and 11% of the herds, respectively. It was rare to take milk samples for bacteriology after calving. Most (85%) of the farmers using ITS thought that treatment improves cow udder health in the coming lactation while 61% and 66% stated that the milk production and cow longevity, respectively, improves (20–30% answered that they did not know). Just under half (43%) of the farmers did not know if calf health was affected by ITS while 31% did not think it had any effect.As can be seen in Table 5, five herd variables, but especially milking system, were associated with the responses given by the farmers. Detailed information on the results is given in Supplementary Table S3. Here, only a short summary with some examples is given. The use of ITS varied between regions and was, for example, more common in herds in North Middle Sweden and South Sweden than in Norrland. It was also more common to use ITS in herds producing >11,000 kg ECM than in herds with lower production, and in herds with AMS, milking parlour, or milking rotary than in herds with tie-stall milking. In herds with BMSCC <200,000 cells/mL it was more common than in herds with higher BMSCC to give the reason that the udder health was so good that ITS was not needed. It was also more common to give this explanation in herds with tie-stall milking or milking parlour than in herds with AMS, and in herds with <138 cows than in herds with more cows.3.7. Questions about ITS to Veterinarians and Associations between Veterinary Variables and Answers Given by VeterinariansDescriptive statistics on advice given on the use of ITS by veterinarians are given in Table 6. In short, almost half of the veterinarians never prescribed ITS. Among those who did, less than 15% stated that they always/rather often use bacteriology before such prescription. If prescribed, most (77%) veterinarians sent the milk samples to an accredited laboratory. It was most common to prescribe ITS to the occasional cow in the herds but approximately one-third said they prescribed ITS to all cows in a herd. The most common reason for this was herd problems with Gram-negative IMI. When selecting cows for ITS, the variable mostly used (69%) was the UHC at the last milk recording before DO. Other variables used for selection were having no case of clinical mastitis (42%), the CSCC at last milk recording (38%), and no CMT reaction at drying off (42%). Almost all veterinarians prescribing ITS recommended treatment of all four udder quarters and 85% of those veterinarians gave advice on how to perform the treatment. Those who did not give such advice stated that they did not perceive any demand for this. Among the veterinarians, 92% thought that clean gloves were important, 87% recommended wiping the teat end with provided serviette, 73% to wash hands before treatment, and 60% to wipe the teats with a moist single-use towel. Twelve percent recommended the use of full insertion while 63% recommended partial insertion of the tip of the tube into the teat canal. Almost all (93%) stated that there are risks or difficulties with treatment, most commonly lack in hygiene/risk of introducing bacteria into the udder. Most veterinarians recommended to sometimes (60%) or always (10%) combine ITS with DCT. To evaluate the effects of ITS treatment by CMT after calving was always to rather often recommended by 60% of the veterinarians while the corresponding numbers for CSCC at first milk recording was 55% and taking milk sample for bacteriology 11%. Half (52%) of the veterinarians thought that ITS treatment improves cow udder health in coming lactation while 28% thought that the milk production improves, 35% that the cow longevity improves, and 9% that calf health improves. Approximately half of the veterinarians stated that they did not know if ITS treatment had any effect on udder health (43%), calf health (56%), milk production (54%), or cow longevity (54%).As can be seen in Table 6, three veterinary variables, but especially post-graduate training, were associated with the responses given by the veterinarians. Detailed information on the results is given in Supplementary Table S4. Here, a summary of the results with some examples is given. The likelihood that the veterinarians prescribed ITS varied between regions and was, for example, higher in Småland and the islands or West Sweden than in East Sweden. Prescribing ITS was more common among veterinarians with post-graduate training and veterinarians treating 9–15 mastitis cases per month than among veterinarians without training or treating fewer cases, respectively. Veterinarians with post-graduate training gave advice about the routines at ITS treatment more often than veterinarians without training. It was also more common that veterinarians with such training stated that ITS treatment improves udder health after calving compared to veterinarians without training who often stated that they did not know if ITS affects udder health. The same attitude was observed for veterinarians treating >1 mastitis case/month compared with those treating <1 case/month.4. DiscussionThis study is the first on farmer routines for DCT and ITS treatment of dairy cows in Swedish herds. Studies, covering some of the questions raised in this study, have previously been performed in Finland and Germany [13,15]. In addition, other studies have reported on proportions of cows treated in different countries, e.g., [16,17,18,19]. To our knowledge, however, this is the most comprehensive study of farmer routines and the first to investigate attitudes of the farmers to the effects of DCT and ITS, and the first to study advice and attitudes of field veterinarians on DCT and ITS treatment of dairy cows.4.1. Use of DCT and Comparisons with National and International RecommendationsThe results clearly showed that farmers and veterinarians followed the national recommendations to use SDCT. The results at herd level were similar to those in a recent Finnish study [13]. In contrast, most herds in Canada, Germany, and US used BDCT [15,16,17,19]. The Nordic model has always been to use SDCT to minimize prophylactic use of antibiotics [2], and this is reflected, for example, in the recommendations and guidelines in Sweden [3] and Denmark [8]. In many other countries, BDCT has been the norm for many years, but in recent years, more countries have introduced SDCT with good results reviewed by [20]. According to a recent meta-analysis the results varied between studies comparing SDCT and BDCT, but the combined data indicated a slightly higher risk for IMI after calving when using SDCT [20]. However, as the number of studies that did not use ITS in combination with DCT was small, it is difficult to distinguish the effect of the use of DCT alone from that of the combined use. In a recent Finnish study [21], significant differences in SCC and milk production were not found at herd level between herds using SDCT and herds using BDCT or no DCT.When using SDCT it is important to select cows that are likely to benefit from the treatment to avoid unnecessary use of antibiotics. In the present study, the most common variable used by farmers and veterinarians for selection of cows was the UHC at the last milk recording before drying-off, which accords with national recommendations. However, only two-thirds of the farmers stated that they used this variable, indicating a need for improvement. Good knowledge about types of IMI and their antimicrobial susceptibility is also important when selecting cows for DCT to ensure selection of a suitable DCT product. In the present study, however, only a third of the veterinarians recommended milk sampling for bacteriology before DCT. Moreover, the results indicated that this is seldom done by the farmers, so improvements are also needed in this area. To use some measure of the CSCC to identify cows likely to have an IMI when selecting cows for DCT is also recommended by branch organisations in other countries, but the CSCC cut-offs used, and the number of milk recordings included vary [7,8]. They also state that it is important to know the pathogens present in the herd and their antimicrobial susceptibility. Several studies have been performed on how to select cows for DCT; however, despite this, the optimal selection method cannot be considered scientifically proven. Examples of tools suggested for identification of infected/non-infected cows at drying-off are: checking the CSCC at the last 1–3 milk recordings before drying off; whether or not the cow has had clinical mastitis during the lactation; and the performing of bacteriological examination of milk samples [10,20,22,23]. Other studies indicate that CMT can be used to identify udder quarters with subclinical IMI at drying-off [24,25,26]. The most common CSCC cut-off used for infected/non-infected is 200,000 cells/mL, but in other studies the cut-off has varied between 100,000 and 300,000 cells/mL [20]. As a high CSCC at the last milk recording before drying-off can indicate poor prognosis of DCT cure, the recommendation may be to cull such cows instead of using DCT [22].According to the national recommendations, thorough cleaning of the teat end before DCT is important, preferably with alcohol, but more details were not given. However, almost none of the farmers and only a small proportion of the veterinarians stated that they use/recommend cleaning with cotton moistened with alcohol. Instead, most used, or recommended, cleaning with the provided serviette. Teat dipping or spraying directly after treatment, and control of the udder and teat dipping or spraying morning and evening for 24 h, were also included in the recommendations. These recommendations were, however, not given as alternatives in the questionnaire and only a few and none, respectively, added this routine under comments. It was clear that the description of hygienic measures when using DCT in the national recommendations was not detailed enough and needs updating. In comparison, the UK recommendations provide much more information using both pictures and videos [7].Compliance with the recommendation to evaluate effects of DCT after calving varied markedly within both groups and the results indicate room for improvement. For example, only 60% of the farmers stated that they always examined the CSCC and only one-fourth used CMT after calving. The proportion of veterinarians that always/almost always recommended evaluation of the effects of DCT after calving was also low. Moreover, around two-thirds of the farmers did not think there were any risks or difficulties associated with DCT while the veterinarians were of the opposite opinion. The fact that so few farmers knew about the risks is especially noteworthy as the farmers are those performing the actual treatments. It was also clear that many of the veterinarians needed to improve their knowledge on, and compliance with, the national legislation on use of DCT.At the time of the study, advice on pros and cons with full or partial insertion of the tip of the intramammary tube into the teat canal or on massage of the udder quarter after treatment was not included in the recommendations. Although this information was not provided by the pharmaceutical company responsible for the DCT products at the time of the study, their recommendation was to use partial insertion and massage the teat and udder after infusion (Manske T., personal communication, 2020). To use partial insertion of the tip of the tube rather than full insertion is also recommended in several other countries [5,6,7]. Few studies comparing partial and full insertion have, however, been published. In one study, partial insertion led to fewer new IMI and increased treatment efficacy, and in another study an association was found between being a herd with a low BMSCC and herd use of partial insertion [27,28]. The theory is that the risk for mechanical damages in the teat canal and the risk to introduce unwanted bacteria into the udder decreases when using partial insertion/short tip. It is also considered favourable that the teat canal is exposed to antibiotics as bacteria can colonize/infect the teat canal.4.2. Use of ITS and Comparisons with National and International RecommendationsIn the present study, less than one-fifth of the farmers used ITS and it was not so common that the veterinarians prescribed ITS. At the time of the study advice on ITS was not included in the national recommendations or guidelines. The results were somewhat lower than those reported from Finland and Germany where around one-third of the farmers reported the use of ITS [13,15]. In contrast, most farmers used ITS in UK [18], which was in line with the UK recommendations [7]. The use of ITS is also recommended and common in North America [5,6,17]. Several studies on ITS have been performed in other countries, and a recent meta-analysis indicated that ITS reduces the occurrence of new IMI at calving and the presence of clinical mastitis; although, the results vary between studies, especially if the herd has problems with environmental bacteria or not [12]. In Sweden, the main mastitis-associated pathogens are contagious bacteria making the interest of ITS smaller. Moreover, a couple of Swedish case studies did not indicate any positive effects of ITS on udder health [29,30].According to the questionnaires, ITS was only used or recommended by a relatively small proportion of the farmers and veterinarians. Thus, the results of the questionnaires should be interpreted with care as they are based on small numbers of respondents. Almost half of the farmers using ITS were, however, aware of the risks and difficulties with ITS treatment, and the proportion was numerically higher than for the same question regarding DCT. Although the use of ITS was low, there is still the need to include ITS treatment in the national recommendations in the future.4.3. Attitudes to the Importance of DCT and ITS TreatmentThe attitudes to the importance of DCT for cow udder health and milk production in the beginning of the next lactation, cow longevity, and calf health varied among farmers and veterinarians. The variation within group was smallest for cow udder health and most of the respondents thought that it would improve due to DCT. Substantial proportions of farmers and veterinarians were, however, uncertain on the effects of DCT on cow longevity and calf health. Associations between DCT and better udder health have been clearly shown in several studies, e.g., [1,20,31]. It is also clear that cows with healthy udders produce more milk than cows with IMI and mastitis. Cows with healthy udders also have a lower risk for pre-mature culling improving the longevity of the cows [32]. However, whether there is an effect of DCT on calf health is less clear. It is likely that other factors are more important for calf health, but studies have shown that the absorption of colostral antibodies is reduced if the colostrum contains high numbers of bacteria [33]. This may in turn increase the risk for calf diarrhoea and reduced growth.The attitudes to the importance of ITS for cow health and production, or for calf health, did also vary markedly within the groups. Overall, the uncertainty was rather high among the veterinarians, probably reflecting their limited experience of ITS. Among farmers responding to these questions, most thought that ITS improves cow udder health, milk production, and longevity. However, as the number of farmers using ITS was low the results must be interpreted with care. According to studies from other countries the use of ITS is associated with improved udder health and thus with better milk production and a lower risk for culling [12]. The Swedish experiences are very limited but have so far been less favourable [29,30].4.4. Associations with Demographic VariablesAmong the herd variables, milking system and milk production were significantly associated with the largest number of questions on DCT and ITS while production form was associated with the lowest number of questions. Among the veterinary variables, post-graduate training and number of mastitis cases per month were significantly associated with the largest number of questions, and year of degree and gender were associated with the lowest number of questions. In both groups, more associations were found between the variables and the questions on DCT than between the variables and the questions on ITS, probably reflecting the smaller number of respondents for the questions on ITS. Several of the herd variables were most likely influenced by each other; for example, where both milking system and milk production are associated with the number of cows in the herd, it is difficult to evaluate if a single variable or a combination of variables is important. Moreover, the number of cows per herd varies between regions of the country [34]. In the present study, it was not possible to perform multivariable analyses to further elucidate the effects.All three variables, i.e., milking system, milk production, and number of cows per herd, were, for example, associated with, if, or how often DCT was used. Overall, use of DCT was more common in herds with relatively higher milk production and number of cows per herd. In addition, it was more common in AMS than in herds with tie-stall milking. These findings are in line with national data from the milk recording scheme showing that herds with more cows and herds with AMS have higher bulk milk SCC than smaller herds and other milking systems, respectively [35].As expected, the BMSCC was associated with several questions on both DCT and ITS. The answers were in line with the larger need for DCT in herds with higher BMSCC as those have more cows with subclinical mastitis. Interestingly, herds with low BMSCC evaluated the effect of DCT directly after calving, possibly indicating higher awareness of the benefits of preventive measures.Few differences were observed between organic and conventional herds. However, more organic farms thought there were risks or difficulties with DCT. The reasons for this finding are not clear but it is possible that organic herds are more concerned by withdrawal times and antimicrobial resistance.For both farmers and veterinarians, region was associated with some of the questions. For example, the choice of DCT product varied between regions in both groups. For the veterinarians, differences between regions were also observed, for example, for how often they prescribe DCT and ITS. The reasons for these differences are not known but aspects like tradition among farmers and veterinarians may be of importance. The through-put of veterinarians may also vary between regions.Year of degree and years in cattle practice are both indicators of experience and probably also type of education and tradition. It was therefore not surprising that the variables had similar associations with factors such as compliance with legislation and evaluation of the effects of DCT. Other factors indicating differences in experience but also in interest in cattle diseases were post-graduate training and number of mastitis cases/month. It was not surprising that both variables were associated with the answers to several questions and that the answers indicated better knowledge about DCT and ITS, and better compliance with recommendations among those with post-graduate training and many cases per month.4.5. Methodological ConsiderationsUnfortunately, the proportions of respondents were rather low for both farmers (14%) and veterinarians (25%). However, a similar response rate (13%) was also observed in a web-based study on farmers in Finland [13] while a web-based study on mastitis sent to Swedish veterinarians had a higher response rate (36%) [36]. The reasons for the low numbers of respondents are not known but lack of time is probably an important factor. It is also possible that a web-based questionnaire results in lower response rate than other types of questionnaires. For example, Bertulat et al. [15] had a response rate of 49% for a questionnaire to milk producers performed in connection with a physical meeting and McDougall et al. [28] had a response rate of 44% when sending the questionnaire via postal mail.Questionnaire studies must always be interpreted with care as respondents may not be representative for the population in question. In our study, the farmers participating were well spread geographically in the country but the proportion of herds with free-stalls and with AMS as well as the number of cows per herd was larger than the average among herds affiliated to the official cow control scheme for that year (55%, 33%, and 92 cows, respectively) [34]. Moreover, the results indicate that participating herds also had higher milk production and lower BMSCC than the average herd (10,417 kg ECM/cow and 211,000 cells/mL, respectively) (Nyman, A.-K., personal communication 2021). Unfortunately, we were not able to control if the responding veterinarians were representative for the target group. However, in a previous web-based questionnaire study to veterinarians [36], the distribution of gender and year of veterinary degree among respondents did not differ from the target group. Given the facts mentioned above on representativity in combination with the relatively low response rate, the results, including proportions of cows treated with DCT and/or ITS, must be interpreted with care.The results should also be interpreted with caution as the risk for type I errors, to find significant results even though there are no true associations, increases when many risk factors are tested. Hence, some of the associations found might be just due to chance. As it is impossible to know which of these associations could be due to chance, additional studies are needed to further confirm the findings in this study.5. ConclusionsThe routines used by the farmers and the advice given by the veterinarians responding to the questionnaires were in many areas in line with the recommendations available at the time, but the answers also indicated room for improvement in some areas. We also found interesting associations between routines used and advice given and the tested herd and veterinary variables, respectively. The results, as well as those on attitudes to the effects of DCT and ITS on animal health and production, indicated a need for more education. We also found that the existing recommendations were insufficient and in need of an up-date. Therefore, new national recommendations were produced and spread among target groups.

animals : an open access journal from mdpi

10.3390/ani13050817

PMC10000054

As the price of fishmeal continues to rise, it is urgent to seek new protein sources to decrease fishmeal inclusion in aquafeeds. However, anti-nutrient factors limit the application of plant proteins in aquafeeds. Among alternative protein sources, cottonseed protein concentrate (CPC) and Clostridium autoethanogenum protein (CAP) contain are considered promising substitutes because of their few anti-nutrient factors and complementary effect in amino acid composition. In this study, we evaluated the effects of replacing fishmeal with a mixture of CPC and CAP (1:1) on growth performance, nutrient utilization, serum biochemical indices, and intestinal and hepatopancreas histology of rainbow trout (Oncorhynchus mykiss). Based on the results, the mixture of CPC and CAP could replace 50% of dietary fishmeal without negative effects on rainbow trout.

The purpose of this study was to develop the potential of cottonseed protein concentrate (CPC) and Clostridium autoethanogenum protein (CAP) in the diet of rainbow trout (Oncorhynchus mykiss) by evaluating the effects of substituting fishmeal with a CPC and CAP mixture on growth performance, nutrient utilization, serum biochemical indices, intestinal and hepatopancreas histology. In a basal diet containing 200 g/kg fishmeal (Con), the mixture of CPC and CAP (1:1) was used to reduce dietary fishmeal to 150, 100, 50 and 0 g/kg, to form five diets with the same crude protein and crude lipid contents (CON, FM-15, FM-10, FM-5 and FM-0). Then, the five diets were fed to rainbow trout (35.00 ± 0.05 g) for 8 weeks. The weight gain (WG) and feed conversion ratio (FCR) of the five groups were 258.72%, 258.82%, 249.90%, 242.89%, 236.57%, and 1.19, 1.20, 1.24, 1.28, 1.31, respectively. FM-5 and FM-0 groups showed significantly lower WG and higher FCR than the CON group (p < 0.05). In terms of whole-body composition, such as moisture, crude ash, and crude protein, no significant difference was observed among all the groups (p > 0.05), except that significantly higher crude lipid content was detected in the FM-0 group than in the CON group (p < 0.05). In the FM-5 and FM-0 groups, protein efficiency, protein retention, intestinal protease activity and amylase activity were significantly lower than in the CON group (p < 0.05). Compared to the CON group, the serum contents of glucose and total cholesterol in the FM-0 group as well as MDA in the FM-5 and FM-0 groups were significantly increased, and catalase, superoxide dismutase, and total antioxidant capacity were decreased (p < 0.05). In intestine and hepatopancreas histology, the intestinal villus height in the FM-5 and FM-0 groups and villus width in the FM-0 group were decreased significantly (p < 0.05), while no significant difference in hepatopancreas morphology was observed among all the groups except that some vacuolization was observed in the FM-0 group (p > 0.05). In summary, the mixture of CPC and CAP can effectively replace 100 g/kg fishmeal in a diet containing 200 g/kg fishmeal without adverse effects on the growth performance, nutrient utilization, serum biochemical, or intestinal and hepatopancreas histology of rainbow trout.

1. IntroductionIn aquaculture, more than 50% of production costs are incurred by feed [1], and fishmeal is the primary source of protein due to its balanced amino acid composition, high crude protein content, and unknown growth factors [2,3]. Due to the shortage of marine resources and the growing demand for fishmeal in aquafeeds, fishmeal price has been rising continuously. Therefore, new protein sources must be sought to replace or decrease fishmeal inclusion in aquafeeds. Previous studies have shown that some plant protein sources can partly or completely replace dietary fishmeal, including soybean meal [4], cottonseed meal [5], peanut meal [6], canola protein isolate [7], and corn protein [8]. However, the anti-nutrient factors in plant proteins may adversely affect the performance of fish [9]. Therefore, exploring new plant protein sources with low anti-nutrient factors seems to be critical for their wide application in aquafeeds.Cottonseed protein concentrate (CPC) is derived from shelled cottonseed after oil extraction, dephenolization, and low-temperature drying. Generally, the crude protein content of CPC can reach 60% to 70% with plenty of arginine, histidine, and phenylalanine, and the free gossypol content is very low [10]. CPC was reported to replace 385 g/kg fishmeal in a basal diet containing 700 g/kg fishmeal without adversely affecting the growth and flesh quality of largemouth bass (Micropterus salmoides) [10]. The CPC prepared from glandless seeds completely replaced dietary fishmeal (400 g/kg) without adverse impacts on growth performance and feed utilization of black sea bass (Centropristis striata) [11]. Moreover, CPC has also been reported to partially replace fishmeal in the diets of golden pompano (Trachinotus ovatus) [12], hybrid grouper (Epinephelus fuscoguttatus × E. lanceolatus) [1], and large yellow croaker (Larimichthys crocea) [13].As a new type of single-cell protein, Clostridium autoethanogenum protein (CAP) is obtained by liquid fermentation of Clostridium ethanolum with CO and ammonia as carbon and nitrogen sources. After the centrifugation and drying, the CAP product contains high levels of crude protein (80%) with high digestibility, high lysine content, and fewer anti-nutrient factors, but the arginine level is relatively low [14]. The successful replacement of fishmeal with CAP has been reported in some aquaculture species. For example, no adverse effects on the growth performance, feed utilization, or intestinal histology were found in largemouth bass when CAP replaced 150 g/kg fishmeal in a basal diet with fishmeal inclusion of 350 g/kg [15]. Similar results have also been reported in obscure pufferfish (Takifugu obscurus) [16], large yellow croakers [17], Pacific white shrimp (Litopenaeus vannamei) [18,19], juvenile turbot (Scophthalmus maximus L.) [15], and black sea bream (Acanthopagrus schlegelii) [20].Rainbow trout (Oncorhynchus mykiss) is a cold water fish, widely cultured throughout the world. Rainbow trout is a fast growing fish with tender flesh, no intermuscular spines, and plenty of DHA and EPA [21]. Among salmon and trout, rainbow trout production is the second largest after Atlantic salmon (Salmo salar) [22]. As a carnivorous fish, the commercial diet for rainbow trout usually contains a high level of fishmeal, thus, reducing the fishmeal inclusion in feed is of great significance to develop sustainable aquaculture of rainbow trout. To date, it has been reported that some plant protein sources can replace fishmeal in rainbow trout diets, such as enzymatical soybean meal [23], fermented soybean meal [24], corn protein concentrate [8], canola: pea [25]. Recently, Zhao et al. [26] reported that CPC successfully replaced 50% of dietary fishmeal without adverse impacts on growth performance or antioxidant capacity of rainbow trout. However, the control diet in that study contained a high level of fishmeal of up to 400 g/kg. Additionally, the inclusion of CAP in the diet has not yet been investigated in rainbow trout.In terms of amino acid composition, CPC contains much arginine and little lysine and methionine, while CAP contains plenty of lysine and methionine and relatively low arginine levels. From this point of view, the two protein sources may have complementary effects in amino acid composition. Therefore, we evaluated the effects of substituting fishmeal with a mixture of CPC and CAP (1:1) on the growth performance, nutrient utilization, serum biochemical indices, and intestinal and hepatopancreas histology in the present study. The findings will develop the potential of CPC and CAP in carnivorous fish feeds and provide a low fishmeal diet for rainbow trout. 2. Materials and Methods2.1. Ethics StatementAll the procedures for handling animals involved in this experiment are in accordance with the regulations of the Experimental Animal Ethics Committee and the Institutional Animal Care Committee of Shanghai Ocean University (Approval code: SFI 2020-23 of 20 May 2020).2.2. Experimental Diets and DesignFirst, a control diet was formulated to contain 200 g/kg fishmeal (CON). Then, the mixture of CPC and CAP (1:1) was used to reduce dietary fishmeal to 150, 100, 50 and 0 g/kg to form five diets with the same contents of crude protein (430 g/kg) and crude lipid (100 g/kg) (CON, FM-15, FM-10, FM-5 and FM-0). According to the method described by Xu et al. and Yao et al. [10,19], the protein ingredients such as fishmeal, CAP, CPC, etc., were ground and screened through a 60-mesh sieve, then mixed with liquid ingredients (fish oil, soybean oil, soybean lecithin). The five diets were prepared to form sinking pellets with a diameter of 3 mm using a single-screw extruder (LX-75 extruder, Longxiang Food Machinery Factory, Hebei, China). The extruding and air-drying temperatures were 90 ± 5 °C and 30 ± 2 °C, respectively. Then, all diets were stored at 4 °C until use. Table 1 shows the formulation and proximate composition of the diets.CPC and CAP were provided by Chengrun Jinlan Biotechnology Co., Ltd. (Xinjiang, China) and Shoulang New Energy Technology Co., Ltd. (Tangshan City, China), respectively. The crude protein content of CPC and CAP was 615.1 g/kg and 842.1 g/kg, respectively, and the crude lipid content was 23.6 g/kg and 1.9 g/kg, respectively. The free gossypol content in CPC was 172.8 mg/kg. Fishmeal was steam-dried fishmeal (TASA), and the crude protein and lipid contents were 682.1 g/kg and 90.0 g/kg, respectively. Table 2 shows the proximate composition and amino acids profile of the four ingredients.2.3. Management of Experimental Fish and FeedingRainbow trout were obtained from a local aquaculture farm in Sichuan (China). After transportation to the aquaculture station (Binhai, Shanghai, China), all fish were adapted to the environment and diets by feeding control diets for 2 weeks. A total of 300 juvenile rainbow trout (35.00 ± 0.05 g) were selected and randomly distributed into 15 indoor polyvinyl tanks with diameter, height, and water capacity of 1.0 m, 0.8 m, and 650 L, respectively. Three replicates (tanks) were designed for each treatment and each tank contained 20 fish. The system ran on the recirculation regime with a flowing rate of 10 L/min for each tank. During the feeding period, all the fish were fed at 09:00 and 16:00, and the feed intake was adjusted to ensure that no diet residue was found within 5 min when feeding was finished. The feces waste was siphoned out at the 2nd hour after feeding, and the cultured water was renewed (about 1/3) twice a week. During the feeding period of 56 days, the water quality was measured by measuring temperature (12–14 °C), dissolved oxygen content (6–8 mg/L), ammonia nitrogen content (<0.1 mg/L) and nitrite content (<0.1 mg/L).2.4. Samples CollectionAt the beginning of the feeding trial, ten fish were sampled and stored at −20 °C to determine the whole-body composition of initial fish. At the end of the feeding trial, all the fish were deprived of diets for 24 h, and then counted and bulk weighed for each individual tank to calculate weight gain (WG), feed conversion ratio (FCR), specific growth rate (SGR), survival, and feed intake (FI). Six fish per tank were randomly selected, of which three fish were used for the measurement of final fish whole-body composition. For the other three fish, body length and weight were measured to calculate condition factor (K), and then the blood was collected from the cordial vein with a syringe. After centrifuging at 3500 rpm for 10 min (4 °C), the supernatant was stored at −80 °C to determine serum biochemical indices. Then, the three fish were dissected, and the weight of visceral and hepatopancreas were measured to calculate the viscerosomatic index (VSI) and hepatosomatic index (HSI). Then, hepatopancreas (1 cm × 0.5 cm × 0.5 cm) and foregut (1 cm) tissues were stored in Bouin’s solution to observe tissue structure. The rest of the intestines were kept at −80 °C to determine digestive enzyme activity.2.5. Measurement Indicators and Methods2.5.1. Growth Performance and Physical IndicesThe growth performance and physical indices were calculated as follows:Survival (%) = 100 × (final number of fish/initial number of fish) WG (%) = 100 × [final weight (g) − initial weight (g)]/initial weight (g) FCR = feed consumption (g)/weight gain (g) SGR (%/day) = 100 × [ln final weight (g) − ln initial weight (g)]/days K (g/cm3) = 100 × [body weight (g)/body length3 (cm)3] VSI (%) = 100 × [visceral weight (g)/body weight (g)] HSI (%) = 100 × [hepatopancreas weight (g)/body weight (g)] FI (g/fish/d) = feed intake (g)/[(final fish number + initial fish number)/2]/days2.5.2. The Diets and Whole-Body CompositionAOAC method was used to measure moisture, crude ash, crude protein, and crude lipid contents in diets and whole-body composition [27]. To determine the moisture content, samples were dried in an oven at 105 °C to constant weight. The combustion method, the auto Kjeldahl system (2300 Auto analyzer, Foss Tecator, AB, Hoganas, Sweden), and the chloroform–methanol method were used to determine the contents of crude ash, crude protein, and crude lipid, respectively. The dietary amino acid composition was determined according to the description by Xu et al. [10]: The dried sample (70 mg) was added with 6 mol/L hydrochloric acid, then hydrolyzed at 110 °C for 24 h in a vacuum. After dilution (1:10), the amino acid content was determined by automatic analyzer (S-433D, Sykam, Munich, Germany).2.5.3. Serum Biochemical IndicesThe serum activities of aspartate aminotransferase (AST), alanine aminotransferase (ALT), total antioxidant capacity (T-AOC), superoxide dismutase (SOD), catalase (CAT), and serum contents of malondialdehyde (MDA), triglyceride (TG), glucose (GLU), total cholesterol (TCHO), and total protein (TP) were determined by kits supplied by Nanjing Jiancheng Biological Co., Ltd., Nanjing, China.2.5.4. Nutrient RetentionThe protein efficiency ratio (PER), protein retention (PR), and lipid retention (LR) were calculated as follows:PER (%) = 100 × [final body weight (g) − initial body weight (g)]/protein intake (g) PR (%) = 100 × [final body weight (g) × crude protein content of the final whole fish − initial body weight (g) × crude protein content of the initial whole fish]/protein intake (g) LR (%) = 100 × [final body weight (g) × crude lipid content of the final whole fish − initial body weight (g) × crude lipid content of the initial whole fish]/lipid intake (g)2.5.5. Intestinal Digestive Enzyme ActivityAfter thawing at 4 °C, intestinal samples (0.1 g) were homogenized with nine times volume of ice-cold normal saline (1:9 w/v), and then centrifuged at 3000 rpm for 10 min at 4 °C. The supernatant was stored at −20 °C for digestive enzyme activity.Protease activity was measured using the Folin phenol method [28], and the amount of enzyme that hydrolyzes casein and produces 1 μg tyrosine per minute (1 mg tissue protein) at pH 7.2 and 37 °C was defined as one unit (U/mgprot). Amylase activity was measured by iodine–starch colorimetry [29], and tissue protein (1 mg) reacted with the substrate at 37 °C for 30 min to hydrolyze 10 mg starch was defined as one unit (U/mgprot). The above indices were measured using kits supplied by Nanjing Jiancheng Biological Co., Ltd., Nanjing, China.2.5.6. Intestinal and Hepatopancreas HistologyFor the embedding of foregut and hepatopancreas, we referred to the method described by Yang et al. [23]. A slicer (Leica RM2235, Heidelberg, Germany) was used to cut the sections (7 μm), which was then stained by hematoxylin-eosin. The microscope (Nikon YS100, Melville, NY, USA) was used to observe and photograph morphology of intestine and hepatopancreas, and then the villus height, villus width and muscle thickness were measured with image analysis software (Image J14.0), referring to the description of Zhao et al. [26].2.6. Statistical AnalysisAll data were presented as mean ± standard deviation (mean ± SD). Excel and SPSS 26.0 software were used to conduct a one-way ANOVA. Statistical significance among groups was determined using the Tukey’s multirange test if significant differences were detected.3. Results3.1. Growth Performance and Physical IndicesIn Table 3, all the groups showed high survival, up to 100%. In the FM-15 and FM-10 groups, WG, FCR, and SGR were similar to those in the CON group (p > 0.05), while the FM-5 and FM-0 groups showed significantly lower WG and SGR (−6.12% and −3.51%, −8.56% and −4.82%), and higher FCR (+0.09, +0.12), than the CON group (p < 0.05). All the groups presented similar K, VSI, and HSI (p > 0.05).3.2. Whole-Body CompositionThe whole-body composition is shown in Table 4. The crude lipid content in the FM-0 group was significantly higher than in the CON group (p < 0.05). There were no significant differences in the content of moisture, crude ash, and crude protein among all the groups (p > 0.05). 3.3. Nutrient Utilization and Intestinal Digestive Enzyme ActivityIn Table 5, the increasing replacement of fishmeal with the mixture of CPC and CAP tended to decrease the protease and amylase activity, and PER. The protease activity in the FM-10, FM-5, and FM-0 groups and the amylase activity in the FM-5 and FM-0 groups was significantly lower than in the CON group (p < 0.05). The FM-5 and FM-0 groups also showed significantly lower PER and PR than the CON group (p < 0.05). The LR of the FM-5 group was significantly lower than that of the other groups (p < 0.05).3.4. Serum Biochemical IndicesTable 6 shows the results of biochemical indices in serum. Serum contents of GLU and TCHO in the FM-0 group were significantly higher than those in the CON group (p < 0.05). Compared to the CON group, the FM-5 and FM-0 groups showed lower activities of SOD, T-AOC, and higher MDA content (p < 0.05). No significant difference in the activity of CAT, AST, ALT, or in the content of TG or TP was detected among all the groups (p > 0.05), but the increasing replacement of fishmeal with the CPC and CAP mixture tended to decrease the activity of CAT. 3.5. Intestinal MorphologyTable 7 and Figure 1 depict the intestinal morphology. The villus height and muscle thickness in the FM-5 and FM-0 groups, as well as villus width in the FM-0 group, were significantly smaller than in the CON group (p < 0.05).3.6. Hepatopancreas MorphologyAs shown in Figure 2, vacuolation and lipid droplets were observed in some hepatocytes in the FM-0 group, while the nucleus and cell structure of hepatocytes in the other groups showed no abnormality.4. Discussion4.1. Growth Performance, Whole-Body Composition and Nutrient UtilizationThe high replacement of fishmeal with plant protein sources may have adverse effects on the performance of aquatic animals [8]. In basal diets with fishmeal inclusion of 250 g/kg, 450 g/kg, 452 g/kg, and 400 g/kg, fermented soybean meal [24], solvent-extracted cottonseed meal [30], corn protein [8], and CPC [26] could reduce dietary fishmeal content to 150 g/kg, 225 g/kg, 338.1 g/kg, and 200 g/kg, respectively, without affecting the growth performance of rainbow trout. In the present study, the fishmeal content in the basal diet was only 200 g/kg, and the replaced fishmeal reached 100 g/kg. Both were much lower than the reported values in the above studies, realizing a low fishmeal diet for rainbow trout feed. It indicates that the CPC and CAP mixture is an excellent plant protein source to replace fishmeal in aquafeeds.However, the WG and PR of rainbow trout were significantly decreased, and FCR was significantly increased, when dietary fishmeal was reduced to 50 g/kg by inclusion of the plant mixture (Table 3). In largemouth bass, the WG and feed utilization were significantly decreased when CPC was used to substitute 70% of fishmeal in the basal diet containing 700 g/kg fishmeal [10]. In Pacific white shrimp, the WG and PER were also decreased by the substitution of 45% of fishmeal with CPC in a diet containing 250 g/kg fishmeal [31]. Similarly, the WG of largemouth bass was significantly decreased when dietary fishmeal was reduced from 350 g/kg to 150 g/kg by CAP inclusion [15]. The substitution of 45% of dietary fishmeal with CAP also significantly decreased the WG and PER of Pacific white shrimp [19]. There are several reasons connected with the decreased growth and nutrient utilization by the excessive replacement of fishmeal with the CPC and CAP mixture in the present study: (1) The mixture contains a lower level of methionine than fishmeal (Table 2), and the high substitution of fishmeal may lead to deficiency of methionine. Methionine is a precursor of protein synthesis, which can be converted into taurine to promote fish growth [32,33]. (2) The high level of arginine in CPC may be another factor affecting the growth of rainbow trout. As excessive arginine causes antagonism with lysine, a high arginine level in the diet may negatively affect the lysine utilization and the growth of fish [34]. (3) Fishmeal contains high levels of taurine, hydroxyproline, and some unknown growth factors that are essential for growth and intestinal health, while bacterial and plant proteins lack these active compounds [15,19]. mTOR is a key signal molecule that regulates growth [35], and the efficiency of activating the mTOR signal pathway by fishmeal was found to be significantly higher than that by other protein sources such as CPC [36], corn gluten meal [35], and CAP [14,37]. Taurine is a conditionally essential amino acid in fish, which can promote the growth of fish [38]. The dietary supplementation of taurine reduced fishmeal inclusion from 240 g/kg to 160 g/kg without negative effects on the growth performance of largemouth bass [39]. In a low fishmeal diet (70 g/kg), the supplementation of taurine and selenium yeast increased the growth and protein utilization of black sea bass [40]. Hydroxyproline is a semi-essential amino acid in aquatic animals, which is a characteristic amino acid in collagen. In a turbot diet with high plant protein source inclusion, the supplementation of hydroxyproline (6 g/kg) increased the replaced ratio of fishmeal from 40% to 60% [41]. Aksnes et al. [42] also reported that the supplementation of hydroxyproline (2.9 g/kg) in a high plant protein source diet increased the WG of Atlantic salmon (Salmo salar L.) by 14%. (4) CAP and CPC have lower digestibility than fishmeal. The high fiber content in CPC and the presence of the cell wall in CAP may be an important factor affecting nutrient digestibility [43,44,45]. Li et al. [46] reported that the apparent digestibility coefficient of dry matter and crude protein of CPC were lower (−11.75% and −8.23%) than fishmeal in Pacific white shrimp, and high inclusion levels of CAP may reduce digestibility in animals.The present findings showed no significant difference in whole-body composition such as moisture, crude protein, and crude ash when CPC and CAP mixture was used to substitution fishmeal. Similar results were also reported in rainbow trout [26] and black sea bream [20]. However, the FM-0 group presented significantly higher crude lipid content than the CON group. Liu et al. [47] found that the substitution of 75% of dietary fishmeal with CPC significantly increased the whole-body content of crude lipid in largemouth bass. Similar results were also described in southern flounder (Paralichthys lethostigma) [48]. Due to the decrease in feed utilization (including protein utilization), the intake of protein can not be effectively utilized to synthesize body protein, but rather is converted into lipid for deposition.4.2. Intestinal Morphology and Digestive Enzyme ActivityThe primary organ for digesting and absorbing nutrients of fish is the intestine, which impacts normal growth and development [49]. Intestinal protease and amylase are the main digestive enzymes in the intestine, and the villus height and villus width determine the contact area between mucosal epithelial cells and chyme, while muscle thickness is beneficial to intestinal peristalsis and chyme propulsion [50]. In the study, the intestinal protease and amylase activity tended to decrease with the decreasing inclusion of fishmeal, and the digestive enzyme activity, villus height, and muscle thickness in the FM-5 and FM-0 groups were significantly lower than those in the CON group. Wu et al. [51] found that the replacement of 45% of dietary fishmeal with CAP significantly decreased intestinal protease and amylase activity, resulting in intestinal damage and inflammation in large yellow croaker. In Pacific white shrimp, the intestinal villus height was significantly decreased when 40% of dietary fishmeal was replaced with CAP, but the supplementation of phosphorus improved the intestinal health [52]. Similar results were also observed in silver sillago (Sillago sihama Forsskál) [53] and large yellow croaker [13] when CPC replaced a high proportion of fishmeal. With the decrease in fishmeal inclusion, the bioactive substances such as taurine and trimethylamine oxide tend to decrease, which may reduce the ability of the pancreas and intestinal glands to secrete digestive enzymes. In juvenile cobia (Rachycentron canadum) [54] and golden pompano [55], the supplementation of taurine in the diet was reported to increase the trypsin activity and improve the intestinal structure, respectively.4.3. Serum Biochemical IndicesAST and ALT are two important transaminases in amino acid metabolism, which exist in the heart and liver. When the liver is damaged, the cell membrane permeability of hepatocytes increases, and the two transaminases in the cytoplasm will be released into the blood, resulting in increased activity in the blood [56]. No significant difference in serum activities of AST and ALT was observed in the present study, although the hepatocytes in the FM-0 group presented some vacuolation and lipid droplets, indicating lipid accumulation in the liver. Such a result was consistent with the increasing whole-body crude lipid content in this group. However, in hybrid grouper [57] and largemouth bass [37], AST and ALT activity in serum were significantly increased when CPC and CAP replaced 36% and 50% of dietary fishmeal, respectively. It could be that the combination of CAP and CPC in the present study produced complementary effects and overcame their respective shortcomings. In addition, CPC is rich in arginine, and arginine could protect the liver by reducing the production of pro-inflammatory cytokines and free radicals [58].Physiologically, blood GLU reflects the health of fish, which was also measured as a stress marker to evaluate the stress response caused by dietary changes [59]. The present study showed that blood GLU content was significantly increased only in the FM-0 group. In grass carp (Ctenopharyngodon idllus), the serum content of GLU was also significantly increased when 100 g/kg CAP was used to replace soybean meal [60]. In general, a high blood GLU level would lead to a stress response and affect the growth performance of fish [61]. Serum TCHO is an important indicator of body lipid metabolism. The content of TCHO in the FM-0 group was also significantly higher than that in the CON group, which was consistent with the increase in whole-body crude lipid content in this group.Generally, the production and scavenging of active oxides are in dynamic equilibrium, and excessive free radicals lead to lipid peroxidation [62]. MDA is the toxic product of lipid peroxidation, and it can adversely affect health. SOD is the primary substance for organisms to scavenge oxygen free radicals, and CAT scavenge hydrogen peroxide to protect cells from toxicity, while T-AOC is usually used to evaluate the antioxidant capacity of animals [49,50,63,64]. In the present study, the activities of SOD and T-AOC in the FM-5 and FM-0 groups were all significantly lower, while the MDA content was significantly higher than those in the CON group, indicating that the high substitution of fishmeal by CPC and CAP mixture decreased the antioxidant capacity and broke the dynamic balance of free radicals. It has been reported that the high substitution of fishmeal (36%, 57.1%) with either CPC or CAP significantly increases the serum content of MDA in hybrid grouper [57] and largemouth bass [15]. Dietary supplementation of taurine alleviated the oxidative damage, such as decreasing the content of MDA [65], which has been confirmed in rice field eel (Monopterus albus) [66] and spotted knifejaw (Oplegnathus punctatus) [67]. In addition, the free gossypol in CPC may adversely affect the antioxidant system, although the free gossypol level in CPC is much lower than that in cottonseed meal. Gossypol can easily bind to proteins in the electron transfer chain of mitochondria, interfere with the function of mitochondria, and lead to excessive production of reactive oxygen species, thereby inhibiting the activity of various enzymes and causing oxidative damage [68]. 5. ConclusionsIn the present study, the mixture inclusion of CPC and CAP successfully decreased dietary fishmeal from 200 g/kg to 100 g/kg without adverse effects on the growth performance, nutrient utilization, serum biochemical, or intestinal and hepatopancreas histology of rainbow trout.

animals : an open access journal from mdpi

10.3390/ani13071221

PMC10093325

Research on the microorganisms in the reproductive tract of cows has become increasingly popular. Reproductive pathogens, including bacteria, caused uterine disease and decrease fertility. Using sequencing techniques endometrial microbiomes in healthy animals and those with metritis were compared. Our study has identified uterine microbiome profiles that are positively and negatively associated with uterine health. Since it is important to know which bacteria live in healthy or diseased animals, this information will enable the development of treatment options for cows that not only reduce antibiotic use but improve fertility. An improved understanding of changes to the bacteria communities will help to identify animals that can successfully become pregnant again after calving.

The bovine genital tract harbors a dynamic microbiome. Genital tract microbial communities in healthy animals have been characterized using next-generation sequencing methods showing that microbe compositions differ between the vagina and uterus, more so during the postpartum period. Pre-calving fecal and vaginal, and endometrial swabs at the different postpartum intervals were collected from dairy cows. Microbiomes in these samples were determined based on bacterial 16S amplicon sequencing and compared between healthy (H; n = 10) control animals and cows that developed metritis (M; n = 10) within 21 days postpartum (DPP). Compared to healthy animals the pre-calving fecal and vaginal microbiomes of metritis animals were more abundant in sequences from the phylum Fusobacteria and the bacterial genera such as Escherichia-Shigella and Histophilus. In addition, compared to healthy animals, metritis cows harboured low microbial species diversity in the endometrium, as well as decreasing Proteobacteria and increasing Fusobacteria, Firmicutes, Actinobacteria, and Bacteroidetes abundances. The greatest taxonomic compositional deviations in endometrial microbial communities between the metritis and health cows were detected between 7 and 10 DPP. There was high taxonomic similarity detected between postpartum endometrial microbiomes and the prepartum vaginal and fecal microbiomes suggesting that colonization through bacteria ascending from the rectum and vagina to the uterine cavity might play a major role in establishing the endometrial microbiome postpartum. A deeper understanding of the establishment and dynamics of postpartum endometrial microbial communities in cows will thus provide crucial basic knowledge to guide the development of genital microbiome manipulation strategies for preventing uterine disease and improving fertility in dairy cows.

1. IntroductionHigh reproductive efficiency in the dairy cow requires a disease-free postpartum period. Endometrial and vaginal microbiomes are critical in the study of endometritis, which is an important cause of infertility in cattle. Following calving, a common characteristic of uterine involution is microbial influx into the endometrium. Thus, in addition to the increasing energy demands for lactation, a dampened immune status may lead to uterine infection in approximately 40% of the animals postpartum [1]. The resultant uterine diseases perturb fertility [2,3,4]. Cows with metritis suffer from infection and inflammation of the uterus while endometritis is an inflammation of the superficial layer of the uterus. Clinically, a distinction is made between puerperal metritis, which usually occurs in the first 10 days after birth and is associated with general clinical symptoms and fetid discharge, and clinical endometritis, which is characterized by (muco)purulent vulval discharges [1,5]. There exists a relationship between scoring uterine secretions and the abundance of microbes present in the uterus identified using culture-dependent methods [6,7,8]. However, given that bovids possess a complex yet dynamic microbiome, only a small proportion of the bacteria present in the uterus are detected using this approach [9]. Endometrial and vaginal microbiomes are critical in the study of endometritis, which is an important cause of infertility in dairy cattle. Thus, culture-independent approaches to unravel endometrial microbiomes are warranted [2]. Previously, metagenomic sequencing of the 16S rRNA genes in cattle with or without metritis, described 28 different phyla, including Bacteroidetes, Proteobacteria, Fusobacteria, and Firmicutes [10]. Compared to healthy animals, cows with metritis harbored more abundant Bacteroidetes. At the genus level, there were 824 genera detected, with Fusobacterium, Bacteroides, Coxiella, and Porphyromonas being the most abundant [10]. The genera Fusobacterium and Bacteroides are highly abundant in animals with copious vulval discharges and severe uterine disease. Similar bacterial communities were found in the vagina and postpartum uterine secretions; however, it was not possible to predict the occurrence of metritis from the vaginal microbiome [11]. In other studies, quantitative qPCR did not show a link between a diagnosis of metritis and the detection of Escherichia coli and Truepurella pyogenes [12]. This is contrary to other studies where these pathogens were frequently isolated from cattle with metritis [6,13]. The origins of the various pathogens are not yet fully known. Although the uterus is not free of bacteria even during pregnancy [14,15], it is thought that endometrial microbiota originates from the gut, the vagina, and the environment [1]. This study aimed to interrogate vaginal, endometrial, and fecal microbiomes in dairy animals and determine whether these profiles predict the risk of uterine disease postpartum.2. Materials and Methods2.1. Examined AnimalsOn a dairy farm with 100 Holstein Frisian cows, a total of 39 animals that calved were examined over a 9-month period. All cows calved spontaneously or with minor traction support from a maximum of one person. The animals were clinically examined on days 1, 4, 7, 10, and 21 DPP. There were no placenta retentions recorded. All animals were housed on hay bedding and fed Total Mixed Ration (TMR) comprising maize, grass silage, hay, and concentrate. The animals were clinically examined and their general condition, rectal temperature, heart and respiratory rate, mucous membranes, and gastrointestinal tract were assessed prior to sampling. Then, palpation per rectum and vaginal exploration were performed. Uterine size, the opening of the cervix, and cervical secretion were assessed and scored as previously described [16]. Briefly, vulval discharges were scored on a scale from 1 to 5, where: 1 is clear or translucent mucus; 2 is mucus containing flecks of yellowish white coloured pus; 3 is discharge containing ≤50% mucopurulent material; 4 is discharge containing >50% purulent material; and 5 is foul-smelling watery, reddish, or brownish discharge [17]. Animals showing signs of poor general health status and fever (>39.5 °C) in addition to an enlarged uterus and watery, foul-smelling discharge within 10 days postpartum (DPP) were deemed puerperal metritis animals [18,19]. Animals with a distended uterus, and purulent vulval secretions, but no signs of clinical illness 21 DPP were classified as clinical metritis cows. After classification, animals that received antibiotic treatment were excluded from the study (n = 19). The 20 animals selected for microbial analysis, were grouped into metritis (n = 10) and healthy control animals (n = 10).2.2. Sample CollectionSeven to 0 days prepartum, swabs of vaginal secretions were obtained using a double-guarded sampling system (uterus culture swab with introduction pipette, Minitube, Tiefenbach, Germany) as previously described [20]. Sampling was repeated in animals that did not calve 7 days after initial sampling. Briefly, a fecal swab was obtained by swabbing feces collected per rectum (sterile swab BD CultureSwab, Becton Dickinson Allschwil, Switzerland). Then, endometrial swabs were obtained 1, 4, 7, 10, and 21 DPP using double-guarded swabs to prevent contamination. Briefly, the guarded swab was introduced into the vagina and then guided through the cervix by transrectal manipulation [20]. Swabs were obtained from the uterus (uterine body). Within the uterine body lumen, the inner brush was pushed through the outer guard and rotated 3–5 times against the mucosa. The swab was then retracted into the inner cover and withdrawn. The swabs were then capped and placed on ice before transportation to the lab. At the lab, the swabs were stored frozen at −20 °C.2.3. Sequencing of MicrobiomesGenomic DNA was extracted from the swabs using QIAamp DNA (Qiagen, Hilden, Germany) (vaginal, endometrial) and QIAamp DNA Stool (fecal) Mini Kits, respectively, according to the manufacturer’s instructions. Quantus Fluorometer and the QuantiFlour® Double DNA (dsDNA) system (Promega, Madison, WI, USA) kit was used to determine DNA yields from the samples. Bacterial 16S rRNA V4 regions in the DNA samples were amplified using the 515F and 806R primer combination [21] and subjected to Illumina sequencing at StarSEQ (Mainz, Germany, https://www.starseq.com/ (accessed 30 March 2023)). Sequence data (deposited under Bioproject number PRJNA914879) was processed and analysed in R (R-project-org) using the phyloseq package version 1.40.0 [22]. Read quality filtering, error correction, dereplication, merging, and construction of an operational taxonomical unit (OTU) table was performed in the Dada2 package version 1.24.0 using the settings recommended by the package authors [23]. The 16S rRNA amplicon sequences were annotated using the Silva SSU database version 138.1 and assigned to OTUs with a 97% similarity cut-off. Diversities were calculated using the vegan package 2.6-2 [24], and phylogenetic relations were calculated using the ape package [25]. The alpha diversity of the samples was estimated using the observed OTUs and Shannon diversity metrics, and the statistical significance of the differences was assessed using Kruskall-Wallis tests. Beta diversity was estimated using the Bray-Curtis dissimilarity index. Permutational multivariate analysis of variance (PERMANOVA) tests were used to test for differences in microbial beta diversity between the helathy (n = 10) and metritis (n = 10) sample groups. Differential OTU abundances between healthy and metritis animals were calculated using the EdgeR normalization [26]. A cluster dendrogram based on the averaged taxa presence and absence composition for cows in each group was generated using CLC genomics (Qiagen, Prismet, Aarhus, Denmark).2.4. Statistical AnalysisClinical parameters were evaluated using the Mann-Whitney test. Differences in metritis scores were then analyzed using ANOVA for repeated measures and calculated using the Fisher’s post hoc test with a statistical program (StatEL, Paris, France) in Excel (Microsoft, Wallisellen, Switzerland). Significance was set at p-value < 0.05. One-way Analysis of similarities (ANOSIM) was performed between multiple groups based on the Bray-Curtis distance metric with 9999 permutations.Univariate Repeated-measures ANOVA was performed on both H and M groups between all time points. Analyses were performed using R version 4.2.2, utilising ‘vegan’ and ‘heplots’ packages. LDA Effect Size (Lefse) [27] was performed on microbial abundance data for healthy and metritis cattle. Lefse uses a non-parametric factorial Kruskal-Wallis sum-rank test to detect features with significant differential abundance with respect to the class of interest, followed by pair-wise tests of sub-classes using the Wilcoxon rank-sum test.3. Results3.1. Clinical FindingsHealthy animals were on average 5.4 years old and calved after an average gestation length of 283 days. Compared to healthy controls, metritis animals had an average age of 4.6 years and a gestation length of 282 days. Metritis animals showed a significantly higher average vaginal discharge score and rectal temperature compared to healthy animals (p < 0.0001 and p < 0.05, respectively). The course of vaginal discharge scoring over the study period also differed significantly between the two groups (p < 0.0001; Figure 1). Compared to healthy animals, the mean vaginal discharge scores of metritis animals were statistically significantly higher (p < 0.01) on 7, 10, and 21 DPP.3.2. Microbiome Comparison3.2.1. Sequencing Results and Overall Composition of Vaginal, Endometrial, and Fecal MicrobiomesUtilizing bacterial 16S amplicon sequencing the microbiome compositions of prepartum vaginal and fecal as well as postpartum endometrial samples of metritis (n = 10) and healthy (n = 10) animals were determined. Of the 140 samples (70 metritis and 70 healthy) sequenced, 7674 to 285,339 sequence reads per sample were obtained. Post filtering the sequences were assigned to 5200 OTUs that belonged to 45 different phyla and 625 genera. Overall, Proteobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Actinobacteria were the predominant phyla detected among the sequenced fecal, vaginal, and endometrial microbiomes (Supplementary Table S1). In fecal and vaginal microbiomes Bacteroidetes, Firmicutes, and Proteobacteria predominated, whereas the order of the most predominant phyla in uterine microbiomes varied depending on the uterine disease condition and the examined time point postpartum (see Supplementary Table S1). Analysis of similarity (ANOSIM) for comparison of microbial communities in healthy and metritis animals detected significant (p = 0.001; see Supplementary Table S2) differences in microbiome taxonomic composition between the different postpartum time points. This was further interrogated using univariate repeated-measures ANOVA assuming sphericity, which showed there were significant differences in taxonomic composition between healthy (p = 0.0013) (see Supplementary Table S2) and metritis (p = 0.00016) animals (see Supplementary Table S2). To characterise the differences between these groups, Lefse was performed, however, possibly due to natural biological variability between individual animals in each group and the small sample sizes, no statistically significant differences between the groups were discovered.3.2.2. Comparison of Vaginal and Fecal Microbiomes PrepartumExamining samples collected from cows prior to delivery revealed no differences in microbial species richness and alpha diversity between the fecal and vaginal microbiomes of metritis and healthy animals (Figure 2A,B). Using taxonomic composition-based clustering, the fecal microbiomes of the examined animals were less compositionally distinct than vaginal microbiomes within and between the two groups (Figure 2C,D). Meanwhile comparing the overall core taxa distributions between the two groups showed that there were more variably distributed genera between the vaginal than fecal microbiomes of the two animal groups (Supplementary Figure S1). A comparison of the relative abundance of the topmost abundant phyla (n = 5) and genera (n = 10) showed that fecal and vaginal microbiomes are similar taxonomically between metritis and healthy animals. Despite this, the sequences from the phyla Fusobacteria, and genera such as Bacteroides, Staphylococcus, Histophilus and Escherichia-Shigella, were significantly more abundant in vaginal and fecal microbiomes of metritis animals (Figure 2E,F and Supplementary Table S1).3.2.3. Comparison of Postpartum Endometrial MicrobiomesEndometrial microbiomes detected at the beginning (1 DPP, 4 DPP) and end (21 DPP) of the sampling period presented similarly in microbial species richness and alpha diversity levels between metritis and healthy animals (Figure 3A,B). Compared to healthy animals, endometrial microbiomes of metritis cows at 7 and 10 DPP were significantly lower in microbial species richness and alpha diversity (Figure 3A,B). Grouped according to similarities in an averaged group (Figure 3C and Supplementary Figure S2) as well as individual (Figure 4A) cow taxonomic compositions, the 1, 4, and 21 DPP endometrial microbiomes formed no distinctive clusters, whereas the 7 and 10 DPP uterine microbiomes formed distinct clusters according to the uterine disease status. Based on averaged taxonomic composition, both 7 and 10 DPP uterine microbiomes clustered closely within each group but formed separated clusters between the metritis and healthy microbiomes (Figure 3C and Supplementary Figure S2). The 7 and 10 DPP uterine microbiomes showed the highest compositional similarity between individual cows in each group and represented the time points at which the highest compositional shift between metritis and healthy endometrial microbiomes was observed (Figure 4A). Taxonomic composition differences were further analyzed by comparing the total number and distribution of all the bacteria genera present in metritis and healthy endometrial microbiomes at different times postpartum (shown on Venn diagrams, Figure 5). Metritis endometrial microbiomes differed from healthy in both the total number and distribution of the overall bacteria genera found at different times postpartum. The greatest discrepancies in the total number and distribution of genus taxa between the animals were in endometrial microbiomes at 7 and 10 DPP (Supplementary Figure S3). In addition to revealing that there was an overall lower number of genus taxa detected in metritis animals, this comparison also revealed that animals from the two groups had the lowest numbers of commonly shared genera (Supplementary Figure S3). Comparing the five topmost abundant phyla on the other hand showed that endometrial microbiomes of metritis animals declined in Proteobacteria (51 to 6%) while gradually increasing in both Actinobacteria (2 to 24%) and Bacteroidetes (6 to 8%) relative abundances as postpartum time progressed from 1 to 10 DPP (Figure 4B). In healthy animals, on the other hand the Proteobacteria abundance declined modestly (63 to 39%), whereas Bacteroidetes (8 to 4%) abundance decreased, and there was a significantly lower increase of Actinobacteria (4 to 7%) abundance during this period (Figure 4B). The relative abundances of Firmicutes and Fusobacteria similarly fluctuated between 1 and 10 DPP for animals in both groups. Both phyla were however more abundant in metritis than healthy endometrial microbiomes except on 10 DPP, where there were similar Fusobacteria abundances detected for both groups. Taxa relative abundances examination at the genus level showed that the Proteobacteria abundance declined between 1 and 10 DPP within healthy animals largely due to decreasing abundance of an unclassified Proteobacteria genus (OTU1941) (Figure 4C). Increasing Actinobacteria and Bacteroidetes abundances were associated with rising Trueperella, Bacteroides, and Porphyromonas abundances in metritis animals (Figure 4C). Reduced Proteobacteria relative abundances between 1 and 4 DPP in metritis cows were replaced by increased Fusobacteria and Actinobacteria (4 DPP only) levels, whose relative abundances were significantly higher than those in healthy animals. The 7 and 10 DPP uterine microbiomes showed the most similar taxonomic profiles between the different postpartum days in both groups. Relative to healthy animals, the 7 and 10 DPP metritis animals were less abundant in Proteobacteria but higher in Firmicutes, Bacteroidetes, and Actinobacteria abundances at these timepoints (Figure 4B). Trueperella, Bacteroides, Enterococcus, Porphyromonas, Escherichia-Shigella, and Histophilus were among the genera that displayed higher mean relative abundances within metritis endometrial microbiomes at 7 and 10 DPP (Figure 4C and Supplementary Table S1). The unclassified Proteobacteria genus OTU1941, in contrast, dominated in healthy endometrial microbiomes and was significantly more abundant than in metritis animals (Figure 4C and Supplementary Table S1). Compared to healthy cows, the 1 and 4 DPP endometrial microbiomes of metritis animals also displayed a higher abundance of other Proteobacteria genera including Histophilus (Figure 4C). Endometrial microbiomes in metritis animals on the other hand had rebounded and stabilized on 21 DPP as their microbial species diversity (Figure 3A,B) and relative abundances (Figure 4B,C) of predominating phyla were like those in healthy cows. Despite this, 21 DPP microbiomes between cows of the two groups still maintained some differences in abundance and distribution of various bacterial phyla and genera (Figure 4B,C, Supplementary Table S1 and Supplementary Figure S3). In addition to variable distribution for several of the less abundant genera, the 21 DPP metritis endometrial microbiomes, compared to healthy animals, were more abundant in several shared genera including Trueperella, Bacteroides, Porphyromonas, Prevotella, and Histophilus.3.2.4. Distribution of Bacterial Taxa between Vaginal, Endometrial, and Fecal MicrobiomesThe impacts of metritis on endometrial microbiome composition were also probed by comparing the distribution of bacterial genera between metritis and healthy core endometrial microbiomes postpartum. These were comprised of bacterial genera detected among cows of each group at different time points during the postpartum. period Based on these criteria, core postpartum endometrial microbiomes of 45 and 78 genera, respectively, were identified in metritis and healthy endometrial microbiomes (Figure 5A,B and Supplementary Table S3). Nine and 42 genera were found unique to the core endometrial microbiomes of postpartum metritis and healthy animals, respectively (Figure 5C and Supplementary Table S4). Genera such as Fusobacterium, Mycoplasma, and Enterococcus were unique to the metritis core, whereas Prevotella and Lactobacillus were among the genera unique to the healthy core endometrial community. As much as 76% (34/45) of the metritis and 68% (53/78) of the healthy postpartum core endometrial microbial communities were common with their corresponding core prepartum fecal and vaginal microbiomes (Figure 6A,B). Eight of the genera including Enterococcus and Escherichia-Shigella were only shared between core vaginal, endometrial and fecal microbiomes in metritis animals. There were 10 and 22 genera, respectively, within metritis and healthy core endometrial microbiomes postpartum that were only shared with the corresponding prepartum core vaginal microbiomes (Figure 6A,B). These genera included 5 and 17 genera, respectively, that were exclusive to metritis and healthy animals (Figure 6D; Supplementary Table S6).4. DiscussionThe natural genital microbiomes are beneficial to the host through different mechanisms including production of protective biofilms as well as other host defense molecules. Both endometrial and vaginal microbiomes are critical in the development of endometritis, which is an important cause of infertility in dairy cattle. Bacterial 16S rRNA sequencing previously showed that the normal vaginal microbiome comprises bacterial phyla Bacteroidetes, Fusobacteria, and Proteobacteria [2]. In addition, it is known that postpartum endometrial microbiomes in cows have important roles in the modulation of uterine health and the development of metritis [2]. Thus, metritis in cows is commonly associated with a dysbiosis that causes less complexity and altered taxonomic composition of endometrial microbial communities [2,10,17,28,29]. Compared to healthy animals, metritis cows harbour notable endometrial taxonomic compositional changes of their microbial communities showing more Fusobacteria, Bacteroidetes, and Actinobacteria, but less Proteobacteria abundances [2,10,11,15,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38]. Fusobacterium, Trueperella, Bacteroides, Escherichia-Shigella, Histophilus, and Porphyromonas are among bacterial genera that are reported to increase their abundance in metritis compared to healthy cows [2,32,34]. Notably, these genera include species of specific endometritis pathogens such as Escherichia coli, Trueperella pyogenes, Fusobacterium necrophorum, Prevotella melaninogenica, and Porphyromonas levii¸ which have been isolated using culture-dependent approaches from metritis infections [7,29,39,40].Bacteria deriving from fecal and vaginal microbiomes contribute to the post-calving uterine microbial communities in cows [2,9,41,42]. Bacteroidetes, Fusobacteria, Firmicutes, Proteobacteria, Actinobacteria, and Tenericutes sequences were previously found to predominate among the fecal and vaginal microbiomes detected in cows [9,41,42,43,44,45]. Similar composition of predominating bacterial phyla was determined in the fecal and vaginal microbiomes found prior to calving in metritis and healthy cows examined in this study. Previously, pre-calving vaginal microbiomes in cows diagnosed with metritis were reported to contain higher Proteobacteria loads than those of healthy cows [33]. In the current study, compared to healthy animals, metritis cows also displayed some taxonomic compositional differences. In particular, we found that fecal and vaginal microbial communities in cows that developed metritis were higher in an abundance of some of the bacterial taxa that are known to include metritis infection-associated pathogenic bacterial species such as the phyla Fusobacteria and the genera Escherichia-Shigella [2,6,13]. Metritis cows further differed from healthy cows in the occurrence and distribution of various other less abundant bacteria taxa. Taken together, some of these microbial community differences may have contributed to postpartum endometrial microbiomes in ways that impacted uterine health or the development of metritis postpartum.Overall, our analysis showed that the endometrial microbial communities are similarly dynamic within both animal groups. In line with previous observations, we also found that endometrial microbial communities in metritis cows shifted their structure and composition when compared to healthy animals. Lower Proteobacteria and higher Fusobacteria, Bacteroidetes, and Actinobacteria abundances in uterine microbial communities have been associated with metritis development whereas increased Proteobacteria and Tenericutes abundances have been associated with a healthy uterus [2]. In our study, the metritis cows similarly showed an overall gradual depletion of Proteobacteria, which was replaced through increasing Fusobacteria, Firmicutes, Bacteroidetes, and Actinobacteria populations between 1 and 10 DPP. Although Fusobacteria and Firmicutes’ relative abundances similarly fluctuated within both groups both phyla in most cases were more abundant within metritis animals. Differences of varying magnitude in endometrial microbial community composition between metritis and healthy cows were associated with changes in distribution and relative abundance of different bacteria taxa detected. Statistically significant changes with respect to species richness and alpha diversity of uterine microbial communities were however only apparent at 7 and 10 DPP between cows from the two groups. Endometrial microbial populations on 7 and 10 DPP were most distinct between healthy and metritis animals. During this period, the increasing divergence in the endometritis score observed between the metritis and healthy animals is in line with many incidences of metritis detected at 5 and 7 DPP as previously reported [46].5. ConclusionsA deeper understanding of endometrial microbiota structure and compositional dynamics will pave the road for a new era in research for diagnosing uterine disease, improving fertility, and enhancing the productivity of dairy animals. Our results suggest that microbial communities found in feces and the vagina prepartum may play a key role in composing the endometrial microbiome postpartum. Thus, an improved understanding of the relationship between pre-calving fecal and vaginal microbiomes and postpartum endometrial microbial communities would open possibilities of predicting uterine disease in cows based on pre-calving fecal and vaginal microbiomes. Harnessing this knowledge may revolutionize monitoring and treating uterine disease in dairy animals. More research is needed in order to develop robust and comprehensive protocols incorporating pre- and probiotic products that support the proliferation of ‘good’ endometrial microbiomes.

animals : an open access journal from mdpi

10.3390/ani11061518

PMC8224646

Feed efficiency is becoming an important selection tool in the beef cattle industry. Traditionally, feed efficiency of beef cattle has been expressed as the ratio of feed intake to body weight gained; however, selection for high growth rates inevitably increases the maintenance requirements, feed requirements, and intake of cattle, with subsequent higher feed costs. In contrast, net feed efficiency, or residual feed intake (RFI), is defined as the difference between an animal’s actual feed intake and its expected feed requirements for maintenance and growth, with low-RFI animals being more efficient at converting forage intake into kilograms of production than high-RFI animals. This study evaluated the impacts of cow age and RFI on body weight (BW) and body condition score (BCS) change, supplement intake, grazing behavior, and resource use of grazing beef cattle grazing mixed-grass rangelands. Heifer post-weaning RFI had little effect on subsequent performance (BW or BCS), grazing behavior, supplement intake behavior, or resource use. However, cow age significantly influenced subsequent performance, grazing behavior, supplement intake behavior, and resource use. In summary, post-weaning RFI had minimal effects on beef cattle performance, grazing behavior, or resource utilization; however, cow age impacted both grazing behavior and resource use.

The objectives of this study were to evaluate the influence of RFI and cow age on the supplement intake and grazing behavior of beef cattle. Average daily supplement intake (kg/cow/d) displayed an RFI × cow age interaction (p < 0.01), with a linear increase in average daily supplement intake with increasing RFI of 3-year-old cows (p < 0.01). Average daily supplement intake (g ∙ kg BW−1 ∙ d−1) displayed an RFI × cow age interaction (p < 0.01), with a quadratic effect on supplement intake of 3-year-old cows (p = 0.01). Cow age displayed a quadratic effect on variation of supplement intake (p < 0.01), where 1-year-old cows had a greater CV of supplement intake than all other cow ages (p < 0.01). Distance traveled displayed a cow age × RFI interaction (p = 0.02), where high-RFI 5-year-old cows traveled further per day than low 5-year-old RFI cows. The probability of grazing site selection was influenced by cow age (p ≤ 0.03). In summary, heifer post-weaning RFI had minimal effects on beef cattle performance, grazing behavior, or resource utilization; however, cow age impacted both grazing behavior and resource use.

1. IntroductionThe greatest operating cost for commercial cow-calf producers is providing adequate nutrition for animals is where supplemental feed can account for 65% of the annual expenses [1,2,3]. Selection pressure for feed efficient beef cattle that have lower feed intake while maintaining production could have a significant impact on cow-calf profitability [3]. It has been reported that roughly two-thirds of mature cow energy requirements are utilized for maintenance [4,5,6]. However, substantial animal-to-animal variation, independent of body size and growth, exists in maintenance requirements of cattle [7,8,9]. Consideration of the lower maintenance requirements for low-residual-feed-intake (RFI) cattle becomes much more important as cattle move into times of negative energy balance, such as dormant, late-season grazing [10]. Therefore, improving feed efficiency through genetic selection holds significant opportunity for the beef industry.Currently, RFI is being used as a selection tool for purchasing bulls and/or retention of heifers. Research evaluating the efficacy of using post-weaning RFI values as selection criteria for beef cattle that fit Western rangeland systems is currently lacking. Most RFI studies utilize energy-dense diets and rations focusing on feedlot performance [11]. Research pertaining to RFI in cattle offered forage-based diets is limited [12], with even fewer data available related to beef cattle forage-based production systems [3,10,13]. As a result, more research is needed to evaluate the utility of RFI estimates on the selection of heifers for extensive forage-based systems [10,14,15].Western US cow-calf production systems rely heavily on rangeland forages to supply nutrients for both cows and calves [16]. The primary goal in a forage-based livestock production system is to obtain optimal animal performance while effectively utilizing the forage resource base [17]. Seasonal nutrient deficiencies associated with dormant rangeland forages often require protein supplementation to maintain animal performance, production, and provide increased economic returns [18,19]. However, the reported effectiveness of protein supplementation programs on grazing beef cattle performance has been inconsistent, likely due to variation in animal-to-animal protein supplement intake behavior [18,19,20]. Recent research has demonstrated that cow performance is related to supplement intake behavior [21], therefore, RFI, a potential proxy for cow efficiency may also be linked to protein supplement intake behavior. However, information relating cow RFI to protein supplement intake behavior does not currently exist.Although a central aspect of domestic livestock ecosystems, the spatial component of livestock grazing has remained difficult to interpret [22]. Mechanisms that influence grazing distribution of grazing cattle can be characterized as follows: exogenous, the physical environment (e.g., topography [23,24]), or endogenous (e.g., age and experience [23,25,26,27]). Thus, cattle grazing the same pasture can have different grazing distribution patterns [28]. Recent research has demonstrated that low-RFI cows (more efficient) have larger distribution patterns and outperform high-RFI (less-efficient) cows [10,29]. Therefore, it is possible that grazing behavior may vary with cow RFI (efficiency) and cow age while grazing nutrient deficient forages. However, relationships between exogenous factors with endogenous attributes on grazing behavior are less understood [26]. Therefore, the objectives of this study were to evaluate the influence of cow RFI and age on (1) beef cattle performance; (2) supplement intake behavior; and (3) grazing behavior, resource use, and distribution patterns on winter mixed-grass prairie rangelands. We hypothesized that cow RFI and age interact to influence animal performance, supplement intake, and grazing behavior.2. Materials and MethodsThe use of animals in this study was approved by the Agricultural Animal Care and Use Committee of Montana State University AACUC #2018-AA12. All animals used in this study were provided by the Montana Agricultural Experiment Stations, and the study was conducted at the Northern Agricultural Research Center in Havre, Montana.Two consecutive years of winter grazing studies were conducted on non-lactating, pregnant commercial Angus cows to evaluate the influence of RFI and cow age on beef cattle performance, supplement intake behavior, grazing behavior as well as distribution and resource use patterns. This study was conducted at the Montana State University Northern Agriculture Research Center’s Thackeray Ranch (48°21″ N 109°30′ W), located 21 km south of Havre, MT, USA. Bull Hook Creek, a perennial stream, transects the pasture and is used for livestock watering. Vegetation is dominated by Kentucky bluegrass (Poa pratensis L.), bluebunch wheatgrass (Pseudoroegneria spicata [Pursh] A. Love), and rough fescue (Festuca scabrella Torr.; Reference [20]). Available forage biomass of the study area was estimated by clipping ten randomly located plots to 2 cm in height, immediately prior to grazing, using a 0.25 m2 plot frame. Samples were placed in a forced-air oven at 55 °C for 72 h and then weighed and recorded to calculate dry matter production (kg ∙ ha−1; Table 1). Vegetation samples were weighed individually and composited by year, ground to pass a 1 mm screen in a Wiley mill, and sent to a Dairy One for nutrient analysis (Dairy One, Ithaca, New York).2.1. Animal Performance and Supplement Intake BehaviorA commercial herd of bred Angus cows (205 in year 1 and 203 in year 2) ranging in age from 1 to 10 years old were classified into one of 6 age groups (1, 2, 3, 4, 5 to 7, and ≥ 8 years old) and grazed on rangeland pastures (~1.5 ha ∙ animal unit month−1) from mid-October to early-January each year. Individual cow body weight (BW) and body condition scores (BCS) were obtained following a 16-h shrink pre- and post-grazing (Table 2). Body condition scores were based on a 1–9 scale (BCS: 1 = emaciated, 9 = obese, [30]) and assessed by two trained individuals, then averaged for a final BCS. All heifers, (9–11 months of age), went through a 70-day post-weaning RFI trial [7,29] using a GrowSafe system (GrowSafe DAQ 4000E; GrowSafe System Ltd., Airdrie, AB, Canada) and were classified as either low (<−0.50 SD from mean), or high (>+0.50 SD from the mean) RFI within their contemporary age group [31].A free-choice, self-fed, canola based pressed block supplement (28.7% crude protein (CP; year 1) and 30% CP (year 2)) was provided ad-libitum to cattle. The target daily-recommended intake range was 0.45 to 0.91 kg ∙ cow−1 ∙ d−1 with 23% salt, texture and bitterness to limit daily intake. Individual animals were equipped with an electronic identification tag attached to the left ear for the measurement of daily individual supplement intake (kg ∙ cow−1 ∙ d−1; g ∙ kg BW−1 ∙ d−1), and time spent at supplement feeders (min∙ d−1) using centrally located SmartFeed Pro self-feeder systems (C-Lock Inc., Rapid City, South Dakota) with a total of 2 trailers and 8 feeding stations (4 per trailer). Variation in supplement intake, measured as the coefficient of variation (% CV), was based on daily intake estimates for individual animals.2.2. Grazing Behavior and Resource UseEach year, a subset of 30 cows were randomly selected within age (2, 5, and 8 years old) and RFI (Low, High) and fitted with Lotek GPS collars containing an activity sensor (Lotek Engineering, Newmarket, ON, Canada; 5 collars per RFI × cow age combinations) representing a minimum of 39% of the total cattle population within each RFI × age combination each year. GPS collars were configured to record positions at 15-min intervals, and activity sensor measurements at 5-min intervals to determine distance traveled, location, and duration of grazing activities, as well as resource use [32,33,34]. Due to limited battery life of GPS collars, supplement intake, grazing behavior and resource use was only measured during the last 45 days of grazing each year. Individual cow was considered the experimental unit to evaluate effects of age and RFI on supplement intake behavior. Grazing activities were derived by the binary classification methods developed by Augustine and Derner [35] to evaluate time spent grazing and foraging distribution. Observations were limited to grazing to determine critical foraging areas rather than general occupancy [17,26]. Pasture supplement and water locations were recorded by using a handheld GPS unit (spatial error < 10-m). Spatial layers including aspect, terrain ruggedness (sum change in elevation between a grid cell and its eight neighboring cells; Reference [36]) and distance from supplement and water locations were developed by using the spatial analysis tool in ArcGIS (Environmental Systems Research Institute, Redlands, CA, USA) and a digital elevation model at a 30 m2 resolution.2.3. Statistical AnalysisThe influence of RFI and cow age on cow BCS, BW (Supplementary Materials Table S1), supplement intake behavior (Supplementary Materials Table S2), time spent grazing, and distance traveled (Supplementary Materials Table S3) were analyzed by using a generalized linear mixed model, with a Gaussian error structure, in an ANOVA framework that included RFI, age, and the interactions of RFI and age as fixed effects, and individual cow and year as random effects. An alpha ≤ 0.05 was considered significant with animal considered as the experimental unit. Linear and quadratic effects were determined, using orthogonal polynomial contrasts for each analysis. The Tukey method was used to separate means when p < 0.05. Tendencies were considered when p ≤ 0.10.To model relative resource selection for cattle grazing late season dormant forages, individual GPS-collared cows were defined as the biological unit of interest. To evaluate the response of an individual cow’s space use to pasture level covariates, we used multiple regression in a resource utilization function (RUF) analysis with the ruf.fit package in R [37,38,39]. Resource utilization functions reduce error associated with location estimation and increase sensitivity for detecting resource selection by evaluating within animal variation in resource use and incorporating an individual cow’s entire grazing distribution, independently, while accounting for spatial autocorrelation [37,38,40].Due to cattle grazing distribution being defined by pasture boundary, GPS grazing locations were used to build RUFs to quantify cow selection of environmental covariates within pasture (third-order scale; [41]). Specific utilization density rasters for grazing locations were created for each individual at a 30 m2 resolution, using the adehabitatHR and raster packages in R [42,43]. Relative use values were bound between 1 and 99 for each 30 m2 cell based off of the relative volume of use within the cell compared to all other cells in the pasture [37]. Pasture level spatial covariates anticipated to effect resource utilization included distance to supplement and water, elevation, terrain ruggedness and aspect. Individual relative use and pasture level covariates rasters were stacked and converted to data files, using the raster function in R as input for the ruf.fit package (see Supplementary materials Table S4) [39,40]. To meet the assumptions of multiple regression models, individual relative use values were log-transformed. Standardized β coefficients were developed and evaluated for each cow to determine the influence of the pasture level covariates on cattle resource utilization [37,39,40].Significant predictors of resource use were determined by standardized coefficients with 95% confidence intervals not overlapping zero [37,38]. Significant resource utilization coefficients were determined to be greater or less than expected based on availability of the covariate within the pasture [35,36]. For pasture level covariates displaying high herd-level variability in grazing resource utilization (herd-level SE of standardized coefficients > 0.25; [44]), a post hoc analysis was conducted to evaluate the effects of RFI, cow age, and the interaction of RFI and cow age on resource use coefficients relative to each pasture covariate, using ANOVA with a generalized linear mixed model, with a Gaussian error structure, including year as a random intercept. An alpha ≤ 0.05 was considered significant with animal considered as the experimental unit. For age main effects, linear and quadratic effects were determined by using orthogonal polynomial contrasts for each analysis. The Tukey method was used to separate means when p < 0.05. Tendencies were considered when p ≤ 0.10.Core grazing areas for individual cows were estimated by using the kernel utilization distribution function in the adehabitatHR package in R [42] (see Figure 1). Kernel utilization distributions are three-dimensional representations of estimated distribution of use, used to calculate home range [45]. The estimated values of use for the kernel utilization distributions containing only grazing locations were then used to create a contour representing 50% of the volume of use delineating core grazing areas [46,47,48]. The area of the 50% contours were calculated by using the gArea function in the rgeos package of R (see Supplementary Materials Table S5) [49]. Due to pasture management unit defining the extent of core grazing areas of cattle, contours representing core grazing areas were bound by pasture boundary. The influence of RFI and cow age on core grazing area was analyzed by using a generalized linear mixed model, with a Gaussian error structure, in an ANOVA framework including RFI, age, and the interactions of RFI and age as fixed effects, and year as a random intercept. An alpha ≤ 0.05 was considered significant with animal considered as the experimental unit. For age main effects, linear and quadratic effects were determined by using orthogonal polynomial contrasts for each analysis. The Tukey method was used to separate means when p < 0.05. Tendencies were considered when p ≤ 0.10. All statistical analyses were performed in R [50].3. Results3.1. Supplement Intake Behavior and Animal PerformanceBody weight change over the 84-day grazing period exhibited a tendency for an RFI × cow age interaction (p = 0.08). Specifically, 4-year-old cows BW change decreased linearly with increasing RFI (p < 0.01), averaging 22.69 ± 23.66, 2.59 ± 23.31, and −12.49 ± 23.60 kg weight change for low, average, and high RFI, respectively. No RFI differences (p > 0.44) were observed for change of cow BW within all other cow ages. Cow age influenced (p = 0.02) weight change, over the 84-day grazing period; however, no linear or quadratic responses were noted (p ≥ 0.19) with weight changes ranging from −5.01 ± 22.87 kg (yearlings) to 8.72 ± 22.92 kg (3-year-old cows) with all other cow age being intermediate. Change in BCS exhibited a tendency (p = 0.07) for a quadratic effect of cow age (p = 0.01), with 1-year-old cows losing BCS (−0.10 ± 0.10 units) compared to 4-year-old cows gaining BCS (0.15 ± 0.10 units) with all other cow age being intermediate over the 84-day grazing period.Average daily supplement intake (kg ∙ cow−1 ∙ d−1) displayed an RFI × cow age interaction (p < 0.01; Figure 2), with RFI exhibiting a linear effect on 3-year-old cows (p < 0.01) where supplement intake increased with increasing RFI (Figure 2C). However, RFI had no effect on supplement intake for other cow age (p > 0.28). Supplement intake expressed as g ∙ kg−1 BW−1 ∙ d−1 displayed an RFI × cow age interaction (p < 0.01). Specifically, a quadratic effect of RFI was observed on supplement intake of 3-year-old cows (p = 0.01; Figure 2I), where high-RFI cattle consumed more supplement than low- and average-RFI cattle (p < 0.01). However, RFI had no effect on supplement intake for other cow ages (p > 0.23). Likewise, RFI effects were not observed for variation of supplement intake (% CV; p > 0.69). However, there was a quadratic effect of cow age on variation in supplement intake (p < 0.01; Figure 3), where 1-year-old cows had a larger CV of supplement intake than all other ages (p < 0.01).There was no effect (p > 0.13) of RFI or cow age observed on supplement intake rate, with supplement intake rate averaging 60.6 ± 32.0 g ∙ min−1. Time spent at the feeders was quadratically influenced (p < 0.01) by cow age, where 2-, 3-, and 4-year-old cows spent more time at the feeders than 1-, 5–7-, and ≥8-year-old cows (p < 0.02; Figure 4). However, time spent at the feeders exhibited a tendency for an RFI × cow age interaction (p = 0.08), with a quadratic effect (p = 0.03) of RFI for 5–7-year-old cows, where average-RFI cows tended to spend more time at the feeders than high-RFI cows (p = 0.08).3.2. Grazing Behavior and Resource UseDistance traveled displayed a cow age × RFI interaction (p = 0.02), where high-RFI 5-year-old cows traveled further per day than low-RFI 5-year-old cows (p < 0.03), and low-RFI 8-year-old cows tended (p = 0.08; Figure 5) to travel further than high-RFI 8-year-old cows. No RFI effects (p < 0.71) were observed on distance traveled for 2-year-old cows. Distance traveled ranged from 3.00 ± 0.09 for 2-year-old cows to 2.49 ± 0.09 km ∙ d−1 for 8-year-old cows. Time spent grazing was not influenced by cow age (p = 0.29) nor RFI (p = 0.40), averaging 6.06 ± 0.44 h ∙ d−1.Herd-level grazing resource utilization for cattle on dormant rangeland forage was not impacted by aspect (β¯^ North = 0.02 ± 0.01; β¯^ South = −0.01 ± 0.01; β¯^ East = −0.02 ± 0.02; β¯^ West = −0.01 ± 0.02), elevation (β¯^ = 0.07 ± 0.37), distance to supplement (β¯^ = −0.37 ± 0.51), distance to water (β¯^ = 0.04 ± 0.44), or terrain ruggedness (β¯^ = −0.06 ± 0.03; Figure 6). However, resource utilization relative to elevation, distance to supplement, and distance to water were highly variable among individuals (herd-level SE of standardized coefficients > 0.25; Figure 7). Therefore, we conducted a post hoc analysis evaluating the effects of RFI, cow age, and the interaction of RFI and cow age on grazing resource utilization relative to elevation, distance to supplement, and distance to water.Cow age did not interact with RFI (p > 0.11) when evaluating the probability of grazing site selection relative to elevation, distance from supplement and water. As a result, only RFI and cow age main effects are reported. The probability of grazing site selection relative to elevation was not influenced by RFI (p = 0.62) or cow age (p = 0.16). The probability of grazing site selection relative to distance to supplement was not impacted by RFI (p = 0.68), but was linearly influenced by cow age (β¯^ 2-years old = 0.06 ± 0.08; β¯^ 5-years old = −0.31 ± 0.08; β¯^ 8-years old = −0.90 ± 0.08; p < 0.01) with older cows grazing closer to supplement locations than younger cows (Figure 8A). The probability of grazing site selection relative to distance to water was also not impacted by RFI (p = 0.63); however, it was quadratically influenced by cow age ((β ¯^2-years old = −0.20 ± 0.08; (β ¯^5-years old = −0.10 ± 0.08; (β¯^ 8-years old = 0.44 ± 0.08; p = 0.03) with 8-year-old cows selecting grazing locations further from water than 2- and 5-year-old cows (Figure 8B).Core grazing area (ha) was linearly affected by cow age (p < 0.01), with core grazing area decreasing with increasing cow age (Figure 9). There was no effect of RFI on core grazing area (p = 0.71), but a tendency for a cow age × RFI interaction was observed (p = 0.07). However, there were no differences observed for RFI within cow age (p > 0.10) relative to size of core grazing area.4. DiscussionWhile the estimation of RFI on young bulls and developing heifers is becoming a more popular practice in the US beef industry, the use of RFI of post-weaned heifers as a selection criterion for retention of replacement females for forage-based rangeland environments has not been well studied. Conventional wisdom suggests that cattle with low RFI values are more efficient in converting nutrients to maintenance energy and body weight gain. Few research experiments have investigated the effects of RFI on beef cattle performance, supplement intake and grazing behavior or resource use while grazing late season dormant rangelands. If low-RFI cattle have lower maintenance requirements, selection for low-RFI cattle on low quality late-season mature rangelands could lead to cattle that are more efficient while grazing on low-energy and low-protein diets, assuming that low-RFI cattle have lower maintenance requirements when facing nutritionally stressful periods.Sprinkle and coworkers [10] were the first to examine the effects of RFI on livestock performance while grazing low-quality dormant forages; however, their research focused solely on 2-year-old cows. In contrast, our study focused on the effects of RFI across multiple age groups on supplement intake behavior, grazing behavior, and resource use while grazing low quality dormant mixed-grass rangelands. Most of the research investigating the effects of RFI on cattle performance has occurred on both irrigated and improved dryland summer pastures [11,14,51]. All the above-mentioned research trials can be characterized as having adequate forage quality to meet maintenance requirements with the exception of Sprinkle et al. [10], where their grazing and forage conditions were similar to those experienced during our research trial.Previous research [29] reported that low-RFI cows grazing summer rangelands in central Idaho travel further and graze longer than high-RFI cows at warmer temperatures. Conversely, we observed no effect of RFI on grazing behavior, and resource use during our late fall/early winter grazing trial. However, 5-year-old high-RFI cows did travel further than low-RFI 5-year-old cows, and low-RFI 8-year-old cows tended to travel further than high-RFI 8-year-old cows with no observed differences among 2-year-old cows. Sprinkle et al. [10] reported that, while grazing late season dormant rangelands in Idaho, low-RFI 2-year-old cows lost less weight and body condition compared to high-RFI 2-year-old cows with no difference in daily distance traveled or foraging rate (bites ∙ min−1). Conversely, we observed a decrease in BW change with increasing RFI for 4-year-old cows, with no observed changes in cow BW within all other cow ages. However, we did not observe effects of RFI on BCS change, and the differences we observed in distance traveled and core grazing area were associated with cow age and not RFI. Similarly, Sprinkle et al. [52] also reported no difference in distance traveled or time spent grazing between low-RFI and high-RFI 2-year-old cows grazing supplemented fall pastures in central Idaho. Our results are also consistent with Meyer et al. [3], where RFI did not affect BW and BCS change or supplement intake while grazing during late winter and early spring.Although limited, previous experiments evaluating supplement intake of mixed-age beef herds have reported that younger cows spent less time at the supplement feeders and consumed less supplement than older cows [53,54]. In contrast, Wyffels et al. [44] reported that younger cattle consumed more supplement and visited the supplement feeders more often than older cows. Results from our research are similar to Wyffels et al. [44], since we observed a quadratic effect in supplement intake related to cow age, where 1-year-old cows consumed more and had a larger CV of supplement intake than older cows.Previous research reported that cow age significantly impacted grazing behavior and distribution with older cows grazing further from water and using higher elevations than younger cows [26,55]. Our results agree with these reports, where older cows selected grazing locations further from water than younger cows. Wyffels et al. [55] reported results similar to our study, where older cows selected grazing locations closer to supplement feeders and herd-level resource utilization was negatively related to terrain ruggedness on mixed-grass prairie rangelands. In our study we observed that herd-level resource utilization of cattle grazing dormant mixed-grass prairie rangelands was not impacted by aspect, elevation, or terrain ruggedness.Our research suggests that in dormant forage grazing environments, heifer post-weaning RFI has little effect on supplement intake behavior (with the exception of 3-year-old cows), grazing behavior, or resource use. We observed no RFI impact on BW or BCS change over the 84-day grazing period, with the exception of 4-year-old cows. Our results agree with previous research [56] that suggests post-weaning RFI is independent of mature cow BW and may have little impact on the efficiency of cattle winter grazing on dormant rangelands. However, despite the lack of differences in supplement intake behavior, grazing behavior, and resource use in our study, if cows selected for low RFI have the same performance parameters as high-RFI cows while consuming less feed, selection for low-RFI cattle would still be warranted. As a result, further research is needed to investigate the relationship of heifer post-weaning RFI and dry-matter intake of cows at different ages and stages of production.5. ConclusionsHeifer post-weaning RFI had little effect on subsequent cow performance (BW or BCS), grazing behavior, supplement intake behavior, and resource use. However, cow age significantly influenced cow performance, grazing behavior, supplement intake behavior, and resource use. We also observed high individual variability in grazing site selection, suggesting that individual-level factors may be driving grazing resource use and grazing behavior. Therefore, our research suggests that cow age has more of an impact on resource use, supplement intake, and grazing behavior than heifer post-weaning RFI while grazing dormant-season mixed-grass prairie rangelands.

animals : an open access journal from mdpi

10.3390/ani13081312

PMC10135243

Horses are high-level athletic athletes prone to musculoskeletal injuries. Tendon/ligament injuries are the most frequent types of injuries which that are very difficult to treat. Instead of tissue regeneration, usually, fibrous scar tissue develops which leads to decreased functionality of the injured area and threatens the participation of sport horses. The aim of regenerative medicine is to find a treatment that promotes tissue regeneration and that allows the equine patient to return to the same level of athletic performance in the shortest time period possible. In this study, we developed a solution of equine synovial membrane stem cells and autologous serum, to be injected at the lesion site to promote tissue regeneration. We describe the processes of tissue collection, preparation, isolation of synovial stem cells, expansion, culture, cryopreservation, and posterior preparation with autologous serum. The solution was tested in 16 tendons and ligaments of equines. After treatment, all equine patients underwent a physical rehabilitation program and were monitored with physical and ultrasonographic exams. The results were very promising, and thus, support the use of equine synovial stem cells and autologous serum in the treatment of tendonitis and desmitis.

Tendon and ligament injuries are frequent in sport horses and humans, and such injuries represent a significant therapeutic challenge. Tissue regeneration and function recovery are the paramount goals of tendon and ligament lesion management. Nowadays, several regenerative treatments are being developed, based on the use of stem cell and stem cell-based therapies. In the present study, the preparation of equine synovial membrane mesenchymal stem cells (eSM-MSCs) is described for clinical use, collection, transport, isolation, differentiation, characterization, and application. These cells are fibroblast-like and grow in clusters. They retain osteogenic, chondrogenic, and adipogenic differentiation potential. We present 16 clinical cases of tendonitis and desmitis, treated with allogenic eSM-MSCs and autologous serum, and we also include their evaluation, treatment, and follow-up. The concerns associated with the use of autologous serum as a vehicle are related to a reduced immunogenic response after the administration of this therapeutic combination, as well as the pro-regenerative effects from the growth factors and immunoglobulins that are part of its constitution. Most of the cases (14/16) healed in 30 days and presented good outcomes. Treatment of tendon and ligament lesions with a mixture of eSM-MSCs and autologous serum appears to be a promising clinical option for this category of lesions in equine patients.

1. IntroductionTendonitis and desmitis are defying clinical challenges in equine patients that require long recovery periods, and ineffective tendon repair can affect their sport careers. Tendons operate near their functional limits during maximal exercise, and their ability to adapt to stress and self-repair is limited. A controlled exercise program alone or in combination with a variety of conservative treatments, such as corrective shoeing and nonsteroidal anti-inflammatory drugs (NSAIDs), is still the gold standard therapy for equine tendon disease [1]. Current treatments often do not fully repair or regenerate the injured or affected tendon nor lead to its total functional recovery [1,2].The aim of tendinopathy treatment is to achieve tissue regeneration and return to complete organ function and performance. Recently, tissue engineering approaches have attracted attention for tissue repair. Among the approaches, the use of mesenchymal stem cell-based therapy has increased, since it is a promising approach for tissue repair and regeneration including tendinopathy and desmitis [1,3,4,5,6].Mesenchymal stem cells (MSCs) can be isolated from several tissue sources such as bone marrow, peripheral blood, dental pulp, umbilical cord, and amniotic fluid [7]. MSC characteristics have been defined by the Mesenchymal and Tissue Stem Cell Committee of the International Society for Cellular Therapy (ISCT), and include plastic adherence when maintained in standard culture conditions; expressing clusters of differentiation (CDs), such as CD44, CD90, and CD105; and no expression of major histocompatibility complex (MHC)-class II markers and of hematopoietic-related markers (CD45 and CD34) [8]. Finally, MSCs must be able to differentiate in vitro into, at least, osteoblasts, adipocytes, and chondroblasts, in the presence of adequate differentiation culture media [8].Synovial membrane mesenchymal stem cells (SM-MSCs) were initially isolated, in 2001, by De Bari et al. [9], from human knee joints and showed significant proliferative ability in culture, even after Passage 10 (P10), and multilineage differentiation potential in vitro [9]. These cells represent a good source of MSCs and a promising therapeutic tool mostly for musculoskeletal pathologies [10]. Sakagushi et al. compared the properties of different sources of human stem cells, i.e., bone marrow, synovium, periosteum, skeletal muscle, and adipose tissue, and observed the superiority of synovium as a source for MSCs for treatment of musculoskeletal pathologies as they had more ability to chondrogenesis. Pellets of synovium-derived stem cells were larger and expressed more intense staining for chondrogenic differentiation [11].SM-MSCs have higher chondrogenic capacity than other studied sources of MSCs, such as bone marrow (BM-MSCs) [12,13]. Cartilage pellets from SM-MSCs have been reported to be significantly larger than those from BM-MSCs [12]. SM-MSCs have a higher production of uridine diphosphate glucose dehydrogenase (UDPGD) [13], an enzyme that converts UDP-glucose into UDP-glucuronate, one of the two substrates required by hyaluronan synthase for hyaluronan polymer assembly. In addition, Sox-9, collagen type II (Col-II), and aggrecan, specific markers for chondrogenesis, as well as cartilage-specific molecules such as cartilage oligomeric matrix protein (COMP) have also been found in high amounts in equine synovial fluid-derived MSCs and the extracellular matrix, respectively by reverse transcription polymerase chain reaction (RT-PCR) [13].In a recent study, using a rabbit model, Bami et al. highlighted the superiority of SM-MSCs in terms of chondrogenesis, osteogenesis, myogenesis, and tenogenesis [14]. A study of xenogenic implantation of SM-MSCs in equine articular defects also confirmed better healing of the cartilage of affected knees as well as a higher expression of collagen type II, indicating the presence of hyaline cartilage in the healed defect [15].SM-MSCs have been defined as MSCs due to their phenotypic profile and differentiation potential. Even though there are no specific antibody markers to identify these MSCs, there is general agreement that MSCs should be negative to hematopoietic markers CD34 and CD45 and positive to CD44, CD73, CD90, and CD105 [16]. Mochizuki et al. found that SM-MSCs maintained their proliferative ability, regardless of which region they were collected from in the synovium [17].In 2003, Fickert et al. reported that the markers CD9, CD44, CD54, CD90, and CD166 could be used to identify MSCs isolated from the synovium of human patients with osteoarthritis (OA), and they also confirmed that CD9/CD90/CD166 triple-positive cell subgroups had obvious chondrogenic and osteogenic differentiation abilities [18].Prado et al. confirmed the mesenchymal nature of equine synovial membrane and fluid-derived stem cells through the expression of significant hematopoietic (CD45, CD34, CD117, and CD133), mesenchymal (CD105, CD90), pluripotency (OCT3/4 and NANOG), embryonic (Tra-1-81), inflammatory, and angiogenesis (vascular endothelial growth factor (VEGF-R1) and LY6a) markers [19]. Although the presence of hematopoietic and inflammatory markers was not expected, variations may occur and must be considered to influence acute or chronic stages of osteochondrosis expression and/or inflammatory events [19,20].Nevertheless, the immunophenotype characterization of equine MSCs (eMSCs), as well as in other veterinary species, has not yet been completely established [19]. This is a major challenge, since the expression of certain adult stem cell markers may differ between species. For that reason, it is a need to define a set of CD markers which can be uniformly applied for the identification of eMSCs [8,20].Horses are high performance athletes prone to musculoskeletal diseases, i.e., osteoarticular, as well as tendon/ligament lesions and fractures of various degrees due to sport- and age-related injuries. These pathologies resemble human musculoskeletal conditions, and therefore, horses are a valuable animal model for assessing stem cell and cell-based therapies prior to the translation of results into humans [21]. The use of a therapy that can regenerate these structures and restore their complete functionality instead of ordinary healing is the aim of our study and of equine practitioners throughout the world. Recent studies have suggested that MSCs can self-renew, migrate to injury sites (homing), perform multilineage differentiation, and secrete bioactive factors, thus, increasing proliferation and migration of tendon stem/progenitor cells via paracrine signaling and increasing the regeneration ability of tissues with poor aptitude [1,3,4,5,22,23].In fact, the knowledge of the importance of this paracrine action has opened doors to cell-free therapeutic strategies in regenerative medicine. The soluble factors (cytokines, chemokines, and growth factors) and nonsoluble factors (extracellular vesicles and exosomes) released in the extracellular space by MSCs, commonly known as secretome, have become the focus of novel therapeutic approaches due to their key role in cell-to-cell communication, their active influence on immune modulation, and their pro-regenerative capacity both in vitro and in vivo [23]. Therefore, in this study, secretome was also analyzed with the prospect of being used therapeutically, in the future, in similar clinical cases.In the present study, equines used as show jumping and dressage athletes as well as leisure horses with acute and chronic lesions were treated with intralesional administration of the considered combination, i.e., autologous serum and eSM-MSCs. The treatment consisted of two injections, 15 days apart. Pre- and post-treatment evaluations consisted of clinical, orthopedic, and tendon/ligament ultrasound exams. None of the selected equine patients had previously received any other regenerative treatment.2. Materials and Methods2.1. Study Design and Horse SelectionThis prospective longitudinal study was performed in Portugal between February 2016 and January 2019. Sixteen horses, from 5 to 22 years old with acute and chronic signs of lameness were enrolled in this study (11 males and 5 mares), whose sport activities were distributed over show jumping (14), dressage (1), and leisure (1). The horses were all outpatients from an equine ambulatory clinic. This study included the treatment of 16 tendons, i.e., 14 superficial digital flexor tendons and 2 deep digital flexor tendons, and 4 suspensory ligaments. Lameness was scored based on the American Association of Equine Practitioners (AAEP) scale (Table 1) and confirmed by using a positive regional nerve block. Flexion and pain to pressure tests were also evaluated [24]. 2.2. Inclusion and Exclusion CriteriaIn this study, horses with acute or chronic lameness (Table 2), with diagnosed tendonitis and/or desmitis and with no signs of systemic disease were accepted in the inclusion criteria. Injured horses were treated in acute stages of disease, except for two equine patients (Patients 3 and 6). Patient 3 had an injury the year before this treatment and laser therapy had been performed, without a complete recovery. After that, he had a re-injury and, at this time, this treatment was suggested. Patient 6 was referred by another clinician who tried, unsuccessfully, to treat this patient. Patient 6 was sent to the field for one year and then re-evaluated. At this time, and as its trainer wanted to improve its quality of life, this treatment was proposed by its clinician. The lameness grade of each equine patient is specified in Table 2. Considering the established exclusion criteria, the selected equine patients should not have been under any other medical treatment (including nonsteroidal anti-inflammatory drugs, intra-articular corticosteroids, hyaluronan, glycosaminoglycans, platelet-rich plasma (PRP), and other MSC preparations) for at least 2 months before the allogenic eSM-MSC treatment and did not receive any additional medical treatment (except for that described in the treatment plan) for at least 2 months post the cell-based treatment.2.3. Ethics and RegulationThis study was carried out in accordance with the Organismo Responsável pelo Bem Estar Animal (ORBEA) from ICBAS-UP, project number P289/ORBEA/2018 recommendations and authorization. Treatments were performed with permission and signature of an informed consent from the equine patient’s legal trainer, following a thorough explanation on the procedure itself and possible risks and associated effects, in accordance with national regulations and project approval from the competent authorities. In addition, no animals were euthanized for this study.2.4. Donor Selection and SM CollectionThe eSM-MSC donor was a young and healthy foal, 7 months old, who died accidentally when running in the arena. The trainer authorized synovial membrane collection from the hocks, knees, and fetlocks. The synovial membrane was evaluated and its appearance was transparent, bright, and smooth; in addition, the presence of viscous and transparent synovial fluid confirmed its soundness. The skin covering the incisional field was surgically cleaned with chlorohexidine and alcohol. The skin and subcutaneous tissue were incised, debrided, the articular capsule was opened, and the synovial membrane was isolated and extracted into a Dulbecco′s phosphate buffered saline (DPBS) container. The samples were transported to the laboratory with ice packs for refrigerated temperatures. Figure 1a presents the fresh tissue arrival and Figure 1b shows the preparation at the laboratory. Figure 2 shows a schematic representation of the process from eSM-MSC collection to the administration of the combination, i.e., eSM-MSCs and autologous serum (1 × 106 cells/mL and 1 mL of autologous serum in a total volume of 2 mL).2.5. eSM-MSC IsolationAfter collection, the equine synovial membrane was prepared at the Laboratory of Veterinary Cell-based Therapies from ICBAS-UP. The isolation protocol for eSM-MSCs was developed by patented proprietary technology Regenera® (PCT/IB2019/052006, WO2019175773, Compositions in use for the treatment of musculoskeletal conditions and methods for producing the same leveraging of the synergistic activity of two different types of mesenchymal stromal/stem cells, Regenera®). Fresh tissue was transported to the laboratory facilities in a hermetically sealed sterile container in transport media (supplemented with 3% (v/v) penicillin-streptomycin (Gibco®, Waltham, MA, USA) and 3% amphotericin B (Gibco®) and processed within a period of up to 48 h. The synovial tissue was digested using collagenase and the isolated cells were incubated in a static monolayer culture using standard MSC basal medium supplemented with 10% fetal bovine serum (FBS) and maintained in standard culture conditions (37 °C, 5% CO2, and humidified atmosphere) until they reached confluence. Cells from confluent cultures were cryopreserved in 10% dimethylsulphoxide (DMSO) and FBS, at a concentration of 3 × 10⁶ cells/mL, using a control rate temperature freezer (Sy-Lab Cryobiology, SY-LAB Geräte GmbH, Purkersdorf, Austria). For expansion optimization, cells were cryopreserved in passages (P) between P2 and P3 to generate suitable master cell banks (MCBs). Expansion, thereafter, was analyzed during a maximum of 20 cumulative population doublings (cCPDs). The range of cCPDs chosen allowed for enough expansion to maximize the number of cells in the working cell banks (WCB) but kept the cCPDs within the genomic stability range.2.6. SM-MSC Characterization2.6.1. Tri-Lineage Differentiation ProtocolsFor all the differentiation protocols, cells in P4 were used after thawing.Adipogenic Differentiation and Oil Red O StainingFor the adipogenic differentiation protocol, 1 × 104 cells/cm2 were seeded in the wells of a 12-well plate (cell culture plates, 12-well, VWR®, Suwanee, Atlanta, GA, USA), with the addition of the standard culture medium. The plate was incubated under standard conditions for 4 days. After this period, the culture medium of 10 wells was replaced by complete adipogenesis differentiation medium (StemPro® Adipogenesis Differentiation Kit, Gibco®, Waltham, MA, USA), 2 wells were used as controls and maintained with the standard culture medium. Following the manufacturer’s instructions, the media were replaced every 3–4 days and the cells maintained in differentiation for 14 days. At the end of this period, the oil red O staining protocol was performed using a handmade solution. The culture differentiation medium was removed, and the wells were gently washed with PBS. Cells were fixed with 4% formaldehyde (3.7–4% buffered to pH 7, reference# 252931.1315, Panreac AppliChem®, Darmstadt, Germany) for 10 min at room temperature, and the wells were washed 3 additional times with phosphate-buffered saline (PBS). Oil red O solution was added to each well and the plate incubated for 10–20 min at room temperature. Oil red O was discarded, and any excess dye was removed by several washes with PBS. PBS was added to each well for visualization. The aim of this assay was the identification of rounded cells with intracytoplasmic lipid vacuoles and their red coloration due to the exposure to the oil red O solution.Chondrogenic Differentiation and Alcian Blue StainingThawed eSM-MSCs were automatically counted, and cell viability determined (%). Then, the cells were centrifuged, supernatant removed, and the pellet resuspended in culture medium to generate a cell suspension with 1.6 × 107 viable cells/mL. To generate micro-mass cultures, 5 μL droplets of the cell suspension were placed in the center of 10 wells of a 96-well plate (cell culture plates, 96-well, VWR®, Suwanee, Atlanta, GA, USA) to induce chondrogenic differentiation. The plate was maintained under standard conditions for 2 h. After this time, chondrogenic differentiation medium (StemPro® Chondrogenesis Differentiation Kit, Gibco®, Waltham, MA, USA) was added to 8 wells; the other 2 wells were considered to be controls and to these, the standard culture medium was added. Following the manufacturer’s instructions, the media were replaced every 3–4 days and cells maintained in differentiation for 14 days. At the end of this period, the Alcian blue staining, pH 2.5, protocol was performed (Alcian Blue 8GX, Sigma-Aldrich®, St. Louis, MO, USA). The culture differentiation medium was removed, and the wells were gently washed with PBS. The cells were fixed with 4% formaldehyde for 20 min at room temperature, and the wells were washed 3 additional times with PBS. Alcian blue solution was added to each well and the plate incubated for 30 min at room temperature. Then, the Alcian blue was discarded, and the wells were rinsed 3 times with 3% acetic acid (v/v). For neutralization of acidity and for visualization by inverted phase contrast microscopy, distilled water was added to all wells. The aim of this assay was the identification of chondrogenic aggregates and their coloration in blue due to the exposure to Alcian blue solution.Osteogenic Differentiation and Alizarin Red StainingFor osteogenic differentiation, 8 × 103 cells/cm2 were seeded into the wells of a 12-well plate. The plate was maintained under standard conditions for 4 days. After this period, the culture medium of 10 wells was replaced by complete osteogenic differentiation medium (StemPro® Osteogenic Differentiation Kit, Gibco®, Waltham, MA, USA), and 2 wells were used as controls and maintained with the standard culture medium. Following the manufacturer’s instructions, the media were replaced every 3–4 days and the cells maintained in differentiation for 21 days. At the end of this period, the alizarin red s staining protocol was performed using a commercial solution (alizarin red staining solution, Milllipore®, Burlington, MA, USA). The culture differentiation medium was removed, and the wells were gently washed with PBS. The cells were fixed with 4% formaldehyde for 30 min at room temperature, and the wells were washed twice with distilled water. One ml of 40 mM of alizarin red solution was added to each well and the plate incubated for 30 min. Then, the alizarin red solution was discarded, and the wells were rinsed 3 times with distilled water until the supernatant became clear. For visualization by inverted phase contrast microscopy, PBS was added to all the wells. The aim of this assay was to identify calcium-containing osteocytes stained in red after exposure to alizarin red solution. 2.6.2. Karyotype AnalysisThe eSM-MSCs in two different passages (P4 and P7) were submitted to cytogenetic analysis to determine the genetic stability in terms of chromosome number and occurrence of neoplastic changes. For both passages, 70–80% confluence was reached. Then, the culture medium was changed and supplemented with 10 μg/mL colcemid solution (KaryoMAX® Colcemid™ Solution, Gibco®, Waltham, MA, USA). After 4 h, the eSM-MSCs were collected and resuspended in 8 mL of 0.075 M KCl solution, followed by incubation under standard conditions for 15 min. After centrifugation (1700 rpm), 8 mL of ice-cold fixative comprising methanol and glacial acetic at a proportion of 3:1 was added and mixed. Afterwards, the cells were centrifuged again. Three fixation rounds were carried out. After the last centrifugation, the suspension of eSM-MSCs was spread over glass slides. A karyotype analysis was performed by one scorer on Giemsa-stained cells. For the different passages, a specific number of cells in metaphase were evaluated depending on the number of cells with a normal karyotype identified, guaranteeing a better representation of the population under study.2.6.3. Secretome Cell Conditioned Medium (CM) AnalysisThe eSM-MSCs were harvested from equine synovial membrane and maintained in culture, as previously described. The cells in P4 were subjected to an analysis of their conditioned medium (CM) to identify cytokines and chemokines secreted after conditioning. When in culture, after reaching a confluence of around 70–80%, the culture medium was removed, and the culture flasks were gently washed with DPBS two to three times. Then, the culture flasks were further washed two to three times with the basal culture medium of each cell type, without any supplementation. To begin the conditioning, non-supplemented DMEM/F12 GlutaMAX™ (10565018, Gibco®, Thermo Fisher Scientific®, Waltham, MA, USA) culture medium was added to the culture flasks, which were then incubated under standard conditions. The culture medium rich in factors secreted by the cells (CM) was collected after 48 h. The collected CM was then concentrated five times. After collection, it was centrifuged for 10 min at 1600 rpm, and its supernatant collected and filtered with a 0.2 μm syringe filter (Filtropur S®, PES, Sarstedt, Nümbrecht, Germany). For the concentration procedure, Pierce™ Protein Concentrator, 3k MWCO, 5–20 mL tubes (88525, Thermo Scientific®, Waltham, MA, USA) were used. Initially, the concentrators were sterilized following the manufacturer’s instructions. Briefly, the upper compartment of each concentrator tube was filled with 70% ethanol (v/v) and centrifuged at 300× g for 10 min. At the end of the centrifugation, the ethanol was discarded, and the same procedure was carried out with DPBS. Each concentrator tube was subjected to two such centrifugation cycles, followed by a 10 min period in the laminar flow hood to complete drying. Finally, the upper compartment of the concentrator tubes was filled with plain CM (1 × concentration) and subjected to new centrifugation cycles, under the conditions described above, for the number of cycles necessary to obtain the desired CM concentration (5×). The concentrated CM was stored at −20 °C and subsequently subjected to a Multiplexing LASER Bead analysis (Eve Technologies, Calgary, AB, Canada) to identify a set of biomarkers present in the Equine Cytokine 8-Plex Assay (EQCYT-08-501). The list of searched biomarkers includes basic fibroblast growth factor (FGF-2), granulocyte colony-stimulating factor (G-CSF), granulocyte macrophage colony-stimulating factor (GM-CSF), monocyte chemoattractant protein-1 (MCP-1), interleukins (IL) IL-6, IL-8, IL-17A, and human growth-regulated oncogene/keratinocyte chemoattractant (GRO/KC). All samples were analyzed in duplicate.2.6.4. ImmunohistochemistryEarly passages of eSM-MSCs-P0 and -P3 were maintained in culture until a confluence of 70–80% was reached, and then enzymatic detachment was performed with 0.25% trypsin-EDTA solution. A cytoblock was performed fixing the cells with Sure Thin® (Statlab®, Gerwig Ln Columbia, Columbia, MD, USA). Consecutive sections were cut at 2 μm, deparaffinized, hydrated, and submitted to immunohistochemical analysis using the Novolink™ Polymer Detection Systems (Leica Biosystems®, Vista, CA, USA) kit, according to the manufacturer’s instructions. Information regarding the primary antibodies and antigen retrieval recovery methods used in this study is summarized in Table 3.The antibodies were selected to confirm the pluripotent and mesenchymal origin of the eSM-MSCs’ octamer-binding transcription factor 4 (OCT4), homeobox protein (NANOG), proto-oncogene receptor tyrosine kinase or stem cell factor receptor (c-kit), synovial origin (lysozyme), and non-epithelial origin histogenesis (vimentin). Additionally, pan-cytokeratin (AE1 and AE3), synaptophysin, CD31, and glial fibrillary acidic protein (GFAP) were used to confirm there were no vascular, epithelial, neuronal, and neuroendocrine origins of cells, respectively. For each antibody, appropriate negative and positive controls were included, and all primary antibodies were incubated overnight.The final step consisted of microscopic cell observation, evaluation, and photograph using the microscope Eclipse E600 (Nikon®, Tokyo, Japan) and the software Imaging Software NIS-Elements F Ver4.30.01 (Laboratory Imaging®, prague, mmun republic). A semi-quantitative score was used for mmunoexpression evaluation, consisting of the percentage of labeled cells (<5%, 5–80%, and >80%) and labeling intensity (0, negative; +, weak; ++, moderate; and +++, strong). Immunoreactivity was considered positive when distinct nuclear and cytoplasmic staining was recognized in at least 5% of the cells. 2.7. eSM-MSC Solution PreparationThe eSM-MSC solution for local clinical application in the 16 equine patients, was a combination of allogenic eSM-MSCs suspended in autologous serum. Prior to preparation of the final therapeutic combination, autologous serum was isolated from whole blood. Then, 10 mL samples of whole blood were collected into dry blood collection tubes, and after clotting, they were centrifuged at 2300 rpm for 10 min and their supernatant (serum) collected and transferred to a 15 mL falcon. Then, the serum was inactivated through a water bath at 56 °C for 20 min followed by cooling on ice. Finally, the serum was centrifuged and filtered using a 0.22 µm syringe filter and stored at −20 °C until further use. Cryopreserved P3 eSM-MSC batches were thawed in a 37 °C water bath, and the content was transferred to a 10 mL tube with autologous serum and slowly diluted, followed by the addition of sterile DPBS until reaching 10 mL. Then, the mixture was centrifuged at 1600 rpm for 10 min. The supernatant was discarded, and the cell pellet was re-suspended in a mixture of autologous serum at a ratio of 0.8:1. Cell counting and viability were determined by using the trypan blue exclusion dye assay (Invitrogen TM, Waltham, MA, USA) using an automatic counter (Countess II FL Automated Cell Counter, Thermo Fisher Scientific®, Waltham, MA, USA). Then, the cell number was adjusted to 5 × 10⁶ cells/mL, and then 2 mL of the solution of eSM-MSCs suspended in autologous serum was transferred to a perforable capped vial and preserved on ice until the time of administration.2.8. Treatment ProtocolTwenty structures, tendons and ligaments, were treated with a mixture of allogenic eSM-MSCs and autologous serum. The same treatment protocol was used in every case. All equine patients were submitted to identification, anamnesis, physical examination (cardiac and respiratory frequency, body temperature, mucous membrane examination, inspection of the whole body, and palpation), orthopedic examination (evaluation of the limbs, gait inspection and movements (walk, trot and gallop), and flexion test of the main joints for 60 s followed by trot). Lameness was evaluated at a walk and a trot on hard surface and scored on a scale from 0 to 5, according to the AAEP parameters. Complementary diagnostic exams included regional nerve blocks (to identify the pain area), radiographs, and ultrasound image as reported in other studies [21,24,25,27,28,29,30,31,32].Following the assumptions of the exclusion criteria, the horses did not receive any treatment before or after the administration of the therapy protocol. In the case of adverse events occurring, such as inflammatory/anaphylactic reactions or infections, the horses should be immediately evaluated and treated with anti-inflammatories or antibiotics, in accordance with their clinical status. The equine patients were monitored in the 48 h after treatment and any occurrences were registered. Following the treatment, the equine patients were assessed periodically to control the equine patient’s healing evolution and to provide valid comparative data among equine patients within the same study group. Table 4 presents the lesion type casuistic.2.8.1. Intralesional eSM-MSC InjectionSelected horses were sedated with detomidine (0.02 mg/kg), trichotomized, a regional nerve block was performed with lidocaine 2% (20 mg/mL, 2 mL/point), and the surgical skin was disinfected with chlorohexidine and alcohol. The therapeutic combination was aspired to a 2 mL syringe and homogenized, ultrasound was used to identify the lesion site, and an ultrasound guided injection was performed at the lesion over three different points. Finally, a bandage was applied to the limb. All equine patients were injected with phenylbutazone (2.2 mg/kg, IV, SID) at the end of the treatment. The established protocol included a second eSM-MSC administration 15 days after the first treatment using the same protocol.2.8.2. Clinical Evaluation—Serial EvaluationsTissue regeneration was estimated through a lameness evaluation, pain to pressure test, limb inflammation, sensitivity, and ultrasound image (reduction of hypoechoic area and fiber alignment). Lesion ultrasonographic evaluations were performed using a 7.5 MHz linear transductor probe (Sonoscape A5®, Shenzhen New Industries Biomedical Engineering Co Ltd., Shenzhen, China). For each assessment, a complete examination of the structure was conducted by means of longitudinal and transverse scans. The obtained images were evaluated at each examination for two parameters: lesion echogenicity and lesion longitudinal fiber alignment (FA). The contralateral healthy limb was used as comparison. The evaluation was performed on the treatment day (Day 1) as well as on Days 15, 30, and 45 post-treatments, as presented in Figure 3. According to the classification proposed by Guest et al., this is a short term period study [33]. The rehabilitation program consisted of an exercise-controlled program with stall confinement and increasing the amount of time for exercise. Early mobilization included weight-bearing activities, strengthening, and flexibility, and stall rest alone was used as infrequently as possible, as presented on Table 5 [34,35,36,37,38]. Regular ultrasound evaluations were also performed. 3. Results3.1. eSM-MSC IsolationeSM-MSCs were successfully isolated from equine synovial membrane samples and the average total number of cells isolated from the samples was 1.2 × 105 and 5.6 × 105 at Days 6 and 11, respectively, and expanded from the donor. Cells were observed radiating from the explants and those identified in culture showed clear plastic adherence and mostly fibroblast-like morphology, an essential feature to characterize cells as MSCs (Figure 4a,b).3.2. eSM-MSC Characterization3.2.1. Tri-lineage DifferentiationTri-lineage differentiation was confirmed (Figure 5).Adipogenic Differentiation—Oil Red O StainingAdipogenic differentiation was confirmed by the presence of large red stained lipid vacuoles in the cytoplasm due to exposure of oil red O staining.Chondrogenic Differentiation—Alcian Blue StainingChondrogenic differentiation was confirmed by the presence of proteoglycans’ marked deposition in the extracellular matrix which stained blue, confirming the presence of chondrogenic aggregates.Osteogenic Differentiation—Alizarin Red StainingOsteogenic differentiation was demonstrated by the presence of extracellular calcium deposits stained red by alizarin red solution, which dyes chelate complexes with calcium.3.2.2. Karyotype AnalysisThe cytogenetic analysis revealed the presence of 36% normal cells in P4 and 32% normal cells in P7. Tetraploidy was present in 4% of P4 cells and 8% of P7 cells. Aneuploidy represented 60% of the cells in both passages, hypoploidy being the most representative (56%), as shown at Table 6 and Figure 6.3.2.3. Secretome AnalysisThe analysis of CM revealed the production and secretion of several factors with immunomodulatory functions, capable of intervening beneficially in tissue regeneration. The results of the eSM-MSC CM analysis are shown in Figure 7. Seven biomarkers were identified: keratinocyte chemoattractant/growth regulated oncogene (KC/GRO), monocyte chemoattractant protein-1 (MCP-1), interleukin-6 (IL-6), fibroblast growth factor (FGF-2), granulocyte colony-stimulating factor (G-CSF), granulocyte macrophage colony-stimulating factor (GM-CSF), and interleukin-8 (IL-8). The most expressive were KC/GRO and MCP-1. 3.2.4. ImmunohistochemistryThe eSM-MSCs showed strong expressions of OCT4/NANOG, vimentin, and lysozyme which confirmed marked stem cells, non-epithelial cells, and synovial cells, respectively; weak expression of GFAP; and no expression of CD31, synaptophysin, and pan-cytokeratin, as seen in Figure 8, which confirmed no vascular, neuronal, and epithelial origins of cells. Except for GFAP, in which a smaller number of cells exhibited weaker cytoplasmic immunolabeling in P3 as compared with in passage P0, there was preservation of immunoexpression of all the antibodies between passages P0 and P3. The combination of the positive and negative expressions of these different markers confirmed the expected mesenchymal origin of the cells. Figure 8 presents the immunolabeling of the eSM-MSCs.3.3. Treatment ResultsNo horse had any adverse event that required study cessation, unplanned procedures, or additional treatments. All intra-tendinous injections and follow-up procedures had no adverse reactions (inflammation, infection, deterioration of the lesion, increased lameness), as shown by Godwin et al. (2012) [39]. No horse had abnormalities identified on the weeks following the injection.Tendon/ligament regeneration occurred in a time frame of less than 30 days in 80% of the cases and between 30–90 days in 20% of the cases. In this study, eight horses had a lesion on the right front limb, six horses had a lesion on the left front limb, and two horses had a lesion on the right hind limb. There were 14 acute cases and two chronic cases. Chronic cases were diagnosed 6 months before our approach.After Day 90, meaning they had completed the proposed rehabilitation physical program, the horses started cantering and started to return to their usual work plan. By Day 120 post the first treatment, 87.5% of the horses were back to full work, with the exception of the 12.5% who needed another 30 days to return to full work. All horses returned to the same level of sport activity they had before injury. Table 2 and Table 7 summarize the recovery progress, with the respective ultrasound images in Figure 9 and Figure 10. At Day 30, the group that fully recovered demonstrated both a fulfilled ultrasound cross-sectional area and good fiber alignment. There was also no evidence of pain and lameness. Below, transversal and longitudinal ultrasound images of four cases on Day 1 and on Day 30 are presented. After the eSM-MSC treatment, all horses were submitted to a rehabilitation program, as explained in Table 4.Radiograph exams were performed to rule out the presence of other associated pathologies and regional nerve blocks were performed to better localize the injured region originating the pain.Ultrasound images at Day 1 and at Day 30 clearly illustrate the evolution of tendon regeneration. Changes in echogenicity, fiber alignment, and cross-sectional area are evident, as seen in Figure 10.4. DiscussionRecently, eSM-MSCs have become an interesting subject for those who study cellular and cell-based therapies due to their promising ability to promote tissue regeneration with high capacity of regeneration of articular structures, tendons, and ligaments. Regarding the collection, isolation, expansion, freezing, and thawing protocols used in this clinical trial, it was possible to use these cells in equine tendon regenerative treatments. The full characterization of eSM-MSCs presents a significant challenge since eSM-MSCs are not as well studied as MSCs from other species, namely human MSCs. However, in this study, their stemness and origins were confirmed through different processes: trilineage differentiation, karyotype, secretome, and immunohistochemistry. All the SM-MSC cultures presented monolayer culture, plastic adherence capacity, and fibroblast-like shape [40,41,42,43], accomplishing some of the minimal criteria defined by ISCT. Successful osteogenic, chondrogenic, and adipogenic differentiation was also demonstrated. De Bari et al. [9] were the first group of researchers to isolate MSCs from synovial tissues.The karyotype presented some genomic variations when the number of passages was increased. That was consistent with some studies regarding genomic variations along cell passages [44,45,46,47,48]. DNA replication is a critical event for timely genome duplication. Errors in replication lead to genomic instability across evolution [49]. Prieto Gonzalez et al. considered that genomic instability, incurred during the process of stem cell isolation, culture expansion, and reprogramming, might be the most critical point of a stem cell-based therapeutic approach as a viable option from the clinical perspective [50]. Peterson et al. highlighted that there was very little evidence linking genomic abnormalities, for example, in human pluripotent stem cells (hPSCs) with tumorigeneses [44]. The frequency and effects of variations have increased with the development of even more sensitive methods for detecting genomic variation [45].As reported by Simona Neri, the interpretation of genetic instability and senescence of cultured MSCs is controversial, but the increasing incidence of genetic alterations at advanced culture times clearly indicates that few culture passages correspond to a reduced chance to harbor dangerous alterations. Therefore, prudent behavior is desirable with a reduction in culture times as much as possible to avoid safety concerns [51]. More studies must be performed in this area. During the last decade, it has been shown that the therapeutic effectiveness of MSCs is due mainly to the release of paracrine factors, namely CM, composed of soluble (cytokines, chemokines, and growth factors) and nonsoluble factors (extracellular vesicles) that are primarily secreted in the extracellular space by stem cells [52]. CM’s paracrine signaling can be considered to be the primary mechanism by which MSCs contribute to the healing process, and therefore, their study has become an interesting subject [53,54].In our study, eSM-MSCs revealed a CM with a high level of KC/GRO, MCP-1, Il-6, FGF-2, G-CSF, GM-CSF, and IL-8. This highlights the intense activity of fibroblasts, producing KC/GRO that is chemotaxic for neutrophils during inflammation. MCP-1 is essential for reperfusion and the successful completion of musculoskeletal tissue after an ischemic injury [55]. Macrophages are tissue resident cells involved in tissue regeneration along with their inflammatory and infection responses [56]. IL-6 is a proinflammatory and angiogenic interleukin capable of increasing the expression of growth factors; reactivating, for example, intrinsic growth programs of neurons; promoting axonal regrowth; and creating a link between inflammation and tissue regeneration [57,58]. FGF-2 is a recognized GF responsible for proliferation of tenogenic stem cells. FGF-2 signaling has been reported to produce a tendon progenitor population that expressed scleraxis during somite development [59]. FGF-2 plays a crucial role in cell proliferation and collagen production, becoming a useful GF for tissue regeneration by promoting stem cell proliferation [60]. G-CSF is a cytokine that mobilizes bone marrow-derived cells (BM-DCs) to peripheral blood. A study suggested that injection of G-CSF to promote BM-DC release in the target area, i.e., rotator cuff, effectively enhanced rotator cuff healing by promoting tenocyte and cartilage matrix production [61]. Wright et al. presented a study that confirmed skeletal muscle damage, including muscle damage following strenuous exercise, induced an elevation in plasma G-CSF, implicating it as a potential mediator of skeletal muscle repair [62]. Recent human trials have shown the benefits of G-CSF administration as a treatment for neuromuscular diseases, considering that G-CSF affects skeletal muscle, leading to functional improvements [63,64,65,66,67,68]. GM-CSF is an hematopoietic growth factor with proinflammatory functions [69]. Major sources of GM-CSF are T and B cells, monocyte/macrophage endothelial cells, and fibroblasts. Neutrophils, eosinophils, epithelial cells, mesothelial cells, Paneth cells, chondrocytes, and tumor cells can also produce GM-CSF [70]. Paredes et al. evidenced that elevated levels of proinflammatory factors such as those found at these cells CM (GM-CSF, G-CSF, Il-6, IL-8 and IL-17), were implicated in the activation of resident tendon cells for effective healing, stimulating tendon cell proliferation [71,72]. IL-8 is one of the major mediators of inflammatory response and is a potent angiogenic factor. This is similar to IL-6, but IL-8 has a longer half-life [73].A recent study highlighted that hematopoietic factor promoted tendon healing in aged mouse tendons. Histochemical results demonstrated that vascularization of the injury site was significantly elevated. It was concluded that vascular endothelial growth factor (VEGF) played an important role in decreasing adipocyte accumulation and also improved vascularization of the tendon during aged tendon healing. Active regulation of VEGF may improve the treatment of age-related tendon diseases and tendon injuries [74].Studies with human BM-MSCs using a human-specific proteome profiler array with different angiogenic factors such as VEGF-A, IL-6, IL-8, platelet-derived growth factor A (PDGF-A), endothelin-1 (ET1), and urokinase plasminogen activator (uPA), which had not been previously reported in the CM of human MSCs, were also identified in an equine array, confirming what we found in this study [75]. This factor has been proposed as a modulator of the different neovascularization stages, through the enhancement of VEGF gene promotor activity [75,76]. Schokry et al. [77] reported that BM-MSC therapies have recovery times of 3–6 months and conservative therapeutic methods allow recovery in 12–18 months without regeneration but with formation of fibrous scar tissue. Retrospectively, no re-injuries of tendons have occurred in horses treated with this new approach, during the study frame time. In the literature [78], Smith et al. referred to a low percentage re-injury rate of 27% for SFD tendonitis treated with bone marrow stem cells. Horses returned to “full function” as defined by Cook et al. and modified by Guest et al. [33,79].A study using a murine osteoarthritis (OA) model demonstrated that an injection of MSCs CM, similarly to injection of MSCs, resulted in early pain reduction and had a protective effect on the development of cartilage damage in a murine OA model, by using the regenerative capacities of the MSCs-secreted factors [80].Interestingly, the results accumulated so far have provided evidence that veterinary patients affected by naturally occurring diseases should provide more reliable outcomes of cell therapy than laboratory animals, thus, allowing translating potential therapies to the human field. More recently, a cell-free therapy based on MSCs CM has been proposed. Even though there are very few clinical reports to refer to in veterinary medicine, recent acquisitions suggest that MSC-derived products may have major advantages compared to the related cells, for example, they are considered safer and less immunogenic [52]. As evidenced before, eSM-MSC CM factors are able to promote tendon healing by reducing inflammation and fatty infiltration, stimulating cell proliferation and tenogenic differentiation [81].In this study we used a cell-based therapy instead of CM itself, but we were aware of its effect and potential on cell-based therapies; its advantages and therapeutic effects were the reason why this study was performed.To better characterize the cells under study, we performed immunohistochemistry assays. The choice of markers was based on a previous work [8] and included several of the criteria used for humans, as determined by the ISCT. Results of our study demonstrated the presence of the embryonic stem cell markers OCT4 and NANOG. Detection of these markers has been previously described by Beltrami et al., in multipotent adult stem cells (HMASC) from human bone marrow [82], as well as, by Riekstina et al., who also demonstrated the presence of these markers in HMASC derived from bone marrow, adipose tissue, heart, and dermis [83]. Greco et al. also evidenced elevated expression of OCT4 in P3 MSCs and hypothesized OCT4 expression could be an indicator of MSC differentiation potential in clinical diagnostics [84]. In equine characterization of synovial fluid and membrane-derived MSCs, Prado et al. also evidenced the presence of NANOG and OCT4 markers [19]. In contrast, Fulber et al. had no positive results for these two markers in equine mesenchymal stem cells of synovial tissues [43]. Vimentin, a mesenchymal stem cell marker, was also detected, suggesting the mesenchymal origin of cells. The presence of lysozyme confirmed the synovial origin of cells, as stated by Fulber et al. [43].The immunohistochemistry analysis showed the absence of CD31, sinaptophysine, and pan-cytokeratin expressions, confirming no vascular, neuronal and epithelial origins of cells. GFAP was weakly expressed, being less expressed in P3 than in P0 cells. CD31 was performed to investigate the presence of hematopoietic cells in eSM-MSCs. The expression of VEGF was not found, these results being similar to those from Fulber et al., and to other authors that evidenced the absence of hematopoietic markers [43,85]. The absence of neuronal and dermal markers was also consistent with other studies [19,43].In our clinical trial, we treated mainly early acute lesions; 87.5% of the cases were acute lesions of tendons or ligaments. Therefore, we created a master cell bank of allogenic eSM-MSCs suitable for treatments in early acute phases versus treatments with autologous cells where time of tissue collection, preparation, and cell culture need to be considered. Furthermore, cell harvesting for autologous treatment is an invasive procedure which is unnecessary with this new eSM-MSC solution. The possibility of having a master cell bank enables faster healing of the organ and a quicker return to sport life. Horses spend less time in recovery time and have a regenerated tissue instead of a fibrotic tissue. These are some advantages of the eSM-MSC solution. Another concern is that in the early stages of the lesion there is an inflammatory phase; however, the paracrine factors released by eSM-MSCs also have anti-inflammatory action, reducing inflammation. Chronic cases represented 12.5% of the cases, involving four structures. Three of the horses recovered in 30 days and one of the horses had a delayed recovery time.The delayed recovery time in 20% of the structures, meaning 12.5% of the horses, was due to, in Case 6, an increased number of involved structures (more than one tendon or ligament) and a foot conformation abnormality, as the horse had a fetlock hyperextension that was impairing correct tendon healing. This was corrected with special shoeing. Inappropriate rehabilitation program (Case 7) was another cause of delayed recovery time. As soon as the corrective shoeing was performed, ligament regeneration started.We could also conclude that lameness grade was not directly correlated with lesion cross-sectional area. Horses with ultrasonographic cross-sectional grade 1, 2, and 3 lesions presented lameness grade 4/5, which was observed in 9 of 16 patients. Lameness grade 3/5 was presents in 4 of 16 of equine patients with ultrasonographic cross-sectional grade 1 and 2 lesions. Lameness grade 2/5 was present in 3 of 16 equine patients with ultrasonographic cross-sectional grade 1 lesions.Kamm et al. (2021) concluded that based on the evidence to date, tendons appear to have improved healing when treated with allogeneic MSCs, and the use of these treatments in equine tendon and ligament lesions is warranted [86]. Colbath et al. (2020) claimed that some of the advantages of using allogenic stem cells include the ability to bank cells and to also reduce the treatment time, to collect MSCs from younger donor animals, and the ability to manipulate banked cells prior to administration [87]. Some of the disadvantages focused on the risk of immunological reactions. However, currently, there are several studies in horses accumulating evidence that allogeneic MSCs may be a safe alternative to autologous MSCs [87]. Nevertheless, the donor’s health must always be taken into consideration as well as the donor’s age [88].5. ConclusionsTo sum up, this study accomplishes the criteria for reporting veterinary and animal medicine research for MSCs in orthopedic applications [33] and the ISCT perspective on immune assays for MSC’s criteria for advanced phase clinical trials [89], confirmed by plastic adherence, tri-lineage differentiation, synovial membrane origin, spindle-shaped cells, as well as proliferative and immune modulatory capacity proven by immunohistochemistry and CM. From a clinical point of view, the idea of having an allogenic eSM-MSC cell bank is very interesting. Therefore, the possibility of having a universal donor who can provide a large amount of eSM-MSCs, to culture and preserve non-immunogenic cells whose availability is immediate, allowing a quick and effective therapeutic answer in acute stages of musculoskeletal lesion is the paramount goal of orthopedic medicine.From a “one-health” perspective, equines play an important role as a model for human musculoskeletal disorders; the high-level analogy between human and equine structures may have a great translational value for both species for future clinical aspects [28,90]. There are significant resemblances between equine SDFT and human Achilles tendon with respect to the size of anatomical structure and load, function (energy store), pathophysiology of tendon injury, and the healing response under activity or traumatic rupture compared to other species [90]. Moreover, considering the result of tendinopathy in equine species which reflects the conditions encountered in humans, the horse is accepted as an appropriate model in this area by the research community and by other authorities such as the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA).Based on the clinical, ultrasonographic, and performance outcomes identified in the present study, the use of eSM-MSCs together with autologous serum solution has proven its efficiency for tendon and ligament repair and contributes to reduce the recovery period and subsequent rapid return to athletic activity. The therapy was demonstrated to be safe and had no adverse findings. The clinical results and athletic outcomes of the horses were very positive. Comparing our study with others, using for example BM-MSCs, it seems that our new approach has shorter recovery times and fewer re-injuries [39,77]. These results encourage the use of eSM-MSCs and autologous serum for the treatment of tendonitis and desmitis, since they can regenerate tendon and ligament tissue and regain organ function, enhancing the return to competition in excellent time frames.

animals : an open access journal from mdpi

10.3390/ani13101712

PMC10215144

Troglonectes is a small-body loach endemic to the Guangxi and Guizhou provinces of China, showing a particular affinity for cave areas. Twenty species were recorded in this genus, including one new species. The new species, Tr. canlinensis, can be distinguished from other congenetic species by their morphological characteristics and molecular evidence. In the genus of Troglonectes, the eye, lateral line and scale present or absent, the number of branched pectoral fin rays, caudal fin rays and anal fin rays, and the depth of the upper adipose keel on the caudal peduncle are important identifying characteristics.

A new species of the genus Troglonectes is described based on specimens from a karst cave in Andong Town, Xincheng County, Liuzhou City, Guangxi, China. Troglonectes canlinensis sp. nov. can be distinguished from its congener species by the following combination of characteristics: eye degenerated into a black spot; whole body covered by scales, except for the head, throat, and abdomen; incomplete lateral line; forked caudal fin; 8–10 gill rakers on the first gill arch; 13–14 branched caudal fin rays; 8–9 branched dorsal fin rays; 5–6 anal fin rays; 9–10 pectoral fin rays; upper adipose keel depth mostly 1/2 of the caudal peduncle depth; and caudal fin forked.

1. IntroductionCave loaches of the genus Troglonectes Zhang, Zhao, and Yang, 2016 (abbreviation is Tr. in this study in order to differ from the abbreviation of Triplophysa) are small-bodied fish that mainly occur in the Guangxi and Guizhou provinces of China, showing a particular affinity for cave areas. Troglonectes was separated from the genus Oreonectes Günther, 1838, that Du et al. [1] divided Oreonectes into the platycephalus group, i.e., caudal fin rounded or truncated, and the furcocaudalis group, i.e., caudal fin forked. Subsequently, Zhang et al. [2] proposed the genus Troglonectes and assigned seven nominal species to Troglonectes, i.e., Tr. acridorsalis (Lan, 2013), Tr. barbatus (Gan, 2013), Tr. elongatus (Tang, Zhao, and Zhang, 2013), Tr. macrolepis (Huang, Du, Chen, and Yang, 2009), Tr. microphthalmus (Du, Chen, and Yang, 2008), and Tr. translucens (Zhang, Zhao, and Zhang, 2006), in addition to the type species Tr. furcocaudalis (Zhu and Cao, 1987). Troglonectes can be distinguished from other genera in the Nemacheilidae by possessing narrowly separated nostrils, tube-shaped anterior nostril, tip of the anterior nostril extending into the barbel, dorsal fin origin anterior to the pelvic fin origin, and caudal fin forked or truncated [2]. Except for the species of Oreonectes placed in Troglonectes, some species of Paracobitis and Triplophysa were also placed in Troglonectes based on their morphology and molecular evidence. Chen et al. [3] described P. longibarbatus Chen, Yang, Sket, and Aljancic, 1998 from Libo County, Guizhou, but Du et al. [1] placed it in the Triplophysa, based on the morphological characteristics, elongated barbel-like anterior nostril, and sexual dimorphism present in males. Li et al. [4] and Lin et al. [5] described P. maolanensis Li, Ran, and Chen, 2006 and T. jiarongensis Lin, Li, and Song, 2012 from Guizhou Province, respectively. However, Huang et al. [6] and Luo et al. [7] placed P. longibarbatus, P. maolanensis, and T. jiarongensis in the Troglonectes. Subsequently, Luo et al. [7] treated T. jiarongensis as a synonym of Tr. elongatus. Additionally, Huang et al. [6] mentioned that O. daqikongensis Deng, Wen, Xiao, and Zhou, 2016 and O. shuilongensis Deng, Wen, Xiao, and Zhou, 2016 also belong to the genus Troglonectes due to the forked caudal fin, dorsal fin originating anterior to the pelvic fin origin, and presence of caudal crests. Luo et al. [7] placed T. huanjiangensis Yang, Wu, and Lan, 2011, T. lihuensis Wu, Yang, and Lan, 2012, T. lingyunensis (Liao, Wang, and Luo, 1997), and O. retrodorsalis Lan, Yang, and Chen, 1995 in the Troglonectes based on molecular analysis. Zhao et al. [8] described one new species, T. hechiensis Zhao, Liu, Du, and Luo, 2021, and stated that 17 species were contained within the Troglonectes. Luo et al. [7] established one new genus, named Karstsinnectes Zhou, Luo, Wang, Zhou, and Xiao, 2023 (type species Oreonectes anophthalmus Zheng, 1981), and placed O. acridorsalis and Heminoemacheilus parvus Zhu and Zhu, 2014 in this genus. In conclusion, 19 species of Troglonectes have been recorded in China, including Tr. barbatus, Tr. daqikongensis, Tr. donglanensis, Tr. dongganensis, Tr. duanensis, Tr. elongatus, Tr. furcocaudalis, Tr. hechiensis, Tr. huanjiangensis, Tr. jiarongensis, Tr. lihuensis, Tr. lingyunensis, Tr. longibarbatus, Tr. macrolepis, Tr. maolanensis, Tr. microphthalmus, Tr. retrodorsalis, Tr. shuilongensis, and Tr. translucens.In July 2022, 10 specimens of Troglonectes were collected from a cave in Andong Town, Xincheng County, Liuzhou City, Guangxi Zhuang Autonomous Region (hereinafter referred to as Guangxi), China. Morphological and molecular evidence supported these loach specimens representing a new species of Troglonectes. Hence, the new species is described herein.2. Materials and MethodsAll care and use of experimental animals complied with the relevant laws of the Chinese Laboratory of Animal Welfare and Ethics (GB/T 35892-2018). Specimens of Troglonectes canlinensis sp. nov. were collected by F.G. Luo and euthanized rapidly by an overdose of clove oil anesthetic. The right-side pectoral fin and pelvic fin of one specimen were removed and preserved in 99% ethanol. The specimens for the morphological study were stored in 10% formalin, then transferred to 75% alcohol for long-term preservation in the Kunming Natural History Museum of Zoology, Kunming Institute of Zoology (KIZ), Chinese Academy of Science (CAS), China.Counts and measurements followed Du et al. [1,9], Tang et al. [10], and Lan et al. [11]. Complete mitochondrial genes were sequenced by the Science Corporation of Gene (China) following standard Illumina protocols. Genome sequencing data were submitted to GenBank under Accession No. OQ129618. We retrieved twenty-one complete mitochondrial genomes and five cytb reference sequences of twenty-four Nemacheilidae and two Botiidae species from the NCBI GenBank database for phylogenetic tree reconstruction. Parabotia fasciata Dabry de Thiersant, 1872 and Leptobotia elongata (Bleeker, 1870), two species of Botiidae, were used as outgroups. To test the phylogenetic position of Troglonectes canlinensis sp. nov., Bayesian inference (BI) analysis was performed using MrBayes on XSEDE (v3.2.7a) and the CIPRES Science Gateway [12]. Two runs were performed simultaneously with four Markov chains starting from a random tree. The chains were run for five million generations and sampled every 100 generations. The first 25% of the sampled trees were discarded as burn-in, and the remaining trees were used to create a consensus tree and estimate the Bayesian posterior probabilities (BPPs).3. ResultsTroglonectes canlinensis sp. nov. (Table 1, Figure 1, Figure 2 and Figure 3)Holotype. Kunming Natural History Museum of Zoology, KIZ-GXNU202210, 36.0 mm standard length (SL), Andong Town, Xincheng County, Guangxi Zhuang Autonomous Region, China; 24°18.57′ N, 108°59.61′ E, 179 m a.s.l.; collected by F.G. Luo, 20 July 2022.Paratypes. KIZ-GXNU202207–09, 9 ex., 29.9–54.3 mm SL, collected with holotype.Diagnosis. Troglonectes canlinensis sp. nov., T. duanensis, T. lingyunensis, T. macrolepis, T. hechiensis, and T. retrodorsalis share their whole trunk being scaled, except for the head and area between the pectoral fins and pelvic fins; other species of Troglonectes have scaleless bodies or bodies scaled after the dorsal fin origin in Tr. furcocaudalis. However, the new species can be distinguished from T. duanensis by the incomplete lateral line (vs. absent), from T. lingyunensis and T. macrolepis by the eye being present (vs. eye reduced to black pigment), from T. hechiensis by the 8–10 inner-gill rakers on first gill arch (vs. 14), and from T. retrodorsalis by the tip of the anterior nostril being elongated to barbel-like and the nostril barbel length being nearly twice the nostril tube length (vs. nostril barbel length being less than 1/2 of the tube length).Description. The morphometric data of the type specimens of Troglonectes canlinensis sp. nov. are given in Table 1. Dorsal fin with 4 unbranched and 8–9 branched rays; anal fin with 3 unbranched and 5–6 branched rays; pectoral fin with 1 unbranched and 9–10 branched rays; pelvic fin with 1 unbranched and 5–6 branched rays, caudal fin with 13–14 branched rays; and 8–10 inner-gill rakers on the first gill arch. Vertebrae 4 + 34 (one specimen)Body elongated, slightly flattened in front, strongly compressed in back. Dorsal profile convex and ventral profile straight in live specimen, but it inversed in preserved specimens. From snout to dorsal fin origin, the body depth increases to its maximum, maximum body depth of 18.2–21.3% SL. Head slightly depressed and flattened, maximum head width greater than the deepest head depth. Anterior and posterior nostrils adjacent, distance less than the diameter of the posterior nostril. Eyes reduced, eye diameter 7.5–11.6% of the lateral head length. Mouth inferior, snout obtuse, upper and lower lips with small furrows and without papillae, median of the lower lip with a V-shaped notch. Three pairs of barbels, inner, outer, and maxillary barbels, extend vertically to the posterior margin of the anterior nostril, anterior margin of the eye, and preopercle, respectively.Distal margin of dorsal fin truncates, origin anterior to the pelvic fin origin, predorsal length of 54.2–58.6% SL. Tip of pectoral fin reaching halfway to the pelvic fin origin. Tip of pelvic fin far away from the anus. Anus with close-set anal fin base. Caudal fin forked, upper part slightly longer than the lower part. Upper and lower edges of the caudal peduncle with caudal adipose keels, upper adipose keel height mostly 1/2 of the caudal peduncle depth. Caudal peduncle length 90.2–119.0% of its depth. Body trunk covered by tiny scales, except for the ventral surface before the pelvic fin origin. Lateral line incomplete. Cephalic lateral line system with 3 + 3 supratemporal, 6 supraorbital, 3 + 8 infraorbital, and 7–11 preoperculo-mandibular pores.Stomach U-shaped, intestine long, after stomach, with a bend. Swim bladder divided into two chambers. Anterior chamber covered by dumbbell-shaped bony capsule, and posterior chamber developed.Coloration. Dorsal surface and trunk of body yellowish brown, abdomen gray and translucent, stomach and intestine visible from outside. Fin membrane hyaline.Distribution and habitat. The new species was collected from Andong Township, Xincheng County, Laibin City, Guangxi Zhuang Autonomous Region, China (24°18.57′ N, 108°59.61′ E). Troglonectes canlinensis sp. nov. lives in a karst cave, where water accumulates to form a pool. Most specimens were collected in the rainy season. During the winter, the pool dries up and the cave opening is too narrow for human access. The water temperature was 20 °C during the survey period in July 2022.Etymology. The specific name “canlinensis” is derived from the pinyin of “can” and “lin”, which refer to resplendence and forest, respectively, with “canlin” symbolizing health and tenacious vitality. Troglonectes canlinensis sp. nov. is valuable and rare and requires strong vitality to maintain a viable population. We suggest the common Chinese name “灿 (càn) 林 (lín) 洞 (dòng) 鳅 (qīu)”.Genetic comparisons. The molecular phylogenies based on BI analysis showed that Troglonectes species formed a monophyletic group, sister to the genus Paranemachilus. Troglonectes canlinensis sp. nov. was sister to the clade including T. dongganensis, T. duanensis, T. macrolepis, T. microphthalmus, and T. translucens, with bootstrap values of 100. Additionally, the species of Troglonectes were divided into two sub-clades: sub-clade 1 contained species with truncated caudal fins, i.e., Tr. shuilongensis, Tr. retrodorsalis, and Tr. hechiensis; sub-clade 2 contains species with forked or emarginated caudal fins, i.e., Tr. elongatus, Tr. jiarongensis, Tr. dongganensis, Tr. longibarbatus, Tr. daqikongensis, Tr. barbatus, Tr. furcocaudalis, Tr. duanensis, Tr. donglanensis, Tr. microphthalmus, Tr. macrolepis, and Tr. canlinensis sp. nov.Mitochondrial differentiation. The pairwise comparisons of ctyb revealed that the average uncorrected p-distances interspecies of Troglonectes ranged from 0.2% to 12.2% (average 7.7%, Table 2). The maximum uncorrected p-distance was between Tr. jiarongensis and Tr. barbatus, and the minimum p-distance was between Tr. translucens and Tr. donglanensis. The average uncorrected p-distance between Tr. canlinensis sp. nov. and other congeneric species ranged from 3.0% to 9.0% (average 6.8%). Identification Key to Species of Troglonectes1. Eye present··················································································································································································································2–. Eye degenerated or absent·····················································································································································································52. Body scaled after dorsal fin origin·········································································································································Tr. furcocaudalis–. Whole body scaled except for head and thorax··············································································································································33. Caudal fin forked·············································································································································································Tr. duanensis–. Caudal fin truncated································································································································································································44. Caudal peduncle length 12.0–13.6% SL··································································································································Tr. hechiensis–. Caudal peduncle length 10.8–12.0% SL·····························································································································Tr. retrodorsalis5. Eye degenerated with black pigment····························································································································································6–. Eye absent··················································································································································································································106. Body scaleless·············································································································································································································7–. Whole body scaled except for head and thorax··············································································································································87. Upper adipose keel height larger than caudal peduncle depth·············································································Tr. microphthalmus–. Upper adipose keel height mostly 1/2 the caudal peduncle depth···········································································Tr. donglanensis8. Posterior chamber of swim bladder degenerated············································································································Tr. lingyunensis–. Posterior chamber of swim bladder developed·············································································································································99. Total of 12–13 inner gill rakers on first gill arch·····················································································································Tr. macrolepis–. Total of 8–10 inner gill rakers on first gill arch······································································································Tr. canlinensis sp. nov.10. Caudal fin truncated································································································································································Tr. shuilongensis–. Caudal fin emarginated or forked····································································································································································1111. Caudal fin emarginated······················································································································································································12–. Caudal fin forked····································································································································································································1412. Lateral line complete···································································································································································Tr. jiarongensis–. Lateral line incomplete or absent······································································································································································1313. Lateral line incomplete································································································································································Tr. translucens–. Lateral line absent················································································································································································Tr. lihuensis14. Lateral line absent································································································································································································15–. Lateral line complete or incomplete································································································································································1615. Standard length 2.6–3.5 times the lateral head length·······································································································Tr. barbatus–. Standard length 4.3–4.9 times the lateral head length································································································Tr. huanjiangensis16. Lateral line complete···························································································································································································17–. Lateral line incomplete·························································································································································································1917. Dorsal fin with six branched rays, anal fin with four branched rays······································································Tr. maolanensis–. Dorsal fin with eight or nine branched rays, anal fin with six branched rays···················································································1818. Standard length 10.1–14.0 times the caudal peduncle depth··················································································Tr. daqikongensis–. Standard length 14.5–18.1 times the caudal peduncle depth····················································································Tr. longibarbatus19. Pelvic fin origin opposite the dorsal fin origin···············································································································Tr. dongganensis–. Pelvic fin origin anterior to the dorsal fin origin······················································································································Tr. elongatus4. DiscussionThe genus Troglonectes is currently distributed in the Pearl River system in Guangxi and Guizhou Provinces, and is endemic to China. Although Zhang et al. [2] mentioned that one of the identifying features of the genus is a forked caudal fin, there are truncated, emarginated, and forked caudal fins, three types of caudal fin. The phylogenetic tree indicates that the species of Troglonectes divided into sub-clade 1 contains species with truncated caudal fins and sub-clade 2 contains species with emarginated or forked caudal fins. Hence, the caudal fin shape and phylogenetic tree support that Troglonectes could be divided into two groups; the truncated caudal fin group contains Tr. hechiensis, Tr. retrodorsalis, and Tr. shuilongensis, and the emarginated or forked caudal fin group contains Tr. donglanensis, Tr. microphthalmus, Tr. macrolepis, Tr. canlinensis sp. nov., Tr. lingyunensis, Tr. barbatus, Tr. huanjiangensis, Tr. longibarbatus, Tr. maolanensis, Tr. daqikongensis, Tr. dongganensis, Tr. elongatus, Tr. translucens, Tr. jiarongensis, Tr. lihuensis, Tr. furcocaudalis, and Tr. duanensis. Thus, based on our BI analysis and external characteristics, the genus description for Troglonectes includes the following characteristics: anterior and posterior nostrils separated by a short distance shorter than the diameter of the posterior nostril, tip of anterior nostril elongated to barbel-like, and adipose keels on the upper and lower edges of the caudal peduncle present.Luo et al. [7] treated Tr. donglanensis and Tr. duanensis as synonyms of Tr. translucens, and Tr. jiarongensis and Tr. dongganensis as synonyms of Tr. elongatus based on morphological characteristics and a lack of genetic differences, respectively. Troglonectes dongganensis, Tr. elongatus, Tr. jiarongensis, and Tr. longibarbatus formed a monophyletic group in the phylogenetic tree, and the genetic distance was 0.4–1.0% (average 0.7%). However, they can be morphologically distinguished from each other by the lateral line (complete in Tr. jiarongensis and Tr. longibarbatus, incomplete in Tr. elongatus and Tr. dongganensis, and absent in Tr. huanjiangensis), branched caudal fins (16 in Tr. jiarongensis and 13–14 in other species), and body depth (8.6–10.7% SL in Tr. elongatus, and more than 13% in Tr. dongganensis, Tr. huanjiangensis, Tr. jiarongensis, and Tr. longibarbatus). Hence, we treated T. dongganensis, T. elongatus, T. huanjiangensis, T. jiarongensis, and T. longibarbatus as valid species in this study. Additionally, Tr. donglanensis, Tr. duanensis, and Tr. translucens can be distinguished from each other by the 16 branched caudal fins in Tr. translucens (vs. 13–14 in Tr. donglanensis and Tr. duanensis) and the body being covered by scales and the lateral line being absent in Tr. duanensis (vs. scaleless and incomplete lateral line in Tr. donglanensis and Tr. translucens). Thus, we propose Tr. donglanensis, Tr. duanensis, and Tr. translucens as valid species.Within the genus Troglonectes, 20 valid species were recorded, including the new species. Troglonectes canlinensis sp. nov. can be distinguished from Tr. hechiensis, Tr. retrodorsalis, and Tr. shuilongensis by its forked caudal fin (vs. truncated) and upper adipose keel height being mostly 1/2 of the caudal peduncle depth (vs. 1/4), and it can be further distinguished from Tr. shuilongensis by its degenerated eye with a black pigment (vs. absent), scaled body (vs. scaleless), incomplete lateral line (vs. complete), and 8−10 inner-gill rakers on the first gill arch (vs. 10−12); from Tr. hechiensis by the 8−10 inner-gill rakers on the first gill arch (vs. 14) and 9−10 branched pectoral fin rays (vs. 11); and from Tr. retrodorsalis by the 8−10 inner-gill rakers on the first gill arch (vs. 13−14) and 9−10 branched pectoral fin rays (vs. 11−12). Troglonectes canlinensis sp. nov. is different from Tr. translucens, Tr. jiarongensis, and Tr. lihuensis owing to its forked caudal fin (vs. emarginated) and scaled body (vs. scaleless); it can be further differentiated from Tr. jiarongensis and Tr. lihuensis by its incomplete lateral line (vs. absent in Tr. lihuensis and complete in Tr. jiarongensis) and upper adipose keel height being mostly 1/2 of the caudal peduncle depth (vs. equal to the caudal peduncle depth); and from Tr. jiarongensis by the 13−14 branched caudal fin rays (vs. 16). The new species is different from Tr. furcocaudalis owing to its whole body being scaled, except for the head and thorax (vs. scaled after the dorsal fin origin), 8−10 inner-gill rakers on the first gill arch (vs. 12−13), and 5−6 branched pelvic fin rays (vs. 7); from Tr. duanensis owing to the incomplete lateral line (vs. absent), 8−10 inner-gill rakers on the first gill arch (vs. 13), anal fin with 5−6 branched rays (vs. 6−7), and eye degenerated with black pigment (vs. present); from Tr. lingyunensis by the developed posterior chamber of the swim bladder (vs. degenerated), caudal fin with 13−14 branched rays (vs. 16), dorsal fin with 8−9 branched rays (vs. 6−7), and upper adipose keel height being mostly 1/2 half of the caudal peduncle depth (vs. 1/4); and from Tr. macrolepis by the 8−10 inner-gill rakers on the first gill arch (vs. 12−13), dorsal fin with 8−9 branched rays (vs. 9−11), pectoral fin with 9−10 branched rays (vs. 10−12), and upper adipose keel height being mostly 1/2 of the caudal peduncle depth (vs. equal with caudal peduncle depth). Troglonectes canlinensis sp. nov. is different from Tr. barbatus, Tr. huanjiangensis, Tr. longibarbatus, Tr. maolanensis, Tr. daqikongensis, Tr. dongganensis, Tr. elongatus, Tr. donglanensis, and Tr. microphthalmus owing to its scaled body (vs. scaleless); it can be further distinguished from these species by the eye being degenerated with black pigment (vs. absent in Tr. barbatus, Tr. huanjiangensis, Tr. longibarbatus, Tr. maolanensis, Tr. daqikongensis, Tr. dongganensis, and Tr. elongatus), lateral line being incomplete (vs. complete in Tr. longibarbatus, Tr. maolanensis, and Tr. daqikongensis, or absent in Tr. barbatus and Tr. huanjiangensis), dorsal fin having 8−9 branched rays (vs. 10−11 in Tr. microphthalmus and 6 in Tr. maolanensis), anal fin having 5−6 branched rays (vs. 4 in Tr. maolanensis or 6−7 in Tr. huanjiangensis, Tr. longibarbatus, Tr. daqikongensis, Tr. dongganensis, Tr. elongatus, Tr. donglanensis, and Tr. microphthalmus), and upper adipose keel height being mostly 1/2 of the caudal peduncle depth (vs. equal to the caudal peduncle depth in Tr. barbatus, Tr. huanjiangensis, Tr. longibarbatus, Tr. maolanensis, Tr. daqikongensis, Tr. dongganensis, Tr. elongatus, and Tr. microphthalmus).Species of Troglonectes are highly adapted to survive in cave habitats and are found only in limited regions with relatively small populations. Ma et al. [13] mentioned that cave fish have morphological adaptations to extreme cave environments, including the degeneration or disappearance of the eyes, reduced pigment, and scales. Additionally, cave fish have specialized features including well-developed tentacles and prolonged pectoral fins. Species of Troglonectes have developed barbels, well-developed adipose keel on the upper and lower caudal peduncles, reduced or no eyes, lateral line, scales, and pigment; these characteristics are adaptions to cave environments. As their life histories are limited to caves, these fish are vulnerable to various threats, such as habitat degradation, hydrological alterations, environmental pollution, resource overexploitation, and non-native species introduction [13]. Karst caves and subterranean streams are common geological features in Guangxi. More than 300 freshwater fish species have been recorded in Guangxi, including 61 cavefish [11]. On 16 September 2022, the Department of Forestry of the Guangxi Zhuang Autonomous Region published a list of wildlife under key protection in Guangxi, which included all cavefish species. The new species is currently only known from the type locality, where a few specimens were collected when the water rose from the cave during the rainy season. The discovery of this previously unknown species can hopefully lead to conservation measures to protect this area.5. ConclusionsOne new species of Troglonectes is described herein based on its morphological characteristics and molecular analysis. Additionally, the phylogenetic tree indicates species of Troglonectes divided into two sub-clades, viz. the truncated caudal fin sub-clade and emarginated or forked caudal fin sub-clade.6. Nomenclatural Acts RegistrationThis published work and the nomenclatural acts it contains have been registered in ZooBank LSIDs (Life Science Identifiers) and can be resolved, and the associated information can be viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/ (accessed on 10 February 2023).Publication LSID:urn:lsid:zoobank.org:pub:5DB60B6B-94EC-4E18-B734-9258F9E31D2A.Troglonectes canlinensis LSID:urn:lsid:zoobank.org:act:1FEADE1C-2EBF-4956-A052-F6926AB9124C.

animals : an open access journal from mdpi

10.3390/ani11051232

PMC8146394

Early life and gestational experience influence the behavioural development of the offspring. This study investigated the relationship between gestating gilts’ fear of humans and cortisol levels and their feeding and maternal behaviour, and the personality (coping style, human fear) and growth of their piglets. Gilts were classified as fearful or friendly after four human approach tests performed between d 104 and d 111 of gestation, cortisol level was assessed between d 90 and d 108 of gestation, and maternal behaviour evaluated at d 13 of lactation. Piglets were submitted to a back test at 13 days old, and to a human approach test and an open field test at 20 days old. Fearful gilts had higher cortisol levels than friendly gilts. Piglets from friendly gilts tended to have a more active response to the back test, less freezing reaction in the open field test, and accepted human contact more than piglets from fearful gilts. The results of this study support the hypothesis that the fearfulness of gilts towards humans is related to their stress levels, and that both could influence the behavioural profile of their offspring.

Gestational and early life experiences affect subsequent behavioural and physical development. The objective of the current study was to investigate associations between gilts’ fear of humans, gestational stress level, and feeding and maternal behaviour, as well as how these related to aspects of the personality and growth of their offspring. A total of 37 gilts were used. Four human approach tests were performed between d 104 and d 111 of gestation to classify gilts as fearful or friendly. Gilt feeding behaviour and salivary cortisol concentration was measured between d 90 and d 108 of gestation, and gilt nursing behaviour assessed at d 13 of lactation. Piglets were subject to a back test at d 13 of age, to an open field test and a human approach test at d 20 of age, and growth was monitored to weaning (d 26 of age). Gilts classified as having a fearful behavioural profile had higher cortisol levels than friendly gilts (p < 0.05). Human fear level did not affect reproductive performance or the growth of offspring (p > 0.05). The offspring of friendly gilts tended to have a more active response to the back test (p = 0.09), less freezing response in the open field test (p < 0.05), and received human contact more than piglets from fearful gilts (p < 0.05). The present study shows that gilt human fear level is linked to their stress levels, which can affect the personality of their piglets.

1. IntroductionGestational and early life experiences influence subsequent behavioural and physical development. Historically, most studies on stressors that affect individuals prior to birth (i.e., pre-natal stress; PNS) have been conducted on non-human primates and rodents, and have demonstrated that offspring from chronically stressed mothers have impaired stress coping ability and locomotor and cognitive development (reviewed in [1]). An example of this in pigs is in relation to maternal behaviour; mothers that are stressed during pregnancy produce daughters that subsequently have poor maternal behaviour (piglet-directed aggression, less motivation to nurse) [2,3]. In most studies to date, PNS has been imposed by carrying out procedures likely to reflect farm practices: from daily physical restraint of the sow [4] to mixing of sows [5,6], to investigate whether maternal stressors, which are reasonably regular occurrences on farm, can also affect the offspring. However other causes of PNS, and their specific effects on the offspring, have not yet been fully investigated in pigs.Understanding the causes of PNS and how its intensity affects piglet development will help to improve both animal welfare and performance. Unlike in rodents [7], PNS does not appear to affect the weight of piglets at birth [8]. Interestingly, Brajon et al., (2017) [9] found that prenatally stressed piglets showed signs of behavioural inhibition post-weaning (i.e., less environmental exploration and less play behaviour), even though they did not differ pre-weaning. In sheep, PNS, which was mediated by the emotional reactivity of the mother, also altered the behaviour of lambs in both a human approach and a novel object test [10]. Personality is another non-environmental factor that seems to be related to an individual’s physiological and behavioural responses to stress.One aspect of pig personality is coping style, determined by genetics and early life experience [11]. In pigs, coping style has been most often evaluated in early life using a back test, which assesses the reaction of the pig to restraint (e.g., [12,13,14,15]). Highly responsive pigs, displaying many escape attempts during the test, are considered to have a proactive (or active) coping style, and lowly responsive pigs are considered to have a reactive (or passive) coping style. In stressful situations, proactive pigs are more likely to show an aggressive and less flexible/adaptable behavioural response, and their physiological response (e.g., higher heart rate) is mediated by their sympathetic nervous system. Fear of humans or of novelty are other aspects of pig personality that can affect their productivity (e.g., reproduction, growth rate; [16]) and welfare [17]. Sows that are fearful of humans are more likely to savage their piglets and have stillborn piglets [18], while positive human handling (to promote reduced fear) resulted in shorter farrowing and more rest following farrowing [19]. Human fear level is maintained between parities [18] and although it seems to have a low heritability (e.g., [20,21]), it can be learned by piglets through emotional contagion and social learning [22]. There is a global lack of knowledge on the factors influencing the transmission of human fear from the dam to the piglets. Filling this gap could help improve the human–animal relationship on farm, and therefore promote both animal welfare and productivity.This study investigated associations between gilts’ fear of humans, gestational stress level, and feeding and maternal behaviour, as well as how these related to aspects of the personality (coping style, fear responses) and growth of their piglets. We hypothesised that sows that are more fearful of humans experience more stress, have poor maternal skills, and that their offspring would also show greater fear of humans, lower growth, and a more proactive coping style.2. Materials and MethodsThe experiment was carried out between June 2016 and March 2017 at the Teagasc Pig Development Department, Moorepark, Fermoy, Co., Cork, Ireland. Ethical approval for this study was granted by the Teagasc Animal Ethics Committee (approval no. TAEC120/2016) and the project was authorized by the Health Products Regulatory Authority (project authorization no. AE19132/P051). The experiment was conducted in accordance with Irish legislation (SI no. 543/2012) and the EU Directive 2010/63/EU for animal experimentation.2.1. Animals, Housing and FeedingThirty-seven gilts with the same genetic background (Large White x Landrace; Hermitage Genetics, Sion Road, Co., Kilkenny, Ireland) were used in the study. Gilts selected from two breeding batches were artificially inseminated at the onset of standing estrus and again 24 h later using pooled semen (Danish Duroc; Hermitage Genetics, Co., Kilkenny, Ireland). Batches were inseminated at 3-week intervals, with 25 and 12 gilts per batch. During gestation, gilts were managed in a large dynamic group pen which held 120 breeding animals at any one time. The group pen had insulated concrete lying bays and fully slatted floors. Gestating gilts were fed via two electronic sow feeders (ESFs; Schauer Feeding System (Competent 6), Prambachkirchen, Austria) and gilts had ad libitum access to water from single-bite drinkers in the ESFs and from five drinker bowls located around the pen. On d 38 of gestation, gilts were blocked within batch on the basis of body weight (BW; mean ± S.D.; 179.4 ± 9.87 kg) and back-fat depth (BF; 16.9 ± 3.42 mm) and randomly allocated to 1 of 4 dietary treatments until parturition: (1) control (n = 8), (2) L-carnitine (0.125 g/d L-carnitine; n = 9), (3) sugar beet pulp (40% sugar beet pulp; n = 10), and (4) sugar beet pulp plus L-carnitine (40% sugar beet pulp + 0.125 g/d L-carnitine; n = 10). This study was a component of a larger study by Rooney et al., (2019) [23], in which the effects of the four different diets were evaluated. For further information on both the dietary treatments and the composition of the experimental diets, please see [23]. Gilts were then moved within their farrowing group to a smaller pen at d 90 of gestation and the pen had the same layout and facilities as the larger group pen. Six days before gilts were due to farrow (d 108 of gestation), they were moved into standard farrowing crates (pen dimensions: 2.5 m × 1.8 m) and farrowing rooms accommodated 7 or 14 animals per room. The health and welfare status of all gilts and their offspring was monitored daily by farm personnel.Once farrowed, gilts received a standard lactation diet twice daily for the first 6 days of lactation and three times daily thereafter until weaning. Water was provided to gilts from a single-bite drinker in the feed trough and suckling piglets had access to water from a bowl in the farrowing pen. Suckling piglets received creep feed twice daily from d 13 of lactation. The temperature in the farrowing room was maintained at ~24 °C at farrowing and gradually reduced to 21 °C by d 7 of lactation. Artificial lighting was provided from 08:00 h to 16:30 h each day. Where possible, litter size was standardized during the first 48 h after parturition, based on piglet BW, so that there was an average litter size of 13.4 ± 0.40 piglets per gilt. Cross-fostering was only done within gilt treatment and piglets that had been fostered were excluded from further investigations. Therefore, when a gilt is described as the ‘mother’ of a piglet in this study, it refers to the biological mother. Piglets’ teeth were clipped within 24 h postpartum and tails were docked on d 3 postpartum. All piglets received an iron injection on d 5 postpartum and males remained fully intact. Pigs were weaned on d 26 ± 0.1 of lactation.2.2. Categorisation of the Gilts2.2.1. Human Approach TestA human approach test (HAT) was performed four times, between d 104 and d 111 of gestation, to assess differences between gilts in their reaction to a human. The HAT was performed when gilts were unrestrained in the pen. The experimenter calmly entered the pen and quietly walked among the gilts for 2–3 min before commencing the test. The test was carried out in a randomised order to control for the order of testing, and the effects on each gilt was tested individually using the scoring system outlined in Table 1 and adapted from [24]. If a gilt voluntarily approached the experimenter, the HAT commenced for that animal and they were assigned 0 for ‘approach’. Otherwise, the experimenter walked slowly towards the gilt from the front. After scoring the gilt’s response to an approach, the experimenter then attempted contact, by reaching out and attempting to touch the gilts neck, and again scoring the reaction. Finally, the type of vocalization, if any, was scored.2.2.2. Profile AssignmentGilts were assigned to one of three response profiles each time the HAT was performed: ‘friendly’, ‘fearful’, or ‘unclassified’ (Table 2). Two datasets were then created. The first included only gilts that were categorised as ‘friendly’ or ‘fearful’ in every test (PURE gilts). The second included all gilts. Gilts were categorised as ‘friendly’ or ‘fearful’ if they fit that profile in three out of four HATs, and as ‘unclassified’ if their profile could not be established due to inconsistent scores between tests. In total, 20 gilts were considered PURE: 7 friendly and 13 fearful. When considering all 37 gilts, 15 were classified as friendly, 17 as fearful, and 5 were unclassified.2.3. Gilt Measures2.3.1. Live-Weight, Back-Fat Depth, and Farrowing PerformanceThe back-fat depth and live weight of gilts were recorded three times during gestation, at d 38 (blocking), d 90 (move to smaller pen), and d 108 (move to farrowing crates) of gestation, as well as at weaning according to the methods previously described by [23]. The number of piglets born (total, live, and stillborn) was recorded for each litter at birth.2.3.2. Gestation Feeding BehaviourBetween d 90 and d 108 of gestation (i.e., when in the smaller pen), interactions with the ESF were monitored to determine whether gilt profile influenced feeding behaviour. Each ESF day commenced at 19:00 and concluded at 18:00. Thus, there was an hour during which the ESF was not accessible to the gilts, to enable routine maintenance, etc. The time of each gilts’ first visit after the ESF opened at 19:00 was automatically recorded and downloaded, and from this, the order in which each gilt entered the ESF determined. As there were different numbers of gilts in each batch, the order of entry for each was divided by the total number of gilts in the pen. This value represents the proportion of gilts in the pen that entered the ESF prior to each gilt on each day. The coefficient of variation of the order of entry over all days for each gilt was calculated to provide an estimate of the level of stability of entry order over time. Finally, the total number of ESF visits for each gilt was summed on each day.2.3.3. Salivary CortisolThree saliva samples were collected from gilts and analysed according to the methods previously described by [23]. In brief, samples were collected once every week between d 90 and d 108 of gestation and all samples were collected between 09:00 h and 10:00 h each morning, roughly 9 h after the gilts’ last meal. To obtain the saliva samples, gilts were allowed to chew on a large cotton swab for 30–40 s until it was saturated (Salivette, Sarstedt, Co., Wexford, Ireland). Once collected, the cotton swabs were placed into plastic Eppendorf tubes and centrifuged (400× g at room temperature for 10 min), before being stored at −20 °C until analysis. Salivary cortisol concentration was assessed in duplicate using an enzyme linked immunosorbent assay (Salivary Cortisol kit, Salimetrics Europe Ltd., Suffolk, UK). The minimum detectable concentration of cortisol that could be distinguished from 0 was < 0.003 μg/dL. The intra- and inter-assay CVs were 4.5% and 4.2%, respectively.2.3.4. Nursing BehaviourTo measure the willingness of gilts to nurse their litters, suckling piglets were separated from their mother for 2 h on d 13 post-farrowing. After the 2 h separation period, the gilt was encouraged to stand if she was not already doing so, then her litter of piglets was returned to the farrowing pen. A stopwatch was started when all piglets had been returned to the pen and the time that it took gilts to kneel/lie down and nurse her piglets was recorded. The observation time for each gilt was no longer than 5 min in duration.2.4. Piglet Measures2.4.1. GrowthThe weight and sex of each piglet was recorded at birth, and each piglet was tagged for identification purposes. Thereafter, piglets were individually weighed on d 1, d 6, and d 13 after farrowing, as well at weaning, and the data were used to determine piglet average daily gain (ADG).2.4.2. BehaviourBack Test. On d 13 post-farrowing, each piglet was subjected to a back test as described by [25]. Each piglet was placed on its back on a wooden v-board and restrained for 1 min. The tester held the hind legs with one hand and placed the other gently on the throat (Figure 1). The tester did not change the piglets’ position or move their hands during the 1-min experiment. The piglet was held in this position for 1 min and the number of escape attempts (leg kicks, wriggles) made during this period counted. Each series of wriggles and kicks that was made without pausing was classified as a single escape attempt. The total number of escape attempts was considered the back test score.Open Field Test. One week after the back-test, on day 20 post farrowing, four piglets were selected for an open field test (OFT) from each of the gilts classified as friendly and fearful (i.e., not including unclassified gilts; n = 32 gilts). The piglets selected were those that represented the average BW of the litter and consisted of two males and two females. The back test results were also used in the selection of piglets; piglets were considered a low responder (LR) if the number of escape attempts was less than or equal to 2 and a high responder (HR) if the number was greater than 2 [11,26]. The proportion of high and low responders in the litter was then used to calculate the number of high and low responders to be selected for the OFT (Table 3).The OFT arena was an empty and disinfected farrowing pen (pen dimensions: 2.5 m × 1.8 m) in a room that was visually and acoustically isolated from the piglets’ home pen. The room that contained the OFT arena was kept at a similar temperature as the other farrowing rooms that housed experimental gilts. Piglets were placed together in a trolley and calmly transferred to the test arena. One at a time, each piglet was then taken out of the trolley and placed in a corner of the pen. As soon the piglet was released, the OFT began. Piglet behaviour (Table 4) was continuously recorded for 3 min by a single observer standing outside the test arena, using a Psion Workabout installed with the software package The Observer® XT (Noldus Information Technology, Wageningen, The Netherlands).2.4.3. Human Approach TestsAfter the OFT test was complete, each piglet was subjected to a HAT in the same test arena. The tester calmly entered the pen, walked to the farthest side, and sat on the floor cross-legged for 1 min. The tester did not move or interact with the piglets during the test, and each piglets’ behaviour was scored based on the following scale: 0: the piglet does not touch the tester during the 1 min test; 1: the piglet touches the tester during the 1 min test. The length of time that it took the piglet to touch the tester was recorded. If a piglet touched the tester during the 1 min test, a forced HAT was performed. The tester initiated contact by slowly moving their hand to the top of the piglet’s head and gently touching the piglet. The piglet’s behaviour in response to being touched was scored based on the following scale: 0:the piglet flees from the contact; 1:the tester touches the piglet but the piglet withdraws after contact; 2: the tester touches the piglet and the piglet withdraws after contact but returns to the tester in less than 10 s; 3: the piglet does not withdraw after contact.2.5. Statistical AnalysisStatistical analyses were performed using SAS 9.4 (SAS Inst. Inc., Cary, CA, USA). The experimental unit for analysis was the gilt. Analyses were carried out on two datasets. The first dataset included gilts that were consistently categorized as friendly (n = 7) or fearful (n = 13) in all four of the HATs (PURE gilts). The second dataset included all gilts; gilts were categorized as being friendly or fearful if this was how they responded in at least three out of the four HATs. If gilts were friendly and fearful in two each of the tests, they were considered unclassified. This second dataset was analysed firstly to increase the sample size (friendly = 15; unclassified = 5; fearful = 17), and secondly, to determine whether results would be similar if gilts that did not respond completely consistently were included. When using this second dataset, we were able to compare the fearful gilts to all others (friendly and unclassified combined) or to only friendly gilts.Data distribution and the presence of outliers were initially evaluated by the examination of histograms and normal distribution plots (PROC UNIVARIATE). General linear mixed models (PROC MIXED) were used for most of the analysis. Degrees of freedom were estimated using the Kenwood–Rogers adjustment, and residuals were examined to verify normality and the homogeneity of variances. In cases where repeated measures were used, model fit was determined by choosing models with the minimum finite-sample corrected Akaike information criteria. The Tukey–Kramer adjustment was used for multiple comparisons where least squares means (LS means) were determined (i.e., when we used the second dataset in analysis, and compared all three gilt profiles to each other). Differences were considered statistically significant when alpha was ≤ 0.05, and tendencies were determined when alpha was between 0.05 and 0.1 (inclusive).All models included the main effects of sow profile (friendly, fearful, and unclassified), as well as the fixed effects of fibre level of the diet (high/low), L-carnitine supplementation (yes/no), and batch. The repeated statement was used where necessary, details of which are provided for specific models below. The gilt was considered the experimental unit in all analysis. Additional terms which were not relevant for all models (e.g., piglet sex, measurement specific covariates, etc.) are detailed in the corresponding section below. To investigate the overall hypothesis that both fearful gilts and their offspring would have responses different to all other gilts, contrast statements were used to investigate differences between fearful gilts and unclassified and friendly combined.When data did not conform to normality, a transformation was initially attempted (e.g., log transformation for the number of visits per day to the ESF). If unsuccessful, then non-parametric statistic was used. The number of piglets born dead as analysed using the Kruskall–Wallis test with a Dwass, Steel, Critchlow-Fligner procedure to protect against type 1 error.2.5.1. Sow MeasurementsCortisol measurements from each gilt were averaged over the three sampling days prior to analysis. For analysis of the ESF data, the repeated effect of day was included in the models where relevant.2.5.2. Piglet MeasurementsFor the analysis of piglet performance, the sex of the piglet was also included in the analysis. For birthweight, the number of piglets born was also considered a covariate. Piglet birthweight was also included as a covariate for analysis of growth to weaning. Additional factors included in the model for the analysis of the piglet back test were the fixed effects of sex and the person holding the piglets. Birthweight was included as a covariate.Mixed models were used to analyse the number of low grunts and the duration of standing, walking, and exploration. Aside from the fixed effects included in all models, additional effects included the fixed effect of sex and back-test score.For analysis of the number of elimination events, screams, and jumps, and the duration of freezing and running, there were multiple 0 values, and as such the data could not be normalised. Thus, data were analysed using the Wilcoxon rank test.The number of piglets that made voluntary contact from gilts of each profile (friendly vs. fearful) was compared using a Chi-square test. Further analysis could only be carried out on piglets that made contact (n = 94). The time it took to touch the researcher was analysed using the same mixed model as for the OFT variables, but log transformed for analysis so that the residuals approached a normal distribution. Once p-values were determined, the appropriate model was run using raw data to generate least squares means. The response to the forced human contact test was analysed using the Wilcoxon rank test.3. Results3.1. Sow Measurements3.1.1. Back-Fat and WeightOf the PURE gilts, friendly gilts had greater back-fat than fearful (16.8 ± 0.4 vs. 15.3 ± 0.3; p = 0.01). However, there was no interaction between recording day and profile, which indicates that friendly gilts simply maintained a back-fat advantage which they had at the beginning of the experiment. When all gilts were included in the analysis, there was no difference in back-fat thickness across profiles (friendly = 16.0 ± 0.4, unclassified = 15.0 ± 0.7, fearful = 15.4 ± 0.3). Live-weight did not differ across the profiles, whether only PURE or all gilts were included in the analysis.3.1.2. CortisolThere was no difference in cortisol level between the PURE friendly or fearful gilts (p = 0.15; Figure 2A). However, when all gilts were included in the analysis, there tended to be an effect of gilt profile (p = 0.1), with fearful gilts having higher cortisol levels than the other categories (p < 0.05; Figure 2B).3.1.3. Feeding BehaviourFor the PURE gilts, there was no effect of profile on the order in which the gilts entered the ESF (fearful = 54 ± 5% vs. friendly = 56 ± 6%; % represents the proportion gilts in the group that entered the ESF prior; p = 0.77). When all gilts were included in the analysis, the effect of profile was significant (p < 0.01). However, this was driven by the unclassified gilts, which on average entered the ESF much later (75 ± 8% entered prior to them) than either the friendly (45 ± 4%; p < 0.01) or fearful (56 ± 4%; p = 0.07). There was no difference between the friendly or fearful (p = 0.13).For the PURE gilts, there was no effect of profile on the variation in the order in which the gilts entered the ESF over the experimental period (p = 0.34; Figure 3A). When all gilts were included in the analysis, however, the effect of profile became significant (p < 0.05), with friendly gilts displaying more variation than fearful gilts (p < 0.05; Figure 3B).Of the PURE gilts, friendly gilts entered the ESF more times per day than the fearful (p < 0.01; Figure 3C), and this pattern was somewhat replicated when all gilts were included in the analysis (Figure 3D), as there tended to be an effect of profile (p = 0.1). However, in this scenario there was no significant difference between friendly and fearful gilts (p = 0.17).3.1.4. Reproductive PerformanceThere were no effects of gilt profile on reproductive performance (Table 5).3.1.5. Nursing BehaviourThere was no difference in time to nurse piglets after a 2 h separation, when considering only PURE gilts. However, when including all gilts in the analysis, although there was no overall difference in time to nurse, fearful gilts tended to take longer to nurse than all others (p = 0.1; Figure 4).3.2. Piglet Measurements3.2.1. Piglet PerformanceThere was no effect of gilt profile on piglet birthweight, whether only PURE or all gilts were included. Across all gilts, piglets weighed approximately 1.32 ± 0.34 kg (mean ± std. dev.) at birth. When all piglets were included in the analysis, there was no effect of gilt profile on the piglets’ growth to weaning or weaning weight. However, when only piglets from the PURE gilts were included, piglets from fearful gilts tended to be heavier than the friendly (p = 0.05), and there also tended to be an interaction between profile type and weighing data (p = 0.07). Indeed, on day 13 after birth, piglets from the fearful gilts tended to be heavier than those from the friendly ones (p = 0.06; Figure 5). However, weaning weight did not differ for piglets from sows of divergent profiles. All piglets were weaned at approximately 7.02 ± 1.53 kg.3.2.2. Back-TestWhen piglets from all gilts were included in the analysis, there was no effect of sow profile on back-test scores. However, when only considering PURE gilts, piglets from friendly gilts tended to have a higher back-test score than the fearful (3.47 ± 0.21 vs. 3.01 ± 0.15; p = 0.09).3.2.3. Open-Field TestThere was no effect of gilt profile, whether including all gilts or only the PURE ones, on the duration of standing, walking, exploring, or running. However, when all gilts were included, piglets from friendly gilts spent less time in a freeze position (0 (0–3.53) seconds, median (interquartile range)) than fearful ones (2.61 (0–9.42) seconds; p = 0.01). Likewise, when only considering PURE gilts, piglets from friendly gilts tended to spend less time frozen (0 (0–4.11) seconds) than piglets from fearful gilts (2.51 (0–8.89) seconds; p = 0.06). There were no differences in the incidence of low- or high-pitched vocalisations, eliminations, or jumping.3.2.4. Human Approach TestsAt least one piglet from every sow voluntarily made contact with the tester. When all gilts were included, piglets from friendly gilts tended to be more likely to make contact with the tester than piglets from fearful gilts (n = 127 piglets; 81.4% vs. 67.6%; p = 0.1). However, this was not the case when only PURE gilts were considered, even though the proportions were similar (n = 79 piglets; 81.5% vs. 69.2%; p = 0.29).When data from all gilts were analysed, piglets from friendly gilts took less time numerically to touch the tester than the fearful (17.1 ± 2.9 vs. 19.3 ± 2.7 s; p = 0.15), but this was not significant. When data from only PURE gilts were analysed, there was a tendency for piglets from friendly gilts to take less time, even given the smaller sample size (friendly = 12.8 ± 2.9 vs. fearful = 16.8 ± 2.3 s; p = 0.08).For the forced human contact test, scores of piglets from friendly gilts were higher (i.e., indicative of less fear) than those from fearful gilts (2 (1–3) vs. 1.5 (0–3); p = 0.03). When only considering PURE gilts, the pattern was similar (friendly = 2 (2–3) vs. fearful = 1.5 (0–3); p = 0.04).4. DiscussionIn this study, we investigated the effects of gestating gilts’ reaction to a human and the associated stress level on their feeding and maternal behaviour and on the personality (coping style, human fear) and growth of their piglets. Gilts classified as having a fearful response (determined by HAT scores) had higher basal cortisol levels during late gestation compared with friendly gilts. Contrary to expectations, human fear level did not affect the growth of offspring from birth to weaning. However, a strong association between prenatal stress and the personality or coping behaviour of gilt offspring was observed in the behaviour tests applied to piglets, whereby behaviour responses that indicate an increased level of fearfulness were observed in offspring of fearful gilts.Fear of humans is often considered an important indicator of farm animal welfare, as it is associated with physiological stress and can influence maternal performance [19,29]. Cortisol concentration in saliva is often used to assess the level of physiological stress the animal is experiencing, as it correlates well with circulating levels in blood [30,31]. As such, salivary cortisol concentration has often been used as a measurement to give insight into the effects of PNS and negative handling, and on types of human–animal interaction. Jarvis et al., (2006) [2] demonstrated that sows exposed to an environmental stressor during gestation (stimulated PNS via social mixing) had increased salivary cortisol levels; while several studies have shown that animals that experienced unpleasant handling have higher cortisol concentrations [32,33,34]. In the present study, the higher basal cortisol levels of gilts that were classified as being fearful rather than friendly during late gestation (between day 90 and day 108) suggests that differences in HAT scores may be reflected in physiological responses. This finding agrees with those of a recent study, whereby pigs that were more fearful (based on results from a novel object test) had significantly higher levels of cortisol at slaughter [35]. There are potential limits when interpreting cortisol data, as the average concentrations of cortisol in saliva are influenced by several factors such as age (concentrations decrease with age), sex (higher concentrations in males than in females; [36]), and time of the day (levels peak in the early morning hours and are lowest in the evening and at night) [37]. However, we controlled for these factors by collecting saliva samples from all gilts in each group on the same day and within an hour (usually within 15 min) for all subjects.In addition to the greater variation in entry order, friendly gilts also had a greater number of visits to the ESF per day than fearful gilts. A greater number of entries into the feeding stations could mean that friendly gilts were more ‘optimistic’ that they would be fed each time they entered the ESF. This hypothesis can be compared to judgement bias tests, an alternative measure for evaluating psychological welfare in animals [38]. The theory behind this test suggests that an animal will evaluate a particular stimulus as predicting either a positive or negative outcome, depending on the animal’s affective state [38,39,40]. As such, friendly gilts could have associated each visit to the ESF as having a positive outcome (i.e., receiving feed upon entry to the ESF). If the feeding behaviour of sows and gilts that were housed in large loose housing and fed via ESF stations was studied consistently, the data could then be harnessed to assist with the identification of animals needing specific attention, similar to the use of automatic milking robots in dairy systems [41], or utilised as a non-invasive measure of optimism/pessimism within a breeding herd.Although differences in HAT scores were reflected in cortisol levels during pregnancy, reproductive performance (numbers total born, born alive, and stillborn) did not differ between gilts with different behaviour profiles. Indeed, average cortisol levels across gilts were lower than those recorded in gestating gilts in the same research facility [42], and lower than the levels of the larger pool of gilts from which the current study animals were selected [23]. Thus, there is no indication that fearful gilts were excessively stressed relative to the other profiles, rather just that they had higher basal cortisol levels, which is possibly why foetal growth was not affected. Where foetal growth was previously shown to be affected by increased maternal cortisol levels, it was generally where substances were administered to initiate a state of physiological stress artificially; oral administration of hydrocortisone acetate to sows during early and late gestation resulted in decreased piglet birthweights [43], and the birthweight of piglets born to sows treated with injections of adrenocorticotrophic hormone during the last week of pregnancy was lower than piglets from non-treated sows [44]. Our findings are consistent with previous work on pigs (see review by [45]), whereby stressful conditions during gestation had no effect on the weight of sow progeny at birth. The aforementioned results suggest that the growth of offspring in utero can be influenced by artificially increased maternal cortisol levels during gestation (i.e., glucocorticoid models) but that the experience of stressful situations (e.g., those imposed by management) by the pregnant sow have little or no effect on foetal growth. However, it must be noted that the sample size used in the present study was likely too small to detect significant differences in the reproductive performance of gilts.We used a simple test of nursing behaviour as an indicator of maternal behaviour. Our original hypothesis that gilts that are more fearful of humans have lower maternal skills was not entirely supported, however, as there was little difference in the response to piglets after a separation. Nevertheless, the small sample size in the current study may have precluded the ability to detect a difference, as the numerical pattern was indicative of a longer latency to lie in fearful gilts. There is other evidence in the literature that fearful behaviour during gestation translates to poor maternal ability; Marchant Forde et al., (2002) [18] reported that a fearful behavioural profile during gestation was associated with increased savaging in gilts and Rutherford et al., (2014) [3] observed more abnormal maternal behaviour (e.g., less time spent lying laterally and increased restlessness) in gestating sows that experienced the social stress of mixing, relative to those in stable groups. Janczak et al., (2003) [29] also found that sows showing less fear of humans had more adaptive maternal behaviour and concluded that fear of humans is negatively associated with maternal ability. Future work could aim to validate our simple test, taking into account sample size and precise profiling of sows.The coping style of an individual pig is a reflection of its preferred strategy for reacting to stressors [46]. In the present study, piglets from friendly gilts tended to have a higher back-test score (i.e., they made more escape attempts during the back test and thus were classified as high-responsive piglets) than piglets from fearful gilts. Thus, they adopted a more active coping style in response to the stress of being restrained than piglets from fearful gilts, which displayed a more passive coping style. This result suggests that the coping style of offspring in response to stress or fear might be somewhat influenced by the fearfulness, or stress levels during gestation, of the dam. However, to obtain more robust data on this relationship, more behaviour testing incorporating a wider range of test types that contribute to personality profiling for the mother is advised for future studies [14]. It is not possible to elucidate from our data any causal relationship between maternal fear, maternal cortisol levels, and piglet responses.Nevertheless, we did incorporate into the present study two other behaviour tests for piglets in addition to the back-test. The HAT and the open field/novel environment test allow for the assessment of fear of both humans and novelty and are complementary to the back test. For example, Kooij et al. (2002) [47] demonstrated a correlation between the back-test score of pre-weaned piglets and the behaviour of piglets in a HAT that was conducted after weaning (5–7 weeks of age). In the aforementioned study, proactive piglets were more likely to approach a human faster than a reactive pig during the HAT. Likewise, piglets from friendly gilts in the current study were considered to have a proactive coping style according to the back test, and also tended to be more likely to make contact with the tester during the HAT and be more likely to permit the human experimenter to make contact with them. Therefore, our findings suggest a relationship between the level of fear expressed by gilts, and their offspring, towards humans.There also appeared to be a clear relationship between the response to the back test and the behaviour of piglets in another stressful situation, the open field test; piglets from friendly gilts, which were generally proactive, spent less time in the freeze position during the open field test than piglets from fearful gilts (generally reactive). Again, this is consistent with the results of former studies [12,15,26]. For example, Zebunke et al., (2017) [15] demonstrated that piglets that were classified as high responders showed earlier contact with an unknown human in a human approach test. Furthermore, the high responders exhibited earlier and more frequent but shorter locomotion and standing episodes, longer contact with a novel object, and shorter latency until the first escape attempt in an open field test. Our results are also in agreement with the description of Koolhaas et al., (1999) [48] that proactive animals engage in an active response, also known as ‘fight-flight’, to challenging situations whereas reactive animals engage in a conservation-withdrawal response, also known as ‘freeze’.It is beyond the scope of this study to consider whether the responses of the piglets to the various behaviour tests and their associations with maternal profile were due to genetic, epi-genetic, or environmental effects (e.g., exposure to cortisol, learning from the mother). However, it is possible that it was a combination of all three; as well as the documented effects of exposure to cortisol (listed above), there is evidence that personality type in animals is somewhat heritable. For instance, Dochtermann et al., (2015) [49] found that up to 52% of variation in animal personality could be due to additive genetic variation. Our study does indicate however that associations are present and as such, it is a useful addition to the literature from which future hypotheses can be developed.Finally, our original hypothesis that PNS during pregnancy may be negatively associated with the subsequent growth of offspring was not supported by our results as the overall growth rate of piglets from birth to weaning was unaffected by gilt behavioural profile. Piglets from gilts classified as consistently fearful (i.e., PURE fearful gilts) tended to be heavier than piglets from consistently friendly gilts at 13 days of age, but this did not translate into heavier piglets at weaning.5. ConclusionsGiven the pressure for efficiency in pig farming from both an economic and overall sustainability point of view, transmission of both fearful behaviours and poor maternal skills should be avoided. Our results demonstrate a clear relationship between gilt behaviour and physiology during pregnancy and maternal behaviour during lactation. In turn, the offspring of mothers fearful of humans had a behaviour profile indicating that they were also fearful. We found that the patterns of the results were very similar whether we included gilts which were completely or only partially consistent in their response to a human approach test, which is useful information for the planning of future research where sample sizes may be limited. Further work could build on our results and attempt to validate quick-to-use and easily applied methods to identify pigs that may be more fearful and need special attention/care (e.g., data automatically collected by the ESF). This is particularly the case for replacement gilts, as fear of humans could result in impaired welfare and performance not only for themselves, but also for their offspring.

animals : an open access journal from mdpi

10.3390/ani11123495

PMC8698070

Fasciolosis, caused by the worm parasite Fasciola hepatica (liver fluke), is a global disease of farm animals and a neglected disease of humans. Infection arises from the ingestion of resistant metacercariae that contaminate vegetation. Within the intestine, the parasite excysts as an active larvae, the newly excysted juvenile (NEJ), that borrows through the intestinal wall to infect the host and migrates to the liver. NEJ release, tissue penetration and migration are facilitated by enzymes secreted by the parasite, namely, cathepsin B1 (FhCB1), cathepsin B2 (FhCB2), cathepsin B3 (FhCB3) and cathepsin L3 (FhCL3). While our knowledge of these enzymes is growing, we have yet to understand why the parasites require all four of them to invade the host. In this study, we produced functional recombinant forms of these enzymes and demonstrated that they vary greatly in terms of activity, optimal pH and substrate specificity, suggesting that, combined, these enzymes provide the parasite with an efficient digestion system for different host tissues and molecules. We also identified several compounds that inhibited the activity of these enzymes, but did not affect the ability of the larvae to excyst or survive. However, this does not exclude these enzymes as targets for development of drugs or vaccines.

Fasciolosis caused by Fasciola hepatica is a major global disease of livestock and an important neglected helminthiasis of humans. Infection arises when encysted metacercariae are ingested by the mammalian host. Within the intestine, the parasite excysts as a newly excysted juvenile (NEJ) that penetrates the intestinal wall and migrates to the liver. NEJ excystment and tissue penetration are facilitated by the secretion of cysteine peptidases, namely, cathepsin B1 (FhCB1), cathepsin B2 (FhCB2), cathepsin B3 (FhCB3) and cathepsin L3 (FhCL3). While our knowledge of these peptidases is growing, we have yet to understand why multiple enzymes are required for parasite invasion. Here, we produced functional recombinant forms of these four peptidases and compared their physio-biochemical characteristics. Our studies show great variation of their pH optima for activity, substrate specificity and inhibitory profile. Carboxy-dipeptidase activity was exhibited exclusively by FhCB1. Our studies suggest that, combined, these peptidases create a powerful hydrolytic cocktail capable of digesting the various host tissues, cells and macromolecules. Although we found several inhibitors of these enzymes, they did not show potent inhibition of metacercarial excystment or NEJ viability in vitro. However, this does not exclude these peptidases as targets for future drug or vaccine development.

1. IntroductionFasciolosis caused by the parasite Fasciola hepatica, or liver fluke, is a food- and water-borne disease of humans and their livestock. The parasite has a global distribution and results in major losses to the agricultural community, conservatively placed at EUR 2.5 billion each year, due to the reduced productivity of infected sheep, cattle, water buffalo and goats [1,2]. Infections of humans are estimated to be between 2.4 and 17 million, while >180 million people across 70 countries live at risk of infection [3]. Recently, the disability-adjusted life years (DALYs) for this disease has been estimated at 35,000 per annum [3,4] and it is now recognised as a major neglected helminth infection by the World Health Organisation [3].The parasitic flatworm infects its mammalian host following ingestion of the metacercariae life stage, which is found encysted on vegetation, such as grass or edible aquatic plants and floating in water. A newly excysted juvenile (NEJ) emerges from the cyst in the small intestine and quickly penetrates the wall of the intestinal epithelium and surrounding tissues to make its way to the liver [5]. Here, the parasite spends about two months tunnelling and feeding on the liver parenchymal tissue before moving into the bile ducts, where it establishes a chronic infection, becomes fecund and produces thousands of eggs each day. These eggs are carried with bile secretions to the intestine and then passed within the faeces into the external surrounding environment where they fully develop and hatch to release a miracidium that is infectious to aquatic snails such as Galba truncatula. After a series of asexual cloning and developmental steps within the snail, the parasites emerge as cercariae and then settle as encysted metacercariae to complete the cycle [6,7,8].Parasites that are resistant to frontline anti-liver fluke treatments, such as triclabendazole, have spread globally, causing concern for the control of the disease, particularly since there are no commercially available vaccines. For the development of new control treatments, detailed molecular biological studies are required to identify amenable targets in the parasite. Recently, much research has focused on the NEJ stage and its means of invasion, since it is this stage that initiates infection and leads to the more damaging later developmental stages. Proteomic and transcriptomic studies have revealed the identity of several peptidases that the NEJ secrete to facilitate infection by breaking down the molecules, cells and tissues of the intestinal wall [5,9,10,11]. These peptidases include four papain-like cysteine peptides, three F. hepatica cathepsins B (FhCBs), termed FhCB1, FhCB2 and FhCB3 [9,12,13,14,15], and a cathepsin L, termed FhCL3 [9,16,17].Structural studies of papain-like cysteine peptidases show that they share a common fold [18]. Mature enzymes are bi-lobed molecules consisting of two domains with an active site located in the cleft between them. Substrates bind to the peptidases via a series of active site subsites (S1, S2, S3, etc.) that determine the specificity of the enzyme. The proteolytic activity of cysteine peptidases is conferred by the presence of three active site residues, Cys-His-Asn, known as the catalytic triad, which are part of the S1 subsite [19]. The carbonyl group of the substrate P1 residue is located in an oxyanion hole next to the active site cysteine whose thiol group acts as a nucleophile in enzyme–substrate interactions, while the P1 side chain is orientated outwards toward the solvent. The backbone amides of the substrate residues at positions P2, P1 and P1′ form a network of hydrogen bonds with surface residues in the corresponding substrate subsites S2, S1 and S1′ [18,20,21,22,23,24].The cathepsins B and L are produced as zymogens that possess an N-terminal propeptide region that acts as an inhibitor of their cognate peptidase by acting as a ‘clamp’ to block the peptidase active site. Removal of the propeptide may occur either auto- or trans-catalytically, at acidic pH [25,26,27]. We have shown that the cathepsins B and L of F. hepatica NEJs are located in the gastrodermal cells lining the parasite gut from which they are secreted into the low pH environment of the gut lumen [9,28,29].To date, no comparative study of the secreted cathepsins B and cathepsins L from the infective F. hepatica NEJs has been published. Therefore, in the present study, we conducted a physio-biochemical characterisation of the endopeptidase activity of the recombinantly produced FhCB1, FhCB2, FhCB3 and FhCL3 to probe their specificity and identify differences and novelties that might help elucidate the roles of these enzymes in the invasiveness and pathogenicity of F. hepatica.2. Material and Methods2.1. Production, Purification and In Vitro Autocatalytic Processing of Recombinant FhCBs and FhCL3The recombinant proteins, FhCB1, FhCB2, FhCB3 and FhCL3, were produced in the methylotrophic yeast Pichia pastoris and isolated from the medium by Nickle-chelate affinity chromatography as described previously [27]. Autocatalytic activation of recombinant proenzymes to their mature forms was carried out in C-P activation buffer (0.1 M citrate-phosphate, 100 mM NaCl and 2 mM DTT, at pH 4.5) supplemented with dextran sulphate (10 μg/mL) at a concentration of 100 μg/mL at 37 °C for 2 h. Activated proteins were dialysed against phosphate-buffered saline (PBS; Sigma-Aldrich, St. Louis, MO, USA) and stored at −20 °C.The Schistosoma mansoni cathepsin B1 (SmCB1) was also produced in our laboratory using P. pastoris as described above. The human cathepsin B (HsCB) and cathepsin L1 (HsCL1) were produced in human embryonic kidney cells (Sigma-Aldrich).2.2. Assessment of Enzymatic Activity by Fluorogenic Substrate AssayRecombinant peptidases were assayed for enzymatic activity using a fluorogenic substrate assay. The reactions were carried out in 0.1 M C-P buffer (pH 7.0) with 1 mM DTT in a final volume of 200 μL, in a black 96-well plate. Each well contained activated enzyme (140 nM) and 20 μM fluorogenic substrate, which was added to start the reaction. All assays were run in triplicate at 37 °C for 1 h. Enzyme activity was monitored by rate of hydrolysis and subsequent release of the AMC group on the synthetic peptide substrates at an excitation wavelength of 370 nm and an emission wavelength of 460 nm, as relative fluorescence units (RFU), in a PolarStar Omega Spectrophotometer (BMG Labtech, Aylesbury, UK). All data were plotted and analysed using MARS Data Analysis Software (BMG Labtech, UK) and GraphPad Prism version 5.To determine pH optima for activity, fluorescence assays were performed in 0.1 M C-P buffer with 1 mM DTT at a range of pH of 3.5–8.0. The fluorogenic substrates used were Z-Phe-Arg-AMC (for FhCB1 and FhCB3) and Z-Gly-Pro-Arg-AMC (for FhCB2 and FhCL3).The substrate specificities of the recombinant F. hepatica peptidases were investigated using a panel of six fluorogenic peptide substrates (Bachem, St Helens, UK), Z-Phe-Arg-AMC, Z-Arg-Arg-AMC, Z-Leu-Arg-AMC, Z-Gly-Pro-Arg-AMC, Z-Pro-Arg-AMC and Z-Val-Ile-Arg-AMC in 0.1 M C-P buffer with 1 mM DTT, at the optimum pH for each peptidase.To determine the kinetic constants, Kcat, KM and Kcat/KM of the activated peptidases against the Z-Phe-Arg-AMC, Z-Gly-Pro-Arg-AMC and Z-Val-Ile-Arg-AMC substrates were used at concentrations of 2000–4 μM over a series of 10 dilutions using enzymes, buffer and assay conditions as described above. The kinetic constants, KM and Vmax, were calculated using the Michaelis–Menten equation (Equation (1)) in a non-linear regression analysis of initial reaction velocity against substrate concentration (built into GraphPad Prism 5). The value of KM can then be fitted to Equation (2) to determine the catalytic constant Kcat. (1)ν=d[P]dt=Vmax[S]KM+[S] where ν is the initial reaction velocity, [P] is the concentration of the formed product, [S] is the substrate concentration, Vmax is the maximum rate achieved by the system and KM is the Michaelis–Menten constant, which represents the concentration of substrate at which the reaction has a rate equal to half of Vmax. (2)ν=Kcat[Et]{}[S]KM+[S] where [Et] is the enzyme concentration and Kcat is the catalytic constant, representing the measure of substrate molecules converted to product by enzyme per second.2.3. Determination of Inhibitor Specificity against Recombinant FhCBs and FhCL3Commercially available inhibitors of cathepsin B (Ca074 and Ca074-OMe), cathepsin L (L, L1, LII, LIII and LIV), cathepsin K (KI, KII and KIII) and cathepsin S were purchased from Merck Millipore (Cork, Ireland; Supplementary Figure S2). The broad-spectrum papain peptidase inhibitor Z-Phe-Ala-FMK was purchased from Enzo Life Sciences (Exeter, UK). These were used to determine and compare the inhibition constant (Ki) against the recombinant FhCBs and FhCL3. All the assays were carried out in 0.1 M C-P buffer with 1 mM DTT and 0.01% Brij® L23, at the pH corresponding to the optimum for each enzyme in a total 200 µL reaction volume. The peptidases (140 nM) were pre-incubated with the inhibitor (1000-15 nM) at room temperature, for 30 min, before starting the reaction by the addition of the substrate that was optimal for each enzyme (FhCB1, Z-Phe-Arg-AMC; FhCB2 and FhCB3, Z-Val-Ile-Arg-AMC; FhCL3, Z-Gly-Pro-Arg-AMC). Substrate concentrations were equal to KM (shown in Table 1) and the apparent inhibition constant (Kiapp) values were calculated with GraphPad Prism 5, using the Morrison equation for tight binding inhibition (Equation (3)). For competitive inhibitors, such as those used in this study, Kiapp was fitted to a second equation (Equation (4)) from which Ki can be determined [30]. (3)vivo=1−([E])+[I]+Kiapp)−([E]+[I]+Kiapp)2−4[E][I] 2[E] (4)Kiapp=Ki(1+[S]KM)2.4. Exo-Carboxypeptidase Activity of Recombinant FhCBs and FhCL3The exo-carboxypeptidase activity of the FhCBs and FhCL3 was compared with the activity of the Schistosoma mansoni cathepsin B1 (SmCB1), human cathepsin B (HsCB) and cathepsin L (HsCL1) (positive controls). The exopeptidase activity was examined using two fluorescent substrates, Abz-Phe-Arg-Ala-Lys(Dnp)-OH and Abz-Phe-Arg-Ala-Lys(Dnp)-NH2, where Abz is 2-aminobenzoyl and Dnp is 2,4-dinitrophenyl (Biomatik, Cambridge, Canada). The peptidases were assayed using a similar protocol to that described above for measuring endopeptidase activity. The reactions were carried out in 0.1 M C-P buffer, pH 5.0, with a final substrate concentration of 20 μM. The enzymatic activity of the peptidases was monitored as RFU, in a PolarStar Omega Spectrophotometer, with readings taken at excitation and emission wavelengths of 355 nm and 460 nm, respectively.2.5. Effects of Inhibitors on the Excystment of F. hepatica Metacercariae and Viability of NEJF. hepatica metacercariae (Italian isolate; Ridgeway Research, Saint Briavels, UK) were prepared for excystment as previously described by Robinson et al. [10]. To verify the effect of the cathepsin peptidase inhibitors on the ability of the metacercariae to excyst, 50 metacercariae per well were placed into a 24-well plate and 1 mL of excystment medium (0.6% sodium bicarbonate, 0.45% sodium chloride, 0.4% sodium tauroglycocholate, 0.025 N HCl and 0.4% L-cysteine) supplemented with 100 μM of the inhibitor was added to the well. As controls, we used PBS and dimethylsulfoxide (DMSO, 1:100; Sigma-Aldrich) alone. The plate was then incubated at 37 °C with 5% CO2, for 3 h to allow metacercarial excystment. Using a light microscope (25× magnification), the number of excysted NEJ was determined.NEJ were washed three times in a culturing medium (RPMI-1640 medium supplemented with 30 mM HEPES, 0.1% glucose, 10% foetal bovine serum and 50 μg/mL of gentamicin) and their viability assessed by culturing the parasites in 1 mL of culture medium containing 100 μM of the inhibitor at 5% CO2 at 37 °C for 24 h. After incubation, the number of moribund/dead parasites were counted using a light microscope (25× magnification). The criteria used to assess the NEJ viability were loss of motility and peristaltic movement and any obvious damage to the tegument or internal structures, as previously standardized by [31]. The experiments were performed in triplicate. The average of the replicate experimental data was subjected to a two-way ANOVA comparing the values obtained with each inhibitor against the mean of the DMSO control group, with minimum 95% confidence intervals. All the statistical analyses were performed using GraphPad Prism 5.3. Results3.1. Production and Isolation of Recombinant F. hepatica FhCB1, FhCB2, FhCB3 and FhCL3Utilising the yeast P. pastoris system, the recombinant forms of FhCB1, FhCB2, FhCB3 and FhCL3 were produced as zymogens that could be purified to homogeneity using affinity chromatography at a yield from 5 to 15 milligrams per litre. The peptidases were resolved in SDS-PAGE, producing bands in the expected molecular weight of ~37 kDa (Figure 1A). All four proenzymes could be auto-catalytically activated to functional mature forms under acidic conditions, when incubated in C-P activation buffer with the addition of dextran sulphate at pH 4.5 (Figure 1B). The mature cathepsin peptidases were demonstrated to be active enzymes when assayed against fluorogenic peptide substrates (see below).3.2. pH Dependency of Mature F. hepatica FhCB1, FhCB2, FhCB3 and FhCL3To first establish the pH dependency of FhCB1, FhCB2, FhCB3 and FhCL3, their ability to cleave fluorogenic peptide substrates was determined over the pH range 3.5–8.0 (Figure 2). The various mature peptidases displayed optimum activity at different pH values; FhCB1 and FhCB2 exhibited the greatest enzymatic activity against Z-Phe-Arg-AMC and Z-Gly-Pro-Arg-AMC, respectively, at pH 5.0. By contrast, both FhCB3 and FhCL3 efficiently cleaved Z-Phe-Arg-AMC and Z-Gly-Pro-Arg-AMC at an optimum pH of 7.0 and pH 6.5, respectively.3.3. FhCB1, FhCB2, FhCB3 and FhCL3 Exhibit Distinct Substrate Preference ProfilesThe endopeptidase substrate specificity of FhCB1, FhCB2, FhCB3 and FhCL3 was examined using six different N-terminally blocked fluorogenic peptide substrates at the optimum pH for each enzyme. Under these conditions, FhCB1 showed a clear preference for substrates Z-Phe-Arg-AMC and Z-Leu-Arg-AMC (Figure 3A). By contrast, FhCB2 exhibited low activity against these two substrates, while, at the same time, showing a preference for Z-Gly-Pro-Arg-AMC, Z-Pro-Arg-AMC and Z-Val-Ile-Arg-AMC (Figure 3B). FhCB3 displayed a more promiscuous activity, as it hydrolysed all six fluorogenic peptide substrates examined, although it preferred the substrates Z-Phe-Arg-AMC and Z-Val-Ile-Arg-AMC (Figure 3C). Surprisingly, all three FhCBs showed relatively weak activity against Z-Arg-Arg-AMC, a substrate generally considered to show selectivity towards cathepsins B [32,33].FhCL3 exhibited a unique substrate profile compared to FhCB1, -2 and -3; it effectively hydrolysed Z-Gly-Pro-Arg-AMC (Figure 3D). This peptidase also cleaved Z-Pro-Arg-AMC but with relatively reduced efficiency, highlighting the importance of Gly in the P3 position for binding to the enzyme’s active site. FhCL3 also cleaved Z-Leu-Arg-AMC but not Z-Phe-Arg-AMC.3.4. Kinetic Analyses of the Enzymatic Efficiency of FhCB1, FhCB2, FhCB3 and FhCL3The comparative kinetic parameters, KM and Kcat, of the various enzyme–substrate interactions were determined. The substrates employed for these analyses were the substrates with greater preference for one or more enzymes, as shown in Figure 3: Z-Phe-Arg-AMC (FhCB1, FhCB3), Z-Gly-Pro-Arg-AMC (FhCB2, FhCB3, FhCL3) and Z-Val-Ile-Arg-AMC (FhCB2, FhCB3) (Table 1). Based on the terminology derived by Koshland [34], the catalytic efficiency of the enzyme against each substrate was determined by the ratio Kcat/KM.The hydrolysis of Z-Phe-Arg-AMC was observed to be the most efficient for FhCB1 (Kcat/KM of 13,596 M−1S−1), which exhibited a value at least eight times greater than that obtained with FhCB3 (Kcat/KM of 1708 M−1S−1).The catalytic efficiency of FhCB2 against Z-Gly-Pro-Arg-AMC and Z-Val-Ile-Arg-AMC was similar (Kcat/KM of 4933 and 6876 M−1S−1, respectively) and more efficient than that observed for FhCB3 (Kcat/KM of 579 and 2266 M−1S−1, respectively). For both enzymes, Z-Val-Ile-Arg-AMC was the best peptide substrate.FhCL3 hydrolysed Z-Gly-Pro-Arg-AMC very effectively with a Kcat/KM value of 17,960 M−1S−1, which is 2.3 and 31 times greater than the ratio obtained with FhCB2 and FhCB3, respectively.3.5. Exo-Carboxypeptidase ActivityThe exopeptidase activity of the four recombinant peptidases was assayed against two fluorogenic exopeptidase substrates, Abz-Phe-Arg-Ala-Lys(Dnp)-OH (Exo-OH) and Abz-Phe-Arg-Ala-Lys(Dnp)-NH2 (Exo-NH2), designed to span the S2-S2′ subsites of the active peptidases [35]. Of the four F. hepatica cathepsin peptidases assessed, only the recombinant FhCB1 cleaved Exo-OH and Exo-NH2; however, this activity was similar to that observed for SmCB1, but both were relatively low compared with the positive human control, HsCB (Figure 4).3.6. Inhibitors Screening against FhCB1, FhCB2, FhCB3 and FhCL3A panel of 12 commercially available cysteine peptidase inhibitors that preferentially target mammalian cathepsin B (Ca074 and Ca074-OMe), cathepsin L (L, L1, LII, LIII and LIV) cathepsin S (S) and cathepsin K (KI, KII and KIII) were screened against the FhCBs, FhCL3 and, as controls, HsCB and HsCL1. The broad-range cysteine peptidase inhibitor Z-Phe-Ala-FMK was also included (Figure 5).Each F. hepatica peptidase displayed a distinct profile against the inhibitors tested. However, it is noteworthy that the epoxysuccinyl cathepsin B inhibitors, Ca074 and Ca074-OMe, which we confirmed to be selective for HsCB compared to HsCL1, were poor inhibitors of all three FhCBs and of FhCL3 (Figure 5). Our studies show that the inhibitory constants (Ki) were relatively high for these inhibitors against all FhCBs (Table 2). However, FhCB1, FhCB2 and FhCB3 were most potently inhibited by Z-Phe-Ala-FMK, LII, LIV and inhibitor S (Ki = 345 nM). The compounds KII and KIII abrogated the activity of FhCB3, whilst they were only weak inhibitors of FhCB1 and had no effect against FhCB2.Compared to the FhCBs, the activity of FhCL3 peptidase was potently abrogated by the inhibitor KII (Ki = 0.7 nM). In addition, Phe-Ala-FMK, LII and S also strongly inhibited FhCL3 (Ki =15 nM, 8 nM and 35 nM, respectively).3.7. Effect of Cysteine Peptidases Inhibitors on F. hepatica Metacercarial Excystment In VitroThe excystment of F. hepatica metacercariae was examined in the presence of the most efficient inhibitors, namely, Phe-Ala-FMK, LII, KII, S, Ca074 and Ca074-OMe, each at a final concentration of 100 μM (Figure 6). In the untreated group (PBS and DMSO control), the average excystment rate of metacercariae was 73–83%. While most of the inhibitor compounds just slightly decreased the F. hepatica metacercariae excystment rate, the broad-range inhibitor Z-Phe-Ala-FMK was shown to significantly impact excystment by reducing it by more than 50% (p ≤ 0.005) (Figure 6A).The survival rate of NEJ in culture in the presence of the inhibitors was also examined. A significant effect was only observed with inhibitor S, which reduced the viability of the parasites by ~55% compared to the DMSO group (p ≤ 0.001). None of the other inhibitors tested significantly influenced the survival of this stage of the parasite (Figure 6B).4. DiscussionTo establish infection in their host, NEJ of F. hepatica must migrate from the intestinal lumen, through the intestinal wall and to the liver. During this migration the juvenile parasites encounter several layers of host tissues and various macromolecules that create physical barriers. Immunolocalisation studies using polyclonal antibodies showed that NEJ produce copious amounts of FhCBs and FhCLs by the gastrodermal cells of the parasite gut [36,37,38]. Their secretion from these cells is critical in parasite virulence, as RNAi-induced silencing of these peptidases significantly reduces tissue penetration in vitro [36]. However, it is still unclear whether each peptidase plays similar or distinct roles in virulence. Here, we comprehensively characterised recombinant forms of the four major NEJ peptidases, FhCB1, FhCB2, FhCB3 and FhCL3, and show that they displayed major biochemical differences.The optimal pH for proteolysis by FhCB1 and FhCB2 was observed at pH 5.0, while that observed for FhCB3 and FhCL3 was closer to neutral, pH 7.0 and pH 6.5, respectively. The optimal pH for activity is often indicative of the physiological pH found at the site of action for a given peptidase; therefore, these results suggest that peptidases function in different places within the parasite or host [28]. Significantly, the digestive tract of F. hepatica is maintained at slightly acidic pH, around pH 5.5 [28,39], while the pH of the host intestinal tract is between 6.0 and 7.4 [40]. This could suggest that FhCB1 and FhCB2 function predominantly in the parasite gut lumen, while FhCB3 and FhCL3 play a more prominent role in the penetration of the host tissues. Indeed, FhCB3 and FhCL3 are the most prominent peptidases found in the ES of juvenile parasites maintained in vitro [10,37,41,42].FhCL3, FhCB2 and, to a lesser extent, FhCB3, showed selectivity toward cleavage of the substrate, Z-Gly-Pro-Arg-AMC, that is indicative of collagenolytic activity. Orthologous peptidases of FhCB2 and FhCB3 from F. gigantica, FgCB2 and FgCB3, respectively, were also shown to digest type I collagen [43]. We have previously reported that FhCL3 exhibits a similar substrate specificity to the collagenolytic enzyme cathepsin K and likely aids in the degradation of host tissue in order to facilitate migration of parasite NEJ from the intestinal lumen to the liver [44]. This present study is the first to indicate that NEJ FhCB2 and FhCB3 may function in concert with FhCL3 to degrade the interstitial matrix of tissues to facilitate the parasite’s migration through the intestinal wall. However, kinetics studies show that FhCL3 displays the greatest activity on Z-Gly-Pro-Arg-AMC and its performance constant was found to be 2.3 and 31 times greater than FhCB2 and FhCB3, respectively.By contrast to the other three peptidases, FhCB1 showed little activity towards substrates containing a Pro residue at the P2 position, suggesting that this peptidase is not directly involved with collagen degradation. However, FhCB1 also differed in having very low activity against Z-Arg-Arg-AMC, a substrate typically cleaved by cathepsins B. Greater activity was observed for FhCB1 against substrates containing the peptides Z-Phe-Arg-AMC and Z-Leu-Arg-AMC, indicating a preference for hydrophobic P2 residues, which is more typical of cathepsin L endopeptidase activity [44]. This unique property may suggest a distinctive function worth elucidating in the future.The S2 subsite of the active site of cathepsin B peptidases are of prime importance for holding substrates in position; these are largely conserved in cathepsin B peptidases (Supplementary Figure S1). Of particular note is the residue Glu316, which, in SmCB1, is found at the bottom of the S2 subsite and is shown to interact directly with the substrate/inhibitor P2 residue and change its orientation within the S2 subsite (Glu246 in SmCB) [24]. However, in FhCB1 and FhCB3, this residue is replaced by the smaller hydrophobic Ile and basic Arg, respectively, while FhCB2 retains the acidic Glu. This single residue difference has a significant impact on the substrate specificities of the enzymes as, for example, when Glu316 was replaced by Gln, in rat cathepsin B, it significantly decreased its ability to accommodate an Arg residue in its S2 site [45]. Using the substrate Z-Phe-Arg-AMC, we found that FhCB1 had a value of Kcat (substrate turnover) more than 20 times higher than that of FhCB3, while this substrate was not cleaved by FhCB2. On the other hand, both FhCB2 and FhCB3 preferably cleaved Z-Val-Ile-Arg-AMC over all other substrates examined, whereas this was a poor substrate for FhCB1. Collectively, these data indicate major difference in the substrate specificity of the three F. hepatica cathepsin B peptidases.Cathepsin B peptidases are uniquely characterised by having a flexible loop structure, termed the occluding loop, which is located at the apical region of the active site cleft and confers these enzymes with unique carboxy-peptidyl-dipeptidase activity. This activity brings about the removal of two amino acid residues from the carboxy-terminus of protein substrates [24,35,46,47]. Deletion of the occluding loop sequence not only obliterates this exopeptidase activity but also increases the endopeptidase activity of the enzyme [48]. This exopeptidase activity is facilitated by the presence of two neighbouring His residues in the occluding loop, at positions 110 and 111 (S. mansoni mature cathepsin B1 enzyme numbering in Jílková et al. [24]; Supplemental material Figure S1), which provides a positively charged anchor for the C-terminal carboxy group of the substrate. However, only His110 is thought to be critical for exopeptidase activity, while His111 increases the positive charge on the loop and increases the binding potential. To hold the occluding loop in an appropriate conformation for exopeptidase activity, in human cathepsin B, the His110 forms a salt-bridge interaction with Asp93, which is strengthened by the presence of His181 [35].Sequence alignments show that FhCB1, FhCB2 and FhCB3 possess an occluding loop, but, while these retain His110, they each lack His 111 (Supplemental Material Figure S1). Absence of this residue has been previously shown to deplete the carboxydipeptidyl activity of FhCB2 [15], while F. gigantica cathepsin B5, which possesses both His110 and His111, exhibits exopeptidase activity [49]. Our studies add support to these previous reports by showing that FhCB2 and FhCB3 do not possess carboxydipeptidyl activity, while FhCB1 exhibits low but significant exopeptidase activity. Of further interest is the fact that the compounds Ca074 and Ca074-OMe, that specifically bind and inhibit mammalian cathepsin B by forming H-bonds with His110 and His111 on the occluding loop [21], were very poor inhibitors of the F. hepatica cathepsins B.Interestingly, we found that the occluding loops of the F. hepatica and S. mansoni cathepsins B were not highly conserved. This region only shares 11–23% sequence identity between the FhCBs, whereas it is generally conserved in cathepsins B from mammalian species [24]. This unusual non-conserved region could contribute to the lack of carboxydipeptidyl activity found in the F. hepatica cathepsins B and/or could confer unique, so far undisclosed, exopeptidase activities.A screen of commercially available cathepsin inhibitors revealed that the three FhCBs and FhCL3 exhibited distinct inhibition profiles. While the five cathepsin L inhibitors (L, LI, LII, LIII and LIV) showed specificity for HsCL1 over HsCBs, they exhibited high variability against the FhCBs and FhCL3, again emphasising the difference between the parasite enzymes and their mammalian hosts. For example, the epoxysuccinyl peptide inhibitor, L, which did not inhibit FhCB1 or FhCB2, significantly reduced FhCB3 and FhCL3 activity, whereas L1, containing a Phe-Phe peptide, was the most effective at blocking FhCB1 and FhCB3 activity. LII, which contains the dipeptide Phe-Tyr, was the most effective at reducing the enzymatic activity of all F. hepatica peptidases but this compound also inhibited human cathepsin L and B demonstrating a potent but less specific interaction of this inhibitor with the parasite and host enzymes.The mammalian cathepsin K inhibitors (KI, KII and KIII) were generally poor inhibitors of FhCB1 and FhCB2 but reduced FhCB3 and FhCL3 activity. This observation is in agreement with our earlier finding that FhCL3 exhibits cathepsin K-like collagenase activity [44] and further implies that FhCB3, which can also cleave the substrate Z-Gly-Pro-Arg-AMC, may collaborate with FhCL3 to perform this activity in vivo. However, despite the fact that FhCB2 can cleave Z-Gly-Pro-Arg-AMC, it was not inhibited by the KI, KII or KIII compound.Detailed kinetics studies were performed with six inhibitors and confirmed the poor inhibition of all four parasite enzymes with the inhibitors Ca074 and Ca074-OMe. In general, FhCB1 and FhCL3 were more susceptible to inhibition by cysteine peptidase inhibitors than FhCB2 and FhCB3. Nevertheless, we found that the cathepsin L inhibitor, LII and the broad range inhibitors Z-Phe-Ala-FMK were extremely potent inhibitors of FhCB1 and FhCL3 with Ki values in the low nM range, <15 nM. The cathepsin S inhibitor, S, a Phe-Leu dipeptide with a keto-aldehyde (COCHO) group, which has reported Ki values for inhibition of human cathepsins S and B of 0.185 nM and 76 nM, respectively, ref. [50] was the next best inhibitor (Ki values of 14 nM and 35 nM for FhCB1 and FhCL3, respectively).Our NEJ excystment experiments revealed that the majority of cysteine protease inhibitors examined at a 100 μM final concentration had little effect on the ability of the fluke to emerge from the cysts. The only compound to show significant inhibition of excystment was Z-Phe-Ala-FMK. Previous studies from our laboratory have shown that a final concentration of 1 mM of this compound completely prevented metacercariae excystment [10]. This suggests that the inhibitors do not readily pass through the metacercarial cyst walls and supports the suggestion by of Dixon and Mercer [51] that the metacercarial cyst walls are generally impenetrable to small compounds.When NEJ were cultured for 24 h with each inhibitor, we observed that most of the cysteine peptidase inhibitors did not elicit a significant effect on parasite viability. Our data differ from those of Beckham et al. [15], who reported that the cathepsin B inhibitor, Ca074-OMe, had a sub-lethal effect on NEJ cultured in a concentration of 6.25 μM. In our studies, we found that the inhibitor S, an effective inhibitor of FhCBs and FhCL3, was the only effective flukicide against NEJ within 24 h of in vitro culture. The compound S belongs to a group of inhibitors known as alpha-keto-beta-aldehydes that contain two highly electrophilic carbons (α and β) in their C-terminal group that facilitate strong and irreversible binding to the cathepsin cysteine peptidases [52,53]. However, the potency and indiscriminate nature of alpha-keto-beta-aldehydes in protease inhibition (since they also inhibit serine peptidases) may have implications for their toxicity to host biology and potentially prevent their use as therapeutics.In summary, we undertook a series of experiments on enzyme–substrate/inhibitor kinetics and revealed distinct differences in the specificity of the NEJ FhCB1, FhCB2, FhCB3 and FhCL3. This variation provides evidence for distinct and over-lapping functional roles of the peptidases that would allow the parasite to degrade a multitude of macromolecular substrates. Investigations into the digestion of protein substrates from mammalian hosts susceptible to Fasciola infection have shown that NEJ proteinases catalyse the hydrolysis of gelatin [54,55], collagen [43,44,56], fibronectin [57], serum albumin [12], immunoglobulins [58] and haemoglobin [28]. Collectively, the data support our hypothesis that peptidases secreted by juvenile flukes act together as a powerful hydrolytic mix to facilitate tissue migration through digestion of host tissues, which would otherwise act as physical barriers to invasion, with complementary functions in feeding and immune evasion. The blocking of the peptidase activity in vitro did not affect parasite viability, but this may be because the enzymes predominantly function extra-corporeally. However, we have previously shown that RNAi-mediated knockdown of FhCL and FhCB expression in NEJ did prevent the parasites’ migration through the gut wall in culture [36]. Hence, the blocking of their activities in vivo by chemical or immunological means could still offer a route towards developing new anti-fluke treatments.

animals : an open access journal from mdpi

10.3390/ani11030705

PMC8000346

Current cultural shifts in Western countries have changed the position of the cat to a companion animal, and its traditional role as a pest controller is no longer recognized by city dwellers. In a growing number of theoretical and field studies, the hunting abilities of cats and their high fertility are perceived as environmental risks. Bringing together theoretical perspectives from human–animal studies, animal ethics, population ecology, and biosemiotics, I highlight the existence of two different ecological (and even cultural) communities inhabiting urban environments: the culture of feral cats and the humano–cat culture of pets. Arguments are given for the essential role of feral cats in the population dynamics of the species when a growing number of pet cats are routinely neutered. Whereas neutering is presented by animal shelters and veterinary institutions as a universal means for improving cat welfare, it is at odds with the psychobiological needs of cats as viewed by a laissez-faire approach. This leads us to the conclusion that instead of one type of management of free-roaming cats, individual solutions should be sought to achieve a balance between the welfare of cats, other species, and human cultures in diverse urban environments.

Urban environments are inhabited by several types of feline populations, which we can differentiate as feral cats, free-roaming pets, and confined pets. Due to a shift in the cultural representation of cats from pest controllers to companion animals, cats living semi-independently of humans are perceived increasingly negatively, while the pet population has become the object of intense care. A regulative approach converges with a concern for welfare in the operation and educational campaigns of municipal shelters, which through their implementation of neutering policies have proven to be key players in the contemporary relation of urban cats and humans. The generally widespread notion of cat welfare associated with a secure life comes into tension with the fact that the psychobiological needs of feral cats are significantly different than those of pets. It becomes apparent that individual interactions between humans and cats in urban environments in the Anthropocene are increasingly influenced by the intervention of institutions that can be characterized as seeking to administer the wild.

1. IntroductionThe cohabitation of cats, humans, and other species in urban residences represents a complex social, ecological, and, increasingly, ethical problem. A growing number of studies portray cats as dangerous predators that threaten the stability of bird, small mammal, and reptile populations or as carriers of dangerous diseases. A different perspective sees cats (along with dogs) as the most favored household pets, and concern for the welfare of cats motivates the actions of many individuals and municipal institutions. Cats occupy a dual role as “autonomous predator and ostensibly dependent companion” [1]. This deeply ambivalent attitude toward cats is inevitably mirrored in the problematic practice of trying to regulate the cohabitation of humans and cats in particular towns. This ambivalence is often to be found in the thinking of those who associate a negative image of cats with feral colonies yet reserve a positive one for individual cats who have owners.The Anthropocene epoch can be characterized not only by the increased impact of human activity on the biosphere, but also by the greater determination of people to regulate the ecological relations of species (e.g., conservation programs, intervention against invasive species). Associated with this is a concern for animal welfare, extending ethical considerations from humans to include other animal species. In the case of cats living in urban environments, there arises the question of how to harmonize the interests of individual groups of residents, cats, and the species cats prey upon. Theoretical studies usually assume that from an environmental perspective, cats are an alien species that threatens populations of small vertebrates due to their exceptional predatory skills [2,3]. Their discourse repeats the problems of invasive species ecology [4].From the point of view taken in the present study, this view is problematic for at least two reasons: (i) in regions where domestic cats have been living for several centuries (particularly the “Old World”), they can rightly be considered a natural part of the ecosystem, since there prevails “a high degree of adaptation of local wildlife to cats” [5]; (ii) urban areas are characterized by a high concentration of both cat and bird populations [6,7]; (iii) domestic cats ecologically compete with/prey on the predators of bird nests like the brown rat (Rattus norvegicus), beech marten (Martes foina), and garden dormouse (Eliomys quercinus). Domestic cats are generalist and obligate predators that receive supplementary food, and their population density reflects that of humans more than the density of their prey [8]. In urban ecology, the classical distinction of nature and culture is problematic: rather than “wildlife,” in urban areas it is more accurate to speak of synanthropic species that have inhabited city spaces at different times. From this perspective, the common blackbird (Turdus merula), for example, is a more recent arrival in urban spaces than cats. As this commentary concerns the problem of cats in urban environments, it does not take into account cases of wild populations threatening endemic species (particularly in fragile island ecologies) or farm cats occasionally preying on wildlife. Therefore, it is justified (particularly in the Old World) to view free-roaming cats as a natural part of the character of urban areas rather than an invasive species [9,10,11,12].Free-ranging cats are individuals with the characteristics of semi-wild commensal animals whose important ecological and ethological functions, unlike fully domesticated animals, are not under human control [13]. The legal position in many European countries that defines a cat as a domestic animal and presupposes a distinct owner for each individual thus does not reflect the variability of ecological and social niches in this species [14]. Feral colonies can receive supplemental feeding from cat lovers, but “feeding ladies” are not responsible for the behavior of the individuals they assist. Furthermore, the application of a given welfare concept has different impacts on individuals from different cat populations. For feral cats, veterinary care means removal from its environment and the endurance of significant stress, as they perceive humans primarily as predators, whereas free-ranging pets are acculturated to a degree that a visit to the veterinarian does not represent a decisive intervention in their lives (with the exception of neutering). The situation is further complicated by the fact that animal welfare can be evaluated according to three criteria: (i) affective states, (ii) natural living, (iii) basic health and functioning. Animal welfare involves different components that can be grouped roughly under these headings, which involve considerable but imperfect overlap. It is crucial to understand that the pursuit of any one criterion does not guarantee a high level of welfare as judged by the others [15].For the sake of simplicity, I will group criteria (i) and (ii) in the case of cats under the laissez-faire policy that allows them to fulfill their psychobiological needs, while the veterinary view (iii) is based on the utilitarian perfectionist stance (more on this in Section 2) [16]. While the laissez-faire policy is applied in the professional literature primarily to members of wild species [17], the veterinary view prevails in the approach to the welfare of domestic animals, including cats [18]. It is important to note, however, that the first view is very widespread among cat owners, particularly in rural areas and in poorer urban areas [19]. In the approach to cats living in urban areas, we also see an intense clash of welfare concepts, expressed most forcefully around the issues of the free movement and reproductive possibilities of pets. I aim to demonstrate that the universal labeling of owners with a more liberal approach as “irresponsible” is the result of an excessive simplification of a complex issue. Key to this study is an understanding that proper management of urban cats is not a matter only of scientific facts, but also of cultural and ethical values manifested in the preferences of urban residents and in differences of welfare criteria. Every effort to adjust the relationship of humans and cats in urban environments is thus based on both scientific evidence and value assumptions that are of a different character in the case of feral cats, free-roaming pets, and confined pets, as will be demonstrated in Section 3 (threefold modeling). It will also be shown that the welfare of an individual does not necessarily overlap with the interests of the population to which it belongs—population genetics [20] and group-behavioral specifics (cultures) [21] must be considered here.While conscious that circumstances may differ in individual countries, I proceed on the basis of studies carried out in Great Britain and the United States, which will be supplemented in places with the situation in other Western countries. The aim here is not a comparative study of the development of individual populations, but the utilization of empirical studies to identify the conflicts that arise when applying different approaches to cat welfare.2. Urban Cat Populations as Distinctive CulturesDrawing on the work of Natoli and Sandøe, I will differentiate the following three groups of synanthropic domestic cats (Felis silvestris catus) living in urban areas: feral cats, free-roaming pets, and confined pets [19,22,23]. Feral cats live in loose associations and can be found in public spaces. Urban environments provide them suitable shelter and sufficient nutrients, both from people (food scraps) and through predation (mostly rodents). While they are typically wary of people, they can develop relationships with specific people who feed them (known in the English literature as feeding ladies). Free-roaming pets include cats typically associated with a single household but which have the possibility of free movement in an urban environment. They come into contact not only with other pets, but occasionally with ferals as well. Their degree of dependence on a particular household varies from case to case, but it can generally be said that when appropriately cared for, they seek regular contact with their owner, in connection with the intake of the majority of their nutritional supply. Confined pets are tied to a specific household without the possibility of movement beyond the space of the house (with the possible exception of supervised movement, e.g., around a vacation home). They are fully dependent on people, who also make decisions on their reproductive possibilities. Reproduction is generally allowed only to pedigree cats, as with domestic short-haired cats the behavioral manifestations associated with intact individuals come into conflict with the restrictions of urban apartments. In practice, it is difficult to precisely distinguish individual categories of cats: “stray cats” (or also semi-feral) can be perceived as a transitional category between ferals and free-roaming pets. These individuals are not tied to any particular household but can receive supplemental feeding from residents of households and at the same time transiently join colonies of feral cats. For our purposes, however, it is sufficient to distinguish these three categories of urban cats, whose characteristics are summarized in Table 1.When deliberating on the appropriate management of cats in urban environments, it is important to realize that the set of ecological, social (in relation to other cats and to humans), and behavioral needs of each group varies to a degree that entitles us to speak of different cat or cat-human cultures. The phrase “cat culture” was first used by sociologists Janet and Steve Alger in describing the environment of a cat shelter where humans and cats interacted and caretakers took into consideration the different temperaments and habits of individual cats in assessing their needs [24,25,26]. The concept of the social life and inner cognitive-affective world of cats as a culture can be understood theoretically from a biosemiotic perspective, which emphasizes the ability of animals to actively interpret their surroundings (the concept of Umwelt) [21,27,28]. This concept makes apparent the increased role of the social sphere in contrast to purely genetic dispositions, fully corresponding to the significant ecological plasticity of this species. Paul Leyhausen observed cats in Paris gathering in a single location without displaying the usual territorial aggression and described the tradition as “social gathering” [9]. The generational continuity of behavioral characteristics is exemplified by the fact that during a sensitive period (3–8 weeks of age), kittens adopt from their mothers the manner of relating to other cats and to people, and these early experiences have long-lasting effects into adulthood [29,30,31]. Mistrust of humans is transmitted intergenerationally in feral communities, while among pets the need for physical and social contact with people is an important component of their welfare.While in the case of feral cats, their dependence on humans is indirect, and in the environments of Western cities, they are merely tolerated, pets are connected with humans to an extent that they cannot be considered a separate population. This is most evident with neutered individuals, who in some regions of the West make up the majority of the cat population, but do not contribute genetically to its future composition (in Shirley, Southampton, the estimate in 1994 was 96.8% of adult males and 98.7% of adult females; among all owned cats in the USA, the estimate was 79.8% in 1994 and 80% in 2007) [20,32,33]. In the case of pedigree cats, the appearance and behavioral characteristics of a breed are objects of intensive artificial selection. Conversely, we can say that people adapt to the needs of “their” cats in that they do not stay away from their homes for long periods. We can thus speak of the specific humano–cat culture of pets, whose existence is also reflected in the fact that domestic cats have gradually spread to all continents [21,34].On a theoretical level, there arises the possibility that the need to regulate cat populations would apply exclusively to feral cats, who, because of their limited access to human-mediated diets, pose a greater threat to populations of birds and small mammals. Attempts to extinguish feral cat populations, such as in Australia, are a consequence of such considerations [35]. Concern for the welfare of cats would then be concentrated only on pets. This is problematic on two levels. In practice, it cannot be declared that ferals and cats living in close contact with humans are clearly divisible by group. If we see a tabby cat walking on a street, it is not clear to which category it belongs. This raises a considerable dilemma, as, depending on the categorization, the same individual can be seen as an object of regulation (feral) or of concern for its welfare (pet). On another level lies a distinct ethical problem: is it possible to give different normative valuations of members of the same species?The majority of studies a priori perceive the very existence of feral colonies as a problem that needs to be solved by human intervention [4]. In recent decades there is increasing concern for the conservation of species that fall prey to roaming cats [8,36]. In addition to predation, problematic factors often named include disease transmission (to pets, livestock, wildlife, humans), noise during mating, and the presence of excrement in public spaces [23,37]. From a different perspective, one can assume that urban residents are averse to feral kittens’ high susceptibility to disease and high mortality (87.5% [38]).Some studies are emerging that demonstrate an important function for ferals in the genetic continuity of cat populations [20] or demonstrate their affiliative relationships with particular people [10,11]. Members of animal rights organizations also have a positive relationship with feral cats in that they perceive them as objects of care, which mainly concerns “rescuing” feral kittens. Some supporters of animal rights do not acknowledge the independent right of reproduction, which in domesticated animals is often seen as perpetuating their suffering [39]. This attitude (common in urban shelters) is applied in relation to feral colonies, whose numbers are reduced in the name of preventing unnecessary suffering, through trap–neuter–return (TNR) programs or adoption (which is also associated with neutering) [40]. Such an approach is usually perceived as more humane than a direct eradication of feral colonies (but see [41,42]).Even if we encounter a concern for welfare in the management of feral cat colonies, it must be realized that this is rather about choosing between the preferred outcomes of different groups of urban residents than a direct consideration of the interests of feral cats themselves. In theory, management decisions motivated by an authentic interest in the welfare of feral cats should have two stages: (i) what is in the interest of feral cats as an independent population/culture; (ii) how can these interests be reconciled with the needs of different groups of urban residents (who may also be representing the interests of other animal species)? In practice, however, stage (i) is rarely taken into account. Feral cats are regarded through the prism of welfare as fully domesticated animals (for whom a maximal life expectancy, for example, is considered desirable), or it is automatically assumed that numerical regulation of colonies is necessary and desirable. The significant dispute between proponents of the TNR method and of direct euthanasia of feral cats takes place in stage (ii): the decision is between which set of institutional actions is less stressful for feral individuals, which, however, does not mean that a given type of management is applied primarily in the interest of their welfare.In all cases of feral cat management (euthanasia, TNR, transfer to cat shelters), a consistent implementation would lead to the disappearance of the specific behavioral-social manifestations of urban populations, which significantly mirror the general ecological strategies of domestic cats as a semi-wild species. These can come into conflict with certain anthropocentric leanings in the evaluation of cat welfare. As an example, from the perspective of the natural reproductive dynamics of feral populations, kittens are not pampered playthings; rather, their large litters represent an expendable resource strategy [43]. Assessing the welfare of pets is a difficult task. Unlike with feral cats, we cannot speak of a distinct population, as the movement, behavior, and reproduction of pets differ according the type of cohabitation they have with their owners. An individual cat’s quality of life is linked with the tolerance by members of the household of its behaviors, which may be manifestations of its own well-being but can be viewed negatively by humans. Steps taken in the supposed interest of pets, then, are in reality always a compromise of the interests of the individual actors (this should also include the point of view of species preyed upon) [44,45]. The theoretical and ethical dimensions of the problem are also unclear. On the one hand is a utilitarian perfectionist stance, subordinating the satisfaction of instinctive desires to overall quality of life (i.e., long lifespan) [18]. Such a view typically leads to neutering and confining pets. On the other hand is the traditional laissez-faire policy, which prioritizes the fulfilment of psychobiological needs of animals (e.g., engaging in predatory behaviors) [46]. The latter approach can combine intense care for an animal’s health (e.g., regular visits to the veterinarian) with the possibility of free movement, even though this can potentially bring harm. As research in several European countries has shown, such an approach is taken by most pet-owning households [19,45]. In the matter of pet welfare, the issue of neutering is a chapter unto itself. In contemporary Western society, there is a broad consensus supporting the neutering of cats that are not kept for breeding purposes [47,48]. Taking into consideration that “complications may develop from anesthesia or surgical trauma,” the main arguments for neutering are: (i) the prevention of potentially unwanted kittens; (ii) a reduction in behavioral problems in relation to owners or other people (e.g., increased aggression) [46]. It is important to realize, however, that this argument is not valid from a laissez-faire perspective, as it does not allow for certain key psychobiological needs of the animal associated with mating and nurturing offspring [16]. Routine neutering is particularly problematic, as it has the potential to significantly affect the population dynamics of the entire species.From a population genetics and ecological perspective, the implementation of routine neutering constitutes disruptive selection: from free-ranging cats, only individuals who have learned to completely avoid humans and perceive them basically as predators remain intact. If they are living in groups, these represent a type of wild population whose members are difficult to redomesticate (classified as pseudo-wild in [20]). At the opposite end of the spectrum, then, are more and more individuals whose movement beyond the grounds of their owners is very limited (perhaps in a protected enclosure in the garden), and, regarding their possibilities of reproduction and social contact with other cats, they fall under full control of an owner. In light of this trend, it is necessary to observe that the dynamics of the human–cat relationship are shifting, at least in Western urban areas. At the same time, it is reasonable to suppose that the kind of interaction between humans and free-roaming cats who decide themselves whether to spend the night near a human dwelling or under cover of darkness has remained prevalent throughout the history of human–cat cohabitation, which goes back to ancient Egypt [13].3. Threefold Modeling of Urban Cat-Human RelationsIn understanding the multifarious interactions between humans and cats in urban environments, it is important to keep in mind that there are differences in opinion, not only on interspecies contact, but also on how to view cats themselves. The peak of conflict here are the “cat wars,” in which one side maintains that pet cats should be kept indoors or have restricted outdoor access, while the other side is of the opinion that companion cats should be allowed to move about in public spaces [3]. It would be naïve to believe that this disagreement is based on objective research into the ecological and social role of cats; it arises, first of all, from a change in society, which then sees cats through different eyes. It is important here to observe the shift of assessment criteria from economics or social utility to the problem of welfare. In the Anthropocene, we are witness to a widespread conviction that, in the interest of welfare, active interventions into the ecological relations or even the physiology of a given species are necessary. To an increasing degree, such efforts are propelled by institutions driven by both general demand and scientific studies. To better understand the complex dynamics of these relationships, it will be helpful to distinguish which cat populations are concerned. For each, we will note the mutual influence of three layers: zoosemiotic interactions, institutionalized actions, and cultural representations. Zoosemiotic interactions include intra- and inter-species contacts of an individual nature—for example, greeting rituals between cats or vocal communication between a cat and its owner. From a biosemiotic perspective, the cat and the owner here are both active agents whose situational behavior is modulated by individual experience. In contrast to this are institutionalized actions performed through a mediator who does not have an established relationship with the particular animal and is acting in a professional capacity. Institutions, such as animal shelters or veterinary organizations, regulate the movement and reproduction of urban cats, generally in accordance with public opinion. Given that different residents can have diametrically opposed assessments of the same cat population, we need to include the fluctuation of cultural representations in the model [21]. Confined pets are the source of least social conflict, as their owners have full control over their movement and reproduction. These cats are in compliance with the idea of responsible private ownership and do not come into contact with other people, such as neighbors. Their owners may doubt whether keeping them exclusively indoors is good for their physical condition, but in the case of intact pedigree females, breeders prefer the assurance of maintaining a genetically pure line. With males, the risk of injury from fighting or from passing cars is considered too high. Regard for a cat’s welfare and a possible need to plan its reproduction thus leads to a complete restriction of free movement, which has the side effect of confined cats posing no threat to populations of wild animals, other pets, livestock or humans (e.g., through a transmission of parasitic diseases) [37]. Thus, confined cats do not bother bird watchers, nature conservationists, hunters, and other interest groups [19]. We might add that indoor cat owners are the ideal customers of a large-scale industry serving cat needs—everything from canned food, to litter boxes, to scratching posts. It is the invention of nutritionally balanced cat food and the commercial availability of other pet products since the 1950s that has made it possible to keep cats exclusively indoors [1] (Figure 1). At the other end of the spectrum, here are feral colonies. These cats have a good number of human opponents, due not necessarily to personal experience, but to negative cultural representations, which are further strengthened by some ecological, epidemiological, and veterinary studies. Feral cats are portrayed as effective killers, carriers of parasites and disease, and disturbers of the peace (e.g., loud mating noises). These colonies are the primary focus of institutions of public health and veterinary medicine, frequently in cooperation with animal shelters. There are, of course, countries where this picture is more complicated (in Europe, primarily the Mediterranean countries). Groups of cats living in historical city centers are seen as part of the genius loci, and individual residents are friendly toward them, valuing the positive aspects of their presence. Natoli points to the antidepressant effect of colonies on the individuals who feed them (“gattare” in Italian), the educational effect on those interested in animals, and lastly the aesthetic effect (cats as “living decoration”) [18]. Another positive aspect is predation of urban rodent populations. In Italy, concern for animal welfare extends to feral cats, thanks to law no. 281 (enacted 1991). The prevailing interpretation is that the population of colonies should be regulated, yet individual animals should not be subjected to unnecessary stress. TNR method is usually employed. In sensitive cases, veterinary authorities work with people who have regular contact with colonies to capture less timid individuals [47]. The question of whether sterilization is compatible with a concern for welfare will be discussed below, but here we can note that members of animal rights organizations, as well as opponents of feral cats, can agree on a policy of reducing the number of colonies, but not completely eradicating them (the practice of TNR, with aid from volunteers, is widespread also in Austria, France, Portugal, Spain [14], and in parts of the USA as well [49]). It must be taken into account that shelters and veterinary organizations do not merely carry out the will of urban residents, but actively shape the discussion on welfare and themselves serve as a sort of “executive body” for managing free-roaming cats. Here we should point out the hybrid operation of cat shelters: they are hostile to feral colonies because of concerns over public hygiene or for species preyed upon by cats, while they see the wandering individual cat as an object of care. This inconsistency is sometimes rooted in legislation differentiating the approach to feral and owned cats (e.g., in the northern territory in Australia) [35]. During research in Estonia, all cat shelter managers I spoke with were convinced that a cat should have an owner, whose responsibility it is to “supervise” its movement to a certain degree. This conviction is unambiguously expressed in the shelters’ practice of microchipping their animals, which in some cities (e.g., Tallinn) is a legal obligation [21]. Shelters have effective public relations and present themselves in the media almost exclusively as places where abandoned cats find a home. The downside of their function of providing (temporary) homes to large numbers of animals is that their inhabitants are exposed to communicable diseases, to which stressed animals are particularly susceptible [50]. Municipal shelters, even in good faith, often cannot provide animals the conditions they need, and diseased or behaviorally problematic individuals are frequently euthanized, if only because of limited space. In deliberating on an ethical way of dealing with feral populations, one must face the sad reality that the direct eradication of colonies can be a better solution than the incessant suffering of their individual members in shelters (e.g., if confined in small cages), since their chances of adoption are minimal. TNR would seem to be a compromise here, as it allows the returned cats to continue their lives of freedom, although it also raises concerns about welfare, especially regarding difficulties of social continuity given the changes in psychical, hormonal, and immune function that can result from neutering [40,41] (Figure 2).At this point, it is useful to show how an understanding of the specific needs of particular cat populations can help shelters that wish to consider the welfare of their animals as individuals [24]. Firstly, it must be recognized that representatives of different cat cultures have different social needs when it comes to the other cats and the people at the shelter (cf. Table 1). For feral cats, it is most beneficial if they can move freely around the space, allowing them to maintain contact with other individuals according to their particular preferences. Shelter employees should give these cats their space and let them initiate brief contact with people (excepting, of course, those in need of medical treatment). On the other hand, cats who had an owner can be very stressed in the presence of other cats, and as a rule, need contact with staff and visitors. These individuals have a good chance at adoption, but in the meantime, it is important to ensure that their living quarters offer a hiding place and that they have an appropriate degree of seclusion. Medium-sized cages are a good option, so long as shelter employees provide isolated individuals regular physical social contact [51]. If a cat is taken off the street, the problem arises of identifying which category it belongs to; this, however, can be solved relatively easily by observing its reaction to other cats and to people [52,53]. Trained volunteers play a key role here, as they do also in resocializing cats who were previously owned by humans, but then spent a long time in the streets. Relations of city dwellers to free-roaming pets are highly variable. First, we should note that while the categories of feral and confined cats are conceptually and practicably distinguishable, free-roaming pets do not constitute a group with distinct margins. This group includes neutered and intact cats, cats with apparent owners or those who frequent multiple households, cats who keep to themselves or who go on occasional wanderings. Such individuals are often the source of conflict between neighbors, which reflects the divided approach to free-roaming pets among the general public and among scientists as well. Concern for the welfare of pets can result in keeping cats indoors or allowing them to spend time both indoors and outdoors, and the decision of whether to neuter a cat depends primarily on the owner’s tolerance of the accompanying olfactory and behavioral manifestations. Solicitude towards cat welfare, then, can be seen rather as additional justification of the individual ideas of owners, as the arguments can go either way. This is increasingly compounded by concern for the welfare of species preyed upon by cats (especially birds), for which cat owners in different countries take varying degrees of responsibility [45]. Western city dwellers are generally coming to see the norm of neutering free-roaming pets as the responsible choice for owners and caregivers—it is endorsed by both “intolerant neighbor” types and influential organizations that see it as part of a comprehensive animal care fulfilling the criteria of welfare. As previously mentioned, neutering of males and females is a widespread practice with tremendous support from shelters and veterinary clinics. A routine part of this campaign is the portrayal of owners who do not neuter their cats as irresponsible (with the exception of pedigree owners); there are large numbers of cats in shelters, and every newborn kitten unnecessarily becomes a potential ward of these facilities. With increased public awareness and increases in neutering, however, situations can arise (e.g., in Finland) where the population of free-roaming cats significantly decreases and demand for adoptions must be met by shelters from abroad [21]. Here, we can see that the utilitarian perfectionist stance applied to the life of a particular animal is in direct conflict with the needs of the population of which it is a member. Added to this is the reality that the safety associated with keeping cats indoors has its downsides as well (boredom, obesity, stress), which further calls into question the veterinary conception of welfare [45]. A representative study of pets raised in Denmark has shown that while cats who are not allowed outdoors exhibit an increased degree of behavioral problems, pedigree cats are burdened with a higher incidence of disease [48] (Figure 3).The appropriateness of neutering cats is not a purely scientific question, but to a significant degree also a valuative and sociocultural question. Various conceptions of animal welfare come into collision here—leaving aside the problem of interaction with other species, a positive evaluation of a neutered cat’s welfare stands or falls on the assumption that a potentially longer life is worth more than the hormonal, behavioral, and social processes associated with reproduction. The question of human responsibility for the behavior of cats is complicated by the fact that pet cats have many characteristics of a domestic species (confined pets being fully domestic), whereas feral cats, from a behavioral and ecological perspective, are at most a semi-domesticated species [13]. Added to this are the varying ethical views on the responsibility of humans for the reproductive scenarios of urban populations: a utilitarian perfectionist stance tends to assign full responsibility, while the laissez-faire approach sees cats as independent and free actors. One wonders whether neutering campaigns, at root, is not the reaction of a society that wants to eradicate the wild side of cat life (associated with the cycle of reproduction—instinctual freedom—death) as a projection of its own negative image [21]. We should recall that a condemnatory view of the fertility of cats is a repeating motif in history and played a role, for example, in the witchcraft trials of Tudor England [54]. Based on the examples of threefold modeling given here, it is difficult to avoid the concern that the one-size-fits-all approach to cat welfare supported by contemporary institutions (animal shelters, veterinary clinics) diverges from the actual psychobiological needs of many members of the species Felis silvestris catus. We have seen that the group of free-roaming pets is being exposed to pressure from the changing attitudes and needs of city dwellers. Even greater pressure is exerted on the culture of feral cats, which, to a significant degree, is intentional, given that they, unlike free-roaming pets, are portrayed in a thoroughly negative light. At the theoretical level, we need to subject to criticism the welfare concept for a species that most experts do not classify either as purely domesticated or as wild [55]. The guiding principle here must be an approach specific to the population and the individual.4. ConclusionsIn this study, I have highlighted three main problems of the cohabitation of humans and cats in urban environments: (i) interventions in feral cat colonies; (ii) keeping pets exclusively indoors; (iii) applying across-the-board neutering of all groups of cats except pedigrees. It is characteristic of the Anthropocene era that each of these steps is taken with reference to the declared welfare of individuals, while the specifics of particular cat cultures or cat-human cultures are forgotten [21,26]. From a population genetics perspective, consistent application of all three steps brings about a disruptive selection [20]. The problematic nature of this selection is aptly reflected in the question: “Is a world of sterilized feral cats and fertile cats of valued breeds desirable?” [23]. Natoli answers probably not, but adds that the successful sterilization of an entire urban feral cat population is unlikely anyway.The number of free-roaming cats in urban environments is difficult to estimate. In the 1990s, the number of owned cats in the USA was estimated at 60 million and the number of feral and stray cats at 25–40 million [56]. In theory, data from municipal shelters could be a useful clue to population development, but these data can significantly vary among individual regions (on growth in Denmark between 2004 and 2017, see [57]; on decrease in ASPCA shelters in New York from 1934 to 1994, see [58]). In addition, a positive trend does not necessarily mean a growth in the population of free-roaming cats, especially in higher-income areas where residents “abandon” fewer cats on the street, but rather hand them over directly to the care of shelters [57]. Neutering campaigns are certainly rational in regions where the number of cats is generally considered too high—mainly in urban areas. From the veterinary view of cat welfare, neutering is positively evaluated based on the outcomes of a longer life span and a reduction in behavioral problems toward humans. We must realize, however, that neutering campaigns appear reasonable only because of their limited impact (they generally do not reach lower income households or rural communities) [19]. Otherwise, there would inevitably be an accelerated decline in the population of common short-haired pet cats. Owners who do not have their cats neutered are labeled as irresponsible by animal rights activists and by people who think there are too many cats in a given area (e.g., bird watchers, conservation advocates, or hunters) [59]. However, if neutering were to become an obligatory standard, this viewpoint would change: these same people would be providers of kittens, which (due to regulations) would become scarce commodities. From a global perspective, such a situation sounds like science fiction, but in some countries (England, Finland), it is becoming a local reality [20,21]. The laissez-faire approach to the question of pet reproduction is informed not only by a different assessment of welfare (e.g., the satisfaction of instinctive desires), but also by what type of cat is favored by the majority of people. In the end, it is the common cat owners who favor domestic short-haired cats that answer Natoli’s question in the negative. In the long term, it is preferable and in fact necessary that the respective proponents of the veterinary and laissez-faire views agree on a sensible approach to neutering cats, which in urban environments are neither a purely domestic species, nor a species independent of humans. The issue of keeping pets exclusively indoors is proving to be a case of value conflict. On the one hand is a utilitarian perfectionist stance, subordinating the satisfaction of instinctive desires to overall quality of life (i.e., long lifespan) [18]. On the other hand is a laissez-faire approach, which prioritizes the fulfilment of psychobiological needs of animals (e.g., engaging in predatory behaviors) [46]. While proponents of the first approach argue for keeping pet cats exclusively indoors, the other side points out that pets who are able to move about outdoors exhibit fewer behavioral problems. In this discussion, regard for the welfare of cats is confusingly combined with the issue of the welfare of species preyed upon by cats. This study argues that in the case of populations in European cities, and even in the USA to a significant degree, the negative effect of cat predation is probably overestimated. Through centuries of coexistence, the local fauna has had the opportunity to adapt, and it is likely that the abundance of resources associated with human presence has led to the increased density of both cat and bird populations in Western cities [6,7].The issue of the cohabitation of humans and cats in urban areas is beset by conflicts between various interest groups and ideals of what the relationship between a cat and its owner should look like. Often forgotten is the diversity of the social life of cats and the values specific to each type of cat culture. We face the risk, then, that arguments on the welfare of cats are based on influential cultural representations rather than the interests of animals as members of groups distinguished as feral cats, free-roaming pets, and confined pets. Whether the wild side of cats is admired or perceived as a threat, people should accept it as a natural fixture of this semi-domesticated species. I am of the opinion that a greater awareness of the actual psychobiological needs of cats, including their reproduction, has the potential to clarify discussions on our mutual coexistence in urban spaces. Companion animal ecology should be developed, not only in close contact with veterinary and animal sciences, but should take into consideration the specific local situation, opinions, and habits of the individual participants in the debate [60]. Every city is a hybrid environment in which the mental worlds of humans (and their various interest groups) interact with those of cats and other domestic animals, as well as wild animals. The shape this coexistence takes is a matter of continual compromise, and the ideal of balance should be more important than the application of sweeping regulations tied to simplistic cultural representations that do not respect the complexity of ecological and social ties.

animals : an open access journal from mdpi

10.3390/ani11061508

PMC8224607

When a horse is diagnosed with a locomotor disorder, the veterinarian treats the specific injury to restore the horse to soundness. Even after the injury has healed, however, the horse may not be fully functional due to persistent limitations in movement or strength in specific areas of the body. As in people, rehabilitation seeks to optimize function and reduce any existing disability using a variety of methods including manual therapy, the use of physical and mechanical agents, and specialized exercise regimes. This study has reviewed the scientific literature with the goal of identifying which types of physical therapy have been described in horses over the past 20 years. The most frequently reported techniques were exercise, electrotherapy, and hydrotherapy but there are relatively few publications describing details of their use and outcomes in clinical cases. This study reviews the methodology and outcomes of rehabilitation in clinical cases. The results highlight the paucity of clinically-based reports on the practical applications of equine rehabilitation and physical therapy.

Injuries to the locomotor system are a common problem in athletic horses. Veterinarians address these injuries using appropriate medical, surgical, and pharmacological treatments. During or after recovery from the initial injury, horses may be treated for functional locomotor deficits using specific rehabilitation techniques aimed at restoring full athletic performance. This study reviews the literature to identify which rehabilitative techniques have been used most frequently in horses over the past 20 years, the protocols that were used, and the outcomes of the treatments in naturally occurring injuries and diseases. Publications were identified using keyword selection (Equine Athlete OR Equine OR Horse) AND (Rehabilitation OR Physiotherapy OR Physical Therapy). After removing duplicates and screening papers for suitability, 49 manuscripts were included in the study. The majority of publications that met the inclusion criteria were narrative reviews (49%) in which the authors cited the relatively small number of published evidence-based studies supplemented by personal experience. Observational/descriptive studies were also popular (35%). Randomized control trials accounted for only 10%. The most frequently reported rehabilitation techniques were exercise, electrotherapy, and hydrotherapy. The findings highlight the need for further information regarding type of intervention, parameterization, and outcomes of equine rehabilitation in clinical practice.

1. IntroductionRehabilitation has been defined as the optimization of function and reduction of disability in a patient suffering from a health condition (disease, disorder, injury, or trauma) [1]. Although this definition is based on the human patient, it can be applied equally well to the equine patient and, indeed, to veterinary rehabilitation in general. Thus, equine rehabilitation restores the incapacitated horse to its normal functional capacity and allows the athletic horse to perform at the expected level. Physical therapy employs physical methods to treat pain, disease, or injury by physical means [2]. Physiotherapy is a pseudonym for physical therapy, but since it is a protected term in some countries, this article uses the more general term physical therapy. Physiotherapeutic is the adjective that refers to physical therapy.Interest in equine rehabilitation and physical therapy is growing rapidly, e.g., [3,4]. As far as the authors are aware, there has not been a published review of the interventions used to rehabilitate horses or the suggested protocols and parameters for application of those interventions. One of the tools used to verify the extent of knowledge in a specific area is a scoping review. A scoping review combines the expertise of the author(s) with various literature searches to address the review topics in a much broader way then a systematic review [5]. Because of this, it serves to identify research gaps and to map the key concepts and main sources of the available evidence [5], and to suggest future research directions in that field.Thus, the aim of the present study is to address the question, “What are the most frequently used interventions in equine rehabilitation?”. Specifically, we wish to identify which interventions in the area of equine physical therapy and rehabilitation have been most popular in the field over the past 20 years, and to report suggested protocols for their use. 2. Materials and MethodsTo answer the research question, a scoping review was conducted following the framework proposed by Arksey and O’Malley [5]. This framework consists of five steps: (1) identify the research question; (2) identify relevant studies; (3) study selection; (4) chart the data; and (5) collate, summarize, and report the data [5]. 2.1. Identifying the Research QuestionThe proposed research question addresses the need to assess the published veterinary literature in the area of physical therapy and rehabilitation of horses, as a preparatory step towards a larger project being conducted by Lusófona University regarding equine rehabilitation techniques. The research question “What are the most frequently used interventions in equine rehabilitation?” defines two main search terms, rehabilitation and horse. Under the term rehabilitation, we searched for studies related to rehabilitation or physiotherapy or physical therapy, which are the terms used most often in this field of research, e.g., [3]. Under the term horse, we included the words horse or equine or equine athlete as being the most frequently used. The final keyword selection for the search was (Equine Athlete OR Equine OR Horse) AND (Rehabilitation OR Physiotherapy OR Physical Therapy).2.2. Identifying Relevant Studies2.2.1. Electronic DatabasesThe electronic databases used in our search were MEDLINE complete, Cochrane, and Science Direct Elsevier.2.2.2. Reference ListsReference lists were screened to determine whether any manuscript that met the review criteria was missing from the word search. The included manuscripts are indicated in the flowchart (Figure 1).2.2.3. Manual Search of Other SourcesSome grey literature was screened to check whether important data could be missing from our pool of information. It was also used to understand common word usage in this area as a tool to define keywords.2.3. Study Selection2.3.1. Inclusion CriteriaThe inclusion criteria were all studies (independent of the study type) that address equine rehabilitation from a physical therapy perspective, i.e., using physiotherapeutic methods and modalities for equine rehabilitation.2.3.2. Exclusion CriteriaAll studies addressing other types of intervention (osteopathic, chiropractic, acupuncture, and derivates) were discarded, since the focus was on physical therapy. In addition, studies addressing the benefits of interventions made on sound horses or in vitro models of disease were discarded.Figure 1 illustrates the manuscript selection process at the different levels. The study selection was performed by three of the authors at different levels of the process. A total of 49 studies were selected for inclusion. Six of these presented only partial information, as they were mainly concerned with other intervention methods that were considered as an exclusion criterion. The reasons for exclusion will be explained further in the results and discussion sections.2.4. Charting the DataAfter manuscript selection and reading the full text, the final selected manuscripts were charted by key information, including the following: Manuscript authors and dateType of intervention and comparator (if applicable)PopulationsAimsMethodsSince we included all types of studies (exploratory, descriptive, clinical trials, etc.), not all the fields were applicable to all the manuscripts. To facilitate the reason why some fields could not be used for a manuscript, we added the “type of study” field in the information collected.Information screening was conducted by only one of the authors to ensure consistency. 2.5. Collating, Summarizing, and Reporting DataThis phase involved presentation and overview of the review outcome. This is presented in a narrative way according to the key themes identified during the review process, which were related to the interventions used to rehabilitate the equine patient (athlete or non-athlete). Those interventions are divided into manual therapy (including tissue mobilization and joint mobilization techniques), physical agents (including kinesiology taping, thermotherapy, hydrotherapy, and electrotherapy options), and exercise therapy.3. Results and DiscussionThe synthesis table is presented in Table 1. The manuscripts that were included were oriented towards rehabilitation from the perspective of physical therapy. However, manuscripts that include complementary approaches in addition to physical therapy as part of the rehabilitation program were considered to contribute to the goal of the present review. The results are described according to the methods of intervention, but first we provide an overview of the types of studies included.3.1. Type of StudiesFrom the 49 studies included, the most common type is the narrative review which we defined as a review performed with the aim of describing important themes in equine rehabilitation, but without following a systematic data approach and relying to a large extent on the authors’ experience. The second most frequent is observational/descriptive studies, which describe the effects of a treatment option that was carried out frequently in a veterinary clinic. Some surveys and case reports were also found. The number of clinical trials or randomized clinical trials was very small. Figure 2 shows the distribution of the type of studies in the included manuscripts.3.2. Manual Therapy-Based InterventionsBy definition, manual interventions involve applying the hands to the patient’s body for diagnostic or therapeutic purposes. Manual therapy may involve passive stretching, soft tissue mobilization, or joint mobilization to restore the range of motion. As an example of the potential application of manual therapy it has been shown that when sound horses had one fore fetlock joint immobilized in a cast for 7 weeks followed by cast removal and 8 weeks of progressively increasing exercise, the treated fetlock retained 20% reduction in range of motion at the end of the study [54]. Clinical cases involving contracture or limitation of the range of motion after injury or post-surgically may benefit from manual therapy.3.2.1. Passive Stretching Frick [19] reviewed the use of stretching exercises to improve range of motion, prevent injury, and decrease pain. She presented indications and protocols for stretching exercises such as a series of rear limb stretches (hind limb protraction; quadriceps extension—hind limb retraction; hind limb crossover-to stretch adductors; pelvic rocking). It was recommended that each stretch be performed for 3–5 min, once daily, on 3–7 days per week to provide an adequate stimulus. In a review of manual therapies for pain management, Haussler [20] also included a description of the use of stretching for soft tissue restriction and joint stiffness. In a different study, Haussler [35] reviewed joint mobilization and manipulation in the management of the equine athlete and indicated that it was significantly more effective to hold a stretch for 30 s than for 15 s.3.2.2. Tissue MobilizationTissue mobilization includes the techniques of massage, myofascial release, and neural tissue mobilization to break down myofascial adhesions such as scar tissue, to move blood and tissue fluids, and to relax muscle tension and optimize muscle function. The narrative review of Bromiley [6] described massage techniques performed in equine rehabilitation settings, including effleurage, petrissage, tapotement, friction, and skin rolling. Sessions of 20 to 30 min were recommended to treat back pain with the outcome seeming to benefit from the inclusion of passive mobilization. Ridgway and Harman [7] also recommended massage to treat equine back problems but did not define parameters for its use. In the 2006 narrative review of Buchner and Schildboeck [12], none of the cited literature supported a physiological effect associated with general massage techniques but indicated a promising physiological effect in manual lymphatic drainage. On the other hand, Goff [13], stated that massage techniques and neural mobilization were indicated in soft tissue and neural tissue problems, but no parameters were given. A narrative review dedicated to massage therapy [16] describes different techniques (effleurage, circular friction, muscle pressure and shaking, skin manipulation, tapotement, petrissage, cross-fiber massage, wringing, classical Swedish techniques, and stroking) as described in the literature, and sometimes associated with stretching, as being beneficial when applied in 10 to 20 min sessions. Haussler [20] described the benefits of massage in alleviating muscle hypertonicity, soft tissue restrictions and pain, and the value of soft tissue mobilization for soft tissue restrictions and pain. A descriptive clinical trial in which the application of effleurage was interspersed by a 3 × 30 s circular kneading for 30 min significantly increased passive and active hind limb protraction [22]. A case study [24] that included massage in a multi-modal physiotherapeutic approach to a case of tetanus reportedly produced a good result but without describing massage parameters. At 6 and 12 months follow up, the foal did not have any deficits. In a survey of rehabilitation modalities used in horse treatments, massage was used by 69% of the respondents [49]. 3.2.3. Joint MobilizationJoint mobilization applies a force manually to induce passive physiologic or accessory movements, and active mobilizations of joints. Each joint should be moved in a specific manner, so this technique is best performed by trained professionals. Mobilization techniques include small rhythmic oscillations and gliding movements across the joint directed perpendicular or parallel to the joint’s normal direction of movement to improve motion and normalize joint function with a consequent reduction of stiffness and pain [20]. The first manuscript that refers to this approach is Bromiley [6], in which passive movement is recommended as a good intervention for back problems when associated with massage techniques. Porter [11] recommended 10 repetitions of passive range of motion in the normal physiologic range as a good intervention for joint diseases in horses. Joint mobilization was described as an effective approach for articular, neural, and muscular structures [13]. In rehabilitation of equine articular structures, the recommended techniques were passive mobilization at different amplitudes, velocities, and positions within the available range of motion, integration of both physiological and accessory movements, and the integration of passive accessory mobilization with active movement [13]. Haussler [14] indicated the value of joint mobilization in cases of joint stiffness and pain. The same author described joint mobilization and manipulation as important in equine treatment; mobilization was recommended for more generic use, whereas manipulation was more effective in specific conditions [14]. Haussler [35] defined grade 1–2 mobilizations as being characterized by slow oscillations within 25–50% of range of motion and grade 3–4 as being close to the end feel of the joint. A randomized clinical trial using 24 horses was performed to determine the effect of spinal manipulation on vertebral stiffness when added to spinal mobilization [21]. The authors used rhythmic spinal mobilization at five intervertebral sites within the thoracolumbar region. Vertebral stiffness decreased, and there was a further incremental improvement with the addition of spinal manipulation. In a case study of a radial fracture, manual passive physiologic mobilization starting 8 weeks post-surgery seemed beneficial, but no treatment parameters were reported [28]. Guedes described manual therapy and movement as elective techniques in painful conditions but did not present any parameters of usage [37]. A survey published in 2018 indicated that range of motion therapies were reportedly used in 71.9% of the responses [49].3.3. Physical and Mechanical Agents3.3.1. Kinesiological Taping and BandagesThis section includes all interventions that use bandages or taping, with only three manuscripts mentioning these approaches. The first was Goff [13] who added kinesiology tape to manual therapy approaches but did not present any indications for its use. In a research study, kinesiology tape was applied with the FAN technique at 10% tension for 72 h following tibio-patellofemoral joint arthroscopy. Compared with operated but untaped controls, there was a significant reduction in swelling from 24 to 72 h post-surgery [40]. In the survey of Wilson et al. [49], kinesiology taping was part of the intervention in 33% of cases, whereas compression bandages were used in 89.5% of rehabilitation procedures.3.3.2. Electrotherapy InterventionsAccording to Bromiley in 1999 [6], common equine electrotherapeutic rehabilitation options were magnetic field therapy (applied with a blanket); transcutaneous electrical nerve stimulation or TENS (once daily for 20–30 min); therapeutic ultrasound (maximal suggested parameters of 1.0 to 0.5 W/cm2 with lower powers seeming to be more effective, 3–5 min daily to a maximum of 20 treatment sessions followed by 3 weeks rest); and low-level laser (maximal dosage of 10 J/cm2). In a study to assess the efficacy of iontophoresis as a drug delivery option for articular disease, one group received a single intraarticular injection of 4 mL dexamethasone-phosphate (6 mg/mL), and the second group received iontophoretic administration of dexamethasone-phosphate (6 mg/mL) at 4 mA for 40 min in the treated limb and at 0 mA for 40 min in the contralateral (control) limb. Blood and synovial fluid were evaluated but no drug delivery by electrophoresis was detected [8]. The benefits of electrical stimulation modalities for rehabilitation of equine joint disease include laser therapy (general report with human-based studies, no parameterization for horses) and therapeutic ultrasound (describes methods, no parameterization recommended) [11]. A 2006 narrative review included information about electrotherapy (TENS) (evidence based on humans and dogs, no parameterization reported), magnetic field therapy (with no conclusive evidence and no parameterization), laser therapy (presenting evidence of the low power option in horses without parameterization), and therapeutic ultrasound (with impressive evidence but no parameterization) [12]. Indications for electrotherapy were presented, including therapeutic ultrasound for which the authors discuss the relative merits of normal to long-wave ultrasound in regard to tissue depth penetration, but without information regarding parameters. This paper also included TENS and iontophoresis, again without suggested parameterization [15]. A narrative review described the mechanics of actions, indications, and contraindications for use, and treatment protocols for electrotherapies including neuromuscular electrical stimulation, pulsed electromagnetic field therapy, therapeutic ultrasound, extracorporeal shockwave treatment, laser therapy, and whole-body vibration [36].In 2018 it was reported that shockwave therapy was part of the treatment in 72.4% of the common procedures, vibration in 39.6%, class 4 laser in 39.9%, therapeutic ultrasound in 39%, class 3 laser in 34.3%, neuromuscular electrical stimulation in 31.8%, TENS in 29.2%, and pulsed electromagnetic field therapy in 22.9% [49]. Magnetic Field TherapyIn 2017, Guedes described magnetic field therapy as a common option for pain management but without any parameterization [37]. Pulsed electromagnetic fields (PEMFs) applied with a blanket were used in a randomized clinical trial in 20 polo ponies with back pain [30]. A placebo blanket was applied to horses in the control group. Using the parameters shown in Table 1, the results failed to indicate significant differences between the PEMF intervention and placebo groups [30].Radial Pressure Wave TherapyA descriptive study of radial pressure wave therapy applied according to the manufacturer’s recommendations was reported in 65 horses with recurrent proximal suspensory desmitis [10]. Horses received three treatments at 2-week intervals with the parameters shown in Table 1. After the treatments, they performed a controlled exercise program. It was reported that this therapeutic option seemed to provide better results than the placebo [10]. Extracorporeal Shock Wave Therapy (ESWT)Parameters for ESWT have been recommended by Kaneps [3] for soft tissue and bone injury (tendinitis, desmitis, osteoarthritis, deep muscle pain), and some proposed protocols were as follows: Impulses: small lesions, such as a collateral ligament desmitis of the distal interphalangeal joint, require 1000 impulses per treatment site. Suspensory desmitis lesions most often receive 2000 impulses/treatment. Large areas of the back may require 3000 impulses/treatment.Energy levels: soft tissue injuries <4 cm deep to the skin: 0.2–0.35 mJ/mm; soft tissue and bone in the heel region: 0.35–0.45 mJ/mm; back disorders: 0.4–0.5 mJ/mm; bucked shins and incomplete fractures: 0.35–0.55 mJ/mm; osteoarthritis: 0.15–0.3 mJ/mm; wounds: 0.1–0.15 mJ/mm.Focus depth: the focal point for ESWT should be the average depth of the lesion from the skin. Some ESWT devices use gel standoffs to focus the energy at the required depth; others use hand pieces with different focal depths.For lameness, Contino [42] recommended 2000 pulses to be given with a 12 mm head followed 2 weeks later by 1500 pulses. Compared to the administration of intramuscular polysulfated glycosaminoglycan (PSGAG, 500 mg every 4 days for seven treatments), EWST presented better results and was particularly beneficial for low-motion joints and enthesopathy at the joint capsule insertion [42]. Extracorporeal shockwave therapy was included in the conservative management of foot problems in horses, but without reporting parameterization of usage [43].A study was designed to assess the effects of ESWT on mechanical nociceptive thresholds and cross-sectional area of multifidus muscle in 12 horses with thoracolumbar pain [53]. The authors reported an increment in the mechanical nociceptive threshold that was more evident in the thoracic region but no significant changes in multifidus cross-sectional area.Therapeutic UltrasoundA narrative review [3] described the use of therapeutic ultrasound for heat production with a 1 MHz transducer for deeper penetration or 3 MHz for superficial penetration. Energy levels were from 1 to 2 W/cm2 applied as a continuous wave for 10 min. Low-intensity ultrasound may be applied once daily for 2 to 3 h in acute injuries and 4 to 6 h in chronic injuries. The device was set at 2.75 MHz at 0.85 W/cm2 without the possibility of adjustment. Non-invasive, low-frequency ultrasound has been explored for the treatment of lameness, including parameters for its use in acute, subacute, and chronic injuries [50]. Protocols for habituation and treatment were defined according to injury, using different transducers for pulsed and continuous emission, and with different shapes for different wave emission. Recommended protocols for different stages of injury are as follows: Acute protocol: week 1: 6 days pulsed emission, flat transducer, 70–80% full power for 10 min; week 2: 6 days pulsed emission, flat transducer, 95% full power for 5 min; weeks 3 and 4, if necessary based on clinical and ultrasound examination, 3 times per week continuous emission, flat transducer, 70–85% full power for 6 min followed by pulsed emission, flat transducer at 80–95% full power for 6 min, followed by pulsed emission, convex transducer at 80–95% full power for 5 min.Subacute protocol: week 1: 6 days pulsed emission, flat transducer, 70–80% full power for 10 min; week 2: 3 days pulsed emission, flat transducer, 80–95% full power for 6 min followed by pulsed emission, convex focused transducer, 80–95% full power for 5 min; week 3 and 4 (if necessary by clinical assessment), 2 days continuous emission, flat transducer, 70–80% full power for 6 min followed by pulsed emission, flat transducer, 80–95% full power for 6 min followed by pulsed emission, convex focused transducer, 80–95% full power for 5 min.Chronic protocol: week 1: 6 days continuous emission, flat transducer, 70–80% full power for 6 min followed by pulsed emission, convex transducer, 80–95% full power for 5 min; week 2: 3 days continuous emission, flat transducer, 70–80% full power for 6 min followed by pulsed emission, flat transducer, 80–95% full power for 6 min followed by pulsed emission, convex focused transducer, 80–95% full power for 5 min; weeks 3 and 4 (if needed): 2 days continuous emission, flat transducer, 70–85% full power for 6 min followed by pulsed emission, flat transducer, 80–95% full power for 6 min followed by pulsed emission, convex focused transducer, 80–95% full power for 5 min.A beneficial result was observed in the cross-sectional area of the suspensory ligament after a mean treatment duration of 3.3 weeks [50]. Laser TherapyProposed parameters for laser therapy include the recommended dosage of 4–12 J/cm2 [3]. A study of high-power laser therapy in 150 lame sport horses used the manufacturer’s pre-established protocol for the area of injury [46]. Horses received 250 J/cm2 for 20 min daily for 2 consecutive weeks with an exercise rehabilitation protocol and/or pharmacological treatment prescribed at the veterinarian’s discretion. The results suggested a beneficial contribution of high-power laser [46].Whole Body VibrationA pilot study of the benefits of vibration therapy used a frequency of 50 Hz for 30 min, using a control group with sham procedures, but no differences were found in lameness or gait abnormalities [45]. The use of whole-body vibration for 60 days, five sessions a week for 30 min in horses with chronic lameness, did not produce immediate or long-term benefits [38]. 3.3.3. Thermal TherapyThermal therapy is the application of heat (thermotherapy) or cold (cryotherapy) to the skin to change the temperature of the cutaneous, intra-articular, or other soft tissues as an adjunctive form of therapy for treatment of musculoskeletal and soft tissue injuries. The application of heat may increase skin and joint temperature, improve circulation, relax muscles, and reduce joint stiffness. Additionally, deep heating may decrease the sensitivity of nerves and muscle spindles. Cold applications may reduce pain, decrease swelling, constrict blood vessels, and block nerve impulses. Heat and cold are important treatment methods because they are, for the most part, inexpensive and easy to apply by lay people with only a few precautions. The first included manuscript referring to thermotherapy approaches was Ridgway and Harman [7], who described the use of heat and cold therapy for back problems but did not mention any parameters for usage. On the other hand, Porter‘s narrative review [11] described using cold therapy for 30 min several times daily, with more frequent treatments during the first 72 h of the acute phase. Buchner and Schildboeck [12] stated that prolonged or repeated ice water cooling is particularly effective in the equine limbs, while heat affects primarily superficial tissues. No information was provided about parameterization. In a narrative review, Paulekas and Haussler [15] indicated cold therapy should be performed for 10–20 min every 2–4 h during the first 48 h post injury and was most effective if applied immediately after injury. They described superficial heat therapy as being beneficial if applied by hot packs at 75° or immersion in hot water. There was no information about water temperature or duration of treatment in the immersion option. In another narrative review [3], some parameterization and usage descriptors were presented, and cryotherapy was stated to be particularly effective in the first 24–48 h after injury for reduction of inflammation and edema and pain control. Ice water immersion, or the application of ice or cold packs for 15 min was recommended. Heat therapy was recommended in chronic injuries when it is usually applied for 20–30 min by warm water irrigation, hot packs, or leg sweats [3]. Ice and heat were described as effective methods for pain management, without parameters for their use [37]. A retrospective study of 285 horses described the use of sleeve-style digital cryotherapy to treat distal limb pathological conditions. Three types of application were described: continuous (211 horses), interrupted (57 horses), and intermittent (17 horses) [47]. The cryotherapy was conducted for more than 12 h. The incidence of injury related to this approach was of 7% [47]. Thermal therapy was reported to be used frequently in the 2018 survey [49] in which ice was used in 95.2%, a cold water circulation machine in 48.5%, and heat in 77.6%.3.3.4. HydrotherapyHydrotherapy uses the physical properties of water, including temperature, pressure, viscosity, and buoyancy for therapeutic purposes that include relieving pain, stimulating blood circulation and treating diseases such as arthritis, muscular diseases, and nerve diseases.Bromiley [6] referred to hydrotherapy as a common intervention for back problems but did not present parameters of usage. However, swimming is contra-indicated in horses with back pain or diseases due to the fact that they swim with the neck raised and the thoracolumbar spine extended (hollowed) [15]. Porter [11] defined hydrotherapy to include swimming for active exercise and use of an underwater treadmill for active assisted exercises. She recommended a temperature of 62 °C for treating joint diseases but without further information on parameters of usage. Some evidence for the use of hydrotherapy to treat limb injuries has been reported but, again, parameters of use are missing [8]. Swimming for 100–500 m was found useful in the later stages of the rehabilitation program for a radial fracture [28]. There have been several surveys and reviews of the use of the water treadmill in which the horse walks or trots on a motorized belt in a chamber filled with water to a chosen level. In 2013, King et al. [29] described the potential value of the water treadmill in rehabilitation of horses with osteoarthritis and secondary musculoskeletal injuries. Published studies of the benefits of water treadmill therapy in osteoarthritic people and dogs were described, and the need for comparable studies in horses was highlighted. Subsequently, this research group published a clinical trial in which 16 horses without clinical abnormalities had bilateral arthroscopies of the middle carpal joints [39]. On one randomly selected side, an osteochondral fragment was created to induce mild osteoarthritis and low-grade lameness. Horses were randomly assigned to treatment and control groups (N = 8). All horses performed treadmill exercise at walk on 5 days/week for 8 weeks beginning on day 15 post-surgically. Initial duration was 5 min/day, increasing by 5 min/week to a maximum of 20 min at walk. Control horses exercised without water in the treadmill, treated horses had water to hip height. Evaluation of forelimb kinematics, symmetry of forelimb loading, activation of select forelimb muscles acting on the carpi, and degree of carpal joint flexion supported the value of the underwater treadmill in rehabilitation of experimentally induced osteoarthritis. A review of the use of treadmills in equine rehabilitation [41] presented evidence to support using an overground treadmill to rehabilitate horses with distal limb injuries and back pain. The effects of water treadmill exercise were considered in light of the fact that many horses suffer from multiple musculoskeletal issues, such as lameness and back pain, which may require different approaches. The authors emphasize that successful rehabilitation depends not only on choosing appropriate exercises for the individual case, but also on avoiding inappropriate types of exercise. They list conditions for which exercise on an overground treadmill or a water treadmill are contra-indicated. Guidance is offered regarding the selection of an appropriate depth on the water treadmill. The authors stress the importance of professional monitoring of each horse’s gait pattern during the period of rehabilitation.A survey published in 2018 investigated why, when, and how the water treadmill is used in equine rehabilitation using three questionnaires that were part of an international survey-based approach [48]. Rehabilitation was reported to account for only 40% of water treadmill use with training being a more popular use (60%). Respondents stressed the importance of adequate habituation to the water treadmill with different centers using an average of 2–3 sessions lasting 10–30 min in water to the depth of the hock or fetlock for this purpose. Significant differences were identified between training and rehabilitation sessions. Training sessions most often used water at hock height with horses walking for 20.5 min and trotting for 8.8 min on average. Water treadmills have been used most frequently in rehabilitation of horses with ligament and tendon injuries. The most common reasons to use the water treadmill for rehabilitation were tendonitis and suspensory desmitis (41%). Water height was most often from just above the fetlock to mid-cannon for rehabilitation sessions. Compared to training, rehabilitation used significantly shallower water, a faster walking speed, and shorter duration (p < 0.023 for all three variables). On average the treatment protocol included seven (range 0–14) exercise sessions per week. Therapeutic protocols were similar, but rehabilitation protocols varied significantly between venues [48]. Finally, hydrotherapy was reported to be one of the common interventions in 82.9% of the cases, and the water treadmill was used in 39% [49].The narrative review of Muñoz et al. [52] describes the application of therapy using the water treadmill in horses with injuries of the superficial or deep digital flexor tendons and their accessory ligaments and in back and joint diseases. The suggestions are backed up by biomechanical information. Control of treadmill speed and water depth are important components of the rehabilitation program. Its use is recommended for treatment of subacute and chronic tendon and ligament injuries and chronic osteoarthritis.3.4. Exercise TherapyIn the past, horses were often rested for a prolonged period of time during rehabilitation, but it is now recognized that it is preferable to return athletic horses to an appropriate type and level of exercise as soon it is safe to do so. Exercise therapy describes the use of specific gaits and movements to enhance the horse’s recovery from injury or disease.An early reference describing exercise as a therapeutic approach in equine rehabilitation was in Bromiley’s narrative review [6]. The types of exercise described include the treadmill, the horse walker in which horses walk in circles in individual pens that can be controlled for speed and direction, the use of long reins, weighted shoes starting at around 0.25 kg and increasing to around 2.1 kg, for at least 6 weeks, and exercise without the rider’s weight, including on gradients. The author did not present parameters for each type of exercise. Another study recommended the Tellington Touch Equine Awareness Movement exercises (TTEAM) that involve touching the horse with a “wand” to stimulate awareness of the body parts, together with negotiating mazes, picking up sticks, and star exercises. It also referred to the importance of cross-training in both training and retraining situations [7]. A retrospective study of 17 horses described the conservative management of tibial tuberosity fractures [9]. The progressive exercise program included stall rest for 2 or 3 months with hand-walking starting in the 2nd month and increasing gradually (no amount indicated). Small paddock turn out and walking under saddle (depending on the horse’s attitude) were advised either separately or in combination until the horse was sound at slow trot. The amount of training was then gradually increased [9]. In a study about the use of radial pressure wave therapy, a controlled exercise program was recommended. For the first 6 weeks, the horse was on box-rest with daily controlled walking exercise, gradually increasing from 30 to 60 min. Trotting exercise was introduced and gradually increased until week 10–12 [10]. In a narrative review about manual therapy in equine treatment programs, Haussler [14] stated the importance of adding stretching exercises (active baited stretches for the axial skeleton, passive stretching for the limbs) with stretches being held for 30 s as an important part of the manual therapy approach. A subsequent narrative review on therapeutic exercise included the use of theraband exercises, induced passive and active cervical bending exercises, spinal reflex movements, axial tail traction, walking through a labyrinth, the star obstacle, walking across elevated bridges, walking over ground poles and raised poles, negotiating raised cavalleti at slow trot, standing on a pedestal, stepping down from a pedestal, walking up and over a pedestal, walking uphill, walking downhill, walking uphill and downhill over poles, and backing uphill. No parameterization was presented [15]. Dynamic mobilization exercises or baited stretches involve having the horse follow the path of a treat or a target into specific positions of cervical flexion, extension, or lateral bending. In order to reach the desired positions while remaining balanced, the horse must activate the core musculature. A study of the effects of baited stretches in three neck flexion exercises were studied, namely chin-to-chest, chin-between-carpi, and chin-between-fore fetlocks, in eight riding school horses that were thought to suffer from back pain [17]. The greatest amounts of movement were observed in the most cranial and caudal cervical joints with smaller movements in the mid-cervical region and mid to caudal thoracic regions. A 12-week program based on dynamic mobilization exercises resulted in significant increases in cross-sectional area of the deep spinal stabilizer multifidus muscle on both sides of the spine at five vertebral levels from T10 to L5. Additionally, the muscle cross-sectional areas became more symmetrical on left and right sides [25]. The potential value of dynamic mobilization exercises was explored as a way to increase core strength and vertebral mobilization in horses with back problems [27]. Dynamic mobilization exercises involving cervical lateral bending (chin-to-girth, chin-to-hip, and chin-to-tarsus) showed increased lateral bending of the cervical and thoracolumbar intervertebral joints as the horse stretched further caudally [27].Another exercise-based therapeutic approach used lightweight (55 g) tactile stimulators attached loosely around the hind pasterns to increase the range of joint motion, re-educate the movement, and strengthen the muscles of the hind limbs. Skin stimulation of the pastern and coronet was thought to mimic the effects of the tripping reflex. Speed and stride duration did not differ between conditions, but when stimulators were present, stance duration decreased, swing duration increased, and peak height of the hind hooves increased due to greater flexion of the stifle, tarsal, metatarsophalangeal, and distal interphalangeal joints during the swing phase [18]. A subsequent study [23] compared the effects of four types of pastern stimulators (10 g strap, 55 g tactile stimulators, 700 g weights, 700 g weights with tactile stimulators added) on trot kinematics. At the same trotting speed, stance duration was shorter and swing duration correspondingly longer with all stimulators except the strap. Peak hoof height was significantly higher with tactile stimulators and with weights, and there was a further significant increase in peak hoof height with the combined tactile-weighted stimulator. Individual horses varied in their responses, and the shape of the hoof flight arc differed between stimulators. Overall, the use of pastern stimulators increased swing phase flexions of the stifle, tarsal, and fetlock joints during trotting and can serve as a rehabilitation tool when flexion of these joints has been reduced, for example, by immobilization. It was recommended that the use of lightweight tactile stimulators should precede the use of weights.A study was based on the use of acoustic myography [44] in eight horses in which the superficial gluteal muscle in the left hind limb was slightly but significantly weaker as determined by having a higher ESTi score when the horses circled to the left. (The ESTi score is a measurement specific to acoustic myography for assessing muscle function). Horses were trained every third day for one hour wearing a lightweight (82 g) bell boot on the left hind limb. After training with the bell boot for 6 weeks, acoustic myography showed that the asymmetry in the left hind limb on the left circle had decreased, but on the right circle an imbalance had developed that was thought to represent on over-compensation.A randomized clinical trial assessed whether the contribution of dynamic mobilization exercises and gymnastic training improved the quality of the walk stride and epaxial muscle size in nine hippotherapy horses [31]. Horses that performed dynamic mobilization exercises (cervical flexion, extension, and lateral bending to both sides) showed a significant increase in cross-sectional area of multifidus muscle and a non-significant increase in thickness of longissimus dorsi. Horses that also performed gymnastic exercises (pelvic tilt, backing up, walking tight circles, stepping over a raised pole) showed increases in stride length and tracking distance at walk [31].The same dynamic mobilization exercises, together with core strengthening exercises and balancing exercises, were proposed for use in horses with back pain in a narrative review [32]. This review also explored the benefits of a range of exercises at different gaits and speeds, on circles and gradients, jumping, poles, and unstable footing in rehabilitation of horses with back pain. When available, evidence-based research to support the use of specific exercises was included. A narrative review on rehabilitation assessment and interventions described exercise as one of the most frequently used interventions [33]. It described essentially the same dynamic mobilization, core training, and balancing exercises as the previous review [32] and also recommended exercise at different gaits and speeds, spiraling in and out on circles, changes of gait and speed, use of gradients to selectively load hind or forelimbs, jumping, poles, and unstable footing. A case study of a foal with tetanus included exercise as part of the treatment protocol, starting in weeks 3–4 [24]. During this phase, a walking frame was used to assist the foal with standing and walking and to allow longer periods of weight-bearing. In weeks 5–6, exercises were included to overcome residual problems, in particular the foal’s inability to raise and lower itself independently. Follow-up at 6 and 12 months did not reveal any deficiencies assessed by normal physical examination [24]. An observational study [26] evaluated the effect of athletic conditioning on degenerative suspensory ligament desmitis (DSLD) using six horses exercised on a treadmill for 30 min every other day at an average heart rate close to the anaerobic threshold (Table 1). The results showed that vertical impulse increased after 8 weeks of exercise and 4 months of pasture rest in DSLD-affected horses. The suspensory ligament fiber pattern subjectively improved with exercise in affected horses. Insulin levels significantly decreased from baseline in all horses after 4 and 8 weeks of exercise. The authors conclude that exercise did not seem to exacerbate and may have improved signs of DSLD in mild to moderate cases [26]. A case study that included exercise consisting of hand-walking and longing on a firm surface (5 to 20 min) in the rehabilitation of a radial fracture reported good results [28]. Davidson [34] presented a narrative review to describe controlled exercise protocols commonly used as part of the rehabilitation process to promote healing after muscle, bone, tendon, and ligament injury. Based on the author’s experience and some supporting literature, the following recommendations were made:Muscle injury: week 1: stall rest; week 2: stall rest, walk 15 min; week 3: stall rest, walk 30 min; week 4: stall rest, walk 30 min, trot 5 min; week 5: stall rest, walk 20 min, trot 10 min; week 6: stall rest, walk 20 min, trot 20 min; week 7: stall rest, walk 20 min, trot 20 min, canter 5 min; week 8 onwards: small paddock turn out (6 × 6 m), gradually increase exercise to full training.Bone injury: week 1–4: stall rest; week 5–6: stall rest, walk 15 min; week 7–8 (with radiographic evaluation of bone healing): stall rest, walk 30 min; week 9–16: small paddock turn out (6 × 6 m); week 16 onward: gradually increase exercise to full training.Tendon/ligament injury: week 1–2: stall rest; week 3–4: stall rest, walk 5 min; week 5–6: stall rest, walk 10 min; week 7–8: stall rest, walk 15 min; week 9–10 (with lameness and ultrasound examination): stall rest, walk 20 min; week 11–12: stall rest, walk 25 min; weeks 13–14: stall rest, walk 30 min; weeks 15–16: stall rest, walk 35 min; week 17–18: stall rest, walk 40 min; weeks 19–20: stall rest, walk 40 min, trot 2 min; weeks 21–22: stall rest, walk 35 min, trot 5 min; weeks 23–24: stall rest, walk 30 min, trot 10 min; weeks 25–26: stall rest, walk 25 min, trot 15 min; weeks 27–28: stall rest, walk 20 min, trot 20 min; weeks 29–30: stall rest, walk 20 min, trot 20 min, canter 1 min; weeks 31–32: stall rest, walk 20 min, trot 20 min, canter 5 min; weeks 33–34: stall rest, walk 20 min, trot 20 min, canter 10 min; weeks 35–36: stall rest, walk 15 min, trot 20 min, canter 15 min; weeks 37–38: stall rest, walk 10 min, trot 20 min, canter 20 min; weeks 39–42: small paddock turn out (6 × 6 m), full flat work, no speed work or jumping; weeks 42 onward: small paddock turn out (6 × 6 m), full flat work, gradually introduce speed work or jumping [34]. Kaneps [3] described common rehabilitation approaches to surgical or medical equine conditions. In the exercise portion, he started with walk for 5 to 10 min once or twice a day with incremental increases based on observation of the horses’ soundness. Trot started only after 10–15 min hand walking for warm-up and was initially for short periods of 1–1.5 min. This author emphasized the need for exercise to be oriented toward the needs of the horse’s usual activity. In a retrospective study with 150 horses, exercise was, in some cases, part of the protocol used to test the effectiveness of high-power laser therapy [46]. The exercise program progressed as follows: week 1–2: walk 20 min on hard surface; week 3–6: walk 20 min on soft surface; week 7–10: walk 20 min on soft surface and introduce trotting increasing by 2 min per week; week 11–15: minimum 20 min walk, trot increasing 2 min per week and canter increasing 2 min per week; week 15–18: walk minimum 20 min, trot and canter as in normal flat work, and with increasing dressage exercises or jumping up to full workload. Based on 129 horses, the median time to return to the previous performance level was 6 months In a 2018 survey, exercise was found to be used frequently as part of the rehabilitation process in the form of controlled hand walking in 97.3%, as therapeutic exercises in 84.3%, by means of stretching in 83.3%, using an automatic horse walker in 56.7%, by the application of the Pessoa®® lunging system in 46.2%, using a land treadmill in 39.9%, and with an Equiband in 27.4% of cases [49]. Finally, a retrospective analysis of 62 horses that survived at least 12 months after colic surgery (11 treated, 51 controls) investigated whether a 4-week program of core abdominal rehabilitation exercises (CARE) hastened postoperative recovery and allowed a more rapid return to training [51]. The exercise program started 4 weeks after surgery with dynamic mobilization exercises (chin-to-chest, chin-to-girth), sternal, withers and thoracic, lumbar and lumbosacral lifting; and caudal tail shift. Dynamic mobilization exercises performed in lateral bending and chin-between-fetlocks were added in week 2 with gradual increases in the number of repetitions of the exercises that were maintained through weeks 3–4. The CARE horses returned to work under saddle faster (median 60 days) compared with control horses (median 90 days) and were more likely to compete in some form of sport post-surgically (10/11 CARE, 24/51 controls). All CARE horses completed the program without complications, and they returned to work and training within a significantly shorter time than controls. It was concluded that core abdominal rehabilitation exercises could be safely performed after colic surgery and appeared to facilitate a faster recovery and return to work.4. ConclusionsOverall, there is a lack of randomized clinical trials using large samples that can help describe evidence related to the different approaches cited. The large representation of narrative reviews and observational/descriptive studies, mostly based on the personal experience of the authors or citing the same results of the few studies conducted, needs to be supplemented by rigorously conducted, evidence-based research. Exercise, physical agents, and hydrotherapy appear to be the most commonly used options, but much of the information regarding their potential efficacy is based largely on the results of human studies. Some studies present options and parameterizations that can be useful for equine clinical practice, but it is clear that more evidence is needed with regard to parameters for use and efficacy of different rehabilitation methods in horses.

animals : an open access journal from mdpi

10.3390/ani11071920

PMC8300297

In Australia, feeding grazing dairy cows concentrate and forage supplements is common. Dairy farmers face the challenge of profitably feeding their cows in situations where there is significant variation in feed costs and milk price. We used the results of grazing experiments to develop equations that predict the yield of milk fat and milk protein when different combinations of concentrates and pasture + forage are fed to grazing lactating dairy cows. We applied economic principles to these predictions to estimate the optimal combination of these feeds for given costs and prices. Feed is the largest variable cost in dairying. The allocation of pasture and supplements that are based on better estimates of milk responses to supplements should lead to increased profit for farmers.

Feed is the largest variable cost for dairy farms in Australia, and dairy farmers are faced with the challenge of profitably feeding their cows in situations where there is significant variation in input costs and milk price. In theory, the addition of 5.2 MJ of metabolisable energy to a lactating cow’s diet should be capable of supporting an increase in milk production of one litre of milk of 4.0% fat, 3.2% protein and 4.9% lactose. However, this is almost never seen in practice, due to competition for energy from other processes (e.g., body tissue gain), forage substitution, associative effects and imbalances in rumen fermentation. Pasture species, stage of maturity, pasture mass, allowance and intake, stage of lactation, cow body condition and type of supplement can all affect the milk protein plus fat production response to additional feed consumed by grazing dairy cows. We developed a model to predict marginal milk protein plus fat response/kg DM intake when lactating dairy cows consume concentrates and pasture + forages. Data from peer reviewed published experiments undertaken in Australia were collated into a database. Meta-analysis techniques were applied to the data and a two-variable quadratic polynomial production function was developed. Production economic theory was used to estimate the level of output for given quantities of input, the marginal physical productivity of each input, the isoquants for any specified level of output and the optimal input combination for given costs and prices of inputs and output. The application of the model and economic overlay was demonstrated using four scenarios based on a farm in Gippsland, Victoria. Given that feed accounts for the largest input cost in dairying, allocation of pasture and supplements that are based on better estimates of marginal milk responses to supplements should deliver increased profit from either savings in feed costs, or in some cases, increased output to approach the point where marginal revenue equals marginal costs. Such data are critical if the industry is to take advantage of the opportunities to use supplements to improve both productivity and profitability.

1. IntroductionFeed is the largest variable cost for dairy farms in Australia, and dairy farmers are faced with the challenge of profitably feeding their cows in situations where there is significant variation in input costs and milk price [1]. Pasture is generally considered the cheapest source of nutrients [2], and while there is significant variation in the growth rate of pastures throughout the year, farmers can conserve excess pasture as hay and silage to feed back to the herd in times of pasture deficit. Concentrate supplements are also commonly fed to increase stocking rate and overcome deficits in pasture supply [1].Heard et al. [2] reported new empirical models that predicted the quantitative relationship between milk yield (and milk protein and milk fat yield) and dry matter intake of cereal-based supplements by grazing dairy cows in Australia. Such models are also known as production functions. The models reported by Heard et al. [2] were developed using meta-analysis techniques, and were subsequently employed by Ho et al. [1] to demonstrate the value of applying marginal economic theory to make on-farm, profitable and tactical concentrate feeding decisions. However, as the meta-analysis had only included results from experiments in which grazing cows were fed cereal-based supplements, these models were of limited use because they could not be applied in situations where the cows’ diet also included supplementary hay and silage. While it is difficult to know exactly what proportion of dairy farmers feed their lactating herd both supplementary concentrates and forages as part of the milking ration, on average, hay and silage made up 34% of the total tonnes of DM consumed on the milking area of the farms contributing to the 2019/2020 Dairy Farm Monitor project (C. Waterman pers comm.).We reasoned that empirical models could be developed that include situations where grazing dairy cows are fed both cereal grain and hay and silage (forage) supplements using meta-analysis techniques. These models would be production functions, which could then be combined with production economics principles to determine the optimal combination of feeds for different on-farm costs and prices. This would support farmers in making more profitable choices between alternative feeds in a tactical setting (weekly, monthly or seasonal timeframe). Here, we report on the meta-analysis and the resulting empirical production function model, and demonstrate the application of production economics to determine the most profitable combination of inputs to feed lactating, grazing dairy cows in southern Australia.2. Materials and MethodsData from short-term experiments conducted in Victoria that involved dairy cows grazing pasture and fed supplements (both cereal grains and forages) were collated ([3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27], Table 1). Experiments were included if they met the following criteria: 1. At least two rates of supplement (concentrate and/or forage) were included, 2. For all treatments, the average daily pasture and supplement dry matter intake (DMI) data per cow were available and 3. Daily milk protein and milk fat concentrations and/or yield were measured. In total, this represented 241 lines of data (equivalent to 241 different treatments). A large number of variables and measures from each experiment were included in the database. Their relative contribution to estimates of the marginal milk protein plus milk fat yield response could then be determined via statistical analysis and probabilistic techniques. All variables were included in the database at the outset and disregarded in the statistical analysis as appropriate, rather than collecting an incomplete dataset and overlooking what may be a key variable. The structure of the database allowed for 320 individual parameters to be included, most of which were directly available or calculated from the results of each experiment. This amount of detail allowed a thorough investigation of the factors that most influence marginal milk responses.The data consisted of treatment means within replicated experiments. Experiments reported in some of the publications had two-factor factorial treatment structures for which only the main effects were reported, since interaction effects were non-significant. In these cases, individual treatment means were estimated, assuming additivity of main effects, in the following way:(1)μij=μ0+Ai+Bj where Ai is the main effect of level i of one factor and Bj is the main effect of level j of the other factor. This method of predicting treatment means from the reported main effects was employed for each dependent variable, milk yield and milk fat and protein composition and yields.Individual variation in factors such as seasonal conditions, number of days lactating, body condition score, pasture composition, pasture mass and allowance and amount of supplement consumed by cows meant that no two experiments were the same. The aim of the meta-analysis was to determine the contribution of the input variables to the prediction of milk protein and milk fat yield (kg/cow·day), and to derive predictive equations based on readily available observations of a pasture-based dairy system.A meta-analysis was performed on the data within the database using a mixed effects, random-coefficients model in which the fixed effects described and tested relationships between covariates and production variables. Relationships between covariates and milk protein and milk fat yield (kg/cow·day) were tested. In this analysis, we aimed to derive a two-variable (bivariate) quadratic relationship between yield (milk protein and milk fat) and concentrate and pasture + forage intake. This was so that the resulting models could be coupled with production economics to help farmers determine the optimal combination of feed inputs for a given situation. A bivariate quadratic equation has the following form:(2)Y=a0+a1X1+a2X2+a11X12+a22X22+a12X1X2 where Y = yield (milk protein + fat, kg/cow·day)X1 = concentrate intake (kg DM/cow·day) andX2 = pasture + forage intake (kg DM/cow·day)The first step in the meta-analysis was data exploration using the Lattice package in R software [28], and subsequently, all models were fitted to data using residual maximum likelihood software in Genstat 18 [29]. Production data (yields of milk protein and fat) were plotted against pasture DMI, forage DMI, concentrate DMI and total DMI, classified by other variables such as experiment, season, liveweight and stage of lactation. Observable trends in the plotted data, along with structural considerations (namely treatment means grouped within experiments), suggested an initial (baseline) mixed model that included linear fixed effects of pasture DMI, forage DMI, concentrate DMI, their interactions with season and random effects consisting of linear, random-coefficients for total DMI within experiments, plus residual variance. The random coefficients allowed for unexplained variation in response to DMI by experiment, as well as encoding the nested (i.e., treatment within experiment) structure of the data. Other fixed-effect terms, such as quadratics separately in pasture DMI, forage DMI and in concentrate DMI, pasture + forage by concentrate DMI cross-product, linear effects of days in lactation, liveweight, pasture nutritive characteristics and groupings of these interacting with DMI variables, were variously added and removed from the baseline model in order to test the significance of associations with the dependent (production) data using analysis of deviance F-tests. The sign and magnitude of estimated coefficients were checked for biological plausibility. Terms with strong associations with the dependent variable were retained in a parsimonious ‘best’ model.Few experiments in the database had more than two treatment rates of forage DMI, and even fewer with forage DMI in combination with pasture or concentrate rates of DMI. Consequently, the production response to forage DMI, and its interaction with pasture and concentrates (necessary in such models, since the response to one source of DMI depends on the presence and DMI amount from the other sources), was estimated with poor precision. The coefficients of quadratic and cross-product terms in DMI, therefore, were not always estimated to be negative, a requirement for biological plausibility. However, since responses to forage DMI were consistent with those of pasture DMI, a pragmatic solution was to work with the sum, pasture DMI and forage DMI, as a single variable. Henceforth, in this report, ‘pasture + forage’ refers to a variable being the sum of DMIs from these two sources. This constrains the model for pasture and forage responses to follow a common trajectory.Models for milk protein and fat yields, both individually and as ‘milk solids’ (milk protein plus fat yield), were developed and calibrated separately. This meant that a small degree of inconsistency could be expected. In particular, the sum of yields for milk protein and milk fat individually need not in general equal the same as the results from the direct milk protein plus fat model. This is because a slightly different set of variables can be selected for the different models. There were very small and statistically insignificant differences between the sum of the respective coefficients from the milk protein and milk fat models separately and the milk protein plus fat model. Given this, the simpler milk protein plus fat model was employed. In all cases, predicted values were calculated from the estimated fixed effects only.The goodness of fit of each model to the data within the database was checked using measures of concordance, Pearson correlation, Lin’s concordance coefficient, Nash–Sutcliffe efficiency coefficient and the root-mean-square error [30,31,32,33].EconomicsA production function model was developed, relating the DMI of concentrate supplement and pasture + forage with the output of milk protein plus fat. This production function was combined with information on costs (of feeds) and prices received (for milk protein and milk fat), to assess what combination of inputs would be best for the farmer to use to maximise profit [34]. The broader investigation of production functions is called Production Economic Theory, of which Dillon and Hardaker [34] provided a thorough and concise overview.The two variable input production function surface (Figure 1) describes the relationship between the change in quantity of two variable inputs (X1 and X2, e.g., concentrates and pasture + forage) and the resulting change in output (y, e.g., milk protein plus fat yield). The height of the surface above any point in the (X1, X2) plane shows the amount of output corresponding to that combination of X1 and X2 [34].The primary application of the production function was to estimate the amount of milk protein plus fat output for given quantities of feed input. However, the production function can also be used to determine a number of other key measures. The marginal product, defined as the change in output from an additional unit of feed, was calculated for each input factor [34]. The production function was also used to develop isoquant equations, which describe all the combinations of concentrate and pasture + forage inputs that would yield a specified quantity of milk protein plus fat output. Using the isoquant equations, the rate of technical substitution between inputs could be estimated, identifying the amount by which one variable must be increased if the second variable is decreased by one unit for the level of production to remain the same [34]. The technical substitution rates were, in turn, used to determine the isocline equations, which specify the least cost combination of concentrates and pasture + forage for any feasible amount of output. The final step in our economic analysis was to determine the profit maximising set of inputs, with or without a financial constraint of $3/cow·day. This amount was selected to illustrate application of the marginal economic analysis. The equations for each of these economic analyses are presented (Table 2). It is important to note that the technical substitution between inputs does not factor in biological impacts of changing ratios of feed inputs such as impacts on rumen function. A spreadsheet-based tool was developed to illustrate the production functions and economic concepts. It could also form the foundation for a decision support tool for farmers if the concept was shown to work. Four scenarios were examined to demonstrate these concepts and compare the estimated profit maximising combination of concentrates and pasture + forage in different seasons, with different feed and milk protein and milk fat prices. The input data used for this scenario analysis are given in Table 3. Estimates of pasture intake were necessary and were calculated using the equations published by [35]. Concentrate and forage feed prices represent the 5-year (2014–2018), CPI-adjusted (to 2017/2018 dollars), median values for Gippsland reported by Dairy Australia [36]. Milk protein and milk fat prices represent the CPI-adjusted (to 2019/2020 dollars) seasonal averages across 3 years for a commercial eastern Victorian dairy farm. Levies and charges were not subtracted from these values. An energy-corrected milk composition of 4.0% fat and 3.3% protein was assumed.3. Results3.1. Meta-AnalysisA summary of key animal, pasture and supplementary feed descriptors from experiments included in the database are presented in Table 4.The best fit, parsimonious model for milk protein + fat was:(3)Milk protein +fat yield kg/cow.day=μ+δxSY+θxWeek+αp+fxp+f+βp+fxp+f2+αcxc+βcxc2+γxP+fc+τSeason+αc.Seasonxc+λLWT+ϑxDMD%+Ei+BixDMI+εijEach covariate was centred, that is, x = covariate-mean (covariate). The Ei and Bi represent bivariate normal random mean and slope coefficients in total DMI, for experiment i. The residual error for the datum j of experiment i is denoted εij. These models can be manipulated so that the direct coefficients for the bivariate quadratic equation form can be determined.Definitions of covariates, x, coefficients and standard errors (Table 5) and means of covariates are provided below (Table 6).Using data from Table 5 and Table 6, the significant equation for milk protein + fat yield (kg/cow·day) can, therefore, be written as:(4)Milk protein + fat (kg/cow·day) =1.465 + 0.178 × (pre-experimental milk protein + fat yield − 1.63)− 0.006 × (weeks lactating − 18.18)+ season (Spring = 0, Summer = −0.174, Autumn = −0.312)+ 0.100 × (DMI pasture + forage − 12.82)+ 0.107 × (DMI conc − 2.26)− 0.002 × (DMI pasture + forage2 − 172.90)− 0.005 × (DMI concentrates2 − 11.49)− 0.002 × (DMI pasture + forage x concentrates − 26.13)+ Season × DMI concentrates (Spring = 0, Summer = 0.020, Autumn = 0.030)+ Liveweight group (<500 kg = 0, >500 kg = 0.030)+ 0.014 × (Past DMD% consumed − 72.63)Milk protein plus fat yield was lower in summer (−0.17 kg/cow·day) and autumn (−0.31 kg/cow·day) than in spring. However, there was a positive interaction with concentrate intake and season, with greater milk protein plus fat yield response to concentrates in summer (0.02 kg/cow·day) and autumn (0.03 kg/cow·day) than spring. Milk protein plus fat yield was also strongly related to DMI of concentrate and pasture + forage supplement in that each had significant linear and quadratic terms. The quadratic coefficient estimates for pasture + forage and concentrate DMI were negative, consistent with a diminishing milk protein plus fat yield response as DMI increases. Cows heavier than 500 kg liveweight were determined to produce more milk protein plus fat than cows less than 500 kg liveweight (a difference of 0.01 kg/cow·day). Finally, the digestibility of pasture consumed was significant, with higher digestibility pasture leading to increased yields of milk protein plus fat. Production surfaces for the two-variable quadratic relationship between concentrate and pasture + forage DMI and milk protein plus fat yield for spring and autumn are presented (Figure 2).How well the model fit the data within the database—the ‘goodness of fit’—was tested using measures of concordance (Table 7).The fitted model for milk protein plus fat yield was shown to closely reflect milk protein plus fat yield measured under experimental conditions (r = 0.93). However, this is to be expected, as the same data used to build the models were also used in this instance to test the ‘goodness of fit’. Ideally, the models need to be tested against ‘novel’ data—data that has not contributed to the meta-analysis. Such data are currently lacking, and we did not have enough data in our dataset to hold some back for this purpose.3.2. Economic AnalysisThe predicted profit maximising combination of concentrates and pasture + forage with respect to milk protein plus fat yield (kg/cow·day) was calculated for the scenarios presented in Table 3 (Table 8).Seasonal fluctuations in milk protein plus fat and feed prices, together with changes in animal physiology with the progression of lactation, led to variable responses, and therefore, variable profit maximising combinations of inputs. For the spring scenario, for the given combination of costs and prices, it was predicted that DMI of concentrates be increased to 2.6 kg DM/cow·day from 2 kg/cow·day, and pasture + forage DMI increased by 1.5 kg DM to 12.1 kg DM/cow·day (Table 8, Figure 3a), which would lead to an increase in milk protein plus fat yield from 2.1 to 2.2 kg/cow·day. This would cost $4.49/cow·day, and return $13.42/cow·day, generating a profit of $8.93/cow·day. By contrast, under the autumn scenario, the profit maximising amount of concentrate (6.2 kg DM/cow·day) and pasture + forage (15.0 kg DM/cow·day; Table 8, Figure 3b) was predicted to generate 1.8 kg of milk protein plus fat/cow·day. This would cost $5.23 and return $12.60/cow·day; a profit of $7.37/cow·day.For each of the scenarios, isoquants, least cost isoclines, optimal input combinations and isocost combinations in the face of financial constraint were calculated. For brevity, the spring and autumn scenarios are presented (Figure 3).These isoquants represent all the combinations of concentrate and pasture + forage for a given level of milk protein plus fat output for both the starting scenario and the modelled optimal input combination. Isoclines denote the least cost combination of inputs for a specified quantity of output, based on the unit price of concentrates and pasture + forages. Logically, the point at which the isocline transects any isoquant represents the least cost combination of feeds for the given output. In the situation where there is no constraint on the quantity of outputs to be produced, or on the quantity of inputs available [33], the profit maximising combination of inputs can be calculated, taking into account both the price paid for inputs and the price received for product. The isocost line describes the various combinations of concentrate and pasture + forage in the situation of a financial constraint. The point at which the isocost and isocline lines intercept represents the least cost combination of inputs for the given financial constraint, which was set at $3/cow·day. For the spring scenario, this is predicted to be 8.6 kg DMI of pasture + forage and 1.3 kg of concentrate supplement, and for the autumn scenario, 9.4 kg DMI of pasture + forage and 3.1 kg DMI of concentrate supplement. Milk protein plus fat output for this scenario could then be calculated. The shape of the production surfaces for the two-variable quadratic relationship between concentrate and pasture + forage DMI and profit (total milk income minus the total feed cost) for the spring and autumn scenarios are presented in Figure 4. The production surfaces demonstrate the different combinations of feed inputs to give the same profit.4. DiscussionCurrently, farmers make tactical decisions about how much supplement to feed implicitly. Many strategies are employed; some farmers feed supplements according to current milk production and changes they expect, some according to stage of lactation, some use a strategy of flat-rate feeding and some aim to manage their pastures to a consistent grazing height and use indicators such as overgrazing or wastage to judge the appropriate rate of supplement to feed [1]. However, to profitably feed supplements, it is important that farmers know how much extra milk of a particular composition will be produced for each kilogram of supplement consumed [2]—that is, the immediate marginal milk response. This knowledge can be coupled with the milk price received to determine the most profitable level of supplements to feed, i.e., the point at which marginal revenue from the extra milk just exceeds the marginal cost of the extra feed. Allocating supplements based on well-informed and more accurate estimates of marginal milk responses to supplements has the potential to improve farm profit by reducing feed costs or increasing the amount of profitable output [1].There are numerous and complex interactions that influence the immediate marginal milk response to supplements. Factors such as season, the nutritive characteristics of pastures, pregrazing pasture mass and allowance of pasture on offer [4,15,18,37], amount of pasture consumed [4,18], amount and type of supplement consumed, nutritive characteristics of the supplement [4,5,18,23], amount of substitution [15,18] and animal and management factors, such as stage of lactation [5], body condition score [19], frequency of feeding [38] and genetic merit of cows [39], have all been shown to influence milk production when supplements are fed.In the present study, a newly generated response function of milk protein plus fat yield was developed and used to analyse the economics of tactical feeding decisions where both grain and forage supplements are fed. The question examined was how much supplement plus pasture should be fed over a short time period, such as the next fortnight, to maximise profit given a particular farm situation, incorporating information such as starting milk yield and stage of lactation and what is known about milk and supplementary feed prices. It was also demonstrated how the response function could be used to determine the marginal product, the rate of technical substitution between inputs and to generate isoquants. The response function described here builds on the work reported by [1,2], which described a response function for grazing cows fed concentrate supplements only.The idea of developing a method of predicting the immediate marginal milk response is not new. Ideally, a mechanistic model that incorporates concepts about underlying biology [40] would be available to predict milk solids production when concentrates and pastures + forages are fed. However, it has been shown that for grazing cows, predicted immediate milk responses using commercially available, mostly mechanistic models are often in disagreement. Testing 11 of the most commonly used programs, Little et al. [41] reported that the predicted immediate milk response from an additional 5 kg DM of cereal grain ranged from 1.0 to 14.0 L of milk, corresponding to immediate marginal milk responses of between 0.2 and 2.8 kg milk/kg DM cereal grain. This wide range in predicted responses illustrates the difficulty in modelling complex biological systems.The new empirical model described here was developed from a dataset of experiments offering concentrates and pasture + forage to lactating dairy cows over several decades. The objective of the work was to develop a model that would have easy, on-farm application, drawing on experimental data, without detailed metabolic measurements. Thus, an empirical, predictive approach, requiring uncomplicated inputs was essential. This model has application within boundaries largely defined by the nature of the research used to build it. The model was built using data from short-term experiments, where cows grazing temperate pastures were supplemented with cereal-based or forage supplements. It is, therefore, recommended that this model is only applied in situations such as the one described. The milk response to supplements in primiparous cows could be expected to be less than for multiparous cows; however, from the current dataset, there were insufficient data to distinguish the differences between the two. In the present analysis, milk yields from animals lighter than 400 and heavier than 600 kg are largely untested, as are milk yields from cows beyond 270 days in milk. The influence of the genetic merit of cows [38] is also not captured in this model. Empirical models of this kind, developed on, and calibrated to, a wide diversity of experimental data collated from different locations in different decades, cannot detect and reliably represent the minutiae of biological processes. However, they may detect and summarise the major processes and represent an average response that may be expected to apply, at least approximately, over a wide set of conditions.Yields of milk protein and milk fat from dairy cows are strongly positively correlated with milk yield [42]. However, while total yields of protein and fat in general increase with increasing milk yield, the concentration of these components often decreases. Milk protein concentration and yield can be altered via dietary manipulation, by increasing the overall energy intake with a concurrent reduction in the pasture to concentrate ratio [43]. Decreasing the pasture-to-concentrate ratio has also been shown to reduce milk fat concentration, as has increasing overall DMI [44]. The amount of physically effective fibre, the composition of the carbohydrate within the concentrate being fed, lipid intake and frequency of feeding have also been shown to affect milk fat concentration [44]. In our model, DMI of pasture + forage, or of concentrate, significantly increased milk protein + fat yield, and would be expected to increase milk yield. These both had negative quadratic coefficient estimates, consistent with a diminishing milk protein + fat yield response as DMI increases. However, in the model presented here, it is difficult to uncouple the impacts of changes in milk protein and milk fat concentration and changes to milk protein and fat yield. It is also difficult to categorically uncouple the impacts of season main effects and stage of lactation, as the majority of cows contributing to the database of experimental results calved in spring. The season factor in the model serves to take out some of the between-experiment variation, as well as being necessary before considering DMI interactions with season. Importantly, the information on responses to DMI, which comes mostly from within rather than between experiments, are more reliable, and was the primary focus of this work.A necessary simplifying assumption was to use the sum of pasture and forage DMI as a single variable in the response function. This was because there were few experiments available that had more than two treatment rates of forage DMI or forage DMI in combination with different rates of pasture or concentrate DMI. If additional data became available, the meta-analysis could be revised to develop a three-variable response function where concentrate, forage and pasture DMI were separate inputs. This would also enable pasture and forage supplements to be valued separately in the economic analysis rather than treated as substitutes as done in this study [45].Although the response function used and the analysis performed accounted for the contribution of supplement + pasture to milk protein plus fat production only, there are benefits of feeding concentrates to replenish body tissue, increasing body condition and improving reproduction and animal health. The potential of these longer-term benefits means that farmers will continue to feed supplements when it may appear to be uneconomic on a ‘milk only’ basis.5. ConclusionsWe developed a model that could be used to predict marginal milk protein plus fat response/kg DM intake when lactating dairy cows consume concentrates and pasture + forages. Data from peer reviewed published experiments undertaken in Australia were collated into a database. Meta-analysis techniques were applied to the data and a two-variable quadratic polynomial production function was developed, based on relatively simple inputs. Production economic theory was used to estimate the level of output for given quantities of input, the marginal physical productivity of each input, the isoquants for any specified level of output and the optimal input combination for given costs and prices of inputs and output.Limitations to the application of this on-farm work predominantly lie with the scarcity of appropriate data with which to develop the model. However, given feed accounts for the largest input cost in dairying, allocation of pasture and supplements that are based on better estimates of marginal milk responses to feed inputs should deliver increased profit from either savings in feed costs, or, in some cases, increased output to approach the point where marginal revenue equals marginal costs. Such data are critical if the industry is to take better advantage of the opportunities to use supplements to improve both productivity and profitability. If the model is to be further developed and used in farmer decision tools, it will be important to test whether it can be used to predict novel production data, how it compares with commercially available computer models and also the sensitivity of predictions to changes in key inputs.

animals : an open access journal from mdpi

10.3390/ani11123560

PMC8698065

More than 41 percent of amphibians evaluated by International Union for Conservation of Nature are threatened. It is vitally important to establish scientific and effective protection strategies for these organisms. Leishan Spiny Toad is endemic to China and it has a narrow distribution area. Long-term intentional human use and habitat destruction has caused the species to suffer. Here, we developed newly reliable and efficient molecular markers based on its genome to assess its genetic diversity and population history and provided support for conservation of this toad. Our results show that this toad still possesses high genetic diversity, but population decline may increase the possibility of inbreeding, which could work against persisting survival. Recovering the toad’s habitat and strengthening the publicity and education of wildlife protection can be helpful to its sustainability.

Persisting declination of amphibians around the world has resulted in the public attaching importance to the conservation of their biodiversity. Genetic data can be greatly helpful in conservation planning and management, especially in species that are small in size and hard to observe. It is essential to perform genetic assessments for the conservation of Leptobrachium leishanense, an endangered toad and receiving secondary protection on the list of state-protected wildlife in China. However, current molecular markers with low reliability and efficiency hinder studies. Here, we sampled 120 adult toes from the population in the Leishan Mountain, 23 of which were used to develop tetranucleotide microsatellite markers based on one reference L. leishanense genome. After primer optimization, stability detection, and polymorphism detection, we obtained 12 satisfactory microsatellite loci. Then, we used these loci to evaluate the genetic diversity and population dynamics of the 120 individuals. Our results show that there is a low degree of inbreeding in the population, and it has a high genetic diversity. Recently, the population has not experienced population bottlenecks, and the estimated effective population size was 424.3. Accordingly, stabilizing genetic diversity will be key to population sustainability. Recovering its habitat and avoiding intentional human use will be useful for conservation of this species.

1. IntroductionAmphibians have long been declining on a global scale, and this trend will continue [1]. Furthermore, some amphibians face extinction or have become extinct [2]. There have been reports of massive declines in amphibians in many places, including areas where all species have been actively conserved [3,4]. Although there has been little consensus on the causes of this phenomenon [5], we recognize that amphibian populations are under serious threat and are in desperate need of conservation.The Leishan Spiny Toad (Leptobrachium leishanense) is an endemic amphibian to China and is mainly restricted in Leishan county of Guizhou Province. This species inhabits broadleaf forests at elevations ranging from 1100–1800 m and breeds in slow-flowing streams via larval development [6]. The toad suffers from significant habitat loss and is often harvested for local consumption [7]. Thus, the population size has declined dramatically. It is listed as an endangered species on the International Union for Conservation of Nature (IUCN) Red List and receives secondary protection on the new list of state-protected wildlife in China. Formulating scientific conservation strategies is necessary for this species.Genetic assessment is one of the aims of the conservation of biodiversity [8] and an important measure for amphibian population conservation [9,10,11]. Estimating genetic diversity and effective population size are the main goals of genetic assessments [12]. Genetic diversity reflects the adaptive potential of populations for environmental change [13]. When genetic diversity decreases, the extinction risk of populations increases [14]. Moreover, the levels of genetic diversity are related to population size [15]. It is a consensus that determining effective population size is more vital than measuring census size in populations [16]. In theory, small populations are susceptible to genetic depletion through drift and inbreeding, with adverse consequences for viability [17,18]. Therefore, effective population size can be used to assess the viability of populations.As next-generation sequencing technologies offer new opportunities for conservation genetics [19], microsatellite markers with high mutation rates and genome-wide distributions reveal recent changes in genetic structure and demography critical for population management [20,21]. Although several studies using 10 dinucleotide microsatellites have shown that L. leishanense has high levels of genetic diversity and has not experienced recent bottleneck events [22,23,24], genetic assessments of this species are not nearly sufficient. In addition, dinucleotide microsatellites are considered less efficient and more unreliable than tetranucleotides because of their minimal PCR stutter [25]. Moreover, the traditional methods of microsatellite isolation and characterization are quite involved, costly, and time-consuming [26]. With the publication of a number of genomes, we can obtain sufficient numbers of different types of useful microsatellite loci more efficiently [27]. The genome sequencing project of L. leishanense has provided the opportunity to isolate and characterize microsatellites at the genomic level [28].Here, we totally sampled 120 adult toes of L. leishanense from the population in the Leishan Moutain, and 23 of them were used to develop tetranucleotide microsatellite markers with polymorphisms based on one reference L. leishanense genome. After that, we analyzed the genetic diversity and population dynamics using the microsatellite loci we identified. The goals of this study were to (1) develop microsatellite loci with high reliability and efficiency, (2) evaluate the genetic diversity of the L. leishanense population, (3) detect if the population is experiencing a population bottleneck, (4) estimate the effective population size, and (5) provide molecular support for L. leishanense conservation planning.2. Materials and Methods2.1. SamplingIn 2012, 2013, 2014, 2015, and 2018, we collected 24 L. leishanense adults per year in Maoping village of Leishan County, Guizhou, China (Figure 1), sampled their toes, fixed the toes in anhydrous ethanol, and stored them in a −20 °C refrigerator. All individuals were released immediately after sampling. All experiments involving animals were approved by the Animal Ethics Committee of the School of Life Sciences, Central China Normal University (CCNU-IACUC-2019–008). We have complied with all relevant ethical regulations for animal testing and research.2.2. DNA Extraction and Primer SelectionDNA samples were extracted using the TIANamp DNA kit (Tiangen, Beijing, China) and stored at −20 °C. MicroSatellite identification tool (MISA-web, Gatersleben, Germany) [29] was used to obtain the simple sequence repeats (SSRs) of L. leishanense from its genome. Then, we randomly selected 87 tetrabase repeat microsatellite markers that were repeated more than 10 times and designed 87 pairs of primers according to the flanking sequences at both ends of each primer. With the extracted DNA as a template, we optimized the annealing temperature of the primers and reaction system. Each polymerase chain reaction (PCR) procedure was conducted in a 10 μL volume, in which the premix was 5 μL, each primer was 0.3 μL, template DNA was 0.6 μL, and ddH2O was 3.8 μL. The procedure was performed with initial denaturation at 94 °C for 5 min, followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at temperature Ta for 30 s, extension at 72 °C for 45 s, and extension at 72 °C for 5 min. PCR products were detected by 1% agarose gel electrophoresis. By adjusting the Ta temperature, a product with a clear band was obtained. The Ta temperature corresponding to the product was used as the optimum temperature for PCR amplification. Under the optimal amplification conditions, we used the DNA of three different individuals to detect the stability of primers in different individuals and screened the primers that could be amplified stably.2.3. Polymorphic Microsatellite VerificationThe screened primers were used to synthesize 5′ upstream fluorescent primers (FAM, HEX and TEMED, compounded by Tiangen, Beijing, China). DNA amplification was performed on 23 individuals collected in 2012 and 2013 by PCR with fluorescent primers, and the amplified fluorescence PCR products were sent to Tsingke Biological Company, Beijing, China for SSR scanning and sequenced by an ABI 3730xl analyzer. Then, the products were genotyped and calculated, and the evaluation criterion of the polymorphisms was a PIC value higher than 0.5 [30]. We used Genemarker 1.3 software [31] to read the lengths of alleles, genotyped the microsatellite markers, and selected the sites with obvious polymorphisms for the following analysis. The microsatellite genotyping data in Excel were transformed by using the Microsatellite Toolkit [32]. Cervus 3.0 software [33] was used to calculate the number of alleles (Na), polymorphism information content (PIC), expected heterozygosity (He), and observed heterozygosity (Ho). Micro-Checker 2.2.3 [34] was used to check large allele dropout of the microsatellite markers. GenePop 1.2 software [35] was used to detect the Hardy–Weinberg equilibrium (HWE) and linkage disequilibrium (LD) of the screening microsatellite markers with polymorphisms, and the Bonferroni correction was used for correction. The significance level was p < 0.05.2.4. Genetic Diversity AnalysisA total of 120 DNA samples were amplified by PCR with the screening fluorescent primers described above. The PCR products were sent to Qingke Biological Company for SSR scanning. An ABI 3730xl analyzer was used for sequencing. Data were analyzed using GenAlEx 6.502 [36] to calculate the effective number of alleles (Ne), the mean relatedness of the individuals for every year, and the per year genetic differentiation coefficient (Fst), and Cervus 3.0 software was used again to calculate the values described above. Excel and Microsatellite Toolkit v3.1.1 software were used for preliminary genetic data statistics and data format conversion. FSTAT 2.9.3.2 software [37] was used to calculate allelic richness (Ar), allelic diversity (Hs), and the inbreeding coefficient (Fis).2.5. Population Bottleneck IdentificationBottleneck 1.2.02 software [38] was used to test whether the population had experienced population bottlenecks. Sign and Wilcoxon methods were used to test mutations through three mutation models: the infinite allele model (IAM), the stepwise mutation model (SMM), and the two-phased model of mutation (TPM). TPM was set to 95% SMM, with a variance of 30 and 1,000 iterations.2.6. Effective Population Size CalculationNeEstimator 2.1 [39] was used to calculate the effective population size by selecting the random mating model, and the confidence interval was 95%.3. Results3.1. Distribution of SSR in Genome of L. leishanenseA total of 1,454,145 microsatellite markers were obtained from the genome of L. leishanense. Monobase repeat microsatellite markers and dibase repeat microsatellite markers were the most common among all microsatellite markers. There were 874,773 monobase repeat microsatellite markers and 263,927 dibase repeat microsatellite markers, accounting for 60.16% and 18.15% of the total number of microsatellite markers, respectively, followed by 71,167 tribase repeat microsatellite markers and 23,332 tetrabase repeat microsatellite markers, accounting for 4.89% and 1.60% of the total number of microsatellite markers, respectively. The number of pentabase repeat microsatellite markers and hexabase repeat microsatellite markers was the lowest, with 909 pentabase repeat microsatellite markers and 844 hexabase repeat microsatellite markers, accounting for only 0.12% of the total microsatellite markers (Figure 2).3.2. Polymorphism Microsatellite LociEighty-seven pairs of primers randomly chosen from 23,332 tetrabase repeat microsatellite marker. After primer optimization, 64 pairs of primers were successfully amplified. Then stability detection was used, and 46 pairs of primers were obtained. Employing polymorphism detection, we obtained 12 satisfactory microsatellite loci. The Na of these loci ranged from 6–16. PIC values ranged from 0.537–0.904. Ho and He were between 0.609–0.913 and between 0.622–0.931, respectively (Table 1). All 12 loci were not significant with regard to LD (p > 0.05), and there were no loci deviated from HWE (p > 0.05). According to Micro-Checker 2.2.3, there was no large allele dropout of these microsatellite markers and no scoring error caused by the shadow peak.3.3. Population Genetic DiversityUsing the 12 loci screened above, all 120 individuals were used to study the genetic diversity of the population. HWE detection of the L. leishanense population was performed. After Bonferroni correction, the significance level was p < 0.0042. The results are shown in Table 2. When we used all 120 samples for testing, loci LEA23, LEA7, LEA47, LEA2, and LEA53 deviated from HWE significantly, with Fis values as 0.179, 0.112, 0.119, 0.230, and 0.134, respectively (Table 3). When we separated the samples into each year for testing, locus LEA20 deviated from HWE significantly in 2013, with a Fis value of 0.215. Locus LEA23 and LEA7 deviated from HWE significantly in 2015 and 2018. Fis values of locus LEA23 in 2015 and 2018 are 0.132 and 0.249, respectively. Fis values of locus LEA7 in 2015 and 2018 are 0.161 and 0.369, respectively. Locus LEA47 deviated from HWE significantly in 2018 with Fis value as 0.201, and locus LEA2 deviated from HWE significantly in 2015 with Fis value as 0.218.Then, we calculated pairwise year Fst valus in L. leishanense (Table 3). None of these values is greater than 0.05, suggesting the genetic differentiation between these years is negligible [40]. Further, we calculated the mean relatedness of the individuals for every year (Figure 3). In 2013, 2014, and 2015, mean pairwise relatedness within groups was significantly greater than zero, indicating the samples we collected in these three years have relatively close relationships.Next, we calculated Na, Ne, PIC, Ho, He, Ar, Hs, and Fis of the population (Table 4). The results indicated that the genetic diversity of the population was still high. The positive value of Fis and that of Ho was lower than that of He, suggesting that there was a low degree of inbreeding in the population.3.4. Population BottleneckThe average expected heterozygosity (Heq) of the population in the IAM, SMM, and TPM models was calculated (Table 5). In the IAM model, there were 11 sites where He was significantly higher than Heq (p < 0.05), among which LEA22, LEA25, LEA20, LEA35, LEA14, LEA23, and LEA24 were extremely significantly higher than Heq (p < 0.01). In the TPM and SMM models, only He at LEA5 was significantly higher than Heq, showing heterozygote surplus. The sign test and Wilcoxon test were used to detect the heterozygosity surplus of the population under the three models of IAM, TPM, and SMM (Table 6). The mutation-drift balance of the population was detected under the IAM model, both of the sign and Wilcoxon tests showed significant deviations from the mutation-drift balance of the population. Under the TPM and SMM models, both the results of the sign test and Wilcoxon test showed that the population did not deviate from mutation-drift equilibrium.The analysis of allele frequency distribution in the L. leishanense population showed that the allele frequency was mainly concentrated between 0.0–0.1, which was approximately 84.52% of the total allele frequency. Alleles with a frequency of 0.1–0.2 accounted for 12.69% of the total allele frequency. The proportion of the frequency distribution interval of 0.2–0.3 was 1.80%, while the allele proportions of the frequency distribution intervals of 0.3–0.4 and 0.4–0.5 were 0.5%. The allele frequency showed a typical “L” type distribution (Figure 4), suggesting that the population has not recently experienced a bottleneck effect.3.5. Effective Population SizeAccording to the LD distance, the effective population size was estimated to be 424.3 (95% CI = 272.7–878.2) from NeEstimator 2.1.3.1.4. DiscussionWe isolated and characterized 12 tetrabase repeat microsatellite markers with polymorphisms from one reference genome of L. leishanense. Then, we used these loci to study the genetic diversity and population dynamics of this species. We found that the genetic diversity of the population was high and that there was a low degree of inbreeding in the population. Moreover, the population has not recently experienced bottleneck effects, and the estimated effective population size is 424.3.4.1. Tetranucleotide Microsatellite MarkersAlthough the results above are similar to those of Zhang’s research [24], which used 10 dibase repeat microsatellite markers, the 12 tetranucleotide microsatellite markers we developed are more polymorphic and suitable for genetic diversity research. During PCR amplification, a biological phenomenon called stutter is generated due to chain slippage, resulting in typing errors, and the stutter product has one or more fewer duplicates than the real allele product [41,42]. In general, tetranucleotide repeats tend to stutter less than trinucleotide and dinucleotide repeats and are much more accurate and reliable [43,44]. Therefore, in different types of microsatellite systems, tetrabase repeat microsatellite markers are more common than dibase or tribase markers. Moreover, the PIC values of all 12 loci were higher than 0.5, suggesting that the loci we developed had higher polymorphism. Stable and reliable microsatellite markers are a necessary prerequisite for population estimation in the wild [45]. Thus, after primer optimization, stability detection, and polymorphism detection, we finally obtained 12 satisfactory tetranucleotide microsatellite loci.4.2. Genetic DiversityWe speculated that several loci deviated from HWE (Table 2) mainly caused by sampling from the same family (Figure 3). The sharp decline of the population size has increased the possibility in sampling individuals of same family. The genetic diversity of a population is a long-term process, the population of L. leishanense does not have significant genetic differentiation among these five years (Table 3); accordingly, we considered that these deviating loci were still effective in estimating population genetic diversity. Then, a series of indices were used to measure the genetic diversity of the toad, including Na, Ar, Ne, PIC, Ho, He, and Hs. According to our results, the toad still has high genetic diversity. Threatened species usually have small or declining populations and are prone to loss of genetic diversity due to inbreeding or genetic drift [14]. As an endangered and narrowly distributed toad, the population shows the opposite result. Several studies investigating endangered or narrowly distributed species have obtained similar results [45,46,47,48,49], indicating that endangered species or species with a narrow distribution may also have high levels of genetic diversity. When the earth was in an ice age, some areas with a stable ecological environment became the refuge of organisms, and the populations living in the refuge survived and accumulated rich genetic diversity [50]. The Leishan Spiny Toad is a relatively primitive species, and its formation dates back to the Miocene [22]. The toad survived by staying on Leigong Mountain and retained rich genetic diversity when the ice age came. In addition, two additional distribution sites were found by Zheng et al. [51], suggesting that the toad is not strictly a narrowly distributed species. We may have underestimated the genetic diversity of the species.Although the Fis value of the population is on the low degree, this does not mean that there is no inbreeding between the individuals in the population. According to our year-by-year field work, its population size is declining. This undoubtedly increases the possibility of its inbreeding. Inbreeding has a negative effect on the fitness of the population, including fertility and viability [52], which is not conducive to the long-term development of the population. We could not find more obvious molecular evidence of inbreeding, possibly due to our restricted sampling size and the relatively high number of alleles found. As with high number of alleles, the probability of obtaining homozygote hgenotypes in one locus is very low. Thus, it will influence our detection of inbreeding.4.3. Population DynamicsCombining the results of model simulation with allele frequency distribution, we find that the population has not recently experienced a bottleneck effect. We tested three models, and IAM was significant both in the sign test and the Wilcoxon test. Both SMM and TPM were not significant in the sign test and Wilcoxon test (Table 5). IAM assumes that there is only one mutation of an allele in a population, and each mutation produces a new allele, which is generally used in isozyme or DNA sequencing data. SMM supposes that alleles can mutate upward or downward into new alleles. TPM is the synthesis of the previous two models, and the probability of occurrence of two kinds of mutations can be determined. The principle of allelic mutation in microsatellite data is the increase or absence of repeating units, which is represented by the change in sequence length. Some studies believe that TPM is more suitable for microsatellite data [53]. Therefore, we accept the result of TPM that there is no significant excess heterozygosity in this population. That is, the rate of heterozygosity decrease is approximately the same as the rate of allele loss in L. leishanense, indicating that the population has not recently experienced a bottleneck.However, the ability to detect the population bottleneck based on heterozygosity is limited, and the number of alleles is more sensitive to population fluctuation, so it is more reliable to analyze the distribution of allele rates in the case of heterozygous residues to determine whether the population has experienced the bottleneck effect [54]. To enhance the adaptability to environmental changes, species tend to accumulate many rare alleles with low frequency. Therefore, the frequency distribution of alleles in mutation–drift equilibrium shows an “L” shape. If the species recently experienced a genetic bottleneck, the distribution of alleles with low frequency (0.0–0.1) will change to a mid-frequency distribution (0.1–0.2); thus, the allele frequency distribution will deviate from “L” [55]. In this study, the frequencies of the allele were generally in a typical “L” (Figure 3), suggesting that the population did not experience bottleneck effects.The LD distance between microsatellite markers can be used to estimate effective population size, and this method has been applied to mammals, fish, amphibians, and other animals [56]. Effective population size is a valuable method in population conservation and management research. Maintaining an effective population of sufficient size is a key factor to maintain the rich genetic diversity of the population. Based on the microsatellite loci we developed, the estimated effective population size of L. leishanense is 424.3. Nei et al. [57] deemed that the population size should be 4–10 times of the effective population size to maintain the stability of population genetic diversity. Therefore, to maintain the stability of the population, the number of Leishan Spiny Toads should be 1697.2–4243. However, while LD information is used to estimate the effective population size, the accuracy of the results is significantly correlated with the sample size [58]. More samples may be needed to obtain more reliable and accurate results in L. leishanense.5. ConclusionsOur study has provided 12 reliable tetranucleotide microsatellite loci with polymorphisms, enriching the information regarding the genetic diversity and population dynamics of L. leishanense. Although the genetic diversity is still high based on our results, a low degree of inbreeding indicates that the population is declining. Avoiding habitat fragmentation and intentional human use will be key to the conservation of this species. Furthermore, recovering the streams and woodlands where the species once existed abundantly will also help to stabilize its genetic diversity.

animals : an open access journal from mdpi

10.3390/ani13091530

PMC10177539

Wild animals in captivity need stimuli that increase their well-being. Canids in general have a well-developed sense of smell and are strongly related to environmental stimuli through scent. Therefore, we tested an olfactory enrichment method in five hoary foxes, which was successfully developed in another species of canid. We offered four stimuli (cheese, eggs, meat, and sawdust impregnated with rat urine), and observed the individuals’ reactions that indicated an improvement in well-being before, during, and after exposure to the stimuli. There were no significant changes in behaviors that indicated well-being, although there was no worsening in behaviors suggestive of stress. We suggest that the indifference to stimuli of this little-known species is due to the highly insectivorous diet of the hoary fox.

We have tested a method of olfactory environmental enrichment in hoary foxes used in other wild canids in captivity. The individuals were exposed to four olfactory stimuli (meat, mouse urine, cheese, and egg) that were wrapped in cotton bags outside the enclosures at the zoo for five minutes. Behavioral observations were performed using the focal animal method, and all occurrences were recorded. The pre-exposure phase (Basal), exposure phase (Exp), and post-exposure phase and Basal phase (Pos) were analyzed for a period of five minutes. Behavioral responses were categorized as positive, negative, or other. Positive behavior tended to increase (p = 0.07) from the Basal phase to the Exp phase, but there was no statistical difference (p = 0.31) between the phases. Negative and other behavior did not differ statistically from the Basal phase to the Exp phase (N−, p = 0.32; Ot, p = 0.35) or Basal to the Pos phase (N−, p = 0.18; Ot, p = 0.92). The odors used seemed to elicit positive behavior weakly. Negative behavior was stable for the hoary foxes. The method failed to improve the hoary foxes’ welfare. Because their natural diet is based on insects and fruits, it is suggested that the stimuli used in this study have no appetitive value for hoary foxes. The method used with the same olfactory stimuli that were successful in other canid species is unsuitable for hoary foxes.

1. IntroductionCaptive environments are often monotonous, limited in stimuli, and restrict the performance of behaviors considered normal for the species. Environmental enrichment (EE) is defined as “an improvement in the biological functioning of captive animals resulting from modifications to their environment” [1]. The application of EE must be safe, significant to the individual, and preferably with low administrative costs [1]. As described within the definition itself, EE leads to an increase in animal welfare, which is one of the goals of most zoos [2].EE is characterized by the introduction of stimuli linked to the social, physical, and sensory contexts of captive animals [3]. Despite knowledge of the high level of olfactory acuity of canids in general [4,5,6,7], studies about olfactory enrichment are little explored. Only 3% of the articles published in the scientific literature deal with olfactory enrichment (OE) in canids [4]. Despite the scarcity of studies on olfactory enrichment for South American canids, recently, some authors have presented a successful and less invasive method for crab-eating fox (Cerdocyon thous) [8]. In that experiment, the authors observed 22 crab-eating foxes exposed to four types of food-related odors. Behaviors suggestive of enhanced well-being (“Positive” behaviors) increased. On the contrary, there was a decrement in behaviors considered “negative”, which lowered well-being. These effects remained after withdrawal of stimuli, in the short term [8].While crab-eating foxes are relatively well studied in captivity, there are few studies on the behavior of hoary foxes (Figure 1). In particular, there are no studies of OE for the hoary fox [7,9]. The hoary fox is a species endemic to the Brazilian Cerrado, being considered “Near Threatened” in the International Union of Conservation Nature extinction risk indices [10]. The greatest threats to the hoary fox are habitat loss, predation by domestic dogs (Canis lupus familiaris), and the danger of being run over on the country’s highways [11]. In many cases, animals injured or seized outside of their habitat are taken to recovery centers or zoos, remaining in captivity indefinitely [11].The diet of foxes is well known to be based largely on insects, particularly Coleoptera [11]. The regular acquisition of insects to feed the foxes in captivity is not feasible because it would require an infrastructure that demands high costs. For this reason, a mix of dog food, meat, and some vegetables is regularly offered in the diet of zoo canids, including foxes. Little is known about the social interactions of foxes, appearing to be restricted to the pair’s interaction during the mating season and the mother’s relationship with her young [11]. In a literature review, we did not find systematic studies on the relationships of foxes in captivity when housed in pairs or with more animals. With this scenario, we deduce that the captive environment is not stimulating, consequently reducing the possibility of satisfying the foxes’ behavioral needs.Based on studies on environmental enrichment in canids [4,8], we hypothesized that the introduction of different, non-noxious olfactory stimuli could increase the well-being of captive hoary foxes, as observed in another study on crab-eating foxes [8]. Because of the lack of knowledge on how olfactory stimuli (OS) can be introduced in an EE program for rarely studied canid species, the current study sought to investigate the behavioral response of hoary foxes exposed to different odors in captivity. The ultimate goals of olfactory stimuli are responses with exploratory behaviors, play, non-agonistic interactions, and relaxation; when these behaviors increase, we interpret that there is an increase in the well-being of the foxes.2. Materials and MethodsThe study was carried out with five captive hoary foxes (Figure 1) in the Ecological Zoo Park of São Carlos (PESC), in São Carlos, SP, Brazil. The individuals were adults (one male and four females) between two and eight years old. The foxes were housed in a pair and a trio (2 females and a male), in enclosures with an area of approximately 100 m2, surrounded by wire fences on three sides and a wall in the back. Inside the enclosure, there was a shelter for the foxes to hide and rest. Tree trunks, a bush, and natural stones also structured the exhibit. At the back of the enclosure, there was an indoor area with bowls for drinking water and eating. The foxes were fed a mix of fresh fruits, protein of animal origin, and industrial dog food in the morning. The foxes were healthy, and neither pregnant females nor puppies were present during the study.The OE method tested in hoary foxes in the present investigation was adapted from the study by Figueira et al. (2021) [8] on crab-eating foxes. Due to the absence of studies in the scientific literature on OS for hoary foxes, the odorous stimuli were adapted from the study on crab-eating foxes [8]. The OS were 100 g of fresh minced beef; 100 g of chopped parmesan cheese; two boiled and chopped chicken eggs; and approximately 100 g of sawdust removed from boxes containing rodents. A detailed ethogram for captive hoary foxes was not found in the scientific literature to determine behavioral welfare. For this reason, an ethogram (Table 1) was developed as an adaptation from the description of captive crab-eating fox behavior [8]. In order to better analyze the effects of OS, behavior was categorized as positive (P+), negative (N−), and other (Ot). Based on the scientific literature, the P+ category contains behaviors that increase an animal’s welfare, while the N− category contains undesirable signs of distress [12,13]. The Ot behaviors were considered ambiguous or indifferent to OS and consequently do not influence the welfare of the hoary foxes.The olfactory stimuli were placed inside permeable cotton bags, which allowed the animals to sense their odors without being able to see them. All of the bags were the same color and size and were washed with mild soap after each use. The OS were positioned in front of and outside each enclosure. The observation sessions were in the morning, before food was placed for the foxes and without visitors in the zoo.The behavior of the individuals was recorded with digital cameras (Samsung® ST77, Daegu, Republic of Korea), which were mounted on tripods in front of the cage at a height of 1.5 m. The filming took place between 8 am and 10 am, before the feeding of the animals by zoo staff.The filming sessions of the foxes in each enclosure lasted one morning, on different days (Figure 2). On the day, four OS sessions were conducted, one for each attractive stimulus. The order in which the OS were presented had been previously defined by chance. Each session lasted five minutes, with one-minute intervals between sessions. After positioning the camera, we filmed for 5 min without exposing the subjects to any stimuli. We called this phase “warm up”, so that the individuals would get used to the presence of the film camera and movements of the researcher. The approach and movement of the researcher could scare the hoary foxes. The warmup session serves to not suddenly scare the animals. Soon after, the Basal session began; this was a five-minute session where the animal was filmed without presentation of any OS. Following that, the exposure session (Exp) began when the researcher placed the OS in front of the enclosure and left again. At the end of the Exp session, the researcher entered in front of the enclosure to remove the OS and again left. This marked the beginning of the post-exposure (Pos) session, where the individual was filmed for five more minutes without the stimulus. After that, the session ended.The behavioral responses of the individuals were collected following the focal animal method and recording all behaviors [14]. The total time of each behavior was counted with the aid of the Prostcom behavioral analysis software [15]. During the first minute of each phase (Basal, Exp, and Pos), while the researcher was in front of the animal’s field of view, the behaviors were not recorded in the software. The means of behavioral responses (P+, N−, and Ot) from the five hoary foxes were calculated from the recorded time for each set of stimuli sessions. All comparisons were performed in order to verify whether there were any changes in behavior in the Exp and Pos phases in relation to the Basal session. When changes in behavior occurred, it was assessed whether they increased or decreased and whether they remained after the olfactory stimulus was removed. Due to the small sample size, which distorts to a non-normal distribution of the data, non-parametric analyses were performed, applying the Wilcoxon test for paired samples [16]. All statistical tests followed a two-tailed distribution, with an alpha level < 5%.3. Results and DiscussionThe behavior duration (s) of the P+ category tended to increase (p = 0.07) from the Basal phase to the Exp phase, but it was not statistically different (p = 0.31) in the Pos phase compared to the Basal phase (Table 2). The average duration for the N− and Ot categories did not differ significantly (N−, p = 0.32; Ot, p = 0.35) from the Basal phase to the Exp phase or from the Pos phase to the Basal phase (N−, p = 0.18; Ot, p = 0.92; Table 2).Canids use smell as one of their principal means of communication and exploration of the environment [9], but in this experiment, the olfactory stimuli were not able to alter behavior significantly. The hoary foxes seemed to be indifferent to the olfactory stimuli with the method used. The lack of differences in N− and Ot, together with the weak effect on P+ between phases, strongly suggests that the response of hoary foxes to olfactory stimuli was one of indifference.Odors are a complex mixture of several volatile compounds, whose composition is dependent on concentration and the chemical family of the molecules [17]. Some edible items share common volatile compounds, having odor-like organoleptic characteristics and making them appetizing for an animal species. The stimuli used (meat, egg, mouse urine, and Parmesan cheese) have volatile components that differ in their composition [18,19,20,21], and are items that are not listed in the scientific literature as food ingested by hoary foxes. The hoary fox is the most specialized South American canid [11], with a diet largely based on termites, coleopterans, and fruits [22], having insectivore dentition [23]. Insects and fruits predominate in the hoary fox diet, whose compositions of volatile molecules, such as a high concentration of alkaloids in insects, are apparently different from the OS that were utilized in the experiment [24]. The stimuli to which the hoary foxes were exposed may therefore not have had enough appetitive value to elicit a behavioral response indicative of either increased or decreased welfare.Despite the tendency for positive behaviors to increase during exposure to the stimulus, overall, the mean time spent on these activities was low compared to the category of other behaviors. The mean duration of negative behaviors was also low compared to the category of other behaviors. Looking from another perspective, the average times of the positive or negative behavior occupied a fraction of the session’s time, while the average time of other behaviors occupied between 98.8% (Phase Exp) to 99.9% (Phase Pos) of the time of the sessions. The very low manifestation of positive and negative behaviors suggests that the hoary foxes were indifferent to the stimuli in our method adapted from the schedule proposed by Figueira and collaborators [8] for crab-eating foxes. In that experiment, positive behaviors increased during exposure and remained high after stimulus withdrawal.Environmental enrichment using olfactory cues that simulate food must have an appetitive value for individuals, in order to encourage behaviors from their natural behavioral repertory [1]. The stimuli used in this experiment do not appear to have been reported in the scientific literature for free-ranging hoary foxes, and may not have biological significance to motivate individuals to increase positive behaviors. This study suggests that it is necessary to know aspects of the feeding ecology of each species to expose captive animals to olfactory enrichment.The sample size may have been insufficient to demonstrate the effect of stimuli on positive behaviors, given the statistical trend found between the Basal phase and the Exp phase. The opposite might also be true, that is, an increased sample size might clearly show a lack of significant difference between the phases. Therefore, the results based on a sample of five individuals do not allow the conclusion that the method using the four stimuli is appropriate to increase the well-being of hoary foxes. Although we did not obtain results that indicated an increase in the welfare of hoary foxes, the reporting of the results is recommended for practical and ethical reasons. Results that frustrated expectations or hypotheses are less publicized than results that prove the expectations of the researchers, which can lead to biased conclusions [25]. Resources can be saved by avoiding procedures that do not appear to be adequate for improving animal welfare. Despite the inconclusive results for the environmental enrichment, this behavioral investigation is original, since it is the first study to evaluate olfactory stimuli in hoary foxes.4. ConclusionsThe olfactory environmental enrichment method used in other canid species did not seem suitable for hoary foxes. Due to the highly specialized food biology of hoary foxes, the olfactory stimuli used do not seem to be attractive enough to modify the behavior that indicates an improvement in well-being. Other more appropriate stimuli (e.g., insect odor) could lead to the success of environmental enrichment using the method described in this study. Finally, it is necessary to consider the ecology of each animal species to introduce environmental enrichment to captive individuals.

animals : an open access journal from mdpi

10.3390/ani12030391

PMC8833776

To elucidate the differences in milk protein compositions and mammary gland functions between yaks of standard lactation (TL yaks) and prolonged lactation (HL yaks), iTRAQ technique was used to compare the skim milk proteins in the two yak groups. A total of 202 differentially expressed proteins (DEPs) were revealed, among which 109 proteins were up-regulated and 93 were down-regulated in the milk of HL yaks compared to TL yaks. The bioinformatics analysis revealed that the differences in skim milk protein between the HL yaks and the TL yaks suggests that the mammary gland of the HL yak is at a degeneration stage.

Extended lactation is a common phenomenon in lactating yaks under grazing and natural reproduction conditions. To elucidate differences in milk protein compositions and mammary gland functions between yaks of standard lactation (TL yaks) and prolonged lactation (HL yaks), whole milk samples of TL yaks and HL yaks (n = 15 each) were collected from a yak pasture at the northwest highland of China. The iTRAQ technique was used to compare the skim milk proteins in the two yak groups. A total of 202 differentially expressed proteins (DEPs) were revealed, among which 109 proteins were up-regulated and 93 were down-regulated in the milk of HL yaks compared to TL yaks. Caseins including κ-casein, αs1-casein, αs2-casein, and β-casein were up-regulated in HL yak milk over 1.43-fold. The GO function annotation analysis showed that HL yaks produced milk with characteristics of milk at the degeneration stage, similar to that of dairy cows. KEGG enrichment showed that the metabolic pathways with the most differences are those that involve carbohydrate metabolism and the biosynthesis of amino acids. The present results highlight detailed differences in skim milk proteins produced by HL yaks and TL yaks and suggest that the mammary gland of HL yak is at the degeneration stage.

1. IntroductionThe yak is one of the main unique livestock species in ethnic minority areas of the Qinghai–Tibet Plateau in China [1]. It provides important living materials for local farmers, with milk being one of the major products. There exist some differences in lactation between yaks and dairy cows. The lactation of dairy cows can be extended beyond the standard 305-d lactation through several manipulated strategies. Moreover, these cows are subject to a substantial decline in milk yield and their milk composition changes during the extended lactation stage [2]. Since the feeding and management mode in yaks is natural grazing and reproduction, they are completely dependent on natural grass resources and natural reproduction. Furthermore, considerable proportions of lactating yaks (about 15%) will experience a prolonged lactation stage if they are not pregnant during the calving year [3]. If these yaks are calved in the spring and milked for consumption, can maintain lactation in the following winter by calf suckling only, and can be milked again during the next summer season when there is enough grass available. Thus, they are at a stage of naturally prolonged or extended lactation, and yaks at this lactation stage are called half-lactating (HL) yaks—respective to the standard lactation of yaks, which are called total-lactating (TL) yaks. HL yaks exhibit their second lactation peak during the summer season. Due to the relatively large number of HL yaks, their milk constitutes an important part of yak milk resources.The composition of milk from the extended lactation stage has been studied in dairy cows, and it was found that milk protein and fat concentrations increase significantly [4], which is probably related to the decline in milk yield. Compared to TL yaks, the milk yield of HL yaks decreases significantly, while the concentrations of milk protein and fat increase significantly [5]. Our previous study has discovered some differences in milk composition between TL yaks and HL yaks. For example, lactose level and α-lactalbumin percentage are reduced in HL yak milk [6], which reflects the composition characteristics of milk from late lactation [7]. However, the composition of HL milk has not been studied extensively, and the state of the mammary glands of HL yaks remains unclear.Most of the bovine milk proteins, including caseins, α-lactalbumin, and β-lactoglobulin, are synthesized by the epithelial cells of the mammary gland, while a small part of milk proteins, such as immunoglobulins and albumin, are derived from the blood [8]. Therefore, the protein composition of milk is related to the function of the mammary gland to some extent—a well-known example is the high immunoglobulin level in bovine colostrum [9]. Isobaric tags for relative and absolute quantitation (iTRAQ)-based proteomics is a widely used tool to study the overall protein profile of cells or tissues [10].It is highly precise and has been used in the field of biomedicine, zoology, botany, and microbiology research. Reinhardt used the iTRAQ technique to analyze milk fat globules of membrane proteins and found that 26 proteins were up-regulated, while 19 proteins were down-regulated, in the mature milk compared to colostrum [11]. Zhang L. reported that during the first 9 days after calving, one-third of the proteins, especially the content of immunoglobulin, decreased significantly in the milk [12]. A review article by Roncada P. reported the last progress in proteomic analysis of milk from farm animals [13]. The objectives of the current experiments were to examine the detailed differences in milk proteins between TL and HL yaks and to elucidate the nutritional value of HL milk. Moreover, the current experiments investigated possible functional changes in the mammary glands of HL yaks, which is of significance for the utilization of yaks at this special lactation stage. 2. Materials and Methods2.1. Experimental Yaks and Milk SamplingMilk samples of yaks were collected in August. The experimental yaks were raised on a pasture about 3500 m above sea level in Hongyuan county, Sichuan Province, China. The experimental yaks included TL yaks (n = 15) and HL yaks (n = 15). The TL yaks calved during the spring season (March to May), while the HL yaks calved one year earlier. Their milk secretion was maintained by calf suckling during the winter. All the experimental yaks were 4 to 7 years old and 2 to 4 parities. The experimental yaks and their calves were grazed on the same natural grassland during the daytime and were separated from their calves at night. The lactating yaks were milked by the hands of local farmers in the morning. Approximately 50 mL of whole milk were collected from each yak, which was transferred to the laboratory using dry ice and stored at −80 °C until analysis. All animal care and milking procedures were approved by The Animal Ethics Committee of Southwest Minzu University (No. swun20200138).2.2. iTRAQ Analysis of Skim Milk Proteins of YaksThe whole milk of each yak was centrifuged at 800× g and 4 °C for 20 min to prepare skim milk. The pooled skim milk samples of TL yaks and HL yaks were prepared by mixing equal volumes of 15 skim milk samples of corresponding yaks, respectively. The two pooled samples (2 mL each) were transported using dry ice to Shenzhen BGI Technology Co., Ltd. for iTRAQ analysis.For the iTRAQ assay, the pooled skim milk samples were centrifuged at 25,000× g for 20 min to remove residual fat and cell debris. The supernatant skim milk was removed and mixed with 5 volumes of cold acetone and stored at −20 °C overnight. The mixture was centrifuged again, and the resulting pellet was used for further preparing a protein solution [14]. A total of 100 μg of protein from this solution was digested with Trypsin Gold (protein: trypsin = 20:1) at 37 °C for 12 h. The resulting peptides were labeled using the iTRAQ Reagent 8-plex Kit according to the manufacturer’s protocol, followed by fractionation using a Shimadzu LC-20AB HPLC equipped with a 4.6 mm × 250 mm Gemini C18 column (Phenomenex) [15]. The eluted peptides were pooled as 20 fractions and were then vacuum-dried, dissolved, and loaded on an LC-20AD nano HPLC (Shimadzu, Kyoto, Japan) equipped with a 2 cm C18 trap column. Then, the peptides were eluted into an 18 cm analytical C18 column. Mass spectrometry analysis was performed as described in previous studies [13]. Data was acquired using a TripleTOF 5600 System fitted with a Nanospray III source (AB SCIEX, Downtown Redwood City, America).2.3. Bioinformatics AnalysisIQuant software was applied to the quantification of proteins. Proteins with a 1.2-fold change and a Q-value of less than 0.05 were determined as differentially expressed proteins, and they must be defined in at least 1 replicate experiment. All proteins with a false discovery rate (FDR) of less than 1% proceeded with the following analysis, including Gene Ontology (GO), Clusters of Orthologous Groups (COG), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway. The KEGG database (http://www.genome.jp/kegg/, accessed on 6 November 2021) and the COG database (http://www.ncbi.nlm.nih.gov/COG/, accessed on 6 November 2021) were used to classify and group the identified proteins. Functional annotations of the proteins were performed using the Blast2GO program against the non-redundant protein database in NCBI (www.ncbi.nlm.nih.gov, accessed on 6 November 2021). The pathway analysis was carried out by KEGG (http://www.genome.jp/kegg/, accessed on 6 November 2021) [16]. 3. Results3.1. Identification of Skim Milk Proteins of YaksiTRAQ analysis identified 767 proteins in the skim milk samples of TL and HL yaks based on the 1930 unique peptides obtained (Supplemental Material File S1). The protein coverage was between 0.001 to 0.999, of which 37.1% was identified as using at least two unique peptides. The length of most acquired peptides ranged between 7 and 17 amino acids (Figure 1A), and the molecular weights of approximately 80% of the identified proteins were less than 100 kDa (Figure 1B).3.2. GO and COG Annotation of Identified ProteinsAll identified proteins in the skim milk samples of TL and HL yaks were subjected to the GO analysis and were categorized into biological processes, cellular components, and molecular functions, based on their GO annotations (Figure 2, Supplemental Material File S2). The proteins identified were classified by molecular functions and were enriched in 13 different functional terms. The majority were related to binding and catalytic activity (471 and 276 proteins in total, respectively), followed by structural, molecular activity, transporter activity, and enzyme regulator activity. The GO cellular location classifications of the proteins were involved in 16 categories, and the most obvious differences were in the cell, cell part, organelle, and organelle part. There were 23 biological process categories in which the identified proteins were involved, and the largest number of proteins were observed in the cellular process (475), followed by the metabolic process (388), the single-organism process (385), and biological regulation (299).COG annotation of all the identified proteins revealed 24 functional categories (Figure 3, Supplemental Material File S3). Among these COG categories, the skim milk samples of TL and HL yaks were highly enriched in several major functional COG categories, including post-translational modification, protein turnover, chaperones, energy production and conversion, cytoskeleton, carbohydrate transport, and metabolism. 3.3. Differentially Expressed Milk Proteins of YaksA total of 202 differentially expressed proteins (DEPs) were revealed according to the standards of fold-change ratios ≥ 1.2 and p < 0.05, among which 109 proteins were up-regulated and 93 proteins were down-regulated in the milk of HL yaks compared to TL yaks (Supplemental Material File S4 and S5). The top 50 DEPs with at least 2.18- and 1.54-fold change, respectively, are listed in Table 1 and Table 2. The 14-3-3 protein theta showed the largest fold change (increased approximately 10-fold in HL yak milk). Another 11 proteins were up-regulated by approximately 4-fold, including spliceosome RNA helicase BAT1, dolichyl-diphosphooligosaccharide-protein glycosyltransferase 48 kDa subunit, small nuclear ribonucleoprotein, protein S100-A1, tubulin beta-7, 40S ribosomal protein S3, glycogen phosphorylase, and hemoglobin subunit beta (Table 1). Four kinds of caseins were identified, including κ-casein, αs1-casein, αs2-casein, and β-casein. Compared with those of TL yaks, these caseins were up-regulated in HL yaks by 2.04-, 2.48-, 2.35-, and 1.43-fold, respectively. The enzyme γ-glutamyltransferase2, vitamin D-binding protein, retinol-binding protein 4, lactotransferrin, serotransferrin, keratin, cysteine-rich secretory protein 3 precursor, and vinculin were also up-regulated in the milk of HL yaks (Table 1, Supplemental Material File S4).3.4. GO Enrichment Analysis of DEPsGO enrichment analysis of the 202 DEPs between TL and HL yaks demonstrated that 13, 19, and 22 protein categories were highly enriched in the cellular component, molecular function, and biological process categories with p < 0.05 or p < 0.01, respectively (Figure 4). Among these, the vesicle, cytoplasmic vesicle, and membrane-bounded vesicle were the most abundant categories in the cellular component (Figure 4A, Supplemental Material File S6). The transporter activity, enzyme regulator activity, and enzyme inhibitor activity were the most abundant categories in the molecular function (Figure 4B, Supplemental Material File S6). In the biological process, as many as 65 GO terms were enriched with p < 0.05, and the 22 major categories (p < 0.01) were related to oxidation-reduction processes and the generation of precursor metabolites and energy (Figure 4C, Supplemental Material File S6).3.5. Pathway Enrichment Analysis of DEPsThe 202 DEPs were used for pathway enrichment analysis, and 25 major pathways were highly enriched via KEGG with p < 0.05 (Supplemental Material File S7). In addition, a scatter plot for the top 20 of KEGG enrichment results is shown in Figure 5. The KEGG pathways that the DEPs mainly participated in were: primary immunodeficiency, staphylococcus aureus infection, cytokine-cytokine receptor interaction, and dilated cardiomyopathy. The metabolic pathways with the most differences involved carbohydrate metabolism, the biosynthesis of amino acids, and fat digestion and absorption.4. DiscussionOur previous research discovered some differences in milk between TL yaks and HL yaks [6]. For example, the contents of protein and fat and the activities of several enzymes, such as γ-glutamyltransferase and alkaline phosphatase, in HL yak milk were significantly higher compared to the milk of TL yaks. It has been reported that the yield and composition of milk are influenced by many factors, including human maternal age, time of delivery and maternal diet, and the stage of lactation, which was the most influential factor [17]. The detailed composition and quality of milk from extended lactation have been studied in cows and humans. One study found higher protein and fat concentrations, an unaffected casein to protein ratio, and protein composition of the bovine milk from the extended lactation [18]. Czosnykowska-Łukacka et al. reported that the concentration of carbohydrates in mother’s milk showed a negative correlation with lactation of about two years, while fat and protein concentrations were opposite. Moreover, during prolonged lactation in humans (over 18 months), it was found that the concentration of carbohydrates significantly decreases, and fat and protein concentration significantly increases [19]. The analysis of the composition of prolonged yak milk allows one to assess the nutritional value of milk. In this study, the milk protein profiles of TL yak and HL yak were intensively studied based on the iTRAQ technique. The expressions of some major milk proteins in HL yaks were increased, such as κ-casein, αS1-casein, αS2-casein, and β-casein. In addition, vitamin D-binding protein, retinol-binding protein 4, lactotransferrin, and serotransferrin levels in HL yak milk also increased (Supplemental Material File S4). Proteins are the major nutrients of milk and have many functions except for providing proteins for nutrition purposes. The micronutrients in milk can also affect its function [20]. The present results indicate that HL yaks can provide more nutrients in milk compared to TL yaks.Based on their GO functional annotations, all identified proteins in the skim milk samples of TL and HL yaks were classified in accordance with molecular function, cellular localization, and biological pathways. It was reported that 66% of the DEPs identified in whey from yak colostrum and mature milk were found to be related to binding activity [21]. Approximately 44% and 22.4% of the identified proteins in bovine colostrum were involved in catalytic activity and binding activity, respectively [22]. Our present results are basically consistent with these reports, and the catalytic activity and binding activity account for the largest proportion of the identified proteins (Figure 2).The mammary gland undergoes morphological and functional changes during development. The lactation cycle of cows includes early lactation, middle lactation, late lactation, and dry lactation [23]. Mammary gland degeneration is a key stage in dry lactation, and during this stage, the ability to synthesize milk decreases and cell apoptosis increases [4]. It was shown that the concentration of lactoferrin in mammary secretions during the dry period was significantly increased, which could be used to measure the degree of mammary degeneration in dairy cows [24]. In this study, the expressions of hemoglobin subunit-beta, lactotransferrin, and serotransferrin in HL yak milk increased significantly compared to TL yak milk. Since these components are blood-derived proteins, it suggests that the permeability of mammary gland tight junctions increase in HL yaks, which is consistent with the characteristics of cows in the degeneration stage. In addition, during mammary gland degeneration, the somatic cell count and the apoptosis rate of mammary epithelial cells increase [25]. This study found that HL yaks contained significantly higher levels of several types of keratins in milk than TL yaks (Supplemental Material File S4), which may be a result of increased shedding of mammary epithelial cells in HL yak milk since keratin is a marker of epithelial cells [26]. Moreover, cysteine-rich secretory protein 3 (CRISP-3) precursor, which is a key protein in cell apoptosis [27], and vinculin, which can maintain cell growth and differentiation and promote cell survival [28], were also up-regulated and down-regulated in HL yak milk, respectively. This suggests that the mammary glands in HL yaks are at a state of degeneration, consistent with the characteristics of mammary glands of degenerated cows.During the dry period, the ability of mammary epithelial cells to synthesize lactose, milk fat, casein, α-lactalbumin, and β-lactoglobulin decreases, while the concentration of lactoferrin in mammary secretions increases significantly [4,7]. GO enrichment analysis of DEPs showed that the vesicle and cytoplasmic vesicle were the most abundant categories in the cellular component (Figure 4A). In contrast, the transporter activity was the most abundant category in the molecular function (Figure 4B). Vesicles play a key role in protein transport [29], and the difference in protein transport activity may be related to milk protein components in HL yak milk. The KEGG pathway enrichment analysis, which was based on the DEPs, revealed that the metabolic pathways with the most differences were those that involved carbohydrate metabolism, the biosynthesis of amino acids, and fat digestion and absorption. Previous research found that a high number of proteins in human and ruminant milk serum were related to metabolic processes [30]. In another study, metabolism-related pathways (such as glycolysis/gluconeogenesis and biosynthesis of amino acids) also enriched many differentially expressed whey proteins in human and bovine colostrum and mature milk [31]. The differences in these metabolic pathways were related to the ability of mammary epithelial cells to synthesize lactose, milk fat, and milk protein in HL yak.In the sample, the pooled skim milk samples of TL yaks and HL yaks were prepared by mixing equal volumes of 15 skim milk samples of corresponding yaks, respectively. This procedure clearly blurs the individual differences between the tested yak females. However, the lactation period between yak groups was similar. The same sample processing method was used in previous studies [32].5. ConclusionsThe iTRAQ technique was used to compare skim milk proteins in yaks from standard and naturally extended lactation (TL yaks and HL yaks, respectively). A total of 202 differentially expressed proteins were identified, among which 109 proteins were up-regulated and 93 were down-regulated in HL yaks compared to TL yaks. The GO function annotation and pathway enrichment analysis suggest that HL yaks produce milk with characteristics that reflect the degeneration stage, similar to that of dairy cows.

animals : an open access journal from mdpi

10.3390/ani11092608

PMC8466940

The reproductive potential and longevity of rabbit does are the real determinants of the economic profitability of rabbit production. Another important characteristic is the quality of rabbit’s milk, as it determines the survival and growth of rabbit kits. The goal of this study was to examine the effect of parity order on the milk quality and the reproductive performance of Hycole does housed under intensive conditions of a commercial rabbit farm. The results of this trial and its duration allowed to trace the reproductive performance of Hycole females throughout their lifespan and to indicate the age boundary of a profitable reproductive performance of Hycole does.

The goal of this study was to analyze the reproductive performance of does, growth of their kits, and chemical composition of their milk over nine consecutive parities in order to indicate the boundary of female reproductive profitability. The novelty of this study results from the combinations of three factors: extensive reproductive rhythm, commercial farming conditions, and a period of nine consecutive parities, showing the actual lifespan of a rabbit doe on commercial farms. The data was collected on 60 Hycole females kept at a commercial rabbit farm. Throughout the study, 32 does were excluded due to different reasons (e.g., excluded by means of selection—43.8% and mortalities—25.0%). The does were first inseminated at 28 weeks of age. Following artificial inseminations were conducted 14–15 days after each parturition. All kits were weaned at the age of 35 days. The following characteristics were analysed: body weight of rabbit does at artificial insemination, milk production per lactation, litter size, litter weight, average kit weight, and milk chemical composition. Rabbit does had a significant decrease in kindling rate between the eighth and the ninth parity (by 10.0 percentage points; p = 0.039). The litter size at weaning in the ninth parity was significantly lower to litters weaned at other analysed parities. The amount of milk produced per lactation was affected by the parity order (6.31–6.76 kg; p = 0.042). The litter weights on day 21 and 35 were the lowest at ninth parity. The content of total solids (TS), solids-not-fat, and fat was affected by the parity order on both analysed lactation days. The content of TS and fat in rabbit milk was characterized with a decreasing trend over the analysed period, on both lactation days. The results clearly indicate that rabbit does under extensive reproductive cycles characterize with a very good reproductive performance and can be successfully used for reproduction even up to the eighth parity. However, further research is needed if keeping them longer will not be profitable.

1. IntroductionThe reproductive longevity of rabbit does is listed among the most important traits of parental lines of rabbits and is determined by health and high reproductive performance [1]. Even though there is a common assumption that rabbit does are commercially used for reproduction for a maximum of 8–10 reproductive cycles, only a few studies support this decision statistically [2]. This comes as no surprise since most research on reproduction in rabbits presents results for two to three reproductive cycles, which are only a small part of their reproductive life [3,4,5]. Moreover, the vitality and growth of kits during the lactation period is dependent on the milk production and promoted by the unique chemical composition of rabbit milk. The health, growth rate, and survival of rabbit kits during the nursing period determines the litter size and litter weight at weaning, therefore, the fundamental understanding of the role of rabbit milk in the development of kits is needed. Based on the above, the thorough evaluation of does performance over their lifetime should covers multiple parities and combines the analysis of female reproduction traits with kits growth and milk quality.The goal of the study was to trace the changes in the reproductive performance of does kept under an extensive reproductive cycle, as well as growth of their kits and chemical composition of their milk. The novelty of this study results from the combinations of three factors: extensive reproductive rhythm, commercial farming conditions, and a period of nine consecutive parities, showing the actual lifespan of a rabbit doe on commercial farms. Another unique aspect of this study is examination of rabbit milk chemical composition over nine consecutive parities.2. Materials and Methods2.1. Animals and Housing ConditionsThe study was conducted on a commercial rabbit farm located in the western Poland. The building was equipped with industrial facilities (mechanical ventilation and conventional industrial furnace) allowing to maintain the uniform microclimate conditions all year round (temperature 16–20 °C, humidity 60–75%, and lighting schedule of 16 h light/8 h darkness). The reproductive performance of parental Hycole does was evaluated up to the ninth parity. The females were obtained from crossbreeding grandparental bucks (GPC) with grandparental does (GPD) [6]. At the start of the study, a total of 60 females was included. Rabbit does and their kits were kept in conventional wire-mesh cages including a feeder, nipple drinker, and a closable nest box. A single cage was 40 cm × 85 cm × 35 cm (width × length × height). Each cage was connected to a nest with dimensions of 30 cm × 45 cm × 30 cm. Throughout the study period, rabbit does were fed ad libitum with a commercial pelleted feed. The females and their kits had unlimited access to fresh water. First artificial insemination (AI) of nulliparous does was conducted at 28 weeks of age. This rather late age of first AI was the decision of the farm owner, who based it on observed improved reproductive performance. The farm was managed under a single-batch system. Artificial insemination was repeated each 45 days to inseminate the does 14–15 days after each parturition. The effectiveness of AI was checked by palpation performed 10 days after the procedure. In the case of ineffective AI, does were excluded from the experiment. Also the mortalities and other reasons to exclude the females from the study were recorded (i.e., small litter size, insufficient maternal capacities, severe mastitis). After birth, the litters were standardized by cross-fostering to obtain 10 kits per nest. From the moment of standardization until insemination, the nests were opened only once a day for about 10 min. After AI, the does had unlimited access to the does, drinker, and feeder.2.2. Reproductive Performance CharacteristicsThe reproductive performance during the nine consecutive parities was assessed. The collected and analysed traits were: the live body weight of rabbit does at AI, kindling rate (the percentage of does kindling related to the number of inseminated does), litter size (at birth, and after standardization, on days: 14, 21, and 35 of lactation), and litter weight. The other trait analysed in this study was milk production per lactation (MP) calculated according to the equation developed by De Blas et al. [7]:MP (kg) = 0.75 + 1.75 × W21 (kg)(1)MP—milk productionLW21—litter weight at 21 day of lactation2.3. Rabbit Milk Recovery and ExaminationThe rabbit milk recovery and quality examination were conducted using the methodology described in the study of Ludwiczak et al. [8]. The samples were collected during each of the nine analysed parities, on the 2nd and 21st day post-partum. A quantity of 12 mL of milk per rabbit doe was obtained into a plastic probe by gently massaging the mammary glands, and chilled to 4 °C directly after recovery. The milk was collected from randomly selected 25 females. Samples were transported under chilled conditions for examination of basic chemical composition (total solids, solids-not-fat, fat, protein, casein, lactose, and ash). The chemical composition of milk was determined by automated infrared analysis with a MilkoScan FT 120 analyzer (Foss Electric, Warsaw, Poland).2.4. Statistical AnalysisAll statistical analysis performed in this study were done with the SAS software package ver. 9.4 [9].The PROC MIXED was applied to estimate the effect of the parity order on the body weight of rabbit does, litter weight, milk production and milk chemical composition (model 1), effect of the parity order on the litter size (model 2), as well as the effect of lactation day on the milk chemical composition (model 3). For this, the following three models were used:Yijkl = μ + αi + βj + dk + eijkl (model 1)(2) where:μ—the overall mean of the analysed trait,αi—the fixed effect of the parity (i = 1, 2, …, 9),βj—the fixed effect of litter size (l = 4, …, 15),dk—random effect of the female (j = 1, 2, …, 60),eijkl—random error. Yijk = μ + αi + dj + eijk (model 2)(3) where:μ—the overall mean of the analysed trait,αi—the fixed effect of the parity (i = 1, …, 9),dj—random effect of the female (j = 1, …, 60),eijk—random error. Yijkl = μ + γi + βj + dk + eijkl (model 3)(4) where:μ—the overall mean of the analysed trait,γi—the fixed effect of day of lactation (i = 2, 21),βj—the fixed effect of litter size (l = 4, …, 15),dk—random effect of the female (j = 1, …, 60),eijkl—random error.The random effect of a female was included in each model to correct for repeated observations per doe. Litter size as a fixed effect was added to the model to correct for the number of rabbits in the litter at the time of data collection. The “parity order*lactation day” interaction was initially also included in the model for the milk characteristics, but it was not statistically significant.The kindling rate and the percentage of rabbit does excluded from the study was compared by χ2 test (PROC FREQ procedure in SAS). Tukey–Kramer adjustment was implemented for multiple comparisons of Least Squares Mean (LSM) differences, which was included as an additional analysis next to the models presented above.3. Results3.1. Reproductive Performance of Hycole Does and the Growth of Rabbit KitsIn Table 1 are presented all reasons for excluding the does and the percentage of animals affected by it. The fertility problem was the most often reason to exclude a female (43.8% of all cases), whereas the second most common reason was mortality during lactation (25% of all cases). The fact that does were removed from the farm by means of selection strategy allowed to maintain a high kindling rate up to the eight parity. High fertility was also promoted by controlled nursing. The amount of milk produced per lactation ranged from 6.31 kg to 6.76 kg, and was higher for the third and sixth parity compared to the first, second, seventh, eighth, and ninth parity. The reproductive performance of Hycole does examined in our study is presented in Table 2. The body weight of does at AI, kindling rate, and litter size varied between parities. The body weight of does was highly affected by the parity order (p = 0.001). The lowest body weight was observed at first parity (4.67 kg), whereas from the first to the fourth parity the body weight gradually increased. The total recorded difference in the body weight of females between first and fourth parity reached 8.5% increase. From the seventh parity till the end of the experiment this trait was almost stable. All rabbit does were characterized by a high kindling rate throughout the examined period. A clear decrease of this trait could only be observed between the eighth and the ninth parity (by 10.0 percentage points). The slightly significant effect of the parity order on the kindling rate (p = 0.039) can be explained by the fact that all the females with serious fertility problems (i.e., low litter size, severe mastitis, sore hocks or poor body condition) were excluded over the course of the study.We also observed that the total born kits, born alive, and litter size on 21st and 35th lactation day significantly varied between consecutive parities (Table 3). The number of total born and born alive kits was the lowest at first and ninth parity (11.6 and 11.4 kits.) No significant differences were recorded on day 14 of lactation. The differences between parities recorded in litter size observed on days 21 and 35 were caused by kits mortality, which was significantly affected by the parity order. Litter weight at birth, on days 14, 21, and 35 of lactation significantly varied between the consecutive parities (Table 4). While the average kit weight was affected by the parity order only on day 35 post-partum (p = 0.001). The average kit weight recorded on day 35 of eighth and ninth parity showed a tendency to be lower compared to other examined parities.3.2. Chemical Composition of Rabbit MilkThe effect of parity order on chemical composition of rabbit milk collected on day 2 and 21 is given in Table 5 and Table 6. On day 2 of lactations, the parity order caused a variation in the level of total solids (p = 0.001), SNF (p = 0.012), and fat (p = 0.039) in rabbit milk. The TS at first parity was much higher compared to parities eight and nine, by 2.36%. SNF showed the lowest values at the last two parities, and a decreasing trend over the analysed period. On day 21 of lactations, the parity order affected the content of TS (p = 0.044), fat (p = 0.015), and SNF (p = 0.036). TS and fat decreased over the analysed period. Fat content was significantly lower at parities eight and nine compared to parities one to four. Obviously, the content of chemical compounds in milk was affected by the day of lactation (Figure 1). Milk collected on day 2 was characterized by a greater content of TS, fat, and ash, and a lower content of protein, casein and lactose, compared to milk from day 21.4. DiscussionHycole does characteristics given by the producer emphasize the high reproductive performance of these rabbits: age of reproductive maturity in the range of 17–19 weeks, 89% birth rate, 10.7 kits born alive/kindling, 9.3 kits weaned/kindling, and 4.8–5.0 kg adult weight [6]. We observed an even better reproductive performance of this synthetic line of rabbits, and there are a few factors that promoted these results. Most commercial farms, and therefore, studies performed on rabbit does under intensive farming conditions, present the intensive reproductive rhythm with AI performed up to the 11th day post-partum. In our study, the extensive reproductive rhythm was used with the late age of the first service and as a result, almost had adult body weight [10]. This unquestionably had a positive effect on the reproductive performance of does, which allowed them to maintain high reproductive performance throughout the analyzed period. This can be explained by the already mentioned late age at first service, as well as extensive reproductive rhythm, properly performed culling strategy, and controlled nursing. Eiben et al. [4] also observed the positive effect of controlled nursing on reproductive performance in rabbits with higher fertility rate compared to free nursing (85.5% vs. 71.1%; p < 0.05).Moreover, the literature underlines the relation between reproductive performance and the condition of does. Rebollar et al. [11] stated that nulliparous does characterize with higher fertility compared to the multiparous does, as the latter are exposed to significant energy deficits. In many studies it is highlighted that the inability of young does to meet high energy requirements for pregnancy and lactation during the first litters leads to high culling rates [12,13]. According to Rosell and de la Fuente [14], the average culling age for breeding does is 14.9 months and 6 parities. The authors observed that the first three parities are characterised by the highest risk of culling and mortality. Low productivity was given among the major causes of culling, while mastitis, poor condition, or sore hocks were noted less often. Rosell and de la Fuente [15,16] point to respiratory tract disorders as the main cause of mortality (including rabbits euthanized due to respiratory problems). Although we have noted some mortalities among rabbit does at different stages of reproductive cycle, their reproductive performance was high. This allows to speculate about other reasons for these mortalities than poor condition and energy deficits.We have noted a significant influence of the parity order on the rabbit does’ body weight measured at AI. According to the literature, the changes of doe body weight with consecutive parities may be related to the effect of parity order on the body energy deficit [17,18]. Rabbit does in our study produced over 6.0 kg of milk per lactation. Although the milk production was generally high despite the parity order, the effect of parity on this trait was also clearly marked. The level of milk production per lactation was previously analysed by De Blas et al. [7] and ranged from 5.73 to 6.06 kg in a 30 d lactation. Moreover, the kits of does with the highest milk production (6.06 kg) were characterised with the highest litter weight on day 21 and at weaning. In the study of Xiccato et al. [19], the reproductive performance of rabbit does was analysed over three consecutive parities. The authors observed that the milk production increased with parity order, from 4548 g at first parity to 5410 g at third parity (p < 0.001). If we considered only the first three parities, we could observe the same tendency compared to Xiccato et al. [19]. We have noted an increase in milk production between the first and third parity, by 7.3%. Because in our study the litter weigh was affected by litter size and mortality, the parities with highest milk production did not exactly overlap with the parities with the greatest litter weaning weights.The decrease in litter size from litter standardization till weaning recorded in our study was much lower compared to data in the literature [10]. The authors, Whitney et al. [20] conducted a survey on causes of pre-weaning mortality among young rabbits and noted that 12.4% of rabbits on commercial farms die in the pre-weaning period. From the total number of mortalities (347 kits) registered in the period from 0 to 4 weeks, 32.0% were stillborn; 21.3% deaths were caused by maternal neglect, inanition or hypothermia; 21.0% deaths were connected with inadequate husbandry, culling or fostering; and 18.4% causes of pre-weaning death cases remained not diagnosed. According to Rashwan and Marai [21], the pre-weaning mortality can be reduced through selection for greatest resistance to diseases in rabbits and is strongly related to the milk yield of the doe. Therefore, all the factors that decrease the milk production will lead to increased pre-weaning mortality. Therefore, high litter size at weaning recorded in our study was a result of a group of factors, with major ones being good condition of does and high milk production.Similarly to our results, Mikó et al. [22] found that the parity order affected all the evaluated reproductive traits of rabbit does, including does body weight at AI, litter size, and litter weight. Litter weight is a composite of the number and individual weight of kits. According to the literature, the parity order and physiological status of rabbit does are among the major factors deciding about the birth weight of kits. Research conducted by Parigi–Bini and Xiccato [23] showed that kits of multiparous does were even 10% heavier at birth compared to kits from primiparous does. Rommers et al. [24] highlighted the individual milk intake and the litter size as the major factors deciding about the pre-weaning growth of kits. The effect of parity on litter weight observed in our study on days 14, 21, and 35, was rather related with kits mortality and litter size reduction than with the kit average weight.Our study is unique as it evaluates the milk composition and quality in rabbits over nine parities. The knowledge that was available so far on this topic is very insufficient as the previous studies covered only short time periods, i.e., two or three parities [3,5]. These research neither reflect the true pattern of rabbit milk composition changes with parity order nor the effect of milk quality on kits growth and mortality. Because of the aforementioned, we decided to discuss our results with data obtained from studies on other farmed animals. According to the existing knowledge on farmed dairy cattle, the content of chemical compounds in milk and parity order are in strong relation, and the content of chemical compounds in milk tends to decrease with the parity order [25,26]. Similar observations were made in our study, showing that the rabbit milk composition has the same direction of changes compared to milk of farmed dairy animals, although the lactation in rabbits lasts only 30 days. The effect of lactation day on the chemical composition of rabbit milk observed in our study is consistent with the available literature [5,8,27,28,29].5. ConclusionsTo conclude, Hycole showed fluctuations in fertility, litter size, mortality, litter weight, milk production, and milk chemical composition over nine consecutive parities. The observation that should draw most attention is the kindling rate that is clearly decreasing between the eighth and the ninth parity. This indicates a decrease in reproductive performance of rabbit does around their eighth to ninth reproductive cycle. Nevertheless, the Hycole used in this study had a very good reproductive performance and high level of milk production throughout their reproductive lifetime. These results are partially related to the late age of their first service and an implemented extensive reproductive rhythm.

animals : an open access journal from mdpi

10.3390/ani11113158

PMC8614246

Boar taint is a meat quality issue that results from the accumulation of androstenone in the adipose tissue. During steroid synthesis, steroids such as androstenone undergo sulfoconjugation, a process that involves the attachment of a sulfonate group to enhance polarity. Androstenone sulfate is more abundant in the plasma than free androstenone and is suspected to enzymatically regenerate free androstenone in peripheral tissues such as the fat to indirectly contribute to boar taint development. In this article, we identified a specific membrane transporter that is responsible for the uptake of androstenone sulfate into the fat and confirmed that androstenone sulfate can enzymatically regenerate free androstenone within the adipose tissue. We also identified a positive relationship between the quantity of free androstenone enzymatically produced from androstenone sulfate and fat androstenone concentrations in early maturing boars. These results suggest that the production of free androstenone from androstenone sulfate may contribute to the development of boar taint in early maturing animals.

Boars express high testicular levels of sulfotransferase enzymes, and consequently, the boar taint causing compound androstenone predominantly circulates as a steroid sulfate. Androstenone sulfate is suspected to function as a steroid reservoir that can be deconjugated to provide a source of free androstenone for accumulation. Therefore, the purpose of this study was to characterize the uptake and deconjugation of androstenone sulfate in the adipose tissue of the boar. Real-time PCR was used to quantify the expression of steroid sulfatase (STS) and several organic anion transporting polypeptides (OATPs) in the adipose tissue. Additionally, [3H]-androstenone sulfate was incubated with adipocytes or supernatant from homogenized fat to assess steroid uptake and conversion, respectively. A positive correlation existed between OATP-B expression and androstenone sulfate uptake (r = 0.86, p = 0.03), as well as between STS expression and androstenone sulfate conversion (r = 0.76, p < 0.001). Moreover, fat androstenone concentrations were positively correlated (r = 0.85, p < 0.001) with androstenone sulfate conversion and tended to increase with STS expression in early maturing boars. This suggests that androstenone sulfate uptake and deconjugation are mediated by OATP-B and STS, respectively, which may influence the development of boar taint in early maturing animals.

1. IntroductionBoar taint is an undesirable flavor or odor that develops in heated pork products from entire male pigs and is caused by the accumulation of androstenone (5α-androst-16-en-3-one) and skatole (3-methylindole) in the adipose tissue [1]. Androstenone is a sex pheromone that is synthesized in the testis during steroidogenesis and is sulfoconjugated by the sulfotransferase enzyme SULT2A1 before entering the systemic circulation [2,3]. Androstenone sulfate is the predominant form of androstenone in the peripheral plasma of the boar, accounting for approximately 70% of the total androstenone present in the circulation [3,4]. However, the role of androstenone sulfate in the boar has yet to be elucidated.Sulfoconjugation involves the transfer of a sulfonate group (SO3−) from 3′-phosphoadenosine 5′-phosphosulfate (PAPS), a donor molecule, to the 3-hydroxyl position of an accepting steroid, which functions to inactivate and increase the water solubility of steroids [5]. Originally, sulfoconjugation was regarded as a mechanism to facilitate steroid excretion, and steroid sulfates were considered to be metabolic end products [5]. However, steroid sulfates were later found to function as steroid reservoirs, which are transported by membrane transporters belonging to the organic anion transporting polypeptide family (OATP) into various tissues and deconjugated by steroid sulfatase (STS) to return free bioactive steroids [6,7]. In humans, OATPs such as OATP-B, OATP-E, OATP-A, and OATP-D are encoded by the solute carrier organic anion (SLCO) gene and facilitate the sodium and ATP-independent uptake of sulfated steroids into several tissues [7,8], while STS is a microsomal enzyme that is expressed ubiquitously in small quantities, which hydrolyzes dehydroepiandrosterone sulfate (DHEAS) and estrone sulfate (E1S) [6].The sulfoconjugation of androstenone is thought to require enolisation of the 3-keto group to produce a 3-enol intermediate, which can accept a sulfonate group from PAPS. Recently, a metabolite tentatively identified as androst-3-enol-3-sulfate was detected by liquid chromatography–mass spectrometry from Leydig cell culture and was found to return free androstenone, and not a hydroxylated metabolite, following chemical removal of the sulfate group [9]. Additionally, we have previously demonstrated that androstenone sulfate has a low binding capacity for porcine albumin, which is the carrier protein responsible for the transport of various steroids in the boar including free androstenone [10,11]. Consequently, androstenone sulfate circulates predominantly unbound in the porcine plasma and is presumably readily available for uptake into peripheral tissues [11,12]. On this basis, we hypothesized that androstenone sulfate may function as a steroid reservoir that is transported by OATPs into peripheral tissues such as the fat and hydrolyzed by STS to return free androstenone, which may subsequently accumulate to cause boar taint. Additionally, it is likely that this process could vary between individual animals due to differences in STS expression or hormonal status.Therefore, the purpose of this study was to characterize the uptake and deconjugation of androstenone sulfate in the adipose tissue of boars with varying sulfatase expression and hormonal status to determine if androstenone sulfate can indirectly contribute to the development of boar taint.2. Materials and Methods2.1. Sample CollectionPlasma and backfat samples were obtained from 16 terminal cross [Duroc × (Yorkshire × Landrace)] boars. The boars were housed in pens with slatted floors in groups of approximately 5 beginning at 7 weeks of age and were provided ad libitum access to water and standard starter, grower, and finisher rations, formulated by Flordale Feed Mill Limited. All animals were used in accordance with the guidelines of the Canadian Council of Animal Care and the University of Guelph Animal Care Policy. A single pre-slaughter blood sample was collected from each animal at 120, 130, and 140 kg live weights. Plasma samples were analyzed using an E1S specific radioimmunoassay, previously described by Raeside and Renaud [13], in order to assess hormonal status. At 188 ± 3 days of age and approximately 160 kg live weight, the boars were electrically stunned and exsanguinated, and backfat samples were collected in liquid nitrogen from all boars and stored at −80 °C, allowing for subsequent evaluation of sulfatase activity and expression. Fresh backfat samples were also collected from 6 boars and immediately used for primary adipocyte culture. Fat androstenone concentrations were determined in backfat samples from all boars using an established reverse phase high-performance liquid chromatography (HPLC) technique previously described by Hansen-Møller [14], where dansylhydrazine is used to derivatize androstenone extracted from fat, which allows for subsequent quantification by florescence detection.2.2. RNA Extraction and Gene Expression AnalysisFat tissue was kept frozen in liquid nitrogen and pulverized with a mortar and pestle. Approximately 100 mg of powdered fat tissue was homogenized in 1 mL lysis buffer, and RNA was subsequently extracted using silica-based spin columns (RNeasy Lipid Tissue Mini Kit, Qiagen, Hilden, Germany). The RNA concentration was quantified using a NanoDrop 8000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) and the RNA integrity was assessed using an Agilent 2000 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA).RNA (1 µg) was reverse transcribed in 20 µL final volume with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Waltham, MA, USA) according to the manufacturer’s instructions. After incubation at 25 °C for 10 min, reverse transcription was carried out at 37 °C for 120 min, followed by 85 °C for 5 min. The resulting cDNA was diluted 5× and amplified by real-time PCR using a QuantaStudio Real Time PCR system (Thermo Fisher Scientific) in a 20 µL reaction volume containing 10 µL SsoAdvanced Universal Inhibitor-Tolerant SYBR Green Supermix (Bio-Rad, Hercules, CA, USA), 4.2 µL water, 5 µL diluted cDNA, and 200 nM of the respective primers, which are listed in Table 1. The run conditions were as follows: 3 min at 98 °C for polymerase activation, followed by 40 cycles of two-step qPCR (10 s of denaturation at 98 °C, 30 s of combined annealing/extension at 60 °C). Real-time PCR reactions were run in triplicate and gene expression was calculated using the ∆∆Ct method [15] using β-actin as a housekeeping gene and barrow fat as the calibrator.2.3. Radiolabeled Androstenone Sulfate Synthesis and PurificationHuman embryonic kidney (HEK293FT) cells purchased from ATCC (Manassas, VA. USA) were used to synthesize [3H]-androstenone sulfate as previously described by Bone and Squires [11]. Briefly, confluent HEK293FT cells were transfected with 6 µg/plate of the porcine sulfotransferase SULT2A1 (pSULT2A1) expression vector constructed as previously described by Laderoute et al. [2], and, after 48 h, treated with radiolabeled [3H]-androstenone (8 million CPM, 36 µCi/µmol, 0.1% ethanol). Cell culture media was collected 24 h later and analyzed by reverse-phase C18 HPLC using a Luna 5µ C18(2) HPLC column (250 × 4.60 mm) purchased from Phenomenex (Torrance, CA, USA). The 40 min HPLC profile used to confirm the conversion of free androstenone to androstenone sulfate by the HEK293FT cells was previously described by Laderoute et al. [9] and optimized for the elution of 16-androstene steroids. The elution of radiolabeled androstenone and androstenone sulfate occurred at approximately 32 and 16 min, respectively, and was monitored by a β-RAM model 2 isotope detector (IN/US Systems, Brandon, FL, USA). Radiolabeled androstenone sulfate was then isolated from the media by solid phase extraction using Sep-Pak C18 solid-phase chromatography cartridges (Waters, Milford, MA, USA), as previously described by Laderoute et al. [2]. The sulfated steroid fraction was dried under nitrogen and reconstituted in 100% ethanol.2.4. Porcine Adipocyte IsolationMature adipocytes were isolated and cultured as previously described by Alexandersson et al. [23] with modifications. Briefly, 50 g of fat was minced, and the resulting homogenous mixture was added to 250 mL of digestion buffer (TC 199) containing 4.2 mM NaHCO3, 15 mM bovine serum albumin (BSA), 5.5 mM D-glucose, 1% penicillin–streptomycin, collagenase type I (1 mg/mL), 1.6 mM DNase, and 2.5 mM trypsin inhibitor. The fat was digested in a shaking water bath at 37 °C for approximately 40 min. The digested fat solution was then filtered using a 255 µm nylon mesh filter and transferred into a separation funnel. The remainder of the separation funnel was filled with warm wash buffer (TC 199) containing 4.2 mM NaHCO3, 15 mM BSA, 5.5 mM D-glucose, and 1% penicillin–streptomycin and then was gently inverted to mix the wash buffer and the digested fat solution. After 3 min, the fat layer had separated from the buffer, and the buffer was subsequently drained from the separation funnel. This was step was repeated three times to remove any remaining collagenase from the digestion. After washing, the mature adipocytes were collected into a 50 mL conical tube and centrifuged at 50× g for 8 min, which allowed oil from damaged cells and remaining wash buffer to be separated and removed.2.5. Steroid Transport Studies using Mature AdipocytesIsolated adipocytes (2 mL containing approximately 4 million cells) were suspended in 5 mL incubation buffer (DMEM/F12 mixed 1:1 (v/v) with α-MEM) containing 15.7 mM HEPES, 17.5 nM insulin, 1% fetal bovine serum (FBS), and 1% penicillin–streptomycin in 50 mL flasks and was incubated with [3H]-androstenone sulfate (24,000 CPM, 18.3 nCi/nmol) for 3 to 24 h in a shaking water bath at 37 °C with 95% air and 5% CO2. [3H]-Androstenone sulfate was incubated with media in the absence of isolated adipocytes as a negative control. All incubations were run in triplicate. Following incubation, the adipocytes were frozen, and the media was removed using a syringe. The adipocytes were then rinsed with wash buffer and melted in a hot water bath. Steroids were extracted once with 8 mL methanol, which was dried to a total volume of approximately 500 µL under nitrogen. The media was extracted twice with 4 mL ether, which was dried under nitrogen and reconstituted with 50% acetonitrile (500 µL). The media and fat extracts were then filtered and analyzed by reverse-phase C18 HPLC using the aforementioned 40 min HPLC profile optimized for the elution of 16-androstene steroids [9]. Steroid transport was calculated from the sum of free and sulfated androstenone present in the fat extract, which was quantified from the peak areas detected for each steroid and expressed as a percentage of the total steroid added to the incubation. This was used to calculate total steroid uptake (pmol) and steroid uptake rate (pmol/h).2.6. Sulfatase AssayThe deconjugation of androstenone sulfate in the fat was quantified using a sulfatase assay previously described by Dalla Valle et al. [7] with modifications. Briefly, 2 g of backfat was thinly sliced and homogenized in 5 mL buffered medium (100 mM KCl, 16 mM K2HPO4, 4 mM KH2PO4, 1 mM DTT, 1 mM EDTA, 4 mM nicotinamide). The homogenate was centrifuged at 2000× g for 15 min at 4 °C, and 1 mL total volume of the resulting supernatant was incubated with [3H]-androstenone sulfate (28,000 CPM, 32 nCi/nmol) in a shaking water bath at 37 °C for 3 to 24 h. [3H]-Androstenone sulfate was incubated in buffer as a negative control. All incubations were run in triplicate.Incubations were terminated and steroids were extracted twice with 2 mL ether, which was dried under nitrogen and reconstituted in 50% acetonitrile (300 µL). The extracted steroid solution was then filtered with a 0.2 µm nylon syringe filter (Fisher Scientific, Toronto, ON, Canada) and analyzed by reverse-phase C18 HPLC using the 40 min HPLC profile previously described. Steroid conversion was calculated from the percentage of free androstenone that was produced from androstenone sulfate, which was quantified from the respective peak area detected for each steroid. This was used to calculate the total conversion (pmol) and the conversion rate (pmol/h) of androstenone sulfate.2.7. Statistical AnalysisStatistical analysis was conducted using SAS 9.4 (SAS Institute, Cary, NC, USA). Differences between animals in gene expression, fat androstenone concentrations, and the conversion of androstenone sulfate were evaluated using Student’s t-test with a significance level of p < 0.05. Additionally, Pearson correlation coefficients were determined for the following: (1) STS expression vs. OATP expression, (2) OATP expression vs. the rate of androstenone sulfate uptake into the fat, (3) STS expression vs. the percentage of androstenone sulfate converted to free androstenone, (4) STS expression vs. the concentration of androstenone in the fat. Correlations were calculated using the following model and considered statistically significant at p < 0.05:ρ=σxyσx2σy2 where σx2 is the variance of the x variable, σy2 is the variance of the y variable, and σxy is the covariance between x and y.3. Results3.1. Gene Expression DifferencesRNA was extracted from fat tissue, and STS and OATP expression was quantified by real-time PCR to evaluate differences in gene expression between animals. All adipose tissue samples (n = 16) expressed STS, OATP-B, OATP-E, and OATP-D, while OATP-A was not detected. A moderate positive correlation (r = 0.63, p = 0.01) existed between the expression of sulfatase and OATP-D (Figure 1A), which was the most abundant membrane transporter expressed in the fat. Conversely, the expression of STS was not well correlated with that of OATP-E or OATP-B.To evaluate the relationship between sulfatase expression and the expression of other genes in the fat, we arbitrarily classified boars as high (n = 8, STS expression ≥ 0.70) or low (n = 8, STS expression ≤ 0.67) sulfatase animals. The expression of STS and OATPs in high- and low-sulfatase boars is shown in Figure 1B. Adipose tissue from high-sulfatase boars expressed significantly greater quantities of STS (p = 0.004), OATP-D (p < 0.001), and OATP-E (p = 0.03) mRNA than adipose tissue from low sulfatase boars, while the expression of OATP-B was approximately equal.Hormonal status was assessed by quantifying E1S concentrations in the plasma at 120, 130, and 140 live weights, and animals were classified as early or late maturing on the basis of the cutoff level for high E1S production that was previously described by Zamaratskia et al. [24]. Boars with plasma E1S concentrations greater than 15.7 ng/mL at 130 kg live weight or less were considered to have a hormonal status consistent with early maturation (n = 9), and animals with plasma E1S concentrations of 15.7 ng/mL or less at 130 kg live weight were classified as late maturing boars (n = 7). STS and OATP expression were compared between early and late maturing animals to assess the relationship between hormonal status and gene expression in the fat. There was no significant difference (p > 0.05) in the expression of STS or any of the OATPs between early and late maturing boars. However, a strong positive correlation (r = 0.96, p < 0.001) was observed between the expression of STS and OATP-D in late maturing boars (Figure 1C) and was not well correlated in early maturing boars.3.2. Time Course Analysis of Steroid Uptake and ConversionRadiolabeled [3H]-androstenone sulfate was incubated with adipocytes (Figure 2A) or supernatant from homogenized fat tissue (Figure 2B) for 3, 6, 18, and 24 h to assess the time course for the uptake and conversion of androstenone sulfate by the adipose tissue, respectively. The uptake of androstenone sulfate increased over time and was greatest after 24 h (195.5 ± 40.4 pmol). Similarly, free androstenone production, resulting from the conversion of androstenone sulfate, increased significantly (p = 0.04) from 3 (71.7 ± 14.9 pmol) to 24 (205.2 ± 5.3 pmol) hours. The rate of both uptake and conversion was not linear but was greatest after 3 h (21.6 ± 6.6 pmol/h and 23.9 ± 14.9 pmol/h, respectively) and decreased over time, reaching 8.1 ± 1.7 pmol/h and 8.5 ± 5.3 pmol/h, respectively, after 24 h. Incubation times of 6 and 4 h were determined optimal for quantifying the uptake of androstenone sulfate and conversion of androstenone sulfate to free androstenone, respectively.3.3. Uptake of Androstenone Sulfate by AdipocytesAdipocytes isolated from fresh adipose tissue samples (n = 6) were incubated with [3H]-androstenone sulfate to characterize steroid transport, which was quantified by HPLC. A typical chromatogram depicting the uptake and subsequent conversion of androstenone sulfate to free androstenone is shown in Figure 3. The average uptake of androstenone sulfate by adipocytes over 6 h was 143.0 ± 12.0 pmol, or 23.1 ± 2.0 pmol/h. Additionally, 53.4 ± 10.1% of the androstenone sulfate transported into adipocytes was converted to free androstenone (76.4 ± 16.7 pmol). A strong positive correlation (r = 0.86, p = 0.03) was observed between the uptake of androstenone sulfate by adipocytes and the expression of the membrane transporter OATP-B (Figure 4), while the expression of OATP-D and OATP-E were not well correlated with androstenone sulfate uptake.3.4. Conversion of Androstenone Sulfate to Free AndrostenoneFat tissue samples (n = 16) were homogenized in buffer, and the resulting supernatant was incubated with [3H]-androstenone sulfate and analyzed by HPLC to quantify the production of free androstenone from androstenone sulfate in the adipose tissue. The average expression of STS was 1.13 ± 0.26, and over 4 h incubations, the average production of free androstenone from androstenone sulfate was 123.7 ± 16.4 pmol, or 30.9 ± 4.1 pmol/h. Fat androstenone concentrations ranged from 0.96 to 8.38 µg/g with an average concentration of 3.77 ± 0.63 µg/g, and a strong positive correlation (r = 0.76, p < 0.001) was observed between the expression of STS and the production of free androstenone from androstenone sulfate (Figure 5A).The relationship between fat androstenone concentration and androstenone sulfate conversion was examined in high and low sulfatase boars as well as early and late maturing animals to determine the effect of sulfatase expression and hormonal status on the deconjugation of androstenone sulfate in the fat. The average quantity of free androstenone produced from androstenone sulfate in boars with high sulfatase expression (n = 8) was 169.6 ± 18.6 pmol, which was significantly greater (p = 0.0015) than the steroid conversion quantified in low sulfatase boars (77.8 ± 14.3 pmol, n = 8). Additionally, fat androstenone concentrations were not significantly different (p > 0.05) between high (4.48 ± 1.03 µg/g) and low (3.06 ± 0.69 µg/g) sulfatase boars.The production of free androstenone from androstenone sulfate was positively correlated (r = 0.85, p < 0.001) with fat androstenone concentrations in early maturing (n = 9) boars and was not well correlated in late (n = 7) maturing boars (Figure 5B). Additionally, fat androstenone concentrations in early maturing boars tended to increase with the expression of STS (r = 0.67, p = 0.05, Figure 5C); however, there were no significant differences (p > 0.05) in the conversion of androstenone sulfate or fat androstenone concentrations between early and late maturing boars.4. DiscussionThe accumulation of androstenone in the adipose tissue causes a meat quality issue in heated pork products from entire males, which is known as boar taint. In humans, the adipose tissue functions in an intracrine manner by supporting the uptake and conversion of DHEAS to free dehydroepiandrosterone (DHEA), which serves as a precursor for bioactive androgens and estrogens [7]. Therefore, in the present study, we investigated androstenone production from the conversion of androstenone sulfate in adipose tissue from boars with high and low sulfatase expression to examine the relationship between the deconjugation of androstenone sulfate and boar taint development.OATPs facilitate the cellular uptake of substrates such as xenobiotics, bile acids, and steroid sulfates [25]. Following uptake, steroid sulfates such as E1S and DHEAS are deconjugated by STS to return estrone (E1) and DHEA, respectively [6]. Using real-time PCR, we established that porcine adipose tissue expressed STS and all membrane transporters except OATP-A, which is in concordance with the expression reported in human adipose tissue [7]. We identified a positive correlation between the expression of STS and OATP-D and determined that higher sulfatase expression was associated with greater expression of the membrane transporters OATP-D and OATP-E, but not OATP-B. In humans, E1S is a substrate of all three OATPs, while the cellular uptake of DHEAS is mediated by OATP-B and not OATP-D or OATP-E [25]. Boars express high levels of testicular sulfotransferases and often produce large quantities of E1S [26]. Plasma concentrations of E1S increase as boars reach sexual maturity [27,28], and concentrations of E1 in the fat of sexually mature boars are positively correlated with fat androstenone concentrations [24]. The positive correlation between the expression of STS and OATP-D in late maturing boars suggests that the uptake and deconjugation of E1S to produce bioactive E1 by the adipose tissue may be necessary to promote the onset of sexual maturity in late but not early maturing boars. Therefore, future research should investigate the uptake and deconjugation of E1S by the porcine adipose tissue to further characterize the effect of E1S production on sexual maturation and consequently the development of boar taint.We identified a positive correlation between the uptake of androstenone sulfate by adipocytes and the expression of OATP-B, as well as the expression of STS and the conversion of androstenone sulfate to free androstenone. Additionally, boars that expressed higher levels of sulfatase converted significantly greater quantities of androstenone sulfate to free androstenone; however, the expression of OATP-B in boars with high and low sulfatase expression was approximately equal. These results suggest that the transport and deconjugation of androstenone sulfate is mediated by OATP-B and STS, respectively, and the quantity of androstenone sulfate that is converted to free androstenone in the adipose tissue depends on the expression of STS rather than OATP-B.The STS-mediated hydrolysis of androstenone sulfate did not result in the production of a C3 hydroxysteroid, but rather returned the parent compound (androstenone), which is a C3 keto steroid. Consistent with our results, it has been previously reported that chemical hydrolysis of androstenone sulfate returns free androstenone and not a hydroxylated metabolite [9]. On the basis of the results of the present study, we predict that the STS-mediated hydrolysis of androstenone sulfate results in the production of the same 3-enol intermediate that is suspected to facilitate sulfoconjugation, with stabilization resulting in the movement of a double bond between C3 and C4 to the 3-keto position to produce free androstenone (Figure 6). This is consistent with the idea that androstenone sulfate functions as a steroid reservoir in the boar, which was originally proposed by Laderoute et al. [9]. However, future research is required to confirm the pathway mediating the deconjugation of androstenone sulfate. Furthermore, additional research should further investigate the relationship between OATP and STS expression and the uptake and deconjugation of androstenone sulfate in other breeds and with a larger number of animals.Previous research has identified a positive correlation between testicular SULT2A1 expression and plasma concentrations of androstenone sulfate as well as a negative correlation between testicular SULT2A1 activity and fat androstenone concentrations [29]. On the basis of these results, researchers suggested that the sulfoconjugation of androstenone reduces boar taint development by decreasing the quantity of free androstenone available to accumulate in the fat [29]. However, the present study demonstrated that fat androstenone concentrations tended to increase with the expression of STS and were positively correlated with the production of free androstenone from androstenone sulfate in early but not late maturing boars. The development of boar taint depends on numerous factors that influence the rate of androstenone synthesis and metabolism and vary significantly between different breeds as well as individual boars within the same breed [30]. Our results suggest that the production of free androstenone from androstenone sulfate in the adipose tissue may have a significant impact on the development of boar taint in early maturing boars. Therefore, future research should further investigate the relationship between STS and fat androstenone concentrations in early maturing boars using larger sample sizes to determine if STS is a suitable candidate gene for boar taint.5. ConclusionsThis study demonstrated that the uptake and deconjugation of androstenone sulfate in the adipose tissue of the boar is facilitated by the membrane transporter OATP-B and STS, respectively. Additionally, we have shown that boars expressing higher levels of sulfatase have a greater expression of several OATPs and convert larger quantities of androstenone sulfate to free androstenone than boars with low sulfatase expression, which suggests that the uptake and deconjugation of steroid sulfates varies significantly between individual animals. Fat androstenone concentrations tended to increase with the expression of STS and were positively correlated with the production of free androstenone from androstenone sulfate in early maturing animals. These results suggest that the STS-mediated deconjugation of androstenone sulfate may be a significant cause of boar taint development in early maturing animals. Therefore, future research should further investigate this relationship to determine if STS is a suitable candidate gene for boar taint.

animals : an open access journal from mdpi

10.3390/ani12010050

PMC8749766

Greenhouse gas emission has attracted considerable public attention in recent years, driving the search for genetic, nutritional, and management strategies to reduce methane emissions and increase the sustainability of milk production. Rumination activity has an important function in feed particle size reduction, condition of feeding behavior, and feed intake as well as in stabilizing rumen fluid pH through saliva production. A total of 365 high-yielding Polish Holstein -Friesian multiparous dairy cows were included in the study covering 24 to 304 days of lactation. Next, the data from the cows were assigned to three groups based on daily rumination time: low rumination up to 412 min/day (up to 25th rumination percentile), medium rumination from 412 to 527 min/day (between the 25th and 75th percentile), and high rumination above 527 min/day (from the 75th percentile). We showed that a longer rumination time leads to a lower methane emission level. Therefore, strategies that increase chewing activity may be used to reduce the environmental impact of dairy cows production.

The objective of this study was to determine the effect of the rumination time on milk yield and composition as well as methane emission during lactation in high-yielding dairy cows fed a partial mixed ration based on maize silage without pasture access. A total of 365 high-yielding Polish Holstein-Friesian multiparous dairy cows were included in the study covering 24 to 304 days of lactation. Methane emission, rumination time, and milk production traits were observed for the period of 12 months. Next, the data from the cows were assigned to three groups based on daily rumination time: low rumination up to 412 min/day (up to 25th rumination percentile), medium rumination from 412 to 527 min/day (between the 25th and 75th percentile), and high rumination above 527 min/day (from the 75th percentile). Rumination time had no effect on milk yield, energy-corrected milk yield, or fat and protein-corrected milk yield. High rumination time had an effect on lower fat concentration in milk compared with the medium and low rumination groups. The highest daily CH4 production was noted in low rumination cows, which emitted 1.8% more CH4 than medium rumination cows and 4.2% more than high rumination cows. Rumination time affected daily methane production per kg of milk. Cows from the high rumination group produced 2.9% less CH4 per milk unit compared to medium rumination cows and 4.6% in comparison to low rumination cows. Similar observations were noted for daily CH4 production per ECM unit. In conclusion, a longer rumination time is connected with lower methane emission as well as lower methane production per milk unit in high-yielding dairy cows fed a maize silage-based partial mixed ration without pasture access.

1. IntroductionThe milk production of dairy cows has increased substantially over the last few years due to selection as well as substantially improved nutrition and herd management. High production requires substituting forage with a high starch content concentrate to meet the high nutrient requirement as well as maintain rumen homeostasis. As a consequence, the contribution of crude fiber and physically effective neutral detergent fiber to the diet of high-yielding dairy cows has decreased. In turn, this can affect the rumination behavior. Rumination is desirable, as it takes part in breaking down of the feed particles and stimulates saliva production. Saliva lysozyme through preventing the proliferation of Gram-positive bacteria plays an important function on the rumen microbiota and can also influence the selection of methanogenic microorganisms and affect the rumen ecosystem and modulate methane emissions. Saliva also contains bicarbonate and phosphate buffers and plays an important role in sustaining the rumen fluid pH and cellulolytic microbial activity [1]. Thus, the optimal rumination activity is necessary to decrease the risk of rumen subacute and acute acidosis [2,3] as well as maintain good health status and lower incidences of clinical and subclinical disorders [4,5,6,7]. Rumination impacts the whole digestion process, including the feed passage rate as well as voluntary feed intake in dairy cows [8], while it may impact the cow’s milk performance [9]. Watt et al. [10] showed that a longer rumination time improves feed intake, milk production, and total methane emission, while it also reduces methane emission per milk unit during the 22-day experimental period in grazing dairy cows.Greenhouse gas emission by dairy farms has become the focus of public attention in recent years. The search for nutritional and management methods to reduce methane emission is necessary for sustainable milk production [11,12]. The rumen environment may affect methane synthesis by the rumen methanogens [13]. An increase in acetate and butyrate contents in the rumen fluid can affect the concentration of dissolved hydrogen utilized in methane synthesis [14]. The rise of acetate fermentation is related to the availability of crude fiber and creates a homeostatic environment related to fiber degradation bacteria [1]. As described above, rumination time due to its role in stabilizing pH of rumen fluid is related to the health status of cows and also can indirectly affect the rise of methane emission. In the available literature, the relationships between both phenotypes—rumination time and methane emission—has been described mainly in grazing dairy cows [10]. Despite other studies, which mainly focused on the description of genetics correlations between rumination time and methane emission, there is a lack of a study covering high-yielding dairy cows fed a diet based on maize silage during the whole lactation period. Additionally, results of the published experiments covered only a small part of lactation [10,15] or were conducted on other than Polish Holstein-Friesian breed [16] or aimed to compare different methods of methane measurement [17], whereas the present study analyzed records from 24 to 304 days of lactation on 365 animals to provide a better overview of interactions between rumination activity, performance, and methane production.We hypothesized that a longer rumination time is connected with lower methane emission per milk unit in high yielding dairy cows fed without pasture access.The objective of this study was to determine the effect of the rumination time, milk yield, and composition along with methane emission during lactation in high-yielding dairy cows fed a maize silage-based partial mixed ration.2. Materials and Methods2.1. Animal Management, Experimental Design, and DietAll animal procedures were conducted in accordance with the guidelines of the Polish Council for Animal Care and the Local Ethics Commission of the Poznań University of Life Sciences (Poznań, Poland) with respect to animal experimentation and care of the animals under the study.A total of 365 high-yielding (11,264 kg/305 days lactation) Polish Holstein-Friesian multiparous dairy cows were included in the study covering 24 to 304 days of lactation. In total, 14,274 daily complete (cow and all milk production traits) observations were collected. Data were collected in a production environment. Data from cows were assigned to three groups based on individual cow average daily rumination time (Figure 1): low rumination up to 412 min/day (L, up to the 25th rumination percentile), medium rumination from 412 to 527 min/day (M, between the 25th and 75th percentile), and high rumination above 527 min/day (H, from the 75th percentile).The cows were milked in an automatic milking system (AMS, Astronaut, Lely Industries, NV, Maassluis, The Netherlands).The cows received ad libitum a partial mixed ration (PMR), which was served twice a day and met requirements for 25 kg of milk yield. The animals had free and equal access to the feeding table. The cows were divided into two groups due to the management routine and not based on their characteristics. Each group had one common feeding table whose size was dependent on the number of the animals in the technological group.The nutritional values of the feed components were calculated on the basis of the analyzed nutrient contents using NIRS (InfraXact, Foss, Hilleroed, Denmark) and the MAXTM System for Dairy software (3.19, Cargill, Minneapolis, MN, USA). The diets were balanced according to the NRC (2001) system recommendations and the MAXTM System for Dairy software (3.19, Cargill, Minneapolis, MN, USA).PMR and concentrates ingredients and nutritional value are shown in Table 1.Two concentrates (C standard and C extra) were added according to the requirements of individual cows from 0.5 to 8 kg into AMS during each milking. The proportion of C standard and C extra dispensed in AMS depended on individual milk yields and ranged from 75:25 to 70:30.The silages were analyzed and verified two times per month using the NIRS method.Weekly forage, concentrates, and PMR representative samples were collected, frozen, and stored (−20 °C) for further pooled monthly analyses using wet chemistry methods. On the basis of crude protein (CP, method 976.05; AOAC International, 2005), neutral detergent fiber (NDF, PN-EN ISO 16472:2007), and acid detergent fiber (ADF, PN-EN ISO 13906:2009), feeds as well as the PMR were verified. The PMR values were recalculated monthly or before a new forage from a new silo was used.The particle size distribution of PMR samples was determined weekly by the Penn State Particle Separator system with 3 sieves (19 mm, 8 mm, 1.18 mm) and a bottom pan [18]. The mean retention of particles were: 6% (>19 mm), 48% (8–19 mm), 40.5% (1.8–8 mm), and 5.5% (<1.18 mm).2.2. Rumination Time, Milk Performance, Body WeightRumination time was measured using electronic rumination loggers placed on the neck collars (SCR Engineers Ltd., Netanya, Israel). Loggers recorded rumination data in 2 h intervals (i.e., 12 values per day), and rumination time was expressed in minutes of rumination time recorded within each time interval. The data for rumination with accuracy (rumination mark) were read from the loggers by the readers placed in the barn connected with the Lely T4C. The daily rumination time of cow was calculated by adding 12 measurements from the day. Measurements with low accuracy (rumination mark below 98) were rejected, and all the rumination time observations of a particular cow recorded at that day were removed from the dataset (i.e., 12% of daily rumination time was set to missing).Daily milk production, fat, and protein content were obtained from the farm management system (Lely T4C) and then used for calculating energy-corrected milk (ECM) and fat protein-corrected milk (FPCM). The ECM was calculated according to Reist et al. [19] as [(0.038 × g crude fat + 0.024 × g crude protein + 0.017 × g lactose)] × kg milk/3.14. The FPCM was calculated as [(0.337 + 0.116 × milk fat % + 0.06 × milk protein %) × kg of milk] [20].Body weight was collected in automatic scales, and therefore, some additional filtering of the data was required. For that, data from each cow were checked separately. First, the median body weight (BW) for a cow was calculated. Second, BW values lower than 100 kg than the cow’s median BW were set to missing, as such a difference was assumed to be an erroneous record. This was confirmed by the visual inspection of the data (now shown). Third, the missing BW records were substituted by the cow’s median BW.The AMS identified each animal during milking and saved daily information concerning body weight and milk performance.2.3. Methane MeasurementsThe CH4 concentration (ppm) was measured in the air exhaled by the cows during milking in AMS using an Fourier transform infrared spectroscopy FTIR analyzer (GASMET 4030; Gasmet Technologies Oy, Helsinki, Finland) installed in the feeding bin. The samples were taken continuously, and the gas samples were analyzed every 5 s. The investigated phenotypes were daily averages obtained as described in Pszczola et al. [21]. First, the concentrations from the whole milking were averaged. Secondly, the measurements from all milkings were corrected for the diurnal variation in CH4. Subsequently, the corrected measurements for each cow were averaged within the day. Then, methane production was calculated in L/day following Madsen et al. [22] based on the ratios between CH4 and CO2 concentrations measured during each milking and estimated heat production.The following average daily phenotypes were defined and analyzed: methane production (CH4) (L), the CH4 production in relation to metabolic weight (CH4/BW0.75) (L/kg), milk production (CH4/milk) (L/kg), energy-corrected milk (CH4/ECM) (L/kg), and per concentrate intake (CH4/concentrate intake) (L/g).2.4. Statistical AnalysisRumination time was divided into three groups according to the quartile distribution. Cows below the first quartile of rumination time were assigned to the low-L group, cows between the first and third quartile were assigned to the medium-M group, and cows above the third quartile of rumination time were placed in the high-H group.Differences between rumination groups were assessed for each of the analyzed traits separately.To check whether rumination time has an impact on the analyzed traits, the following model was employed:yijkl=GROUPj+LACk×∑n=15βnDIMln+cowi+errorijkl, where yijkl was one of the following traits (i.e., daily rumination time, body weight, metabolic body weight, concentrate intake, concentrate intake per kg of milk, concentrate intake per metabolic weight, daily milk yield, energy-corrected milk yield, fat protein-corrected milk yield, fat, protein and lactose concentration, fat to protein ratio, daily methane production, daily methane production per metabolic weight, daily methane production per milk production, and daily methane production per concentrate intake) observed on the ith cow assigned to the jth group of rumination level (GROUP). The overall lactation curve was modeled with fourth-order Legendre polynomials separately for first, second, and further lactations. The GROUP had three levels: High, Medium, and Low. The terms cow and error were random terms.The analyses were performed in R software [23]. The model effects were estimated using lme4 package [24], the significance of the difference between estimated marginal means was assessed using lmerTest [25] and emmeans packages [26] using Satterthwaite’s method [27] for approximating degrees of freedom enabling testing for the significance of differences between fixed effects levels. The p-values obtained for the differences between the estimated marginal means for rumination groups were adjusted using Tukey’s method for comparing 3 estimates.3. ResultsDifferences in rumination time were observed between all the groups (H, M, and L) (Figure 2). The average daily rumination time was 195 min longer for cows in the H group in comparison to the L group and 84 min greater compared to cows, which belonged to the medium rumination time group (M) (Table 2). Mean body weight differed significantly between all the groups (H: 543 kg, M: 546 kg, L: 551 kg). The intake of concentrate from AMS was higher in low rumination cows compared to the other groups. High rumination cows were characterized by the lowest concentrate intake per their metabolic body weight (BW0.75, 39.58 g/kg) and differed both from medium rumination (40.34 g/kg) and low rumination cows (40.32 g/kg). Rumination time had no effect on concentrate intake on milk yield.Rumination time had no effect on milk, energy-corrected milk, as well as fat and protein-corrected milk yield (Table 3). High rumination cows had an effect on lower fat concentration in milk (3.75%) compared with the M and L groups (3.94% and 3.80%, respectively). Differences between rumination time groups on protein and lactose concentrations in milk were not confirmed. The fat and protein ratio was lower in high rumination cows (1.14) compared to the low (1.15) and medium (1.15) rumination cows. Rumination time had no effect on the number of milkings in AMS, which were on average 2.85 per day.The significant effect of rumination time on methane (CH4) emission was observed in all the groups. The highest daily CH4 production was noted in low rumination cows (412.47 L), which emitted 1.8% more CH4 than medium rumination cows (404.99 L) and 4.2% more than low rumination cows (395.80 L). The cows from the high rumination group had a lower daily CH4 production per BW0.75 (3.59 L/kg) compared to both groups, medium rumination cows (3.67 L/kg) and low rumination cows (3.68 L/kg).Rumination time had a positive effect on daily methane production per kg of milk. Cows from the high rumination group produced less daily CH4 per kg of milk (11.52 L/kg) compared to medium (11.86 L/kg) and low (12.07 L/kg) rumination cows. Similar observations were noted for daily CH4 production per ECM unit (11.79 L/kg, 12.07 L/kg, 12.26 L/kg). Daily lower methane production per concentrate intake unit was highest in medium rumination cows (0.12 L/g) compared to the L group (0.10 L/g).Daily methane yield (kg) was higher at the beginning of the lactation and decreased toward the end of the milking period (Figure 3), whereas the methane production per kg of milk was low at the beginning of the lactation and increased toward the end of the lactation (Figure 4).4. DiscussionWe hypothesized that a longer rumination time would be connected with lower daily methane production per milk unit in high-yielding dairy cows fed a partial mixed ration based on maize silage without pasture access. Cows from all the groups (H, M, and L) ruminated approximately 458 min per day, which is in the range reported in the literature by White et al. [28], who analyzed 179 cows with a mean rumination time of 436 min/day, ranging from 236 to 610 min/day, as well as Zetouni et al. [16], who recorded 443 min/day as average rumination time during Danish Holstein cows lactation. Cows from the high rumination group ruminated 551 min/day, which was an 84 min increase compared to the medium (467 min/day) and 195 min more compared to the low ruminating cows (356 min/day). Similar differences of rumination time in grazing cows were reported by Watt et al. [10].Rumination time had no effect on milk, energy-corrected milk as well as fat and protein-corrected milk production. Despite a positive relationship between rumination time and milk production in early lactation [29] and mid-lactation [4], which has been reported earlier, Stone et al. [30] noted a weak correlation between both phenotypes (r = 0.30). The positive relationship between rumination time and milk production may be indirectly related to dry matter intake. Nevertheless, dry matter intake may indirectly cause a positive relationship between rumination time and milk yield, and the association between dry matter intake and rumination time can also depend on diet composition [3].Moreover, Stone et al. [30] explained that the reason for the different results shown by various authors was due to differences in methods of rumination activity detection. In an early study, rumination was estimated based on direct visual observations, and results can be different when measured by an automated rumination logging system [30].Watt et al. [10] observed a positive association with rumination time and greater milk production, concentrate intake from AMS, as well as estimated dry matter intake by grazing cows. It is commonly known that the main factors of rumination time are connected with the chemical and physical characteristics of the diets, but according to Beauchemin et al. [3], who described a positive relationship between rumination time and dry mater intake in dairy cows, on this basis, we can assume that high rumination cows were also fed a more PMR-based diet. Schirmann et al. [31] showed that cows that ruminated more time per day spent less time feeding (r = −0.34), and rumination times did not relate to dry matter intake (r = 0.11). In the present study, differences in concentrate intake across the groups were not detected. Rumination time had a slight effect on milk composition; the cows that ruminated longer (H) had only less fat concentration without differences in protein and lactose concentrations in milk. Similarly, a negative correlation between rumination time and milk fat concentration during the first month of lactation in cows older than the third lactation was noted by Kaufman et al. [9]. It would appear that an increase in rumination time should be directly connected with better rumen homeostasis and fiber microbial degradation and an increase in fat percentage [32]. Less milk fat concentration in high ruminating cows (H) may be connected with their higher milk yield, while it may also be a consequence of the enhanced availability of glucose for the synthesis of lactose in milk without any increase in volatile fatty acids or long-chain fatty acids for butterfat synthesis. Rumination time had no effect on protein concentration in milk, which is in agreement with the observations reported by Kaufman et al. [9], who found no association between milk protein and rumination time in dairy cows during the first month of lactation. Different results, i.e., a negative relationship between rumination time and milk production, protein content in milk, but a positively association with milk fat concentration in a study of mid-lactation Holstein and Swedish Red cows were reported by Byskov et al. [33].Rumination time had a positive effect on a decrease in methane production; cows assigned to the high ruminating group produced less methane than medium and lower ruminating groups, and medium ruminating cows produced less methane than cows with a lower daily rumination time. A similar result, negative genetic correlation between methane and rumination time was estimated by López-Paredes et al. [15], who collected methane data from 14 to 21-day periods. This results are different from those of Zatouni et al. [16], who observed a lack of relationship between rumination time and methane emission by high-yielding dairy cows. Phenotypes, methane emission, and rumination activity are affected by many factors that are hard to be accounted for, and therefore, the results from other studies can differ. Additionally, it is known that increasing NDF from forages in the dairy cows diets stimulates rumination activity, increases saliva production, and via buffering rumen fluid increases the production of acetate in the rumen, leading to higher methane production [33]. On the other hand, decreasing NDF from forage and an increase in concentrates intake may be associated with a decreased rumen pH, leading to an increase in the levels of propionate and a decrease in acetate and butyrate levels while decreasing hydrogen equivalents that would be converted to methane and are inhibitors in methanogenesis. Different results from current study, higher methane emissions from high ruminating grazing cows were shown by Watt et al. [10]. An explanation of these differences may be attributed to the different body weights of high and low ruminating cows in both experiments. In a study described by Watt et al. [10], high ruminating grazing cows were heavier than low ruminating grazing cows in contrast to the present study, where high ruminating cows had lower body weight. Additionally, the high ruminating cows had lower methane emissions per metabolic body weight than cows that spent less time on rumination. Lower methane production in high ruminating cows per body weight may be connected with lower body weight as well as lower methane production by cows from this group.In the present study, high ruminating cows had a lower daily methane production per milk unit as well as energy-corrected milk than other cows, which spent less time on rumination. A reduction of methane emission per milk production in high ruminating cows with similar milk yield between the three groups is connected only with the lowest methane emission. A reduction of methane production per unit of product was also observed in high ruminating grazing dairy cows [10]. Knapp et al. [14] described that diets containing more energy or with better digestibility increase net energy intake, and when this energy is partitioned into milk production, a decrease in methane emission per ECM yield can be observed. In addition, Aguerre et al. [34] observed a decrease in methane per ECM production when non-fiber carbohydrates were elevated through an increase in concentrate intake from 32 to 53% in the diet.We collected 14,274 records of daily methane emissions recorded throughout lactation from 24 to 304 days to obtain high reliability of the daily methane production estimates. Including individual dry matter intake levels would provide additional insights; however, they was not possible to collect due to the very large number of collected observations and technical difficulties. Methane emission measurements are highly variable between animals and within the lactation period. Thus, studies on methane emission should be conducted on a large number of animals and cover a long time period and the association of rumination time that best indicates the physiological state of ruminal fermentation at optimal levels to ensure animal welfare and health.5. ConclusionsIn conclusion, the results confirmed the hypothesis that a longer rumination time is related to lower methane emission per milk unit in high-yielding dairy cows fed a partial mixed ration based on maize silage without pasture access.

animals : an open access journal from mdpi

10.3390/ani11113043

PMC8614447

Feeding restriction (FR) is essential to reduce excessive fat deposits caused by overfeeding in hens and to ensure their reasonable nutritional requirements for egg production. Effective FR is particularly crucial for raising hens in the late phase of laying; because hens require lower energy at this stage, overfeeding reduces their feed efficiency and increase feed costs. The gut microbiota is involved in various metabolic pathways of laying hens, including in late-phase age. Thus, changes in feeding interventions can alter the presence of gut microorganisms and the structure of the microbial community, resulting in altered metabolic regulation. In this study, we investigate the microbiota and metabolome responses of late-phase laying hens under FR. Our results provide data to access the profile of the cecal bacteria community, their relevance to cecal and serum metabolites, and their FR biosynthetic pathways related to host nutritional requirements and intestinal nutrient availability. Moreover, understanding the principles of host-microbial interaction is essential for developing cost-effective strategies to improve laying hens’ production.

This study investigated cecal bacterial community profile, cecal and serum metabolites, and its biosynthesis pathway in late-phase laying hens during 6 weeks feeding restriction (FR), using 16S rDNA as gene sequencing and non-targeted LC-MS/MS as metabolomics approach. We used three groups (ad libitum, FR20, and FR40). FR can reduce excessive fat in late-phase laying hens, while egg production rate is not affected, except for the FR40 group. In phylum level, FR20 had more population of Bacteriodetes and Firmicutes amongst groups. The same result is at genus level, FR20 were higher of the predominant genus (Bacteroides and Rikenellaceae_RC9_gut_group). Both of FR20 and FR40 reduced Proteobacteria as potential pathogenic bacteria. Non-targeted metabolomic analysis revealed that FR20 modified 20 metabolites in cecal and 10 metabolites in serum of laying hens, whereas 48 cecal metabolites and 31 serum metabolites has revealed in FR40. KEGG assay showed FR20 and FR40 upregulated lipid, carbohydrate, amino acid, nucleic acid pathway, and FR40 modified steroid metabolism in cecal analysis. In serum, only FR40 modified lipid, amino acid pathway, and carbohydrate biosynthesis were shown. This study showed that FR during late-phase laying hens altered the microbiome composition, modified metabolites profile and biosynthesis of the cecal as well as serum.

1. IntroductionThe intestinal microbiota, the extremely large number of different microorganisms that inhabit the gut, is exclusively responsible for intestinal morphology, nutrient absorption, immunity, and host health, including growth performance [1,2]. This microbiota is also involved in important metabolic functions and numerous host pathways, such as the biosynthesis of lipid and amino acid [1,3]. Therefore, nutritional intervention can induce the presence of intestinal microorganisms [4,5] and the structure of the microbial community [5], which, in turn, affects host immunity and metabolic regulation [6].The means of nutritional intervention for chicken is called feeding restriction (FR) [7]. In the laying hens industry, FR is an important feeding method to increase egg production of hens [4]; for example, to artificially control the feed intake [7,8], the protein and/or energy levels that are applied to chickens are limited to reduce feed costs in the hen’s diet [9,10]. The purpose is to ensure that the hens will not accumulate too much fat to affect their production performance [11,12,13,14]. Specifically, FR imposed in the rearing stage of the hens can properly control their body weight, avoid premature maturity, and reduce mortality [9,12]; additionally, the number of heavy follicles in the hens will be decreased at onset of the egg production [15]. Before the peak of egg production, FR is usually beneficial to hens for egg production; it has a longer-lasting effect on egg size and egg production capacity and lower mortality in the hens, but this effect may depending on the species of hens and the level of FR used [13]. Although there are few associations between FR and egg production after peak production [16,17], the FR carry out on the hens at this stage has always been the subject of much commercial interest, because hens that reach body maturity at late stage of laying cycle require less dietary energy [18,19]; therefore, applying FR to hens can improve the feed efficiency and save feed costs [16,20].Gut microbiota has a role in a variety of metabolic processes in laying hens, including late-phase age. As a result, interventions in dietary treatments can modify the presence of gut microbes and the structure of the microbial community, resulting in changes in metabolic control. FR in laying hens can not only alter the population of intestinal microbes [4,21] and its microbial gene expression [5] but also affect host health [4,22] and metabolite regulation. In addition, due to host-related physiological adaptations, chickens with restricted feeding often exhibit different gut characteristics (e.g., lower ileal and cecal short chain fatty acid (SCFA) profiles, bigger duodenum, and enlarged pancreas size), possibly concerning enhancing the utilization of feed to obtain the greater energy and nutrient requirements resulting from FR [4]. This occurs as a form of compensatory dietary restriction, thereby reducing the availability of intestinal nutrients. Moreover, FR is associated with differences in gut bacteria [4,21]. These changes affect the small intestinal tract, possibly giving the microbiota more time to utilize non-digested feed [4]. Accordingly, bacteria that have functional abilities to degrade nondigestible carbohydrates [21] were enriched in the ileum in chicken with FR. Another underlying mechanism might have been a greater mucus secretion to facilitate the ingesta flow in chicken with FR [4].Recently, metabolomics has become an emerged technique that focuses to identify the functional correlation between the host and the intestinal commensal microbiota. Most studies on the intestinal microbiome aim to understand disease-related metabolites or their dysregulated metabolic pathways [2,23]. This approach is effective for assessing the effect of nutritional interventions, especially when traditional hypothetical methods cannot detect metabolic changes, because they solely focus on nutrient content to maintain a population over improving health and performance [24].The commensal bacterial community in the intestine is important for chicken metabolism. These bacteria not only interact amongst their community, but also interact with the host tissue [1]. This interaction is fundamental for poultry production and health, because these bacteria can protect the intestine from pathogens [1,25]. Layer production starts to fall after 31 weeks of age [26], and due to aging, the egg productivity and immunity of laying hens decline sharply, and hence can affect the metabolism and hormonal status of these hens [27]. Due to the FR in the late laying period, the changes in hen performance, egg production, and cecal microbial community and its metabolite-microbial community, as well as metabolite in serum, have not been discovered before. Herein, we revealed the cecal bacterial community profiles, cecal and serum metabolites profile, and the correlation of cecal microbiome with metabolites (cecal and serum) and its biosynthesis pathway in late-phase laying hens.2. Materials and Methods2.1. Ethical StatementAll research was approved by the Tunghai University Institutional Animal Care and Use Committee (IACUC Approval No. 106-15) prior to the start of data collection.2.2. AnimalsAll experiments were performed in accordance with approved guidelines. The animal protocol was approved by the Tunghai University Institutional Animal Care and Use Committee (IACUC Approval No. 106-15) prior to the start of data collection. A total of 30 healthy, 48 week-old Lohmann laying hens from the same hatch of local commercial hatchery were weighed and randomly allotted to one of the three groups at Tunghai University experimental farm located in Taichung, Taiwan. The hens were reared in galvanized wire cages (25 × 40 × 30 cm, one hen per cage) with a nipple drinker and individual trough-feeder. The lighting schedule was 16 h light, 8 h dark throughout the experiment. Mean ambient temperature was 25 ± 5 °C; the relative humidity was maintained within the range of 60–70%. Feed and water were provided ad libitum over the entire experimental period. All hens were provided with the same diet, which was formulated according to the recommendations of National Research Council [28], as shown in Table 1.2.3. Experimental DesignAt 48 weeks of age, 30 laying hens were allocated randomly into three experimental groups; each group contained 10 repetitions of 1 hen in one replicate. The three groups were (1) ad libitum (AL) as control group, (2) 20% feed restriction group (FR20), and (3) 40% feed restriction group (FR40). The experiments were carried out over a total of a 6 week period. Before the beginning of data collection, the hens were adapted for two weeks. The AL group consisted of a supply of 100 g of the laying diet per bird a day (Table 1), and the other FR group consisted of a supply of 80 g (FR20) and 60 g (FR40) of laying diet per bird a day, compared to the AL group, respectively. The body weight and egg production in each treatment were recorded daily, and the egg weight was recorded every two days. The dead hens were replaced by spare birds maintained under identical treatment.2.4. Sample CollectionAt the end of the experiment, 5 chickens were randomly selected from each group, and cecal contents samples and blood samples were collected for further microbiome and metabolome analysis. Specifically, the hens were injected intravenously with sodium pentobarbital (30 mg/kg body weight), and cervical dislocation was performed. Approximately 2 g of cecal contents were collected, aliquot into two sterilized tubes, and stored at −80 °C. One of these cecal content samples was used for DNA extraction and pyrosequencing, and the other samples was used for metabolomics analysis. Blood was collected from the left brachial vein of the hens; these blood samples were centrifuged at 1500× g for 15 min at 4 °C to collect serum. The serum sample was stored at −80 °C for metabolomic analysis. Liver and abdominal fat were collected for weight determination.2.5. DNA Extraction and 16S rDNA Amplicon Pyrosequencing of Feces SampleThe total bacterial genomic DNA in each sample of intestinal contents sample was extracted using QIAamp Fast DNA Stool Mini Kit (QIAGEN, Hilden, Germany). The extracted DNA was then measured using a SimpliNano spectrophotometer (Biochrom, Cambridge, UK) and agarose gel electrophoresis. The paired-end 2 × 300 bp sequencing was performed using the Illlumina MiSeq platform with MiSeq Reagent Kit (Illumina, San Diego, CA, USA).2.6. Sequence AnalysisQuantitative insights into microbial ecology (QIIME, v1.8.0) pipeline was used to process the sequencing data, as described previously [29]. Briefly, we used FLASH v.1.2.11 for assembling the 300 bp paired-end raw reads derived from the 16S ribosomal amplicon sequencing and barcode identification for de-multiplexing. We discarded a Q score of less than the threshold (Q < 20) in the QIIME 1.9.1 pipeline (as a quality control). Before operational taxonomic unit (OTU) clustering at 97% sequence, identified using the UPARSE function in the USEARCH v.7 pipeline, effective tags were filtered and obtained by UCHIME to investigated chimera sequences. Based on the information retrieved from the Silva database v.132, we used an RDP classifier (v.2.2) algorithm to annotate taxonomy classification for each representative sequence. Alpha diversity, based on 6 criteria from QIIME pipeline (i.e., observed-OTU, Chao1, Shannon, Simpson, abundance-based coverage estimators (ACE), and good-coverage), indicate the complexity of each species within individual samples. Meanwhile, the number of different species represented in the microbial community is referred to as observed-OTU. The Chao1 and ACE indices was used for investigating community richness, and the relative abundance and evenness accounting for diversity were evaluated by the Shannon indices. For representing the number of the observed species, a random selection of a certain amount of sequencing data of each sample was used to construct a rarefaction curve. We then analyzed the differences among samples in terms of species complexity using beta diversity. A principal component analysis (PCA) preluded a cluster (genus) analysis whose function was to reduce the dimensions of the multiple variables using the FactoMineR package and ggplot2 package in R software (v.2.15.3). For enhancing the community distinction, the partial-least-squares discriminant analysis (PLS-DA) was used to analyze and visualize variance based on OTU level of gut microbiota composition among the communities. This PLS-DA can be evaluated using the R package mixOmics. For statistical analysis, a zero-inflated Gussian (ZIG) log-normal model, as implemented in the “fitFeatureModel” function of the Bioconductor metagenomeSeq package, was used to determine the significance of all species among groups at various taxonomic level, as previously described [29]. Furthermore, to determine whether the community structures significantly differ among and within groups, we then used analysis of similarities (ANOSIM) analysis. This analysis also provided microbial phenotype, and this database source could be revealed in the integrated microbial genomes (IMG) database, the Kyoto Encyclopedia of Genes and Genomes (KEGG) and the Pathosystems Resource Integration Center (PATRIC) [30,31]. BugBase was used to predict organism-level microbiome phenotypes; specifically, BugBase uses input from various databases, including IMG, KEGG, and PATRIC, to categorize six main phenotype categories: Gram staining, oxygen tolerance, ability to form biofilms, mobile element content, pathogenicity, and oxidative stress tolerance [32].2.7. Metabolomic Extraction of Feces and Serum SampleWe used 100% methanol extraction method (methanol, with 1 μg/mL internal standard) to obtain metabolites from the cecal contents and blood serum. Briefly, a hundred microliter of serum or sample of cecal contents was mixed with 400 μL of extraction solution in an eppendorf tube. The samples were sonicated in an ice-water bath for 10 min and then incubated at −20 °C for 1 h to precipitate protein. These extracted samples were then centrifuged at 15,000× g for 25 min at 4 °C. Afterwards, each supernatant was transferred to a fresh glass vial for LC–MS/MS analyses. A quality control (QC) sample was then prepared by mixing aliquots of supernatants from all samples.2.8. Metabolomic AnalysisThe cecal contents of LC–MS/MS analyses were determined using an UHPLC system (1290, Agilent Technologies) with a UPLC HSS T3 column (2.1 mm × 100 mm, 1.8 μm) coupled with a Q Exactive mass spectrometer (Orbitrap MS, Thermo). The mobile phase A contained 0.1% formic acid in water (as positive mode) and 5 mmol/L ammonium acetate in water (as negative mode), and the mobile phase B contained acetonitrile. Next, the elution gradient was set: 1% B for 0–1.0 min, 1–99% B for 1.0–8.0 min, 99% B for 8.0–10.0 min, 99–1% B for 10.0–10.1 min, and 1% B 10 for 1–12 min with 0.5 mL/min of flow rate. The QE mass spectrometer was employed due to its ability to acquire MS/MS spectra on the information-dependent acquisition (IDA) mode in the control of the acquisition software (Xcalibur 4.0.27, Thermo). Furthermore, the ESI conditions were set as follows: capillary temperature was at 400 °C, full MS resolution was 70,000, sheath gas flow rate was 45 Arb, Aux gas flow rate was 15 Arb, MS/MS resolution was 17,500, collision energy was 20/40/60 in NCE mode, and voltage spray was 4.0 kV (positive) or −3.6 kV (negative), respectively.2.9. Metabolomic Data ProcessingAfter LC–MS/MS analysis, ProteoWizard was used to convert the acquired raw data into mzXML format and to process with an in-house program, which was developed using R and based on XCMS for peak detection, extraction, alignment, and integration. The normalized total peak intensity was then analyzed by multivariate data, including PCA and orthogonal partial least squares discriminant analysis (OPLS-DA). The importance of each variable in the OPLS-DA model in the projection (VIP) value was further calculated to show its contribution to the classification. Metabolites with a VIP score > 1 and a p value less than 0.05 are considered statistically significant. The functional interpretation of the relevant KEGG pathways in these metabolites was revealed by the MetaboAnalyst, based on hypergeometric testing [33].2.10. Statistical AnalysisStatistical analysis was performed using GraphPad software (version 5 for Windows). Significant differences between each treatment were measured using t-test, and p < 0.05 was considered significant. The asterisks (*) represent the statistically significant difference in the production performance of hens with different FR programs (FR20 or FR40) compared to the AL group (p < 0.05). Differences were considered significant for p values of 0.05 and were considered to represent trends for p values greater than 0.05 and less than or equal to 0.10.3. Results3.1. Effect of FR on Production Performance of Laying HensFigure 1a showed that FR significantly reduced the body weight of FR20 and FR40 treated hens compared to the AL group (p < 0.05). Similar with the body weight, the weight of liver, and abdominal fat in laying hens was also affected by the increase in FR. Specifically, restrictive feeding significantly reduced the weight of liver and abdominal fat in FR20 and FR40 treatment compared to ad libitum feeding (p < 0.05, Figure 1c,d), while FR40 had a significant reduction in liver weight compared to FR20 (p < 0.05, Figure 1c). Although there was no significant difference in the egg production rate between FR20 and AL group (Figure 1b), the egg production rate was significantly reduced in FR40 treatment (p < 0.05, Figure 1b). However, there is no significant difference between the FR20 egg production level and AL group (p < 0.05, Figure 1b), which indicated that the egg production efficiency of hens can be improved under the condition of moderate feed restriction [13,22].3.2. Microbiota SequencingAfter quality trimming and chimera checking through the QIIME pipeline, effective tags were revealed in each sample, and operational taxonomic units (OTU) were picked by clustering at the 97% identity level using UPARSE. Moreover, Venn diagram analysis revealed a total of 996 OTU in the cecum of hens. Among them, 14, 19, and 22 OTU were found in AL, FR20, and FR40 groups, respectively. In addition, in the AL and FR20, FR20 and FR40, and AL and FR40 groups, 26, 13, and 9 OTU were found as common OTU, respectively. Further, Venn diagram showed 893 OTU among these three groups (Figure 2). Species-accumulation curve analysis [34] was then used to estimate the number of species in these samples. It was revealed that a large proportion of species existed in the cecal community (Figure S1). In particular, we showed a sharp increase in the measurement curve, which means there was a substantial increase in species abundance. In addition, the flat curve showed that the number of species did not increase, even though the number of samples increases. The results of this increase in the number of samples indicated that the samples number provided in this study is sufficient.3.3. Effect of FR on Microbial Diversity and Relative AbundanceChao1, ACE, and Shannon are several indices that calculate the abundance or distribution of OTU within a particular population. For example, low richness suggests a low number of species in the community, and low evenness indicates that the sample consists of a few dominating taxa. Chao1, ACE, and OTU-observed species indices are indicators for species richness analysis of gut microbiome; OTU also can classify a group of closely related species [35], while Shannon is used to predict species diversity of a community [35,36]. The diversity of cecum microbiota in three groups, AL, FR20, and FR40, is shown in Figure 3. Compared to the AL group, the restrictive feeding treatments (FR20 and FR40) of the hens had no significant difference in Chao1, ACE, and OTU-observed species indices (Figure 3a,b,d, respectively). A similar result appeared in Shannon, where restrictive feeding did not affect the diversity of species in FR20 and FR40, as compared to AL group (p < 0.05, Figure 3c). Nevertheless, FR resulted in a change in the cecum microbial composition. However, restrictive feeding resulted in changes in the microbial composition of the cecum. Phylogenetic classification of OTUs (Figure 4a) revealed that Bacteroidetes and Firmicutes were the predominant phyla. The relative abundances of Bacteroidetes in FR20 and FR40 were 51.29% and 45.32%, respectively, compared to the AL group (47.08%). Similar to Bacteriodetes, the relative abundances of Firmicutes were higher in the FR group (FR20 was 20.21% and FR40 was 19.52%) than in the AL group (18.57%). On the other hand, the relative abundances of Proteobacteria were higher in the AL group (11.57%) than in both the FR20 group (9.83%) and FR40 group (9.34%); these findings are similar to the abundance of two phyla, Epsilonbacteraeota (AL, FR20, and FR40 were 3.73%, 3.13%, and 1.70%, respectively) and Spirochaetes (AL, FR20, and FR40 were 2.24%, 2.00%, and 2.1%, respectively). Furthermore, the relative abundance of Cyanobacteria (6.46%), Kiritimatiellaeota (4.85%), Verrucomicrobia (3.5%), Synergistetes (2.72%), and Deferribacteres (2.34%) were higher in FR40 than in FR20 and AL. We also predominantly compared the relative abundance of genera (Figure 4b) with the level reached by Bacteroides and Rikenellaceae_RC9_gut_group. FR20 (19.06%) showed the higher relative abundance of genus Bacteroides followed by AL (17.01%) and FR40 (12.73%). In Rikenellaceae_RC9_gut_group relative abundance, the FR40 group (11.67%) was higher than in the FR20 group (10.11%) and AL group (9.9%). The three genera, Helicobacter (3.69%), Parasutterella (2.93%), and Parabacteroides, (1.54%)) of the AL group were higher than in both FR treatments, FR20 (Helicobacter (3.1%), Parasutterella (2.86%), and Parabacteroides (1.5%)) and FR40 (Helicobacter (1.6%), Parasutterella (2.29%), and Parabacteroides (1.27%)). Beyond the genus Bacteroides, FR 20 was also higher in Phascolarctobacterium (3.47%) and Sutterella (2.59%) compared to FR40 (Phascolarctobacterium and Sutterella was 3.14% and 0,84%, respectively). In particular, FR40 showed a higher result in three genera (Synergistes (2.73%), Mucispirillum (2.34%), and Cerasicoccus (2.61%)) compared to FR20 (Synergistes (1.94%), Mucispirillum (1.7%), and Cerasicoccus (0.44%)) and AL (Synergistes (1.33%), Mucispirillum (1.16%), and Cerasicoccus (1.81%)). Parabacteroides slightly decreased in FR40 (12.68%) compared to FR20 (15.13%) and AL (15.39%).Furthermore, we also investigated the details of the main phylum (Figure 4c) and main genus (Figure 4d). In the main phylum (Figure 4c), although Bacteroidetes and Firmicutes were the predominant phyla, there was no significant difference amongst the group. The AL group has a higher relative population of Proteobacteria with significant differences compared to FR20 and FR40 (p < 0.05). The relative abundance of Cyanobacteria, Kiritimatiellaeota, and Verrucomicrobia of FR40 were significantly higher than FR20 (p < 0.05). However, the AL group was significantly increased in Cyanobacteria and Verrucomicrobia compared to FR20 (p < 0.05). FR40 significantly reduced the relative abundance of Epsilonbacteraeota compared to FR20 and AL (p < 0.05). There was no significant difference in phylum Spirochaetes amongst the group. In addition, Synergistetes and Deferribacteres significantly increased in FR40 compared to in the AL group (p < 0.05). Moreover, at the genera level (Figure 4d), the most predominant genus was Bacteroides, and the FR40 group decreased its relative abundance compared to the FR and AL groups (p < 0.05). Four genera (Rikenellaceae_RC9_gut_group, Phascolarctobacterium, Mucispirillum, and Parabacteroides) had no significant difference amongst the group. FR20 significantly increased in Helicobacter and Sutterella compared to FR40 (p < 0.05). In Helicobacter, FR40 also decreased compared to the AL group (p < 0.05), and in Sutterella, AL was lower than FR20 (p < 0.05). Nevertheless, AL and FR40 were sharply increased in Cerasicoccus compared to FR20 (p < 0.05).Principal component analysis (PCA) was used to investigate the distribution of identified OTU (microbiota) in each current taxon. Our result showed that the clusters between the treatments (AL, FR20, and FR40 groups) were significantly separated (p < 0.05, Figure 5a); the principal components PC1 and PC2 had 13.2% and 12.7% variation, respectively. Further, partial least squares discriminant analysis (PLS-DA) was used to reveal the group distinction, indicating that the identified OTU had a good discrimination between groups (p < 0.05, Figure 5b); the variation of PLS1 was 12.45%, and that of PLS2 was 11.91%. These results indicated that PCA and PLS-DA could clearly distinguish the OTU revealed in this study; the OTU in the microbial population were different between the AL, CR20 and CR40 groups.We also explored the similarity of the identified OTU among each group. To this end, a simple agglomerative hierarchical clustering method based on a weighted pair group method with arithmetic mean (WPGMA) was first used to create clusters of similar origin in the treatment (AL, FR20 and FR40 groups; Figure 5c). Then, a statistical test with ANOSIM was performed to examine whether there was a significant difference between the groups. Of these, the R value of the FR20 and FR40 groups was 0.456, while the Bonferroni corrected p value was 0.008. The R value of the AL and FR40 group was 0.536, and the Bonferroni corrected p value was 0.01. The AL and FR20 groups resulted in an R value of 0.208 with a Bonferroni corrected p value of 0.017 (Table S1), suggesting that there was significant dissimilarity among the different groups.The ontology of the microbial phenotype was then revealed. Importantly, as compared to AL group, the restricted feeding group represented higher anaerobic bacteria and less aerobic flora (p < 0.05, Figure 6a,b). The three groups have the same number of anaerobic bacteria that dominate the cecum (Firmicutes and Bacterioidetes), with FR20 having the higher number (p < 0.05, Figure 6a). In addition, Proteobacteria were the only aerobic bacteria that dominate the cecum among the three groups (p < 0.05, Figure 6a). In this analysis, FR20 has highest number amongst all other groups (p < 0.05, Figure 6b). Moreover, the FR40 group observed a decrease in the total population of potential pathogenic bacteria (p < 0.05, Figure 6c). Both FR20 and FR40 reduced Proteobacteria as potential pathogenic bacteria (p < 0.05, Figure 6c), indicating that the phenotype of the intestinal microbiota may be affected by FR in laying hens.3.4. Effect of FR on KEGG Pathway in Cecum and SerumAn untargeted LC–MS-based metabolomics platform was used to analyze the cecum contents and the serum metabolite profiles of chicken fed ad libitum and restricted. According to the variable importance in the projection (VIP) value > 1, in 95% jack-knifed confidence intervals and p < 0.05, detailed information about the different biomarker metabolites has been shown in Table S1. Compared to the AL group, the FR20 group had 20 different metabolites, with 7 LC–MS/MS (ESI+) and 13 LC–MS/MS (ESI−), which were detected in the cecal contents, and 10 different metabolites (8 ESI+ and 2 ESI−), which were detected in serum contents. The FR40 group has 48 different metabolites, with 18 ESI+ and 30 ESI−, which were detected in cecal contents, and 31 different metabolites (16 ESI+ and 15 ESI−), which were detected in serum, as compared to the AL group.The enrichment of relevant KEGG pathways in these metabolites was further revealed by the web-based pipeline MetaboAnalyst [33]. Figure 7 showed the different metabolites in the cecal and serum contents. As shown in Figure 7a, compared to the AL, the FR20 group modified relevant pathways in one more lipid, two carbohydrates, one amino acid, and one nucleic acid. These metabolites upregulated in cecal contents included LysoPC (18:2(9Z,12Z)), carbohydrate-related metabolites isomaltose and D-maltose, amino acid metabolite L-aspartic acid, and thymidine as the nucleic acid metabolite. Three lipid-, four carbohydrate-, two amino acid-, and five nucleotide-related metabolites were downregulated in cecal contents of chicken with FR20 compared to the AL group; further decreased metabolites were AICAR and thymine (nucleotide-related metabolites), sucrose and beta-D-glucose (carbohydrate-related metabolites), and serotonin (amino acid-related metabolites). Additionally, in FR20, three more lipid-related metabolites (stearidonic acid, 13-L-Hydroperoxylinoleic acid, and LysoPC (15:0)) and five amino acid-related metabolites (D-Glutamine, L-Glutamine, L-Tryptophan, N-Methylhydantoin, and Pyrrolidonecarboxylic acid) were increased in the serum of chicken. There were no downregulated metabolites in this group compared to in the AL group (Figure 7c).Furthermore, as compared to the AL group, FR40 modified the relevant pathways of two more lipids, two more carbohydrates, two more amino acids, one more nucleotide, and three more steroids, which were increased in cecal contents. These higher metabolites included carbohydrate-related metabolites such as Dolichyl b-D-glucosyl phosphate and D-maltose, steroid-related metabolites including 11b-Hydroxyandrost-4-ene-3,17-dione and adrenosterone, and prostaglandin (E2) as a lipid-related metabolite. Seven lipids, three carbohydrates, five amino acids, two nucleic acids, and one vitamin and cofactor (pantothenic acid)-related metabolites were decreased in cecal contents. These metabolites included AICAR, histamine, sucrose, beta-D-glucose, gamma-linolenic acid, oleic acid, thymine, and pantothenic acid (Figure 7b). In serum contents, as shown in Figure 7d, the most upregulated metabolites in FR40 were amino acids (ten more amino acid), including D-Ornithine, carnosine, creatinine, and L-Arginin. The other metabolites were two lipids (Stearidonic acid and 13-L-Hydroperoxylinoleic acid) and two carbohydrates (gluconic acid and D-lactic acid) as compared to the AL group. Additionally, three more vitamins and cofactors (riboflavin, biotin, and pantothenic acid), two more lipid-related metabolites (LysoPC(18:1(9Z)) and PA (16:0/16:0)), and three more amino acid related metabolites (N-Acetylserotonin, choline, and L-Phenylalanine) were downregulated in serum contents of chicken with FR40 compared to the AL group chicken.3.5. The Relationship of Different Relative Abundance of Bacteria in the Cecal Microbiota with Cecal and Serum MetabolitesPearson’s correlation analyses showed that the relative abundance of different bacteria (LEfSE) at the phylum level in the cecal microbiota were found to be closely associated with the concentration of specific metabolites in the cecum and serum of chickens (Figure S3). Cecal microbiota of FR20 group showed that Firmicutes were the high proportion of bacteria-correlated metabolites in the phylum level, followed by Bacteriodetes, Proteobacteria, Elusimicrobia, Euryarchaeota, and Verrucomicrobia. Firmicutes were positively correlated with LysoPC(18:2(9Z,12Z)) especially for genus Oribacterium; L-Aspartic acid for genus Ruminiclostridium_9, Oribacterium, and Butyricicoccus; beta-D-Glucose metabolite for genus Ruminococcaceae_UCG_004 and Ruminococcaceae_UCG_014; and AICAR metabolite for genus Ruminococcaceae_UCG_014, Ruminococcaceae_UCG_005, and Lachnospiraceae_NK4A136_group. In addition, this phylum was negatively correlated with LysoPC(18:2(9Z,12Z)) for genus Ruminococcaceae_UCG_014, Ruminococcaceae_UCG_004, Ruminococcaceae_UCG_005, Christensenellaceae_R_7_group, and Ruminococcaceae_UCG_010; Thymidine metabolite for genus Ruminococcaceae_UCG_004 and Christensenellaceae_R_7_group; D-Maltose for genus Ruminococcaceae_UCG_010, Christensenellaceae_R_7_group, Ruminococcaceae_UCG_005, and Ruminococcaceae_UCG_004; L-Aspartic acid metabolite for Ruminococcaceae_UCG_014, Ruminococcaceae_UCG_004, and Ruminococcaceae_UCG_010; beta-D-Glucose metabolite only for genus Ruminiclostridium_9; and AICAR metabolite only for genus Oribacterium. Furthermore, other phylum, Bacteroidetes, had two genus that correlated with cecal metabolites. Genus Prevotellaceae_Ga6A1_group was positively correlated with Thymidine and D-Maltose. On the other hand, genus Rikenella was negatively correlated with Thymidine and D-Maltose. Furthermore, genus Azospirillum_sp_47_25 from phylum Proteobacteria had positively correlated with beta-D-Glucose and negatively correlated with LysoPC(18:2(9Z,12Z)) and L-Aspartic acid. Moreover, genus Elusimicrobium (phylum Elusimicrobia), genus Methanocorpusculum (phylum Euryarchaeota), and genus Cerasicoccus (phylum Verrucomicrobia) were negatively correlated with D-Maltose, LysoPC(18:2(9Z,12Z)), and D-Maltose, respectively (Figure S3).As shown in Figure S4, there was no positively correlation between cecal microbiota of FR20 group and serum metabolites. On other hand, phylum Cyanobacteria, Euryarchaeota, Verrucomicrobia, and Elusimicrobia had negatively correlated with D-Glutamine metabolites but had no significant differences.The relative abundance of phylum microbiota in the cecal of FR40 group was Fusobacterium (p < 0.05) that had positively correlated with oleic acid and negatively correlated with dodecanoic acid (Figure S5). Additionally, microbiota in this group’s cecal, Fusobacterium had positively correlated with choline and negatively correlated with L-Glutamine, L-Arginine_1, L-Histidine, and L-Alanine as metabolites in serum chicken of FR40 group (Figure S6).4. Discussion4.1. Performance ParametersFR is one of the important methods of feeding and management of commercial laying hens. It is an effective way to control the feed intake to limit the energy and crude protein levels of the hens in the diet so as to improve the performance of layer flock. FR for hens during the rearing period was beneficial in maintaining their proper weight, allowing the bones and internal organs to be fully developed, and avoiding premature maturity [9,12], because these conditions have a negative impact on their egg production performance. In the late phase of the laying hens, due to aging [37], the laying performance (egg production and egg quality) of laying hens decrease quickly [38]; also, the mature body weight causes them to require lesser dietary energy [18,19]. Hence, the main purpose of FR for laying hens during this period is to prevent them from overfeeding and reducing the amount of abdominal fat in the laying hens, so that the laying hens can maintain a proper weight.It has been indicated that, when the FR reaches 10%, for 54 weeks, laying hens’ egg production rate is not affected [16]. However, for FR20, the egg production rate of hens was significantly reduced. In our study, FR20 was not observed to have a significant effect on the egg production rate of 48 weeks laying hens compared to AL group. Moreover, in the FR40 group, the egg production rate of hens is significantly decreased compared to that of AL control (p < 0.05). This result is similar to those of other studies [11,39,40]. It indicated that the egg production efficiency of hens can be improved under the condition of moderate feed restriction [13,22]. According to [39,41], the egg production level of laying hens depends on the energy intake. In this way, the more intense the restriction of energy intake, the greater the negative effect on egg production. Although Snetsinger et al. [42] stated that only 6–7% FR can be used in poultry after 40 weeks to significantly reduce the egg production rate, recent evidence has revealed that the effect of FR on hens’ laying performance may relate to the type of layer used and the implementation of FR [13].FR program plays an important role in the growth performance and nutrient utilization of laying hens [11]. Consistent with previous reports [16,18,20], we demonstrated that FR could reduce the body weight of hens in the late-phase of laying; additionally, we recorded a decrease in the abdominal fat weight of these hens. This may indicate that the FR levels used in this study are not sufficient to support unwanted fat deposits, which tends to support the suggestion of Kingsley et al. [17], that FR chickens make more efficient use of ingested feed through higher metabolic efficiency and fewer excess fat deposits.Since FR can reduce the efficiency of liver metabolism, the effect of the intensity and duration of the FR can lead to a reduction in liver weight [43]. Our results showed that, compared with the AL control, FR reduces the liver weight of laying hens, which may further lead to the reduction in abdominal fat in FR20 and FR40; it is correlated to the activity of lipogenic enzyme, which was decreased during the period of FR and gradually declined in the subsequent weeks [44]. In general, healthy laying hens have abdominal fat, as abdominal fat can be an important body resource to maintain the laying when the dietary nutrient supply is insufficient for the egg formation [45]. High levels of abdominal fat may trigger fatty liver degeneration, which leads to negative impacts on body metabolism [46]. Consider the purpose of applying FR to laying hens is to prevent them from overfeeding and reduce the amount of abdominal fat in later laying hens. Therefore, FR measures are beneficial for maintaining the proper weight of laying hens.4.2. Effect of FR on Microbial Diversity and Relative AbundanceIn order to design nutritional strategies to effectively improve chickens’ feed efficiency, the relationship between a chicken’s feed intake, intestinal microbiota, and host nutritional metabolism needs to be clarified to better understand the underlying modes of action for the difference in feed efficiency [1,2,47]. Both chicken feed intake and intestinal bacterial microbiota differ between chickens with high and low feed efficiency [4,21]. Additionally, FR is also related to intestine microbiota [4]. Sergeant et al. [48] explained that intestinal microbiota plays a vital role in the health, production performance, and welfare of chickens, including laying hens. The most dense microbiota population in chicken intestine is in the ceca, a pair of blind-ended sacs between the small and large intestines [49]. In addition to the microbiota, the ecological environment in the cecum of laying hens is also composed of metabolites [50]. Thus, there is a correlation between the gut microbiota and feed efficiency in chicken production. We hypothesized that this approach would be helpful in clarifying whether differences in feed intake between hens with high and low feed efficiency (feed restriction) would have an impact on shifts in taxa abundance in the composition of the bacterial community. We observed a strong impact of restrictive feeding on the intestinal physiology and bacterial communities (microbiota), mainly influencing the predominant bacteria (phyla and genera) in cecal and serum.This study showed that FR increased the relative abundance of phyla Bacteriodetes and Firmicutes, especially for FR20 treatment. This result was consistent with a previous study on FR-enriched phylum Bacteriodetes [50,51] and Firmicutes [4,50,51]. Meanwhile, Bacteriodes and Rikenellaceae_RC9_gut_group were the major abundant genera in laying hens cecal. A previous study [51] showed that genera Bacteriodes and Rikenellaceae_RC9_gut_group were dominant in the cecal of laying hens, even though it had heat stress; furthermore, this mainly microbial composition was related to feed intake.All group experiments investigated the relative abundancy and dominance of anaerobic bacteria. Firmicutes (Gram-positive) and Bacteriodetes (Gram-negative) were relatively abundant in FR20 and FR40 groups; furthermore, Verrucomicrobia and Synergistetes (both Gram-negative bacteria) were less abundant in FR20 than in FR40. The predominant culturable bacteria in the poultry cecal are obligate anaerobes at the level of 1011/g of content [52]. Similar to another study [53], Firmicutes, a Gram-positive and obligat anaerob bacteria, is the predominant phylum found in poultry cecal. Meanwhile, Gram-negative non-sporing anaerobic rod bacteria, also found in poultry, duck, and turkey cecal, is of the family Bacteroidaceae [52]. A small proportion of phyla found in chicken intestines included Cyanobacteria, Spirochaetes, Synergistetes, Fusobacteria, Tenericutes, and Verrucomicrobia [54]. In the normal cecal adult chicken, over 40 different types of anaerobic Gram-negative and Gram-positive non-sporing rods and cocci have been found, as have at least 17 different species of Clostridia [55]. There are many other organisms that are still to be isolated and characterized, including Proteobacteria, an aerobic bacteria, found predominantly in chicken cecal after Firmicutes and Bacteriodetes [52,53,54].The presence of normal flora in the intestine mainly functions to control or eliminate an invading pathogen; it might be considered competition for limiting carbon sources [55]. The pathogen result of these experiment was Bacteriodetes, Firmicutes, and Proteobacteria. Both FR20 and FR40, especially FR40, could decrease Proteobacteria as potentially pathogenic, compared to the AL control. Among the Proteobacteria, the predominant genera were Desulfohalobium, Escherichia, Shigella, and Neisseria [56]. One of them, Escherichia coli, is an intestine gamma proteobacterium. During the whole life cycle of healthy chickens, this bacterium is found in low abundance. However, several strains of Escherichia coli have specific virulence factors may have infected in chickens and cause disease; these strains are known as avian pathogenic Escherichia coli (APEC). APEC is principally associated with extra intestinal infections that affect the respiratory tract [54]. This result indicates that FR could decrease the population of potential pathogenic bacteria. Videnska et al. [57] showed that, in the ceca of mature laying chickens (up to 60 weeks old), the representative microbial communities at the phylum level, in order of their typical abundance, are Proteobacteria, Firmicutes, and Bacteroidetes, which formed the vast majority of microbiota across all age categories. This indicated that Gram-negative bacteria were the abundant phyla that dominate the gut, and Firmicutes become more dominant in the later age of the laying hen [58].4.3. Effect of FR on KEGG Pathway in Cecum and SerumGut microbiota performs a large number of roles of the host through functional microbial pathways. Metabolomic analysis revealed that LysoPC(18:2(9Z,12Z)) upregulated the KEGG pathway of lipid synthesis in the cecal content of the FR20 group. LysoPC(18:2(9Z,12Z)) is Lysophosphatidylcholines, one of the major structural lipids in a eukaryotic membrane cell. Higher phosphatidylcholine demonstrated that the lipid membrane cell structure is under elastic stress. This stress changes a membrane’s physical properties and results in its biological function by modifying the membrane and influencing lipid–protein interaction [59]. This is a normal condition for the membrane to deal with stress, so that it can still control the lipid function of the membrane. Furthermore, at the same location of the FR20 group, isomaltose and D-maltose also upregulated the KEGG pathway within its carbohydrate metabolism. On the other hand, sucrose and beta-D-glucose are, instead, downregulated. Theoretically, this result correlated with Hornbuckle et al. [60], who explained two steps of starch hydrolyzation, i.e., rapid and slower steps. The rapid one results in maltose and maltotriose formation, and the slower one involves hydrolysis of maltotriose into maltose and glucose. The α-amylase is an enzyme produced by the pancreas that functions to catalyze the specific hydrolysis of α-1,4-glucosidic bonds of starch and glycogen. When α-amylase hydrolyze starch, this will produce the principal product with a predominance of maltose (α-1,4-glycosidic bond), isomaltose (α-1,6-glucosidic bond), and small amounts of glucose, which correlated with our result. Moreover, Hornbuckle et al. [60] also stated that the enzymes maltose and isomaltose are integral parts of the microvillus membrane, so that the final hydrolysis of these two compounds (maltose and isomaltose) occurs at the surface of the intestine mucosal cell. The result of the KEGG pathway of carbohydrate synthesis in the cecal content of FR20 group is the same as the result at FR40. This result indicated that, although FR was carried out, it did not interfere with the function and biochemical processes of the intestine, specifically in carbohydrate synthesis.AICAR also downregulated AICAR in the cecum of the FR20 group. AICAR, Aminoimidazole-4-carboxamide-1-b-DD-ribofurano-side, is one of the AMPK activators. It is phosphorylated in cytosol by adenosine kinase and is further converted to AICAribotide (ZMP), which activates AMPK by mimicking AMP [61,62]. Hereafter, AMPK regulates the expression of various genes that are involved in the glucose metabolism [63]. This condition, related to carbohydrate synthesis, results in the feed restriction group’s cecal result. Our study showed that FR caused the AICAR to be downregulated; then, this condition resulted in a decrease in beta-D-glucose. So, it indicated that AICAR effected other physiological responses, such as hypoglycemia [64]. In addition, this study showed that dietary restriction provided another advantage with the decrease in histamine in the cecum of the FR40 group. Histamine, an amino acid metabolism, plays an important role in epithelial protection. Lower concentrations of histamine might be protecting the epithelial, whereas higher concentrations of it might be detrimental to epithelial protection from pathogen infection [65]. The continuously production of small amounts of histamine, which is produced by the gut microbiota, can lead to suppression of intestinal inflammation [66]. In the blood of laying hens of both FR20 and FR40 groups, two lipid metabolites (stearidonic acid and 13-L-Hydroperoxylinoleic acid) enriched fatty acid biosynthesis. This study indicated that FR conditions initiated in chickens at 48 weeks of age resulted in an increase in lipid metabolism. Above all, FR altered cecal and blood metabolic pathways, especially biosynthesis pathways of lipids and carbohydrates.4.4. The Relationship of Different Relative Abundance of Bacteria in the Cecal Microbiota with Cecal and Serum MetabolitesMicrobiota related to cecal metabolite (metabolomic) analysis in the FR20 group in this recent study showed that phylum Firmicutes have a positive correlation with metabolites such as LysoPC (18: 2 (9Z, 12Z)) (through involvement of the genus Oribacterium), L-Aspartic acid (involving the genus Ruminiclostridium_9, Oribacterium, Butyricicoccus), beta-D-Glucose (through involvement of the genus Ruminococcaceae_UCG_004, Ruminococcaceae_UCG_014), and AICAR (involving the genus Ruminococcaceae_UCG_014, Ruminococcaceae_UCG_005, Lachnospiraceae_NK4A136_group). This LysoPC (18: 2 (9Z, 12Z)) metabolite has an important role in the biosynthesis of fatty acids [44], L-aspartic acid and AICAR play a role in amino acid metabolism [61,62], and beta-D-Glucose functions in carbohydrate synthesis [67]. In the human intestine, Ruminococcaceae has a functional ability to degrade nondigestible carbohydrates such as resistant starch, hemicellulose, and cellulose [68].Furthermore, compared to the FR20 group, phylum Fusobacterium was the only phylum in the FR40 group that had a positive correlation with oleic acid in chicken cecum and choline chicken serum. Oleic acid is a monounsaturated fatty acid with several biological functions: it enhances mitochondrial oxidation of saturated fatty acids (by increasing triacylglycerol and by reducing diacylglycerol and ceramide production) and displays the ability to prevent SFA-induced inflammation, thus protecting the cells from inflammation [69]. Choline, a micronutrient often classified with the B-vitamins, is a proven lipotropic agent in several species of animals [70]. A lipotropic agent is a compound that has an affinity for lipids and thus can help to catalyze the breakdown of fat during metabolism in the body [70,71]. Choline testing can be performed using a blood sample, called the plasma concentration of fat-soluble choline biomolecules [70]. All these FR40 correlations indicated that feeding restriction can affect phylum Fusobacterium, triggering both oleic acid for lipid synthesis and choline for the synthesis of vitamins, in cecum and blood, respectively.5. ConclusionsUnderstanding the fundamentals of host–microbial interactions is crucial for creating cost-effective laying hen production techniques. There will undoubtedly be savings under quantitative FR circumstances owing to lower feed prices, particularly for hens in the late phase of life. As a result, this research aids in the identification of intestinal bacteria, the relationship between microbiota and metabolite profiles, as well as their metabolic pathways in connection to host nutritional needs and gut nutrient availability. This research may be utilized as a foundation for future research into how dietary interventions affect the intestinal microbiota, host physiology, egg quality, and feed efficiency in chickens. In the end, this method can be applied as an alternative method in late-phase laying hens to reduce overfeeding which aims to reduce feeding costs.6. PatentsThere is no patent resulting from the work reported in this manuscript.

animals : an open access journal from mdpi

10.3390/ani12040437

PMC8868309

Camera traps acquire visual data in a non-disturbing and round-the-clock manner, so they are popular for ecological researchers observing wildlife. Each camera trap may record thousands of images of diverse species and bring about millions of images that need to be classified. Many methods have been proposed to classify camera trap images, but almost all methods rely on very deep convolutional neural networks that require intensive computational resources. Such resources may be unavailable and become formidable in cases where the surveillance area is large or becomes greatly expanded. We turn our attention to camera traps organized as groups, where each group produces images that are processed by the edge device with lightweight networks tailored for images produced by the group. To achieve this goal, we propose a method to automatically design networks deployable for edge devices with respect to given images. With the proposed method, researchers without any experience in designing neural networks can develop networks applicable for edge devices. Thus, camera trap images can be processed in a distributed manner through edge devices, lowering the costs of transferring and processing data accumulated at camera traps.

Camera traps provide a feasible way for ecological researchers to observe wildlife, and they often produce millions of images of diverse species requiring classification. This classification can be automated via edge devices installed with convolutional neural networks, but networks may need to be customized per device because edge devices are highly heterogeneous and resource-limited. This can be addressed by a neural architecture search capable of automatically designing networks. However, search methods are usually developed based on benchmark datasets differing widely from camera trap images in many aspects including data distributions and aspect ratios. Therefore, we designed a novel search method conducted directly on camera trap images with lowered resolutions and maintained aspect ratios; the search is guided by a loss function whose hyper parameter is theoretically derived for finding lightweight networks. The search was applied to two datasets and led to lightweight networks tested on an edge device named NVIDIA Jetson X2. The resulting accuracies were competitive in comparison. Conclusively, researchers without knowledge of designing networks can obtain networks optimized for edge devices and thus establish or expand surveillance areas in a cost-effective way.

1. IntroductionVisual data are a rich source of information about wildlife and can provide strong support for wildlife conservation and ecological research. One cost-effective way to obtain visual data of wildlife is via camera traps that work in a non-disturbing [1] and round-the-clock manner [2], thus making them ideal for observing wild animals otherwise difficult to monitor [3], e.g., nocturnal mammals [4] and large animals [5]. Because camera traps are noninvasive [6], a single deployment may record a diverse range of species [7]. Consequently, the recorded images have to be processed before being adopted in ecological research [8]. The images may go through several processing stages determined by the research process, a fundamental stage of which is species identification that is usually implemented as automatically and centrally classifying camera trap images at a data center installed with very deep convolutional neural networks (CNNs) [9,10,11,12,13]. In practice, there may be millions of images produced by camera traps [7,12,14,15], so image transfer and processing at a data center is often computationally intensive and costly. Furthermore, the scale of the surveillance area may also be restricted by the processing capability of the data center.Edge computing [16] was ideally developed for such cases, i.e., intensive computation centralized at a data center can be split and localized by edge devices near camera traps [17,18]. Thus, fundamental processing steps such as removing images without animals [6,7,12,19,20,21] and classifying images with animals [9,10,11,12,13] can be automatically conducted on edge devices. However, edge devices are not only heterogeneous [22] but also resource constrained [23]. These limitations of edge devices narrow down the range of available neural networks [23,24]. Hence, lightweight networks [25] designed for edge devices are critical in edge computing for camera trap images. Even so, “deep neural network design is very difficult, and it requires the experience and knowledge of experts, a lot of trial and error, and even inspiration” [26]. Luckily, network design can be automated through neural architecture search (NAS) [27]. However, NAS is often developed regardless of domain knowledge [28] regarding camera trap images of wildlife [29]. Specifically, NAS is often designed based on benchmark datasets such as CIFAR-10 [30] and ImageNet [31], which differ from camera trap images in many aspects, especially data distribution and aspect ratios as described below.The data distribution of benchmark datasets may differ from camera trap images, e.g., in classes, image foregrounds and backgrounds. Camera trap images purely contain animals, but only partial classes in benchmark datasets are relevant to animals. For instance, six out of ten classes in CIFAR-10 are related to animals and 233 out of 1000 classes in ImageNet are relevant to vertebrate [32]. Consequently, NAS based on benchmark datasets may waste resources on designing networks optimized for data irrelevant to animals. In addition to classes, animal images in benchmark datasets also differ from camera trap images in foregrounds and backgrounds. For benchmark datasets [30,31], images are usually artificially preprocessed to guarantee that foreground animals are large and centered and their backgrounds are relatively small and may differ from animal habitats in the wild. In contrast, animals in habitats are photographed by camera traps under various conditions, so the animals may appear at random locations in images and are often closely related with image backgrounds.The image aspect ratios of benchmark datasets differ from camera trap images, e.g., the aspect ratios of CIFAR-10 and ImageNet are both 1:1 (the image width and height are the same), though this ratio may not hold for camera trap images. For instance, the resolutions of camera trap images range from 2048 × 1536 (aspect ratio: 4:3) to 2616 × 1472 (16:9) in North American Camera Trap Images, i.e., NACTI [13], and the resolutions range from 1920 × 1080 (16:9) to 2048 × 1536 (4:3) in Missouri Camera Trap Images, i.e., MCTI [33]. Therefore, networks found by NAS based on benchmark datasets may require that camera trap images be resized to satisfy the aspect ratio 1:1. However, resizing images may alter their aspect ratios and introduce interpolated pixels, often resulting in either misshaped animals or memory waste.In short, images from benchmark datasets adopted by NAS often differ from camera trap images, and this difference potentially implies domain shift [34]. Additionally, it may be hard to modify existing networks in line with the applications [29]. These issues inspired us to develop NAS based on the domain knowledge of camera trap images for edge devices. We used the proposed method to conduct searches directly on camera trap images rather than images of benchmark datasets. The aspect ratios of camera trap images are maintained during the search, which is guided by a loss function particularly derived for finding the lightweight networks. The hyper parameter of loss function was theoretically analyzed and carefully chosen, and lightweight networks found by the search were tested on the NVIDIA Jetson X2 edge device. The experimental results confirmed the validity of the proposed method. The main contributions of this paper are as follows. A method named Domain-Aware Neural Architecture Search (DANAS) was developed regarding the domain knowledge of camera trap images. The proposed method directly searches networks on camera trap images, thus avoiding negative effects such as the domain shift incurred by benchmark datasets in conventional search methods.Aspect ratios of camera trap images are maintained during the search. As part of domain knowledge, the changes of aspect ratio may not be automatically tackled by neural networks. Therefore, the changes are manually eliminated by first finding the most frequent aspect ratio and then padding images whose aspect ratios differ from the most frequent one.A loss function was derived to guide DANAS to find lightweight networks applicable for edge devices. A theoretical analysis of the proposed loss function was conducted, and the analysis revealed the value of hyper parameter in the loss function to boost its guiding effect on the search. 2. Materials and Methods2.1. DatasetsTwo datasets were employed in this study: MCTI and NACTI, containing 24 thousand and 3.7 million camera trap images, respectively, with varying resolutions. Since label errors are found in NACTI and its millions of images require too much computational resources, NACTI was selectively adopted in this study in the form of a subset named NACTI-a containing 29 thousand images with varying resolutions. The species data in NACTI-a and MCTI are illustrated in Table 1.2.2. MethodDANAS was developed within the framework of reinforcement learning [35,36], i.e., the search is implemented on sampling candidate networks from a search space through a sampler [29], as shown in Figure 1. In DANAS, the sampler is long short-term memory (LSTM) [37]. The reason to use LSTM as the sampler is that this sampler does not rely on parameter sharing [38], which may not be helpful for finding high-performance networks (as reported by [39]). Around the sampler, there are five conceptual search steps (from ① to ⑤ in Figure 1). By repeating these steps, the quality of the sampled network is gradually improved via updates of the learnable parameters θ of the sampler. Starting from the first step, all five steps are introduced sequentially next.In Step ① shown in Figure 1, LSTM samples candidate networks from the search space defined by a meta architecture, i.e., a prototype from which all candidate networks are derived. The meta architecture is similar to the ones defined in [35,36,37,38], i.e., a pipeline segmented to groups of layers called cells. There are two types of cells, normal and reduction cells, and the cells of the same type share the same inner structure. Besides the inner structures, the normal and reduction cells differ in the way they process data, i.e., the width and height dimensions of data remain the same before and after normal cells while the width and height of the input are halved through reduction cells. There are N normal cells in the pipeline, and each normal cell is adjacent to two reduction cells. At the end of the pipeline, a global average [40] is appended. In this study, the reduction cell was simplified to a single pooling layer, i.e., an average pooling or a max pooling with a kernel size of 5 × 5 or 3 × 3, and the normal cells were sampled based on the meta cell shown in Figure 2.As shown in Figure 2, a normal cell is a group of blocks whose inputs come from blocks in the same cell or previous B cells. For blocks not serving as inputs of any other blocks, their outputs are concatenated to produce the cell output. Each normal cell has B (a constant) blocks, and each block has M (determined by the sampler) operations. The operation is sampled from the same set of operations as that in [38], e.g., a stack of 3 × 3 depth-wise-separable convolution [41], batch normalization [42], and ReLU [43]. Accordingly, the sampler first determines the operation number M of a block by sampling an integer from some predefined integers, and then it samples inputs and operations for the block, and the sampling repeats for B blocks to form a normal cell. Once the normal cell has been sampled, the sampler samples a pooling layer to form a reduction cell. Both the sampled normal cell and the reduction cell are employed to build the candidate network.In Step ② shown in Figure 1, the candidate network is built based on the sampled cells and the meta architecture, i.e., assembling the cells according to the cell pipeline. The building process is identical with the one introduced in [44], i.e., we applied the adaptive meta-architecture [44] to build candidate networks. Once the candidate network is properly built, its performance is evaluated based on the camera trap images with maintained aspect ratios.In Step ③ shown in Figure 1, the candidate network is trained and validated on camera trap images with the most frequent aspect ratio, i.e., the occurrences of unique aspect ratios of camera trap images are counted and the aspect ratio with the maximal count is chosen as the most frequent one. Images with aspect ratios different the most frequent one are padded by zero pixels. In practice, camera trap images are processed to have the same aspect ratio before the search starts, and the processed images are employed to train the candidate network. The trained network is then validated to yield validation accuracy to compute the loss.In Step ④ shown in Figure 1, an accuracy reward [44] is generated based on accuracies obtained by training and validating a candidate network, and both the produced accuracy and the network parameter number [25] are employed to generate the loss J. The purpose of this step is to train LSTM to sample “good” networks via gradient-based optimization algorithms such as stochastic gradient descent (SGD). The meaning of “good” is twofold, i.e., the parameter number s of the network should be close to the desired parameter number (s∗=1.5 million in our case) and the accuracy reward R of the network should be close to the ideal accuracy (R∗=100, i.e., 100% accuracy). Since the reward is twofold, we need a bivariate reward function fR,s so that the gradient ∇θJ of the total loss J synchronizes with the reward. According to the case of the unary loss function in studies of reinforcement learning [45], we defined Jθ as (1)Jθ=aΣθfR,s, where θ represents learnable parameters associated with the sampler, R is the accuracy reward involving the training and the validation accuracies of the candidate network, and s is the parameter number of the network in millions. The bivariate function fR,s provides the reward based on R,s, and aΣ summarizes probabilities of sampling the candidate network through the sampler, i.e., (2)aΣθ=∑inC∑jBilogPnjθ+∑knjlogPaka1 : k−1,θ, where the notations are similar to [44], i.e., nC is the number of the cells in the candidate network, Bi denotes the block number of the ith cell, nj is the operation number of the jth block, and Pxy is the probability of sampling x under condition y. The details of aΣ can be found in Appendix A. Since (3)∇θR,sJ=fR,s⋅∇θaΣ, and fR,s yields a scalar, the direction of ∇θR,sJ is solely determined by ∇θaΣ. However, aΣ remains unknown due to the unknown probability distributions of nj and ak, which means the direction of ∇θaΣ is out of our control, i.e., we cannot change the direction of ∇θR,sJ to point to promising positions of high rewards. However, we can change its magnitude ∇θR,sJ via fR,s so that ∇θR,sJ synchronizes with the reward. For example, suppose the sampler sampled a network of R,s close to R∗,s∗; we expect the sampler to sample networks alike, which requires that θ should not be largely updated by SGD involving ∇θR,sJ. However, ∇θR,sJ is partially determined by ∇θaΣ, so ∇θR,sJ may not remain small when R,s is close to R∗,s∗. In this case, fR,s should scale ∇θaΣ to ensure that the resulting ∇θR,sJ is relatively small. This requires the reward surface defined by fR,s to be similar to a whirlpool with vortex R∗,s∗. We chose Witch of Agnesi [46] to build fR,s on account of its bell-like curve and the simple mathematical form that only introduces one hyper parameter. Therefore, fR,s is defined as (4)fR,s=R∗−8a3Rs−s∗2+4a22, where a∈ℝ is the hyper parameter introduced by Witch of Agnesi. In practice, R∗ usually equals 100 (100% accuracy) [44], s∗ is determined by the application, and only a remains unknown. The value of a may be discovered by assuming both R and s are restrained within some range, and this assumption may be reasonable under certain search conditions. Specifically, let x=s−s∗ and y=R; then fR,s can be written as (5)fx,y=R∗−8a3yx2+4a22.Assuming x1≤x≤x2 and y1≤y≤y2, the volume V of fx,y within the assumed ranges is given by V=∫y1y2∫x1x2R∗−8a3yx2+4a22dxdy =∫y1y2∫u1u2R∗−2aytan2u+12d2atanudy =∫y1y2∫u1u2R∗−2aycos2u22acos2ududy =∫y1y2∫u1u22aR∗2cos2u−4a2R∗y+8a3y2cos2ududy =∫y1y22aR∗2tanu u2 u1−4a2R∗yu u2 u1+2a3y2sin2u u2 u1+4a3y2u u2 u1dywhere x=2atanu and u1=tan−1x12a≤u≤tan−1x22a=u2. Suppose u2=−u1=u∗<π/2 and 0≤y≤R∗, then, the formula above can be simplified by substituting tanu and sin2u by their Taylor series of order three, i.e., (6)V=∫y1y24aR∗2tanu u∗0−8a2R∗yu u∗0+4a3y2sin2u u∗0+8a3y2u u∗0dy=∫y1y24aR∗2tanu∗−8a2R∗yu∗+4a3y2sin2u∗+8a3y2u∗dy =4aR∗2tanu∗y R∗0−4a2R∗u∗y2 R∗0+43a3sin2u∗+2u∗y3 R∗0 ≈4aR∗3u∗+u∗33−4a2R∗3u∗+43a3R∗34u∗−2u∗33!=43aR∗31−43a2u∗3+4aR∗31−a+43a2u∗=43aR∗33−4a23u∗3+3−3a+4a2u∗, which is equivalent to (7)u∗3+33−3a+4a23−4a2u∗=9V4aR∗33−4a2, which is a special case of monic cubic polynomials, i.e., the depressed cubic: u∗3+c1u∗=c2. According to Cardano’s formula, the solution of the depressed cubic is (8)u∗=c22+c224+c132723+c22−c224+c132723 where c1 and c2 are (9)c1=33−3a+4a23−4a2c2=9V4aR∗33−4a2 .The solution u∗ of the depressed cubic requires (10)c224+c1327≥0, which holds if c1≥0. The numerator of c1 is 3−3a+4a2 and its determinant is Δ<0, so 3−3a+4a2>0 holds regardless of a. The denominator of c1 is 3−4a2, so c1≥0 is equivalent to 3−4a2>0, which leads to a2<a<3/4.In practice, a=3/4−ε, where ε may take a small value such as 10−6. Figure 3 illustrates the surface of fR,s parameterized by a=3/4−10−6, R∗=100 and s∗=1.5 within the ranges 0≤R≤2R∗/3 and 0≤s≤5. As expected, f does have a whirlpool-like surface with the vortex R∗,s∗, and the sampler may be guided by ∇θR,sJ involving f to find lightweight networks.In step ⑤ shown in Figure 1, a selecting and training strategy is employed to find the optimal network. The idea behind this strategy is concentrating computational resources on promising networks found during the search, as the method first samples a relatively large number of candidate networks with small training epochs, e.g., 2 epochs in our case, and then finds the promising ones based on the sampled networks with large training epochs. In practice, we ran a single search to sample 1500 networks, and then networks with parameter numbers ranging from 1 to 1.5 million (the ideal parameter number in our case) were sorted decreasingly by their validation accuracies. If there are more than 150 networks, then only top 150 networks are retained for retraining through 5 epochs, and then the trained networks are sorted based on accuracies. If there are networks with accuracies >90%, then 15 networks are retained and retrained through 10 epochs; otherwise, half of the networks are retained and retrained. We stopped this procedure at 15 epochs and selected the top-1 network. If the difference between the accuracies between the top-1 and the top-2 networks was not large, e.g., less than 1%, then we would increase the epoch number and continue the training.3. ResultsThe performance evaluation of DANAS was individually conducted on the NACTI-a and MCTI datasets. As shown in Table 1, the most frequent resolution of both NACTI-a and MCTI is 2048 × 1536 (aspect ratio 4:3). Accordingly, the images of the two datasets were resized to have the resolutions 85 × 64 (4:3) [44] for the search and 224 × 168 (4:3) for the test; for each dataset, the search was conducted on 85 × 64 images, and then the optimal network discovered by the search was trained and tested on 224 × 168 images. Each dataset was split to three subsets, i.e., the training set, the validation set and the test set, and the search was conducted on the first two subsets. The split was implemented by randomly sampling images from the dataset at a ratio of 0.64:0.16:0.2 of the sample numbers of three subsets, namely, 20% images were randomly sampled from the dataset to build the test set, then 20% images were randomly sampled from the remaining images to build the validation set, and the rest of the images served as the training set. The candidate networks found by the search were trained on the training set and then tested on the validation set, so the test set remained unknown to the search.In searches on NACTI-a and MCTI, the pipeline shown in Figure 2 had three pairs of one reduction cell and five normal cells (N=3) at most. The normal cell had five blocks (B=5), and each block may have had five operations (M=5) at most. The input channels of the normal cell and reduction cell were, respectively, fixed to 20 and 40. The output channel of the reduction cell was fixed to 40, while the output channel of the normal cell was automatically determined by its operations. The candidate network was trained by using AMSGrad [47] with a batch size of 32, two epochs, and a learning rate of 0.005. The training was conducted on 85 × 64 training images via a PyTorch module named Distributed Data Parallel (DDP) that loaded the network and the batches to available GPUs, individually trained networks on GPUs, collected the resulting gradients from all GPUs and synchronized networks based on the collected gradients. The trained network was then tested on 85 × 64 images of the validation set on each GPU, and the resulting accuracies were retrieved via PyTorch module named Manager. The retrieved accuracies were then averaged to yield the training and the validation accuracies that were used to generate the accuracy reward. Finally, the loss was computed based on the accuracy reward through the loss function whose hyper parameters were set as a=3/4−10−6, R∗=100 and s∗=1.5. All searches were done on a workstation installed with 4 GPUs of NVIDIA TITAN Xp, Ubuntu 20.04, PyTorch 1.7.0 and MySQL 8.0.13.In tests, several networks famous for their lightweight designs or performance were chosen for comparison with DANAS, i.e., MobileNet-v2 [48], EfficientNet [49], DenseNet [50], Resnet-18 [51], ResNext [52] and Wide ResNet [53]. Each network was trained by using SGD [54] of Nesterov momentum [55] with a batch size of 10, 20 epochs, and a learning rate ranging from 0.005 to 0.0001. The learning rate was changed by cosine schedule [54]. The training was conducted on 224 × 168 images from both the training and the validation sets via DDP, and the weights of the network at the last epoch were saved on the hard disk. During tests, the weights were read from the disk and employed to populate the network, and the network was tested on 224 × 168 images of the test set. All networks in comparison were trained and tested on the workstation, and the optimal networks found by DANAS were additionally tested on an NVIDIA Jetson X2 edge device installed with Ubuntu 18.06 and PyTorch 1.1.0.Since the camera trap images differ widely between MCTI and NACTI-a, DANAS found different networks, which led to distinct accuracies and misclassifications for two datasets. The detailed results are discussed in the following sections.3.1. Search and Test on NACTI-aThe search on NACTI-a consumed roughly 74 hours and found a network with 1.36 million parameters. The search performance was compared with a random search via steps like those shown in Figure 1. Specifically, the sampler in step 1 of Figure 1 was replaced by random sampling, and both memory constriction [44] in step 2 and sampler updating in step 4 of Figure 1 were removed. However, the memory constriction could not truly be removed due to the limited physical GPU memory, and the constriction was thus alleviated by resampling the networks until the pipeline shrinkage [44] did not happen. The training and test configurations of networks explored by the random search were the same as those in DANAS. The search procedures of the random search and DANAS are visualized in Figure 4 and Figure 5, respectively.As shown in Figure 4 regarding the random search, there were 57 networks with parameter numbers exceeding 2.5 million and 79 networks with validation accuracies exceeding 60%.As shown in Figure 5 regarding DANAS, there were 32 networks with parameter numbers exceeding 2.5 million and 140 networks with validation accuracies exceeding 60%, and one of them was chosen as the optimal network according to step 5 in Figure 2. The optimal network is highlighted by a yellow star in Figure 5 and its normal cell is depicted in Figure 6; its reduction cell was simply a max pooling with a 3-by-3 kernel.The detailed network structure based on the normal cell shown in Figure 6 is illustrated in Figure 7, which shows how the data flowed through the normal and the reduction cells. The connections between cells are denoted by arrows. In Figure 7, cells labeled “normal cell i−4”, “normal cell i−3” …“normal cell i+1” correspond to cells labeled “Cell i−4”, “Cell i−3” …“Cell i+1” in Figure 6. If we rotate Figure 6 clockwise by 90°, then cell labels and arrow colors in Figure 6 will match labels and arrow colors in Figure 7. For instance, yellow arrows between “cell i−1” and “normal cell” in Figure 6 correspond to the yellow arrow between “3×3 max pool” and “normal cell i” in Figure 7, purple arrows between “cell i−2” and “normal cell” in Figure 6 correspond to the purple arrow between “normal cell i−2” and “normal cell i” in Figure 7, and so forth. For each normal cell shown in Figure 7, its inputs are signified by “a direct arrow running from the previous cell” and “three curved arrows running from another three previous cells”, and each arrow in Figure 7 corresponds to a group of arrows with the same color in Figure 6.The input channels of the normal and reduction cells were fixed to 20 and 40, respectively. The output channel of the reduction cell was fixed to 40, and the output channel of the normal cell was automatically determined by its operations. The fixed channel numbers served as element-wise additions within blocks, i.e., only tensors of the same dimensions could be added element-wise. Therefore, channels of any block input were assumed to be 20. If the input channel differed from this constant, then the input was fed to an additional stack of 1-by-1 convolution, batch normalization and ReLU for changing the channel number to 20. Accordingly, the channel numbers of all block outputs were the same, i.e., 20, due to the fact that no operation within a block affected input data dimensions. Besides the channel numbers of inputs, if an input to a block differed in widths or heights, then all inputs were resized to have the minimal width and height found among the block inputs. Therefore, all block inputs shared the same dimensions, and element-wise additions worked in any block. As shown in Figure 6, the cell output was obtained by concatenating block outputs, which required that outputs to concatenate had the same width and height. If the outputs differed in width or height, then they were resized to the maximal width and height found among outputs to concatenate. The number of cell output channels could thus be easily derived by counting the number of outputs to concatenate, e.g., for the normal cell in Figure 6, its output channel number was 60=3×20, i.e., three block outputs were concatenated to yield the cell output. For reduction cell, since the pooling layer only halved the width and height dimensions of inputs, the output channel was the same as the input channel, i.e., 40. If inputs of a reduction cell had different channel numbers other than 40, then the inputs were fed to the channel-changing stack the same as the normal cell. All convolutions in normal cells had strides set to 1 and paddings set to 1 or 2, respectively, for 3×3 or 5×5 convolutions. All poolings had strides set to 2 and paddings set to 1 or 2, respectively, for 3×3 or 5×5 poolings.As shown in Figure 7, the output of the last normal cell was fed to a global average, i.e., a w×h average pooling where w and h refer to the input width and height, respectively. Here, a w×h×c tensor was pooled to c scalars via the global average where c denotes the class number. If the input to the global average had channels other than c channels, then the input was fed to an additional 1-by-1 convolution of stride set to 1 and padding set to 0 before the input was fed to the global average. The results of the network shown in Figure 7 are illustrated in Table 2, and the best accuracy within each row is highlighted by bold texts.As shown in Table 2, although the parameter number of the optimal network discovered by DANAS was small, the average test accuracy associated with DANAS was the third best of the compared networks. However, there were eight species accuracies in DANAS that were the best (bold digits in Table 2) compared to other networks, and there were eight best species accuracies in Resnet-18, which demonstrated the best average test accuracy.There were 155 images misclassified by DANAS. Among all misclassifications, 78 were color images and the rest were night-vision images, i.e., about half misclassified images were night-vision images. The image samples of typical misclassifications are illustrated in Figure 8, i.e., the partial animal body in the left sample, the small region occupied by the animal in the middle left sample, and visually similar animals in the right and the middle right samples.Among all misclassifications, about 64% (99 samples) were misclassified due to the visual similarity of animals, and these misclassifications mainly originated from the deer and canine species. Samples of deer and canine misclassifications are shown in Figure 9. The misclassifications were mainly made among red deer (29 samples) and red fox (23 samples). For red deer samples, 14 samples were grayscale images without colors (the left sample in Figure 9), and the remaining color samples always contained red deer whose heads were obscured due to camera view limitations, body orientations (the middle-left sample in Figure 9), etc. For red fox samples, 11 samples were grayscale images (the middle-right sample in Figure 9), and the remaining color samples always contained foxes occupying small image regions (the right sample in Figure 9).Samples of misclassifications other than deer are shown in Figure 10. The misclassifications were made among bobcat, cougar, coyote, moose, etc., due to reasons similar to those of the deer and red fox misclassifications. Additionally, misclassification samples only containing animal heads are shown in Figure 10.3.2. Search and Test on MCTIThe search on MCTI consumed roughly 62 hours and found a network with 1.43 million parameters. The search performance was compared with a random search whose configuration was the same as the one introduced in the previous section. The search procedures of DANAS and the random search are visualized in Figure 11 and Figure 12, respectively.As shown in Figure 11 regarding the random search, there were 66 networks with parameter numbers exceeding 2.5 million (62 points on the right of vertical line at 2.5 in the figure; four points are not shown due to limited space) and 16 networks with validation accuracies exceeding 50%.As shown in Figure 12 regarding DANAS, there were 15 networks with parameter numbers exceeding 2.5 million (13 points on the right of vertical line at 2.5 in the figure; two are not shown due to limited space) and 93 networks with validation accuracies exceeding 50%; one of them was chosen as the optimal network according to step 5 in Figure 2. The optimal network is highlighted by a yellow star in Figure 12, its normal cell is depicted in Figure 13; its reduction cell was simply a max pooling with a 3-by-3 kernel.The network structure based on the cell in Figure 13 was the same as the one shown in Figure 7 because normal cells found on both NACTI-a and MCTI involved all previous cells, and the pipeline in Figure 7 illustrates how data flowed at the cell level (in contrast to the data flow at the block level that is shown in Figure 6 and Figure 13). The test results are shown in Table 3, and the best accuracy within each row is highlighted by bold texts.As shown in Table 3, the parameter number of the optimal network discovered by DANAS was small, and the average test accuracy associated with DANAS was the best throughout the networks in comparison. There were 167 images misclassified by DANAS. Among all misclassifications, 45 were color images, and the rest were grayscale images. Samples of typical misclassifications are shown in Figure 14, i.e., vagueness due to dirty camera lens in the left sample, similar backgrounds and species in the middle samples, and partial animal body in the right sample.3.3. Tests on Jetson X2The optimal networks discovered by DANAS with the MCTI and NACTI-a datasets were tested on the NVIDIA Jetson X2 edge device shown in Figure 15. Because the versions of PyTorch installed in the workstation and the Jetson X2 are different, the format of network weights saved in the workstation was incompatible with Jetson X2. This issue was tackled by loading and resaving weights in Pickle-based format through PyTorch’s built-in function torch.save() with the parameter “_use_new_zipfile_serialization” set to False. The resaved network weights and 224 × 168 test images were copied to Jetson X2 through secure copy protocol as in [44]. The test results are shown in Table 4.As shown in Table 4, the average accuracies on Jetson X2 were 92.91% and 94.31% for NACTI-a and MCTI, respectively. The average accuracies on Jetson X2 were slightly lower than the corresponding accuracies obtained on the workstation, i.e., 92.86% and 94.02% for NACTI-a and MCTI, respectively.3.4. Comparisons between DANAS and other Search MethodsSince comparisons of search methods based on custom-defined search space and various hardware may introduce bias, we compared our method with other methods via Nasbench-201 [39]. Nasbench-201 provides a database and application programming interfaces (APIs) for comparing search methods with the same search space and hardware. In Nasbench-201, all candidate networks in a specific search space were trained, validated and tested on the CIFAR-10, CIFAR-100 and ImageNet-16-120 datasets [39]. The training, validating and testing accuracies were saved in databases and could be programmatically retrieved via an API, a network code encoding the network architecture. There were five operations scattering in three cells, i.e., the first cell containing one operation, the second containing two, and the last containing three. For each operation, there were five options available for sampling, i.e., “nor_conv_3 × 3”, “none”, “nor_conv_1 × 1”, “avg_pool_3 × 3” and “skip_connect” [39]. Nasbench-201 does not distinguish networks of different operation inputs (i.e., for all networks of operations arranged in the same encoding order, only one network is trained, validated and tested on the aforementioned datasets.), so there are 5×52×53=15,625 [39] networks in Nasbench-201, and a sampler tested on Nasbench-201 is restricted to sample operations only. Accordingly, we simplified our sampler and applied Bayesian optimization [21] to automatically find values of the sampler hyper parameters, i.e., the embedding dimension, the hidden unit number, the layer number, and the learning rate were set to 19, 33, 1, and 0.005, respectively. The rest of the configuration was the same as that of DANAS.According to [39], there are two types of search methods tested on Nasbench-201, i.e., methods dependent on or independent of parameter sharing. Parameter sharing often means weights of a newly-sampled network are initialized by using weights from the previously-sampled networks trained on the dataset, so the weights of previously trained networks are not abandoned during the search. In [39], parameter-sharing-dependent methods were repeated three times and other methods were repeated 500 times. For each run of the method independent of parameter sharing, the method continued to run until the simulated training time [39] of its sampled networks reached a predefined limit called time budget, i.e., 12,000 s. [39]. The simulated training time of the sampled network was obtained by adding its training and validation time saved in Nasbench-201.Since our method (DANAS) is independent of parameter sharing, DANAS was tested according to the configuration of search methods independent of parameter sharing, i.e., the search based on DANAS was repeated 500 times and each search automatically stopped once the time budget was reached. Different from methods in [39], our method requires an additional hyper parameter, i.e., the ideal parameter number s∗. This parameter is set to the parameter number of the candidate network with the optimal validation accuracy. Accordingly, DANAS was tested against three datasets available in Nasbench-201, i.e., CIFAR-10 (s∗=0.87 in millions), CIFAR-100 (s∗=0.86 in millions), and ImageNet-16-120 (s∗=1.29 in millions). Because network weights required by parameter-sharing-dependent methods were not available at the time of paper submission, we only tested parameter-sharing-independent methods with Nasbench-201 on our own hardware. Specifically, all search steps except for training, validating and testing sampled networks were conducted on our hardware, and network accuracies and parameter numbers were directly retrieved from Nasbench-201. The configurations of all methods except DANAS were the same as [39]. The results are illustrated in Table 5.As shown in Table 5, five search methods were compared on three benchmark datasets, i.e., CIFAR-10, CIFAR-100 and ImageNet-16-120 [39]. Among methods in comparison, i.e., REA [56], RS [57], REINFORCE [45] and BOHB [58], our method (DANAS) achieved the second best test accuracy on CIFAR-10 and the third best test accuracy on both CIFAR-100 and ImageNet-16-120.4. DiscussionDANAS was evaluated on two datasets, NACTI-a and MCTI. For both datasets, the random searches significantly differed from DANAS in changes of validation accuracy and parameter number over time.In the case of NACTI-a, the number of networks with parameter numbers exceeding 2.5 million in the random search was almost twice that of DANAS, and the number of networks with validation accuracies exceeding 50% in the random search was roughly half that of DANAS. More importantly, the distribution of points from DANAS in Figure 5 illustrates a growing trend towards networks with few parameter numbers and high validation accuracies, i.e., the search tended to find Pareto solutions good for both accuracy and parameter number, while no such trend can be seen in Figure 4 regarding the random search.In the case of MCTI, the ratio between the numbers of networks with 2.5 million parameter numbers or above for random search and DANAS was higher than the case of NACTI-a, i.e., about 4:1, and the ratio between the numbers of networks with validation accuracies exceeding 50% for the random search and DANAS was lower than the case of NACTI-a, i.e., about 1:8. The distribution of points from DANAS in Figure 12 illustrates the same trend as the case of NACTI-a, and the random search showed no such trend, as depicted in Figure 11.The performance of the networks found by DANAS was evaluated by comparing the test accuracies with seven CNNs with parameter numbers ranging from 0.7 to 66.8 million on two datasets, NACTI-a and MCTI. Although the parameter numbers of networks found on both datasets were lower than 1.5 M, the test accuracy was the third best for NACTI-a and the best for MCTI. These results reveal the benefit of designing CNNs with structures highly customized for studied data and used device. Generally, the experimental results confirmed the validity of DANAS.The search efficiency of DANAS was compared with search methods reported in [39] based on Nasbench-201, and the search methods with parameter sharing were retested on our hardware. For all benchmark datasets of Nasbench-201, our method outperformed all parameter-sharing-dependent methods reported in [39] and most of parameter-sharing-independent methods including the random search. Generally, DANAS outperformed NAS methods with parameter sharing and was competitive compared with NAS methods without parameter sharing.5. ConclusionsIn this study, DANAS is proposed to automatically design lightweight CNNs for ecological research powered by camera traps and edge computing. DANAS was developed based on domain knowledge of camera trap images, i.e., the search is conducted on camera trap images whose resolutions are lowered while the original aspect ratios are maintained. Therefore, the data distribution of the original dataset is preserved during the search, so the data distribution difference incurred by using benchmark datasets in traditional NAS is reduced in DANAS. Furthermore, the search in DANAS is guided by a loss function designed based on Witch of Agnesi whose hyper parameter was theoretically derived. In experiments, DANAS was shown to successfully find lightweight networks for two datasets of wildlife camera trap images. The found networks were then trained on a workstation and tested on both the workstation and an edge device. In comparison with CNNs of classical lightweight designs and good performance, the networks found by DANAS had low parameter numbers and competitive test accuracies. Generally, researchers without knowledge of designing CNNs can obtain lightweight CNNs optimized for edge devices through DANAS and thus expand surveillance areas in a cost-effective way.

animals : an open access journal from mdpi

10.3390/ani13101581

PMC10215336

Precision livestock farming has been shown to not only be beneficial to producer profitability and production efficiency, but also in improving social acceptance and sustainability in livestock production. The research illustrated here utilizes an advanced computer vision system (NUtrack) that allows for the individual identification and tracking of pigs in a standard commercial setting. Traits that would otherwise be impractical to measure and therefore absent from selection criteria, such as the activity level, are easily obtained with this system. The data here suggest that progeny activity levels are impacted by sire EBVs for production traits. In the current study, pigs from boars with superior growth and feed intake EBVs were less active and faster growing compared to all other groups. Further work may lead to genetic selection for activity traits to decrease RFI levels and ultimately improve production rates through reduced calorie expenditure. Activity differences observed across sexes may contribute to managerial improvements in feeding barrows vs. gilts. Ultimately, furthering the objective tracking of activity patterns with NUtrack may help producers overcome difficulties that prevent them from maximizing feed efficiency; with potential for additional benefits to both the producer and the animal.

Feed cost accounts for over two-thirds of the variable cost of production. In order to reduce feed costs without sacrificing production numbers, feed efficiency must be improved. Calorie expenditure has been difficult to quantify in the past but is understood to impact residual feed intake (RFI) greatly. The objective of this work was to utilize an advanced computer vision system to evaluate activity levels across sex and sire groups with different expected breeding value combinations for growth and feed intake. A total of 199 pigs from four different sire groups (DNA Genetics Line 600) High Feed Intake/High Growth (HIHG), Low Feed Intake/High Growth (LIHG), High Feed Intake/Low Growth (HILG), and Low Feed Intake/Low Growth (LILG) were utilized at the UNL ENREC farm over 127 days. The NUtrack system allowed for individual monitoring of pigs in group housing to track daily activity traits. In total, HIHG pigs travelled less (p < 0.05; 139 vs. 150 km), spent more time lying (p < 0.05; 2421 vs. 2391 h), and less time eating (p < 0.05; 235 vs. 243 h) when compared to LILG pigs across time. The results suggest variation in activity occurs across the progeny of the sire groups selected to differentiate in growth and feed intake.

1. IntroductionThe variable cost of producing market hogs is largely centered around feed cost, with diet formulation and feed conversion being the two largest contributors. Feed cost has been estimated to account for 60 to 70% of production costs, with efforts to reduce it being at the forefront of competitive systems [1]. Diet cost can be manipulated in many ways, such as utilizing alternative feed ingredients, replacing traditional protein sources with crystalline amino acids, altering feed particle size, and/or through enhanced management of animals and their environment. Altogether, the inclusion of dried distiller grains with solubles (DDGS; a major co-product from the production of ethanol from grain) is one of the most common ways to mitigate feed costs [2]. However, inclusion rates above 15% of the diet can have diminishing returns towards efficiency [3,4]. Feed conversion ratio (FCR) is commonly described by the proportion of average daily feed intake (ADFI):average daily gain (ADG) with improvements being made through genetic selection, management practices, and nutrition. An increase in FCR has traditionally been an area of focus for genetic improvement. Unfortunately, diminishing returns have been observed in response to the intense selection of FCR due to underlying correlations between growth and carcass traits [5]. For example, Godinho et al. [6] reported negative genetic correlations between FCR and ADG (−0.32) and between FCR and protein deposition (−0.47). Additionally, FCR showed positive genetic correlations between backfat depth (BF) (+0.16) and average daily feed intake (ADFI) (+0.54) [6]. Furthermore, roughly a third of the phenotypic variation associated with ADFI is accredited to residual feed intake (RFI) [7,8]), making the accurate prediction of growth performance and potential genetic selection difficult. Residual feed intake is defined as the feed consumed above expected requirements for production and maintenance [9]. Genetic selection for lower RFI has been emphasized to create offspring that consume less feed without sacrificing growth [10]. Selection for increased feed efficiency based on RFI results has been shown to improve efficiency in the grow-finish stages of production but is suggested to be combined with the growth rate in the selection scheme in order to be applicable to producers [11]. Without this integral step, producers would most likely observe decreases in growth performance due to unfavorable associations. Factors affecting RFI include digestion efficiency, metabolism, maintenance, and activity. Each of these factors has been shown to have a high level of variability among individuals [12,13]. Specifically, maintenance represents a significant portion of the daily energy intake in a pig, with roughly 34% of the daily energy intake of a 70 kg pig being directed to maintenance [14]. Typically, the activity component is disregarded due to the lack of technology to objectively measure it. To be more precise in estimating expected feed intake and to more accurately select for RFI associated traits, further understanding is needed in the area of calorie expenditure due to activity patterns.In the past, swine production systems relied solely on human observation for monitoring behavior and activity trends within group-housed animals, with activity being a secondary concern in most circumstances [15]. Most of these observational times are dedicated to health and welfare, with caretakers focusing on behaviors that can be indicative of injury, sickness, and/or stress. It has been estimated that in modern commercial farms, human observation per animal each day can be as low as only a few seconds [16]. Even if more human interaction/labor were dedicated to observing general behavior patterns that may indicate activity levels, the presence of a human can significantly impact the adaptive behavior of the pigs [17]. Furthermore, human observation is often heavily subjective, with technicians varying in their observations. Detection of changes in behavior and activity patterns of individual pigs housed in groups is crucial to maintaining animal health and wellbeing standards. Such changes can be indicative of internal/external stressors impacting the pig. Animal-based indicators offer the most centralized measure of animal wellbeing [18]. Rapid detection of such indicators may lead to faster response time in treatment of unwell animals.One developing solution to overcome the constraints of human observation is the progression of precision livestock technology (PLT). A primary objective with PLT is to develop a real-time/on-line system that has capabilities to maintain the individual animal ID, track traits of interest on an individual basis, and provide accurate predictions of variable change [19]. Not only will PLT allow producers to more accurately measure and assess key traits of interest, PLT can be used to improve the health and well-being of the herd. Advancements have shown that such technologies may be able to identify an animal in need of treatment prior to human observation. In the beginning, attempts to solve these PLT goals included attaching a wide array of electronic devices to individual animals, such as radio-frequency identification (RFID) collars and tags [20,21,22]. Unfortunately, these devices not only require significant maintenance and financial commitment to utilize, but also can be invasive to the animal, creating a welfare concern. Thus, recent approaches are gravitating toward non-invasive, vision-centered methods [23]. Recent developments in advanced computer vision systems, such as NUtrack (NUtrack Livestock Monitoring), now allow for activity traits to be quantified and tracked individually in a traditional commercial setting [24]. NUtrack is a fully convolution machine learning program for the long-term location and activity monitoring of individual pigs [17]. Therefore, the aim of this study was to utilize NUtrack to objectively measure nursery and finisher activity levels of pigs from sires with different estimated breeding value (EBV) combinations for growth and feed intake. The EBV is the calculation of the animal’s genetic worth based on performance records of itself and other relatives and is reported as a deviation from the population mean [25]. Additional objectives included determining variation across activity traits of all animals as well as evaluating the impact of sex on performance and activity traits.2. Materials and MethodsAll procedures involving the use of animals were approved by the University of Nebraska Institutional Animal Care and Use Committee protocol number 2089. A total of 25 litters [Duroc x (York x Landrace)] were utilized at the Eastern Nebraska, Extension and Education (ENREC) swine farm (May 2021). All litters were sired by Line 600 boars (Duroc terminal line boars) from DNA Genetics (Columbus, NE, USA) that had previously met selection criteria for use in commercial boar studs. A total of 287 duroc boars were available for use following selection at the boar stud. Additionally, these boars were actively being collected and used in commercial production. Sires were subdivided into one of four different sire groups based on their EBVs for average daily gain (ADG) and feed intake (FI) with the groups named as follows: High Feed Intake/High Growth (HIHG), Low Feed Intake/High Growth (LIHG), High Feed Intake/Low Growth (HILG), and Low Feed Intake/Low Growth (LILG). Each sire group was composed of pooled semen from 8 boars and boars were placed into groups based on their EBVs for feed intake and then growth. The resulting differentiation of high vs. low amongst groups shall be looked at in regard to intake firstly, and then growth secondly within intake, with the understanding that the categorization between “High” and “Low” across groups may be somewhat dissimilar. The averages of all boars available for selection for growth and feed intake EBVs were 32.2 and 16.0, respectively. The average EBV for growth and feed intake for each group can be found in Table 1.One day post-partum, all piglets received individual identification via barcoded ear-tags. Prior to weaning, 192 pigs were selected for the trial to assess activity for both the nursery and finisher stages. Inclusion criteria included viability level, sire group, and sex. Twenty-four males and 24 females were selected per sire group. Pigs to serve as potential replacements were also identified prior to weaning, however these animals were placed in a separate nursery without the NUtrack system and thus without nursery activity data. Nursery pens (n = 12) in a single nursery room were stocked after the collection of individual weaning weights, with two barrows and two gilts per sire group represented per pen (16 pigs/pen; 0.42 m2/pig). Random allocation was utilized in placing pigs in their respective pens. Prior to entering the pen, pigs were given an individual ear tag (NUtrack tags) that was unique from pen contemporaries in either its letter/number and/or color combination [26]. Pigs were individually weighed and transferred from nursery to finisher rooms (n = 3 rooms, 8 pens per room) after 42 days in the nursery. Each room shared like environments. Each nursery pen was split evenly into two groups upon entry to the finisher, with one barrow and one gilt of each sire group being represented per pen (n = 8; 0.92 m2/pig). Room and pen allocation per nursery group to finisher pens was also random. To keep pens balanced for pigs/pen at the transition to the finisher, 7 pigs that were preselected as replacements at weaning were utilized to substitute for pigs that died in the nursery or were not thrifty enough to be placed into the finisher. Replacement pigs were selected to maintain the sex and sire group of the fallout pig. These animals were eventually disregarded in the compiled activity data due to their lack of nursery tracking. Pigs were off-tested after 85 days in the finisher, making the total time under analysis 127 days. At off-test, individual weights along with backfat depth (BF) and loin eye area (LEA) were recorded utilizing an ExaGo veterinary ultrasound scanner (IMV Imaging, Rochester, MN). Feed was supplied on an ad libitum basis using a standard multi-hole feeder (n = 8 holes in nursery and 4 holes in finisher) thus not allowing for individual feed intake records. Pen feed intake was not recorded, as pigs from all sire groups were represented in each pen evenly.The NUtrack system was designed specifically for animals living in fixed group-housing environments via the multi-object tracking method [27]. In both the nursery and finisher rooms, each pen was equipped with a Lorex 4K Ultra HD (Lorex Technology, Inc., Markham, ON, Canada) camera mounted from the ceiling above the center of the pen, along with two infrared lights (placed on both sides of the camera) to allow for tracking to take place at night. Each camera was positioned so the field of view included the entire living space. Cameras were hardwired back to a Network Video Recorder (NVR; Lorex Technology, Inc.), which had internet capabilities to allow for the video feed to be pushed out for data processing via an ftp connection. This method of long-term tracking of individual animals within a group-housed setting is feasible via deep convolutional neural networks. These networks locate individual targets (animals) within a fixed living space and classify their identity [27]. This detection utilizes deep learning from annotated images, in which it joins body parts (left ear, right ear, shoulder, and tail) into “instances” via part association vectors [27]. Additionally, activity tracking is processed via the Bayesian multi-object tracking method. Utilizing frame-to-frame movement probabilities with images being captured at 5 frames per second, the probabilities of individual identification (by the NUtrack ear tags) can be assessed. Across a variety of environments, this method has shown to be over 90% accurate in maintaining individual identification [27]. In short, NUtrack does not classify pig behavior, instead it tracks activity and body posture determined by body part position and change in location based on frame-to-frame movement analysis. Currently, activities tracked on an individual pig basis with NUtrack include standing time, sitting time, lying time (sternal and lateral), time spent engaged at the feeder, velocity when walking, rotations, and distance traveled. Time spent engaged at the feeder tracks the amount of time an individual’s head remains positioned appropriately for eating within the designated area surrounding the feeder. Activity traits used for analysis included standing time, lying time, and distance travelled. All other traits not further discussed in this manuscript were not significantly associated with traits included in this study’s objectives. Variables were analyzed on a per day basis, with the periods considered being those of nursery, finisher, and cumulative. Cumulative totals included individual pig as the unit of replication. Pigs in this analysis were of the same genetics and housed in the same rooms/pens with the same lighting and stocking density in which the NUtrack system was originally trained on. Furthermore, NUtrack has been shown to show greater specificity and sensitivity than trained human observers while providing continuous coverage and analysis of pig activity [28] that is not feasible for human observers to accomplish in a study of this scale.RStudio v. 1.1.456 was used for data editing and analysis [29]. Together, the lm() and emmeans() functions were used to derive estimated marginal means. Analysis of variance (ANOVA) testing was conducted utilizing individual pig data as the unit of measure with the use of the summary() function. Pigs that suffered mortality, lameness, and/or ruptures were removed from the data set. A total of 175 pigs were available for analysis after editing. A decrease in observations was the result of mortality, removal from pen for health/lameness, and non-sensical outliers such as when it was confirmed that a pig had died the day before it was noted by farm staff, as the death occurred after they had exited out for the day. Video data was disregarded for the day of entry to nursery as well as the day of transition from nursey to finisher due to incomplete 24-h time frames in their respective pen; although future investigation into these time periods may prove to be informative. Categorical fixed effects for activity traits included sire group, sex, and location (room and pen). Average daily gain was calculated as the difference in off-test weight from weaning weight divided by 127 days. The analysis for nursery growth included the categorical fixed effects for sire group and sex with the covariate of weaning weight. The analysis for finisher growth included the categorical fixed effects for sire group and sex with the covariate of nursery exit weight. For carcass traits, the analysis included the categorical fixed effects of sire group and sex with the covariate of off-test weight. No carcass data was available for the investigation into carcass quality given the activity levels and sire group.3. ResultsDuring the nursery phase, HIHG pigs travelled less (p < 0.05) and spent more time lying per day (p < 0.05) compared to the other three groups. HIHG pigs spent more time at the feeder per day compared to HILG and LIHG pigs (p < 0.05) but this was less (p < 0.05) than that observed in both HILG and LIHG pigs. Furthermore, HILG pigs travelled more per day (p < 0.05) compared to all other groups. The estimated marginal means activity traits in the nursery can be found in Table 2.Throughout the finisher stage, HIHG pigs travelled less per day (p < 0.05) compared to all groups and spent more time lying per day (p < 0.05) and spent less time engaged at the feeder (p < 0.05) compared to the LIHG and LILG groups. Conversely, LILG pigs travelled the most (p < 0.05), spent the least amount of time lying down (p < 0.05), and spent the most time engaged at the feeder (p < 0.05) out of all groups. Estimated marginal means activity traits in the finisher can be found in Table 3.In total, when combining nursery and finisher stages, HIHG pigs travelled less (p < 0.05) and spent more time lying (p < 0.05) than the LIHG and LILG groups as well as spent less time engaged at the feeder (p < 0.05), compared to all other groups. Contrarily, LILG pigs travelled more (p < 0.05) than all other groups throughout both stages. The estimated marginal means for cumulative totals for both stages combined can be found in Table 4.When looking at the performance data across groups, total ADG was the highest (p < 0.05) in HIHG pigs and the lowest (p < 0.05) in LILG pigs; whereas HILG and LIHG pigs were similar. HIHG pigs had the smallest (p < 0.05) LEA, whereas LILG pigs had the largest (p < 0.05) LEA; and HILG and LIHG pigs were intermediate in comparison. LILG pigs had the least (p < 0.05) amount of backfat, whereas LIHG pigs had the most (p < 0.05) backfat and HIHG and HILG were the same. The results showing the ADG, LEA, and BF estimated marginal means can be found in Table 5.Activity and performance averages between the sexes were also calculated. Barrows had a greater (p < 0.01) total ADG and more (p < 0.01) BF at off-test, yet gilts had a larger (p < 0.01) LEA, as expected. In terms of activity, gilts travelled more (p < 0.01) throughout the day and spent less (p < 0.01) time engaged at the feeder compared to barrows. No significant differences were seen between the sexes for time spent lying per day. The estimated marginal means for performance and activity traits for each sex can be found in Table 6.4. DiscussionIn the current study, we utilized an advanced computer vision system (NUtrack) to detect activity differences in pigs from sires with different EBV combinations for growth and feed intake in the hopes of understanding if these EBVs are predictive of activity differences. The sire groups were composed of boars with contrasting EBVs for both growth and feed intake; two traits commonly measured and selected upon to increase performance of terminal animals. These results show clear differences in activity between the most extreme groups for both feed intake and growth (HIHG vs. LILG) but showed inconsistencies when comparing the intermediate groups (HILG vs. LIHG). A more comprehensive comparison among these groups could be made by including a larger sample size, as the EBV difference achievable in this population between these groups for ADG was quite small. Alternatively, an unselected population with greater genetic variation could be used to form EBV groups. The sires used for this study were boars that had previously met selection criteria for use as an AI sire; thus, the variation of performance traits was decreased. Furthermore, some overlap was present in the categorization of sire groups for growth and feed intake, potentially hindering the resulting differences amongst them. Together, further work is needed to properly delineate the association between individual EBVs and the corresponding activity trait(s).Pigs sired by HIHG boars spent more time lying, travel less, and spend less time engaged at the feeder on average compared to pigs from most other sire groups. Differences across sire groups were largely consistent between the nursery to the finisher stage, with some variation explained by the larger number of days and the changing rate of growth and feed intake. In total, HIHG pigs walked 10.92 km less over both phases, as well as spent 29.63 more hours lying down and 8.46 less hours engaged at the feeder compared to LILG pigs. It may appear counter intuitive that HIHG pigs spend less time at the feeder than other groups; however, it has been demonstrated that pigs with a higher intake per second of feed are faster growing and have a greater deposition of adipose tissue [30]. These differences, especially the difference in distance traveled and time spent lying down, can be assumed to impact calorie expenditure and may explain a portion of the variation observed in performance. the non-independence amongst activity traits such as distance travelled and lying time should also be noted. If an animal spends more time lying down, there will be less time available for compiling movement. This time allowance for activity may explain some of the variation. Differences observed in time spent at the feeder across sire groups may be due to biological effects. Literature has shown that pigs will spend up to half of their active time on “foraging” and eating, with significant variability amongst individuals [31]. The rate at which individuals were eating is most likely the largest factor contributing to the variation observed within and across groups. Additionally, the frequency in which pigs consumed feed may impact carcass and other performance traits [32]. Activity differences were less clear between HILG and LIHG groups. Differences between these two groups were often small/non-significant. Perhaps the smaller differences in mean EBVs, most notably in growth, can explain some of the similarities in the activity patterns. The variability present across and between sire groups may represent an opportunity for genetic selection of an individual trait or an index of activity traits in the future. As individual feed intake data was not collected and pigs of all groups were equally represented in each pen, we could not determine differences in feed efficiency among these groups nor individuals within groups.The sample size of the study is small relative to traditional studies involving EBVs, yet the difference in growth between the HIHG (0.88 kg/d) and LILG (0.81 kg/d) groups of 0.07 kg/d was half of the difference in the intake EBVs across respective sire groups. This is what was predicted from the difference in EBV for growth, as half of the progeny’s breeding value for any given trait is assumed to come from the sire and half from the dam. HIHG pigs also had more (p < 0.05) backfat at off-test (1.48 vs. 1.42 cm), which suggests that the activity levels of the pigs could partially explain fat and lean deposition in growing pigs. A larger deposition of carcass fat can be explained by the rate of fat synthesis being greater than the rate of fat utilization [33]. Significant differences were also observed for LEA, with LILG pigs (52.7 cm2) being greater (p < 0.05) than HIHG pigs (49.66 cm2). This agrees with previous work, as physical activity has been shown to positively impact the leanness of carcass composition [34]. No difference was seen in ADG between HILG- and LIHG-sired pigs which again may be attributed to the small difference in the growth EBV.The comparison of performance between barrows and gilts coincides with what is commonly seen in the industry. The results here show that barrows have greater ADG, more backfat, and a smaller LEA on average compared to gilts. This agrees with the review done by Kansas State University [35] of 34 different trials showing gilts having a 5.9% lower ADG, 11.7% less backfat and a 4.5% increase in lean percentage compared to barrows. However, activity differences associated between sexes have yet to be quantified or managed. These results show that gilts are more active (50.9 m/d) and spend less time at the feeder (−7.1 min/day) compared to barrows over both phases of production. No meaningful difference was detected between lying time per day (2 min/d). As mentioned earlier, individual feed intake data is needed to accurately estimate feed efficiency between sexes. However, these results show variability among the sexes, again representing opportunity for selection in the future and the potential to create different management strategies such as providing more feeder space per pig for barrows versus gilts given their increased time engaged at the feeder. Overall, a more expansive study is needed to better understand the significance of all associations mentioned above.5. ConclusionsCalorie expenditure due to activity in market-driven animals has been difficult to quantify in the past due to the absence of technology to objectively measure it in a commercial setting, but it is understood to impact feed efficiency. The results herein suggest that the progeny activity level is impacted by the sire EBVs for production traits and could be detected in the progeny of boars already selected for use in commercial production where the genetic variation is less than the full population’s genetic variation. In the current study, pigs from boars with superior growth and feed intake EBVs were less active and faster growing compared to their contemporaries. In addition, pigs sired by boars with the lowest EBVs for growth and feed intake were the most active and slowest growing. Further work may lead to genetic selection for activity traits to decrease RFI levels and ultimately improve production rates through reduced caloric expenditure. In addition, the results show differences amongst sexes in terms of activity and growth, with barrows being less active and faster growing when compared to gilts. Further work in the area of activity and behavior differences amongst sexes may result in different management strategies for feeding and housing of different sexes. Ultimately, furthering the objective tracking of activity patterns with precision livestock farming systems, such as NUtrack, may help producers increase the rate of genetic gain for feed efficiency and contribute to improved pig management.

animals : an open access journal from mdpi

10.3390/ani11092624

PMC8466178

Although cryopreservation techniques for use in bird species have advanced greatly over recent decades, especially in relation to domestic species, major gaps in our knowledge and technical capacities remain due to the complexity of the process and the unique particularities of sperm from different species. The hatchability of chicks is the decisive parameter that demonstrates the quality of a frozen–thawed sperm. Since very little information has been published about the common pheasant, a total of six artificial inseminations (AIs) were performed at 3–4-day intervals with doses of 35 × 106 of normal live thawed sperm on a total of 40 females. The inseminated sperm were collected from pheasants fed either a basal diet or an antioxidant-enriched diet and were then processed using a pellet freezing–thawing method, in which dimethylacetamide was used as a cryoprotectant. Regardless of the male birds’ dietary group, the resulting fertility rate from frozen–thawed sperm was approx. 30%, with 8–9 chicks hatching for every 100 eggs incubated.

A widely used approach to preserving genetic diversity in birds involves the cryopreservation of semen. In this process, cells are subjected to physical and chemical stresses, but not all cell species respond equally. Many studies have been published on the freezing–thawing of sperm cells from a wide variety of domestic and wild species, on issues ranging from the sperm quality to different protocols, fertilisation success rates, etc. Nevertheless, very little information is available on the common pheasant. To fill this gap, the aim of this study was to describe the pheasant semen collection method, evaluate some qualitative parameters of sperm from males fed an antioxidant-enriched diet, and to test the in vivo fertilising capacity of the cryo-preserved semen. The freezing protocol employed involved pellets thawed by the hotplate method. Dimethylacetamide was used as a cryoprotectant at a final concentration of 6%. A total of six AIs were performed at 3-4-day intervals on a total of 40 females with doses of 35 × 106 of normal live thawed sperm. Males receiving the enriched diet produce more abundant and concentrated ejaculates. Freeze–thawed sperm lost 85% of their initial mobility, and diet influenced neither sperm mobility nor viability. The enriched diet did improve the number of normal freeze–thawed cells and was associated with a lower sperm fracture incidence. Regardless of the dietary group, frozen–thawed sperm resulted in a fertility rate of 30%, with 8-9 chicks hatching for every 100 eggs incubated.

1. IntroductionSince the first successful attempt to cryopreserve chicken sperm more than seventy years ago, many studies have been performed to improve this reproductive technology [1]. Sperm cryopreservation has been investigated in many domestic bird species, including, for example, turkeys, ducks [2], geese [3], guinea fowl [4], quail [5] and pigeons [6]. The literature now documents sperm cryopreservation protocols for a number of bird species used in animal farming as well as fancy species [7,8]. The success of the cryopreservation of bird semen depends on several factors, including bird species, breed, collection method, dilution and diluent, sperm concentration, cooling rate, cryoprotectant and the freezing and thawing methods employed [7,8].Bird sperm are highly differentiated cells, the job of which it is to fertilise the oocyte and pass on their genetic information. They possess a low cytoplasm content and a large cellular membrane [9]. The cell membrane plays an important role in sperm maturation and in the egg fertilisation process, in which the cell undergoes a series of biochemical and functional changes [10]. Indeed, the process of fertilisation requires a membrane with specific adaptations at its points of fusion with the oolemma and in its sperm organelles such as, for example, the acrosome, to prepare it for fertilisation [9]. The sperm membrane is characterised by its permeability, fluidity and lipid content [9,10,11], which is high in phospholipids and sterols [12], mainly in polyunsaturated fatty acids [13]. This feature renders the cell membrane of avian sperm highly vulnerable to lysis and lipid peroxidation. Avian semen naturally contains antioxidants and enzymatic defences that protect the cell structure [14]. Stress factors acting on semen during manipulation, liquid storage or cryopreservation endanger and compromise sperm integrity. Therefore, it is necessary to offer the bird additional molecules and/or mineral trace elements to sustain a highly efficient antioxidant system, which consequently diminishes the damage incurred during the freezing/thawing process and artificial insemination (AI) [10]. Strategies for protecting sperm against excessive reactive oxygen species (ROS) production during cryopreservation can be carried out by dietary means [13,15,16] or by supplementing the freezing medium with antioxidants [17,18,19], similarly to protocols used for mammals, for example, rabbit [20] or bull [21].Cell antioxidant systems include natural fat-soluble antioxidants such as vitamin E, and antioxidant enzymes are also involved, such as superoxide dismutase and glutathione peroxidase; selenium is an integral cofactor of the latter enzyme [13]. Moreover, the interaction between vitamin E and selenium may increase glutathione peroxidase production [13,22]. It is well established that dietary supplementation with selenium, vitamin E and carotenoids can modulate the antioxidant defences in poultry [13].Among game birds, the common pheasant represents one of the most popular and economically important species, kept mainly for hunting purposes. Moreover, its eggs and meat offer a valuable food source to the human diet [23,24]. Breeding programs are followed either to increase the productive performance of populations bred for their meat or for hunting purposes, and to maintain specific traits, mainly those found in wild populations for genetic conservation and release into the wild [23]. Sperm cryopreservation provides a key tool for supporting the execution of these programs by helping to spread genetic diversity and through its contribution to the development of reproductive technologies [9,25]. To date, very few studies have been reported on the cryopreservation of semen from the common pheasant [26,27,28].Semen collection in birds is traditionally performed by the dorso-abdominal massage technique, as first described more than 80 years ago [29]. The collection method largely influences the quality of the collected semen. The ability to avoid contaminants in the semen and to minimise handling stress to the birds are two fundamental aspects of a collection technique [30,31,32,33,34]. Therefore, semen collection should be considered the first essential factor for the cryopreservation process and successful AIs. This aspect was recently highlighted in relation to chickens, for which the number of penetrating sperm, evaluated using the inner perivitelline membrane penetration assay, was unexpectedly greater in frozen–thawed sperm compared with fresh sperm, indicating a better performance in the cryopreserved sperm [35]. These authors attributed this result to the benefits obtained from the training of male: while the fresh sperm were obtained from males who were also allowed to mate naturally, and only occasionally subjected to massage collection, the cryopreserved sperm came from roosters kept exclusively as donors, and were thus habituated to the massage collection [35]. The aim of the present study was to describe the pheasant semen collection technique, to evaluate some qualitative parameters of sperm from males fed an antioxidant-enriched diet, and to test the fertilising capacity of the cryo-preserved semen in vivo.2. Materials and Methods2.1. ReagentsAll chemicals were purchased from Sigma Aldrich (Milano, Italy), with the exception of Accudenz, a cell separation media that was purchased from Accurate Chemical and Scientific Corp. (Westbury, NY, USA).2.2. BirdsThirty male common pheasants (Phasianus colchicus mongolicus) were housed for their first reproductive season in a peaceful location far from public roads in open-air aviaries containing perches; each male was kept alone in a 6 m2 allocated space [36]. Birds were randomly divided into 2 groups and fed ad libitum one of two diets which differed according to their α-tocopheril-acetate (Vit. E) and selenium contents: the control group (CON) diet constituted a basal commercial feed (11.51 MJ/kg of M.E., 19% of C.P., 1% fish oil; plus 40 mg α-tocopheril-acetate and 0.1 mg selenomethionine per kg of feed); the antioxidant-enriched group (E-Se) diet was fed the same basal commercial feed supplemented with additional α-tocopheril-acetate and selenium to a final concentration of 200 mg α-tocopheril-acetate and 0.3 mg selenomethionine per kg of feed. All pheasants received the diets for one month. Semen was collected twelve times (every 3–4 days) over a 39-day period (May–June).2.3. Semen Collection and Semen Quality ParametersSemen was collected by means of the dorso-abdominal massage technique by two operators in the following way: one operator captures the bird with a net to minimise potential stress [36]. The operator then assumes a sitting position, placing the bird with its abdomen on his lap, such that the bird’s legs are dangling downward. At this point, the second operator gently restrains the bird by placing the animal’s legs through the circle of space created by touching his thumb and middle finger together. In this way, the bird is able to move forward and backwards its legs. The first operator can then use both hands to stimulate the bird. The massage initiates from the bird’s back, stroking the bird with one hand starting from behind the level of wing attachment and moving backwards towards the base of the tail. This movement is performed 3 or 4 times. The bird responds immediately by lifting his tail. This movement is repeated a second time, but with the second hand under the cloaca. This is repeated 2 to 3 times, at this point the bird everts the cloaca and ejaculates voluntarily. The only intervention made by the handler is to prolong the bird’s eversion for a few seconds to facilitate the second operator who collects the semen. The second operator collects the semen by means of an aspirating device. During this process, the bird can move its wings and legs as he desires. The objective is to permit the bird as much liberty of movement as possible, such that he feels comfortable and not constrained.Semen was collected from each male and directly introduced into the collection tube with 50 µL of pre-freezing Lake’s diluent [37] plus 50 mM glycine diluent (except in tubes destined for pH measurements). Immediately thereafter, it was placed inside a portable refrigerator set to 18 °C until all donations had been collected. Within 40 min from collection, clean ejaculates were chosen, and semen was pooled according to the diet into two samples (CON and E-Se). Ejaculate volume was assessed by weighing the tubes, before and after collection (Sartorius BL 150S, ±0.001 g). The pH was measured for three random samples without diluent for three non-consecutive days of collection (Hamilton 238140, Hanna Instruments, Italy). Sperm concentration was assessed in duplicate using a Bürker–Türk counting chamber (in a 5% formalin and 0.9% NaCl solution). The sperm viability percentage and sperm morphology were evaluated in triplicate in samples of 500 cells using the eosin-nigrosin staining technique [38]. Viable cells did not stain at all, whereas cells considered dead appeared totally or partially stained pink. The sperm viability percentage was calculated relative to total counted sperm (V-CS). From the live sperm sample, abnormal cells were grouped according to anomalies of the head or the tail. The percentage of normal sperm was calculated relative to total live sperm (N-LS). The percentages of abnormal heads (Ah-LS) and abnormal tails (At-LS) were calculated relative to total live sperm. Spermatozoa mobility (SM) was assessed in triplicate using Accudenz methodology [39], which measures the ability of these cells to penetrate a viscous medium at a temperature of 41 °C, and this penetration was measured by spectrophotometry at an absorbance of 550 nm.2.4. Freezing Methodology in PelletsSimultaneously to the assessment of the semen quality parameters of a pool sample (900 µL), pools were divided into 300 µL aliquots and processed for the freezing procedure into pellets [40] adapted for pheasants [28,41]. Briefly, samples were diluted 1:3 (v:v) in pre-freezing Lake’s diluent [37] supplemented with 50 mM glycine, and rapidly cooled inside a device set to −6 °C until the sample temperature indicated 5 °C. Samples were left for 10 min at 5 °C, dimethylacetamide cryoprotectant (DMA) added to obtain a final concentration of 6%, mixed manually for one minute, left to stabilise for 4 min, and then 100 µL semen aliquots were dropped directly into liquid nitrogen. The resulting frozen semen pellets were collected and stored inside cryovials in liquid nitrogen for one year.AIs were performed during the successive pheasant reproductive season. Pellets were melted employing the hotplate method at 75 °C; one at a time, the pellets were placed on an aluminium dish and pushed gently using a micropipette tip in order to collect the melted semen quickly [41].2.5. Thawed Semen QualityImmediately after the semen was thawed, the qualitative parameters were evaluated following the same protocols previously described for fresh semen. The percentage of viable thawed cells was calculated considering the total number of counted thawed sperm as 100% (V-CTS). The mobility results are reported as percentages with respect to fresh sperm cell mobility (SM-FS). The percentage of normal cells was calculated considering the total number of thawed live sperm as 100% (N-LTS). The viability rate, mobility and percentage of normal cells were compared between fresh and frozen–thawed (F–T) sperm. Within the portion of live F–T sperm, cells were grouped according to the presence of lesions of the head or the tail. The percentages of abnormal thawed heads (Ah-LTS) and abnormal thawed tails (At-LTS) were calculated relative to total live thawed sperm. The head injuries identified were: bent, fractured, coiled, swollen-detached, knotted, or headless. The tail injuries identified were: looping, fractured, and coiled. Data are reported as percentages relative to total live thawed sperm (LTS).2.6. Females and Artificial InseminationsForty common female pheasants (Phasianus colchicus mongolicus) in their first reproductive season were housed in a peaceful location far from public roads in open-air sandy aviaries containing perches; each female was provided with a 3.6 m2 space [36]. Two weeks prior to commencing the AIs, all females were fed ad libitum on a breeders commercial feed (11.51 MJ/kg of M.E., 19% of C.P.). The massage method applied to females was substantially the same as described above for male birds, with changes related to the insemination. Additionally, in this case, the eversion of the cloaca was exclusively carried out by the bird; no force was applied by the operator. Vaginal insemination was performed by means of an adapted Gilson pipette (Gilson, Pipetman P200, H23601S64092K) for bird insemination (IMV Technologies, L’Aigle, France). During the release of semen, the female performs cloacal movements, thus taking up the inseminated dose.The fertilising ability of the F–T semen collected the previous year was evaluated through AIs performed on forty females, divided randomly into two groups according to the bird group from which the inseminated pellets were generated: the control group (CON) and α-tocopheril-acetate-selenium group (E-Se). All females received 35 × 106 normal live thawed sperm in each insemination. Inseminations were performed on six days, designated days 1, 2, 6, 9, 13, and 16. Egg fertility was determined on eggs collected from day 2 to day 21 after the first AI. Eggs were weighed daily and set the day after laid. They were candled on day 7 of incubation, and those judged as infertile were broken up for macroscopic examination of the germinal disc. Egg laying, fertility and hatchability were calculated according to the following formulas:Egg laying% = (total laid eggs/females * days) * 100(1) egg fertility% = (total fertile eggs/total incubated eggs) * 100(2) hatchability% = (total hatched chicks/total fertile eggs) * 100(3)2.7. Statistical AnalysisOne way ANOVA was applied by the GLM procedure using SAS software (SAS Studio, v. 3.8) to evaluate the qualitative parameters of fresh and thawed semen and egg parameters. The correlation between the semen qualitative parameters was evaluated by the CORR procedure (SAS Studio, v. 3.8). The statistical model considered the diet as the fixed effect. Percentage data were normalised through √x Arcsine transformation. A p < 0.05 was considered statistically significant.3. ResultsThe sperm qualitative indicators in the fresh and F–T semen from the two pheasant groups fed the control diet (CON group) vs. the diet enriched with α-tocopheril-acetate and selenomethionine (E-Se group) are reported in Table 1. Ejaculate volume and sperm concentration in the fresh semen were positively affected by diet enrichment (p < 0.01). The males in the E-Se group produced more abundant ejaculates, which were a third larger in volume compared with those from the CON group. Moreover, the E-Se group ejaculates had a higher sperm concentration (1.63 × 109 more cells/mL).In the sperm subjected to the F–T procedure, a higher number of normal cells was observed in the E-Se group (+17%), and the rate of abnormal heads and tails was higher in the CON group than in the E-Se group (p < 0.01). The viability rate, mobility, and percentage of normal cells were significantly different between fresh and F–T sperm, as expected (p < 0.01). The number of sperm in F–T samples from the CON group was 40% lower with respect to the number of fresh sperm from these birds, whereas the same comparison in the E-Se group gave a negative difference of 29%. The viability of F–T sperm was 25% in both dietary groups; the reduction in F–T sperm mobility was 84% in the CON group and 82% in the E-Se group with respect to the fresh material. Fractures to tails and heads were the most frequent injuries detected in both groups, and both were significantly higher in the CON group (p < 0.01). Looping tails and bent heads were the next most common injuries detected in both dietary groups.Table 2 reports the correlations between the qualitative parameters in the fresh sperm. The CON group fresh sperm shows positive correlations between pH and sperm concentration (SC; p < 0.01), N-LS (p < 0.01), At-LS (p < 0.01) and Ah-LS (p < 0.01). Positive correlations also exist between SC and N-LS (p < 0.05) and between SC and Ah-LS (p < 0.05), as well as between At-LS and N-LS (p < 0.01) and between Ah-LS and N-LS (p < 0.01) and At-LS (p < 0.01). In the E-Se sperm group, positive correlations are observed between Ah-LS and pH (p < 0.01), N-LS (p < 0.01) and At-LS (p < 0.01). Positive correlations are also observed between At-LS, pH (p < 0.05) and N-LS (p < 0.05).Table 3 reports the correlations between the qualitative parameters in the F–T sperm. In the CON group sperm, a positive correlation was observed between N-LTS and L-CTS (p < 0.05). A negative correlation was observed between At-LTS, L-CTS (p < 0.05) and N-LTS (p < 0.01), as well as between Ah-LTS and N-LTS (p < 0.01). In the E-Se group sperm, a positive correlation was observed between L-CTS and Abs.-F (p < 0.01) and between Ah-LTS and At-LTS (p < 0.01). A negative correlation was observed between N-LTS and At-LTS (p < 0.01) and Ah-LTS (p < 0.01).Table 4 reports the performance of the two pheasant groups inseminated with F–T semen pellets from males fed on the CON or E-Se diet. The fertility percentage is identical in the two groups, reaching 30% in both, and the resulting incubation process output resulted in 8-9 hatched chicks for every 100 eggs incubated.Figure 1 illustrates the evolution of the percentage of fertile eggs following AI with F–T semen. Bell-shaped trends were observed. In both groups, an increase in the percentage of fertile eggs is observed on the day following each AI (indicated by asterisks). After the first and second consecutive AIs, both groups show a peak in fertility rate three days later. Thereafter, peaks are observed the day after AI in the E-Se group, whereas this trend only commenced in the CON group after the fourth AI. Longer fertility persistency is observed in the final bell for the E-Se group. No significant difference was observed in the daily fertility rate between the CON and E-Se groups (p > 0.05).4. DiscussionThe poultry industry has been using the massage technique for semen collection ever since it was first described by Quinn and Burrows [29]. Since its introduction, the technique has been adapted for use in non-domestic birds, and it continues to be the most common method used for semen collection [42]. It has even been reported possible to collect semen from uncooperative birds using this method, but the samples collected tend to be of poorer quality [42]. Whilst we can confirm the last statement from our experience with pheasants and chickens, our policy is to only collect semen from cooperative birds.Here, for the first time, we describe our massage method for the collection of pheasant semen, which differs significantly to the one reported by Quinn and Burrows [29] as well as that by Gee et al. [42], since we never force the male bird to produce semen or the female to evert the cloaca. In fact, by permitting the birds to move as freely as possible, our method permits the collection of an abundance of good quality ejaculate the majority of the time. Considering that the first important step in the achievement of AI success is semen collection, this aspect is of extreme importance. On the contrary, the disadvantage of this technique is that it requires advanced bird hander experience and dexterity.The ejaculate volumes observed in this study are consistent with our previously reported data for male pheasants of the same reproductive age, for which volumes ranged between 106 and 140 µL [36,41,43]. Males fed the E-Se diet in this study produced more abundant ejaculates compared with birds in the CON group (138 vs. 93 µL). This result contradicts our previous findings in which males fed the standard diet produced higher ejaculate volumes [44]. The positive influence of the diet enriched with E-Se on the sperm concentration agrees with the results a previous study [45]. Regarding the sperm concentration, it seems clear that diet is a key factor affecting cell number, as also reported for other species, such as roosters [46,47,48] and ganders, for which changes in diet were also shown to be capable of improving the volume and viability of the ejaculate [49]. Nevertheless, it is important to highlight that other factors (e.g., breeding conditions, health status, diet, environmental conditions, genotype, etc.) may also influence the production of high concentrated ejaculates, as evidenced by studies in which sperm concentrations per mL varied across years: 9.5 × 109 in 2007, 12.5 × 109 in 2008 [36], 10–12 × 109 in 2009 [27,41], 5 × 109 in 2010 [44] and 7.3–8.45 × 109 in 2012 [45]. Comparing fresh semen parameters from birds belonging to the E-Se group in this study (9.11 × 109/mL) with other pheasant species, we can note that the mean sperm concentration is more than 7.6-fold and 1.4-fold smaller in Pucrasia macrolopha and Syrmaticus mikado, respectively [50], and the average ejaculate concentration in the Tragopan caboti [51] is more than 3.2-fold and 3.9-fold smaller compared with our results for the CON and E-Se groups, respectively.The fresh sperm viability percentages observed here lie within the 81–90% range reported in previous studies [27,36,43]. It seems that sperm survival remains quite stable regardless of external factors, whereas sperm concentration is more susceptible to influence by external factors. The sperm viability did not differ between the dietary groups for either fresh or F–T sperm. The viability data for F–T sperm agreed with previous data on common pheasants [43] as well as Polyplectron emphanum [50].No difference was observed in fresh sperm mobility between the dietary groups. Similarly, Marzoni et al. [44] did not observe any differences in sperm motility between pheasants fed diets characterised by the same E-Se ratios (E-40/ Se-0.1 and E-200/ Se-0.3) as used in this trial. Sperm mobility is determined as the capacity of a sperm to accomplish a forward movement through a viscous medium [39], and this movement is measured by changes in the absorbance value. As expected, the freezing process decreases the mobility value, an 85% reduction compared with that achieved by fresh material. In fact, the consequences of oxidative damage occurring during the F–T process are numerous; it has been associated with the disruption of mitochondrial activity, the enhanced efflux of intracellular enzymes and damage to axonemal proteins. All these events result in the loss of the sperm motility [52]. The mobility of the F–T sperm was not affected by the diet. This result was unexpected considering that in the F–T E-Se semen a higher number of normal cells was observed, as was a positive correlation between cell viability and mobility. However, although a more complex scenario obviously occurs in vivo, the reliability of this “mobility test” was confirmed by the in vivo results of this trial since the fertility rates of the two dietary groups were identical.The number of normal sperm cells after the F–T process varied between the dietary groups, with a higher number being detected in E-Se sperm. As expected, the F–T process brought about a large decrease in sperm number relative to the fresh sperm. In chickens, a decrease in the number of normal sperm after in vitro storage was associated with a decrease in the total lipid content of semen [9,53]. This occurs due to the high proportion of polyunsaturated fatty acids, which renders the avian sperm membrane susceptible to lipid peroxidation; thus, a decrease in the concentration of polyunsaturated fatty acids takes place [53]. Studies on the fatty acid composition of pheasant sperm have reported relatively low percentages of PUFAs (19–21%) and total UFAs (37–49%) [54] compared with other avian species such as the chicken (51 and 68%, respectively) [55], and similar levels compared with turkey sperm stored for one hour: 33 and 51%, respectively [56]. Throughout the F–T process, the lipids of sperm membranes and subcellular organelles are affected not only by the lipid peroxidation but also by their specific temperature phase transition, which differs from that of the water [9]. In this study, in the F–T sperm, the higher percentage of normal cells in the E-Se group compared with the control group (54 vs. 37%) might be the result of a more resistant and performing cell membrane. For example, in ruminants, disorders in the sperm cell antioxidant system during cryopreservation and the activation of L-amino-acid oxidase in defective or dead cells contribute towards a higher production of ROS, resulting in sperm plasma membrane damage [52,57,58]. In fact, high amounts of ROS induce the oxidative attack of PUFAs, resulting in the formation of toxic by-products such as malondialdehyde. This lipid assault, in conjunction with the conformational changes of the plasma membrane during the F–T process, contributes to the loss of membrane integrity and the disruption of its permeability properties [52,59].Fractures were the most common injuries detected in the F–T sperm. The antioxidant-enriched diet appeared to exert its protection effect on the head-midpiece and tails of E-Se sperm, as revealed by comparing the damage incurred with that seen in the CON sperm. During the cryopreservation process, the sperm membrane integrity may be compromised due to alteration of the membrane fluidity, changes to the asymmetry of the phospholipid bilayer, and therefore its functional condition [52,59], and this is related to the membrane’s susceptibility to lipid peroxidation [13]. A higher susceptibility to oxidation has been associated with a decrease in cytoplasm volume, which is, at the same time, related to a reduced amount of antioxidant enzymes [60]. In fact, the antioxidant systems of cells include the natural fat-soluble antioxidants, such as vitamin E. Antioxidant enzymes are also involved, such as, for example, glutathione peroxidase (GSH-Px4), and selenium is an integral cofactor of this enzyme [13]. Additionally, the interaction between these two elements (vitamin E and selenium) may increase the production of GSH-Px4 [13,22]. The importance of Se in the diets of male birds was also reported for 14-week-old roosters, in which a dietary deficiency of this element retarded semen production by seven weeks and resulted in low relative testis weights, a high percentage of abnormal sperm in the ejaculates and a higher number of abnormalities in the sperm head midpiece [47]. In a study of male pheasants fed the same basal diet as used in the present study, the n-6/n-3 fatty acid ratio of sperm membranes was found to be lower in the group receiving the E-Se enriched diet (E-200/Se-0.3) [54]. According to another study in the chicken, diets rich in n-3 PUFA are associated with a higher requirement for vitamin E [61]. In the present study, the higher number of abnormal sperm cells in the CON group animals might have correlated with a possible difference in the plasma membrane composition, due to greater protection generated by the higher level of vitamin E.The semen pH observed in this study did not differ between the two dietary groups and was consistent with the values reported in previous reports on pheasants (7.5–8.5) [45,50,51]. Considering pH values previously reported for pheasant semen (7.76–8.7) [27,43] and those in the chicken (6.8–7.85) [62,63], it appears that pheasant semen tends to be more alkaline, while that of the chicken remain slightly acid neutral. Indeed, large variations in pH are not expected considering the crucial nature of this physiological parameter [62]. Consequently, in the preparation of diluents, the pH should specifically be adapted for pheasants. Nevertheless, according to Saint Jalme et al. [50], increasing the diluent alkalinity did not improve sperm viability in a pheasant species characterised by a particularly alkaline semen. In the present study, a high positive correlation was observed between pH and At-LS and between pH and Ah-LS in the fresh sperm of both groups, suggesting that higher pH values render the environment less ideal for these cells, even if a positive correlation between pH and N-LS was seen in the sperm from pheasants from the CON.Studies have reported accelerated lipid peroxidation to occur in abnormal sperm [64]. Furthermore, lipid peroxidation plays a role in regulated cell death [65], and an increase in the pH is related to a rise in the number of dead sperm [66]. In this study, no correlation was observed in the fresh semen between the viability of cells and the pH. Another study using a different pheasant species observed that higher pH values in fresh semen correlated with lower survival rates following the F–T process [50]. The positive correlations observed between the normal and abnormal cells in the fresh sperm of both groups in the present study was an interesting finding, and this correlation became negative in the F–T sperm, emphasizing the effect of the F–T process.The egg weights observed in the present study are consistent with the mean data reported by other authors [67,68]. On the contrary, the mean egg weight in this study (33–34 g) was greater than those found in a previous study in which females were fed a vitamin E-enriched diet (E-200; 31.19–32.25 g) [69]. In natural mating, egg weight has no impact upon the fertility rate, but an effect of egg weight has been reported by other authors on egg hatchability, where eggs heavier than 35.99 g were less likely to hatch [67,68]. In this study, the egg weight had no influence on the fertility or the hatchability.As expected, egg production did not differ between groups, since all females received the same feed. The egg production rate in the present study ranged from 85% to 88% and was higher compared with the 70% reported for females fed on diets enriched with vitamin E over a 4–12-week laying period [69]. This suggests that either the present study was performed at the time of maximum egg production or that the egg-laying performance of these females was actually better. Other authors have reported egg production rates of 64 and 74% in 1- and 2-year-old pheasants, respectively [70].After one year of cryopreservation, the fertilising capacity of the F–T sperm had dropped to 30% in both dietary groups; considering the low insemination doses of only 35 × 106 NLTS at 3–4-day intervals, this outcome is very promising. As expected, a substantial lower fertility rate was observed compared with the 83% obtained in pheasants inseminated once a week with a dose of 150 × 106 fresh sperm, and the 40% fertility rated observed for 24-h stored sperm [71]. Comparable fertility values (26–37%) for F–T sperm have been reported under similar AIs conditions [27,45]. In the Combatiente chicken, doses of 500 × 106 F–T sperm had a reported fertility of just 9.4% [72], whereas 87% fertility has been achieved in the turkey with doses of 400 × 106 F–T sperm [73]. Increasing the sperm dose in the above-described AI protocol for pheasants would probably result in higher fertility percentages. The fertility trend observed after AI was similar in both female groups: the initial increase in fertility rate was gradual but was then followed by a sharper rise. Finally, the decrease in fertility rate tended to be slower for F–T sperm obtained from birds in the E-Se group.Overall, no difference was observed in the reproductive performance of the two female groups inseminated with F–T sperm obtained from CON or E-Se birds. This result is supported by the finding that the mobility of the sperm cells did not differ between the two dietary groups either.5. ConclusionsDietary supplementation with α-tocopheril-acetate and selenomethionine improved the number and integrity of F–T sperm, resulting in a higher number of insemination doses. However, in addition to including antioxidants in the birds’ diet to protect cells from peroxidative damage, antioxidant inclusion in the extender might also help improve cell performance in vitro.Doses of only 35 × 106 live normal thawed sperm per AI generated fertility rates of 30%. The method of freezing sperm in pellets combined with hotplate thawing is an effective method for cryopreserving performing cells in pheasants. Here, it was able to guarantee an incubation process output of 89 hatched chicks for every 100 eggs incubated. Further studies are needed to identify additional and/or alternative strategies for improving the cryopreservation of sperm in pheasants.

animals : an open access journal from mdpi

10.3390/ani13071256

PMC10093747

Acanthocephalans, commonly known as spiny-headed or thorny-headed worms, are a small group of endoparasites with veterinary, medical and economic importance due to their ability to cause disease in domestic animals, wildlife, and humans. In recent decades, great progress has been made using mitochondrial genome data to clarify the phylogenetic relationships of acanthocephalans. However, the current mitochondrial genome database for acanthocephalans remains very limited. Herein, the characterization of the mitochondrial genome of Pseudoacanthocephalus bufonis (Shipley, 1903), the first representative of the family Pseudoacanthocephalidae, is reported. Phylogenetic analyses using the amino acid sequences of 12 protein-coding genes supported the validity of the family Pseudoacanthocephalidae and suggested a close affinity between Pseudoacanthocephalidae and Cavisomatidae. Our phylogenetic results also showed that the families Polymorphidae and Centrorhynchidae have a closer relationship than Plagiorhynchidae in the Polymorphida. These findings contribute to revealing the patterns of mitogenomic evolution in this group and represent a substantial step towards reconstructing the classification of the phylum Acanthocephala.

The phylum Acanthocephala is an important monophyletic group of parasites, with adults parasitic in the digestive tracts of all major vertebrate groups. Acanthocephalans are of veterinary, medical, and economic importance due to their ability to cause disease in domestic animals, wildlife, and humans. However, the current genetic data for acanthocephalans are sparse, both in terms of the proportion of taxa surveyed and the number of genes sequenced. Consequently, the basic molecular phylogenetic framework for the phylum is still incomplete. In the present study, we reported the first complete mitochondrial genome from a representative of the family Pseudoacanthocephalidae Petrochenko, 1956. The mitogenome of Pseudoacanthocephalus bufonis (Shipley, 1903) is 14,056 bp in length, contains 36 genes (12 protein-coding genes (PCGs) (lacking atp8), 22 tRNA genes, and 2 rRNA genes (rrnL and rrnS)) and two non-coding regions (NCR1 and NCR2), and displayed the highest GC-skew in the order Echinorhynchida. Phylogenetic results of maximum likelihood (ML) and Bayesian inference (BI) using the amino acid sequences of 12 protein-coding genes in different models provided further evidence for the resurrection of the family Pseudoacanthocephalidae and also supported that the order Echinorhynchida is paraphyletic. A monophyletic clade comprising P. bufonis and Cavisoma magnum suggests a close affinity between Pseudoacanthocephalidae and Cavisomatidae. Our phylogenetic analyses also showed that Polymorphidae has a closer relationship with Centrorhynchidae than Plagiorhynchidae in the monophyletic order Polymorphida.

1. Introduction Acanthocephala is an important group of obligate endoparasites, with more than 1300 species parasitizing the digestive tracts of all major lineages of vertebrates and their larvae developing in arthropods [1,2,3,4]. According to the current classifications based on a combination of morphological and ecological traits, the phylum is divided into three classes, Archiacanthocephala, Eoacanthocephala, and Palaeacanthocephala, which include 10 orders, 26 families, and over 160 genera [1,5,6,7]. Some previous studies have made efforts to establish a basic molecular phylogenetic framework for Acanthocephala using various nuclear sequence data and mitochondrial genes [8,9,10,11,12,13,14,15,16,17]. Recently, mitochondrial genomic data were used to infer the phylogenetic relationships of the higher taxa in Acanthocephala [15,18,19,20,21,22,23,24,25]. However, to date, mitochondrial genome data are available for only 23 species of acanthocephalans, representing 13 families belonging to 6 orders. Several acanthocephalan families and orders were not represented in the above-mentioned phylogenetic studies, due to the paucity and inaccessibility of suitable material or genetic data for these groups. Echinorhynchida is the largest order in the phylum Acanthocephala, containing more than 470 nominal species, which mainly parasitize teleost fishes but also occur in amphibians and reptiles [1,5,26,27]. Amin (2013) listed 11 families in the Echinorhynchida, including Arhythmacanthidae, Cavisomatidae, Echinorhynchidae, Fessisentidae, Heteracanthocephalidae, llliosentidae, Isthmosacanthidae, Pomphorhynchidae, Rhadinorhynchidae, Transvenidae, and Sauracanthorhynchidae. Later, two new families, namely Gymnorhadinorhynchidae and Spinulacorpidae, were erected [6,28]. Additionally, two families, Paracanthocephalidae and Pseudoacanthocephalidae, were resurrected [16]. However, the phylogenetic relationships of these families are still uncertain. Additionally, the non-monophyly of Echinorhynchida was also revealed by some previous phylogenetic studies [8,9,10,11,12,13,14,17,18,29,30].In order to further test the monophyly of the orders Echinorhynchida and Polymorphida, assess the validity of the recently resurrected family Pseudoacanthocephalidae, and clarify the evolutionary relationships of the Pseudoacanthocephalidae and the other families in Palaeacanthocephala using mitogenome data, the complete mitochondrial genome of Pseudoacanthocephalus bufonis, the first mitogenome from the Pseudoacanthocephalidae, was sequenced and annotated for the first time. Moreover, phylogenetic analyses of the protein-coding genes of all available acanthocephalan mitogenomes were performed using maximum likelihood (ML) and Bayesian inference (BI) in different models.2. Materials and Methods 2.1. Parasite Collection and Species IdentificationA total of 14 spot-legged tree frogs, Polypedates megacephalus Hallowell (Anura: Rhacophoridae), were caught by hand at night in the Diaoluo Mountains, Hainan Island, China, and euthanized by injection of an overdose of pentobarbitone sodium solution. The acanthocephalan specimens were collected from the intestine of the host. For light microscopical studies, acanthocephalans were cleared in glycerine. Photomicrographs were recorded using a Nikon® digital camera coupled to a Nikon® optical microscopy (Nikon ECLIPSE Ni-U, Nikon Corporation, Tokyo, Japan). The specimens were identified as P. bufonis based on morphological features reported in previous studies [31,32,33,34,35,36]. The terminology is according to the previous study [37]. Voucher specimens were deposited in the College of Life Sciences, Hebei Normal University, Hebei Province, Shijiazhuang, China (HBNU-A-2022A001L).2.2. Molecular ProceduresFor molecular analysis, the genomic DNA was extracted using a modified CTAB (pH 8.0)-based DNA extraction protocol as described in Zhao et al. [38]. The genomic DNA library was constructed, and a total of 20 GB of clean data were generated using the pair-end 150 sequencing method on the Illumina NovaSeq 6000 platform by Novogene (Tianjin, China).The complete mitochondrial genome was assembled using GetOrganelle v1.7.2a [39]. Protein coding genes (PCGs), rRNAs, and tRNAs were annotated using the MitoS web server (http://mitos2.bioinf.uni-leipzig.de/index.py, accessed on 20 January 2022) and MitoZ v2.4 [40]. The open reading frame (ORF) of each PCG was confirmed manually by the web version of ORF finder (https://www.ncbi.nlm.nih.gov/orffinder/, accessed on 10 March 2022). The “lost” tRNA genes ignored by both MitoS and MitoZ were identified using BLAST based on a database of the existing tRNA sequences of Acanthocephala. The secondary structures of tRNAs were predicted by the ViennaRNA module [41], building on MitoS2 [42] and RNAstructure v6.3 [43], followed by a manual correction. MitoZ v2.4 was used to visualize and depict gene element features [40]. The base composition, amino acid usage, and relative synonymous codon usage (RSCU) were calculated by a Python script, which refers to codon adaptation index (CAI) [44]. The total length of the base composition included ambiguous bases. The base skew analysis was used to describe the base composition of nucleotide sequences. The relative values were calculated using the formulas: ATskew=A−TA+T and GCskew=G−CG+C. The complete mitochondrial genome sequence of P. bufonis obtained herein was deposited in the GenBank database (http://www.ncbi.nlm.nih.gov, accessed on 5 October 2022).2.3. Phylogenetic AnalysesPhylogenetic analyses were performed based on concatenated amino acid sequences of 12 PCGs using maximum likelihood (ML) and Bayesian inference (BI). Gnathostomula armata and G. paradoxa (Gnathostomulida) were chosen as the out-group. The in-group included 7 species of rotifers and 24 species of acanthocephalans. Detailed information on representatives included in the present phylogeny was provided in Table 1. The phylogenetic trees were re-rooted on Gnathostomulida. Genes were aligned separately using MAFFT v7.313 under the iterative refinement method of E-INS-I [45]. Ambiguous sites and poorly aligned positions were pruned using BMGE v1.12 (m = BLOSUM90, h = 0.5) [46]. The aligned and pruned sequences were concatenated into a matrix by PhyloSuite v1.2.2 [47]. The pruned alignments were then concatenated into the “AA” matrix with the amino acid sequences of PCGs (2087 sites). Bayesian inference (BI) was implemented under the CAT + GTR + G4 model, using PhyloBayes-MPI 1.8 [45,48,49,50,51]. Two independent Markov Chain Monte Carlo (MCMC) runs of 8000 generations each were executed. A consensus tree was simultaneously built by pooling the remaining MCMC trees from both runs. Convergence was evaluated with the “bpcomp” and “tracecomp” procedures in the PhyloBayes package with a burn-in of the first 1000 generations. The maximum discrepancy in the convergence result is 0.017. The maximum likelihood (ML) inference was conducted in IQTREE v2.1.2 [52]. Substitution models were compared and selected according to the Bayesian Information Criterion (BIC) by using ModelFinder [53]. Additionally, a profile mixture model (C60) was used based on the best-fit substitution model of the NP datasets of amino acid datasets [54]. An edge-unlinked model was specified for both the full partition and the merged partition schemes. The best run is selected from the four independent runs based on log-likelihood. A total of 1000 Ultrafast bootstraps were used to evaluate the nodal support of the ML tree [55], and to estimate the consensus tree. For the “AA” matrix, three partition schemes were applied for ML (Table 2): (1) no partition (NP); (2) full partition (FP) that provides the best-fitting model for each individual gene; and (3) merged partition (MP) that implements a greedy strategy starting with the full partition model and subsequently merging pairs of genes until the model fit does not improve any further. The phylogenetic-terrace aware (PTA) data structure was used to facilitate the efficient analysis of the “AA” matrix under each partition model [56]. We selected the best final maximum likelihood and consensus trees according to the Akaike Information Criterion (AIC). Phylogenetic analyses ranked nodes with posterior probabilities (PP) and bootstrap support values (BS) = 1/100 as fully supported, 0.98–0.99/95–99 as strongly supported, 0.95–0.97/90–94 as generally supported, and <90/0.95 as weakly supported. The phylogenetic trees were visualized in iTOL v6.1.1 [57].3. Results and Discussion3.1. Morphology of Pseudoacanthocephalus bufonis (Figure 1, Table 3)Trunks are medium-sized, smooth, and cylindrical. Females are much larger than males. Proboscis is nearly cylindrical, armed with 16–20 longitudinal rows of 3–5 rooted hooks each. Proboscis receptacle is double-walled with the cerebral ganglion at the posterior of the proboscis receptacle. The neck is short. Lemnisci are more or less equal, slightly longer or shorter than the proboscis receptacle. Morphometric data of the present specimens and morphometric comparisons of P. bufonis between our specimens and previous studies are shown in Table 3.The morphology and measurements of the present material are more or less identical to the previous descriptions of P. bufonis [31,32,33,34,35,36], including the morphology and size of trunk and proboscis, the number of the longitudinal rows of proboscis hooks and the hooks per longitudinal row, the number and length of testes and cement-glands, and the morphology and size of eggs. However, the lengths of the proboscis receptacle and lemnisci are slightly ser than those of the previous studies. Additionally, we also sequenced the ITS region (OQ550505, OQ550506) of our specimens. Pairwise comparison of the ITS sequences of our specimens with the available ITS data (KC491878–KC491883) of P. bufonis reported in the previous study [70], displayed only 0.17% nucleotide divergence. Thus, we confirmed our specimens to be P. bufonis.3.2. Gene Content and Organization of the MitogenomeThe complete mitogenome of P. bufonis is 14,056 bp in length and includes 36 genes, containing 12 PCGs (cox1–3, nad1–6, nad4l, cytb, and atp6), 22 tRNA genes, 2 rRNA genes (rrnS and rrnL), and two non-coding regions (NCR1 and NCR2) (Figure 2, Table 4). The lack of atp8 in the mitogenome of P. bufonis is typical for most of the available mitogenomes of acanthocephalans, except for Leptorhynchoides thecatus, which has two putative atp8 genes [55]. All genes in the mitogenome of P. bufonis are encoded on the same strand and in the same direction. Furthermore, the highest GC-skew (0.53) and the second lowest AT-skew (−0.28) of the mitogenome of P. bufonis in the order Echinorhynchida show its preference for G and T nucleotides (Figure 3), which was possibly a result of the propensity for low use of A-rich codons in their PCGs (Table 5). A similar situation also occurred in Polyacanthorhynchus caballeroi and some species of Polymorphida [19,21,25].3.3. Protein-Coding Genes and Codon UsageThe length of 12 PCGs is 10,114 bp. 12 PCGs encode 3358 amino acids and include 3358 codons, excluding termination codons. The longest PCG is nad5 (1620 bp), while the shortest PCG is nad4l (243 bp) (Figure 2, Table 4).The composition and usage of codons in the mitogenome of P. bufonis were shown in Figure 4 and Table 6. ATN (i.e., ATA, ATG, and ATT), GTG, and TTG are used as start codons for the 12 PCGs in the mitogenome of P. bufonis, whereas TAA, TAG, and incomplete codons of T or TA are used as termination codons, in accordance with those of other acanthocephalans [15,18,19,20,21,22,23,24,25,58,59,60,61,62,63,64]. GTG is the most common start codon, being used for six PCGs (cox1, cox3, nad2, nad4l, nad5, and nad6), followed by ATN for four PCGs (ATA: atp6; ATG: nad3 and nad4; ATT: nad1). Two genes (cytb and cox2) were inferred to use TTG as the start codon. Among the 12 PCGs, six genes (atp6, cytb, nad1, nad2, nad3, and nad4l) are terminated with complete stop codon TAA, while three genes (cox1, cox2, and nad5) were inferred to terminate with complete stop codon TAG. The incomplete stop codons T and TA are used for cox3 and nad4, and nad6, respectively.In the PCGs of P. bufonis, the codon with the highest RSCU value is AGG (Ser), while the rarest codon is CTC (Leu). Val is the most frequently used amino acid (16.66%) in 12 PCGs of P. bufonis. Gln is the least commonly used amino acid (0.59%). The high frequency of Val (encoded by GTN) is associated with the high proportions of G and T in their protein-coding sequences (Figure 4 and Table 6).3.4. Ribosomal and Transfer RNAsA total of 22 tRNAs were identified, ranging in length from 42 bp (trnC) to 68 bp (trnI) (Table 4). The anticodons (Table 4) and secondary structures (Figure 5) of the 22 tRNAs were identified. Of the 22 tRNAs, four (trnA, trnN, trnL2, trnS1) have a short dihydrouridine (DHU) arm, six (trnA, trnE, trnG, trnI, trnL1, trnY) lack a TψC (T) arm, and two (trnD and trnC) have lost both arms (Figure 5). Moreover, the trnT has a short amino acid acceptor (AA) arm. The other nine tRNAs were predicted to be folded into typical cloverleaf secondary structures, as found in other acanthocephalans [19,23].In the mitogenome of P. bufonis, two rRNAs, rrnL, located between trnY and trnL1, and rrnS, located between trnM and trnP, were identified. The rrnL is 908 bp in length, with 62.86% A + T content, whereas the rrnS is 572 bp in length with 64.37% A + T content (Figure 2 and Table 5).3.5. Gene OrderIn the mitogenome of P. bufonis, gene arrangement of PCGs and rRNAs is in the following order: cox1, rrnL, nad6, atp6, nad3, nad4l, nad4, nad5, ctyb, nad1, rrnS, cox2, cox3, and nad1, a pattern which appears to be relatively conserved in acanthocephalans [22,23,24,25,64]. However, some tRNAs (i.e., trnS1, trnS2, trnM, trnV, trnK, trnR, and trnC) show more variability in translocation [21,65]. There are up to three translocations in tRNAs in the mitogenome of acanthocephalans reported so far (i.e., trnS1, trnS2, and trnK) [18,20,22,23,24,66] (Figure 6). There are two main gene arrangements of trnS2: type A (trnS2, atp6, nad3, trnW, trnV, trnK, trnE, and trnT) and type B (atp6, nad3, trnW, trnV, trnK, trnE, trnT, and trnS2). The trnK has two arrangements: type C (trnK and trnV) and type D (trnV and trnK). The gene arrangement of trnS1 has three order types: type E (trnS1, trnM, rrnS, trnF, cox2, trnC, cox3, trnA, trnR, and trnN), type F (trnM, rrnS, trnF, cox2, trnC, cox3, trnA, trnR, trnN, and trnS1), and type G (trnM, trnS1, rrnS, trnF, cox2, trnC, cox3, trnA, trnR, and trnN) [18,20,22,23,24,59,66,67]. In the mito-genome of P. bufonis, trnR is situated between trnA and trnN, while trnS2 lays between trnD and atp6 (Figure 6).The gene order of trnS2 in the mitogenome of P. bufonis is type A (trnS2, atp6, nad3, trnW, trnV, trnK, trnE, trnT). The gene arrangement of trnK is of type D (trnK, trnV). The trnS1 of P. bufonis is of type F (trnM, rrnS, trnF, cox2, trnC, cox3, trnA, trnR, trnN, trnS1) (Figure 6).3.6. Non-Coding RegionsIn the mitogenome of P. bufonis, there are two non-coding regions (NCR1 and NCR2). NCR1 is located between trnI and trnM, is 610 bp in length. NCR2, located between trnW and trnV, is 503 bp. Their A + T contents are 61.97% and 56.86%, respectively (Table 5).3.7. Molecular PhylogenyPhylogenetic trees generated from BI and ML methods under different models have similar topologies and indicate that the Acanthocephala are monophyletic, which was widely accepted in previous studies. However, the evolutionary relationships of the Acanthocephala and the three subtaxa of Rotifera (Monogononta, Bdelloidea, and Seisonidea) have been under debate for a long time [19,71,72,73]. The present phylogenetic results showed that the Acanthocephala is sister to Bdelloidea (Rotaria rotatoria, Philodina citrina) and rejected the monophyly of Eurotatoria (Monogononta + Bdelloidea), which are identical to the previous phylogenetic results using EST libraries [72] and mitogenomic data [19], but conflicted with some other phylogenies based on 18S rDNA and transcriptomic data [71,73].Our phylogeny also supported the division of the phylum Acanthocephala into three large clades (Clade I, Clade II, and Clade III) (Figure 7). Clade I, including Macracanthorhynchus hirudinaceus and Oncicola luehei (Oligacanthorhynchida: Oligacanthorhynchidae), represents Archiacanthocephala, a monophyletic group located at the base of the phylogenetic trees of Acanthocephala (Figure 7). The present results agree well with some previous phylogenetic studies [8,9,11,12,13,18,20,22,23,24,25,64,66]. The representative of Polyacanthocephala (Polyacanthorhynchus caballeroi) nested with species of Eoacanthocephala (Pallisentis celatus + Acanthogyrus bilaspurensis + Neoechinorhynchus violentum + Paratenuisentis ambiguus), forming Clade II. The present phylogenetic results challenged the validity of Polyacanthocephala, as have some previous molecular phylogenetic studies [19,22,23,25,64,66].The representatives of Palaeacanthocephala formed Clade III. The monophyly of the order Polymorphida, including the representatives of Plagiorhynchus transversus, Polymorphus minutus, Southwellina hispida, Centrorhynchus clitorideus, C. milvus, C. aluconis, Sphaerirostris lanceoides, and S. picae, is strongly supported in our phylogenetic results. In Polymorphida, the Polymorphidae are more closely related to the Centrorhynchidae than the Plagiorhynchidae, in accordance with other recent mitogenomic phylogenies, but inconsistent with some previous phylogenetic studies based on nuclear and mitochondrial genetic markers [17,68,69,74]. Our phylogenetic results showed that the order Echinorhynchida is paraphyletic, which is consistent with previous molecular phylogenetic studies [12,14,28,30,69]. Furthermore, they supported the resurrection of Pseudoacanthocephalidae [16]. In the present mitogenomic phylogeny, P. bufonis clustered together with Cavisoma magnum, suggesting an affinity between Pseudoacanthocephalidae and Cavisomatidae. Our results agreed well with some recent phylogenetic studies based on nuclear gene sequences [16,75]. These indicated that the current classification of Echinorhynchida is based on unique combinations of characteristics, not shared derived features [13]. The systematics of Echinorhynchida needs to be revised so that its constituent families, subfamilies, and genera reflect the underlying lineages. This will require phylogenetic analysis of both nuclear and mitochondrial DNA datasets from representatives of a more diverse range of taxa than are currently available. It is essential to sequence mitogenomes from yet unrepresented taxa for constructing the molecular phylogenetic framework of Acanthocephala and further exploring the unusual patterns of mitogenomic evolution in this group. The complete mitogenome of P. bufonis obtained herein represents a valuable building block for future work.4. ConclusionsIn the present study, the complete mitochondrial genome of P. bufonis, the first representative of the family Pseudoacanthocephalidae, was characterized. Phylogenetic analyses based on the amino acid sequences of 12 protein-coding genes further confirmed the sister relationship of the Acanthocephala and Bdelloidea and rejected the monophyly of Eurotatoria (Monogononta + Bdelloidea) and Pararotatoria (Seisonidea + Acanthocephala). Our phylogeny also revealed that the order Echinorhynchida and the family Echinorhynchidae are both paraphyletic in the Acanthocephala. The current systematic status of Pseudoacanthocephalus in the Echinorhynchidae is challenged. The present phylogenetic results supported the recent resurrection of Pseudoacanthocephalidae and showed a close affinity between Pseudoacanthocephalidae and Cavisomatidae. Phylogenetic analyses also strongly supported the monophyly of the order Polymorphida and indicated that the Polymorphidae and Centrorhynchidae have a closer relationship than the Plagiorhynchidae. The present phylogenetic studies provided a new insight into the evolutionary relationships of higher taxa within Acanthocephala.

animals : an open access journal from mdpi

10.3390/ani11041047

PMC8068115

Dairy consumption is growing, and both the Italian production and the importation of dairy products are increasing to meet demand. As a first step toward understanding the environmental impacts of water use in the expanding dairy industry, the water footprint approach was used to compute the virtual water and water sustainability of dairy system in Trentino Alto Adige, a region characterized by small-scale farms and typical production. The results highlight that dairy products can be produced with minimal potential to contribute to freshwater scarcity. However, dairy production systems vary, both in production patterns and local environmental context. The development of dairy farming systems with high consumptive water requirements should be avoided in water-stressed regions and supported in particularly suitable regions, as Trentino Alto Adige.

Dairy products play a significant role in the human diet, but they are often associated with high freshwater resource depletion. In Italy, the dairy sector represents more than 12% of the total turnover of the agri-food sector. Trentino Alto Adige is the first Italian region in terms of number of dairy farms, but it does not register a quantitatively consistent dairy production. Notwithstanding, it is characterized mostly by small-scale farms whose strengths are the Protected Designations of Origin and typical mountain productions. The present study aims at: (i) accounting for the virtual water VW of the main dairy products (milk, butter and cheese) produced in Trentino Alto Adige; (ii) estimating the renewable water resources based on the water flow assessment of the study area; (iii) assessing water sustainability comparing the virtual water consumption of the dairy sector at a regional level to the water availability. The findings show that the consumptive virtual water related to dairy production represents about 1% of the water availability in Trentino Alto Adige. Italy’s domestic dairy production is expanding to meet the growing demand, but the expansion of dairy farming in water-stressed regions should be avoided, preferring instead suitable mountain regions where small-scale farms represent a lively entrepreneurial substrate.

1. IntroductionFood production and consumption have been recognized as major sources of environmental impacts [1]. The challenge of meeting the dietary requirements of an increasing world population is stimulating a strong debate about the sustainability of current food production systems, especially referring to meat and dairy productions [2,3,4].Dairy products play a significant role in human diet due to their uniqueness, desirability, and economical and nutritional value. However, these products are often associated with high environmental impacts, mostly in terms of freshwater resource depletion [5,6]. Moreover, the dairy industry is responsible for the production of wastewaters and effluents that could have a significant environmental impact due to their pollutant characteristics [7,8].In Italy, the dairy sector represents more than 12% of the total turnover of the agri-food sector [9]. In the European milk production ranking, Italy is fifth after Germany, France, the United Kingdom, and The Netherlands, but it is the largest producer of typical Protected Designation of Origin (PDO) dairy products [10]. In fact, the Italian dairy supply chain has a strong economic importance on the national agri-food system, thanks to its high level of know-how in management, technology, and genetics. In particular, the Italian alpine area is characterized by small-scale dairy farms whose cows are fed rations based on farm-produced forage [11]. In this area, dairy farming has a strong traditional character, with farms mainly associated to cooperative dairies that produce typical and PDO [12].In this context, considering the importance to pursue economic growth in the dairy sector, while preventing environmental damage, virtual water (VW) is emerging as a relevant sustainability indicator. It is defined as the direct and indirect amount of water that is used in the production processes of commodities during their entire life cycle [13] and can be used as a tool for sustainable freshwater management and governance [14].In particular, virtual water sustainability assessments of livestock production systems and products have received some attention in recent years in countries such as Ireland [15], Australia [16], China [17], Germany [18], Argentina [19], New Zealand [20], and India [21]). VW assessments, through life-cycle approaches, have been conducted analyzing several livestock products and livestock-based production systems at various spatial and temporal scales, quantifying the demand for freshwater resources of the livestock sector [16,20,22,23,24]. Recently, some authors have evaluated and compared virtual water in different types of livestock farming [5,25,26,27,28]. Other studies found that water consumption is strongly influenced by the agroecological characteristics (soil, landform, climate and year type) of the farming system [29,30,31,32,33].In order to assess the actual impact of the dairy sector on the water resources, it is necessary to take into account hydrogeological characteristics and water availability [34]. Palmieri et al. [1] assessed the environmental sustainability of the Italian mozzarella cheese production in a traditional dairy chain, using Life Cycle Assessment, a widely recognized methodology, which allows for the identification of environmental pressures, and also of dairy production [35].Recent studies have evaluated the economic and environmental sustainability of different mountain dairy farms in Northern Italy [36,37,38]. To date, no study has assessed the sustainability of the Italian dairy sector taking into account the water availability of the area in which production is carried out.The present work has the following main objectives: (i) to account for the VW of the main dairy products (milk, butter and cheese) produced in Trentino Alto Adige; (ii) to estimate the renewable water resources based on the water flow assessment of the study area; and (iii) to assess water sustainability comparing the virtual water consumption of the dairy sector at a regional level to the water availability.The Common Agricultural Policy (CAP) recognizes the importance of livestock farming in mountain areas and addresses its programs towards the support of multi-functionality, with contributions, financial incentives, and, in particular, through payments deriving from agri-environmental measures of Rural Development Plans, which are more pronounced than in the past [39].2. Materials and Methods2.1. The Study AreaThe study was conducted in Trentino Alto Adige, an Alpine region located in North-Eastern Italy (Figure 1). Its surface extension is nearly 13,607 km2 and is almost entirely mountainous [40]. Forests dominate the landscape, accounting for 65% of the surface, while cultivated fields only cover 4.3% and pastures 19%. The Utilized Agricultural Area (UAA) thus amounts to 23.6% of the total. While limited in extension, farming in Trentino Alto Adige has a long historical persistence [41]. The climate is alpine, with cold and snowy winters, and short, warm summers. Precipitation is abundant: over the last 10 years, the average rainfall of the region was about 895 mm, exceeding the national average (765 mm) by about 15% [42]. The study area is crossed by numerous freshwater streams and represents the region with the highest distribution of drinking water in Italy [43].Trentino Alto Adige is also the first Italian region in terms of number of dairy cow farms. The latter are characterized by being very small size, with an average of about nine animals per farm [10,36]. Dairy farms in Trentino Alto Adige are small-scale farms, characterized by productions typical of the mountain regions [44]. The small size of the farms has allowed for the conservation of traditional production methods, which are often economically disadvantageous because they involve a lower level of productivity and efficiency, limiting competitiveness on the national and international market [38,44]. However, contrary to the production techniques of intensive farming, dairy mountain farms are of great importance as the pasturing of animals and forage production prevents reforestation [45]. This has positive impacts on the environment and biodiversity but also ensures preservation of traditional landscapes and increases the regional tourist attractiveness [11,46]. In the study area, most of the produced milk is transformed into dairy traditional products, often recognized as Protected Designation of Origin (PDO) [47]. Although Trentino Alto Adige is not very significant quantitatively, producing only 4.5% of Italian milk, its production is mostly characterized by certified quality [48].In particular, the mountain farming system is based on local forage resources, with a combination of fresh pasture in the summer period, and local conserved forages in the rest of the year [49]. The local forage-based diets are part of the basic link between dairy products and their original “terroir”, a notion at the basis of the PDO labelling and image of the product quality from sensory, nutritional, or safety points of view [50,51]. The forages are known to confer specific organoleptic and nutritional qualities to the milk products [52].2.2. Data Sources and DescriptionIn this section, an overview of the data inputs and sources is provided and summarized in Table 1.Data regarding the Italian water footprint (WF) and related components for the dairy products considered in this study were extracted from Mekonnen and Hoekstra [22] (2010) and are expressed in m3/ton. These data, frequently used in the scientific literature [53,54,55,56,57], are based on a 10-year period of monitoring and consequently are less affected by annual weather conditions and/or climatic changes, representing the most affordable choice for macroeconomic studies.Data about the main dairy production in Trentino Alto Adige, provided by the Italian National Institute of Statistics (ISTAT) for the year 2018, were converted into tons and assigned to the product categories established by the Harmonized System and used by Mekonnen and Hoekstra [22]. In particular, these data concern three main product macro-categories: milk, cheese, and butter. Milk was split into three further product categories based on fat content: (i) skimmed milk (containing less than 1% fat), (ii) low-fat milk (containing 1.5–1.8% fat), and (iii) whole milk (containing at least 3.5% fat). Cheese instead was split into two categories based on consistency: (i) fresh and soft cheese, and (ii) semi-hard and hard cheese.Finally, the meteorological data were provided by the Agro-climatic Observatory of the Italian Ministry of Agriculture (MiPAAF). The weather and climate statistics used are estimated based on daily time series of meteorological stations of the National Agrometeorological network (RAN) using a non-stationary geostatistical model that takes into account the location of the stations of both the trend and the geographical correlation of climatic factors. For the purpose of the analysis, the annual precipitation and real evapotranspiration data in Trentino Alto Adige region were used.2.3. Metodological and Empirical FrameworkThe methodology used in this analysis is divided into three different phases. The first phase consisted in identifying the water footprint values and the related component (green, blue and grey) values associated with the production of the dairy products p (skimmed milk, low-fat milk, whole milk, butter, fresh and soft cheese, and semi hard and hard cheese) in the Trentino Alto Adige region, as in the following Equation (1):(1)WF[p] = WFGreen[p] + WFBlue[p] + WFGrey[p]Conceptually, the surface and groundwater utilized for producing any of the selected dairy products is indicated as the WFBlue of the product p. The rainwater utilized for producing any of the selected dairy products p, excluding the rainwater that runs off [59], is indicated as the WFGreen of that dairy product. The quantified amount of water theoretically required to dilute pollutants obtained along the production processes and to bring back the water quality to its acceptable standard is indicated as the WFGrey [59]. The sum of green and blue WFs is also called consumptive WF, while the WFGrey is also called degradative WF [60].Through the evaluation of the individual WF components (green, blue, grey), this first phase aimed at highlighting quantitative differences in terms of water consumption but also at establishing the different impacts on the environment caused by the production of dairy products.In the second phase of the methodological framework, the virtual water (VW) related to dairy products of Trentino Alto Adige region was calculated by multiplying the production volumes of each product p considered in the analysis by its associated water footprints [22]. In particular, the Italian weighted average water footprint was used, whose value takes into account grazing, industrial, and mixed farming systems. In order to calculate the virtual water volumes, the following Equation (2) was adopted:(2)VW[p, t, c] = WF[p, c] × DP[p, t] VW[p, t, c] denotes the virtual water (m3/y) related to each WF component c used to produce the quantities of product p in the year t, and DP represents the dairy production of Trentino Alto Adige in the year t.In the last methodological phase, a measure of water availability (WA) [61] related to the study area was calculated. The available volume of water in the study area was established, analyzing the water balance and in particular the effective infiltration from precipitation. The latter can be considered as a proxy of the volume of renewable water resources, i.e., available water [61]. In fact, due to the Italian morphology and geographical features, the hydrological system can be considered as a closed water system and the precipitation as the major contributor to water availability [62].The volume of precipitation flowing into the study area must be known in order to calculate the overall inflows and outflows. The average annual values for precipitation (P) and those of the real evapotranspiration (ETr) were extracted from MiPAAF, allowing for the analysis of water exchange between the ground and the atmosphere.The effective precipitation (Pe) was estimated as the difference between precipitation (P) and the real evapotranspiration (ETr). After having obtained values for effective precipitation (Pe) and runoff (R), the latter estimated using the method developed by Kennessey [63], effective infiltration (Ie) values were calculated, thus providing an estimation of the recharge and hence the potentially available volume of water (Equation (3)) [64]. (3)WA = (P − ETr) − RThe result of Equation (3), expressed in mm, was converted to m3, considering the territorial extension of the study area [40]. Comparing the VW with the WA, as illustrated in Figure 2, the effective impact of dairy products on the local water resource in Trentino Alto Adige can be evaluated and the water sustainability of the dairy sector discussed.3. ResultsThe results showed that in 2018, the total virtual water of the dairy sector in Trentino Alto Adige was 326,836,622.9 m3, with relevant differences between the three VW components. The VW estimates for each dairy product are provided in Figure 3.For all product categories, the green VW had the largest contribution to the total VW, while blue VW was extremely low in the production of skimmed milk, 1% of the total water, and for all the other categories did not exceed 10%. The grey VW was higher than the blue VW for all the analyzed products with greater values for butter, fresh and soft cheese, and semi-hard and hard cheese. This finding was similar to the results of Palhares and Pezzopane [27], Roibás et al. [65] and Owusu-Sekyere et al. [53], which also indicated green VW as the major component of the total VW for dairy products.The results revealed that the highest total VWs were associated with cheese. In fact, looking at the sum of the two types of cheese considered (fresh cheese and hard and semi-hard cheese), cheese was responsible for about 60% of the virtual water consumption of the entire sector, while representing less than 30% of total production. On the other hand, although about 68% of dairy production was represented by milk (whole, partially skimmed and skimmed), its incidence in terms of virtual water consumption was extremely lower than that of cheese (Figure 4). This result was in line with previous studies that evaluated the environmental impact of cheese production compared to other dairy products, finding that the cheese supply chain had the highest demands for raw materials, energy, and water [66,67]. Cheese production processing involves numerous steps that are both time-consuming and environmentally expensive [68], which make it the main product in the dairy industry due to its greater economic value.On the contrary, drinking milk, requiring fewer steps along the supply chain, necessitates a lower virtual water consumption (3–4% for each component) than all other dairy products.Figure 5 shows that cheese had the highest green, blue, and grey VW values, highlighting a high environmental impact in terms of virtual water consumption. The cheese production process requires a lot of blue water in different stages: Raw milk needs to be pre-treated through pasteurization, separation, and standardization, before beginning the actual cheese production. Afterwards come the phases of coagulation, breaking of the curd, cooking, extraction of the curd, shaping, salting, and maturation. Much of the water used during cheese production is dispersed in the form of wastewater and only a part is incorporated into the finished product, from 40–70% depending on the type of cheese [68].Cheese contributed greatly even to the grey VW of the Trentino Alto Adige dairy system, followed by milk. The grey component provides an indication of the level of pollution generated during the production process. In the case of the dairy supply chain, the increase in the grey water component is mainly due to animal waste, use of fertilizers and pesticides for forage crops, and sediments from eroded pastures [27,69].All products contribute greatly to the green VW of the Trentino Alto Adige dairy industry (Figure 3). Until it becomes blue water, green water does not contribute to environmental flows, which are needed for the freshwater ecosystem protection, and it is not accessible for other human uses [70]. Although the green VW value is particularly high, it does not contribute per se to water scarcity and has a less invasive impact on water resource balances than the blue water consumption [53].The VW related to consumptive water use includes green and blue VWs [59]. The consumptive VW represents about 87% of the total VW of dairy production in Trentino Alto Adige, with a value of about 200 million m3.In order to assess the water sustainability of dairy production, the consumptive VW was compared with the water availability (WA) of the study area (Figure 1). In 2018, the effective infiltration in Trentino Alto Adige amounted approximately to 16 billion m3, allowing for a consistent recharge.The consumptive VW related to dairy production, represents about 1% of the WA measured by the effective infiltration [61], taking into account that much of the runoff water, given the high steepness of the study area and its distance from the sea, ends up in the catchments, feeding them. Considering the effective infiltration as a water availability measure allows for the obtaining of an overestimated precautionary result. Since effective infiltration represents a flow measure, its comparison with the consumptive VW of dairy products results in a measure of the impact of dairy activities exclusively on the regenerative capacity of water, and not on the actual stock. Therefore, we can conclude that the dairy production of the study area, consuming only 1% of the regenerated water, does not cause water stress in the region.Although the dairy sector involves generally water-intensive activities along the whole supply chain, Trentino Alto Adige region has hydrogeological and meteorological characteristics suitable for developing such productions without impacting significantly on local water resources, thus ensuring a strong competitive advantage over the other Italian regions in terms of water sustainability.4. ConclusionsAs the first application of virtual water assessment to the dairy industry in Trentino Alto Adige, this study demonstrated that dairy products can be produced without great potential to contribute to freshwater scarcity, respecting the hydrological cycle (Figure 2). Thus, the generalization that the growing demand for dairy products is one of the major driving factors for water scarcity is not supported in this case. Since Italy is greatly heterogeneous, the opportunities for water footprint reduction are multiple, and examining the regional differences in water footprints of all major dairy commodities is strongly needed.In order to reduce the Italian internal and external virtual water, respectively associated with dairy domestic production and imported products from other countries, environmental policies should focus on the development of the Protected Designations of Origin and typical dairy productions, particularly in suitable mountain areas where small-scale farms represent a lively entrepreneurial substrate.Until today, water has mostly been considered a resource to be managed preferably locally or regionally. This approach does not consider that many water issues are related to remote consumption elsewhere [71]. For dairy products, it is difficult to establish whether they relate to remote water depletion or pollution, because animals are often fed a variety of feed ingredients, whose supply chains are difficult to trace.In Trentino Alto Adige, this problem is rather irrelevant because most of the traditional dairy products are recognized as Protected Designation of Origin (PDO) and produced by mountain traditional small-scale animal farms. Milk, cheese, and butter, in fact, must originate mandatorily from animals raised locally and that grazed locally or were otherwise fed with locally grown feedstuffs. Notwithstanding, from an economic point of view, the small-scale mountain dairy farms are at a greater disadvantage than intensive farms because of their limited productivity [72]. In this regard, during the last five decades, agriculture and livestock systems in the Alps experienced an important structural transformation: The number of traditional and small farms has been decreasing, and they are being replaced by larger, more modern and specialized farms [73], characterized by intensive livestock and crop productions. This transition, the intensity of land use, and the decline of traditional grazing in the highlands could have negative effects on landscape quality and biodiversity, as well as on water resources. In particular, according to Mekonnen and Hoekstra [5], livestock products from industrial farming systems consume and pollute about 97% more surface and groundwater (blue and grey) than livestock products from pastures or mixed systems.Moreover, the increasing complexity of the animal product system hides the existing links between the food we eat, the resources used to produce it, and the associated supply chain impacts [74]. Interventions to reduce water footprints, although desirable, should not be taken without considering the potential consequences for other resource impact categories, as well as social and economic factors.This study can be considered as a basis for further scientific efforts addressed at the evaluation of the dairy production sustainability within particularly suitable areas and regions.

animals : an open access journal from mdpi

10.3390/ani11061734

PMC8230354

Epibionts are organisms that live or grow attached to other living beings, and sea turtles can be suitable habitat for these organisms because they provide a large and diverse substrate. They usually have interspecific relationships of the commensal type; however, some species become parasitic and may cause severe damage, mainly in soft areas. Epibionts provide us with information on the migratory habits of sea turtles and can indicate health status. There are several studies on epibionts and their relationships with sea turtles; however, it is essential to expand research to increase the knowledge that will allow us to comprehend these relationships and their implications. In this study, we analyze the richness, abundance, diversity, prevalence, body distribution, and interspecific relationships of epibionts with Lepidochelys olivacea turtles nesting in the Mexican South Pacific, relate turtle size with the presence of epibionts, characterize the body distribution of epibionts, determine the affinity in species composition, and document the interspecific relationships.

The present study contributes to the knowledge of epibionts recorded on sea turtles that nested on a beach in the South Pacific of Mexico. A total of 125 Lepidochelys olivacea turtles nested on Llano Real beach, Guerrero, Mexico, were examined. We collected 450 conspicuous organisms from 8 species from 43 turtles. The corresponding data analysis was carried out to obtain the relative abundance, the relationship between turtle sizes and the presence of organisms, the similarity of species between the sampling months, and the interspecific relationships between the epibionts and the turtles observed. Chelonibia testudinaria was the most abundant species, while Remora remora was the least abundant species. The turtles were divided into six body sections, with the greatest abundance of these organisms located in the head–neck section of turtles, and there was a significant difference in the size of the turtles that presented epibionts and those that did not. C. testudinaria showed greater similarity between sampling months, and the interspecific relationships recorded were commensalism, parasitism, amensalism, and protocooperation. This research contributes the first record of epibionts in L. olivacea nesting in Guerrero, Mexico.

1. IntroductionEpibionts, being organisms that grow and live attached to other species, are useful to those seeking knowledge of the biology and ecology of the living being that serves as a substrate in this association [1]. Many studies have created lists of epibionts that have been identified on different hosts [2,3]. The body surface of sea turtles is often used by epibiont fauna as a settlement substrate and as a means of dispersal and food procurement [4]. Analysis of epibionts may provide information on sea turtle biology and ecology, indicating types of environments passed through, migration times and depth, regional occurrence, habitat use, health, seasonality, behavior, gender-based patterns, and signs of climate change [5,6].The epibionts of mobile organisms also encounter unfavorable conditions, such as morphological and physiological changes of the basibiont, or friction with other species [7]. They can also be eaten by its basibiont’s predators [8] and suffer abrupt environmental changes, especially those epibionts that live on organisms that have large movements in distance and depth. One of the most extreme examples is the case of the epibionts that live on the carapace of the loggerhead turtle (Caretta caretta), since this basibiont passes through coastal, oceanic, and even terrestrial environments in tropical and subtropical areas [9].Sea turtles are possibly the marine species with the most diverse epibiont communities due to the great variability in movement patterns and feeding preferences among individuals. Therefore, sea turtles and their epibionts are useful as biological models to investigate factors influencing interspecific variation in epibiont community structure [10,11,12]. Comprehensive taxonomic analyses of epibiotic fauna may reveal clustering patterns and distinguish groups of sea turtles [13]. Further, because sea turtles nest on tropical beaches around the world, they are helpful for examining how epibionts respond to hosts leaving the water.Epibionts have been considered indicators of health in sea turtles. In the first instance the intraspecific relationship therebetween is considered commensalism, but this relationship may change to amensalism or even parasitism depending on the number of organisms and the effects of these on a single turtle [14,15]. In Lepidochelys olivacea (Eschscholtz 1829), areas most affected by epibiosis are the carapace, flippers, head and neck [16], and the diversity of species is lower than in others, such as the loggerhead turtle (Caretta caretta) and hawksbill turtle (Eretmochelys imbricata), of which more than 100 epibiont species have been described [17]. On the Mexican coasts, 21 species of conspicuous epibiont fauna have been reported for L. olivacea [18] and studies on epibionts in L. olivacea in the Mexican Pacific have been conducted by Hernández and Valadez [19], Gámez et al. [16], Ayala and González [20], and Frick et al. [21]. The aims of this study were to: (1) determine the richness, abundance, diversity, and prevalence of epibionts in L. olivacea turtles nesting in the Mexican South Pacific; (2) relate turtle size with the presence of epibionts; (3) characterize the distribution of epibionts on turtles’ bodies; (4) determine the affinity in species composition; and (5) document interspecific relationships between Lepidochelys olivacea and its epibionts.2. Materials and MethodsThe fieldwork was carried out in Guerrero, Mexico, on the beach Llano Real where the Sea Turtle Conservation and Protection Center is located (coordinates 17°04′00.4′′ N, 100°26′56.8′′ W). This center is administered by the Universidad Autónoma de Guerrero (Figure 1).Night tours were conducted from July to November 2017 to locate nesting turtles, and once the turtle was found, we examined it carefully. The same two collectors conducted the epibiont assessments, for which the turtle’s body was divided into six sections (head/neck, front flippers, rear flippers, carapace, plastron, and cloaca), following almost the same standardization suggested by Lazo-Wasem et al. [2]. For each body section, the conspicuous fauna were collected and deposited in labeled plastic vessels. After collection, turtle curved carapace length (CCL) and curved carapace width (CCW) biometrics were recorded. The conspicuous epibionts were preserved in alcohol (70%) and transferred to the laboratory for identification and quantification. Identification of epibiont species was carried out using specialized literature [22,23,24], and nomenclature was reviewed using the World Register of Marine Species (WoRMS) site.Epibiont species richness was determined using the Margalef diversity index. Relative abundance was estimated by dividing the individuals of each epibiont species between the total numbers of organisms collected, and expressed as a percentage. Diversity was estimated with the Shannon–Wiener index.To determine prevalence, the number of turtles with epibionts was divided by the total number of turtles analyzed and expressed as a percentage. We used the Mann–Whitney U test to test differences between sizes (CCW) of turtles that showed epibionts and those that did not. Additionally, we determined the body distribution of epibionts, species richness, abundance, and frequency of occurrence per body section. To test affinity in epibiont species composition between months, a dendrogram was formed using the Bray–Curtis dissimilarity (transforming the data to square root and grouping data to determine the similarity between sampling months). Specialized literature was consulted to determine the interspecific relationship between epibionts and turtles [16,19,21]. In addition, to define interspecific relationships, we recorded direct observations of species of epibiont, abundance per turtle, and the type of interaction at the time of collection.3. ResultsWe examined a total of 125 females of L. olivacea nested on the beach (Sea Turtle Conservation and Protection Center). Conspicuous epibiont fauna were observed in 43 turtles, from which 450 specimens were collected, and 3 phyla, 5 families, 8 genera, and 8 species were identified (Figure 2). The prevalence of epibionts was 34.4%, the species richness according to the Margalef index was 1.146, and the estimated Shannon–Wiener diversity index was 2.198 bits/individual.The epibiont with the highest abundance was the barnacle Chelonibia testudinaria (195 specimens), followed by the annelid Ozobranchus branchiatus (110 specimens). The species with the lowest abundance and prevalence was Remora remora, the only vertebrate present within the epibiont group; this species was collected in two females (Table 1).Mean estimated CCW in turtles in which no epibionts were observed was 66.8 ± cm (σ: 8.29), for turtles where epibionts were found, the mean estimated CCW was 69.3 ± cm (σ: 2.95). According to the Mann–Whitney U test, the existence of a significant statistical difference between turtle CCW averages was estimated (U = 1334.0, p = 0.025).For the body distribution of epibionts, the body section with the highest species richness was the carapace, where seven of the eight species reported in this research were found, followed by the head–neck and front flippers sections, where six species were recorded. However, the highest abundance of epibionts was observed in the head–neck (291 specimens), followed by the carapace section (117 specimens, Table 2). No epibionts were observed in the cloaca. The species C. testudinaria was presented as the one with the broadest body distribution since it was observed in five of the six body regions into which the turtle was divided. It was also the most abundant, presenting high abundances in the body of the turtle.Concerning the affinity in species composition, the Bray–Curtis dissimilarity indicated the formation of six groups according to clustering analysis of the species and abundance of epibionts collected from July to November 2017 (Figure 3). Groups e and f showed >50% similarity; group e included species collected during July and August, while group f were species collected during September and November. Group d had only one species collected in July and one in November (40% similarity). Group c had organisms collected in September, October, and November (>30% similarity). Group b obtained <20% similarity, and finally group a was represented by one species collected only in August, obtaining 0% similarity with the other months (Table 3).Based on the specialized literature [16,19,21,25] it was recognized that C. testudinaria, L. hilli, C. virgatum, P. hexastylos, and S. elegans were related to commensalism; S. muricata to amensalism; O. branchiatus to parasitism; and R. remora to protocooperation, with observed sea turtles.4. DiscussionIn Mexico, 21 species of conspicuous epibionts have been reported for L. olivacea [18]. Eighteen epibiont taxa have been reported from turtles [2,16,19,26,27,28], observed at various sites in ecoregion No. 17 [29], where the state of Guerrero is located (Table 4). In the present study, we documented only 38% of the epibiont fauna reported for the Mexican Republic, and 44% of that reported for ecoregion 17 were found in the turtle population analyzed.The prevalence of epibionts in the population studied was low and given that epibionts have been considered as indicators of sea turtle health, since a sick turtle increases the probability of a greater load of epibionts [14], the low prevalence of epibionts is a good health indicator of the population of nesting turtles in the study area. In addition, the low species richness, the low estimated diversity index, and the abundance recorded strengthen the aforementioned assumption; however, it is necessary to test this hypothesis, given that the diversity of epibionts in L. olivacea has been reported to be lower when compared to other sea turtle species [17].According to Márquez [30] and Zug [31], in L. olivacea, as age increases, the ratio of width to length increases, so that older turtles will be wider. The significant difference estimated in CCW size between turtles with and without epibionts in the present study is evidence that strengthens the assumption that the older the turtle, the greater the epibiont load.Of the species reported by this research, only S. muricata had not been found in ecoregion 17; however, there are reports [21,32] of this species in the state of Sinaloa (Table 4).C. testudinaria has been considered an “obligate” epibiont species because it has been found in six species of sea turtles [16,32]. This species secretes a substance that allows it to adhere to areas with rigid substrate and has been found on the heads, noses, and carapaces of sea turtles [3]. In addition to being found in the body sections, it was also observed in the nails in the population studied. Furthermore, this study reports this species as the most widely distributed among turtle body sections and the most abundant in the site in ecoregion 17 [2,26,28]. Further, in addition to being found in 5 of 6 turtle body sections, it was also observed on the claws on the front flippers.The annelid O. branchiatus is hematophagous [33] and was presented as the species that occupied the second place in relative abundance, coinciding with what Gámez et al. [15] reported; however, Hernández-Vázquez and Valadez-González [18] reported it as the most abundant epibiont in L. olivacea. This species was found mainly on the neck and front flippers in the population studied and was also observed moving on the carapace.The species L. hilli and C. virgatum belong to the Lepadidae family, which occupies habitat associated with floating objects and secretes an adhesive substance through the peduncle, so they can be found on any part of the turtle, whether hard or soft, and their presence on turtles implies that they frequent shallow depths close to the surface layers of the sea [34,35]. These two species were found in four of the turtle specimens analyzed, but the greatest abundance was recorded in two: the first had a piece of rope entangled in a front flipper, and the second was observed with damage to the carapace, both considerably injured and weakened, so it is assumed that their swimming was slow, and they spent more time floating on the surface, facilitating the adherence of these organisms.It was reported that members of the genus Stomatolepas are common commensals of sea turtles; these small barnacles often attach themselves to the soft skin areas of their hosts and are frequently located in the neck area [23], which coincides with what was observed in the field for S. elegans. For other barnacles of the same family, such as S. muricata, it has been described that it has a screw shape and finely serrated ornamentations with which it pierces soft areas such as the neck and fins and develops inside the skin [21]. Moreover, for P. hexastylos it was indicated that this species adheres more easily to hard substrates by using a membranous base with grooves that allow it to attach to turtles [36]. Our study corroborates the abovementioned, due to finding S. muricata located in parts of the neck and front flippers inside the skin and P. hexastylos with greater abundance near the mouth and nose of the turtles. The vertebrate R. remora was the least abundant epibiont and was found on the turtle carapace; it has also been reported as the least abundant, and has been found in turtle nests, so presumably, these were attached to the plastron [19].It has been reported that the turtle’s neck, being soft-skinned, is more vulnerable to colonization by epibionts [22]. This research corroborates this since the highest abundance of epibionts was recorded in the head/neck section.In the plastron section, a low abundance of organisms was observed because it is likely that the epibionts are detached from the turtle when it comes out to lay eggs, due to sand friction. The only section where no attached organisms were found was the cloaca; in other studies, coprophagous crabs of the species Planes major have been found [2].Of the eight species collected, it can be seen that some had affinity among them during the collection period: C. testudinaria was the one that occurred in greater abundance and therefore obtained greater similarity between months. From September, October, and November onwards, different species were collected with >50% similarity, unlike July and August, where the species affinity was <20%. The only species for which no similarity was obtained was R. remora since it was only collected in August. This indicates that turtles with greater species similarity may have the same migration routes, feeding, and breeding areas and that the turtles with greater size were those with greater epibiont similarity.The importance of studying symbiotic relationships between epibionts and sea turtles has been pointed out as it aids in determining the effect that epibionts have on their hosts [37]. Species such as C. testudinaria, P. hexastylos, L. hilli, C. virgatum, and S. elegans found in this research were cataloged as commensal species, similar to the conclusion of Casale et al. [38] and Pinou et al. [24], who reported that these species only use the turtle as a substrate and adhere to it from their cypris larval stage, until completing their life cycle. Based on the size, weight, and abundance of the species mentioned above, it is considered that they are not harmful to the turtles analyzed. It was reported that R. remora establishes a proto-cooperative relationship; this interaction presents advantages for the remora that include transportation without energy expenditure, obtaining food, and mating opportunities, in addition to “cleaning” the turtles from other epibionts [39]. The leech O. branchiatus was the only epibiont that was considered a strict ectoparasite. Oceguera-Figueroa and León-Règagnon [33] report that this species feeds directly on the turtle and can cause considerable damage, as it weakens it by consuming its blood. The barnacle S. muricata does not feed on the turtle, but it does cause considerable damage because it pierces the soft parts, such as the flippers and neck. Ross and Frick [40] report that this species has fine lateral serrated ornamentation, which it uses to hold onto turtles and with which it opens the skin; even when the turtle produces a layer of fat around the organism, it breaks through this to be able to develop. Zardus [25] considered that the relationship established between S. muricata and sea turtles is amensalism, which comports with our observations, including that the turtles did not suffer significant damage in this research from those organisms.5. ConclusionsWe report the first record of epibionts in L. olivacea turtles nesting in Guerrero, Mexico. Other reports confirm that the diversity of epibionts in L. olivacea females is low compared to other sea turtle species. In the study area, only 44% of the epibiont species reported in different sites of ecoregion 17 were found to be establishing some type of interspecific relationship with the nesting females of L. olivacea.We report for the first time the presence of the species S. muricata for ecoregion 17 (Mexican Transitional Pacific) as an epibiont on L. olivacea. The presence of C. testudinaria as an “obligate” epibiont in the population studied was corroborated, but it was also the most abundant and widely distributed species in the turtle’s body in the study area. The head/neck section was the one that was inhabited mainly by epibionts without significant effects of their presence in the turtle population.Due to the similarity of epibionts recorded between sampling months, it was considered that the turtles nested at the study site might have different feeding and breeding areas. The epibionts recorded in this study have interspecific relationships of commensalism, parasitism, amensalism, and protocooperation with sea turtles.

animals : an open access journal from mdpi

10.3390/ani12050603

PMC8909170

The provision of appropriate bedding is important for the welfare of dairy cows. Before bedding can be selected, it is critical to understand the properties of the bedding, including its impact on milk microbiota. The objective of this article was to evaluate the influence of three materials for use as bedding on physicochemical properties, bacterial counts and colostrum microbiota of cows. Our results demonstrate that peanut shells appear to be a suitable bedding material for cows. These experiments provide empirical support for the use of peanut shells and rice husks as bedding material for dairy cows and illustrates the effects of bedding types on the colostrum microbiota of dairy cows.

The aim of this study was to evaluate peanut shells and rice husks as bedding for dairy cows. We analyzed material properties including dry matter, water holding capacity, pH level and bacterial counts. Bedding treatments were compared with a one-way ANOVA using twelve cows split into three groups. Colostrum microbiota was analyzed by sequencing of the V3–V4 region of the 16S rRNA gene. Dry matter content was higher in rice husks compared with peanut shells. No treatment effects were found for water holding capacity and pH level. Streptococcus agalactia counts in peanut shell bedding were lower than in rice husk bedding, and Pseudomonas aeruginosa counts were not different between beddings. A significant enrichment for Enhydrobacter and Pantoea were detected in the colostrum of cows that used peanut shells compared with other beddings. Colostrum of cows housed on a peanut–rice combination had a greater relative abundance of Pseudomonas and Corynebacterium than those housed on peanut shells or rice husks. Higher numbers of Bacteroides, Akkermansia, Alistipes, Ruminococcaceae_UCG-014, Coriobacteriaceae_UCG-002 and Intestinimona were found in the colostrum of cows housed on rice husk bedding over other bedding types. These results suggest that bedding types were associated with the growth and diversity of colostrum bacterial loads. In addition, dry matter in peanut shells was lower than found in rice husks, but there was also a lower risk of mastitis for peanut shell bedding than other beddings.

1. IntroductionBedding plays an important role in the welfare of housed dairy cows [1], as they usually lie on it for 8–16 h/d to rest [2]. Good bedding improves cows’ cleanliness, behavior, and udder health, and reduces hoof injuries [3,4,5]. Cows lie down for longer when the bedding is dry, soft and clean [2,3,6]. Management of bedding to maintain properties such as dryness, is also an important consideration. According to Rowbotham et al. [7], providing new bedding to stalls more than once weekly can reduce the bulk milk somatic cell score. This may have a positive benefit for the farm, because increased somatic cell count is related to milk loss [8].The effect of bedding on dairy cows’ hygiene, udder health and milk quality differ between bedding material type. Manure solids used as bedding have high bacterial counts, leading to dirtier udders and higher coliform and streptococci counts than when sand or organic non-manure materials are used [9]. The diversity and abundance of the milk microbiota is affected by the environment where cows are housed (including their bedding [10]), and milked (with milk microbiota being similar to the microbiota on teat liners and teat dip cups [11]). High bacterial counts in bedding translate to high bacterial counts on cows’ teats, adversely affecting udder health and milk quality [5,12,13]. Sand used as bedding has several welfare advantages: as an inert substance, it does not support bacterial growth, and therefore it reduces the prevalence of mastitis [5]. It also has a lower surface temperature compared to wheat straw, rice husk and sawdust, which is beneficial in hot weather [14]. While it is important to consider microbial load in bedding materials, it is important to weigh this against the comfort and preferences of cows using these materials. Sand is not as comfortable as straw and hence cows lie down on sand for less time [15]. These effects on cows may be caused by differences in the physicochemical properties of the bedding materials [16,17].Bedding materials that have traditionally been used in Chinese dairies include manure solids, sand, wheat straw, rubber mats, rice husks, and sawdust. Organic materials (e.g., wheat straw and dry sawdust) tend to improve the welfare of cows compared to inorganic materials (e.g., sand) [15,18]. However, for some dairy farms in China, the cost of recycled manure solids is prohibitive, and producers prefer to purchase other bedding materials. In Pennsylvania, USA, one survey found that bedding costs comprise 5% of the total cost of rearing replacement heifers [19]. In recent years, rising prices of raw materials have caused dairy farmers to consider alternative materials, including forest biomass (e.g., tree bark and vegetal fibers), conifer forest litter, seagrass (e.g., posidonia oceanica), and flax straw [17,20]. In China, peanut shells are a possible bedding material for dairy cows, supporting cow cleanliness to at least as high a degree as rice husk, which is another alternative bedding shown to be preferred by cows for lying on [21]. Despite this, the use of peanut shells and rice husk bedding in dairies has received little empirical scrutiny.Scientific studies of the characteristics of bedding materials typically focus on the physical (e.g., dry matter content and water holding capacity) and, chemical properties (pH, total organic carbon, the C: N ratio, which may be important for compost-bedded pack) of the bedding, as well as microbial load (e.g., content of Pseudomonas aeruginosa, Escherichia coli, and Streptococcus agalactiae) [14,17,20]. Effect of bedding types on milk microbiota composition should also be explored. The present study aimed to evaluate the physicochemical properties of three bedding types (peanut shells, rice husks and a peanut–rice combination) and their composition of colostrum microbiota for use as bedding for dairy cows.2. Materials and MethodsThis study was conducted at the research dairy farm of Henan Agricultural University, in Zhengzhou, China, between January and April 2020. Seven primiparous and five multiparous (0 to 3 lactations), nonlactating and pregnant Holstein cows (mean 38 ± 11 d before parturition) between 2 to 6 years old, from the research dairy were used. Before the experiment, all cows were housed in one barn and bedded with dry manure. Cows were spilt into three bedding treatments with 4 cows per treatment. Each treatment comprised of two pens with the same bedding materials. Cows of each pen were divided by parity and calving date (two primiparous heifers in the last pen). The resting area in each pen was 5.9 m long and 5.15 m wide with a curb height of 15 cm. Cows were presented with one bedding material per pen. The three bedding treatments were with a sand base plus one of the following bedding materials to a depth of 10 cm: (1) peanut shells; (2) peanut–rice combination (a ratio of two parts of peanut shell to one part of rice husk by weight); (3) rice husks. The average time of cows kept in different bedding treatments was 32 d for peanut shells, 28 d for the peanut–rice combination, and 31 d for rice husks before calving. The effects of bedding types on cow behavior and welfare have already been reported and the management of the bedding has already been described [21]. In brief, the bedding was levelled daily, and feces removed while the cows were eating in their pen. An average 30 kg of fresh bedding material was added in each pen once weekly. Mean daily temperature inside the barn ranged from −1.2 to 18.1 °C measured using two portable weather stations recording at 10 min intervals.2.1. Physical, Chemical and Biological Property AnalysesSurface samples of bedding materials were collected on days 0, 5, 13 and 20 for physical, chemical and biological analysis after bedding use. Analyses of the bedding materials were carried out in the Animal Husbandry Laboratory of Henan Agricultural University. The physical (e.g., dry matter and water holding capacity), chemical (e.g., pH level) and biological (e.g., Streptococcus agalactia content, Pseudomonas aeruginosa content and Escherichia coli content) properties of the bedding materials were analyzed.To measure the dry matter (DM) content of the bedding, 200 g bedding samples per bedding treatment were collected in the sampling period. Each sample consisted of 5 subsamples (approximately 40 g per subsample) of the surface materials following the procedure described by Li et al. [21]. The samples were dried for 72 h at 65 °C using an electric heating constant temperature (blast) drying oven (DHG-9030A, Shanghai Jinghong Experimental Equipment Co., Ltd., Shanghai, China), and the percentage of dry matter content (%) was calculated using equation (1) (1)DM=MwetMdry×100 where Mwet is the mass of the new material (g), Mdry is the dry mass of the material (g).To measure water holding capacity (WHC), samples were divided into 10 g sub-samples per sample. They were first packed into a nylon bag, and then soaked in distilled water for 1 h. Any surface moisture was wiped from the nylon bag and it was then left on an iron grid for 30 min until water had ceased dripping from the bag. The water holding capacity was calculated from Equation (2). To measure pH levels of each bedding type, bedding material was mixed with water in a ratio of 1:10 based on weight for 30 min. A handheld pH meter (Testo 206 PH1, Testo International Trade (Shanghai) Co., Ltd., Shanghai, China) was then used to measure the pH level of the resulting solution. (2)WHC=Mw−MdMwet, where Mw is the mass of the saturated material and nylon bag (g) and Md is the dry mass of the material and nylon bag (g), Mdry is the mass of the new material (g).The bacterial content of bedding materials was determined by polymerase chain reaction (PCR). The genomic DNA of the bacteria in the bedding was extracted using a stool genomic DNA extraction kit (DP328, Tiangen Biochemical Technology (Beijing) Co., Ltd., Beijing, China). The abundance and purity of the genomic DNA obtained were detected using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, Waltham, MASS, USA). Specific primers (Table 1) for genomic DNA were designed and synthesized by Sheng-gong Bioengineering (Shanghai, China) Co., Ltd. These primers were used to amplify bacterial DNA by PCR. Real-time PCR was performed using the LightCycler®96 (Roche, Basle, Switzerland) in 50 µL reaction volumes containing 25 µL of SYBR® Premix Ex Taq™ II (2×; TaRaKa, Dalian, China), 1 µL of cDNA, 2 µL of forward and reverse primers, and 22 µL of RNase-free water. The PCR product was purified, recovered and cloned into the PMD18-T vector in 10 µL connection volumes containing 1 µL of PMD-19T, 5 µL of solution I, Vµl of DNA (V = 0.3 × 0.66 × the product length of gene/the concentration of the PCR product), up to 10 µL of RNase-free water, and the combination was then transformed into E. coli. Trans5a competent cells in accordance with the instructions of the manufacturer. The transformed product was added to a liquid medium containing antibiotics (Ampicillin Sodium) and incubated at 37 °C for 21 h. To check for a single colony in the culture medium, DNA sequencing at Sheng-gong Bioengineering (Shanghai, China) Co., Ltd., and DNAMAN software was used to compare the sequencing results with the target gene sequence on the NCBI website. A single colony of the bacteria was confirmed at similarity > 96%. The successfully transformed monoclonal positive bacteria had plasmid DNA extracted using endotoxin-free plasmid small extraction kit, and resulting samples were stored at −20 °C. Plasmid DNA was diluted in a 5-fold gradient for the preparation of calibration curves. The bacteria and plasmid DNA were detected by fluorescence quantitative PCR. Real-time PCR was performed using the LightCycler®96 (Roche, Basle, Switzerland) in 20 L reaction volumes containing 10 µL of SYBR® Premix Ex Taq™ II (2×; TaRaKa, Dalian, China), 2 µL of cDNA, 2 µL of forward and reverse primers, and 6 µL of RNase-free water. The thermal cycling conditions were as follows: 5 min at 95 °C for one cycle, 35 PCR cycles (30 s at 95 °C, 30 s at 60 °C and 32s at 72 °C) and 10 min at 72 °C. This was followed by melt curves at 95 °C for 10 s and 60 °C for 5 s; cooling at 50 °C for 30 s. Reactions were stored at 4 °C. The cycle threshold (Ct) value of plasmid DNA was converted to copy number using the following formula: Copies·g−1 = DNA concentration (ng·µL−1) × 6.02 × 1014/(sum of base pairs of vector and PCR product)/650.2.2. 16S rDNA Amplicon Sequencing of Colostrum MicrobiotaSampling the colostrum of cows that were housed on the three bedding types was undertaken in the milking parlor within 24 h of calving. For all cows, teats were wiped using a clean cloth towel and sanitized by iodine pre-dip. The first 3 streams of milk were not collected and the next 10 mL of milk per teat was poured into a sterile 50 mL tube. Microbial DNA was extracted from colostrum samples using HiPure Soil DNA Kits (Magen, Guangzhou, China) according to manufacturer’ protocols. The V3–V4 region of the 16S rDNA was amplified using primers 341F (CCTACGGGNGGCWGCAG) and 806R (GGACTACHVGGGTATCTAAT). The 16S rDNA target region of the ribosomal RNA gene was amplified by PCR (95 °C for 5 min, followed by 30 cycles at 95 °C for 1 min, 60 °C for 1 min, and 72 °C for 1 min and a final extension at 72 °C for 7 min) using primers 341F and 806R. Amplicons were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer’s instructions and quantified using ABI StepOnePlus Real-Time PCR System (Life Technologies, Foster City, CA, USA). Purified amplicons were pooled in equimolar amounts and paired-end sequenced (PE250) on an Illumina platform according to standard protocols.Raw reads were further filtered using FASTP (version 0.18.0, Chinese Academy of Sciences, Shenzhen, China). Paired end clean reads were merged as raw tags using FLSAH (version 1.2.11, Johns Hopkins University School of Medicine, MD, USA) with a minimum overlap of 10 bp and mismatch error rates of 2%. Noisy sequences of raw tags were filtered under specific filtering conditions to obtain the high-quality clean tags. The clean tags were clustered into operational taxonomic units (OTUs) of ≥97% similarity using UPARSE (version 9.2.64) pipeline. All chimeric tags were removed using the UCHIME algorithm and finally effective tags were obtained for further analysis. The tag sequence with the highest abundance was selected as a representative sequence within each cluster.2.3. Statistical AnalysisSamples of bedding material were considered the experimental unit. Assumptions of normality and homoscedasticity were checked using the Shapiro–Wilk test and Levene’s test, respectively, in SPSS. Bacterial count data of bedding were base-10 log-transformed before analysis. There was no correlation between sampling time in GLM covariance using initial data as a covariant. Properties of bedding material were analyzed using a one-way ANOVA with SPSS (v.26, IBM, Armonk, NY, USA). The bedding treatment was included as an independent variable; all other variables as a dependent variable. When a significant effect was found, the Tukey post hoc test was used to determine differences between treatment means. For analysis of colostrum microorganisms, Chao1, ACE, Shannon, Simpson, Good’s coverage was calculated in QIIME (version 1.9.1) [22]. Alpha index among groups was calculated by Tukey’s HSD testing R Project Vegan package (version 2.5.3) [23]. Principal Coordinates Analysis (PCoA) based on the weighted unifrac distances was generated in R Project Vegan package (version 2.5.3) and plotted in R project ggplot2 package (version 2.2.1) [24]. The experimental unit was the colostrum sample of microbiota analysis. Figures were generated with GraphPad Prism 8 software (GraphPad Software, Inc., Sacramento, CA, USA). A statistical difference was assumed at p Value < 0.05.3. Results3.1. Physical and Chemical Property AnalysesProperties of bedding material are shown in Table 2. The dry matter content for peanut shells, the peanut–rice combination and rice husks were 73.3%, 78.6%, and 79.3%, respectively. Bedding dry matter was lower for peanut shells than for rice husks but did not differ from the peanut–rice combination. Water holding capacity and pH level did not differ between bedding treatments.3.2. Biological Property AnalysesIn this experiment, three bedding bacteria were tested. The phoA gene of Escherichia coli was not detected. For the other two bacteria, Pseudomonas aeruginosa and Streptococcus agalactia were found in all bedding samples (Figure 1). No differences between bedding materials were found for Pseudomonas aeruginosa. Streptococcus agalactia counts from peanut shell bedding was lower than from rice husks but did not differ from the peanut–rice combination.3.3. Microbial Flora Analysis for Colostrum of Dairy Cows Housed on Three Bedding MaterialsTwelve cases of bacterial 16S rDNA amplification products in colostrum were subjected to high-throughput sequencing. A total of 488,574, 496,487 and 478,391 original sequences were obtained for peanut shell, the peanut–rice combination and rice husk, respectively, and the effective sequences were 473,373, 483,123 and 452,088, respectively. The sequences were clustered into OTUs based on 97% identity. After filtering the chimera, a total of 2691 OTUs, 2274 OTUs and 1983 OTUs were obtained in the three treatments. The number of 591 OTUs shared by the three treatments accounted for 12.33%, and the number of OTUs unique to peanut hulls, peanut–rice combination and rice husk accounted for 26.43%, 19.15% and 21.78% of the total, respectively (Figure 2A). Tag numbers and sequencing depth are shown in Figure 2B.3.3.1. Alpha and Beta Diversities of Microbiota from Colostrum SamplesThe Chao 1 richness, ACE richness, Shannon and Simpson diversity of colostrum samples did not differ between the three bedding materials (p > 0.05) (Table 3). For Shannon diversity the range was 7.50 to 7.69 with a mean of 7.57, and for Simpson diversity the range was 0.97 to 0.98 with a mean of 0.98.PCoA is a dimensionality reduction analysis based on a distance matrix, which evaluates the degree of explanation of the overall difference in bacterial colony structure by each coordinate axis as a percentage. In the analysis results, the more similar the samples are, the closer the distance is reflected in the PCoA. PCoA analysis of the colostrum of cows housed on different bedding materials indicated differences in bacterial community composition, although there were similar levels of richness and diversity (Figure 3). A large overlap was observed between the colostrum from cows on peanut shells and those on the peanut–rice combination, but separate clusters were observed between the colostrum from cows on rice husks samples and peanut shells or the peanut–rice combination.3.3.2. Composition of Colostrum MicrobiotaAcross the three groups, Proteobacteria (33.0%), Firmicutes (18.6%), Cyanobacteria (14.9%), Bacteroidetes (13.9%) and Actinobacteria (7.3%) were the five most abundant bacterial phyla in colostrum from cows bedded on peanut shells. Proteobacteria (38.3%), Firmicutes (27.5%), Bacteroidetes (13.8%), Actinobacteria (9.3%) and Cyanobacteria (4.9%) were the five most abundant bacterial phyla in colostrum from cows bedded on the peanut–rice combination. Firmicutes (38.1%), Bacteroidetes (38.0%), Proteobacteria (10.1%), Verrucomicrobia (3.4%) and Actinobacteria (2.7%) were the five most abundant bacterial phyla in colostrum from cows bedded on the rice husks (Figure 4A). Colostrum from cows on the peanut shells had higher (p < 0.05) Proteobacteria, Actinobacteria and lower (p < 0.05) Bacteroidetes, Verrucomicrobia and Epsilonbacteraeota compared with rice husks, but did not differ from the peanut–rice combination.At the genus level, the predominant bacterial genera in colostrum were Enhydrobacter (4.9%), Bacteroides (3.3%) and Acinetobacter (2.0%) for peanut shell bedding, Acinetobacter (12.3%), Pseudomonas (7.1%), Ruminococcaceae_UCG-005 (4.3%) and Rikenellaceae_RC9_gut_group (2.8%) for the peanut–rice combination bedding, and Bacteroides (16.3%), Lachnospiraceae_NK4A136_group (4.1%), Faecalibacterium (3.9%) and Rikenellaceae_RC9_gut_group (3.4%) for rice husk bedding (Figure 4B). Colostrum of cows bedded on peanut shell had a higher abundance of Enhydrobacter and Pantoea of colostrum microbiota (p < 0.05). For colostrum of cows on the peanut–rice combination, proportions of Pseudomonas and Corynebacterium_1 were higher compared with those in peanut shell or rice husk. Relative abundances of Bacteroides, Akkermansia, Alistipes, Ruminococcaceae_UCG-014, Coriobacteriaceae_UCG-002 and Intestinimonas were significantly higher (p < 0.05) for cows on rice husk.4. DiscussionThe present study was to evaluate the physicochemical properties of three organic materials as bedding for dairy cows, and to understand the effect on colostrum microbiota of these cows. Our results indicate that peanut shells may be promising bedding material for dairy cows. In this study, a difference in dry matter between bedding was observed. Higher dry matter content was found in rice husks bedding compared to peanut shells. The dry matter of bedding materials is an important factor affecting bacterial counts and cow comfort. Cows usually prefer to lie down on a dry surface, and the wet bedding material impairs their welfare by affecting their health and lying time. Higher bacterial counts were detected in manure solids with lower dry matter content reported by Patel et al. [9], which is similar to that reported by Wolfe et al. [25]. Additionally, the wet bedding can dirty the udder of cows, which may increase mastitis risk [3,9]. On the other hand, Reich et al. [26], reported that the lying time of cows is 10.4h/d on wet bedding (34.7% DM) compared to 11.5 h/d on dry bedding (89.8% DM). According to Reich et al. [26], the lying time of cows remains stable when the dry matter content of bedding is above 62.2%. In this experiment, the dry matter of all three bedding types were above 73.3%. Rice husks had the highest dry matter content at 79.3%, followed by the peanut–rice combination (78.6%). The bedding materials in our study can be used for dairy cows, and rice husk bedding was superior in dry matter content. In the current study, the pH value was not affected by bedding treatment. Pseudomonas aeruginosa can survive at 5.6–9.0 pH but the maximum growth observed is between 6.6–7.6 The highest biofilm formation rate of Streptococcus agalactiae is at pH 6.5 [27]. Our mean pH for rice husks of 9.2 was in line with the published report by Patel et al. [9,14], who found the pH was 8.7 for rice husks. A low pH level of bedding materials promoted bacterial growth. After use, the pH of peanut shells and the peanut–rice combination bedding in the study increased to above 9.0 but pH for rice husks was high throughout. This change may be attributed to ammonia from urine and differences in properties of the beddings [14]. Quality bedding material needs to absorb moisture, maintain dryness, and allow animals to display natural behaviors [28]. Water holding capacity is an important property of bedding material as it shows the moisture content that the material absorbs and stores [28]. Additionally, Ahn et al. [29], reported that the water holding capacity of materials increases with increasing moisture content. Similar water holding capacity of three bedding materials were found in this study. Our results agree with those of Ferraz et al. [17], who reported that the water holding capacity was 1.62 for barley husks and 1.65 for spelt husks. However, Ferraz et al. [17], found that Posidonia oceanica (7.32) in wood shavings (4.89) and barley straw (4.13) had higher values compared with ours. This may be attributed to the moisture, texture, and structure of the materials [30]. Compared with non-organic material, organic bedding provides a more comfortable surface, improvement in cow welfare, and is favored by cows. However, the high load of pathogenic bacteria in organic material is a concern for farmers. Escherichia coli, Streptococcus agalactiae, Klebsiella spp., Pseudomonas aeruginosa and Staphylococcus aureus are common microorganisms that endanger the health of cow udders and cause mastitis [31,32,33]. Major reductions are observed in performance (cumulative milk, fat and protein yield) and gross profit for cows with mastitis [34,35]. Cows with mastitis also have a decreased percentage of pregnancies for first artificial inseminations and increased pregnancy loss compared with healthy cows [36]. Bedding materials to which cows’ teats are exposed have significant effects on the prevalence of mastitis [37,38]. Robles et al. [5], found that, bedding material of manure solids presented higher gram-negative bacterial counts compared with sand, straw and wood shavings, indicating that there are differences in bacterial counts between different materials. Consequently, reducing bacterial counts on the surface of bedding is important to reduce the risk factor of mastitis of cows. In the current study, the presence of Pseudomonas aeruginosa and Streptococcus agalactia was noted in all bedding materials. No differences were observed on Pseudomonas aeruginosa counts between bedding types, although they were lower in peanut shells. However, peanut shells had lower Streptococcus agalactia counts compared with rice husks. Our results agree with those of Fávero et al. [39], that peanut shells have lower bacterial counts. This illustrated that bacterial counts were better controlled in peanut shells than in rice husks, and had a minimal risk of udder infection for peanut shell bedding.The proportions of OTUs in the milk of cows kept on different bedding is likely to differ [40]. Wu [10] and Nguyen [41] et al., reported that milk microbiotas are associated with both the bedding and airborne dust microbiota. Bacteria types and counts on the teats of cows are affected by bedding types, and manure solids with high bacterial counts cause more Escherichia coli in the milk tank [9,40,42]. These papers indicate that differences in milk microbiota may be attributed to the environment in which the cows are housed, especially the teat dip cup, airborne dust and bedding types directly exposed to teats. In the current study, no significant differences were found in Alpha diversities of the colostrum microbiota of cows on different bedding types, which is similar to the report by Metzger et al. [40]. With regard to PCoA of milk microbiota, samples from cows housed on peanut shell bedding had a large overlap with the peanut–rice combination, but differed from the cows bedded on rice husks. Metzger et al. [40], reported that there were differences in milk microbiota composition for cows kept on manure solids, recycled sand, sawdust and new sand. It has been demonstrated that the milk microbiota is a dynamic community with the shift of environment [11,40,41,43]. Nguyen et al. [44], found that differences in milk microbiota derived from two farms at the family level. In our study, bedding treatments did affect the relative abundance of the prevalent bacteria, e.g., Firmicutes, Bacteroidetes, Proteobacteria and Actinobacteria. The proportion of dominant bacteria was similar to the report that Proteobacteria and Firmicutes were two prevalent phyla by Nguyen et al. [44]. Ganda et al. [45] reported that the milk microbiota from healthy cows were also dominated by Proteobacteria and Firmicutes. In the current study, different bedding materials (pens with different bedding types can be considered as a distinctive environments) provided a potential difference in colostrum microbiota at the phylum level. Similarly, differences of dominant and non-dominant bacteria in colostrum samples were observed at the genus level. Significant enrichment for Enhydrobacter and Pantoea in the colostrum of cows housed on peanut shells were found. Enhydrobacter and Pantoea belong to gram-negative bacteria. According to Aldrete-Tapia [46], and Metzger et al. [47], Enhydrobacter is considered to be the subdominant genus with a relative abundance of 0–21.52%, which varies with season. Pantoea has been confirmed to exist in the barn and has biological control activity against fungal pathogens [48,49].Colostrum samples from cows using the peanut–rice combination had greater relative abundance of Pseudomonas and Corynebacterium. Corynebacterium is an exogenous pathogen, belonging to gram-positive bacteria, which mainly causes subacute inflammation through udder trauma [50,51]. Psychrophilic bacteria of the genus Pseudomonas have been found to cause milk spoilage. However, the Pseudomonas bacterium is not unique in healthy milk; it is also the only bacteria that exhibits a random distribution in milk microorganisms and is usually inhibited by the probiotic network in healthy milk cows [52,53]. However, Pseudomonas which was significantly enriched in colostrum of cows housed on the peanut–rice combination, as a potential pathogen, is still worth noting as a risk factor for udder health. An increased prevalence of Bacteroides, Akkermansia, Alistipes, Ruminococcaceae_UCG-014, Coriobacteriaceae_UCG-002 and Intestinimona in colostrum of cows on rice husks was found. In this experiment, our results that the highest abundance for Bacteroides was in the colostrum of cows on rice husks is similar to the report by Oikonomou et al. [54]. Akkermansia, belonging to the phylum Verrucomicrobial, is considered to be a beneficial microorganism and is negatively related to certain metabolic disorders [55]. Alistipe of the phylum Bacteroides is not present in abundance in milk with mastitis compared with healthy milk [52]. Derakhshan [50], and Doyle et al. [56], reported that Ruminococcaceae, as the dominant flora in colostrum, is easily affected by the environment. No studies have been reported of Coriobacteriaceae_UCG-002 and Intestinimonas. A significant enrichment for Corynebacterium and Pseudomonas in the colostrum of cows using the peanut–rice combination may be an increased risk factor for mastitis compared with other beddings. Adverse risks factors for the colostrum microbiota of cows using peanut shells and rice husks have not been observed.5. ConclusionsBased on the results of these experiments, differences in the bacterial community composition of colostrum were affected by bedding types. The physicochemical properties exhibited small differences among the bedding materials, except for dry matter content. However, peanut shell bedding reduced bacterial growth and had no effect on colonization of colostrum by mastitis pathogens which illustrates that it appears to be a potential bedding material for dairy cows. Further research is needed to determine more properties of peanut shells and rice husks, including particle size, bulk density and cost.

animals : an open access journal from mdpi

10.3390/ani11030758

PMC8000179

This investigation explored the impact of the COVID-19 pandemic lockdown and following alert levels on pets in New Zealand. Pet owners were surveyed during the last week of the first Alert Level 4 lockdown (highest level of restrictions) and then three months later during Alert Level 1 (lowest level of restrictions). During lockdown, just over half of those surveyed thought that their pet’s wellbeing was better than usual, and most owners could list at least one benefit of lockdown for their pets. These included more company, play and exercise. Owners expressed that they were concerned about their pet’s wellbeing after lockdown, with pets missing company/attention and separation anxiety being major themes. The Alert Level 1 survey indicated that owners continued to play with their pets more but that higher levels of exercise were not maintained. Just over one-third of owners took steps to prepare their pets to transition out of lockdown. The results indicate that pets may have enjoyed improved welfare during lockdown due to the possibility of increased human-pet interaction. The steps taken by owners to prepare animals for a return to normal life may enhance pet wellbeing long-term if maintained.

The influence of the COVID-19 pandemic on human-pet interactions within New Zealand, particularly during lockdown, was investigated via two national surveys. In Survey 1, pet owners (n = 686) responded during the final week of the five-week Alert Level 4 lockdown (highest level of restrictions—April 2020), and survey 2 involved 498 respondents during July 2020 whilst at Alert Level 1 (lowest level of restrictions). During the lockdown, 54.7% of owners felt that their pets’ wellbeing was better than usual, while only 7.4% felt that it was worse. Most respondents (84.0%) could list at least one benefit of lockdown for their pets, and they noted pets were engaged with more play (61.7%) and exercise (49.7%) than pre-lockdown. Many respondents (40.3%) expressed that they were concerned about their pet’s wellbeing after lockdown, with pets missing company/attention and separation anxiety being major themes. In Survey 2, 27.9% of respondents reported that they continued to engage in increased rates of play with their pets after lockdown, however, the higher levels of pet exercise were not maintained. Just over one-third (35.9%) of owners took steps to prepare their pets to transition out of lockdown. The results indicate that pets may have enjoyed improved welfare during lockdown due to the possibility of increased human-pet interaction. The steps taken by owners to prepare animals for a return to normal life may enhance pet wellbeing long-term if maintained.

1. IntroductionThe World Health Organisation declared Corona Virus disease 2019 (COVID-19) caused by the Severe Acute Respiratory Syndrome coronavirus 2 (SARS-CoV-2) a pandemic on 11 March 2020,WHO has been assessing this outbreak around the clock and we are deeply concerned both by the alarming levels of spread and severity, and by the alarming levels of inaction. We have therefore made the assessment that COVID-19 can be characterized as a pandemic [1].In New Zealand, the first COVID-19 case was reported on the 28th of February 2020, and the country’s borders were closed to all non-residents from the 19th of March 2020 [2]. The Government introduced the 4-tiered Alert Level system to help combat COVID-19 on the 21st of March 2020, which in brief is Alert Level 4—Lockdown, Alert Level 3—Restrict, Alert Level 2—Reduce, and Alert Level 1 Prepare [3]. A nationwide lockdown (Alert Level 4) began at 11.59 pm on the 25th of March 2020, requiring anyone not deemed an essential worker to go into self-isolation with their household or ‘bubble’ until the 27th of April 2020 11:59 pm at which time the country moved to Alert Level 3, partially lifting some of the lockdown restrictions [2]. On the 13th of May 2020, New Zealand moved to Alert Level 2 lifting the rest of the lockdown restrictions while maintaining physical distancing and gathering size limits [2]. On the 8th of June 2020, the Ministry of Health reported that there were no active cases of COVID-19 in New Zealand, which moved to Alert Level 1 removing all remaining restrictions except border controls [2].Many studies have explored the impact of quarantine and self-isolation due to disease outbreaks (including the COVID-19 pandemic) on humans for reviews see [4,5,6], but few studies have examined the impact on our companion animals (pets). Recent studies have reported that while pets provided their owners with a positive presence and support during the COVID-19 pandemic, some pets also showed stress-related behavioural changes due to the disruptions to their usual environment and routine [7,8,9,10]. Emerging pet behavioural issues along with other unique hardships pet owners face due to COVID-19 lockdowns (eg lack of access to animal supplies and services, exacerbated financial/economic issues) increases the risk of inadequate care and relinquishment or abandonment, resulting in negative impacts on the well-being of both the owner and pet [7,11]. This study investigated the impact of the nationwide lockdown on pet wellbeing in New Zealand as perceived by pet owners and discusses strategies to mitigate issues experienced by pet owners and to reinforce positive human behavioural changes that strengthen the human-animal bond, thus improving both animal and human wellbeing. Given that there is evidence that positive relationships with companion animals can be beneficial to the health and wellbeing of their owner and mitigate some of the detrimental effects of the COVID-19 lockdowns [7,8,9,10], this One Welfare approach promotes healthy communities during a global pandemic.2. Materials and MethodsData Collection: Online questionnaires were used to collect a convenience sample of respondents. The questionnaires used in Survey 1 and 2 were developed by Companion Animals New Zealand (CANZ) in consultation with One Welfare researcher Professor Natalie Waran from the Eastern Institute of Technology, Hawke’s Bay, New Zealand. Survey 1 was designed to inform CANZ’s social media information inventions to address pet owner concerns during and after the Alert Level 4 lockdown that were highlighted by the data collected. Both questionnaires included multiple-choice, Likert-like scale, and free text questions. A full copy of the questionnaires used for each survey can be found in the supplementary material (Figures S1 and S2). The questionnaires were administered online by distributing the survey link via social media and other networks. Survey 1 was open from the 24th of April to the 30th of April 2020, while Survey 2 was open from the 9th of July to the 28th of July 2020. The respondent inclusion criteria were New Zealand resident, pet owner and over the age of 18. Survey completion was anonymous and therefore it could not be determined if a person had completed both surveys or not.Data Analysis: The quantitative data analysis was performed using SPSS Statistics Version 25 [12]. The data have been described using frequencies, percentages. Where appropriate, column proportions were compared using z tests (independent proportions) to identify any significant differences. All statistical analyses were performed with a significance level of 0.05.The qualitative data were either analysed for emergent themes by at least two of the authors independently and then collaboratively categorised using a general inductive approach [13] or analysed using conventional content analysis [14]. Using content analysis, the presence (frequency) of reoccurring descriptions or concerns were quantified.3. Results3.1. Survey 1—During Alert Level 4 (Lockdown)The survey was completed by 686 respondents. Table 1 summarises the percentage of pet type/s owned.In 2020, cats are reported to be the most popular companion animal in New Zealand being part of 41% of households, with dogs being part of 34% [15], therefore, the convenience sample obtained does not appear representative of New Zealand pet owners.Respondents were asked on a scale of 1 to 5 if they were spending more time at home during lockdown (1 = No more than normal, 5 = Far more than before). The results are shown in Table 2, with 82.5% (n = 566) of respondents selecting 4 or 5 indicating the adults in the household were spending a lot more time at home compared to their pre-lockdown normal. Thirty-two percent of the respondents (32.4%, n = 222) also had children at home during lockdown that would have normally been at school.When asked if respondents felt that their pets had experienced any positive benefits from lockdown, 681 answered and 84.0% (n = 572) responded affirmatively. The themes that emerged from the perceived benefits for their pets described by the respondents were: differences between cats and dogs, for example, “the dog likes having me home but the cats couldn’t care less”; having company (subthemes: fewer movement restrictions such as not being crated or locked outside, more mental stimulation); more/better quality attention (subthemes: exercise, play, training, physical affection, grooming, treats, health surveillance/care, bonding), environment benefits (subthemes: fewer cars on road/less chance on being hit, less outside noise for example from vehicles and roadworks, more movement freedoms, access to heated areas on cold days).When asked if respondents felt that their pets had experienced any negative impacts from lockdown, 681 answered, and 52.1% (n = 355) of respondents felt their pets had. The following themes emerged from the respondent explanations: Reduced quality of dog exercise (subthemes: less socialisation, restricted to on lead, restricted exercise areas/more usage by community members); disrupted routines (subthemes: less time to themselves, interrupted sleep/relaxation, excess attention, no car rides); environment changes (extra inside noise, arguing humans); health concerns (subthemes: restricted access to animal services and products, overfeeding, excess exercise injuries, more anxiety/stress, less money to provide for animal needs).Using a Likert-like scale (1 = wellbeing is far worse, 5 = wellbeing is much better), respondents were asked to rate pets’ wellbeing. The results are shown in Table 3, with 54.7% (n = 375) of respondents selecting 4 or 5 indicating better wellbeing during lockdown while 7.4% (n = 51) of selected 1 or 2 indicating they though their pets’ wellbeing was better pre-lockdown. No significant differences were detected between cat and dog owners.Of the 680 respondents that answered the question “Have you noticed any changes in your pets’ behaviour during lockdown?”, 56.9% (n = 387) responded that they had. Content analysis revealed that being more needy/clingy was the most often described behavioural change (n = 75). This was followed by pets being more affectionate (n=33), calmer/more relaxed (n = 30), and happier (n = 21). Conversely, some pets became more nervous/anxious/worried/stressed (n = 26), made more noise such as barking, whining, meowing (n = 30) and were more demanding (n = 17).For most of the respondents (74.1%, n = 508) veterinary care was not required for any of their pets during the Alert Level 4 lockdown. Veterinary care for at least one of their pets was obtained by 26% (n =178) of the total respondents, with 30.3% (n = 54) of the consultations occurring via phone or email.Respondents were asked on a scale of 1 to 5 how much exercise had their pets been getting during lockdown compared with pre-lockdown (1 = A lot less, 5 = A lot more). The results are shown in Table 4, with 44.8% (n = 307) of respondents selecting 4 or 5 indicating their pets got more exercise during lockdown while 12.5% (n = 86) selected 1 or 2 indicating their pets got more exercise pre-lockdown. Similar results are obtained when the data from only those respondents owning dogs is analysed. Most of the respondents (84.4%, n = 579) felt that the guidelines for dog walking during lockdown were reasonable, with 7.7% (n = 53) of respondents being unsure and 7.9% (n = 54) feeling the guidelines were not reasonable.Using a Likert-like scale (1 = a lot less, 5 = a lot more), respondents were asked how much time they spent playing with their pets compared to pre-lockdown. The results are shown in Table 5, with 61.7% (n = 423) of respondents selecting 4 or 5 indicating they played more with their pets during lockdown while 2.3% (n = 16) of selected 1 or 2 indicating they played with their pets less.Figure 1 shows how respondents changed how they feed their pets during the Alert Level 4 lockdown. Most of the respondents (67.3%, n = 462) did not change the way they feed their pets during lockdown. Eighteen percent of the respondents (18.2%, n = 125) reported giving their pets more treats during lockdown but only 7.1% (n = 49) thought they were feeding more food overall.Of the 682 respondents that answered the question “Are you concerned about the future wellbeing of your pets after lockdown?”, 40.3% (n = 275) expressed that they were. Content analysis identified that the most commonly described concerns were that pets would miss having company and attention (n = 100) and experience separation anxiety (n = 83). Other common concerns included access to animal or veterinary services and products (n = 32) and issues with dog socialisation around other animals and people (n = 23). Sixty-four (n = 437) percent of the 682 respondents expressed that there were concerned about the future wellbeing of New Zealand pets generally after lockdown. The respondents’ primary concerns were separation anxiety (n = 125) along with loneliness and boredom (n = 104), the impact of financial hardship on caring for animals (n = 56), an increase in unwanted and inadequately cared for animals including pets that were acquired during lockdown and unwanted kittens and puppies due to a lack of desexing during lockdown (n = 47) and a decrease in exercise levels especially for dogs (n = 22).3.2. CANZ Social Media InterventionsIn response to pet owner feedback and the results of Survey 1, social media information interventions were provided via Facebook by CANZ (https://www.facebook.com/CompanionAnimalsNZ/, accessed on 10 March 2021) targeted at specific concerns that had been identified. Selected examples included the following:(1)CANZ-Accredited Animal Behaviour Consultants Help with Lockdown Behaviour Problems—a 55-min-long livestream talking about behaviour problems related to lockdown. This reached 2494 news feeds with 610 people clicking on the post and 106 likes/comments/shares. The average video watch time was 0.39 s.(2)Lockdown Behavioural Changes—a 7-min-long video talking about behaviour problems related to lockdown. Reached 3201 newsfeeds with 199 people clicking on the post and 97 likes/comments/shares. Average video watch time of 17 s.(3)Figure 2 shows an example of a CANZ post. This example reached 11855 newsfeeds and had 33 shares, 138 likes, 19 emoji ‘loves’, 24 emoji ‘sads’, and 31 general comments.(4)The “Eventually life will go back to normal” infographic (Figure 3) was CANZ’s most engaged with post with 110627 people being reached, 265 likes, 1765 loves, 282 comments, and 613 shares.3.3. Survey 2—During Alert Level 1 (Post-Lockdown)The survey was completed by 498 respondents. Table 1 summarises the percentage of pet types/s owned.Table 6 shows the answer selected when respondents were asked: “Do you feel that your pets’ wellbeing was better during Alert Level 4 lockdown or during Alert Level 1?”. Forty-one percent (n = 204) of respondents felt their pets’ wellbeing was better during Alert Level 4 while 16.8% (n = 84) felt their pet’s wellbeing was better during Alert Level 1 post-lockdown.Table 7 summarises the percentage of respondents that observed specified pet behaviours during Alert Level 4 (Lockdown) and Alert Level 1 (post-lockdown). Forty-three percent of respondents did not report any change in the pet behaviours listed at Alert Level 4, while 50% did not report any change in behaviour for Alert Level 1. The most frequently selected behaviour changes observed during Alert Level 4 were “Being calmer/more relaxed than normal” and “Being more affectionate than normal” with both these behaviours being observed by significantly more respondents when compared to Level 1 (Table 6). Conversely, significantly more respondents observed the following behaviours in Level 1 post-lockdown when compared to Level 4: “Being more needy/clingy than normal”, “Toileting issues” and “Showing signs of separation anxiety” (Table 6). Twelve percent (11.8%, n = 59) of respondents sought advice after Alert Level 4 for pet behavioural issues. Of these, 30.5% (n = 18) consulted with an animal behaviour professional, 22.0% (n = 13) sought advice from family and friends and 69.5% (n = 41) sought advice from other information sources such as the internet and social media.When asked, “What, if anything, did you do to prepare your pet to transition from Alert Level 4 to Alert Level 1?”, 64.3% (n = 313) respondents wrote that they did nothing while 35.7% (n = 174) described how they prepared their pets. Similar percentages were observed if calculated for cat owning respondents and then dog owing respondents (30.8 and 39.8%, respectively). Many of those that prepared their pets for the transition gradually increased time away from their pets (n = 74), while some respondents either maintained their normal routine throughout the various alerts levels or gradually returned to their pre-lockdown routine before returning to work (n = 21), other respondents chose to work more hours from home (n = 21). Some of those returning to work noted deliberately spending more time with their pets when they were home (n = 12) and enriching their pets’ environment, for example, with toys and interactive treat feeders (n = 12). Dog owners also described gradual re-socialisation with other dogs and people (n = 20), gradual re-introduction to doggy daycare (n = 7) and hiring dog walkers (n =2).Respondents were asked on a scale of 1 to 5 how much exercise have their pets been getting during Alert Level 1 post-lockdown compared with pre-lockdown (1 = A lot less, 5 = A lot more). The results are shown in Table 8, with 15.8% (n = 69) of respondents selecting 4 or 5 indicating their pets got more exercise during Alert Level 1 when compare with pre-lockdown, while 11.6% (n = 58) of selected 1 or 2 indicating they though their pets got less exercise. The majority of the respondents indicated that during Alert Level 1 (72.5%, n = 361) their pets exercise levels were similar to pre-lockdown. Similar results are obtained when the data from only those respondents owning dogs is analysed.Using a Likert-like scale (1 = a lot less, 5 = a lot more), respondents were asked how much time they spent playing with their pets compared to pre-lockdown. The results are shown in Table 9, with 27.9% (n = 139) of respondents selecting 4 or 5 indicating they played more with their pets during Alert Level 1 post-lockdown compare to pre-lockdown, while 6.6% (n = 33) of selected 1 or 2 indicating they played with their pets less.Figure 4 shows how respondents changed how they feed their pets during Alert Level 1 post-lockdown compare to Alert Level 4. Most of the respondents (75.7%, n = 377) did not change the way they feed their pets post-lockdown. For those respondents that selected one of the options that indicated they did change the way they feed their pets, 116 explanations were provided. A content analysis revealed those feeding less overall food were doing so because their pets needed to lose weight gained over lockdown (n = 19). Health issues that arose during lockdown required diet changes for some pets (n = 11) as did life stage changes (puppy to adult, adult to senor diets; n = 5). Some respondents changed to cheaper pet food (n = 2), while others changed their pet’s diet due to availability (n = 10) or because they found a better diet option (n = 4). Some respondents reported changing feeding times to fit with their work schedules (n = 27), feeding more treats/food when at home either after work or due to working from home (n = 31), while other respondents noted they were feeding their pets less treats now they were back at work and not home as often (n = 10).Of the 489 respondent that answered the question “Are you concerned about the impact of the COVID-19 pandemic on your pet’s wellbeing?”, 25.7% (n = 128) expressed that they were. Most of these respondents chose to explain what their concerns were. Content analysis identified that the most common concerns were that pets would miss having the company and attention they had during lockdown (n = 24) and experience separation anxiety (n = 15). Another common concern was the lack of access to animal or veterinary services and products during lockdown and the ongoing impact of this (n = 21). Specific to dogs were concerns about a lack of socialisation around other animals and people (n = 31) and the lack of off-lead exercise (n = 7). Other pet concerns mentioned included the impact of financial hardship (n = 2), the impact of routine changes (n = 8), pets being more needy/clingy (n = 5), pets being more anxious/jumpy around people (n = 4), toileting issues (n = 2), pets being overweight (n = 3), pets being COVID carriers (n = 2) and a lack of car sense (n = 1).Sixty-seven (n = 326) percent of the 487 respondents also expressed that they were concerned about the impact of the COVID-19 pandemic on the wellbeing of New Zealand pets generally. An inductive thematic analysis of the descriptions of the concerns identified the following main themes: change of routine related issues (subthemes: missing company and affection, confusion, separation anxiety post-lockdown, pet acquire going into lockdown being neglected or abandoned/rehomed post-lockdown, less care of pets post lockdown including exercise, increased domestic violence/abuse during lockdown including pets), financial hardship related issues (subthemes: not being able to supply pet needs/neglect, increased human stress affecting pets, unable to afford veterinary and animal services and products, pet abandonment/rehoming), reduced access to veterinary and animal services and products (less desexing/more unwanted pregnancies, less grooming, feed changes), concern about pets carrying COVID-19, and the dog-specific issues due to lack of socialisation (eg due to lack of club activities, leashed walks only).4. DiscussionThe results from this study indicate that pets may have enjoyed improved welfare during the Alert Level 4 lockdown in New Zealand due to the possibility for increased human-pet interaction, with both adults and children spending a lot more time at home in line with the self-isolation directive from the Government. Most respondents could list at least one perceived benefit of lockdown for their pets. These included more company, more or better-quality attention (exercise, play, training, physical affection, grooming, treats, health surveillance/care, bonding), and environmental benefits (fewer cars on the road, less outside noise, more movement freedoms, and access to home comforts). Almost half of the respondents indicated that their pets got more exercise during lockdown and almost two-thirds of respondents indicated they spent more time playing with their pet. Just over half of the respondents thought that their pet’s wellbeing was better than usual during lockdown, with most of the remainder indicating that their pet’s wellbeing had not changed. Pet feeding practices were largely unchanged across the different alert levels; however, animals having gained weight was acknowledged as a common concern among respondents.The findings of this study are consistent, in part, with a study of the effects of the Spanish COVID-19 lockdown on pets and their owners by Bowne and colleagues [8] which found the 57.3% of cat owners perceived their pets’ quality of life to be better during lockdown and 34.3% indicating that did not think their cat’s quality of life had changed. However, in contrast to our study, Bowne and colleagues [8] found that 62.1% of the dog owners surveyed thought their dog’s quality of life had got worse with a reduction in the number of walks and duration being reported. The Spanish lockdown was very similar in its restrictions (including those around dog walking) to the Alert Level 4 lockdown in New Zealand, but a notable difference between the two countries is that three-quarters of the Spanish respondents lived in apartments and the level of community transmission and thus the risk of infection was far greater in Spain. This may mean that while the restrictions on dog walking were similar in both countries, the level of impact of COVID-19 on dog walking may have been greater in Spain than in New Zealand. This is supported by Brown and colleagues reporting that before their lockdown, dogs went on an average of 3 walks per day, while during lockdown, they went on an average of 2.5 walks per day. By contrast, our study shows 49.7% of respondents with dogs were exercising them more than before lockdown. Our study also contrasted with the study by Christley and colleagues [16], which showed that in the UK during lockdown dogs were walked both less frequently and for shorter durations. However, this can likely be attributed to the difference in restrictions on outdoor exercise between New Zealand and the UK, where people were limited to one form of outdoor exercise per day.Although the New Zealand Alert Level 4 lockdown was generally viewed as beneficial for pet wellbeing, just over half of the respondents identified negative impacts on their companion animals which included a reduction in the quality of dog exercise, disrupted routines, social and environmental changes, and health concerns. One driver of health concerns was less access to animal services and products. These results are consistent with the observations of Bowne and colleagues that found the most common concerns for dog owners was dog walking restrictions followed by loss of routine, and for cat owners, access to veterinary care and medication [8]. Approximately half of the New Zealand respondents noticed behavioural changes in their pets during lockdown, with pets commonly being described as more needy/clingy/attention-seeking, followed by pets being described as more affectionate. Similar numbers of respondents described pets as being calmer and more relaxed and/or happier compared to those that described their pets as making more noise, becoming more nervous/anxious/worried/stress and or being more demanding. The Spanish COVID-19 lockdown study by Bowne and colleagues also observed an increase in stress-associated behavioural changes, especially in the group of owners that were coping less well, implying that household stresses and people’s abilities to cope with them have an impact on pet wellbeing [8]. Morgan and colleagues also found there was an association between dog owners who felt their quality of life was impaired and the perceptions of reduced quality of life for their pet and the development of new behavioural problems [11]. Collectively, these findings are similar to those of a study on pet owners in the U.S. that focused on the negative impacts of COVID-19 on caring for pets, which concluded that pet owners experience unique COVID-related hardships, which need to be addressed in order to manage pet owner expectations, prevent problem pet behaviours, improve owner well-being and pet welfare [7].Just over two-thirds of respondents expressed that they were concerned about their pet’s wellbeing after lockdown, with pets missing company/attention and separation anxiety being major themes. This level of concern is higher than that observed in the Spanish study, where approximately 40% of cat and dog owners surveyed were concerned their pet would not adapt to the situation after confinement ended [8]. In this study, respondents were also concerned about the wellbeing of New Zealand pets generally after lockdown. In addition to concerns about separation anxiety, loneliness and boredom, respondents were concerned about the impact of financial hardship on caring for animals and also an increase in unwanted and inadequately cared for animals.Research has shown that companion animals can help some people cope with challenging situations and this was also found to be true during the COVID-19 lockdown [7,8,10,17]. However, the human-animal bond is complex and social exchange theory [18] has been used to explain this dynamic relationship, the product of which is the balance between perceived cost and benefits [8,10]. Globally many people chose to take on new pets going into lockdown [11,17,19]. Presumably, this decision was a consequence of the perceived social and health benefits outweighing the perceived costs of pet ownership. However, pets are a vulnerable sector of society as they are almost fully dependent on human care, and concerns about the long-term care of newly adopted animals have been raised in the literature [19]. Changing alert levels can alter the perceived costs and benefits of pet ownership [17], thus impacting on pet wellbeing. Vincent and colleagues highlight that the cost of caring for an animal during economic hardship or the burden of a daily care routine may result in the relinquishing of animals to a shelter both during and post-lockdown [17]. These pet-specific costs are potentially increased by animal behavioural issues that might have developed during lockdown or in groups of vulnerable people such as those infected with COVID-19, the elderly, those with mental health disorders or those with chronic illnesses, who might need additional support to care for their pets adequately. Thus, a One Welfare approach should be considered for the maintenance of the human-animal relationships formed or developed during the lockdown period. It has also been suggested in the academic literature that noticing and amplifying micro-practices that make a positive difference to non-human animals via social media may facilitate improve animal welfare practices [20].In keeping with this, CANZ provided educational information via Facebook to address pet owner concerns, manage pet owner expectations and promote positive practices during the COVID-19 pandemic. Social networks allowed animal professionals to provide users with important COVID-19 related information impacting on animal welfare. To ensure user confidence, on 17 March 2020 the major social network companies declared that they would work closely together to combat false and misinformation and promote information from official sources [21]. Four levels of social network engagement have been acknowledged: (1) observer, (2) follower, (3) participant and (4) defender which can be further condensed into two levels of: passive engagement (level 2) and cognitive participative engagement (levels 3 and 4) [22]. Of the CANZ posts, infographics were engaged with by users more extensively than video resources. This contrasts with marketing research which reports that video posts perform best on Facebook [23]. COVID-19-specific research in Canada found that the type of embedded media used in Facebook posts was a determinant of engagement with public health communications, and that posts with simple, concise messages and high-quality media embedded, such as animations or infographics, were especially well shared [24]. Conversely, this research found that that extensive policy-related messages and video and infographics of lower quality were associated with less engagement.Facebook users engaged with the CANZ infographics on the participatory level with likes, loves, shares, and comments. Compared to the infographics, the videos had less engagement and required longer viewing times (7–55 min) with the highest average watch time being 17 s. Interestingly, TikTok, a video sharing platform that has gained prominence during the COVID-19 pandemic uses a short-form format of 15 s [25]. Pérez-Escoda and colleagues highlight that social audiences need to be part of the media flow [22] which was reinforced by Raamkumar and colleagues, who observed that Facebook users engaging with content from health authorities were more likely to share a post aiding the dissemination of information than they were to react or comment [26]. It could be speculated that the anticipated uptake by one’s followers influences how users interact with content and in this respect those resources that require less time to view may be perceived as having better uptake by other social network users gaining more positive reactions (e.g., like, care, love), comments and shares. Further research into social network user engagement and uptake of information regarding companion animal wellbeing is needed.In the Alert Level 1 survey, most respondents indicated that they thought their pet’s wellbeing was the same as it was in the Alert Level 4 lockdown or better. In Survey 1 more than half of the respondents thought their pet’s wellbeing was better during lockdown (Level 4) than it was pre-lockdown. It, therefore, suggests that Alert Level 1 life for many pets was perceived by their owners to be better than that of pre-lockdown. Half of the respondents did not observe any significant behaviour changes between Alert Level 4 and Alert Level 1 and just over one-third of respondents took steps to prepare their pets to transition out of lockdown with the internet and social media being a major source of information. It can, therefore, be speculated that pet-specific information provided via organisations such as CANZ using social media helped to re-adjust pet owner expectations and facilitate practices that resulted in pet-friendly Alert Level transitions.In Survey 2, a quarter of the respondents expressed they were concerned about the impact of the COVID-19 pandemic on their pet’s wellbeing, significantly less than in Survey 1, suggesting that concerns had been mitigated. Interestingly, a similar percentage of respondents still expressed concern about the impact of the COVID-19 pandemic on the wellbeing of New Zealand pets generally, as did about the general impact of the lockdown on pets. Most concerns related to issues brought about by changes in routine (for both humans and their pets), financial hardship, and reduced access to veterinary and animal services and products. Dog-specific issues due to lack of socialisation were also highlighted as a concern. Again, this is reflective of the concerns of pet owners from other parts of the world [7,8]. There was very little concern about pets being involved in the transmission of COVID-19 among the New Zealand survey respondents which is in line with the current research evidence [27,28] and suggests accurate uptake of this information.A significant determinant of an animal’s wellbeing is the balance between their experience of negative and positive affective states [29]. Several of the changes made by owners as they came out of lockdown, such as spending more time in play with their animal, working more from home, spending more time with pets, providing environmental enrichment, and employing dog walkers provide opportunities for pets to express positive emotions such as excitation and playfulness, the feeling of vitality and pleasure that comes with exercise, and the rewarding feelings that come from spending time with bonded humans. It is, therefore, reasonable to assume that at the time these changes were made the owners were enhancing their pet’s wellbeing. Further research is required to determine whether these changes and their associated positive effects on animal wellbeing, will be maintained long term.LimitationsDemographics were not collected therefore conclusions about sample representation cannot be made. Caution needs to be taken when extrapolating these results to all New Zealand pet owners as online surveys are typically taken by white women of higher socioeconomic status [30,31]. In these surveys, the respondents were free to answer questions taking all their pets into consideration as opposed to selecting only one of their pets as being the focus of their responses.5. ConclusionsThe results from this study indicate that pet owners perceived their pet’s wellbeing to be better during and after the Alert Level 4 lockdown in New Zealand when compared to pre-lockdown. This may have been due to improved welfare facilitated by the possibility for increased human-pet interaction along with the rapid increase in social media use providing animal professionals with the opportunity to promote positive human-pet interaction messages using a platform such as Facebook. The steps that were taken to prepare animals for a return to normal life may enhance pet wellbeing long-term if maintained.

animals : an open access journal from mdpi

10.3390/ani11102787

PMC8532674

Neck and back pain are common ailments in animals. While there are medical and surgical treatment options available for select patients, conservative care is the most common form of management of pain, stiffness and muscle spasms. Physical therapists, osteopaths and chiropractors use mobilization and manipulation techniques to evaluate and treat muscle and joint problems in both humans and animals, but there seems to be little scientific evidence available to support their use in veterinary medicine. This study reviews the scientific literature with the goal of identifying the clinical indications, dosages, outcome parameters, and efficacy of mobilization and manipulation techniques in dogs and horses. Fourteen articles were included in this review of which 13 were equine and one was a canine study. There was a large variability in the quality of evidence that supports the use of joint mobilization or manipulation in treating pain, stiffness and muscle hypertonicity in horses. Therefore, it was difficult to draw firm conclusions despite all studies reporting positive effects. Future studies need to establish standardized methods to evaluate the optimal dosages of mobilization and manipulation for use in animals.

Mobilization and manipulation techniques are often used in small animal and equine practice; however, questions remain concerning indications, dosing and efficacy. A bibliographic search was performed to identify peer-reviewed publications from 1980 to 2020 that evaluated the clinical effects of musculoskeletal mobilization and manipulation techniques in dogs, cats and horses. The search strategy identified 883 papers for review. Inclusion and exclusion criteria were applied. The clinical indications, dosages, outcome parameters, and reported efficacy within each publication were recorded and categorized for comparison with scientific quality assessed according to a standardized grading system. Fourteen articles were included in this systematic review of which 13 were equine and one was a canine study. Seven of these were cohort studies and seven were randomized controlled clinical trials. The canine study involved carpal immobilization-remobilization and all equine studies focused on the effects of passive mobilization (n = 5) or manipulation (n = 8) of the axial skeleton. Study quality was low (n = 4), moderate (n = 7), and high (n = 3) and included a wide array of outcome parameters with varying levels of efficacy and duration of therapeutic effects, which prevented further meta-analysis. Therefore, it was difficult to draw firm conclusions despite all studies reporting positive effects. Optimal technique indications and dosages need to be determined to improve the standardization of these treatment options.

1. IntroductionManual therapy is defined as the application of the hands to the body with a diagnostic or therapeutic intent [1]. Of the different types of manual techniques that have been used in veterinary medicine, soft tissue massage and joint mobilization or manipulation are the most common techniques applied to animals for the relief or pain, stiffness or muscle hypertonicity [2,3,4,5,6]. Mobilization techniques use graded forces to displace musculoskeletal tissues and can generally be categorized into soft tissue or articular-based approaches [7]. Soft tissue mobilization typically focuses on restoring physiologic motion to the skin and underlying fascia, ligaments, and myotendinous structures with the aim of reducing pain, increasing tissue extensibility, and improving function [8]. Soft tissue mobilization techniques are also used to diagnose and restore normal mobility to neural tissues (i.e., peripheral nerves) [9]. Joint mobilization is characterized as repetitive passive joint movements with the purpose of restoring normal and symmetric articular motion [7]. Manipulation is characterized by the application of a non-repetitive, high-velocity, low-amplitude thrust (HVLA) directed at spinal or appendicular articulations [8].The incorporation of manual therapies into veterinary practice has become a common approach for addressing neck, back and pelvic pain and dysfunction in both equine and small animal patients [10,11]. Individuals trained in chiropractic, osteopathic and physical therapy techniques use both mobilization and manipulation to address musculoskeletal and neurologic issues in animals [1]. As with most integrative therapies in veterinary medicine, there is often wide-spread clinical use without a strong body of evidence-based support. While there is a growing body of evidence to support the use of mobilization and manipulation techniques in equine practice, there is substantially less published within the small animal literature [12]. General reviews do exist for musculoskeletal mobilization and manipulation use in veterinary medicine; however, no systematic reviews have been completed to date [1,13,14,15]. Analysis of the current scientific literature would provide insights into the clinical indications and effectiveness of mobilization and manipulation in an effort to improve guidelines for their application in managing musculoskeletal disorders. The objective of this systematic review is to describe the literature that has been published relative to mobilization and manipulation techniques in dogs, cats, and horses as a sole treatment modality. The research questions under investigation included: what are the (1) clinical indications, (2) dosages used, (3) outcome parameters, and (4) perceived efficacy of musculoskeletal mobilization and manipulation.2. Materials and MethodsA systematic review process was conducted as outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16]. Studies were located by professional librarians who performed systematic electronic database searches of the Web of Science, CABI and PubMed in August 2020 for articles published between the years of 1980 and 2020. The following keywords were used in combination with Boolean operators for database searches: dog OR cat OR horse, AND veterinary medicine OR veterinary, AND therap* OR treatment*, AND mobilization OR manipulation OR chiropractic OR osteopathy. Publication date limitations were not implemented. Duplicate references were removed.Articles were screened for relevance by a single author (KH) and studies unrelated to musculoskeletal mobilization or manipulation were excluded. There were no restrictions regarding the language of publication at the initial search stage. After the first stage of screening, articles deemed potentially relevant were accessed from open-access sources or via inter-library loan via Colorado State University. The resulting list of potential studies was screened by all authors against inclusion and exclusion criteria. The inclusion criteria were publications that must (1) be a primary research publication in a peer-reviewed journal or conference proceedings; (2) report on the treatment effects of a single treatment modality; and (3) describe treatment efficacy for a single clinical disorder or related outcome parameters. The exclusion criteria were studies that were (1) single case reports, textbook chapters, editorials and reviews; (2) involved more than one type of treatment; (3) basic science research exploring mechanisms of action; (4) related to the manual reduction of joint luxations; and (5) focused on rodent models or other animal species (e.g., sheep or pigs). Additional hand searching of the bibliographies of eligible records and review articles were examined for studies not retrieved by database and repository searches.Each study included in the full-text review was screened for relevance and categorized according to species (canine, feline, or equine) and the type or application of mobilization or manipulation used. Data extracted included the name of first author, year of publication, study design, species, number of animals included, inclusion and exclusion criteria, intervention (dose, interval, duration), controls, follow up period, dropout rate, and treatment results. When more than one outcome parameter was available, the parameter which provided the most clinical relevance was extracted for analysis. The overall study quality was scored using a checklist that included: sample size, confounding factors, selection bias, deviations from planned therapy, dropout rate, blinding, and external validity.3. ResultsA total of 5529 records were identified via the three combined electronic database searches (Figure 1). Following removal of duplicate records, 883 records were screened for relevance to the review. After title and abstract screening, a total of 149 publications that investigated treatment of musculoskeletal issues with mobilization or manipulation techniques were evaluated in full. Hand searching bibliographies from eligible records and review articles provided 51 additional records of which 2 were judged relevant based on inclusion-exclusion criteria. A large proportion of studies were excluded as review articles and textbook chapters; however, the largest number of excluded articles involved basic science research using cats to explore neurophysiologic mechanisms of action for spinal manipulation (n = 40) [17,18].After completion of the selection process, 14 articles were retained of which 13 were equine and 1 was a canine study (Table 1). There were seven experimental studies and seven observational studies. The canine study involved induced carpal immobilization with subsequent remobilization [19]. All equine studies focused on the effects of passive mobilization (n = 5) or manipulation (n = 8) of the axial skeleton involving either naturally occurring disease (n = 12) or induced back pain (n = 1). Of these equine studies, two were characterized as whole body or generalized treatments and 11 involved regional or local articular treatments.Most mobilization studies were cohort-based study designs (5 of 6) with four of these studies being prospective in nature and two retrospective (Table 2). In contrast, the spinal manipulation studies were mostly randomized, controlled clinical trials (6 of 8). Objective outcome parameters were used to assess treatment efficacy in 11 studies and owner surveys were used in the three equine spinal mobilization studies [20,21,22].The overall quality scores were graded low (n = 4), moderate (n = 7), and high (n = 3) across studies. The three equine osteopathy studies were judged to be of low quality due to their retrospective design and the sole reliance on unspecified owner questionnaires collected 6–18 months after treatment [20,21,22]. The randomized, controlled clinical trials that used objective measures provided the highest quality evidence regarding clinical efficacy [26,27,29].3.1. Clinical Indications3.1.1. MobilizationThe canine study evaluated the experimental effects of carpal immobilization with remobilization [19]. The equine mobilization studies included three osteopathic reports on the treatment of axial skeleton pain and stiffness [20,21,22]. Medical histories often included behavioral or temperament changes, apprehension when saddled, and reduced ridden performance [22]. Static physical examination findings included the inability to stand squarely on all four limbs, epaxial muscle hypertonicity or atrophy, and signs of pelvic asymmetry. Subjective observation was used to assess changes in stride length, inconsistent limb placement, asymmetric pelvic motion, head and neck elevation combined with trunk lordosis, and the inability to back up in a straight line [22]. A whole body mobilization study in horses with acute back pain assessed the effect of caudal tail traction indicated by active trigger points localized within the longissimus and middle gluteal muscles [24]. In a second whole body mobilization study, the effects of induced caudal weight shifting with manual force applied at the point of the shoulder was evaluated in normal horses [28].3.1.2. ManipulationThe equine spinal manipulation studies primarily evaluated changes in thoracolumbar nociceptive thresholds (i.e., back pain) and concurrent trunk stiffness and epaxial muscle hypertonicity [10,23,25,26,27,29,30,31]. Experimental studies of equine manipulation included HVLA thrusts applied bilaterally at standardized thoracolumbar locations in actively ridden horses participating in collegiate programs [26,27,29] or with experimentally-induced spinous process pain [30]. In contrast, clinical studies typically addressed HVLA treatment at variable vertebral locations with localized signs of pain or joint stiffness within individual horses [10,23,25,31].3.2. Dosages3.2.1. MobilizationIn the canine immobilization-remobilization study, carpal flexion, traction, and craniocaudal translation was applied using 3 sets of 20 oscillations, daily for 4 weeks [19]. Equine osteopathic treatments included the use of intravenous sedation (82% of cases) to improve patient compliance and spinal mobilization under general anesthesia (17% of cases) for intransigent pain or dysfunction [21]. The number of treatment sessions using standing sedation averaged 6 treatments (range 1–14) [22]. Cervical mobilization under anesthesia typically involved 1–2 treatment sessions (range 1–6) [20,21]. Post-treatment recommendations often included stall rest and NSAIDs administration for 3–5 days, with restrictions on ridden exercise, and work in hand for 4–8 weeks before return to full work in 4–6 months [21,22]. Caudal tail traction in horses was applied in line with the slope of the croup with three repetitions of a steady applied force of 4.5 kg for 20 s, followed by release for 10 s within a single treatment session [24]. In a second study, caudal weight shifting was induced by applying a caudally-directed manual force to the point of the shoulder in an unweighted forelimb until firm resistance was achieved and the force was then held for 5 s [28].3.2.2. ManipulationThe frequency for applied HVLA treatments varied from a single session [10,25,27,30,31], to once daily over 3 to 5 days for acute back pain [23], and once weekly for 3 weeks for horses with chronic back pain or stiffness [26,29].3.3. Outcome Parameters3.3.1. ObservationThe clinical examination of horses often included static observation of limb positioning or stance, spinal posture, and epaxial muscle asymmetries [22]. Dynamic observation included gait evaluation in hand while moving in straight lines, circles and while backing up.3.3.2. Physical ExaminationDetailed spinal evaluations of the trunk and pelvis were completed before and after a series of three HVLA treatment sessions in horses with acute back pain [23]. The number of affected thoracolumbar and sacral vertebral segments and the severity of epaxial muscle pain and hypertonicity and segmental trunk stiffness in lateral bending were recorded. Firm digital pressure was used to identify painful sites over thoracolumbar (T4–L6) and sacral (S2–S5) spinous processes and to localize epaxial muscle pain and tone within the thoracolumbar and gluteal regions (T4–S5). Trunk stiffness was identified using low amplitude lateral spinal oscillations applied segmentally at each thoracolumbar (T10–L6) vertebral level. Left-right asymmetries in the prevalence and severity of the spinal examination findings were recorded [23]. In a second spinal manipulation study in horses, myofascial sensitivity localized with a diagnostic acupuncture examination was used [10].3.3.3. Muscle Tone and ActivityChanges in thoracolumbar epaxial muscle tone were assessed using a tissue indenter in an equine spinal manipulation study [27]. Electromyography (EMG) of the longissimus muscles was recorded during standing and walking to assess changes per and post spinal manipulation [27]. In a second spinal manipulation study, static bioimpedance and dynamic acoustic myography were used to measure changes in epaxial muscle activity in actively ridden horses [10].3.3.4. Spinal ReflexesSpinal and pelvic responses to applied digital stimulation were used to assess active spinal mobility, coordination, and core strength in horses with acute back pain that were treated with spinal manipulation [23]. Graded responses to applied truncal stimulation were scored based on the quality, amplitude, and the ability to statically hold the induced postures. Digital stimulation was applied along the ventral midline over the sternum or cranial portion of the linea alba to induce elevation of the cranial thoracic region. Bilateral digital stimulation adjacent the lateral tail head was used to induce a combined reflex of pelvic flexion and trunk elevation (i.e., kyphosis). The response to firm medial compression of both tuber sacrale was scored based on the presence of a pain avoidance response and unilateral or bilateral unlocking of the stifles. Applied axial traction to the tail was used to theoretically assess core stability and neuromuscular coupling of the lumbosacral region [23].3.3.5. Mechanical Nociceptive ThresholdsThe effect of caudal tail traction on back pain in horses was measured using pressure algometry to detect mechanical nociceptive thresholds (MNTs) across lumbopelvic landmarks [24]. Three of the equine spinal manipulation studies also used pressure algometry to measured pre- and post-treatment changes in thoracolumbar MNTs [23,25,26].3.3.6. Passive Joint Range of MotionThe effects of experimentally induced carpal immobilization-remobilization in dogs was measured with manual goniometry to assess passive joint range of motion and cinematographic analysis of peak flexion and extension angles of the carpus during active walking [19].3.3.7. Thoracolumbar FlexibilityThe vertical displacement of vertebral segments within the thoracolumbar region were measured using a cable extensometer mounted to an overhead rail system during both spinal mobilization and manipulation procedures in horses [29,30]. The inclusion of a pressure-sensitive mat to record applied forces during both spinal mobilization and manipulation allowed the calculation of changes in segmental stiffness pre- and post-treatment.3.3.8. Motion AnalysisAn equine study evaluated the effect of induced caudal weight shifting of the trunk measured cinematographic changes in the vertical displacement and dorsal trunk angles (i.e., lordosis) of skin markers placed along the dorsal midline in normal horses [28]. In a second equine study that evaluated the effects of spinal manipulation, spinal and limb kinematics during overground walking and trotting were measured using high-speed cameras to track three-dimensional skin marker displacements [31].3.3.9. Visual Analog ScalesAdditional clinical measures included visual analog scales (VAS) for assessing overall pain and spinal function in horses with acute back pain [23]. Owners and trainers were asked to score the overall severity of their horse’s back pain based on a VAS that was numbered from 0 to 10, with 0 representing the best case (e.g., no pain) and 10 representing the worst case (e.g., worst possible pain). A VAS was also used for examining the veterinarian’s perception of the global severity of back pain and the overall quality of spinal and pelvic function [23]. 3.3.10. Owner Surveys of PerformanceThe equine osteopathic studies relied on owner assessments of improvements in neck mobility and overall performance reported 6–18 months post-treatment [20,21]. The return to regular ridden work as reported by the owner was also used to determine the success of osteopathic treatment [22].3.3.11. ThermographyThermographic imaging of the cervical and thoracolumbar regions was used in one of the equine osteopathic studies where temperatures ≥ 1.5 °C cooler than surrounding areas or left-right temperature asymmetries, and loss of the normal dorsal midline thermal demarcation were considered abnormal [22].3.3.12. Lameness EvaluationAn inertial sensor system was used to assess objective signs of limb lameness on a straight line at a trot in horses treated with spinal manipulation [10].3.4. Clinical Efficacy3.4.1. Physical ExaminationThe authors of an equine osteopathic study reported that attempts to classify the degree of pain and dysfunction was considered too subjective to be useful, as the range and severity of clinical signs were too broad and nonspecific [22]. In Western performance horses with acute back pain, a nonsignificant increase (23%) in muscle pain was reported after a series of spinal manipulation sessions applied over 3–5 days, compared to a significant decrease in epaxial muscle pain in horses treated with low-level laser therapy [23]. There was an 83% increase in stiffness as measured by the number of affected vertebral levels in horses treated with a series of HVLA treatment sessions, indicating aggravation of clinical signs. There were no significant changes in the severity of spinous process pain between treatment groups or across treatment sessions [23].3.4.2. Muscle Tone and ActivityEpaxial muscle tone as measured by a mechanical tissue indenter decreased significantly after a single HVLA treatment session by 13%, compared to 0% change within control horses [27]. Similarly, muscle activity as measured by EMG showed a significant decrease of 21 ± 7% within the treatment group, compared to 6 ± 5% in the control group. A single HVLA treatment session abolished myofascial sensitivity assessed with digital palpation and improved measures of muscle function using static bioimpedance and dynamic acoustic myography for up to 3 days post treatment [10].3.4.3. Spinal ReflexesIn horses with acute back pain, significant improvements in the quality and amplitudes of spinal motion associated with induced thoracic (28%) and pelvic flexion (28%) reflexes were reported after spinal manipulation [23]. There were no significant treatment effects on the ability to resist axial traction applied to the tail or the response to tuber sacrale compression.3.4.4. Mechanical Nociceptive ThresholdsCaudal tail traction induced significant changes in MNT values across the 10 lumbopelvic landmarks with an overall increase of 11.6 N/cm2 (range 8.7–16.6 N/cm2) [24]. Within actively ridden horses without overt signs of back pain, MNT values were significantly increased by 27% after a single instrumented HVLA treatment session 1 week post-treatment, compared to <1% change within both the active and inactive control groups [25]. Spinal manipulation applied weekly for 3 weeks increased MNT values within the treatment group by 11 ± 7% and in the control group by 5 ± 6% [26]. Within the treatment location (i.e., T13–L6), the average increase in MNT values was a 11 ± 4% difference between the treatment versus control horses [26]. In horses with acute back pain, three spinal manipulation sessions produced a significant treatment group effect of 2.3% across pooled MNT values, compared to no significant improvement (−3.9%) when manipulation was combined with low-level laser therapy [23]. However, there was a significant combined manipulation and laser group difference in pooled MNT values from baseline to the third treatment session, but no significant percent change was noted within the spinal manipulation group.3.4.5. Passive Joint Range of MotionIn dogs, carpal motion after immobilization and remobilization also produced significant increases in passive range of motion amplitudes and peak carpal flexion-extension angles measured during while walking (i.e., active joint range of motion); however, the changes could not be definitively attributed to treatment [19].3.4.6. Thoracolumbar FlexibilityIn horses with induced back pain, vertical trunk displacements increased 15% (range, 7% to 25%) after HVLA, compared to 0% (range, –4% to 7%) in the control group [30]. In actively ridden horses without overt back pain, spinal flexibility as measured by passive vertical displacement increased 16 ± 7% immediately after HVLA treatment across 5 thoracolumbar sites (T14-L6), compared to 0 ± 3% with mobilization alone [29]. However, vertical displacement measured 1 week after spinal manipulation or mobilization, showed an increase in spinal flexibility of 10 ± 5% within the mobilization group, compared to 5 ± 4% in the spinal manipulation group. These findings suggest an immediate effect due to HVLA treatment versus a delayed effect on spinal flexibility associated with spinal mobilization. After 3 weeks of once weekly HVLA treatment sessions, vertical displacement increased by 40% from baseline values versus 19% with spinal mobilization alone [29].3.4.7. Motion AnalysisCaudal weight shifting in horses caused significant flattening (i.e., reduced lordosis) of the dorsal trunk contour as measured from T10 to L3 with overall changes in vertical displacement of 11 mm (range 1–20 mm) and thoracolumbar angles of 3.4° (range 0.2–7.2°) [28]. In ridden horses with back pain, a single session of spinal manipulation had minor, variable effects on vertebral and pelvic kinematics as measured at the walk and trot [31]. Thoracolumbar and pelvic range of motion tended to increase directly after treatment but was decreased 3 weeks later compared with baseline values. Specific changes included increased thoracolumbar sagittal motion and symmetry of pelvic rotation. No significant changes were noted in stride parameters or cervical vertebral motion patterns [31].3.4.8. Visual Analog ScalesIn horses with acute back pain, veterinarian-derived VAS of back pain severity and spinal function showed no treatment effect over three spinal manipulation sessions, compared to a significant improvement in these parameters within the laser therapy group [23]. Owner reported VAS scores decreased (i.e., reduced back pain) across sessions; however, the changes were not significant for spinal manipulation or combine manipulation and laser therapy.3.4.9. Owner Surveys of PerformanceThe treatment of equine neck pain and stiffness using mobilization techniques under anesthesia produced clinical improvement based exclusively on owner reports as early as 2 days, with 95% of horses improved within 2 weeks post treatment [20]. Based on owner reports, 95% of horses are reported to be improved at least 6 months post-treatment using osteopathic techniques with complete resolution in 74% and partial improvement in 26% [21]. In another equine osteopathic study using owner surveys, return to work was reported at 6–12 weeks in 90% of horses, which had all undergone prior unsuccessful conventional treatment [22]. Longer-term follow-up (>12 months) based on rider assessments showed 53% of horses continued in normal work, 31% worked at a lower level, and 16% were unrideable. These authors suggested that the success of treatment depends on owner’s ability to return horses slowly to work, reestablished patterns of normal tissue function with therapeutic exercises or rehabilitation, and repeated treatment using manual therapies [22].3.4.10. ThermographyIn an equine osteopathic study that reported treatment of a wide variety of clinical signs, normal thermographic patterns were noted in horses that had returned to regular work [22].3.4.11. Lameness EvaluationThere were no significant changes in objective measures of limb lameness in actively ridden horses treated with spinal manipulation [10].4. DiscussionThe objective of this systematic review was to analyze the small animal and equine literature for clinical indications, dosages, outcome parameters used, and the perceived efficacy of musculoskeletal mobilization and manipulation techniques. While there is a plethora of review articles addressing the use of manual therapies in veterinary medicine, there is a clear lack of available primary research [7,15]. Joint mobilization and manipulation are known to have effects via biomechanical and neurophysiological mechanisms in humans [32,33]. Numerous basic science studies have evaluated the biomechanical and neurophysiologic mechanisms of spinal manipulation in feline [17,34,35,36], ovine [37,38], and porcine [39,40] models, which made up the largest proportion of excluded literature within this review. Another factor that limited study inclusion was that mobilization and manipulation techniques are often combined with other modalities (e.g., acupuncture) [41,42].This systematic review reveals that there is a growing body of evidence that supports the use of spinal mobilization and manipulation in horses; however, there remains a critical deficit of published clinical trials in dogs. It is surprising that there are so few studies that have evaluated the effects of joint mobilization or manipulation within the distal limbs of dogs, cats and horses given the high prevalence of appendicular joint disease and associated stiffness that is often addresses with mobilization techniques and the relative ease of which goniometry can be applied in these body regions [43,44,45]. Regrettably, there are also few validated outcome parameters for both dogs and horses for assessing spinal examination findings such as palpable sensitivity, stiffness, and muscle hypertonicity in the clinical setting [23,46]. While functional questionnaires for assessing musculoskeletal or neurologic pain and dysfunction have been validated for use in dogs [47,48,49] and horses [50,51,52,53], these tools have not yet been applied to evaluating the clinical efficacy of mobilization or manipulation techniques.4.1. QualityThe overall study quality scores were low-to-moderate in 11 of 14 studies using the prescribed criteria. Within the observational studies, the largest variability in scoring was within the confounding factors, selection bias and external validity. Within the experimental studies the limiting factor was mostly the lack of blinding.4.2. Treatment MethodsA multimodal rehabilitation approach is common in managing musculoskeletal disorders in veterinary clinical practice and is often judged to have the greatest clinical impact; however, this makes the evaluation of the clinical efficacy of a single treatment modality difficult [41]. Most included studies addressed the effects of either mobilization or manipulation in isolation; however, two equine studies did include both forms of therapies [29,30]. In these two studies, spinal mobilization was required to provide measures of vertical trunk displacement and stiffness across both treatment and control groups whereas, spinal manipulation was only applied with the intent of inducing a treatment effect. Therefore, the treatment group received both spinal mobilization plus HVLA thrusts, which provided insights into the synergistic effects of these combined therapies over 3 sessions at weekly intervals [29]. Both spinal mobilization and manipulation were effective at increasing spinal flexibility; however, HVLA treatment produced larger increases within sessions, whereas the effects of spinal mobilization was delayed as evidenced by changes between sessions. These finding suggested that two possibly different mechanisms of action for spinal mobilization versus manipulation, which is consistent with the human literature [54].Manual therapy implies using the hand to diagnose or treat. However, instrument-assisted and electromechanical forms of manipulation have also been developed for use in humans to manage musculoskeletal disorders [55]. One of the equine studies included in this review used instrument-assisted manipulation (i.e., Activator), which had a peak effect at the last evaluation 7 days post-treatment [25]. Spinal mobilization and manipulation had peak effects after three treatment sessions at weekly intervals [29]. In humans, comparisons of the effectiveness of mobilization, instrument-assisted and manual manipulation have been reported with no one type of therapy shown to be more effective than the others [56,57]. To date, there have not been any direct comparisons between the efficacy of manual versus instrument-assisted manipulation in animals.The equine osteopathic studies included in this review incorporated sedation or general anesthesia, which limited any direct comparisons to other spinal mobilization studies that did not use anesthetic agents [20,21,22]. These studies did not clearly report if horses treated under sedation responded differently from horses treated under general anesthesia. General statements were provided by one author that suggested that treating under general anesthesia produced more favorable results and was the preferred technique despite the inherent risks and added costs [21]. However, the authors also reported that general anesthesia was used for horses with more intransigent or chronic issues, which would suggest that the prognosis for these cases might be worse than more subacute cases. Similar indications exist in humans for the use of mobilization or manipulation under anesthesia [58].4.3. Clinical IndicationsThe subjects included in the studies had naturally occurring axial skeleton pain or stiffness (n = 8), no overt back pain (n = 4), and experimentally induced pain or stiffness (n = 2). In humans, spinal mobilization and manipulation are both reported to have therapeutic effects on neck and back pain [59]. Across the studies included in this review, the treatment areas included the thoracolumbar region (n = 11), cervical region (n = 3), tail (n = 1), and carpus (n = 1). The clinical indications across studies included pain (n = 8), stiffness (n = 6), muscle hypertonicity (n = 3) and lameness or poor performance (n = 3), which are comparable indications within the human mobilization and manipulation literature [60]. However, it is often difficult to isolate the primary limiting disability or spinal dysfunction specifically to pain, stiffness, or muscle hypertonicity as all three of these clinical issues often occur concurrently to varying degrees within an individual patient [1,7]. The lack of specific musculoskeletal indications prevented the inclusion of several studies in this review [61,62].4.4. DosagesA single treatment session was reported in 64% (n = 9) studies. Recurrent treatment sessions for spinal manipulation included three treatment sessions over 3–5 days [23] or treatment once weekly for three weeks [26,29]. Osteopathic treatment under sedation was applied every 2–6 weeks for an average of 6 treatment sessions [22]. Due to the wide diversity in applied treatment techniques (e.g., trunk displacement, tail traction), treatment sites and the total number of applied mobilizations or HVLA thrusts per treatment session, it is difficult to synthesize the available information into clinically useful dosage recommendations. Some general guidelines have been provided in review articles on joint range of motion and stretching exercises [63]; however, specific guidelines based on this systemic review was not possible due to the paucity of data.Data from the equine mobilization and manipulation studies suggest that clinical effects are often noted after a single treatment session [10,24,25,27,31]. Most spinal manipulation studies assessed changes in outcome parameters immediately post-treatment; however, short-term effects were reported at 2 to 6 days [10] and up to 1 to 3 weeks after a single HVLA treatment session [25,31]. The effects of a single treatment session are likely to not be useful in formulating practice guidelines where several treatment sessions may be required to achieve the desired therapeutic effects [31].4.5. Outcome ParametersThe outcome parameters reported across studies assessed measures of joint motion (n = 6), nociception (n = 5), muscle tone or activity (n = 3), and performance (n = 3). Detailed spinal evaluation procedures have been widely used in veterinary medicine; however, standardization of the techniques and quantitative and qualitative scoring is still in the early stages of development [23,64]. Passive joint range of motion (n = 3) and kinematic analysis (n = 3) were the most common methods used to evaluate changes in stiffness [19,23,28,29,30,31]. While measures of normal passive joint range of motion has been reported for the appendicular skeleton [44,65], similar measures are not readily available within the axial skeleton due to a diverse array of measurement methods for assessing spinal stiffness [66].Using pressure algometry to measure MNTs was the most common method used to assess changes in nociception [23,24,25,26]. While there are not well-defined normative MNT values, changes pre-and post-treatment within an individual patient are reported to be reliable [67]. The objective assessment of muscle tone and activity in axial skeleton disorders is challenging [68]. Epaxial muscle tone or activity were assessed using a wide range of methods in this review, which included tonometry and EMG [27], static bioimpedance and dynamic acoustic myography [10], and soft tissue palpation [23].The ability to return to work based on owner surveys was used in the three equine osteopathic studies [20,21,22]. While owner reports may be useful for global assessments of health or performance, they are limited in quantifying the presence, localization and severity of pain, stiffness or muscle hypertonicity [69,70]. Unfortunately, there are very few validated functional questionnaires or standardized owner surveys in veterinary medicine that have been designed to capture measures of musculoskeletal function and specific responses to applied therapies [47,71]. The appropriate timing and delivery of these tools are also important considerations, which were substandard (i.e., cases over a 19-year period) in most included studies [22].4.6. Perceived EfficacyAcross equine studies, MNT values within the thoracolumbar region increased (i.e., less pain) from 11% to 83%, which suggests a clinically significant improvement [24,25,26]. However, long-term follow up in these three studies was not completed. In one spinal manipulation study, longer-term follow up was provided at 3 weeks post treatment and most of the reported positive clinical effects from 1 h post-treatment had dissipated [31]. Two studies compared spinal manipulation to other forms of therapy [23,25]. In ridden horses without overt signs of back pain, a single treatment of HVLA thrusts was more effective (27%) in reducing pain after 7 days than massage therapy (12%) or 7 days of oral phenylbutazone (8%) [25]. However, in Western performance horses in active competition that presented with signs of acute back pain, epaxial muscle hypertonicity and stiffness, HVLA thrusts produced no significant effects compared to low level laser therapy or combined HVLA and laser treatments [23]. Anecdotally, it appears that spinal manipulation in horses is more effective for treating chronic back pain and stiffness, compared to acute pain syndromes [30]. Similar findings are reported in systematic reviews of spinal mobilization and manipulation for treatment of acute and chronic neck or back pain in humans [72,73].Most studies reported positive or beneficial effects of musculoskeletal mobilization and manipulation as applied using the described techniques; however, only the experimental studies included control group comparisons. The cohort studies reported changes pre- and post-treatment within individual patients and often did not provide any long-term follow up. In the canine carpal immobilization-remobilization study, there is moderate evidence that repetitive, cyclic joint motion improved passive joint range of motion [19]. Across these three osteopathic studies, the reported efficacy ranged from 75–95% in clinical improvement and return to work in 90% of horses at 6–12 weeks post-treatment, which decreased to 53% after 6 months [22]. However, it is difficult to evaluate true clinical efficacy in the face of low study quality or design.4.7. LimitationsThe primary limitation of this systematic review is the large heterogeneity in the indications, applied techniques, treatment protocols, and outcome parameters between studies, which prevented a meaningful interpretation of the overall clinical efficacy of joint mobilization and manipulation. Osteopathic techniques include a diverse array of diagnostic and treatment approaches that range from articular, myofascial, vascular, lymphatic and neural techniques, which makes categorization of the type of applied therapy difficult [74]. It is also difficult and may not be clinically useful to categorize manual therapies into ‘stretching exercises’ versus ‘mobilization’ procedures as the definitions are typically poorly described and there may be a large overlap in the applied techniques [75]. Therefore, it is likely that some studies that might be viewed by others as evaluating ‘mobilization’ were judged by the author to fall more into the ‘stretching’ category and thus were not included in this systematic review.5. ConclusionsThere is low-to-moderate quality evidence based on the selected study criteria that various types of joint mobilization or manipulation will reduce pain, stiffness and muscle hypertonicity. The studies are highly heterogeneous in terms of interventions, dosing, duration of treatment, outcome parameters and follow up, which prevented further meta-analysis. Therefore, it is difficult to draw firm conclusions despite all studies reporting positive or therapeutic effects. Future studies need to establish quantitative and qualitative methods to specifically evaluate the effects of mobilization and manipulation, incorporate adequate control groups, provide longer-term follow up, and to include the evaluation of appendicular articulations.

animals : an open access journal from mdpi

10.3390/ani11041045

PMC8068116

The camel milk market was limited for a long time by its almost exclusive self-consumption use in nomadic camps. Significant development has been observed for the past two or three decades, including internationally, boosted by its reputation regarding its health effects for regular consumers. Such emergence has led the stakeholders in the sector to offer diversified products corresponding to the tastes of increasingly urbanized consumers, more sensitive to “modern” products. Thus, traditionally drunk in raw or naturally fermented form, camel milk has undergone unprecedented transformations such as pasteurization, directed fermentation, cheese or yoghurt processing, and manufacture of milk powder for the export market. However, the specific characteristics of this milk (composition, physical properties) mean that the technologies applied (copied from technologies used for cow milk) must be adapted. In this review, some technological innovations are presented, enabling stakeholders of the camel milk sector to satisfy the demand of manufacturers and consumers.

Camel milk is a newcomer to domestic markets and especially to the international milk market. This recent emergence has been accompanied by a diversification of processed products, based on the technologies developed for milk from other dairy species. However, technical innovations had to be adapted to a product with specific behavior and composition. The transformation of camel milk into pasteurized milk, fermented milk, cheese, powder, or other products was supported, under the pressure of commercial development, by technological innovations made possible by a basic and applied research set. Some of these innovations regarding one of the less studied milk sources are presented here, as well as their limitations. Technical investigations for an optimal pasteurization, development of controlled fermentation at industrial scale, control of cheese technology suitable for standardized production, and improvements in processes for the supply of a high-quality milk powder are among the challenges of research regarding camel milk.

1. IntroductionFor a long time, only fresh camel milk was consumed by pastoralists and was regarded as a gift for the hosts coming under the tent of the nomads. Consequently, it was not considered a commodity and its sale was often taboo. Moreover, it did not undergo any transformation, except for fermentation [1] to prolong its shelf life in desert conditions where the cold chain could not be respected. The introduction of camel milk to the regular market at a national or even international level is a recent feature [2]. Such development of the camel milk market was concomitant with a deepening of the knowledge of its fine composition [3] and of its transformation processes, allowing the marketing of a more diversified dairy product [4]. Recent findings are effectively available, allowing this important product renowned for its true or supposed “medicinal” virtues [5] to remove camel milk from marginality. Indeed, the production of camel milk at a world level is experiencing a considerable annual growth, exceeding 8% in the period 2009–2019 [6], testifying to the growing interest for this product.Moreover, despite its proximity with cow milk in term of gross composition (fat, protein, lactose, and mineral proportions), camel milk shows many specificities [3] such as functional proteins, predominant medium-chain fatty acids in fat matter, low lactose intolerance, and richness in vitamin C or iron, among others. These proper characteristics have an obvious impact on the “behavior” of camel milk during processing.Thus, the present paper proposes the state of the art regarding knowledge on camel milk processing by focusing on four main dairy products having different success on the market, i.e., pasteurized milk, fermented milk, camel cheese, and camel milk powder. Other products will rapidly emerge. In addition to the review regarding the current status of knowledge on camel milk processing, the more recent studies (2019–2020) focusing on this topic are presented in Table 1.2. Pasteurized Milk2.1. Current Global Conditions for Pasteurization of Camel MilkPasteurization of camel milk is a commonly used technique in camel countries. Nevertheless, the conditions of pasteurization implemented by each holder were often decided without considering the specificity of camel milk, with the rules being mainly based on the standard issued for pasteurization of cow milk. Some data regarding the conditions of camel milk pasteurization in scientific literature are quite variable: 60 °C for 30 min [33]; 75 °C for 15 s [34]; 63 °C for 30 min [35,36]. At the same time, many private companies in the United Arab Emirates (UAE), Saudi Arabia, Mauritania, Kazakhstan, Algeria, Tunisia, Morocco, and Niger are producing and selling pasteurized camel milk on the local market. All of these companies apply different conditions for pasteurization. It is important to mention that national/regional/international standards are not yet elaborated for camel milk or they are simply copied from cow milk. In some camel countries, no standard is proposed by the government or, at least, it is suggested to apply the same conditions as for bovine milk.2.2. Indicators of Camel Milk PasteurizationThe pasteurization procedure for camel milk should have its own conditions and indicators. Indeed, a preliminary study showed that alkaline phosphatase (ALP), traditionally used for cow milk [37], was not a convenient indicator of successful pasteurization of camel milk, because camel ALP is heat-resistant, still showing activity at 90 °C [38]. Loiseau et al. [39] suggested using glutamyltranspeptidase (75 °C for 30 s) or leucine arylamidase (75 °C for 28 s or 80 °C for 7 s) as an indicator of pasteurization for camel milk. If camel milk must be pasteurized at 72 °C for 20 min, the most appropriate indicator could be gamma-glutamyl transferase (GGT) according to Wernery et al. [40]. However, later in 2011, Lorenzen et al. [41] concluded that GGT was still present in pasteurized camel milk, whereas lactoperoxidase (LPO) could be a more appropriate indicator of pasteurization. Tayefi-Nasrabadi et al. [42] confirmed that camel LPO was less heat-resistant than bovine LPO. Until now, no sufficiently in-depth studies have been done in this field, although pasteurized camel milk was introduced to the international market. Such doubts on a convenient indicator of pasteurization for camel milk are a constraint for the establishment of an international standard. Thus, the pasteurization of camel milk at an industrial scale is possibly achieved in the wrong way and its heat treatment may be incorrect.2.3. Impact of Pasteurization on Physical Properties of Camel MilkThe characteristics of camel milk flow in pipes during the pasteurization process, the cleaning procedures, and the conditions of transfer and pumping also need to be studied. The behavior of fat globules and casein micelles in camel milk is different from that in cow milk, presenting an absolute viscosity of 1.72 mPa·s at 20 °C vs. 2.04 mPa·s at the same temperature in bovine milk. Thus, in dairy plants, the camel milk should not necessarily be transferred at 20 °C. According to Kherouatou et al. [43], camel milk in fresh conditions (without acidification) can resist some mechanical strains without changing its microstructure. The apparent viscosity was quite stable (between 1.6 and 2.0 mPa·s) at pH values between 5.2 and 6.7. Therefore, camel milk can be manipulated in dairy plants, e.g., pumping, agitation, skimming, homogenization, and bottling, for producing a pasteurized product.Few studies have focused on the sensory characteristics of heat-treated camel milk. Comparing three treatment conditions (63 °C for 30 min, 72 °C for 15 s, and 100.5 °C for 10 min), Lund et al. [44] observed lower taste score, texture, and overall acceptability in treated samples of camel milk compared to control (nontreated); however, there was surprisingly no significant difference between pasteurization protocols, although the milk at 63 °C/30 min had the highest mean sensory score. However, these authors did not check the hygienic level of their samples. Indeed, to compare the different protocols, it is important to cross the data regarding pasteurization conditions and sensory characteristics with the hygienic level of the raw milk.Some studies regarding behavior during heat treatment observed that camel milk could give an important amount of dry deposit on a stainless-steel plate during the pasteurization process from 60 to 90 °C for 1 h or 2 h [45]. This study showed that such a deposit is probably not of protein origin, because free thiol groups are in lower quantity than for cow milk treatment. A similar design of experiment with camel whey protein showed that, after 63 °C, whey proteins started to be denatured; this was especially evident at 98 °C [46]. It is possible that camel milk is producing a higher quantity of “milk stones”, i.e., the deposit of milk residues accumulated in an insufficiently cleansed dairy equipment where bacteria can be multiplied, contributing to bad flavors in milk. To the best of our knowledge, there are no available data in the literature regarding this aspect.2.4. Camel Milk Protein Behavior Following Heat TreatmentIn-depth studies on camel milk started relatively recently, especially regarding its technological properties. Even if some data on camel milk composition date from 1905 [47], research regarding the heat treatment impact on camel milk components started being implemented at the end of the 1980s. Thus, in a first trial, camel milk was heated at 63, 80, or 90 °C for 30 min. The rate of heat denaturation of whey proteins was twofold lower than for cow whey proteins [48]. After this study, another trial focused on heat coagulation at 100–130 °C in a pH range of 6.3–7.1. Only the 100 °C variant showed relative stability at pH 7, comparable to cow milk [49].Because milk stability during heat treatment is the most important point and because the size of casein micelles could impact milk preservation in a homogeneous solution, the micelle size of camel milk was measured before and after pasteurization. They are broader than those of cow and human milk. However, an important point must be kept in mind, i.e., the samples were analyzed 36 h after sampling [50]. Regarding the mineralization and citrate quantity in camel casein micelles, the proportions were significantly different than for cow casein micelles after pasteurization. It was observed that, at pH 5, significant changes occurred in the dromedary casein micelle structure from milk to coagulum [51].Some studies [35,52] stated that the pasteurization of camel milk could change its chemical composition. However, in these experiments, the microbial quality of raw camel milk was not assessed, no standardized starter culture was used, and the microbial contamination risk was not taken in account. Yet, in trials where microbiological status was tested before processing, pasteurization at 63 °C for 30 min improved the bacteriological quality of camel milk without changes in the composition compared to raw camel milk [36,53].The observation of camel whey protein using SDS-PAGE showed that some proteins sensitively decrease after heat treatment, albeit to a lesser extent than for cow proteins. The pattern of camel whey proteins and the global composition of proteins are not the same for camel and cow milk. Accordingly, the major proteins of whey in bovine are serum albumin (SA), α-lactalbumin (α-La), and β-lactoglobulin (β-Lg), whereas those for camel are SA, α-La, and three other fractions not reported in cow milk [48,54,55,56,57].In a recent study [58], the effect of heat treatment according to different protocols (65 °C for 30 min, 72 °C for 30 s, 75 °C for 5 min, 85 °C for 5 min, or 90 °C for 5 min) on camel whey proteins was assessed comparatively to cow milk. A lower denaturation of α-lactalbumin was observed in camel milk; after treatment at 90 °C for 5 min, 67% α-lactalbumin remained intact vs. only 5% in cow milk. However, camel serum albumin (CSA) was not totally detected after treatment at 85 °C for 5 min, although this was a less rapid process than with cow serum albumin, which disappeared after 75 °C for 5 min [58]. Similar figures were observed in a previous paper [45].Some authors tried quantifying the changes in the concentration of camel whey proteins during the pasteurization process. However, debatable results occurred especially because different methods were used, needing clarification. Indeed, the quantification of whey proteins in camel milk can be done using electrophoresis [59], radial immunodiffusion [56,60,61,62], high-performance liquid chromatography [56], fast protein liquid chromatography [60], and the more recent techniques of liquid chromatography/tandem mass spectrometry and liquid chromatography/electrospray ionization mass spectroscopy [57].The heat resistance of camel milk can also be revealed by the heat coagulation time [59]. Compositional difference plays an important role in heat resistance, particularly the absence of β-lg, different ratios in the casein complex (higher quantity of αs1-, αs2-, and β-caseins and lower quantity of κ-casein compared to cow milk [63]), and the presence of a higher quantity of other whey proteins. Whey proteins include three protein fractions described as common fractions of immunoglobulins (IgG1, IgG2, and IgG3), α-lactalbumin, lactophorin (which is closely related to the bovine proteose peptone component 3 (PP3)), the innate immunity peptido-glycan recognition protein (PGRP), and the whey acidic protein (WAP) [64].2.5. Sterilized MilkSterilization of camel milk using an ultra-high-temperature (UHT) treatment is yet to be achieved, despite private companies trying to establish an appropriate method. Some studies on the heat resistance of whey and casein proteins, fat globules, vitamins, or other compounds of camel milk are expected to contribute to a technical solution [42,48,49,52,55,65,66,67].After UHT treatment, camel milk presents a separation into two phases. During heat treatment at 65°/30 min, 72°/30 s, 75°/5 min, 85°/5 min, 90 °C/5 min [48] or at 63, 80 and 90 °C for 30 min and 72 °C for 15 s [52], it was observed that the whey proteins were overly sensitive and started denaturing. To stabilize camel milk proteins after UHT treatment, different protocols including the addition of chemicals (sodium hydroxide, calcium chloride, ĸ-casein from cow, sodium dihydrogen phosphate anhydrous, disodium hydrogen orthophosphate, or ethylenediaminetetraacetic salt) were tested but with disappointing results [68]. Further in-depth studies need to be implemented before being able to produce UHT camel milk.However, it is possible to obtain sterilized milk after reconstitution of liquid milk using camel milk powder. This camel dairy product is available on the market in the Middle East. Moreover, ultrafiltration to yield concentrated skim camel milk was tested experimentally [69]. By using an α-alumina ultrafiltration membrane (pore size 1.4 µm) at an operating temperature of 50 °C, it was possible to obtain a product with an extended shelf-life up to 60 days and to decrease the germ count by 99%. The low filtration temperature did not damage the heat-sensitive proteins [70].2.6. Antimicrobial Activity and Pasteurization of Camel MilkMany people believe that camel milk has sufficient natural antimicrobial activity to preserve it from natural adulteration at ambient temperature for a long time. As such, they often consider that camel milk collection does not require the same requirements in terms of hygiene practices. Therefore, in many cases, potential microbial contamination is not checked before implementing the trials. Nevertheless, it is important to recall the results obtained by Sela et al. [71], whereby the thermal death time of Escherichia coli in camel milk was the same as in cow milk. The presence of E. coli in camel milk is common [72] and flouting hygienic rules will provoke digestive disorders in consumers. At the same time, raw and pasteurized camel milk has the capacity to inhibit Cronobacter sakazakii. As this bacterium can grow in powder milk, some authors think that it could be possible to use it in the production of infant formula [73], but such a possibility requires further investigation.Obviously, pasteurization prolongs the shelf-life of milk and several data were obtained using different protocols. In the study of Lund et al. [44], quality was maintained after storage at ambient temperature for 24 h for raw milk vs. 76 h for milk treated at 100 °C/10 min, 64 h at 63 °C/30 min, and at 72 °C/15 s. The bacterial level appeared higher (2.77 log colony-forming units (CFU)/mL) with a protocol of 72 °C/15 s than 75 °C/10 min (2.65 log CFU/mL) and 65 °C/30 min (2.57 log CFU/mL), with the lowest content (2.45 log CFU/mL) being observed with a protocol of 80 °C/5 min [74].In their study assessing the effect of two protocols (63 °C/30 min and 72 °C/15 s), on total bacterium, coliform, and mold counts, Mohamed and El-Zubeir [75] found that the first protocol was non-significantly more efficient in terms of total bacterial and coliform count, but less in terms of yeast and mold count (Figure 1). However, the protocols were applied on highly contaminated milk. Accordingly, it is worth reiterating that pasteurization does not guarantee sterilization and requires respecting the hygienic rules at the milking, storage, and transport stages of the raw material to correct process the milk.Camel milk pasteurization is achieved at an industrial scale, and pasteurized camel milk can be directly provided to consumers with a shelf-life of around a week. Some dairy plants have extensive experience to producing pasteurized camel milk, for example, Tiviski (Mauritania), Camelicious (UAE), Al-Watania (KSA), and Tedjane (Algeria). However, further research is necessary, especially to study the rheological properties of raw and pasteurized camel milk, which are poorly documented.3. Fermented MilkThe fermentation process is commonly used for the preservation of food. The process of milk fermentation, including camel milk, is a traditional ancestral method all over the world. It consists of the transformation of lactose into lactic acid by the natural microflora in milk dominated by lactic bacteria and, in some cases, by yeasts. To understand this process, numerous investigations on the microflora of raw camel milk were performed. As many fermented products from camel milk have been produced by spontaneous fermentation for centuries, we start our analysis of the literature by considering the global microflora (pathogenic or otherwise) of raw camel milk.3.1. Microflora of Raw Camel Milk3.1.1. Nonpathogenic Microflora in Raw MilkThe normal microflora in camel milk has been widely investigated, and the results testify to high diversity across countries. In Tunisia, [76] found in raw camel milk 2.7 × 102 CFU/mL of total mesophilic aerobic bacteria, 1.7 CFU/mL × 102 of yeasts/molds, and 1.3 × 103 CFU/mL of lactic acid bacteria (LAB). In total, 60 LAB strains were isolated. In camel milk samples from Morocco [77,78], 120 bacterial strains were isolated and identified as LAB by morphological and biochemical characterizations, as well as 16S ribosomal RNA (rRNA) gene amplicon sequencing. Among the described strains, some having probiotic properties were detected. Some of those strains were isolated in raw camel milk [79]. They belong to the genus Lactobacillus (L. acidophilus, L. rhamnosus, L. gasseri, and L. delbrueckii) [80]. Other strains belonging to the same genus (L. plantarum, L. pentosus, and Lactococcus lactis) were reported by Yateem et al. [81]. The diversity of microflora in camel milk also includes yeast as observed in Algeria where 12 species, dominated by Trichosporon asahii, Pichia fermentans, and Rhodotorula mucilaginosa, were identified [82].Some strains isolated from raw camel milk such as Leuconostoc mesenteroides have shown specific antimicrobial activity against Listeria [83]. A similar finding was observed with Enterococcus faecium [84]. Abo-Amer [85] described the ability of Lactobacillus acidophilus strains isolated from raw camel milk to produce substances with antimicrobial activities. In a study achieved in Iran, among 64 LAB strains, 11 (belonging to genus Enterococcus, Lactobacillus, or Pediococcus) presented significant antibacterial activity against Staphylococcus aureus subsp. aureus or Bacillus cereus [86]. In Kenya, the predominant strains belonged to species E. faecalis, S. agalactiae, Weisselia confuse, Rhodotorula mucilanginosa, Cryptococcus albidus, and Candida lusitaniae [87,88]. In Kazakhstan, natural microflora including lactic bacteria (E. durans, E. faecalis, E. faecium, Lb. casei, Lb. curvatus, Lb. kefiri, Lb. paracasei, Lb. sakei, Lc. lactis, and Lc. mesenteroides) and yeasts/molds (Kazakhstania unispora, Saccharomyces cerervisiae, and Kluyvermyces marxianus) were identified in dromedary and Bactrian milks [89,90]. In Bactrian milk collected in China [91], 72 strains of lactic bacteria were isolated including Lb. paracasei, E. italicus, E. durans, Lc. Lactis, W. confuse, and E. faecium.Such diversity has high technological interest, with the microflora playing important role both in terms of antimicrobial activity as emphasized above and in terms of acidification, which is essential for fermented products and cheese processing. However, due to the antimicrobial properties of camel milk proteins being greater than those in cow milk [92] and, in some cases, due to the low hygienic status of camel milk samples, the acidification process appears to be slower than for cow milk [93]. The starters used in camel milk processing for fermentation or cheesemaking (mesophilic, thermophilic, or their mixture) led to an acidification rate at 37 °C between 33% and 79% lower than for cow milk [88].3.1.2. Pathogenic Microflora of Raw Camel MilkSubstantial data have been dedicated to a preliminary bacteriological description of raw camel milk using different methods from all camel countries. Sela et al. [71] previously reported that camel milk could contain E. coli if elementary rules of hygiene were not applied. Camel milk could also contain other pathogenic strains such as Streptococcus or Staphylococcus species [94]. According to the latest study in Algeria, 58% of commercialized samples of raw camel milk were of satisfactory quality, 8.33% were acceptable, and 33.3% were unacceptable. However, in their samples, they did not find any Salmonella sp. and Shigella sp. Another study achieved in Sudan reported the presence of E. coli, Klebsiella spp., Pseudomonas spp., Proteus spp., Enterococcus spp., Micrococcus spp., Streptococcus spp., and Staphylococcus spp. [95]. The presence of coliforms can be observed despite the concurrent development of lactic acid bacteria. In Tunisia for example, a study on raw camel milk focused on the numeration of mesophilic count, total LAB, and coliforms, reporting values of 7 × 103, 1.37 × 102, and 1.8 × 101 CFU/mL, respectively [96]. Therefore, it is important to check all raw camel milk samples for microbiological quality before processing. Indeed, when significant changes in camel milk composition are observed during the process, these could be linked to the microbiological status of the raw samples. The lack of systematic checks of microbiological quality in initial raw milk could impact all processing steps.3.2. Diversity of Fermented Camel MilkThe microflora biodiversity leads to a rich diversity of fermented beverages prepared from camel milk. Moreover, fermentation is one of the oldest methods of consuming camel milk. Camel milk producers living in different regions of the world have their own varieties of fermented products with specific taste, texture, and flavor. Each camel country has described their traditional fermented milk in terms of microbiological, physicochemical, and chemical properties, as well as volatile organic compound profiles in some cases. The most known fermented camel milk products described in the literature are shubat in Kazakhstan [97] and China [98], khoormog in Mongolia [99], garris in Sudan [100,101], suusac in Kenya [102], laben (lben) in Arabic countries [103], and ititu and dhanaan in Ethiopia [104,105] (Table 2). Other traditional fermented beverages based on a mixture of camel milk and water are available in Mauritania known as zrig [106], in Morocco known as Lfrik [107], and in Iran and Turkmenistan known as chal [108].However, in most cases, the fermentation process occurs spontaneously upon using previously fermented milk to inoculate raw milk [112]. The microflora in the fermented product is consequently more diversified than that in the raw milk samples [113]. For example, traditional suusac (Kenyan fermented camel milk) contains 45 LAB and three yeasts as identified by API50CHL and API20AUX. LABs were mainly represented by Lactobacillus curvatus, Lb. plantarum, Lb. salivarius, Lactococcus raffinolactis, and Leuconostoc mesenteroides subsp. mesenteroides, and yeasts were mainly represented by Candida krusei, Geotrichum penicillatum, and Rhodotorula mucilaginosa [111]. Other strains such as E. faecalis, Lb. fermentum, Lc. lactis, Cryptococcus laurentii, Candida lusitaniae, Saccharomyces cerevisiae, Trichosporon mucoides, and T. cutaneum were also identified more recently [88]. Traditional garris from Sudan contains the following LAB based on the API39CHL identification system: Lactobacillus animalis, Lb. brevis, Lb. divergens, Lb. plantarum, Lb. rhamnosus, Lb. gasseri, Lb. paracasei, Lb. fermentum, Lactococcus raffinolactis, and Lc. alimentarium [114].Such a complex microflora ecosystem in fermented milk could lead to very variable final products, hardly compatible with obtaining a product with standard organoleptic quality. For example, shubat, which is prepared mainly in Kazakhstan, Uzbekistan, Russia, the Xinjiang region of China, and the western part of Mongolia, is made by spontaneous fermentation, often leading to the production of gas and foam and sometimes resulting in a particularly acidic product, provoking some reluctance in urban consumers [115]. Moreover, spontaneous fermentation can be affected by the presence of pathogenic strains of E. coli because the initial pH is not sufficient to suppress their growth [116]. To solve these problems, it is convenient to use starter cultures, i.e., a preparation containing a limited number of identified live microbial strains (single or mixed) inoculated in raw milk. Such management of controlled fermentation can lead to the expected sensorial properties of the final product by contributing to flavor and texture adapted to the urban consumers’ taste. This also contributes to a standardized and safer product on the market. Unfortunately, despite the high biodiversity of camel milk microflora, starter cultures used in the camel industry are mainly taken from bovine milk. In Kazakhstan, among the 104 isolates from shubat samples, sampled in different regions of the country, 79 were maintained in a pure culture (71 bacteria and 8 yeasts), while three of the strains were selected because of their biochemical properties (Lactobacillus fermentum К5, Lactobacillus fermentum К6, and Lactobacillus plantarum К7), before being tested for their growth kinetics and characteristics to measure their produced biomass [117,118]. This was important for developing an industrial culture in fermenters. Then, each culture was tested for technological suitability, acidity, bacterial colonization, and organoleptic properties [118]. The transfer to an industrial level was achieved thanks to the acquisition of a high-capacity bioreactor, wherein the media used were selected based on optimal bacterial growth and a convenient ratio of cost/growth potential. The final product (packaged starters in powder after lyophilization) was commercialized in two volumes (5 and 10 g per pack). However, industrial transfer remains difficult and requires supplementary studies regarding the technological properties of the numerous strains available in natural fermented milk [119]. In addition to the commercial interest for the development of fermented milk with standardized organoleptic properties, there is the public health benefit of such products thanks to their potential probiotic effect. For example, camel milk inoculated with LAB strain Lactococcus lactis KX881782 isolated from raw camel milk showed higher α-glucosidase inhibition (antidiabetic effect), antioxidant activity, angiotensin-converting enzyme inhibition (antihypertensive effect), and antiproliferative activity (anticancer effect) than cow milk [120].4. Camel CheeseTechnical innovations regarding fermented camel milk have been applied to beverages known from prehistoric times to extend their shelf-life [121]. Furthermore, the making of camel cheese itself was an innovation. Indeed, the difference in casein proportions (notably the lower content of κ-casein) between cow and camel milk should explain the clotting difficulties observed in this process: 3–4% κ-casein vs. 13–15% in cow milk [122]. Moreover, bovine chymosin used in the dairy industry does not allow the optimal clotting of casein micelles from camel milk, leading to a weak curd. Thus, obtaining a firm coagulum was the first challenge faced by camel scientists and dairy factories processing camel milk [123].4.1. The Challenge of CoagulationThe first trials were achieved in the 1980s using bovine rennet enriched in chloride and calcium phosphate with Rhizomucor miehei as coagulant (commercialized under the name of Camifloc®); however, the coagulum remained fragile and brittle [124]. Different vegetal coagulants were also tested as extracts from Zingiber officinale [125] or from Moringa oleifera [126], as well as abomasum extracts from young or adult camel [127]. The solution to getting a firm curd followed the sequencing of camel chymosin achieved by Kappeler et al. [122], which allowed introducing the coding gene for camel chymosin into a mold (Aspergillus niger). Later, this recombinant enzyme was produced by Chr. Hansen© at an industrial scale and marketed from 2008 under the trade name Chymax-M1000®. However, despite a good curd being a necessary condition, it did not guarantee a “good” cheese adapted to the consumers’ preference in terms of taste, especially because camel cheese was often generated by scientists in a laboratory rather than cheese technicians in dairy plants, except in Mauritania [128].With great variety being possible in cheese, many trials are necessary to propose a large panel of products to the consumers. Different cheeses based on the technology for making gruyere [129], mozzarella [130], or feta and halloumi [131] were tested, but the texture, taste, and flavor of the final product did not correspond to the bovine equivalent. Indeed, the behavior of the camel’s “proteinic–lipidic matrix” during cheese processing differs from cattle milk. Such discrepancies among milk from different dairy species require more fundamental investigations of rheological properties to understand the changes during the different steps of acidification, coagulation, draining, brining, and refining, as well as the effect of various starters and thermal treatments [123].4.2. Rheological and Microstructural StudiesAs camel cheesemaking is a recent achievement all over the world, basic studies on the rheology or microstructure of the final products remain scarce. Investigations on curd tension and syneresis were achieved on soft camel cheese [132]. Gel firmness and gelation properties were tested according to different levels of temperature and pH [133]. The parameters of cheese viscosity (storage and loss moduli, loss tangent) after camel milk coagulation by camel chymosin were also reported [134]. To the best of our knowledge, while several studies on texture measurements (hardness, adhesiveness) have been published (for example, [135]), the microstructure properties of camel cheese require more investigations. In Iran, the effects of different mixtures of coagulants (Rhizomucor miehei protease and camel chymosin) on the microstructure and rheological properties of white cheese were compared, reporting a more compact protein network and firmer structure in the cheeses made with camel chymosin, but they used cow milk as a model [136]. In other studies, a comparison of the viscoelastic structure and the global gelation properties at different temperatures between cow and camel milk showed the higher difficulties of getting cheese with pasteurized camel milk [137].However, only studies achieved at a laboratory scale are available. True investigations at the industrial scale are lacking. Moreover, in the published literature regarding the physicochemical and rheological characteristics of camel cheese, the microflora (pathogenic or otherwise) was not systematically considered.4.3. Comparative “Behavior” with Cow CheeseA comparison of the changes observed in the “proteinic–lipidic matrix” of camel and cow milk during cheese processing is a common feature in scientific literature to understand the specificity of their respective “behavior”. However, this comparison is of low interest when the gross composition of each specific milk differs significantly. To avoid this, Konuspayeva et al. [131] adjusted cow milk to obtain the same fat and total protein concentrations. Although they got similar cheese raw yields (7.4 ± 0.15 vs. 7.3 ± 0.55 kg/100 kg for camel and cow milk, respectively) and calcium recovery, camel cheese presented a higher recovery in terms of total nitrogen, while cow cheese retained more fat. Significant differences were also observed in lactoserum composition: camel lactoserum contained more fat and total nitrogen (9.0 ± 1.73 g/kg and 9.21 ± 0.23 g/kg, respectively) compared to cow (7.7 ± 1.61 g/kg and 7.30 ± 0.02 g/kg, respectively) despite similar dry matter (68.9% ± 3.2% and 68.1% ± 1.15% in camel and cow whey, respectively).Two main technological difficulties are faced in camel cheese processing: (i) the continuous removal of serum from curd, and (ii) the slow acidification of the curd. Indeed, contrary to cow cheese where ripening is started after curdling with the rapid appearance of crust formation, camel cheese is characterized by a weak crusting and a continuous loss of moisture, due to serum release, leading often to a very dry curd which hinders correct ripening. Such behavior can be favorable for some types of cheese (such as feta), but not for hard cheeses [123]. This slow acidification was first reported by Farah and Bachmann [137], who found that 10 h was necessary to decrease the pH from 6.6 to 5. Such a delay can be reduced at 36 °C compared to 20 °C [130]. Thermophilic starters such as Lactobacillus helveticus, L. lactis, or Streptococcus thermophilus, known for their high acidifying power, can also be used to improve the acidification process [131]. However, for the management of fermentation to get specific fermented products, camel cheese manufacturing does not use starters made with lactic bacteria isolated from camel milk. The challenge for scientists and cheesemakers could be the identification of LAB strains specific to camel milk, allowing the potential to provide the typical aroma of cheese and to enrich the variety of camel cheese proposed to consumers.Investigations on the interactions between minerals and camel cheese processing are also lacking. It has been established that there was no effect of the addition of phosphate or calcium chloride on improving gelation by camel chymosin [130] contrary to cow milk [138]. There is also a lack of information regarding camel cheese in terms of the analysis of volatile organic compounds responsible for aroma which supporting the typicity of the cheeses. Currently, such an analytical approach has only been developed for cow cheese [139].4.4. The Challenge of Industrial DevelopmentUp to now, it seems that scientists and cheesemakers aim to make cheese by using the same methodology as for cow cheese. However, when feta- or mozzarella-type cheeses are prepared using camel milk, the results are often disappointing because the taste, texture, and consistency differ with respect to the same cheese made using cow milk. In the Middle East, where consumers prefer products with a neutral taste, it is difficult to expect the development of cheeses with a strong character as found in western Europe. To achieve products adapted to the local consumers, different trials were proposed, for example, cheese spread [140] or white soft cheese [141,142]. However, few cheeses have given rise to sensory analyses assessing the acceptability of the product by the local consumers [131]. Yet, such research would be useful for the dairy industry to develop convenient cheeses. To the best of our knowledge, few industrial initiatives have seen the light of day [128].The industrial development of camel cheesemaking is limited not only by the technological difficulties, but also by the hygienic quality of raw milk as mentioned above, notably because it is difficult to coagulate camel milk after pasteurization, forcing manufacturers to work with raw milk. The use of halloumi-type cheese technology [131] is an interesting alternative because the coagulum is pasteurized in lactoserum for 10 min at 80 °C after pressing. A second difficulty is linked to the cost of camel cheese due to the high price of the primary matter [143]. An alternative could be the valorization of lactoserum, which represents 88% to 90% of the initial milk volume.5. Camel Milk PowderAs outlined in Section 1, camel milk is a recent development on the international dairy market, made possible by the development of powder milk production, which is the best way to preserve this highly perishable product for later consumption. Moreover, camel milk is often produced in remote places far away from the consumption basin, whereby the only solution to transport a high quantity of milk is by removing the water it contains (88% to 90% of the weight). Another advantage of this process is the conservation of the nutritive value of liquid milk. To make camel milk powder, two main modern technologies are used: spray-drying (hot-drying) and lyophilization (freeze-drying).5.1. The Drying TechnologiesThe first reported trial aimed at making camel milk powder was recent, where freeze-drying technology was implemented to study the thermal characteristics of camel milk and its main components [7]. However, these tests were carried out in a laboratory (not at an industrial scale) with a freeze-dryer, allowing drying from −40 to 20 °C with a vacuum of 100 Pa. The resulting powder was stabilized at 11.3% humidity. A second, more recent study [144] pursued similar objectives, namely, to assess the effect of freeze-drying on the nutritional properties of camel milk, i.e., how the procedure affects the fine composition of camel milk in comparison to fresh milk. Analyses indicated a relative stability of most components (including minerals and vitamins) and concluded that the nutritional properties of camel milk powder were maintained. However, this was also achieved using laboratory equipment with limited capabilities. Moreover, nothing was said about the solubility of the powder obtained.The spray-drying process was also the subject of a few scientific publications. In a study comparing the physical properties of powdered camel and cow milk obtained by spraying, [145] used a two-step process; milk was first concentrated to 20–30% dry matter using a rotating evaporator at 80 °C and then passed through a sprayer. The equipment used (FT80/81 Tall Form Spray Dryer) allows treating small amounts with the same effects of an industrial sprayer. In their protocol, the drying conditions were as follows: air intake temperature of 200–220 °C, air outlet temperature between 98 and 105 °C, pump speed at 3–5 arbitrary units, and air outlet humidity between 1.2% and 5.8%. In their conclusion, the authors indicated that this process allows obtaining powder with less than 1.8% water, thus allowing a long shelf-life. The drying temperature should be well controlled, as too high temperature results in an increase in the insolubility index due to protein denaturation. Overall, the solubility of camel milk powder is lower than that of cow milk. Another criterion used by manufacturers to assess the quality of milk powder is fluidity. This is the ratio between the density of the untamped powder and its density when compacted. This fluidity appears to be lower for camel milk compared to cow milk but remains at a fairly good level [145]. In a recent study regarding the acid gelation of fresh and powdered camel milk acid [146], the dry-spraying process was achieved using a laboratory sprayer (Buchi B-290), allowing an air entry temperature of 190 °C and an output temperature of 90 °C with an input flow of 600 mL/h. In their publication, [147] also aimed to test the effect of spray-drying, i.e., how the temperature of the air intake (160, 140 and 120 °C), atomization pressure (800, 600, and 400 bars), and feeding flow (5.4 and 3 arbitrary units/s) affect the nutritional components of camel milk (vitamin C, fatty-acid profiles). The equipment used was the same mini-laboratory sprayer cited in previous publications. Powder yield increased with the highest air intake temperatures and the lowest feed flows. High temperatures and high spray pressures reduced vitamin C levels. Lastly, the atomization pressure increased the fatty-acid content.5.2. Interests and Limitations of the Technologies Used for Spraying Camel MilkThe spray-drying method seems preferable to make camel milk powder for a better reconstitution of liquid milk, but the investment for the dairy industry is more important as it requires the procurement of a costly milk drying tower and sprayer. However, the powder obtained by freeze-drying (lyophilization) could be used by agro-food industries (pastry and chocolate factories).Nevertheless, the main limit is that drying consumes a high level of energy. In the dairy industry, the spray-drying process has a higher energy demand per ton of end-product, despite the recent technical improvements and novel equipment decreasing the energy consumption per ton of finished product. Moreover, the high investment necessary for obtaining and using a drying tower to make milk powder requires a sufficient volume of raw matter, which is only possible in certain contexts such as large camel dairy farms or collecting centers with a large network of camel farms.Lastly, it appears that trials on camel milk (in any case, those that are published) are limited in number and that industrial trials are poorly documented. It is also apparent that the spraying of camel milk requires an optimization of the input parameters (temperature, pressure, and flow) to maintain the nutritional properties of the product and the functional characteristics (solubility, hygroscopy, and fluidity) of the powder. As camel milk studies were all carried out using materials dealing with small quantities, the “translation” of these results to an industrial scale involving large volumes is not possible, although the practice of industrial hot spraying of camel milk is implemented in the United Arab Emirates, China, and Europe.5.3. The Challenge for Camel Milk Powder DevelopmentThe main problem with camel milk during high-temperature heat treatment, as happens during spraying, is the denaturation of proteins (especially whey proteins), which explains the difficulty with obtaining UHT milk. To maintain powder milk in the best possible conditions, and to facilitate the solubility of the powder to replenish the liquid milk, the surface composition of the powder is essential [148]. This surface of the spray-dried emulsion is naturally composed of mostly fat (mostly triglycerides), in addition to some proteins. Thus, the denaturation of serum proteins at a high temperature increases the fatty surface content of the powder and makes it difficult to replenish liquid milk. For a better emulsion during this reconstruction, it is proposed to perform an “encapsulation” using sodium caseinates, thereby ensuring stability of the powder. Such encapsulation is improved by the presence of lactose. For example, surface fat decreases from 30% to less than 5% if lactose is present in a 1:1 ratio relative to sodium caseinate.In Table 3, possible improvements are reported to obtain a powder of stable quality due to the specificities of camel milk.6. Other Products6.1. YogurtAmple literature is available on the possibility of making yoghurt with camel milk [149,150]. Several strains of conventional lactic bacteria have been tested such as Lactobacillus bulgaricus or Streptococcus thermophilus [151], as well as L. acidophilus, L. casei, and Bifidobacteria [152].However, the manufacture of camel milk yoghurt poses a texture problem, with the product appearing sticky and ultimately unpleasant to the palate [153]. Indeed, the viscosity of the product does not change during the gelling process compared to the milk of other dairy species. This constraint is related to protein composition [154] and to antibacterial factors naturally present in camel milk [155]. Another reason could be linked to the foaming properties of camel milk. The foam in this milk is stable, but it leads to a weak structure of the gel, which becomes unstable [156].To obtain a better texture, trials with the addition of gelatin, alginate, or calcium were attempted [157], whereas ferments producing exo-polysaccharides were also used [158]. The application of a high-pressure treatment could have a positive effect on the texture, but no trials have been conducted to date with camel milk [159].Other authors have attempted to improve the manufacture of camel milk yoghurt by mixing it with milk from other species [160] or by introducing 0.75% biosynthesized xanthan, albeit with moderate results in terms of organoleptic properties [161]. In any case, the final product corresponds at best to a “drinking yoghurt” without having the taste qualities, even when natural or synthetic aromas are added [162]. These difficulties explain why there is limited industrial production of camel milk yoghurt at present. Some researchers proposed frozen yogurt as a product that is between yogurt and ice cream [163]. The optimal composition from a texture point of view would be allowed with several ingredients such as fat (5%), sugar (13%), gelatin (0.5%), and 14% banana [164]; however, such a proposal has never gone beyond a laboratory scale.6.2. Butter and SweetThe fat in camel milk contains less than 0.5% butyric acid [165] compared to almost 5% in cow milk. In addition, fat cells are smaller than in cow milk [166]. As a result, butter yield is low [167] with disappointing organoleptic properties [167,168]. To obtain fat cells at the time of butter production, it is necessary to implement vigorous hot shaking (22–23 °C), which allows recovering about 80% of the fat [168]. Ghee (clarified butter), a popular product in India, has also been attempted using camel milk [169]; however, in addition to very low yield compared to buffalo or cow milk, the final product was found to be more susceptible to rancidity. Transformation into butter, therefore, does not seem fundamentally interesting in the context of an industrial valuation of camel milk. In fact, apart from trials in Ethiopia where butter consumption, including rancid butter for certain recipes, is popular, the production of camel butter has little future.Making ice creams with different flavors is an easy technology. Ice cream made from camel milk is commercialized in the United Arab Emirates, Morocco, and Kazakhstan. The same technology is used as for other milks. Ice cream is highly popular among consumers and, above all, provokes less reluctance than other products. However, very few studies on the texture and sensory properties have been conducted [170].There is no reference for processing camel milk into sweets. However, traditional products are available. For example, in Kazakhstan, a caramel called Balkailmak is obtained after a long thermal treatment of about 10 h at boiling temperature. The introduction of milk powder to chocolate as proposed in the Emirates can also be mentioned. A study on the acceptance of camel milk in a panel including 470 Emirati students showed a higher score with chocolate-flavored milk [171].6.3. Non-Alimentary Processing of Camel-MilkThe manufacture of soaps and other cosmetic creams with camel milk is now common practice in many countries (Morocco, Mauritania, Saudi Arabia, India, Holland, China, Australia, etc.), whether on a semi-industrial scale or on a handicraft scale. China sells cabinets containing various cosmetic products from lipsticks to moisturizers, shampoos, and various lotions. The interest in the use of camel milk for the cosmetic industry benefits from the hypoallergenic properties of its proteins [172].7. ConclusionsThe “modernized” processing of camel milk is a recent feature compared to the milk from other dairy species. However, the technologies used to transform milk into pasteurized or fermented products, cheese or yoghurt, powder, or various sweets face two main challenges: (i) the systematic application of already proven technologies for cow milk is not necessarily suitable for camel milk and requires adaptations based on more fundamental research on the behavior of milk components during processing; (ii) the transfer of laboratory results already relatively numerous to an industrial scale remains insufficient, especially for products such as cheese or yoghurt, and it requires additional technical and economic analyses. The worldwide interest in camel milk, which is largely due to its expected health effect for consumers, is prompting basic research and development to continue investigations in order to translate technological innovations into products available on a large scale.Despite these technological constraints, the global camel milk market is strongly changing. These changes are visible through two main structural innovations: (i) the emergence of intensive production systems under an entrepreneurial approach appearing more or less disconnected from pastoral dynamics; (ii) the development of periurban camel production systems, often initiated by pastoral breeders, contributing significantly to the urban supply in camel milk and keeping significant relationships with the pastoral economy. Such trends emphasize that the technical innovations in camel milk processing are coupled with socioeconomic dimensions. However, such changes raise the question of the sustainability of these new methods of producing milk and their environmental impact for the planet. Obviously, they contribute to a revaluation of the place of camel in national livestock economies, and it requires ensuring (i) the establishment of an international standard recognized by all the stakeholders, and (ii) the sustainability of “modernized” production and processing systems by avoiding the shortcomings noted in other modern livestock sectors.

animals : an open access journal from mdpi

10.3390/ani12070863

PMC8997151

In the European Union (and elsewhere), the overall use of animals in laboratories has failed to undergo any significant decline, despite six decades of purported adherence to the “3Rs” principles of replacement, reduction, and refinement. In the EU, the 1986 adoption of a legal requirement to use scientific methods not entailing the use of live animals, rising public opinion against the use of animals and the almost exponential rise in development and application of non-animal new approach methodologies (NAMs) signals a readiness to end animal testing. Indeed, the European Parliament recently carried an almost unanimous vote to adopt an action plan to phase out the use of animals in research and testing. This article explores what is needed to make this action plan a success, considering all stakeholders and their needs.

In September 2021, the European Parliament voted overwhelmingly in favour of a resolution to phase out animal use for research, testing, and education, through the adoption of an action plan. Here we explore the opportunity that the action plan could offer in developing a more holistic outlook for fundamental and biomedical research, which accounts for around 70% of all animal use for scientific purposes in the EU. We specifically focus on biomedical research to consider how mapping scientific advances to patient needs, taking into account the ambitious health policies of the EU, would facilitate the development of non-animal strategies to deliver safe and effective medicines, for example. We consider what is needed to help accelerate the move away from animal use, taking account of all stakeholders and setting ambitious but realistic targets for the total replacement of animals. Importantly, we envisage this as a ‘phase-in’ approach, encouraging the use of human-relevant NAMs, enabling their development and application across research (with applications for toxicology testing). We make recommendations for three pillars of activity, inspired by similar efforts for making the shift to renewable energy and reducing carbon emissions, and point out where investment—both financial and personnel—may be needed.

1. IntroductionIn September of 2021, the European Parliament voted overwhelmingly in favour of a resolution to phase out animal use for research, testing and education, through the adoption of an action plan [1]. This resolution stems from several observations. In the European Union (and elsewhere), the overall use of animals in laboratories has failed to undergo any significant decline (Figure 1), despite enshrining the “3Rs” principles of replacement, reduction and refinement in the European Directive 2010/63/EU on the Protection of Animals for Scientific Purposes; (Article Four ‘Principle of replacement, reduction and refinement’ [2]); rising public opinion against the use of animals with almost three-quarters of Europeans in favour of phasing out animal testing [3] and the increase in development and application of non-animal new approach methodologies (NAMs) [4,5,6]. The resolution acknowledged the role that animal research has played, stating “whereas previous animal testing has contributed to advances in developing treatments for human health conditions, as well as medical devices, anaesthetics and safe vaccines, including COVID-19 vaccines, and has also played a role in animal health” but also, importantly, reflects the rise in development and possibilities of the non-animal approaches: “whereas the toolbox of non-animal testing models is growing and shows the potential to enhance our understanding of diseases and accelerate the discovery of effective treatments; whereas this toolbox includes, for example, new organ-on-chip technology, sophisticated computer simulations, 3D cultures of human cells for drug testing and other modern models and technologies”. As the Parliament’s resolution text further acknowledges, this means that the vision in the European Directive 2010/63/EU on the Protection of Animals for Scientific Purposes to reach a “full replacement of procedures on live animals … as soon as it is scientifically possible to do so” could be accelerated in the EU. Thus, despite strong language advocating for the replacement of animals, it is clear that existing legislation and strategies are not sufficient to achieve the shift.Here we explore the opportunity that the action plan could offer in developing a more holistic outlook for fundamental and biomedical research, which accounts for around 70% of all animal use for scientific purposes in the EU [7]. We specifically focus on biomedical research, which we define here as research that is not carried out to satisfy any regulatory requirement(s). We consider how mapping scientific advances to patient needs (considering the vital role of basic science in feeding the discovery pipeline [8] and the possible advantages to the inclusion of patients in basic research [9]), taking into account the ambitious health policies of the EU, would facilitate the development of non-animal strategies to deliver safe and effective medicines, for example. We consider what is needed to help accelerate the move away from animal use, taking account of all stakeholders including scientists, irrespective of their animal use, the public and patients, and setting ambitious but realistic targets for total replacement of animals. Importantly, we envisage this as a ‘phase-in’ approach, encouraging the use of human-relevant NAMs, enabling their development and application across research (with applications for toxicology testing). We make recommendations for three pillars of activity, inspired by similar efforts for making the shift to renewable energy and reducing carbon emissions, and point out where investment—both financial and personnel—may be needed.2. The Current Situation—Sustained Reliance on Animals across the Research SpectraEuropean Directive 2010/63 on the Protection of Animals for Scientific Purposes articulates the need for animal welfare at its heart and throughout, stating “the final goal of full replacement of procedures on live animals for scientific and educational purposes as soon as it is scientifically possible to do so…”. However, it is apparent when studying the available data on animal use for scientific purposes (Figure 1 and see also ALURES database) that effective implementation of the Directive alone is not sufficient to drive meaningful reduction, let alone full replacement.For the purpose of this commentary, we focused on the use of animals for biomedical research. For this, we combined data on animal use for “Basic research” with animal use for “Translational and applied research”, as recorded in the ALURES database. ALURES was launched in 2020 and collates annual data on animal use across the EU, starting with data from 2015. In contrast to animal use for “Regulatory use and Routine production”, basic research and translational and applied research are not associated with a legal mandate to use animals and are therefore may be likely to capture “experimental” animal use, prior to any regulatory submission. Although the Directive does not provide a definition of these categories, statistical returns indicate that animals used for oncology, study of organ systems, sensory organs, metabolism, multisystemic, and animal behaviour are returned under the category ‘Basic research’. For translational and applied research, animal use includes human diseases, animal diseases, animal welfare, disease diagnosis and non-regulatory toxicology and ecotoxicology. However, in combination, animal use for basic research and applied and translational research accounts for almost 70% of the total animal use, representing around seven million animal uses each year, and has done for the past four years for which such records are available online, through the ALURES database (2015–2018). We focused on the data in ALURES as this is freely available although this has limited our analysis to animal use between 2015 and 2018. However, we note that some member states have published additional statistics on animal use for 2019, but these reports are not available for all member states that contribute to ALURES and therefore we used ALURES as the most complete dataset. Note that earlier data on animal use are available through the European Commission portal at https://ec.europa.eu/environment/chemicals/lab_animals/reports_en.htm (accessed on 20 February 2021).Figure 1 summarises the data on animal use for Basic research, Translational and applied research, and Regulatory use, submitted to the European Commission and available through the ALURES database. These data illustrate the fact that reliance on animals as surrogates for humans in biomedical research (represented by Basic research plus Translational and applied research; grey bars) has not significantly declined over this timeframe. The reasons underpinning this lack of a sustained decrease are likely complex, possibly controversial and multi-factorial, and a detailed analysis of them is beyond the scope of this commentary. Briefly, animal use may be justified in terms of the possible benefit to human health, animal health, or the environment, but the evaluation of potential benefits is problematic, with a tendency to “over-promise” the likely advantage (to humans) [10]. Additionally, the choice of animal model has been shown to depend on the historical use of animals rather than the most valid scientific approach [11]. Perhaps it is not surprising that a survey of animal researchers revealed their perception of several roadblocks hampering the development and implementation of animal-free tools [12], and this may be (partly) associated with a lack of formal education in, and exposure to, the innovative, non-animal approaches, as we discuss later. Thus, we suggest that all of these factors, and probably more, are underpinning the continued use of animals.Reuse of animals has accounted for around 2% of total animal use for each year in which reuse data are available (2015 to 2018, via ALURES database), and therefore does not significantly impact the observed lack of decline. It is also not appropriate to suggest that increasing animal re-use be employed as a strategy in order to reduce the absolute number of animals. Animal reuse is rightly, strictly governed, and must occur on a case-by-case basis. The Directive states that “the benefit of reusing animals should be balanced against any adverse effects on their welfare, taking into account the lifetime experience of the individual animal” [2].The data presented in blue bars in Figure 1 show the total animal use for all purposes—namely research, testing and education. The data in Figure 1 also suggest that animal use for regulatory use and routine production (grey bars) has been stable at around two million uses per year, and this is true for the last decade (data not shown). Note that there are differences in data submission (reporting requirements changing, misunderstanding of requirements, the addition of Member States) that make accurate annual comparisons challenging, but even with this variability, there is no indication that animal use is undergoing a sustained significant decrease, despite the clear legal requirement laid out in the 2010 Directive and its predecessor. Note that the increase in animal use from 2017 to 2018 is in part due to the contribution of data from Norway for the first time.The animals most commonly used for research and testing are rodents and fish. However, mice bear the brunt of biomedical research (Figure 2), for reasons of scientific tractability (e.g., ease of genetic manipulation) and cost-effectiveness, rather than scientific rationale. In fact, for translational and applied research at least, the availability of the (animal) model is more likely to underpin model choice than similarity to human pathology [11]. On average, between 2015 and 2018, over 4.5 million uses of mice were recorded each year for biomedical research (defined here as animal use for Basic research combined with animal use for Translational and applied research according to ALURES categories). This is in contrast to mouse use for Regulatory use and Routine production (dotted line, Figure 2), which has undergone a year-on-year decrease and represents less than 10% of all animal use—whereas mouse use for basic, applied and translational research remains at just under 50% of all animal use.The most recent data collected for the 28 Member States of the European Union and Norway are from 2018 (when the United Kingdom was still part of the European Union and therefore UK data are included in these statistics). These data indicate that 4.46 million mice were used for biomedical research, with a further one million uses of mice for regulatory use and routine production. When calculated as a percent of all animal use for all purposes, we see that biomedical research using mice, at around 80% of total mouse use, represents the vast majority with no signs of significant decline since the introduction of the Directive.The decrease of almost half a million mice between 2017 and 2018 (when taking account of the 28 Member States only and not considering the data from Norway), appears promising. However, scrutiny of the changes in animal use indicates fluctuations of hundreds of thousands in either direction across the years—further evidence of the lack of a sustained downward trend. Between 2015 and 2017 there were no changes made to reporting requirements or number of Member States submitting data, and the number of animals did not change significantly in either direction during this time period; thus, it would be a major assumption that a decrease in animal use for one year (as seen between 2017 and 2018) will translate to a sustained, cumulative decrease over many more years without accompanying policy efforts. One of the reasons for this could be that, despite a vision of full replacement of procedures on live animals for scientific purposes, Directive 2010/63/EU focuses its attention on raising animal welfare standards and providing rules governing the use of animals in scientific procedures. This includes, for instance, stating provisions on the use of certain animals, such as non-human primates and endangered species; guidelines on classification of the severity of procedures; rules on anaesthesia; and project evaluation and authorisation.Directive 2010/63/EU is required to encourage consideration of raised standards for animal welfare across the EU, but it appears that implementation of the Directive alone is not enough to address the scientific challenge of entirely moving away from animal use. Article 47 “Alternative approaches” is the only article in the sixty-six that comprise the Directive that specifically describes the requirement for application of the 3Rs. It is therefore timely that the Directive is complemented with an ambitious and proactive action plan. The Parliament’s near-unanimous Resolution from September 2021 offers the opportunity to do just that, and to achieve a sustained decrease in animal use while contributing to the EU health initiatives by accelerating the shift to more human-relevant approaches to research into health and disease.3. Transitioning Biomedical Research towards Human-Based, Non-Animal Methods Represents an Essential Step in Achieving the EU’s Public Health ObjectivesThe EU health research initiatives https://ec.europa.eu/info/research-and-innovation/research-area/health-research-and-innovation_en (accessed on 4 January 2022) encompass several key research areas where the human-relevant approaches are already proving illuminating. For example, for disparate conditions ranging from cancer [13,14,15] through brain diseases [16,17] to rare diseases [18], researchers are successfully applying methodologies no longer reliant on the use of animals as ‘disease models’ in order to develop potential new treatment options and offer a clearer understanding of the human condition.Rare diseases offer an ideal opportunity to harness the impressive advances in genetic and technological science over the last decades for the identification of new “druggable” targets, to understand the mechanism(s) of disease progression and for faster repurposing of existing drugs as potential treatment options for people living with these conditions. The ongoing COVID-19 pandemic has revealed the potential for microphysiological systems (organ chips) to provide fast and effective drug repurposing [19]. It may be possible to apply organ chips, developed using patient cells, to screen approved drugs and discover potential personalised treatments. As research using the chips, and advances in other cell-based NAMs, evolve so that knowledge and experience accumulate, it becomes apparent that chips can be used to measure increasingly complex parameters such as immune cell recruitment [20,21]; thus, these modern, human biology-based NAMs offer a useful platform to look at drug/combination efficacy as well as safety.It is also apparent that these rare diseases do not lend themselves to the creation and use of genetically altered animals as surrogates for those diseases genetic in origin, which comprise the vast majority of rare diseases [22]. This necessitates a different approach, whereby data mining [23], existing data [24], and use of patient biological materials [25,26] are utilised for “pre-clinical” testing, instead of expensive and time-consuming animal models. As one recent success illustrating where NAMs may help to decipher rare diseases, Chou and colleagues developed a human bone marrow chip that recapitulated hematopoiesis [27]. Importantly, chips created using cells from patients with the rare disease Schwachman-Diamond syndrome demonstrated a hypoplastic phenotype, with impaired maturation and aberrant surface marker expression of neutrophils—mimicking the neutropenia and other clinical aspects of the disease. Screening drugs in the chips could be a time- and cost-effective way to get much-needed treatments to the people who require them, and could offer a route to personalised medicine.Overall, the European Commission, through its European Health Initiatives and funding programmes, recognises the inability of a single country or methodology to address these issues. In the area of brain research, for example, there are several initiatives designed collectively to address the need for a better understanding of brain function and also to diagnose and treat brain diseases. However, threaded throughout these projects there remains a reliance on in vivo models employing rodents, and thus a failure to fully embrace human-centric methodologies. This despite mounting scientific evidence that rodent brains fail to fully recapitulate the complexity or functionality of the human organ [28,29], and even declarations that “… species-specific features emphasize the importance of directly studying the human brain.” [30].Nonetheless, there is much to celebrate in these programmes. One example is captured in the strategy for the EU Joint Programme on Neurodegenerative Disease Research (JPND), which is the largest global research initiative aimed specifically at neurodegenerative diseases. The JPND Research and Innovation strategy includes several workplans focused on the use of patients, patient data, phenotypic screening and patient stratification, along with the development of cellular methods using human tissue or stem cells, [31] indicating the value of these non-animal approaches in addressing the issues. However, there remains some reliance on “improving” animal models and an emphasis on “reverse translation”, despite countless failures and the dismal translational success of animal models of neurodegenerative disease. In 2019 alone, 132 agents were in a clinical trial for the treatment of Alzheimer’s disease [32], yet the treatment options remain extremely limited, with either high costs [33] or adverse side effects [34], proving disadvantageous to patients and carers.Currently, preclinical testing, the stages of drug development that occur prior to testing in human volunteers or patients, relies heavily on animal models, and it is no coincidence that—partly due to the insurmountable species differences between rodents, dogs and humans—current drug failure rates are around 95%, often as a result of unexplained toxicity or lack of efficacy [35,36]. It is therefore likely that more innovative human biology-based approaches can offer a more relevant, predictive and cost-effective path for preclinical drug discovery. The pharmaceutical industries should be applauded for their efforts in supporting the development and use of NAMs [37], but it seems that their hands are effectively tied against the greater use of these methods until regulations prioritise NAMs above animal data. As stated in the JPND research and innovation strategy: “To accelerate translation of basic findings to clinical benefit, the validity of model systems used for target identification and therapeutic development needs improvement” [38]. This should be accompanied by a clear acknowledgement as to when and where specific animal models are failing to translate, and a commitment to no longer fund further research using these models. The former Innovative Medicines Initiative (IMI) noted the need to “eliminate poorly predictive animal models”, including those for Parkinson’s, depression, autism and schizophrenia [39], yet continued to fund projects developing or applying animal models in these areas. A review of the closed calls for IMI2 (https://www.imi.europa.eu/apply-funding/closed-calls (accessed on 7 January 2022)) reveals many methods and tools that employ human cells, patient data, in silico tools and thus an increase in research projects and programmes using non-animal models to improve the predictivity and human relevance of these important topics. However, it is also true that many of these non-animal models are being developed and used in parallel to the animal models, despite the acknowledged limitations of the animal-based approaches and their inability to translate.Despite some efforts to invest in more human-relevant approaches, particularly within the portfolio of work funded by IMI and IMI2, there remains a need to define human-based, non-animal models as the preferred method of investigation in health research wherever possible. In June 2021, the EU Innovative Health Initiative was announced [40] as the successor of the IMI for the advancement of medical technology, digital health and diagnostics. Among its strategic priorities is to “… strive to pursue the aims of Directive 2010/63/EU on the protection of animals used for scientific purposes and, in particular, the principle of the Three Rs to replace, reduce and refine the use of animals.” [41]. However, there remains a lack of clear emphasis on prioritised funding for human biology-based non-animal approaches throughout the research agenda. The adoption of unambiguous language prioritising development and application of human-based, non-animal tools in all funding calls will help to accelerate further development of human-predictive technologies, appreciation of where more input may be required (i.e., gap analysis), and further increase confidence and familiarity with the processes and data that will help to embed these non-animal approaches—and by extension animal replacement—in biomedical research across the EU.4. Planning an Inclusive Transition to Non-Animal ResearchA transition away from tradition towards innovation is nothing new— it is what humans have been doing since they began to walk upright—and therefore shifting from the use of animals as models in research and testing to more human-relevant, non-animal approaches can be seen as one more advance. Although historically, research on animals has offered a repository of scientific information which has been applied to better our understanding of human (patho)physiology, there are still many unknowns that require human biology-based approaches. Additionally, of course, animals have been used for the development of effective medicines, since all drugs have to be tested on animals, but it seems timely to consider what we might be missing—where drugs that are toxic to animals are lost from the development pathway [42]—and so we are calling for a more rapid shift towards animal replacement. Also, we can draw from previous experiences and ongoing initiatives to map out a safe, effective and inclusive route toward full replacement that ensures that all stakeholder needs and requirements are considered (Table 1). The concept of a “Just Transition” was recently articulated by the European Commission with regard to its intention to make the move away from fossil fuels in the context of the European Green Deal [43]. This was followed by an announcement from the European Research Executive Agency of a budget increase to 111 million EUR with which to support collaborative research and the development of breakthrough technologies to enable the shift away from fossil fuels [44].Of course, the shift away from using animals in research and testing is not on the same scale as the move to net-zero emissions. However, even with this change in magnitude, there are similarities in terms of the intentional, holistic and methodical approaches applied for carbon neutrality that could be adopted to initially reduce, and ultimately replace, animal use. In line with other initiatives from the European Commission, we are suggesting that targeted reduction in animal use could be divided into three pillars: (Pillar 1) “Promoting innovative science with human biology as the gold standard” encompasses the scientific and technological advances underpinning the development and use of advanced, human-based non-animal NAMs; (Pillar 2) “Agile regulations” moves beyond fundamental biomedical science to address what is required to drive increasing confidence, trust and use of NAMs; (Pillar 3) “Knowledge transfer” includes the vital elements of education and (re)training that are needed to create a fit-for-purpose workforce and to enable stakeholders invested in animal-based research to pivot away from this without losing their livelihoods.To facilitate an inclusive transition away from animal models in biomedical research and toward the more human-relevant tools, it is important to consider all the stakeholders, beyond the researchers themselves. There is also the need to consider the wider population, who may be destined to become patients and therefore to think of disease prevention strategies, which could have an indirect impact on animal use but are more health policy-focused, so are not included here as part of our science-led Pillars. However, we recognise that many of the IHI future plans are placing patients at the centre of their strategies and that this, together with ensuring adequate support for geographical areas with either higher disease burden or reduced public spending capacity, are vital for improving lifestyle education and disease prevention.Thus, Table 1 is not an exhaustive list, for example, besides patients with life-limiting or life-threatening conditions, we must also consider the stakeholder with most to lose in this—the animals themselves—and thus the ultimate purpose of adoption of these pillars is to ensure multi-stakeholder co-operation to enable full replacement of animals across the research and testing spectra. Table 1 compiles many of these actors, suggests the pillars into which they would fit and offers some suggestions of what may be needed to encourage the transition.5. Pillar 1—Promoting Innovative Science with Human Biology As the Gold StandardMeaningful progress toward the replacement of animals in the EU is unlikely to occur until the European Commission and Member States formally recognise that human biology-based tools and methods are the quintessential model for human health research. This recognition should then be reflected in all strategic science priorities, funding calls, and grants awarded. This would lead to a natural redirection of funding away from animal models with low human predictivity/translation towards more predictive, human-relevant NAMs.Within the European Union’s most recent funding programme, Horizon Europe (https://ec.europa.eu/info/research-and-innovation/funding/funding-opportunities/funding-programmes-and-open-calls/horizon-europe_en (accessed on 15 December 2021)), 270 million euros funding was directed toward NAMs [45,46]. This equates to roughly 45 million euros per year across the six-year programme. Assuming a modest annual redirect of 5% earmarked specifically for human-based NAMs methods and research infrastructures would create a cumulative shift of nearly 120 million euros by 2040. A more ambitious redirect of 10% per annum would create a NAMs budget of 187 million euros by 2035. Both of these options are still modest (and therefore easily achieved) given overall Horizon Europe funds of €95.5 billion [45], and taking into account that all framework programmes have increased their overall budgets since their creation in 1984. It appears that investment in NAMs lags behind somewhat: while the overall budget from Framework Programme 7 to Horizon 2020 (which were the funding programmes prior to Horizon Europe and were active between 2007–2013 and 2014–2020, respectively) increased by 40% [47], spending on NAMs has remained stable over the same period according to the Commission [48]. With the budget from Horizon 2020 to Horizon Europe increasing again by almost 24%, it is time for more substantial investment in NAMs [49].More clarity is needed to fully and fairly assess the EU’s budgetary commitments to NAMs. Within the CORDIS database, which collates EU research funding, there is currently no tracking mechanism to identify grants awarded to animal compared to non-animal research. A sound funding strategy will require tracking mechanisms to be developed and implemented. In addition, an approach based on the one described by the IMI which will “eliminate poorly predictive animal models” should be formally adopted in order to promote the development and use of human biology-based tools, promoting translation and maximising return on investment.Along with restricted, ring-fenced funding for NAMs, a shift in application “focus” may be necessary here to ensure success. This could mirror recent calls for multi-disciplinary research applications that have seen engineers, mathematicians, clinicians, and biologists, etc., collaborating to great success [50,51,52]. Offering specific funding for collaborations between the existing NAMs developers and users with those researchers who need to make the transition could create the appropriate incentives, and confidence, to drive the shift away from animals. In addition, creating pools of experts on NAMs approaches who can critique these applications will be vital in ensuring that European science remains cutting edge and will provide the necessary return on investment.There may also be a need for change in infrastructure and funding programmes should reflect this. As the transition away from animals occurs, there will be a greatly reduced need for dedicated animal facilities. Initially, identifying opportunities for resource sharing (under Pillar 3) in line with recent advances in the UK [53] and according to other initiatives such as ShARM [54] and SEARCHBreast [55,56] could begin the shift to reduced animal use without compromising research or careers during the initial stages of the transition. Ultimately, however, it would be advantageous to provide infrastructure grants to allow full conversion of the animal facilities, perhaps to create Centres of Excellence for human-relevant research. For example, in 2019, the Centre for Predictive Human Model Systems (CPHMS; https://aic.ccmb.res.in/cphms/ (accessed on 12 February 2022)) was set up by the Government of India’s Atal Incubation Centre—Centre for Cell and Molecular Biology (AIC-CCMB) in collaboration with Humane Society International/India, as India’s first think tank for human-relevant methods.Success for the activities under Pillar 1—driving widespread acceptance of human biology as a gold standard—will require coordinated incentives for researchers to transition away from animals, through dedicated financial support for NAMs alongside education and training.6. Pillar 2—Agile RegulationsAlthough the use of animals for regulatory purposes is outside the scope of this article, for this pillar we offer some suggestions to explore how developing universally accepted standards and harmonising regulations could increase the use and acceptance of NAMs.The need for common standards for NAMs is necessary for their implementation and for improving the confidence in them which will drive their wider adoption, and discussions around standardisation and qualification of NAMs are underway. In 2017, the International Consortium for Innovation and Quality in Pharmaceutical Development (IQ) microphysiological systems (MPS) Affiliate group was formed. One of the goals of this initiative is to qualify MPS—identifying contexts of use and defining the key characterisation data needed to allow the incorporation of MPS-derived data in pharmaceutical safety screening [57]. For biomedical ‘big data’, the lack of standardisation (in terms of ontology, terminology, data format, etc.) and the existence of (often incompatible) heterogeneous databases complicate the application of these data and prevent effective data sharing and data mining [58,59]. The Horizon 2020 programme STANDS4EU aims to “evaluate strategies for data integration and data-driven in silico modelling approaches to develop standards, recommendations and guidelines for personalized medicine” [60]. In silico approaches could include the use of databases, machine learning, artificial intelligence, molecular modelling, along with quantitative structure activity relationships and network analysis tools that permit the development, and crucially, subsequent testing of a model(s). There are additional efforts to apply in silico approaches in medical device and drug development: both the US Food and Drug Administration and the European Medicines Agency are developing guidelines for the use of in silico methods for regulatory purposes [61,62]. In 2021, the European Joint Research Centre, together with the European Standards Organisations CEN and CENELEC organised a workshop entitled “Putting Science into Standards—Organ on Chip: Towards Standardisation” as part of the Putting Science into Standards initiative, which aims to identify areas where standardisation could enable innovation of emerging technologies.Global harmonisation of testing strategies, and standardisation of the tools, will be necessary to improve confidence in NAMs data and enable full replacement of animals. This will require some degree of flexibility in the regulations to adapt to, and accept data from, NAMs.7. Pillar 3—Knowledge TransferThe EU already supports several training and education initiatives. This includes structured collaborations between Member States such as the European Education Area or the European Research Area, the allocation of European Regional Development Funds to universities, or the creation of training courses launched by the Commission itself. These provide an effective vehicle to promote the acquisition of knowledge necessary for broader use, creation, development and application of NAMs.We acknowledge the Commission’s ongoing efforts toward developing training and education resources for scientists as a first step towards the achievement of Pillar 3 by widening participation of key stakeholders through education, training and (re-)training [63]. However, these are centred around the 3Rs, and in order to facilitate a full, just transition, there needs to be a specific focus on replacement. This should include the inclusion of teaching on NAMs approaches across life sciences curricula at higher education, along with the introduction of the more human-relevant, innovative methods at secondary school level and possibly even earlier, as is being addressed for the 3Rs by the Commission with the European Schoolnet collaboration [64]. At the post-graduate level, more practical, hands-on training could be encouraged, perhaps even culminating in a formal qualification, equivalent to those currently used for animal handling [65,66]. In the UK, there is already some evidence of this shift, with the creation of Centres for Doctoral Training (CDT) dedicated to NAMs approaches for biomedical research. For example, lifETIME is the Engineered Tissues for Discovery, Industry and Medicine CDT, a partnership between the University of Glasgow, the University of Birmingham, Aston University and CÚRAM—Science Foundation Ireland. This aims to develop “bioengineered humanised 3D models, microfluidics, diagnostics and sensing platforms” in order to innovate biomedical research and drug discovery, offering formal training in these advanced, human-relevant tools and creating researchers with a clear understanding of the use and application of NAMs approaches (https://lifetime-cdt.org (accessed on 20 October 2021)).We also note the efforts of projects coordinated by the Joint Research Centre that have resulted in the development of various valuable knowledge sources collating non-animal approaches for human diseases. These currently cover respiratory tract diseases [67], breast cancer [68], immuno-oncology [69], and neurodegenerative diseases [70]. These projects offer a snapshot of the state-of-the-art NAMs in use or under development and should be made available to researchers, project reviewers, ethical approval boards and even competent authorities responsible for approving animal research, as a valuable collection of methods that do not require animal use and therefore could contribute greatly to a reduction strategy without adversely impacting research (or researchers).Of course, there is also a need for continuing education beyond graduation. The NAMs tools are evolving at a rapid pace, necessitating lifelong learning programmes of use not just to the researchers developing and using these tools, but also for grant reviewers, publishers, editors, educators and regulators. We see a facilitatory role for the learned societies in formalising and encouraging this within their continuing Professional Development programmes. There could be a requirement for a defined number of hours of “Innovation Engagement” to ensure familiarity with state-of-the-art NAMs methods as these continue to change, improve and are used across more fields of research.Recently a highly cited paper has been analysed as having a worth of around 14,000 USD per annum [71]. Academic researchers exist under the cloud of “publish or perish”, and therefore we must ensure that they have confidence that a shift in methodology to the NAMs approaches will not preclude publication in these high-impact journals. However, there is an increasingly wider recognised issue—that “reviewer three” would like to see additional in vivo data as a condition of publication [72]. Any attempts to reduce animal use have to come with assurances that the careers of researchers will not suffer as a consequence of a shift. This requires coordinated efforts at the levels of grant reviewers, editors, peer reviewers, academic promotion boards, etc. Thus, there is a need for training and education in NAMs tools to provide expert input into grant, ethical and paper review, such that applications or papers that are entirely dedicated to NAMs approaches are not overlooked or “marked down” as a consequence of the inexperience of the reviewers regarding methods presented that prevents a valid critique of the science (Prof. L. Harries, personal communication, 12 December 2020). The expert body currently curated by the European Commission is ideally placed to offer this input [73].Training and education—for every career stage and across all stakeholders—will be crucial to develop and maintain the workforce needed for the success of Pillar 3. There are many existing initiatives that could be implemented and expanded to achieve this.8. Tracking Progress by Developing MetricsIn terms of basic and applied and translational research, and looking at data from 28 Member States only (without Norway), between 2015 and 2018, the average annual percentage decrease in animal use for basic, applied and translational research was 1.6%. Assuming that a 1.6% annual decrease could be sustained, total animal use would reach 50% of current levels by 2061 and would still be above one million uses per year by the year 2095. These are conservative estimates based on the Directive alone, and it is apparent that taking account of the rapid evolution in NAMs development would allow the Action Plan to accelerate the decline in animal use. We have adapted this to examine what three different (worst case, mid case and best case) entirely hypothetical scenarios would look like, using a linear decrease for simplicity, although we appreciate that this is unlikely to reflect the more complicated real-world picture (Figure 3). For the worst-case scenario, we transformed the average percentage annual decrease in use for basic, applied and translational research to the number of animals (100,000) and this gives a shallow decline in animal use such that use does not near zero till 2081. The mid-case scenario (grey line in Figure 3) is based on the average annual reduction in total animal use between 2015 and 2018 and represents a drop of 150,000 uses per year. Under these conditions, animal use reaches half the 2018 levels by 2039 and zero by 2060. The more ambitious best-case uses an annual reduction of 200,000 uses and here we see animal use halving by 2029 and getting to zero by 2040.Using a linear projection is a simplistic view and it is perhaps more likely that a decline in animal use could be less straightforward and will reflect changes in circumstances, funding etc. For example, we have already seen that total animal use in Great Britain dropped by 15% in 2020, with over 30,000 fewer animal uses reported for basic, applied and translational research [74] as a consequence of national lockdowns due to the SARS-CoV-2 pandemic. It seems unlikely that this reduced animal use could be sustained “under normal circumstances” and indeed, it seems that many animals were culled as researchers could not access laboratories [75], rather than a sign that researchers are shifting away from animals toward non-animal approaches. Also, given that an annual reduction of 30,000 animal uses represents 32% of the total animal use for basic, applied and translational purposes in Great Britain alone. If this could be adopted across all member states then a 32% reduction equates to around two million uses annually and therefore would represent a far higher reduction target than even the best-case scenario presented in Figure 3 (200,000 animals a year, or around 3% total animal use) and may be too ambitious as a starting figure.Additionally, compared to the current trend of an annual decrease of 1.6% of animal use, decreasing animal use by 150,000 or 200,000 animals per year represents a considerable “saving” of almost 26 million or 53 million animals, respectively, over the next three decades. Such long-term predictions are hypothetical, but given that over 60 years of “3Rs implementation” has yet to bring us anywhere near the ultimate goal of full replacement, they are worth exploring, in the context of an ambitious and pro-active EU action plan.Some members of the pharmaceutical industry have already achieved substantial reductions in their use of animals for research and development and regulatory purposes. For example, Sanofi has estimated that since 2013, its animal use has decreased by about 45% [50]. Although this is one example, this is a much steeper decline than the one suggested above, and illustrates that dramatic reductions in animal use can be achieved in a short period of time with the appropriate strategy and commitment.It is worth noting that reducing the number of animals used, in isolation, is not an indicator that NAMs have been more widely adopted. A drop in animal use could reflect, for example, extreme conditions whereby laboratories could not operate at full capacity due to a pandemic, a change in reporting requirements, or budget cuts, and this is particularly true where numbers decline for a one-year reporting period only. It is therefore important that we do not rely solely on animal use statistics, despite the phase out of animals being the ultimate goal; we must also consider other metrics, such as funding. These metrics could also exploit information collected as part of Pillars 1 and 3 to record education, training, continuing education, grant applications, publications etc. associated with a shift to NAMs methods. For example, it may be useful to report the number and monetary value of grants dedicated to NAMs, and even track the numbers of patents, publications, citations, etc. as measures of “success” for transitioning biomedical research to human-based NAMs. Defined metrics are therefore needed to monitor these developments and track their impact.It is also true that, although this commentary is focused on the European Union, the activities and initiatives described in the Pillars could have global resonance. Animal use for research and testing is not solely a European issue. It is not possible to accurately quantify the numbers of animals used for scientific purposes globally, but with estimates varying from around 112 million in the United States alone [76], to 192.1 million globally in 2015 [77], it is clear that animals still bear the brunt of biomedical research and testing. Adoption of a phase-in approach across the European Union sends a clear signal that animal use is outdated and could help to drive global changes.9. Concluding RemarksThe historic resolution from the European Parliament provides a call to action to revolutionise the health research paradigm in Europe, recognising human biology as the gold standard, and prioritising funding for the development and application of more predictive tools, based on human biology. We suggest three pillars of activity would be helpful to ensure that innovative science, education and training and regulatory flexibility are taken fully into account. We envision that a redirection of 10% of the EU’s annual research budget, increasing year on year, towards non-animal research employing NAMs, with a reduction of 200,000 animal uses per annum would bring animal use in basic, applied and translational research to a halt in around thirty years, whilst maintaining the EU’s thriving research environment. It is important to consider that phasing-out the use of animals in research is not only about taking animals out of the biomedical research paradigm, it requires the creation of a scientific environment where NAMs, such as microphysiological systems, computational modelling or ‘omics technologies, are accepted as the ‘new normal’ in laboratories; where researchers are equipped with the requisite skills to effectively apply these methods; and where research becomes human biology-focused. If this is accompanied by an increase in funding dedicated to NAMs as we describe, then the developments in science should keep pace with the reduction in animal use. Where poorly predictive animal models have been identified and effectively defunded, this frees-up funding to be shifted to human-relevant approaches. This offers a win-win for science and animals, adheres to the wishes of European citizens, and a better understanding of human biology should have an additional impact on drug discovery and development.

animals : an open access journal from mdpi

10.3390/ani12040504

PMC8868329

Cancer is the leading cause of death in humans and is one of the most common canine diseases. The similarities in pathological features and tumor behaviors between spontaneous canine tumors and their human counterparts make dogs ideal models for comparative cancer research. Legumain is a novel asparaginyl endopeptidase that is overexpressed in numerous types of human tumors. Furthermore, legumain-targeted cancer therapy has been proposed, and the treatment efficacy is well-tolerated. Previous studies have shown that legumain regulates extracellular matrix degradation and triggers the invasion and the metastasis of tumors. However, in dogs, the role of legumain in the progression of tumors remains largely unknown, and few investigations have described the expression levels of this protein in canine tumors. The present study was carried out to evaluate whether legumain is expressed in ten different types of canine neoplasms. We found that heightened signals of legumain were expressed in all canine tumor samples in the study, and, notably, the non-mesenchymal types of tumors harbored relatively high expression levels. This study is the first to describe the legumain distribution pattern in a series of canine tumors. Though further investigation is needed, the current study has provided large-scale pan-screening data on legumain as a potential biomarker, or a therapeutic target, in veterinary oncology.

Legumain, a novel asparaginyl endopeptidase, has been observed to be overexpressed in several types of human solid tumors. Elevated levels of legumain are found in human cancers, and this oncoprotein may facilitate tumor invasion and metastasis when overexpressed. These findings suggest that legumain plays a malignant role in cancer biology. However, currently, no publications have identified the role of legumain in the development of canine cancers. The present study first compared the expression patterns of legumain in paraffin-embedded canine tumor tissues, with those of normal tissues, by immunohistochemistry. A total of 100 canine tumor samples, including mast cell tumors, soft tissue sarcoma, hemangiosarcoma, lymphoma, mammary gland carcinoma, hepatoid gland tumor, squamous cell carcinoma, trichoblastoma, and melanoma were evaluated. Compared with the normal tissues, all tumor samples displayed high intensities of legumain expression. Mesenchymal-type tumors displayed immunoreactivity for legumain, with an average expression of 40.07% ± 1.70%, which was significantly lower than those of epithelial tumors and other types of tumors, which had median expressions of 49.12% ± 1.75% and 47.35% ± 2.71%, respectively (p < 0.05). These findings indicate that legumain has a high potential to be a candidate for distinguishing tumors from normal tissues. Although further studies on a larger number of cases are necessary to clarify the clinical application of legumain, the overexpression patterns of legumain in canine tumor tissues are reported, for the first time, in this study.

1. IntroductionLegumain is a lysosomal protease and a cysteine endopeptidase of asparaginyl endopeptidase (AEP), which displays a high specificity for the hydrolysis of the C13 family [1]. Legumain is overexpressed in several types of tumors and exhibits a high level of mRNA expression in tumors [2]. Limited legumain expression can be detected in normal tissues [3], whereas numerous human cancers, such as ovarian cancer [4], colorectal cancer [5], prostate cancers [6], gastric carcinoma [7], and breast cancers [3] are known to express elevated levels of legumain. Furthermore, its over-expression has been found to be associated with the tumor invasion and metastasis [2], which may result from the characteristics and functions of this protein. Legumain is not only located in the intracellular regions of tumors; it is also highly expressed on the surface of tumors and tumor-associated endothelial cells, where it is colocalized with integrins [8]. The heightened expression of legumain activates gelatinase A zymogen, a vital regulator of extracellular matrix degradation, and, thereby, makes the tumors’ more malignant counterparts [9,10]. Therefore, an increased level of legumain on the cell surface could trigger the invasion and metastasis of tumors [10]. The malignant behaviors of tumors are also proposed to be related to the proteolytic function of legumain, which can activate other protease zymogens. Some of these activated proteases are linked to angiogenesis and tumor cell proliferation. Legumain-induced protease cascades are highly correlated with coagulation, apoptosis, complement cascades, and other tumor-promoting biological pathways [2,5]. Thus, legumain has been regarded as a candidate tumor antigen that is overexpressed in several types of cancers and facilitates tumor malignancies.Companion dogs are ideal models for comparative oncology because they live relatively close to, and share highly similar tumor behaviors with, humans. In investigations of tumor biology, a canine model can mimic the actual immune status of humans because they are outbred, compared with other laboratory animals, providing a diverse gene background similar to that in human populations [11]. In dogs and humans, cancer initiation and development are influenced by several factors, including age, nutrition, immune status, and environmental stimulation [12,13,14]. Furthermore, a diverse range of cancers can be found in dogs and humans, and the histopathological features are highly similar [11]. In addition to these properties, the most important property is that the deciphering of the canine genome shares high similarities with that of the human genome [15,16], and the genetic molecular alterations that drive cancers in dogs and humans are highly analogous. The oncogenes and tumor suppressors that have been identified in human cancers, such as the mutation of KIT [17] in mast cell tumors, and the alteration of p53 in mammary gland tumors [18,19], also contribute to canine neoplasms. Numerous tumor-related genes and/or antigens were initially identified in humans and were later found to have similar tumor promotion roles in canine cancers.Legumain, a tumor-associated antigen, has been widely investigated in human oncology to determine its malignant roles and its potential tumor-promoting mechanisms; however, few studies have considered the expression of legumain in tumor-bearing dogs. This study aimed to evaluate the expression of legumain in tumor-bearing dogs, because tumor antigens are useful tools in cancer diagnoses, monitoring, prognoses, and even as therapy targets. In addition, whether the anti-legumain polyclonal antibody can cross-react with dogs needs to be considered. Therefore, the main purpose of this study was to compare the legumain expression in various canine tumors with that in normal tissues. Second, this study also assessed the utility of an anti-legumain polyclonal antibody that can specifically recognize canine tumors.2. Materials and Methods2.1. Selection of SamplesAll the tumor samples were collected for diagnostic purposes at the National Taiwan University Veterinary Hospital (NTUVH), Taipei, Taiwan, and the investigators had no influence on the execution of any clinical procedures. The owners provided informed consent to use the clinical data and the excised tumors for teaching and research purposes. A total of 100 tumor samples obtained during surgery between 2016 and 2021 were selected, including the mast cell tumor (MCT, 20 cases), soft tissue sarcoma (STS, 10 cases), hemangiosarcoma (HAS, 10 cases), lymphoma (10 cases), mammary gland tumor (MGT, all were carcinomas, 10 cases), hepatoid gland tumor (HGT, 10 cases), squamous cell carcinoma (SCC, 10 cases), trichoblastoma (10 cases), oral melanoma (OM, 5 cases), and skin melanoma (SM, 5 cases). According to their origin, these tumors were further allocated into three groups: mesenchymal-type tumors (MCT, lymphoma, HAS, and STS), epithelial-type tumors (SCC, MGT, HGT, and trichoblastoma), and melanomas (OM and SM). Mesenchymal cells originate from the mesoderm [20], whereas melanomas are neoplasms of neuroectodermal origin [21]. Therefore, melanomas were excluded from the mesenchymal tumor group. Samples of the control canine bodies were immediately collected upon body donation. No causes of death in the control group were related to tumors. The owners also signed informed consent to use the clinical data and the samples for research purposes. All tissues were fixed in 10% neutral formalin and were paraffin-embedded. All the tumor samples were histologically diagnosed by two veterinary pathologists of the Graduate Institute of Molecular and Comparative Pathobiology, National Taiwan University, Taipei, Taiwan. Histopathology slides and formalin-fixed, paraffin-embedded (FFPE) tissue sections were retrieved, and all slides were further reviewed by the same veterinary pathologist (Wei-Hsiang Huang) to increase the validity of diagnoses in this retrospective study. This study was approved by the Institutional Animal Care and Use Committee, National Taiwan University (NTU-105-EL-00007).2.2. ImmunochemistryTo evaluate the legumain expression, blocks were sectioned at 4 µm and deparaffinized in non-xylene (Muto Pure Chemicals Co., Ltd., Tokyo, Japan), and then they underwent antigen retrieval in 1× Trilogy solution (pH = 7.0) (Sierra College Boulevard, Rocklin, CA, USA, diluted with deionized water) at 114–121 °C for 5 min in a pressure cooker (Montage Opus™ (Diagnostic BioSystems, CA, USA)). The Novolink Polymer Detection System (Leica Biosystems, Wetzlar, Germany) was applied for the immunohistochemistry. To block non-specific endogenous binding, sections were treated with a peroxidase block solution (Novolink™ Polymer Detection System, Newcastle, UK) for 5 min and were then incubated with a protein block solution (Novolink™ Polymer Detection System, Newcastle, UK) for 5 min. Slides were incubated with a primary antibody, the rabbit anti-legumain polyclonal antibody, diluted by 1:200 (LSBio, catalog no. LS-B15611, Seattle, USA) at room temperature for 1 h, followed by a Post Primary solution at room temperature for 30 min (Novolink™ Polymer Detection System, Newcastle, UK), and then the Novolink Polymer at room temperature for 30 min. An AEC (3-Amino-9-ethylcarbazole) substrate was used to develop the staining reaction, and the nuclei were stained with hematoxylin. The sections were stained with an anti-legumain antibody, and in the negative controls, sections were incubated with the normal serum (diluents) instead of the primary antibody.2.3. Evaluation of ImmunochemistryTo evaluate the legumain expression in all tissue sections, the estimated percentages of the immunopositivity of each slide were independently analyzed in five random microscopic fields at 40× objective magnification (0.1 mm2) and were then separately analyzed by two veterinary pathologists, W.-H.H. and C.-S.L. (NTUVH), who were blinded to the experimental history. To avoid artifacts, sections in the areas with necrosis, and areas in the margins of the tissues, were not considered.2.4. Cell CultureCanine diffuse large B-cell lymphoma (CLBL-1) [22] and T-cell lymphoma (UL-1) [23] cell lines were maintained in RPMI-1640 (Thermo Fisher Scientific, catalog no. 11875093) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, catalog no. 10100147) and 1% antibiotic–antimycotic (Simply, catalog no. CC501-0100). Canine mammary gland tumor cell lines, CMT-1 [24] and MPG, and canine melanoma cell lines, KMec [25] and M5 [26], were cultured in Dulbecco’s Modified Eagle Medium (DMEM, HyClone, catalog no. SH30243.02) containing 10% fetal bovine serum and 1% antibiotic–antimycotic. All the cell lines were cultured in a humidified incubator with 5% CO2 at 37 °C.2.5. Protein Extraction and Western Blot AnalysisTo validate the specificity and cross-reactivity of this polyclonal antibody with canine tumors, a Western blot analysis was performed using different tumor cell lines. All canine tumor cell lines were harvested, washed three times with PBS, and were lysed in a RIPA lysis buffer (Sigma-Aldrich, St. Louis, MO, USA, catalog no. R0278). After 30 min of incubation at 4 °C, the proteins were collected by centrifugation at 12,000× g for 20 min. The concentration of proteins in each sample was quantified by a BCA protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Then, the protein samples (20 µg) were separated into 10% acrylamide SDS-PAGE gel and were transferred to a PVDF membrane. The membrane was blocked with a blocking buffer (5% skim milk in TBS) for 1 h at room temperature and then incubated with an anti-legumain antibody or anti-alpha tubulin (Abcam, catalog no. ab7291) at a 1:1000 or 1:5000 dilution in the blocking buffer overnight at 4 °C. The membrane was then washed with TBST (0.05% Tween-20 in TBS) for 30 min and incubated with a horseradish peroxidase (HRP)-conjugated anti-rabbit (Invitrogen, catalog no. 65-6120) or anti-mouse (Invitrogen, catalog no. 62-6520) secondary antibody, diluted to 1:5000 in 1% skim milk for 1 h at room temperature. After washing, the signal was developed by chemiluminescence using the ECL prime kit (Bio-Rad Laboratories, Hercules, CA, USA).2.6. Statistical AnalysisThe data were described as the mean ± standard error of the mean (SEM). Statistical analyses were performed in GraphPad Prism version 8. To determine significant differences, the Kruskal–Wallis test was utilized. Differences were considered statistically significant at a p-value of less than 0.05.3. Results3.1. Specificity of the Anti-Legumain Polyclonal AntibodyThe results of Western blot analysis are shown in Figure 1. The anti-legumain polyclonal antibody recognized a dominant band near 54 kDa (NCBI reference sequence: XP_038528856.1) in the canine tumor cell lysates, indicating that this polyclonal antibody can identify the correspondent proteins in the canine species.3.2. Low Expression Levels of Legumain Were Identified in Normal TissuesTo probe the primary antibody that would identify the normal tissues in dogs, we examined IHC in normal tissues as the control groups. The expression levels were shown as relatively negative signals of the whole figures. Figure 2a shows that there was no significant signal of legumain in the stroma of normal tissues, including the skin (Figure 2b, 3.44% ± 0.68%), thymus (Figure 2b, 7.88% ± 0.31%), muscle (Figure 2b, 0.67% ± 0.19%), and nerve (Figure 2b, 4.14% ± 0.49%) tissues. Of note, in the epidermis and in hair follicles, histiocytes demonstrated moderate-to-strong cytoplasmic legumain positivity, but the muscles, fibroblasts, and lymphocytes showed weak-to-moderated cytoplasmic immune expression.3.3. High Expression of Legumain in Ten Types of Canine NeoplasmsThe statistical analysis revealed that the expression percentages of legumain were significantly higher in all tumors (Figure 3) than in the normal tissues (Figure 2). Based on the source of tumor differentiation, we further divided these tumors into three groups: mesenchymal tumors (MCT, lymphoma, HAS, and STS), epithelial tumors (SCC, MGT, HGT, and trichoblastoma), and melanomas. The statistical analysis is summarized in Figure 3, and the representative figures of the tumors are separately provided in Figure 4 and Figure 5. Low-grade (Figure 4a) and high-grade MCTs (Figure 4b) expressed positive signals of 38.73% ± 3.51% and 44.05% ± 3.86%, respectively, without a significant difference (p > 0.05). In lymphoma cases, legumain expression was 34.23% ± 4.31% (Figure 4c). Legumain expression was observed in the cytoplasm of these round cell tumors (MCT and lymphoma). The low-grade MCTs exhibited majorly diffuse cytoplasmic patterns but occasionally exhibited vesicular patterns. In contrast, the high-grade MCTs and the lymphomas exhibited majorly vesicular and diffuse patterns. Notably, numerous neoplastic cells of the high-grade MCTs and lymphoma had occasionally moderate nuclear staining. HAS (Figure 4d), and STS (Figure 4e) had positive signals of 44.95% ± 3.49% and 38.38% ± 3.46%, respectively. The immunopositivity of legumain was also located in the diffuse cytoplasm, while STS exhibited weak vesicular positivity. The immunopositivity of legumain was also located in the diffuse cytoplasm, while STS exhibited occasionally vesicular positivity. SCC (Figure 5a) and MGT (Figure 5b) had legumain expression signals of 40.83% ± 1.95% and 49.89% ± 1.78%, respectively. Among all the samples, the HGT (Figure 5c) had the highest expression level of legumain at 57.09% ± 5.10% in the tumor sections. Trichoblastoma (Figure 5d) had a legumain signal of 48.68% ± 2.28%. SCC and MGT exhibited vesicular and diffuse patterns. The HGT and trichoblastoma exhibited majorly diffuse and tiny dotted, but occasionally vesicular, patterns. The stomal cells of these tumors exhibited almost negative signals. Oral melanoma (Figure 5e, 49.21% ± 3.80%) and skin melanoma (Figure 5f, 45.49% ± 4.11%) had no significant differences in legumain expressions. Both these melanomas exhibited strongly vesicular cytoplasmic positive signals. In summary, legumain is overexpressed in tumors; however, no significant differences in the expression of legumain were found among the ten types of tumors. To probe the possibility of different patterns among mesenchymal-type cancers, epithelial-type cancers, and melanomas, we further analyzed the expression levels among these groups.3.4. Heightened Expression of Legumain in Non-Mesenchymal TumorsA total of 50 cases, distributed in four types of tumors (MCT, lymphoma, HAS, and STS), belonged to the mesenchymal-type tumor, with an average legumain expression of 40.07% ± 1.70%. Epithelial-type tumors, including SCC, MGT, HGT, and trichoblastoma, had significantly higher legumain expressions (on average, 49.12%; SEM, 1.75%) compared to mesenchymal-type tumors (Figure 6, p = 0.0177). The average expression of 47.35% ± 2.71% in melanomas was significantly higher than that in mesenchymal-type tumors (Figure 6, p = 0.0284). These results indicate that legumain is overexpressed in tumors, especially in non-mesenchymal-type tumors.4. DiscussionThe discovery of ideal tumor antigens that can monitor and/or predict the progression of tumors remains an important issue in human and veterinary medicine. Many tumor antigens with a prognostic value have been widely used in human medicine due to the urgent need for precision medicine. Likewise, in veterinary medicine, the identification of reliable tumor markers has gradually increased. The malignant role of legumain has been proposed; however, few publications to date have considered the malignancy of legumain in veterinary medicine. Here, we reported, for the first time, that legumain, a widely-used cancer biomarker in human oncology, is also significantly expressed in ten types of canine neoplasms.To the authors’ best knowledge, this is the first study to evaluate legumain expression in ten types of canine tumors. Thus far, little is known about the biological functions and processes involving legumain in tumor development in dogs. However, in human oncology, the overexpression, or the potent malignant mechanisms, of legumain have been observed in ovarian cancer [4], colorectal cancer [5], prostate cancers [6], gastric carcinoma [7], and breast cancers [3]. Even though the direct correlation between tumor progression and legumain has not been widely described, the presence of cysteine endopeptidases, such as cathepsins B and L, which are activated by the legumain-mediated hydrolysis of asparaginyl bonds, may contribute to tumor malignancy [8]. This legumain-induced hydrolysis also leads to the activation of protease zymogens, which are highly correlated with angiogenesis and other tumor-facilitating functions [2,5]. Furthermore, legumain regulates the remodeling of the extracellular matrix through the initiation of the gelatinase A zymogen which, thereby, promotes tumor progression [9,10]. These underlying mechanisms may explain the overexpression of legumain found in the canine tumor tissues in the current study. We speculated that, first, through the overexpression of legumain, tumors possibly acquire the ability to display tumor-promoting phenotypes (through the extracellular matrix remodeling) and, thus, the legumain-mediated hydrolysis results in the malignant growth of tumors, especially in epithelial-type tumors.In human medicine, Gawenda et al., reported three staining patterns of legumain, namely, diffuse positivity, tiny dots, and vesicles, in the cytoplasm, and found that breast tumors with legumain, expressed as a vesicular positivity, were correlated with a worse prognosis [3]. Ohno et al., also reported similar findings for prostate cancers, which showed that the vesicular staining patterns of legumain in these tumors had a positive correlation with tumor invasion and aggressiveness [6]. Similarly, in gastric tumors, the vesicular positivity of legumain was found in the cytoplasm of these neoplasms, whereas diffusely positive staining patterns were displayed in the normal mucosa tissues [7]. Taken together, these results indicated that the vesicular positivity of legumain in the cytoplasm could serve as a potential predictor for tumor malignancy. In the current study, we found that, in all the canine tumors, legumain was expressed in the cytoplasm. Interestingly, epithelial-type tumors and melanomas displayed a predominant vesicular positivity of legumain. Vesicles were scattered in the cytoplasm. In contrast, the mesenchymal-type tumors expressed predominant diffuse cytoplasmatic patterns, especially in the mast cell tumors, lymphoma, and hemangiosarcoma. Previous studies focused on the expression patterns in one specific type of tumor using a large sample size. In this pilot study in veterinary medicine, however, we failed to find a correlation between the tumor malignancy and the expression patterns of legumain in one type of canine tumor. Further investigations in veterinary medicine, with larger sample sizes, are needed to clarify whether the expression pattern of immunoactivity located in the cytoplasm is also correlated with the prognosis, as widely proposed in human medicine.Our study shows that legumain is overexpressed in non-mesenchymal-type tumors. The underlying mechanisms by which these tumors displayed high levels of legumain remain unclear, even in human medicine. One previous study reported that the expression of legumain in tumor epithelial cells is associated with the potential epithelial cell-intrinsic role of early-stage tumors in the extracellular matrix degradation that facilitates tumor progression [27]. Furthermore, legumain could induce endopeptidase activity [8,28], which is most intense in the epithelial region [29]. Lastly, the expression patterns of legumain in the cytoplasm are correlated with tumor malignancy [3,6,7]. The majority of non-mesenchymal-type tumors exhibited strong positivity, and predominantly vesicular positivity, in the cytoplasm, especially malignant tumors, such as SCC, MGT, OM, and SM, and, thus, these tumors were expected to express more tumor-associated molecules. Even though no direct evidence of a connection between tumor development and legumain expression has been widely presented, we speculate that these characteristics may explain why legumain was overexpressed, especially in the epithelial tumors. However, further functional studies are highly recommended if the functions of legumain expression in canine carcinogenesis are to be uncovered.The high levels of legumain expressed by tumor cells suggests that the inhibition of legumain is a potential strategy against tumor progression, based on its facilitation of tumor growth and its specific expression in tumors [5]. Legumain-specific targeting therapy, using nanoparticles, has been effective for breast cancer in a mouse model without significant toxicity [30]. The relatively limited expression of legumain in normal tissues [3,31] and its overexpression in tumor-associated endothelial cells [8,31] were also found in the current study. These findings indicate that legumain has the potential to serve as an ideal therapeutic target in tumor-bearing dogs, especially in dogs with epithelial-type tumors. Though we did not identify either the intensity or the percentage of legumain expression in any specific tumor type, we did find that non-mesenchymal tumors expressed an elevated level of legumain. However, it would be more prudent to verify these findings using a larger sample size to support our results. Taken together, the findings show that legumain was overexpressed in the canine tumor samples in this study, and further studies are needed to elucidate these findings with a more comprehensive investigation.In the original Western blot analysis (Figure S1), there were two bands in the M5, KMec, CMT-1, and MPG cell lines, since, in UL-1 and CLBL-1 cell lines, this does not happen. Besides, there is a positive signal in the near 75kDa band. To our best knowledge, this is the first study to investigate legumain expression in canine tissues; we cannot find any clues why there are variations among the different cell lines. One of our possible speculations for the 75kDa band may be the different legumain isoforms. From the NCBI protein database, at least nine isoforms were recorded (while it seems only three complete isoform amino acids were shown with the corresponding molecular weight around 50 kDa). Therefore, we hypothesize the 75kDa band might be one of the isoforms while more data are needed to verify the hypothesis. However, for the antibody we used in this study, the immunogen is the partial amino acid sequence of human legumain. This sequence identity between human and canine species is 91% to prove this antibody could specifically react with canine legumain. In this study, we found increased levels of legumain in canine tumors; however, this study had some limitations. First, this was a preliminary investigation that involved screening ten types of canine neoplasms with a relatively small sample size (n = 5–20) per tumor, and further investigations that focus on certain types of tumors are needed to support our findings. Then, the isoform of the canine legumain should be elucidated with further investigation. Studies with more biological experiments, such as examining functional assays using the knock-down or overexpression of legumain in canine cancer cell lines, could possibly validate the roles and even the mechanisms of legumain in canine tumorigenesis. Overall, even though more studies are needed to address these issues, the current study is the first to demonstrate that legumain has a high potential to serve as a biomarker in dogs with cancers.5. ConclusionsIn conclusion, this study first identified the heightened levels of legumain in canine tumors, which suggested that legumain may also act as a tumor antigen in the promotion of tumor development. Additionally, the levels of legumain significantly increased in epithelial- and other non-mesenchymal-type tumors. The overexpression of legumain potentially facilitates tumor formation in dogs with cancer, indicating that legumain could be a potential tumor biomarker and/or therapeutic target, with more investigation. Though further studies are necessary to define the clinical value and the usefulness of legumain as a therapeutic candidate, the elevated patterns of legumain in canine tumors are reported, for the first time, in this study.

animals : an open access journal from mdpi

10.3390/ani12070855

PMC8996969

Electrically powered devices and power lines generate electromagnetic fields. Technological development has resulted in environmental pollution with anthropogenic electromagnetic fields. One of its components is the magnetic field. Its impact on living organisms is still under investigation, but there are reports suggesting that the direction of change is negative. Pollinators are very important for the environment and are also exposed to this factor. In this study, we wanted to investigate the impact of magnetic field exposure on the behavior of one of the key pollinators: the honeybee. The frequency of the tested field corresponded to those present under high voltage lines, as honeybees often forage in these areas. The results showed that the magnetic field caused a distribution in behavioral patterns, which may have a direct impact on foraging efficiency and pollination success.

Earth’s magnetic field (MF) plays an important role for many species, including the honeybee, in navigation. Nowadays, much larger alternating fields are emitted by miscellaneous electric infrastructure components, such as transformers and power lines, and the environment is therefore polluted by an anthropogenic electromagnetic field, though little is known regarding its impact on living organisms. The behavior of animals is the first and easiest way to establish the impact of stress. It shows if the animal can detect the exposure and react to it. To investigate this, one-day-old bees were exposed to a 50 Hz magnetic field of induction at 1 mT and 1.7 mT for 10 min, 1 h, and 3 h under laboratory conditions. All groups exposed to the magnetic field showed differences in behavioral patterns. What is more, they presented a behavior absent in the control: loss of balance. There were differences, both in the ratio of behaviors and in the number of bouts—exposed bees more often changed behavior. Occurrence of differences is an indication of the reaction of the honeybee organism to the magnetic field. Loss of balance is a disturbing symptom, and behavior changes indicate a disturbance of the honeybee by the electromagnetic field.

1. IntroductionPollination forms the basis of complex ecological systems and is essential for agricultural production. It is estimated that 75% of main crops need animal pollinators [1]. The financial benefits of pollinators are estimated at USD 153 billion, or 9.5% of the total value of the world food market [2]. Bees are the main pollinators in terrestrial ecosystems, and it is through them that both biological and genetic diversity is maintained [3,4]. An invaluable role in this process can be attributed to the honeybee (Apis mellifera L.). Nowadays, the honeybee population faces many harmful factors and stressors: parasites and pests, such as Nosema ceranae and Varroa destructor mite [5,6,7]; monocultures [8], plant protection products [9]; environmental pollution [10]; climate changes [11]; and artificial electromagnetic fields [12,13].An electromagnetic field (EMF) consists of two components: an electric field (EF), with intensity expressed in V/m, and a magnetic field (MF), with magnetic induction expressed in tesla (T).Natural and anthropogenic sources of EMF in the environment can be distinguished. Natural EF can be visible as storms and lighting prompted by the interaction of electrically charged air masses [14]. Natural MF is the cause of magnetosphere occurrence. The magnetosphere is the space above the ionosphere, around the Earth. It extends tens of thousands of kilometers into space and forms a shield that deflects streams of high-energy solar particles (solar wind). It protects the ozone layer and thus the planet’s surface from overexposure to ultraviolet radiation. The absence of this buffer zone would be disastrous for life on the planet’s surface. The source of Earth’s magnetic field is due to the structure of the planet’s interior. It is assumed that, due to the friction between the inner nucleus and the outer liquid nucleus, an electrostatic charge is created. The value of the magnetic field induction depends on the latitude and is, for example, 65 μT at the magnetic poles and 27 μT at the magnetic equator [15]. As electric charge movement induces electromagnetic radiation, every electrical device generates EMF. Radio-frequency radiation is intentionally generated and targeted from space satellites and ground localized transmitters for communication. The frequency of the generated EMF depends mainly on device type, but intensity and magnetic induction mostly depends on the distance from the device [16,17].Migratory animals are most affected by magnetic fields. As well as birds, there are many migratory animals, including dolphins [18], sea turtles [19,20], salmon [20], or insects such as the monarch butterfly (Danaus plexippus L.) [21]. Over the years, these animals and many others have developed their own navigation systems that allow them to reach their destinations, often thousands of kilometers away. A magnetic sense has also been discovered in invertebrates: insects, including bees and wasps [15], and mollusks [22,23]. Honeybees (Apis mellifera L.) have the ability to remember the position of a food source in relation to magnetic field lines. This information is communicated to other members of the hive through waggle dances [15,24]. It was suggested that artificial MF greater than 500 µT can disturb honeybee magnetic navigation [17,25].EMF is a very common factor in the environment, but its influence on living organisms is still poorly understood. It was observed that EMF affects biological systems [26]. Electric power transmission lines generate EMF with a frequency of 50 Hz or 60 Hz, depending on local regulations. Therefore, EMF of this frequency is very common in terrestrial environments, including areas where bees are foraging and hives are situated. As early as 1981, Greenberg et al. [12] observed significant disturbances in bee colonies under EMF exposure. Hives were long-term exposed to a 7 kV/m 59 µA and 85 µA field (corresponding to a field under high-voltage power lines). This led to hive entrances aberrant propolisation, queen loss, and a decrease of capped brood (while the number of eggs and larvae were normal). Additionally, winter survival was decreased in EMF-exposed colonies. These findings led to a recommendations in the USA to not keep bee colonies under power lines [17].As honeybees, unlike most other farm animals, live in an uncontrolled environment, both the life of the individual honeybee and that of the whole colony is dependent largely on themselves: on foraging success, parasite control, raising offspring, seasonal changes in colony structure, and answers to other environmental challenges. Therefore, the ability to proceed with all complex and connected activities is necessary to keep the colony alive. What is more, all of this translates into pollination success, making it crucial for the existence of the environment as we know it. The results of behavioral studies are well visible and have direct impacts, both on individual life or death and on whole colony integrity, even when mechanisms of behavior determination have not yet been fully investigated. This is why behavioral studies on honeybee are so important. Behavioral changes are investigated as a response to stressors, such as parasites and diseases [27] or pesticides [28].The aim of this study was to investigate behavioral changes after exposure to 50 Hz MF of 1 and 1.7 mT magnetic induction and various exposure time variants in honeybee workers.2. Materials and Methods2.1. Research MaterialIn the experiment, one-day-old worker honeybees were used. For research, colonies of Apis mellifera carnica were chosen as a brood source. Frames with a brood capped at 20 days of age were taken from the hive to an incubator with a temperature of 34 °C and humidity of 70–80%. Newly emerged worker bees were gently carried by hand to wooden cages; 100 bees to one cage. Each group consisted of 10 cages. Cages with available bee sugar syrup 1:1 (w:v) were maintained in an incubator for 24 h. Worker bees were fed ad libitum.2.2. Exposure to Magnetic Field (MF)A uniform alternating 50 Hz magnetic field was generated in a solenoid with a diameter of 350 mm and a length of 350 mm. The solenoid was supplied by a stabilized and controlled mains-powered sinusoidal current source. The distribution of magnetic induction in the testing area was measured and controlled using an ESM-100 S/N 972153 m calibrated by an accredited calibration laboratory, AP-078 (calibration certificate LWiMP/W/85/21). The measurements were carried out by the accredited testing laboratory, LWiMP AB-361. Induction non-uniformity in the whole measuring area did not exceed 5%. Due to the fact that the inner and outer surfaces of the coil winding were electrostatically shielded, the 50 Hz electric field inside the coil did not exceed 100 V/m.Cages with dimensions of 200 mm × 150 mm × 70 mm were taken to the magnetic field (MF) emitter, and then emission began. There was only one cage in the emitter during exposure. There were three different times of exposure and two magnetic field intensities (6 experimental groups in total and 1 control group).Times of exposure:10 min—time responding to a short flight, such as to collect water or for defecation1 h—the mean time of a forage flight3 h—long forage time, necessary to bring a heavy load over a long distanceMagnetic field intensities:1 mT1.7 mT2.3. Behavioral AnalysisImmediately after exposure, six randomly chosen bees from each cage were taken to a glass container with a height of 15 cm and a diameter of 20 cm. Bees were filmed for 300 ± 1 s using a SONY HDR-CX240E camera (Sony Mobile Comunications, Lund, Sweden). During the whole experiment, 420 worker bees were used. The videos were later analyzed offline. Seven types of behavior were distinguished:Walking—walking on the base surface or walls of the containerFlight—short episodes of lifting up by wing movementBody cleaning—cleaning own body by legsContact between individuals—any kind of near contact between bees, including touching antennas, trophallaxisWings movement—the rapid movement of wings, used for ventilation, does not cause lifting upStillness—staying motionlesslyLoss of balance—bees from the walls of container fall and land on the bottom of the container upside down.All analyzed types of behavior were mutually exclusive. A 300 s sample from each video was analyzed three times, one per bee. For analysis, we chose the mean total time per bee (how much time bees from one group spent on the behavior) and the number of individual behavior occurrences (how many times during the observation individuals from the group displayed the behavior). Each behavior was immediately marked from its occurrence to its end. The end of one behavior was the start time of another behavior. The recording of the bees came immediately after the end of exposure to the magnetic field.2.4. Statistical AnalysisData were analyzed in RStudio (R Core Team), using packages “dplyr”, “tidyr”, “agricolae”, and “ggplot2” for visualizations. A Shapiro–Wilk test was used to verify the normality of data distribution, a Kruskal–Wallis test with Holm correction was used for multiple comparisons, and α = 0.05 was used to check the significance of differences between groups.3. ResultsDetailed data of total time spent on behaviors per bee and the mean number of behavior occurrences are presented in Table 1 and Table 2. Schemes of behavior patterns are visualized in Figure 1 and Figure 2. Figure 3 shows the number of bees that presented particular behavior.Four out of the seven distinguished behaviors were presented by almost all observed individuals in all groups: walking, flight, body cleaning, and contact between individuals (Figure 3). Walking was the main behavior within all groups, as this presented the highest values, both in total duration time and number of occurrences. All exposed groups, except 1 h 1.7 mT, had significantly more occurrences of walking compared to the control (Table 2), but total duration time presented no differences (Table 1).In the case of total time spent on flight, body cleaning, and contact between individuals, all of the EMF-exposed groups presented significant differences compared to the control, but there were some differences between exposed groups: group 3 h 1 mT had a lower value of time spent on contact between individuals compared to all other treatment groups, and less time spent on flight, but only compared to the 10 min 1.7 mT group (Table 1). Average times spent on body cleaning differed slightly between groups, but varied between individuals, so the differences were not statistically significant.If we consider the number of behavior occurrences in the case of flight, body cleaning, and contact between individuals, the differences are more visible. Group 3 h 1 mT presented significantly more occurrences of flight compared to the control. More occurrences of body cleaning than the control were seen in all groups exposed to the 1.7 mT field, as in the case of contact between individuals, but this time the 10 min 1 mT group also presented significantly higher values.All of the behaviors present in the control were also present in all groups exposed to MF, but there were behaviors present in the treatment groups that were absent or very poorly inherent in the control (Figure 3). When all individuals in the group did not present behavior, the value was marked as “NO” (not observed) and excluded from statistical analysis as non-numerical data (Table 1 and Table 2). Wing movement was generally the rarest behavior and was absent or very rare in most of the groups, including the control. Only in the 10 min 1 mT group was the situation different—wing movement was only slightly often. Stillness was presented in the control by only 1 bee, while in each treatment group this behavior was quite often and presented by more than half of the observed individuals, but in total time spent on this behavior and the number of bouts, significant differences did not appear. The most time on this behavior was seen in groups 3 h 1 mT and 1 h 1.7 mT. Loss of balance was a behavior totally absent in the control, but present and quite common in all exposed groups.Group 10 min 1 mT, on average, spent a significant amount of time on wing movement, but as the behavior was poorly presented in other groups and was not very often seen in this group, it can be considered as marginal behavior. A significant amount of time was spent on stillness in groups 3 h 1 mT and 1 h 1.7 mT, group 3 h 1 mT spent very little time on contact between individuals, and group 1 h 1 mT spent only a little time on body cleaning. Groups exposed to 1 mT MF presented more appearances of loss of balance behavior than groups exposed to 1.7 mT MF. All of these differences were well visible on plots and were distinctive for the treatment groups, but not significant statistically.All groups exposed to MF presented significant differences in behavior compared to the control. In particular, the number of bouts mostly diverged and a much higher number occurred after exposure in all groups, considering both single behaviors and total number of bouts. Another well-visible effect of exposure is the appearance of loss of balance—a behavior absent in the control group.4. DiscussionIn this study, the effects of EMF on honeybee behavior were evaluated. Through observation of bee behavior after MF exposure, the behavioral patterns of young workers exposed to 1 mT and 1.7 mT MF for 10 min, 1 h, and 3 h were assessed. Exposure to MF proved to significantly affect the behavioral patterns. It is difficult to determine the direction of changes with increasing time or intensity of exposure, while each group presented individual behavioral patterns. In many cases, however, all exposed groups had similarities significantly distinguishing them from the control.Well-visible differences occurred in the number of behavior occurrences: exposed groups presented a much higher number of bouts compared to the control, while the ratio of behavior duration did not differ so significantly. This means that bees changed behavior significantly more often, and the frequency of changes was higher. Considering the number of bouts, the impact of an electric field (EF) might be distinguished as oppositely directed: given exposures to a 5 kV/m, 11.5 kV/m, 23 kV/m, and 34.5 kV/m field by 1 h, 3 h, 6 h, and 12 h, it could be seen that there was a slight general tendency for a decrease in the number of behavior occurrences and an increase of mean duration time [29,30].An alternated mobility associated with MF influence has been demonstrated for invertebrates. Shepherd et al. [16], by tethered flight experiments on honeybees, showed that during exposure to 0.1 mT, 1.0 mT, and 7.0 mT MF, wingbeat frequency was increased, with a greater effect at higher exposure levels. The influence of MF on wingbeat frequency has also been demonstrated in locust [31]. A static 50 mT MF modulated the motor behavior of Tenebrio molitor and T. obscurus, but interestingly, even in such closely related species, the effects were remoted [32]. In our studies, the mobility-associated behaviors were walking and flight. Neither MF-exposure variant showed alterations in the ratio of total time spent on these behaviors. The differences were, however, noticeable in the number of bouts, where exposure prompted an increase in the number of occurrences. Therefore, a given behavior lasted much shorter but occurred more often.Clearly differentiating from the control is the appearance of loss of balance behavior after MF exposure. In each group exposed to MF, this behavior was presented by more than half of the observed individuals, except in the 1 h 1.7 mT group, where, in 4 out of 9 individuals, this behavior occurred, which is still almost half. In general, this behavior both occurred more often and took more time in the behaviors ratio in bees exposed to 1 mT compared to 1.7 mT, while the time of exposure did not have such a visible impact. However, these differences are not statistically significant as wide variations between individuals occurred. Similar behavior, described as problems with movement coordination, trembling, tumbling, and lying upside down, has been observed as a result of poisoning after oral administration of neonicotinoids pesticides [28,33,34]. Williamson et al. [28] also observed this behavior in the control, but it intensified after pesticide exposure. In research on the impact of 50 Hz EMF on honeybee behavior, but with natural magnetic components and generated increased electric components, such behavior was not noted, either in the control or in the EF-exposed groups [29,30].In our study, only unrestrained behavior was observed in non-demanding and simple environments. This proves that, even if we do not enforce specific reactions on honeybees, the impact of MF on behavior is noticeable. Other studies have also focused on the analysis of behavior, such as reactions to stimuli and success in task achievement, under MF exposure. It was demonstrated that MF of magnitudes 100 µT and 1000 µT reduce honeybee olfactory learning. Bees were exposed to MF for 1 min, and then immediately proboscis extension response (PER) was examined; this was repeated five times. Bees exposed to 100 µT and 1000 µT MF had a significantly lower level of response compared to the control. Bees exposed to 20 µT MF were not so disrupted. This led to the conclusion that MF can disturb foraging efficiency, which was also evaluated in field experiments. A zone of 100 µT MF exposure was situated between the hive entrance and the feeder in a restricted area. During 15 min of experiment duration, fewer bees flew out from the hive; a decrease in the number of bees returning to the hive was also observed, compared to the non-exposed control, but generally bees that successfully reached the feeder returned to the hive. Therefore, a decrease in forage efficiency as a result of MF influence was displayed [17]. Long-term exposure to MF (17 h) was also demonstrated to have an impact on aversive learning and aggression levels. In the sting expression response (SER) experiment, exposure to 100 µT and 1000 µT MF reduced aversive learning. The aggression level was investigated as a reaction to bees from foreign colonies; bees after 100 µT MF exposure presented much higher aggression scores [35].Honeybees in the environment can be exposed to different stressors at any one time. The results of their coexistence can be difficult to predict. As described above, magnetic field exposure causes changes in honeybee wingbeat frequency and reduced olfactory learning. Interestingly, it has been demonstrated that simultaneous exposure to low-dose neonicotinoids and MF can attenuate this effect [36]. Investigating the impact of EMF in combination with other potentially harmful factors is an interesting issue for further studies.5. ConclusionsExposure to a 1 mT and 1.7 mT magnetic field significantly affected the behavioral pattern of young honeybee workers. It is difficult to denote the direction of the changes, but it was clearly shown that both time of exposure and the magnetic induction of the field makes a difference. The number of behavior bouts increased after all the tested exposures.

animals : an open access journal from mdpi

10.3390/ani13091456

PMC10177123

The digestibility of ingredients in fish diets must be known to use the least-cost feed formulation. However, there is very little published information on lipid (fat) digestibility in fish. With this information, commercial feed producers can choose lipids based on high utilization by the fish, as well as cost. We conducted a feeding trial with catfish fed diets with different lipids: soybean oil, soybean oil with conjugated linoleic acids, catfish offal oil, flaxseed oil, menhaden fish oil and poultry fat. After feeding, fish feces were collected for nutrient analysis and compared to the lipid content of the diets. Lipid and fatty acid digestibility were high overall for all of the lipids tested. However, the digestibility of certain fatty acids was different from overall lipid digestibility. This information can be used to choose the best lipids to meet catfish needs, enhance healthy fats in the fish for human consumers, and produce a cost-effective feed.

Lipid and fatty acid digestibility is presumably high in Channel Catfish, but data is lacking. We determined the lipid and fatty acid digestibility of traditional and alternative dietary lipids in Channel Catfish to inform lipid choice for commercial diets. Six diets contained 4% of different lipids: soybean oil (SBO), soybean oil containing conjugated linoleic acids (CLA-SBO), catfish offal oil (COO), flaxseed oil (FXO), menhaden fish oil (MFO) and poultry fat (PF). Diets were fed to Channel Catfish (150–200 g) maintained at 26.5 °C in each of six 110 L aquaria. Six hours post-prandial, feces were collected for analysis. Total lipid, crude protein and fatty acids of lyophilized feces were analyzed, and apparent digestibility coefficients (ADCs) were calculated. ADCs of lipid, saturated and monounsaturated fatty acids, linoleic acid and protein digestibility were similar among diets. CLA isomers (cis-9, trans-11 (84.1%) and trans-10, cis-12 (90%)) in the CLA-SBO diet were highly digestible. Oleic acid digestibility was highest in the PF diet. ADC was high for α-linolenic acid in the FXO diet, and for arachidonic acid and n-3 LC-PUFA in the MFO diet. Overall, total lipid digestibility was high, but ADCs of individual fatty acids differed by source.

1. IntroductionThe catfish industry is the largest commercial aquaculture industry in the USA, with producer sales of USD 447 in 2022 [1]. Feed costs have increased steadily over time, prompting a search for cheaper alternative ingredients. However, exchanging conventional dietary lipid sources for alternative lipid sources in fish feeds to gain an economic or sustainability advantage might also affect lipid and nutrient digestibility [2]. Lipid and fatty acid digestibility studies allow screening of new ingredients for nutrient bioavailability to the fish [3]. Digestibility in catfish typically does not include lipid sources [4], because it is assumed that lipid and fatty acid digestibility are high. Due to the wide variety of lipid sources used in channel catfish feeds [5,6,7,8], there is a need to understand the digestion and absorption of dietary nutrients when using different dietary lipid sources. This information could assist feed producers in considering different lipid sources for inclusion in commercial Channel Catfish feed formulations.Animal fats are high in saturated fatty acids, while plant oils are rich in polyunsaturated fatty acids (PUFA) and monounsaturated fatty acids (MUFA). The dietary lipid source [9] and fatty acid profile affect the digestibility of dietary lipids [10]. The degree of saturation of fatty acids [11], fatty acid chain length [12] and saturated fatty acid inclusion level in the diet [13] all affect lipid digestibility.Diets for Channel Catfish are formulated assuming that there are no differences in the digestibility of different lipid sources. Lipids must meet the essential fatty acid (EFA) requirement of the fish [4]. Channel Catfish require both n-3 and n-6 PUFA in their diet [14,15]. However, the constituent fatty acids of the dietary lipids also affect the fatty acid composition of the fish, which affects the nutritional value of the fish, as well as organoleptic and storage properties [16,17]. Marine fish oils are not major components of commercial Channel Catfish diets, which results in low levels of n-3 long-chain polyunsaturated fatty acids (n-3 LC PUFA) in catfish fillets. These fatty acids, including eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), are considered beneficial for human health [18]. Marine fish oils were minimized or eliminated from commercial Channel Catfish feeds due to their cost and negative effect on the peroxide values of the feeds. In addition, catfish can synthesize n-3 LC-PUFA from 18-carbon precursors, so preformed n-3 LC PUFAs are not required in the diet. Alternative lipid sources are available that can support fish performance while supplying other healthy fatty acids, such as conjugated linoleic acids (CLAs).Lipids containing CLAs have been used successfully in diets for Channel Catfish [17,19,20,21]. The CLAs have been implicated in mitigating cancer, heart disease, obesity, diabetes and bone pathologies in animal models, as well as in vitro studies [22]. Similar to other dietary fatty acids, the CLAs accumulate in the fish fillet. However, their use does not increase diet cost as much as fish oil, or reduce sensory characteristics of fish compared to catfish fed diets with fish oil [17]. Differences in the digestibility of traditional and novel oils should also be considered in formulating commercial catfish feeds. In this study, the digestibility of CLA-SBO was compared to other practical lipids to determine the nutrient bioavailability of the lipid sources. The null hypothesis is that there would be no differences in the digestibility of lipids or fatty acids among lipid sources.2. Materials and Methods2.1. Diet PreparationA basal diet (sinking pellet) was formulated to meet the nutrient requirements of Channel Catfish [4] (Table 1). The diets were prepared by Dr. Kurt Rosentrater at Iowa State University (Ames, Iowa). The proximate and fatty acid composition of the diets are shown in Table 2 and Table 3, respectively. Yttrium oxide (Yt, 0.5%) was included as the inert marker. The lipid sources tested included some that are commonly used in Channel Catfish diets as well as some novel sources. The tested lipids and their distinctive fatty acids were: soybean oil (SBO; 18:2n-6), soybean oil enhanced with conjugated linoleic acids (CLA-SBO; CLA isomers and 18:2n-6), catfish offal oil (COO; 18:1n-9), flaxseed oil (FXO; 18:3n-3), menhaden fish oil (MFO; n-3 LC-PUFAs) and poultry fat (PF; 18:1n-9). Collectively, the lipid sources contrasted in their fatty acid makeup. Ethoxyquin (0.0125%) was added to the supplemental lipid in all diets to limit lipid peroxidation.All ingredients were milled with a hammer mill (Glen Mills, Clifton, NJ) and thoroughly homogenized in a mixer (Kobalt, Greenfield Center, NY). Each blend was adjusted to a desired pre-extrusion moisture content of ~45% by adding appropriate quantities of water, then mixed again for 15 min to ensure complete moisture distribution. The extrusion processing of each blend was performed using a single screw commercial extruder (InstaPro, Model 600, Des Moines, Iowa), which had a screw compression ratio of 1:1. The die assembly was conical and contained 88 circular 3-mm die holes. The screw speed operated at 600 rpm during extrusion. After extrusion, the pelletized feed blends were dried in a laboratory oven (Thelco Precision, Jovan, Winchester, Virginia) at 50 °C for 24 h. After drying, the diets were broken and sieved into proper pellet size, then stored at −15 °C.2.2. Fecal Collection System, Feeding and Fecal CollectionStocker catfish from the UAPB aquaculture research station were collected and moved to the nutrition wet lab for the feeding trial. They were maintained in 240 L aquaria and fed the control diet for 1 week prior to sorting by size and stocking into experimental tanks. Catfish weighing about 150–200 g individually were stocked in each of six 110 L tanks modified for fecal collection at a rate of 10 fish/tank (Figure 1). A plexiglass board was secured diagonally in the tank using silicone, and fish were held in the upper portion. Smaller plexiglass boards were affixed at the water outlet to direct any feces produced into the drain water. Feces that flowed out with the drain water sunk into the fecal collection chamber.Using a series of two taps, the feces were collected in a plastic cup. The tanks were covered with black sheeting to prevent startling the fish. Each 110 L tank was supplied with dechlorinated municipal tap water. Water flow rate was maintained at 1.1 L/min. Individual air stones aerated each tank. Water temperature was maintained around 26.5 ± 0.3 °C (mean ± SE), which is close to optimum for catfish feed intake [24].The six diets were assigned randomly to each tank. Fish were fed their respective diets to satiation at 8 am each day for 7 days prior to beginning the fecal collections. Feeding and fecal collection for each diet were conducted in triplicate over time in the following manner. After feeding the fish their respective diets, any uneaten food was drained through the collection chamber prior to fecal collection. Feces production began after 5–6 h. Feces were collected from the fecal collection chamber every 1.5–2.0 h and pooled by tank. All feces were collected and stored in a single container (one per tank).The collected fecal material was lyophilized using a Labconco Freezone 4.5 freeze-dry system (Labconco Corporation, Kansas City, Missouri) and stored at −20 °C until analysis. After sufficient feces were collected for analysis in the first collection period, diet-tank assignments were re-randomized such that each tank received a different diet for 7 days prior to the second collection period. Fecal collection commenced thereafter, as previously described.After the second collection period, diets were randomly reassigned as previously described for a third fecal collection period. When incidences of aggression occurred during the study [25], as indicated by external bite marks appearing on the fish, the entire culture system was treated with 2 gm/L sodium chloride to prevent infection. Any severely wounded or dead fish were replaced. After the introduction of a new fish, feeding continued with the respective diets for 2–3 days to allow new fish to acclimate to the new diet. During this period, the feces were discarded to avoid contamination because the newly introduced fish may have consumed other food sources.2.3. Proximate and Fatty Acid Analysis of Diets and FecesThe diets and feces were homogenized prior to analyses. Dry matter and ash were determined according to established procedures [26]. The crude protein in the diets and feces were determined using a Leco Truspec N™ nitrogen analyzer (Leco Inc., St Joseph, MI, USA) at the USDA/ARS–Harry K. Dupree Stuttgart National Aquaculture Research Center, Stuttgart, Arkansas. The results were expressed as g/100 g of total crude protein. Total lipids were extracted from samples and weighed to determine the percentage of lipid [27]. The proximate composition of the finished diets is shown in Table 2.Lipid extracts were then methylated [28] to produce fatty acid methyl esters (FAME) for fatty acid analysis. The FAME were analyzed with a gas chromatograph equipped with a fused silica capillary column (15 m × 0.25 mm ID; Varian CP select for Fame #CP8510), and a flame ionization detector (Varian, Model CP-3800 fitted with a CP-8200 autosampler, Walnut Creek, CA) using helium as the carrier gas. Injection volume was 1 µL, with an injector and detector temperature of 250 °C and 315 °C, respectively. The column temperature was held initially at 100 °C for 10 min, increased to 160 °C at a rate of 15 °C/min and held for 4 min, then increased to 250 °C at a rate of 2.5 °C/min. Each sample had a total analysis time of 81 min. The FAME in diets and feces were identified and quantified by comparing the retention time and peak area to those of serially diluted mixtures of reference standards (trans-10, cis-12 CLA; cis-9, trans-11 CLA, Matreya LLC, Pleasant Gap, PA; GLC-473b, Nu-Check Prep, Elysian, MN). The results of the individual fatty acids were expressed as g/100 g of total identified FAMEs. The fatty acid composition of the finished diets is shown in Table 3.2.4. Analysis of Inert Marker (Yttrium) and Calculation of Apparent Digestibility Coefficients (ADC)Concentration of the inert marker, yttrium (Yt), in the diets and the feces was determined using inductively coupled plasma emission spectrometry (ICP-OES; WBA Analytical Laboratories, Springdale, Arkansas). Apparent digestibility coefficients (ADC) of nutrients in the test diets were calculated [4] as follows:ADC (%) = 100 − 100 [(% Yt in feed)/(% Yt in feces) × (% nutrient in feces/% nutrient in feed)]. Negative digestibility coefficients were adjusted to zero [29].3. Data AnalysisThe ADC of crude protein, lipid and fatty acids in the test diets were analyzed by one-way ANOVA using the SAS® 9.2 software program PROC GLM (SAS Institute Inc., Cary, NC, USA) to test for differences among dietary treatments. Data were analyzed with tank as the experimental unit by averaging across each tank. Fisher’s least significance difference test [30] was used to identify specific differences among treatment means with significance declared at p ≤ 0.05.4. Results4.1. Fatty Acid Composition of FecesSaturates (SFA) were higher in feces of fish fed PF or MFO diets compared to the other diets (Table 4). Feces of fish fed the COO diet had a higher concentration of monounsaturates due to elevated levels of 18:1n-9 (24.4%) in COO. Fecal content of 18:2n-6 was consistently high across diets, and it was higher in the CLA-SBO and SBO diets than in other diets.Feces of fish fed the FXO diet contained the highest concentration of 18:3n-3. The highest concentration of LC-PUFAs, such as 20:5n-3 and 22:6n-3, were found in feces of fish fed the MFO diet. These changes led to a decrease of the n-3/n-6 ratio in feces of fish fed the FXO or MFO diets (1.1 to 0.3 and 0.9 to 0.2, respectively).4.2. Apparent Digestibility Coefficients for Crude Protein, Lipid and Fatty AcidsCrude protein digestibility of the diets ranged from 91.6–92.9%, and there were no differences among diets (p = 0.748). The digestibility of total lipids (82.8–88.4%) was also not different among diets (Table 5). However, individual fatty acid digestibilities were affected by dietary lipid source. The digestibility of 14:0 was higher in fish fed the MFO diet (82.3%) than in the PF and COO diets (55.2 and 53.9%, respectively); however, the digestibility of 16:0 and 18:0 were not different among the diets. The digestibility of 16:1n-7 was higher in the MFO (85.7%) and PF (85.7%) diets than in the COO diet (68.5%). The digestibility of 18:1n-9 was lower in fish fed the MFO diets than all other treatments. The CLA isomers cis-9, trans-11 (84.1%) and trans-10, cis-12 (90%) were efficiently digested in the CLA-SBO diet, while the digestibility of 18:2n-6 was not different among the diets. The highest digestibility coefficient for 18:3n-3 was observed in fish fed the FXO dietary treatment (90.3%), which had the highest concentration of 18:3n-3. High digestibility coefficients (≥58.5%) were observed for the LC-PUFAs (20:4n-6, 20:5n-3 and 22:6n-3) that were mainly present in the MFO diet (94.2, 93.1, 90.6%, respectively). Arachidonic acid (20:4n-6) was digested more efficiently in fish fed the MFO diet than the COO diet. A dash (no digestibility coefficient shown) signifies that the computation of digestibility coefficients was not possible because these fatty acids were detected in the initial diets and not detected in the feces, or vice versa. No coefficients could be generated for the n-3 LC-PUFA in fish fed diets other than MFO, so statistical comparisons were not possible.5. DiscussionFatty acid composition of the diet is the main determinant of lipid digestibility in fish, though lipid class composition is also a factor [31]. The CLA-enhanced soybean oil and the other dietary lipid sources did not affect lipid or protein digestibility in Channel Catfish, which also occurred in Australian Shortfin Eel [32] and Red Hybrid Tilapia [33]. However, lipid digestibility was higher in Atlantic Halibut-fed vegetable oils (93–95%) compared to those fed animal lipids (90%) [9], which was attributed to the high concentration of SFA in the animal lipids. Lipid digestibility decreased with increasing PUFA to LC-PUFA ratio in European Sea Bass [34]. The degree of unsaturation, carbon chain length and the concentration of fatty acid in the lipid can all influence fatty acid digestibility [9,33,35,36]. Saturated fatty acids have a higher melting point than unsaturated fatty acids of the same chain length, resulting in lower digestibility of SFA in fish [37]. Endogenous fatty acid secretions in catfish in this study were observed for fatty acids present in low dietary concentrations. For example, 14:0 was not present in the CLA-SBO diet, and dietary concentrations for the SBO and FXO diets were 0.1 and 0.13%, respectively; whereas fecal concentrations of 14:0 from fish fed those diets were much higher (0.6–0.9%). The current data might suggest that endogenous fatty acids were secreted to compensate for fatty acids that were present in low dietary concentrations. Alternatively, the digestibility of certain fatty acid groups may be underestimated due to the presence of endogenously produced fatty acids, as previously suggested in Rainbow Trout [37]. Endogenously produced fatty acids can affect the computation of digestibility coefficients of fatty acids, leading to very low or negative computed coefficients. This can occur when fatty acids are present at low dietary concentrations and have low digestibility [4].In this study, the digestibility of SFA and other fatty acids was not dependent on the concentration of dietary SFA. The digestibility of SFA was below 61% and lower than MUFA, which had a digestibility of between 64 to 78%, while the digestibility for n-3 and n-6 fatty acids was between 53 to 91% and 55 to 70%, respectively. These results are contrary to observations in other fish species. For example, the digestibility of SFA in Rainbow Trout-fed diets formulated from a combination of capelin or anchovy oil with soybean, rapeseed, palm or olive oil decreased with increasing dietary concentration of SFAs [38]. The digestibility of MUFA, n-3 and n-6 fatty acids also decreased with increasing SFA, but to a lesser extent. High dietary concentrations of SFA also reduced the digestibility of SFA in Atlantic Salmon [13]. However, SFA had little effect on the digestibility of SFA and other fatty acid groups in Atlantic Salmon fed diets with soybean oil with a low concentration of SFA [39].Interestingly, pigs have a similar gut morphology to Channel Catfish [4,40]. Jørgensen, et al., [41] fed pigs (35 kg) a basal diet or diets containing 15% of either fish oil, rapeseed oil or coconut oil. All the diets contained 22% fish meal as a protein source. The higher ileal digestibility of 18:1 and 18:3n-3 was related to the higher dietary concentration of these fatty acids in diets with fish oil and rapeseed oil, as well as higher 18:2n-6 in the rapeseed oil diets. Moreover, the digestibility of n-3 LC-PUFAs was also high in all the swine diets. The negative digestibility coefficient of 18:0 in the swine trial was attributed to the low dietary concentration, low intake and endogenous gut secretions. Similarly, in growing pigs (40 kg) fed diets containing 10% of either tallow, high oleic sunflower oil, sunflower oil, linseed oil or a fat blend of tallow (5.5%), sunflower oil (3.5%) and 1% linseed oil [42], the ileal digestibility of 18:3n-3 in the linseed oil diet and 18:2n-6 in the sunflower oil diet was higher in the diets in which they were present in the highest concentrations. However, the digestibilities of 16:0 and 18:0 were lower in pigs fed the tallow diet, even though the dietary concentration of these fatty acids was higher in this diet. The lower digestibility of 18:0 was determined to be due to the use of dietary animal fats in pigs [43].In the present study, the digestibility of n-3 fatty acids was positively correlated with dietary concentration. The 18:3n-3 and n-3 LC-PUFA were well absorbed from the FXO and MFO diets, respectively, in which they were found in higher proportions. The digestibility of 18:3n-3 was highest in fish fed the FXO diet (90.3%), and digestibility coefficients of 20:5n-3, 22:5n-3 and 22:6n-3 were high (90–100%) in fish fed the MFO diet. Higher digestibility of 18:3n-3 and n-3 LC-PUFA also occurred in Atlantic Halibut with the inclusion of flaxseed oil and fish oil, respectively, in the diets [9]. In cod, lipase specificity was also high for 20:5n-3, 20:4n-6 and 18:2n-6 compared to 22-carbon fatty acids, SFA and MUFA (except 18:1n-9) [44]. Lipolysis is a rate-limiting step in fatty acid digestion and the rate of release from triacylglycerol is higher for PUFA [45] than for 18:0 and MUFA. Lipolytic activity in juvenile turbot indicated preferential absorption of PUFA over MUFA, followed by SFA as digesta moved from the stomach to the rectum [46].We observed some interesting trends in fecal fatty acids that suggested differences in metabolism mediated by key dietary fatty acids. For instance, there were no n-3 LC PUFA in either the SBO or CLA-SBO diets, yet both 20:5n-3 and 22:6n-3 appeared in the feces of fish fed the CLA-SBO diet only. Increased 20:5n-3 and 22:6n-3 has been observed in tissues of other freshwater fish fed diets with CLA [47,48,49], indicating increased delta-6 desaturase activity. The COO diet also lacked n-3 LC PUFA, yet feces of catfish fed that diet contained 22:6n-3. The low level of 18:3n-3 present in the diet (3%) appeared to stimulate delta-6 desaturase activity and production of n-3 LC PUFA. In contrast, fish fed the FXO diet (with no LC PUFA but about 34% 18:3n-3) produced feces without LC-PUFA. We did not measure desaturase activity in our study. However, delta-6 desaturase is the rate-limiting enzyme in LC-PUFA synthesis, and the balance between n-3 and n-6 fatty acids affects the amounts and types of LC-PUFA produced. Catfish can convert 18:3n-3 into n-3 LC-PUFA, and fish fed the FXO diet had the highest ADC for 18:3n-3. The lack of LC-PUFA in feces of those fish might be explained by reduced desaturase activity. These speculations can be addressed with more detailed studies of catfish lipid metabolism that include measurements of desaturase activity.It is difficult to reconcile differences in the results of lipid digestibility studies of different fish species, because they cover a wide range of trophic levels (herbivorous, omnivorous and carnivorous). Experimental and commercial diets for fish range widely in nutrient composition as a result. Diets for omnivorous Channel Catfish rarely contain more than 10% total lipid, whereas those for carnivorous species can exceed 25% [4]. The EFA requirements also vary considerably—omnivorous fish such as catfish can synthesize LC-PUFA from PUFA, whereas carnivorous fish generally require preformed LC-PUFA in the diet.Fish behavior can also affect feed utilization. For instance, stress can reduce feed intake, though specific effects on nutrient digestibility are not well documented. The aggressive behavior noted in catfish in this study sometimes reduced feed intake. However, fish that were not actively feeding were replaced, and no differences in behavior were observed among diets. Aggression in catfish held in aquaria has been noted previously [50]. There is no specific stocking density that will prevent aggressive behavior in catfish; it tends to escalate with fish size, and larger fish are preferred for digestibility trials to ensure sufficient production of feces.Research demonstrating the benefits of plant oils in catfish diets is now substantial, which is likely to result in increased inclusion in commercial diets. The CLA isomers in CLA-Soy oil were well-digested in catfish in this study. Previous results showed that CLA-Soy oil also promotes overall catfish performance, does not compromise the sensory characteristics of the fillet and is beneficial for human health. The combined results indicate the unique potential of CLA-Soy oil to contribute to the sustainability of Channel Catfish production.6. ConclusionsThe efficient absorption of 18:3n-3 and n-3 LC-PUFAs from the FXO and MFO diets, respectively, is in agreement with the n-3 EFA requirements of Channel Catfish (Satoh, et al., 1989). Substitution of fish oils with plant oils in aquatic animal diets is likely to increase due to concerns over environmentally sustainable production practices, including diet composition. The CLA-soybean oil is a novel plant lipid with good potential to enhance healthy fatty acids in Channel Catfish because of its high digestibility compared to other lipid sources. In particular, the high digestibility coefficient of CLA isomers cis-9, trans-11 (84.1%) and trans-10, cis-12 (90%) confirms that they were efficiently absorbed from the CLA-SBO diet.

animals : an open access journal from mdpi

10.3390/ani13081405

PMC10135048

Any abnormalities in the physical properties of an egg can decrease the hatching rate. The embryos of oviparous reptiles obtain Ca from the eggshell. This motivated us to analyze the microstructure and Ca in the eggshells of Chinese alligators. We found that the shells of the eggs with high hatching rates were thicker than those of the eggs with low hatching rates. There were also fewer erosion-crater pores on the surfaces of the eggs with high hatching rates than on the surfaces of the eggs with low hatching rates. Moreover, the shells’ Ca contents were significantly higher in the eggs with high hatching rates than in the eggs with low hatching rates. Cluster modeling indicated that the highest hatching rate occurred when the eggshell thickness was 200–380 µm and there were 1–12 pores. These results suggest that eggs with adequate Ca contents, thicker shells, and less air permeability are more likely to hatch. Furthermore, our findings can inform future studies.

The Chinese alligator (Alligator sinensis), found only in a small region in southeastern Anhui Province, is listed as critically endangered (CR) by the International Union for Conservation of Nature (IUCN) due to its current declining population trend. Any abnormalities in the physical properties of an egg can decrease the hatching rate. In particular, eggshells play an essential role in embryo development, motivating us to analyze the microstructures of the eggshells of Chinese alligators. In this study, we categorized the eggshells into two groups, based on the hatching rates, and analyzed the relationship between the eggshell parameters (eggshell thickness, calcium content, and number of pores in erosion craters) and the hatching rate, as well as the relationships between the eggshell parameters. We found that the shells of the eggs with high hatching rates were thicker than those of the eggs with low hatching rates. There were also fewer erosion-crater pores on the surfaces of the eggs with high hatching rates than on the surfaces of the eggs with low hatching rates. Moreover, the shell Ca content was significantly higher in the eggs with high hatching rates than in the eggs with low hatching rates. Cluster modeling indicated that the highest hatching rate occurred when the eggshell thickness was 200–380 µm and there were 1–12 pores. These results suggest that eggs with adequate Ca contents, thicker shells, and less air permeability are more likely to hatch. Furthermore, our findings can inform future studies, which will be vital for the survival of the critically endangered Chinese alligator species.

1. IntroductionReproductive success is critical to species survival [1]. In oviparous species, the physical properties of eggs play an essential role in embryo development and successful hatching [2]. Any abnormalities in an egg’s physical characteristics can lead to a decrease in the likelihood of its hatching [1]. Eggshell parameters (e.g., porosity and number of pores) are essential factors influencing the successful hatching of eggs [3,4,5]. The eggshell represents a trade-off between two important functions during embryo development: it must be thick and strong enough to protect the embryo from external damage but not so thick that it is an obstacle to hatching. Furthermore, the eggshell must have sufficient pores to provide oxygen to the developing embryo, but a large number of pores allows pathogenic microorganisms to invade the egg [3].As mentioned above, the eggshell thickness significantly affects the hatching rate [3]. The thickness of shells substantially differs among hatched, unhatched, and unfertilized eggs [6]. Specifically, eggs that fail to hatch have significantly thinner eggshells than those that successfully hatch. Moreover, the hatching rate of thin-shelled eggs is 3% to 9% lower than that of thick-shelled eggs. Most researchers have reported that thick eggshells provide the following advantages: (1) the full use of nutrients by the embryo (in chicken eggs) [7], (2) a reduced likelihood of bacteria entering the egg [8], (3) a reduced likelihood of egg dehydration [9], and (4) better protection against mechanical damage [10].The eggshell buffers against changes in the external environment by allowing the exchange of water and gases essential to embryonic development [11,12]. Most squamates lay eggs with leathery shells with high breathability [13]; indeed, the fiber layers of snake eggs, which have leathery shells, is reduced during incubation to increase breathability for the embryo [14]. In contrast, most turtles and all crocodiles lay hard-shelled eggs with very low breathability. To compensate for this low breathability, the eggs harden and become opaque during the second half of incubation. During this period, the embryos rapidly develop, and oxygen consumption and carbon-dioxide production increase exponentially [15,16]. Once hardened (i.e., when they become opaque), the shells of turtles and crocodiles have lower water contents. Removing water from porous eggshells significantly improves their permeability to oxygen and carbon-dioxide gas, thereby promoting the diffusion and exchange of oxygen and carbon dioxide [17]. Conversely, hypoxia reduces the heart rate of American alligators [18] and some birds [19], and it can lead to a decline in both embryo growth and survival in reptiles (turtles and crocodiles), as well as birds [20,21].The maternal consumption of micronutrients, macronutrients, and fatty acids may influence the hatching rate of alligator eggs [22]. Inadequate provision or the pathological metabolism of calcium (Ca) are common in captive reptiles [23]. Crocodile hatchlings are prone to metabolic bone disease caused by inadequate Ca, of which the main symptoms are kyphoscoliosis and jaw softening (“rubber jaw”), accompanied by changes in tooth structure (“glassy teeth”) [24]; alligator hatchlings fed a Ca-deficient diet are also prone to long-bone and spinal compression fractures [25]. Importantly, the embryos of birds and oviparous reptiles obtain Ca from the eggshell [26].In this study, we categorized eggshells into two groups (low hatching rate vs. high hatching rate) based on their hatching rate and analyzed the relationship between the eggshell parameters (eggshell thickness, calcium content, and number of pores in erosion craters) and the hatching rate, as well as the relationships between the eggshell parameters. We expected that our results would provide important insights into the factors determining the hatching of alligator eggs, as well as the supplemental feeding of female alligators, and thus improve the hatching rate of Chinese alligator eggs.2. Materials and Methods2.1. MaterialsThe eggshells of Chinese alligators were collected at the Xuancheng Alligator Breeding Research Center. Eight alligator nests with significantly different hatching rates were selected, for a total of eighty eggs (n = 10 eggs/nest). The eggs were divided into two groups, with a hatching rate of 50% as the dividing line. Eggs with a hatching rate higher than 50% were assigned to the high-hatching-rate group, and those with a hatching rate lower than 50% were assigned to the low-hatching-rate group (high hatching rate = an average of 92.5% of eggs in the nests hatched, low hatching rate = an average of 39.1% of eggs in the nests hatched [27]). For sampling, three pieces of shell (approximately 1 cm2 each) were cut, including samples from the upper surfaces of the middle and both ends of each egg: the end where the young alligator naturally emerged from the shell was referred to as the cracked end (CE), the opposite end as the intact end (IE), and the middle as the middle section (ME). The eggshell membrane was removed, and pieces of shell were rinsed in distilled water and dehydrated in a graded series of alcohol. The outer surfaces of the pieces of shell were then sputter-coated with gold palladium (KYKY gold sprayer) and viewed at 20 kV under a scanning-electron microscope (SEM). Starting at an edge of each piece of shell at a magnification of 37×, open pores were counted on four consecutive SEM display screens along the middle of the piece of shell. Next, the pieces were oriented end-on, and micrographs (×100) were taken. Data from the three samples were pooled [28].The average number of pores in the CE, ME, and IE regions was counted. When counting stomata on the eggshell surface, first, the stomatal numbers N1, N2, and N3 of the three microscopic images were obtained, and then the average stomatal number (N) was calculated.We used MATLAB to measure the thickness of the cross-section of each eggshell and to calculate the mean. Using MATLAB (MathWorks Inc., USA), a GUI tool for measuring eggshell thickness from microscopic images of eggshell cross-sections was prepared. During the measurement, the length L0 of a 100-μm ruler in the image was measured by drawing a straight line on the microscopic image, and then the image widths L1, L2, and L3 at three different positions of the cross-section were measured by drawing a straight line. Subsequently, the average width (L) of the cross-section was calculated by the following formula. L = (L1 + L2 + L3) / L0 × 100 / 3 (μm)An appropriate amount of sample was weighed and placed into polytetrafluoroethylene liner of a high-pressure digestion tank. After addition of HNO, the sample was heated in a drying oven at 150 °C for 2 h. After cooling, the sample was transferred to a 25-mL volumetric flask and diluted to volume with ultrapure water as the test solution for metallic elements. A total of 25 mL of sample solution was accurately taken from the volumetric flask, placed into a 25-mL volumetric flask, diluted with ultrapure water to scale as a sample solution for testing metal elements, and directly tested for the content of each element by an AVIO200 (PerkinElmer, Shanghai, China) inductively coupled plasma emission spectrometer [29].2.2. Data AnalysisMicrosoft Excel was used to preprocess all data. First, the Kolmogorov–Smirnov test was performed to determine eggshell thickness and the number of pores in erosion craters. Next, a t test was performed to compare eggshell thicknesses between the groups, and the Mann–Whitney U test was used to compare the number of erosion-crater pores. Linear fitting was performed to determine the relationship between eggshell thickness and the number of erosion-crater pores. In addition, a Gaussian mixture model was used to cluster eggshell thicknesses and the number of pores in the CE, ME, and IE regions.3. Results3.1. Scanning Results for the Eggshells’ Outer SurfacesThe scanning-electron microscopy revealed that the outer surfaces of the Chinese alligator eggshells were covered with erosion craters and pores. A typical erosion crater spirals inwards and downwards, with a depression in the center. The erosion craters and pores on the eggshell surfaces were randomly distributed; while each pore was in the center of an erosion crater, not all the erosion craters contained pores (Figure 1a).3.2. Effect of Alligator-Eggshell Thickness on the Hatching RateThe eggshell thickness followed a normal distribution (p > 0.05; Supplementary Table S1). A t test was performed to compare the eggshell thicknesses between the two groups. The results showed that the eggshell thickness significantly differed between the high- and low-hatching-rate groups (p < 0.05; Supplementary Figure S1, Table 1, Figure 1c,d).The shells of the eggs with high hatching rates were significantly thicker than those of the eggs with low hatching rates (Figure 2a,c). The eggshell thickness was also positively correlated with the hatching rate (R = 0.912; Figure 3a).Finally, we clustered the shell thicknesses of the CE, ME, and IE areas of the eggshells with a Gaussian mixture model. We found that eggshell thicknesses of 250–380 µm resulted in the highest hatching rates (Figure 4a,b, Supplementary Figures S2 and S3).3.3. Effects of the Number of Erosion-Crater Pores on the Surfaces of Alligator Eggs on Their Hatching RatesThe distribution of the number of erosion-crater pores on the surfaces of the alligator eggs was unknown (p < 0.05; Supplementary Table S1). The numbers of pores in the CE, ME, and IE regions were compared between the two hatching-rate groups using a Mann–Whitney U test. There were significant differences between the numbers of erosion-crater pores on the surfaces of the eggshells with different hatching rates (p < 0.05; Table 2, Figure 1a,b).The results show progressive significance; the significance threshold was 0.05 and, thus, values lower than 0.05 indicated significance.Specifically, the alligator eggs with high hatching rates had fewer erosion-crater pores on their surfaces in the CE, ME, and IE regions than the eggs with low hatching rates (Figure 2b,d). The number of erosion-crater pores was also negatively correlated with the hatching rate (R = −0.632; Figure 3b). Next, we compared the numbers of erosion-crater pores per unit area between the two hatching-rate groups. There was a significant difference (p < 0.01) between the numbers of pores per unit area on the surfaces of the eggshells with different hatching rates. Specifically, the eggs with high hatching rates had fewer pores per unit area on their surfaces than the eggs with low hatching rates (Figure 2b,d, Supplementary Figure S4). Next, a Gaussian mixture model was used to cluster the number of erosion-crater pores per unit area; we found that the presence of 1–12 erosion-crater pores on the surfaces of the eggshells resulted in the highest hatching rates (Figure 4a,b, Supplementary Figures S2 and S3). The number of erosion-crater pores was also negatively correlated with the eggshell thickness (R = −0.839; Figure 3c).3.4. Eggshells’ Ca Contents Are Associated with the Hatching RateThe variance in the Ca content in the two groups of samples with different hatching rates was homogeneous (p > 0.05). The Ca contents in the eggshells of the Chinese alligators were significantly correlated with the hatching rates (p < 0.05) and significantly differed between the hatching-rate groups (p = 0.025) (Table 3); specifically, the high-hatching-rate group had the shells with the highest Ca contents (Supplementary Figure S5).4. Discussion4.1. Effect of Eggshell Thickness on the Hatching RateEggshells that are excessively thin are at an increased risk of fracture, water evaporation, and bacterial penetration, which affect embryo development. However, eggshells that are overly thick have lower water contents during incubation and make it difficult for embryos to break during hatching, leading to embryo death. We found that eggshell thickness significantly affects the hatching rate of fertilized Chinese alligator eggs. The eggs with high hatching rates had significantly thicker shells than those with low hatching rates (p < 0.05). In addition, we found that eggshell thicknesses of 200–380 µm resulted in the highest hatching rates. Previous studies showed that the hatching rates of eggs with thick eggshells are significantly higher than those of eggs with thin eggshells, and that increased eggshell thickness increases the hatching rate of eggs, such as those of turkeys and geese [30,31]. Eggs with thin shells are more susceptible to bacterial infections and excessive dehydration [3]. Therefore, the eggshell thickness significantly affects the hatching rate. The eggshell thickness during incubation may vary between individuals [1]. Studies have shown that eggshell thickness decreases as females age; thus, eggshell thickness may indicate the age of the breeding individual, but this hypothesis needs to be further confirmed by future studies [1]. The eggshell thickness in Chinese alligators may similarly indicate the age of the female, but this hypothesis also needs further confirmation.Eggshells vary in thickness across their surface. In the Chinese alligator, we found that ME is the thickest area of the eggshell, followed by CE, and that IE is the thinnest. Studies of eggshell thickness typically average the measurements of intact eggs at three locations: the eggshell tip, the middle circumference, and the blunt end [1]. However, in the present study, all the eggs had already hatched; thus, the end from which the alligator hatched was defined as the CE. The eggshell thickness of the CE was significantly greater than that of the IE. Moreover, maternal Ca is transferred directly to the shell during shell formation; in the eggshell, Ca is found in the form of calcium carbonate [12].Eggshell thickness and egg weight are strongly correlated [32]. Eggshell thickness is difficult to measure; thus, future research would benefit from the determination of whether eggshell thickness can be estimated from egg weight. Moreover, future studies should determine the impact of egg weight on the hatching rates of Chinese alligators.4.2. Effect of the Number of Erosion-Crater Pores on the Hatching RateAs incubation progresses, the dense calcification layer of alligator eggs exhibits a gradual dissolution of crystals, producing concentric, stepped erosion craters. At the bottom of these erosion craters, the underlying honeycombed layer, which contains many gaps interconnected with other parts of the shell, is exposed [11]. The erosion is due to the byproducts of acid metabolism from bacteria in the nest material. This external degradation gradually increases the number of eggshell pores, reducing the integrity of the eggshell [11]. In the present study, the eggs with high hatching rates had significantly fewer erosion-crater pores on their surfaces than the eggs with low hatching rates. Thus, the eggs with low hatching rates probably experienced greater shell degradation and greater shell-weight loss, as well as exhibiting significantly more pores. Increases in the number of pores result in significantly greater gas and water exchange (leading to water loss) between the embryo and the external environment. Excessive water loss causes the embryo to become dehydrated, resulting in malformation; severe cases of dehydration result in embryo death, decreasing the hatching rate. Therefore, excessive pores on the eggshell surface are unfavorable for embryo development and the hatching of Chinese alligator eggs. Similarly, eggshell thickness and the number of pores significantly affect the development of crocodile embryos [33].The number of pores is one of the most important characteristics of the shell structure. Nevertheless, both smaller and larger numbers of pores have negative impacts on embryo survival. The presence of excessively small numbers of pores leads to impaired oxygen exchange and, thus, increases embryo mortality [10], whereas overly high numbers of pores result in excessive gas and water exchange (and, thus, excessive water loss) between the embryo and the external environment, with severe negative impacts due to dehydration [9].The eggshell cuticle is an uneven organic layer that covers the outer surface of the eggshell. The cuticle consists of an inner calcified layer and an outer noncalcified water-insoluble layer directly deposited on the calcified layer of the eggshell [34]. The deposition of the cuticle is highly important for preventing the invasion of microorganisms. Bacterial penetration is dependent on natural variation in the cuticle produced during deposition. Eggs with thick cuticles are not easily penetrated by bacteria [35], whereas the thinning of the eggshell cuticle causes embryo death (and, thus, a decreased hatching rate) due to bacterial penetration of the eggshell. Additionally, the cuticle serves as a barrier to prevent water loss and excessive dehydration of the egg material. Studies have shown that dehydration leads to a decline in the embryo’s innate immune function; moreover, the impact of dehydration on immune function may be greater than that on energy supply [36].If the external humidity during incubation is excessively low, excessive water loss may occur, leading to the dehydration and death of the embryo. Gas exchange is also key to successful embryo development. During the gas-exchange process, external hypoxic conditions significantly affect the embryo’s yolk utilization and the size of the offspring [10]. Oxygen diffuses from the air into the embryo and is necessary for aerobic metabolism, while water vapor and carbon dioxide diffuse from the embryo into the air [37]. Gaseous oxygen must pass through the eggshell and the two shell membranes [38]. After gas exchange occurs, oxygen is transported to the tissues through the blood vessels and cardiovascular system [39]. If the hard eggshell surface is not corroded by bacterial byproducts in the nest material, the surface displays few pores, thus increasing the difficulty of meeting the gas-exchange requirements of the embryo. However, eggshells that are overly corroded and thinned by bacterial byproducts in the nest material may experience excessive water loss. Therefore, for alligator eggs with hard shells, moderate numbers of surface erosion-crater pores are conducive to gas exchange and the retention of water. However, an excessive number of erosion-crater pores thins the eggshell and leads to severe dehydration, which hampers the later growth and development of the embryo.4.3. Effect of Eggshells’ Ca Content on the Hatching RateDetailed data on Ca metabolism in avian embryos are available, but there are few data regarding Ca metabolism in crocodilian embryos. The shell contributes substantially to embryonic Ca via transport across the chorioallantoic membrane in Crocodylus novaeguineae [40]. Clinical Ca deficiency has been reported in crocodiles fed red meat without supplementary Ca, as well as in animals housed under cover (which are thus unable to synthesize vitamin D3). Additionally, maternal Ca deficiency may translate into reduced Ca deposition in the eggshell, as suggested by Lance et al. (1983); however, no such effect has been formally or even anecdotally reported in crocodilians. In Alligator mississippiensis, most of the Ca utilized by the embryo originates in the shell, but the yolk may also serve as a temporary store of shell-derived Ca [41]. The present study showed that the Ca content of alligator eggshells has a significant effect on the hatching rate because the alligator eggs with high hatching rates had higher eggshell Ca contents than those with low hatching rates. The Ca in the eggshell is transported across the chorion to supply the embryo. Insufficient dietary Ca intake by the mother leads to a reduction in eggshell Ca deposition in crocodiles. Therefore, a substantial portion of the maternal Ca transferred to the embryo is provided through the eggshell.The results of this study can inform future research. For example, the embryonic absorption of Ca from the shells of newly laid Chinese alligator eggs has not been determined. To assess the effect of maternal diet on eggshell Ca content and subsequent embryo viability, comparisons between captive and healthy wild populations should be conducted to identify shortcomings in the rearing of captive crocodiles.5. ConclusionsThe population status of the Chinese alligator, which is an endangered reptile species, is extremely worrisome. In the process of restoring the population of Chinese alligators, the hatching conditions of alligator eggs constitute an important factor. According to our research, the characteristic properties (thickness and erosion-crater pores) of alligator eggs are related to the egg quality and have a significant impact on the hatching rate. Additionally, the calcium content in the eggshell also has a positive correlation with the hatching rate. This project suggests that focusing on the hatching of Chinese alligator eggs will be an important direction for future conservation efforts.

animals : an open access journal from mdpi

10.3390/ani11102774

PMC8532692

Animal behaviors are critical for survival, which is expressed over a long period of time. The emergence of computer vision and deep learning technologies creates new possibilities for understanding the biological basis of these behaviors and accurately quantifying behaviors, which contributes to attaining high production efficiency and precise management in precision farming. Here, we demonstrate that a dual-stream 3D convolutional neural network with RGB and optical flow video clips as input can be used to classify behavior states of fish schools. The FlowNet2 based on deep learning, combined with a 3D convolutional neural network, was first applied to identify fish behavior. Additionally, the results indicate that the proposed non-invasive recognition method can quickly, accurately, and automatically identify fish behaviors across hundreds of hours of video.

The rapid and precise recognition of fish behavior is critical in perceiving health and welfare by allowing farmers to make informed management decisions on recirculating aquaculture systems while reducing labor. The conventional recognition methods are to obtain movement information by implanting sensors on the skin or in the body of the fish, which can affect the normal behavior and welfare of the fish. We present a novel nondestructive method with spatiotemporal and motion information based on deep learning for real-time recognition of fish schools’ behavior. In this work, a dual-stream 3D convolutional neural network (DSC3D) was proposed for the recognition of five behavior states of fish schools, including feeding, hypoxia, hypothermia, frightening and normal behavior. This DSC3D combines spatiotemporal features and motion features by using FlowNet2 and 3D convolutional neural networks and shows significant results suitable for industrial applications in automatic monitoring of fish behavior, with an average accuracy rate of 95.79%. The model evaluation results on the test dataset further demonstrated that our proposed method could be used as an effective tool for the intelligent perception of fish health status.

1. IntroductionDue to rapid growth in the global population, the demands for aquatic products have increased rapidly for the last two decades [1]. In response to this rapid increase in demand for aquatic products, in many countries, fishery products supply has been dominated by farming fish. However, this industry worldwide is facing huge losses due to a lack of adequate fish health and welfare monitoring and management practices. Animals typically respond to and interact with the environment using a series of rich behaviors. The ability to recognize behavior activities from video data creates a need for tools that automatically quantifies behavior and simultaneously provides a deeper understanding of the relationship between behavior and the external environment [2]. Although significant progress has been made, the challenge of fish behavior recognition is still far from being solved due to the lack of research on the specific posture of fish. Automated recognition and classification of fish behavior will aid in this task, which is considered an efficient method for long-term monitoring and assessment of fish health and welfare [3].One common approach is to manually observe behavior videos over a long period of time, which is labor intensive, and inefficient. Researchers have developed and designed a large number of sensors and algorithms to automatically and accurately identify fish behavior. The biological telemetry method has excellent stability in identifying the behavior of single and fewer fishes, which has been used for animal welfare monitoring and behavior recognition, such as environmental stress behavior and swimming behavior [4]. This integrated micro-sensor has the advantage of simultaneously measuring multiple motion parameters such as movement speed, acceleration, and turning angle. However, the sensor needs to be implanted into the skin or body of the fish, which affects the behavior of the fish and also damages fish welfare. Therefore, hydrophone and sonar technologies based on active acoustic were used for fish monitoring and tracking, behavior monitoring, and biomass estimation in the deep sea [5,6]. The acoustic imaging method has the advantage of capturing high-resolution video even in a muddy and dark underwater environment, but this method relies on expensive acoustic equipment.The emergence of computer vision technology in aquaculture has opened new possibilities to observe fish behavior effectively and efficiently. Traditional computer vision techniques for identifying fish behavior relied on extracting important features such as motion patterns, clustering index, and trajectory. Zhou et al. (2017) precisely quantified the feeding behavior of fish by calculating the clustering index of fish in the near-infrared image with an accuracy of 94.5%, providing an effective method for analyzing the feeding behavior of fish [7]. The video tracking algorithm creates the possibility of automatically identifying fish behavior, and has been widely used to measure the swimming parameters (speed, acceleration, and rotation angle) of fish. A machine vision system with a monocular camera was used to evaluate the swimming performance of fish under different water temperatures [8]. In addition, 3D imaging systems based on stereo vision have been applied to fish detection and tracking [9]. In particular, 3D images can provide more spatial information to overcome the problem of occlusion during movement, and the tracking accuracy rate reached 95%. In addition, the stress level of fish has been assessed by monitoring the abnormal trajectory of the fish [10,11]. This method is helpful for the early diagnosis of fish disease and optimizes management practices in aquaculture.Deep learning methods have revolutionized our ability to automatically analyze videos, which have been used for live fish detection [12], species classification [13], biomass estimation [6], behavior analysis [14], and water quality monitoring [15]. The advantage of deep learning is that it can automatically learn to extract image features and shows excellent performance in recognizing sequential activities. Zhou et al. (2019) proposed an automatic classification method of fish feeding intensity based on a convolutional neural network (CNN) and machine vision [16]. The classification accuracy of this method reached 90% for fish appetite (strong, medium, and weak). In recent years, researchers have discovered that it is essential to combine spatial and temporal features to identify fish behavior. A two-stream recurrent network composed of a spatial network, a 3D convolutional network, and a long short-term memory network (LSTM) was used for salmon behavior recognition [17], which integrates spatial and temporal information, and the accuracy of behavior prediction was 80%. Aiming to monitor the local abnormal behaviors of fish in intensive farming, a method based on an improved motion influence diagram and recurrent neural network (RNN) was proposed [18]. The detection, location, and recognition accuracy rates of this method were 98.91%, 91.67%, and 89.89%, respectively.Although deep learning methods have achieved high accuracy in target detection and tracking, fish behavior recognition methods need to be further improved to adapt to the complex underwater environment and rapid target movement. Fish behavior usually reflects the stimulus of the external environment through changes in swimming behavior. For two different types of behaviors, a video sequence may exhibit a similar activity, which leads to a decrease in recognition accuracy. For example, both feeding behavior and frightening behavior include the rapid movement of fish after being stimulated. In addition, one type of behavior contains a variety of different activity states. For example, fish exhibit the behavior of sinking on the side of the water tank when experiencing hypoxia but also actively gather at the water injection port. Therefore, subtle changes between different behavior types and multiple actions of the same behavior are the main challenges for the application of deep learning in fish behavior recognition.To meet the above challenges, in this paper, a DSC3D model based on temporal and spatial feature information is proposed to recognize the behavior states of the fish school from the collected video data. The main contributions of our work are as follows: (1) a novel method based on deep learning is used to recognize fish activities; (2) proposed framework recognizes the five behaviors of fish through the feature fusion of RGB images and optical flow images; (3) proposed framework shows excellent performance and satisfactory results. Our proposed method provides an effective strategy for automatically identifying fish behavior from video sequences in real time, laying the foundation for intelligent perception in aquaculture.2. Materials and Methods2.1. FishIn this experiment, we used Oplegnathus punctatus as an experimental subjects, which is a farmed fish located in Yantai, Shandong Province, China. Before the experiment, the fish had lived in the recirculating aquaculture system for 6 months. All fish tanks were equipped with automatic bait feeding machines, which were set to feed bait 3 times a day.2.2. Experimental SetupThe experimental system consisted of a recirculating aquaculture system and a computer vision system, as shown in Figure 1. The fish tank was a cylindrical shape with a diameter of 5m and a height of 1m. The cultured water was supplied from the nearby seawater that was processed and filtered by the factory. Water quality parameters including dissolved oxygen (DO) at (5 ± 0.6) mg/L, total ammonia nitrogen (TAN) < 0.5 mg/L, water temperature (15 ± 2) °C, pH(7.2 ± 0.5) can be fine-tuned to obtain the behavior changes of fish as for the machine vision system. A high-definition digital camera (Hikvision DS-2CD2T87E(D)WD-L) with a frame rate of 25 fps (1920 × 1080) was deployed on a tripod with a height of about 2 m to captured the video data, as shown in Figure 1. The deep learning model was implemented by writing code based on the Pytorch framework on the Linux Ubuntu 18.04.3 LTS environment. It was trained with an SGD optimizer on a PC with 4 Nvidia RTX 2080 Ti. 2.3. Data AcquisitionTo better experiment and verify the effectiveness of our proposed method, we collected 24 h of video daily for approximately 3 months, resulting in a collection of behavior videos of different sizes, moments, and numbers of fish. All individuals were captured through the top camera, as the behavior of fish is usually characterized by group activities. The video was stored in the memory card of the network camera and regularly downloaded from the memory card to our computer. Our research was focused on recording the changes in fish behavior at different stages within 24 h. In addition to some common behaviors such as feeding behaviors, other behaviors that interacted with the environment were also collected, such as hypothermia, hypoxia, and startle stimulation of fish.With the help of expert experience and knowledge, these behaviors were classified into feeding behaviors, hypoxic behaviors, hypothermia behaviors, frightening behaviors, and normal behaviors, as shown in Table 1. Our research aims to identify these five behaviors, which provide intelligent monitoring of fish health by identifying the active state of the fish when it is stimulated by food, sound, light, etc. In contrast, the hypoxia and hypothermia behaviors of the fish school were inactive, implying risks that may occur during the breeding process.2.4. Data AnnotationsTo perform our dataset more diverse and representative, we selected as many different types of behaviors as possible from the videos. A trained observer manually annotated five types of actions for 2 weeks. In total, 1066 video clips were annotated. Automated analysis of these video clips was requisite for the full characterization of fish behavior in this limited dataset. As a result, we further manually classified the behavior on the keyframes in order to obtain the real situation, as shown in Figure 2.Due to the large displacement of fish and rapid changes in body posture, accurate understanding of fish behavior on keyframes is the main problem in the process of dataset labeling. We invited aquaculture experts to label our behavioral data. In the end, we completed the extraction and annotation of keyframes for behavioral videos at a frame rate of 4FPS. Since we collected the data on the actual farm, occlusions, and reflections of lights on the water surface were also taken into consideration.2.5. Behavior Dataset StatisticsThe number of videos for each behavior in our dataset is shown in Table 2. We manually cropped video clips, each containing an average of 8 s of behavior. The demand for a large amount of data used for deep learning was solved by data enhancement. Data enhancement was achieved by rotating and flipping the image, and the number of video clips was expanded by 2 times. In the end, our dataset had a total of 3198 videos. Among them, 70% of the data was used for model training, and 30% of the data was used for model evaluation. Since the behavior of fish after domestication usually exhibits group characteristics, our dataset was mainly used for the behavior recognition of the fish school. A full dataset was established through a larger-scale investigation of fish behavior and communication with fishery experts to overcome the lack of knowledge about fish behavioral understanding.3. Fish Behavior RecognitionThe purpose of our research is to recognize and understand fish behavior through intelligent methods. Inspired by human behavior recognition, we attempted to apply deep learning methods to the automatic recognition of fish behavior. In this section, we introduce the proposed network architecture, and then we present image preprocessing, spatiotemporal feature extraction, and image fusion schemes.3.1. Overview of the Network FrameworkWith the rapid development of deep learning, the C3D network has shown excellent performance in human temporal action recognition [19,20]. Inspired by the spatiotemporal network method based on deep learning, we improved the dual-stream convolutional neural network to realize the behavior recognition of the fish school. In this work, we proposed the DSC3D network to extract spatiotemporal features and motion features from RGB video and optical flow video, respectively. The spatiotemporal features describe the spatial distribution, appearance, and fish movement in consecutive frames. The motion features specifically representing the direction and speed of the fish swimming. Moreover, our network completes the fusion of spatiotemporal features and motion features.The network framework of this article is shown in Figure 3. Our behavior recognition process was based on video analysis and included the following three parts: (1) image preprocessing and motion detection: we effectively removed water surface noise through image preprocessing methods and applied the FlowNet2 method based on deep learning for motion detection; (2) feature extraction based on a spatiotemporal network: our framework used improved C3D as the backbone network to extract spatiotemporal features and motion features from fish behavior videos; (3) feature fusion: a feature map combined spatiotemporal features and motion features for behavior recognition.3.2. Image Preprocessing and Motion DetectionImage preprocessing is an essential process to eliminate image noise and improve the accuracy of feature extraction and image segmentation. Image preprocessing methods include grayscale, geometric transformation, and image enhancement. As explained in Section 2.3, we processed the dataset through image enhancement methods. As a requirement of the C3D model, we adjusted the frame rate of all input videos to 16 frames and the image size to 112 × 112.The imaging quality underwater is easily affected by water surface fluctuations and light refraction compared with human behavior videos. To eliminate the image flare formed by the reflection of the water surface, the contour of the fish body was segmented by the grayscale method. Then, to avoid interference unrelated to fish activity, we used the Gaussian kernel convolution filtering method to remove the noise caused by water surface fluctuations in the image. We obtained a 3 × 3 Gaussian convolution kernel through the Gaussian function for smoothing filtering of the neighboring pixels of the picture pixel.The Gaussian convolution kernel is used to calculate the weighted average of each pixel in a certain neighborhood:(1)Hi,j=12πσ2e−(i−k−1)2+(j−k−1)22σ2 where parameter σ represents the variance, and k represents the dimension of the convolution kernel matrix.As an effective method of detecting moving objects, the optical flow method has been widely used in behavior recognition. Traditional optical flow estimation methods are divided into the dense optical flow and sparse optical flow. The sparse optical flow based on Lucas–Kanade has made certain progress in the research of fish movement detection and tracking [21]. However, both the detection accuracy and detection speed restrict the application in behavior recognition. Recently, the optical flow estimation method based on deep learning has shown excellent performance in terms of accuracy and speed [22].In this work, the FlowNet2 network based on deep learning was applied to fish motion optical flow estimation [23], which provided motion feature information for behavior recognition. In particular, the behavioral characteristics of fish are mainly represented by the state of movement speed and direction. To this end, the optical flow videos were generated by FlowNet2 for motion estimation on the RGB videos. As shown in Figure 4, the optical flow image highlights the movement characteristics of the fish, and the movement direction is represented by the HSV image. The purpose of our method was to detect the movement of multiple objects in the RGB image through optical flow. To this end, we improved the original C3D model into a dual-stream network in which RGB video and optical flow video are simultaneously input. Feature fusion images contain appearance features and motion features, which provide more information about the fish movement.FlowNet2 is an optical flow prediction method based on CNN, which models optical flow prediction as a supervised deep-learning problem:(2)W=CNN(θ,image1,image2) where W is the predicted optical flow, the parameter θ is the parameter to be learned in the CNN, and image1 and image2 are the images to be input.3.3. Spatiotemporal Feature ExtractionRecently, one possible solution for automatized analysis of behavior has been developed that trains a deep learning model on larger datasets for the purpose of predicting behavior types. For example, the 3D convolutional neural networks have been able to successfully applied to the recognition of human behavior, as they can extract the spatiotemporal features of moving objects between consecutive frames [20]. Inspired by human behavior recognition, we proposed a DSC3D network to extract spatiotemporal features of fish RGB videos and motion features in flow videos, respectively, as shown in Figure 5.The original C3D network structure consists of eight convolutional layers, five pooling layers, two fully connected layers, and one softmax layer. The optimal convolution kernel size used by the eight convolution layers is 3 × 3 × 3. The numbers of convolution kernels used in each convolution layer are 64, 128, 256, 256, 512, 512, 512, and 512, respectively. In order to retain more temporal information in the initial stage of the network, the kernel size used by the first pooling layer is 1 × 2 × 2. After the convolution and pooling layer, a 4096-dim vector is generated and mapped to the predicted label after the softmax layer. The video clips are considered with a size of C × T × W × H, where C is the number of channels, T is the length of consecutive frames, W is the width of the image, and H is the height of the image. To unify the video format, we select consecutive frames with a length of 16 as the input of the model. The image size is adjusted to 1280 × 720 and then cropped to a size of 112 × 112. Since we converted the optical flow prediction result to the HSV color space, the channels of RGB video and Flow video are both 3.When the quantity of training data is large, training deep neural network parameters from scratch is a time-consuming task. The transfer learning method was used to fine-tune the parameters of the pre-trained model trained on the UCF-101 data set (including 101 categories and a total of 13,320 videos). The number of parameters in our model was the same as the number of parameters in the original model. In each iteration, the batch size of the shuffled data input to the network was set to 16. Our network used the stochastic gradient descent method for training with a learning rate (lr) of 0.00005, a momentum of 0.9, and a weight attenuation of 0.01.3.4. Image Fusion SchemeImage fusion methods are designed to combine multiple images into a fused image, which further makes the image more informative [24]. Image fusion methods are divided into data-level fusion, feature-level fusion, and decision-level fusion. As far as our dataset is concerned, it is not necessary for data fusion due to the inaccurate registration of RGB images and flow images. The decision-level fusion requires multiple classifiers to perform a weighted average of results, such as voting and blending methods, but its results depend on the model results with higher weights. One major benefit of feature fusion is that it improves the performance of target detection and recognition by perfectly fusing feature information such as multi-level features, spatial features, and temporal features.The feature fusion method of adding was used for image fusion, which is obviously helpful for image classification due to the increase in the amount of information on the feature maps. Feature fusion methods are divided into two types—namely, serial feature fusion (concat) and parallel feature fusion (add). The concat feature fusion method is realized by increasing the number of feature maps, while the add method superimpose the information of the feature maps and the number of channels remains unchanged. Therefore, the computational complexity of the concat method is much greater than that of the add method. We obtained the spatiotemporal features in the RGB image and the motion features in the optical flow image in the feature extraction stage, and then the two features were fused using the max pooling.3.5. Experimental Evaluation IndexAccording to [16], accuracy, precision, recall, and specificity are used to evaluate the performance of the model. Accuracy is a measure of the accuracy of recognition of all behaviors, and it is the ratio of correct recognition samples to all samples. For a specific behavior, the accuracy is the ratio of correctly classified behavior samples to all predicted behavior samples. Recall rate is the ratio of correctly classified samples to the number of samples that belong to this behavior. Specificity is the proportion of genuinely negative samples among the negative results of the test. These four indicators are calculated as follows:(3)Accuracy=TP+TNTP+FP+TN+FN×100% (4)Precision=TPTP+FP×100% (5)Recall=TPTP+FN×100% (6)Specificity=TNFP+TN×100% where true positives (TP) indicate that the positive class was judged as a positive class, false positives (FP) indicate that the negative class was judged as a positive class, false negatives (FN) indicate that the positive class was judged as a negative class, and true negatives (TN) indicate that the negative class was judged to be the positive class.4. Experimental Results4.1. Result Analysis4.1.1. Comparative Analysis of ExperimentsThe average accuracy rate of the proposed algorithm for distinguishing the five behavior states of the fish school reaches 95.79%. As shown in Figure 6, the accuracy and loss curves of model training and verification tend to be in a stable close-companion state at the 20 epoch, which indicates that the model training results are robust and suitable for recirculating aquaculture systems. Our dataset includes the activity states of five fish schools that are identified by specific behaviors. Table 1 shows the description of each behavior. More details of the dataset are introduced in Section 2. For better experimental performance and model evaluation, we divide the dataset into 70% training set, 20% validation set, and 10% test set.The results of precision, recall, specificity, and accuracy of the model on five different behaviors of fish are shown in Table 3. In addition to the misjudgment of a small number of samples, our proposed method realizes the accurate identification of fish behavior. To verify the effectiveness of the method, we selected 642 fish behavior videos as the test set. For example, the recognition accuracy of hypoxia and frightening behavior is low, achieving 96.9% and 97.2%, respectively. The confusion matrix of five behaviors is shown in Figure 7, including feeding, hypoxia, frightening, hypothermia, and normal state. From Figure 7, we derive the number of correctly predicted behaviors described on the diagonal line. The dark color indicates high confidence in prediction, while the light color shows low confidence in behavior misrecognition. This further shows the feasibility of our method for the accurate recognition of fish behavior.We compared and analyzed the current main deep learning methods for fish behavior recognition on our dataset, including LeNet5 [16], 3D residual networks [25], and convolutional neural networks–long short term memory (CNN-LSTM) [17], as shown in Table 4. LeNet5 is a CNN-based method, which was used to evaluate the feeding intensity and appetite of fish by extracting image features of feeding behavior. In addition, 3D-ResNet is a deep convolutional neural network that extracts information in the temporal and spatial dimensions of the video and has been used to identify fish feeding and building behaviors. The CNN-LSTM architecture combines the feature extraction of CNN and the learning ability of recurrent neural network (RNN) for time series, which has been used for animal behavior recognition. Inspired by the dual-stream network, we directly extracted the spatiotemporal features in the RGB dataset and the motion features in the optical flow dataset through the C3D network and realized feature fusion in the last layer of feature extraction. The results show that the average accuracy of the algorithm is 2.46% higher than that of CNN-LSTM and 6.97% higher than that of the classic 3D residual networks. We found that the 3D convolutional network based on feature fusion as a feature extractor is more conducive to fish behavior classification.4.1.2. Comparison of Spatiotemporal Feature ExtractionIn this section, different feature extraction networks were compared and analyzed, as shown in Table 5. Input resolution represents the size of the input network image, Params represents the model parameters of the neural network, and MACs represents the numbers of multiply-accumulate operations required by the algorithm. The results show that the C3D network achieves the best accuracy with moderate Params and MACs. More broadly, even if the C3D network is only used as a feature extractor without feature fusion, it still achieves an accuracy of 92.8%.We compared and analyzed the I3D network and our method, both of which are dual-stream networks that take RGB videos and optical flow videos as input [26]. Although the improved network based on C3D has higher parameters and model complexity than I3D, our method achieves the highest behavior recognition accuracy rate of 95.79%. This shows that as a stable, simple, and efficient model, C3D is more capable of gathering the target, scene, and motion information in the video. Furthermore, I3D simply averages the results after training the network separately on RGB videos and optical flow videos, while our method considers the fusion of spatiotemporal features and motion features for accurate recognition of actions.4.1.3. Comparative Analysis of Feature FusionIn this work, we attempted to further improve performance through feature fusion, and the fusion result is shown in Figure 8. To emphasize the importance of feature fusion, we also trained on the original RGB and optical flow data, respectively. It is evident that the training result of feature fusion is 2.99% higher than the single RGB data and 34.89% higher than the optical flow data. This shows that the fusion of spatiotemporal features and motion features can more effectively improve the performance of the model for detecting fish motion information.We compared the effects of the early fusion and the late fusion on the model recognition performance by adding fusion features after Conv1 and Conv5b, respectively. As shown in Table 6, the results show that the recognition accuracy of high-level feature fusion is 3.89% higher than that of low-level fusion, and the training loss is reduced by 0.116. This indicates that the high-level features with strong semantic information further improve the detection ability of the region of interest in the image, while the feature image is low in resolution and poor in the perception of details. In addition, our method also adds more spatial motion information to the high-level features, which improves the efficiency of feature fusion.4.1.4. Visualization of Feature MapThe different motion patterns of the 3D CNN features of the five behavior sequences are shown in Figure 9. The feature visualization results help us understand the temporal and spatial regions that the model focuses on and display them in the form of heat maps. From Figure 9, the visualization results show the feature maps extracted by our method on the five behaviors, and the regions of interest in the feature maps corresponding to each behavior are disparate. The reason for this difference is that the fish exhibit very vigorous activities after being exposed to food and other external stimuli, while the fish exhibit an inactive state under conditions such as hypoxia and hypothermia. The similarities between fish behaviors are inevitable, while computer vision technology could still accurately distinguish the subtle changes in the temporal and spatial regions of interest. The experimental results show that the CNN feature map effectively distinguishes the difference in fish movement patterns, which is more conducive to the result of behavior classification.4.2. Discussion4.2.1. The Basis of Fish Behavior Recognition Based on Video AnalysisThe purpose of our research was to intelligently recognize five behaviors of fish, including feeding, hypoxia, hypothermia, frightening, and normal behavior. The recognition of typical behaviors is useful to understand the real activities of fish, which directly helps to discover the health status and diseases of fish. For example, video-based behavior analysis can also accurately quantify fish activity that is closely related to the health of fish and provides help for the diagnosis of an environmental stress response. In related work, researchers mostly use acceleration sensors and other equipment implanted in fish skin and body, combined with machine learning methods to realize fish behavior recognition. However, this method requires high accuracy of the acceleration sensor underwater, and this operation is likely to cause stress to the fish. Furthermore, the method based on acceleration sensors still has difficulty in identifying more complicated fish behaviors. Therefore, the advantages and potentials of fish behavior recognition based on video analysis are incomparable.The automatic recognition of fish behavior states was realized by deep learning methods, and the main behaviors of fish in aquaculture were collected. However, fish behavior recognition based on video analysis is still a challenging task under conditions of insufficient light. Researchers using near-infrared machine vision to quantify fish feeding behavior creates a new solution for behavior recognition under insufficient light conditions [7,27]. As described in Section 4.1.1, our proposed method realized the behavior recognition of fish under different environmental conditions, which lays the foundation for more complex behavior recognition of fish.4.2.2. The Problem of Identification of Similar BehaviorsThe recognition method based on deep learning was implemented by extracting rich feature information from video clips. We highlighted the movement behavior of fish in the feature image by fusing spatiotemporal features and motion features. The results show that our proposed method further improves the accuracy of behavior recognition and meets the needs of behavior classification in aquaculture.However, fish behaviors are susceptible to external environmental stimuli, showing a series of complex behavior states. Therefore, many similar behavior fragments lead to the misjudgment of the model, as shown in Figure 9. The most significant is that both the feeding and frightening behaviors stimulated by the external environment exhibit the aggregation state of the fish school. This behavioral change involves the rapid movement of the fish to a certain position. The similarity of behavioral activity is the reason for the misjudgment of feeding behavior and frightening behavior. In addition to floating at the inlet of the circulating water, fish also exhibit slow swimming behavior under hypoxia. This inactive behavior often occurs under normal conditions, and its appearance is similar to hypoxic behavior. Therefore, the research should focus on how to extract deep feature information and accurately identify behaviors in different environments.4.2.3. Other Behavior RecognitionIn our study, only five basic fish behaviors were identified, including feeding, hypoxia, hypothermia, frightening, and normal behavior. However, fish also exhibit higher-level behaviors such as fish reproduction and fighting in real farming. Reproductive behavior is an essential behavior to optimize the breeding level in aquaculture. We focused on the study of fish movement behavior, while reproductive behavior is difficult to distinguish from normal behavior. Fighting behavior is a competitive behavior in which two or more fishes compete for resources, but it is not easy to collect behavioral data. Furthermore, a general network is difficult to recognize multiple behaviors at the same time, which will reduce the confidence of the network prediction result. Therefore, future research will consider more types of behaviors and further improve the ability to predict multi-classification tasks. In the foreseeable future, the fusion of machine vision methods and acoustic imaging methods will realize the intelligent recognition of fish behavior in aquaculture.5. ConclusionsIn this paper, we proposed a DSC3D network based on deep learning for fish behavior recognition with spatiotemporal information and motion information. This research further expands the idea of fish schools’ behavior recognition through video analysis and specifically focused on animal group behavior. Our proposed method used C3D as a feature extractor to extract spatiotemporal features of RGB videos and motion features of optical flow videos, and feature maps fused with depth feature information were used to classify fish behaviors. This method accurately identifies the five behavioral states of the fish school, including feeding, hypoxia, hypothermia, frightening, and normal behavior. The average accuracy of this algorithm for fish behavior recognition has reached 95.79%, which significantly proves the effectiveness of our strategy on the test dataset. The RGB camera used in our method can be operated in real farming, which is helpful to promoting daily management work. More broadly, we also discussed the challenges of understanding and recognizing fish behavior to improve the efficiency of farming and animal welfare. In the future, our work needs to consider identifying and tracking the behavior of each individual to understand more complex fish behaviors.

animals : an open access journal from mdpi

10.3390/ani12030266

PMC8833334

Observing the health and wellness of livestock is time consuming and costly. Sensor technologies can identify changes in animal activity, providing the potential to remotely monitor livestock health status and welfare. As part of another study, 10 ewes in a pen setting were monitored with near real-time accelerometers manufactured by Herddogg. Ewes were inadvertently fed moldy corn silage. The moldy feed was removed the following day and ewes displaying symptoms, such as reduced intake and difficulty walking, were treated under the direction of a veterinarian. Accelerometers showed a distinct decrease in activity for 4 days after the ewes were exposed to moldy feed. Accelerometers also showed an increase in activity of symptomatic ewes after treatment. Real-time and near real-time accelerometers have the potential to remotely detect changes in sheep activity that occur when animals become ill from mold contaminated feed and perhaps other illnesses, which could help producers monitor livestock health and provide a more timely response when they become ill.

Sensor technologies can identify modified animal activity indicating changes in health status. This study investigated sheep behavior before and after illness caused by mold-contaminated feed using tri-axial accelerometers. Ten ewes were fitted with HerdDogg biometric accelerometers. Five ewes were concurrently fitted with Axivity AX3 accelerometers. The flock was exposed to mold-contaminated feed following an unexpected ration change, and observed symptomatic ewes were treated with a veterinarian-directed protocol. Accelerometer data were evaluated 4 days before exposure (d −4 to −1); the day of ration change (d 0); and 4 days post exposure (d 1 to 4). Herddogg activity index correlated to the variability of minimum and standard deviation of motion intensity monitored by the Axivity accelerometer. Herddogg activity index was lower (p < 0.05) during the mornings (0800 to 1100 h) of days 2 to 4 and the evening of day 1 than days −4 to 0. Symptomatic ewes had lower activity levels in the morning and higher levels at night. After accounting for symptoms, activity levels during days 1 to 4 were lower (p < 0.05) than days −4 to 0 the morning after exposure. Results suggest real-time or near-real time accelerometers have potential to detect illness in ewes.

1. IntroductionBehavior is a basic indicator of an animal’s health and wellness state [1], which highlights the importance of knowing normal behavior patterns of an individual animal. Producers observe animal behavior to assess health status and make management decisions, in an effort to maintain animal welfare and increase productivity of their operation. Animals experiencing illness or painful conditions typically demonstrate changes in appearance, appetite, posture, and behavioral patterns [2,3]. However, prey animals like sheep tolerate pain and injuries without displaying an overt change in behavior as a means to limit vulnerability towards predators and, subsequently, increase their chances of survival [4,5]. The ability of sheep to mask their pain by maintaining a stoic demeanor may hinder early detection of subtle behavior changes. Behavioral irregularities related to painful events can be difficult to identify by observation alone in an intensive system, where space is limited, and an animal can blend within a group of animals. Monitoring livestock may also be difficult in a pasture setting due to low frequency of human observation. Human observation of livestock can be labor intensive; however, it is crucial to detect subtle changes in behavior that might be associated with illness to allow prompt treatment before an animal’s health is further compromised.A great deal of recent research has focused on remote monitoring of livestock. Most of the research has focused on on-animal sensors such as global position system (GPS) tracking and accelerometers [6]. These systems can be used to identify changes in normal activity patterns of livestock, which can be an indication of illness or other animal well-being concerns [7]. Tracking has been used successfully to detect parturition in sheep [8,9] and simulated water failure in cattle [10]. However, value of GPS tracking is limited in intensive and pen situations because the spatial movements of livestock are constrained.The application of accelerometers has been utilized to across livestock systems to identify normal behavior and changes due to shifts in well-being and health status. Accelerometers have been used to detect parturition in sheep [11,12]. Accelerometers have also been used to monitor rumination [13], grazing activity [14], and drinking behavior [15] in cattle. Multiple studies have utilized accelerometers and have shown reductions in activity and movement intensities. Tobin et al. [16] used accelerometers to detect the reduction of movement intensity due to bovine ephemeral fever in heifers. Ikurior et al. [17] deployed tri-axial accelerometers to detect changes in lamb activity due to gastrointestinal nematode infection.For intensive operations, more options for remote monitoring are available than for rangeland and pasture-based systems [7]. Thermal imaging has potential to remotely monitor livestock health and well-being [18]. Rumen boluses can also be used to remotely monitor changes in body temperatures of cattle [19]. Indwelling vaginal temperature sensors can be utilized to determine internal temperatures [20]. The usage of multiple sensors is to provide automatic, continuous monitoring of individual animals [21].Deviation from normal activities can also be a consequence of an abrupt change in diet. Exposure of livestock to toxins present in feed can cause immediate and lethal consequences that require prompt intervention. Molds and mycotoxins are common contaminants in feedstuffs, which are most frequently observed in hay and silages [22]. Mycotoxins impact on animal health is based on a number of factors including level of exposure to feed contamination and animal sensitivity due to immunosuppression [23]. Differentiating molds and mycotoxins’ effects on animal health and performance are challenging [24], as molds may be present without producing toxins [25]. Mycotoxins are a large and complex group of secondary metabolites, which are produced by fungi and certain varieties of molds that stimulate a multitude of toxic responses in humans and animals [22,23]. These low-molecular-weight metabolites are toxic when consumed even in low concentrations [25]. Mycotoxicosis from consumption of feedstuffs containing toxic metabolites typically comes about without producer awareness and transpires over a long period of exposure [22]. Exposure to mycotoxins can also have escalating and undetectable consequences on animal health in early stages [22]. Reduced feed intake and prolonged duration of feeding intervals may occur after exposure to moldy feed due to decreased feed digestibility [26]. Numerous mycotoxins are able to modify the rumen microflora, resulting in decreased milk production, mild diarrhea, and poor feed conversion [26].Low levels of chronic exposure to mycotoxins are not always noticed and behavioral change may be challenging to detect due to the rumen’s ability to degrade, deactivate, and bind toxic molecules [22]. However, detecting behavioral changes at the onset of those animals exhibiting acute mycotoxicosis, where detrimental signs of disease are present, is critical to allow the manager to respond before health is too heavily altered. Sensor technologies possess the ability to detect minute changes in activity, to aid in monitoring livestock. Studies suggest sensor technologies are capable of detecting a variety of abnormalities in behavior linked to changes in animal health [16]. The aim of this case study was to investigate the potential to remotely monitor changes in behavior associated with illness caused by mold-contaminated feed.2. Materials and Methods2.1. Site and AnimalsAll procedures were approved by the New Mexico State University Animal Care and Use Committee (2019-007). This analysis was based on a separate study evaluating parturition and was conducted on the campus of New Mexico State University at the West Sheep Unit research facility. Twenty-five fine wool Debouillet, ewes averaged 3 years (± 0.2 SEM) of age and weighed 79 kg (±3.2) at the start of the parturition study. All animals were observed daily and were healthy at the onset of this study. Ewes were housed in a single pen (18.3 m × 9.1 m) to evaluate the potential of accelerometers to detect parturition-related behavior events [11]. Each ewe was fed 1.8 kg of chopped alfalfa in the morning (0800 h) and supplemented with 0.22 kg of cracked corn both morning (0800 h) and afternoon (1600 h). An unexpected ration change to corn silage occurred on 16 April 2019. Ewes were either in late gestation or in the first 21 days of lactation at the time of the unexpected feed change (16 April 2019) described below. Throughout the trial ewes had ad libitum access to water, mineral, and salt.2.2. AccelerometersA tri-axial Axivity AX3 MEMS accelerometer (Axivity Ltd., Newcastle, UK) was attached to an Allflex ear tag (Allflex USA Inc., Dallas, TX, USA) with shrink wrap tubing, then randomly placed in either the left or right ear of 13 ewes prior to parturition. Accelerometer ear tags were placed on ewes on 13 March 2019 and were removed on 12 May 2019. Accelerometers were charged prior to deployment to last a duration minimum of 30 days (study duration). Axivity accelerometers were configured to collect acceleration signals at a sample rate of 12.5 Hz measuring longitudinal movements of the horizontal x-axis (left and right), longitudinal y-axis (forward and backward), and vertical z-axis (up and down). The dimensions of each accelerometer were 23 mm × 32.5 mm × 7.6 mm and weighed 11 g. Accelerometer movements were subsequently stored on the NAND Memory within the device. Accelerometers were removed post-study to retrieve data via USB connection to the OmGui Axivity computer software. The OmGui program downloads data from the accelerometer, allows for manipulation for desired study period, and stores raw data as a .CWA file, not compatible with Microsoft Excel (Microsoft Corporation, Redmond, WA, USA).A HerdDogg biometric accelerometer ear tag (HerdDogg, Inc., Ashland, OR, USA), was attached to 25 ewes prior to parturition. On the 13 ewes that received an Axivity AX3 accelerometer, a HerdDogg accelerometer was attached on the opposing side ear. HerdDogg accelerometers were configured to collect tri-axial acceleration signals at a sample rate of 24 Hz. The accelerometer signals from the X, Y and Z axes were processed in the HerdDogg tag and condensed into one proprietary index value every 6 min. The raw X, Y, and Z data is not stored on the tag and is only used to calculate the index value. Ambient temperature and temperature measured by a sensor next to the ear were also transmitted in 6 min epochs. Data transmitted from the ear tag was gathered by HerdDogg’s “DoggBone” receiver, which transmits the data via cellular network technologies to a website and smart phone app where it can be viewed. The dimensions of the HerdDogg ear tag were 76.2 mm × 38.1 mm × 12.7 mm and weighed 25 g.2.3. Daily Animal Observation ProtocolEwes from the West Sheep Unit research facility were regularly checked at 0800 h at time of feeding and 1300 h daily for any signs of abnormal behavior or illness within the animals. However, when symptoms were first presented, ewes were checked for change in behavior and locomotion four times daily. Symptomatic ewes were then treated with a protocol directed by a veterinarian (discussed below).2.4. Ration ChangeEwe’s ration was unexpectedly shifted overnight on 16 April 2019 from an alfalfa-corn mixture to a corn-silage. No observable presence of mold or other contaminants were noted upon feeding the corn-silage. The entire flock from the West Sheep Unit were exposed to the corn-silage feed, which was later determined to have been contaminated with mold, including the ewes with a HerdDogg ear tag and Axivity AX3 accelerometer. The unintentional exposure of the ewes to the contaminated feed occurred during late pregnancy or early lactation. Contaminated feed was removed from feed bunks the following day (17 April 2019) in the afternoon after overt detrimental changes to health status were observed. A feed sample was collected for feed analysis and sent to be performed by SDK Laboratories (SDK Laboratories Inc., Hutchinson, KS, USA) (Table 1). Ration was switched to 100% alfalfa on 18 April 2019, and ewes were treated based on the protocol below. Animals were moved to a separate pen for treatment.2.5. Symptoms and Resulting TreatmentsWithin 24 h of exposure; one ewe demonstrated difficulty standing, kept her head down and ultimately died within the first day after exposure (this ewe was not included in the current study). Within two days after exposure, seven of the 25 ewes in the pen discontinued eating and had difficulty walking. Three days post-exposure, five ewes persisted with symptoms of difficultly walking and not eating, even after treatment. Four days after exposure, all but one symptomatic ewe returned to a normal state. Symptoms recorded included: no feed intake, difficulty or weakness in walking, head held down, knuckling of the feet (Figure 1), and continuous shifting of body weight from one leg to another.Treatments began on the morning of 18 April 2019. All ewes showing symptoms were treated with 60 mL of sodium bicarbonate, 10 mL of Bismuth subsalicylate, 3 mL of Banamine®, and were placed into a separate sick pen.2.6. Data Collected and EvaluatedAll ewes in the flock were evaluated opportunistically after unexpected exposure to mold contaminated corn silage. Observation data were recorded for ewes exhibiting abnormal behavior each day after feed exposure until symptoms subsided. Daily treatments were also recorded for each ewe.The unexpected feeding of moldy feed occurred near the end of the expected battery life for the Axivity accelerometer batter life (30 days). Only five ewes had nearly complete data (>95% of expected data recordings) during this study period. The Herddogg accelerometer tags were an earlier version of the technology and sometimes did not transfer all the index values to the Dogbone receiver. Fifteen of the 25 ewes with HerdDogg ear tags could not be used in the study due to inconsistent missing index values within the data sets.Accelerometer data recorded by HerdDogg ear tags were used to evaluate ewe behavior 4 d prior to exposure to the moldy corn silage (days −4 to −1), the day moldy corn silage was first introduced (day 0), and 4 d post-exposure (days 1 to 4). This data set allowed us to evaluate ewe behavior in this case study using a before and after analysis. Ten ewes had complete HerdDogg accelerometer data during this 9 d period and were included in the study. Three of these 10 ewes displayed symptoms and were treated. HerdDogg accelerometer ear tags provide an index value every 6 min that the manufacturer states is related to animal activity. These index values were averaged each hour for statistical analyses.To gain a better appreciation of the proprietary HerdDogg accelerometer index, the correlation between the accelerometer index and metrics calculated from the Axivity accelerometer were calculated. Five ewes had both Axivity and Herddogg data for 8 days at the time of the study (12 April to 20 April). Accelerometer data were retrieved using the Axivity proprietary software Version 1.0.0.37, OmGui, and condensed into 10 s epochs using Anaconda Python programming. Movement intensity (MI) and signal magnitude area (SMA) was calculated for accelerometer reading. (1)MI=1T∑t=1T(Ax2)+(Ay2)+(Az2)(t) (2)SMA=1T(∑t=1T|Ax(t)|+∑t=1T|Ay(t)|+∑t=1T|Az(t)|) where Ax, Ay, and Az are the Axivity accelerometer readings from the x, y, and z axes, respectively as described in detail by Gurule et al. [11] and Tobin et al. [16].The mean MI, range of MI, standard deviation of MI, mean of signal magnitude area (SMA), mean of the x axis, mean of the y axis, and mean of the z axis were calculated for each 10 s epoch. These metrics calculated from the Axivity data were averaged by hour and paired with corresponding hourly averages of the HerdDogg accelerometer index for correlation analyses. We paired Axivity metrics and the HerdDogg index data (by hour each day) for each ewe and pooled the data from all ewes. We used MI as a metric for the Axivity data, because the Herddogg manufacturer reported to us that their proprietary index was a compilation of data from all 3 axes and was similar but not the same as MI [27].The placement of the Axivity and HerdDogg accelerometer in the left or right ear could potentially affect the values accelerometer readings. However, there were only 5 ewes with complete data so there was not sufficient data to compare the effect of placing the tag in the left or right ear. In addition, we used a before and after analysis of individual ewes so the impact of placement of the tag was accounted for by the subject of the repeated measures model (see below).2.7. Statistical AnalysesThe hourly averages of the HerdDogg accelerometer index were classified into four periods, morning (0800 to 1100 h), midday (1100 to 1700), evening (1700 to 2000), and night (2000 to 0800). Due to symptoms including reduced feed intake and difficulty walking, periods were classified into periods where ewes typically had different activity levels based on visual observations. Morning was the period when ewes were fed and were typically active. Midday had lower activity than feeding, activity often increased during the evening, and night had the lowest activity level. These diurnal patterns in sheep activity are typical of domestic sheep [28] had to be accounted for in the analyses in order to accurately compare before and after the moldy forage was fed.The HerdDogg accelerometer index data were analyzed using repeated measures in PROC MIXED of SAS [29]. The fixed effects were day (−4 to 4), period (morning, midday, evening, and night), hour within period, and the period × day interaction. Day 0 was the day ewes were first exposed to the moldy feed. Days −4 to −1 were the four days prior to feeding the moldy corn silage, with day −1 being the day immediately prior to feeding. Days 1 to 4 were the days following the exposure to the moldy feed, with day 1 being the day that moldy corn silage was removed from the feed bunks midday, and day 2 being the day treatment for the symptomatic ewes began following the veterinarian directed protocol discussed above. The subject of the repeated measures model was ewe. The covariance structures evaluated were compound symmetry, autoregressive order 1, and unstructured [29]. The structure used of the three was based on the lowest Akaike Information Criterion (AIC). Pairwise comparisons of days, periods, and the day × period interactions were evaluated using the PDIFF function of PROC MIXED.A second similar repeated measures analysis was completed with the addition of the fixed effect of symptom presence. Specifically, the model included day, period, hour within period, and symptom presence. The two-way interactions and three-way interactions of these fixed effects were also evaluated. Ewe was the subject and the covariance structure with the lowest AIC was autoregressive order 1. A three-way interaction of day, period, and symptom presence was not detected (p = 0.49), so it was dropped from the model.Simple (Pearson) correlation coefficients were calculated between the HerdDogg accelerometer index and the metrics calculated from the Axivity accelerometers. Pairs of the HerdDogg index and Axivity metrics were based on hourly measures collected over 8 days (see above). Separate correlation coefficients were calculated for each of the five ewes with complete data during this period. In addition, correlation coefficients were collected for the pooled data of all five ewes.The HerdDogg accelerometer index was used to evaluate the changes in behavior before and after feeding the moldy corn silage, due to more ewes being monitored with HerdDogg during the period that the silage was fed (10 versus five). Only one of the ewes monitored with Axivity accelerometers showed any symptoms versus three ewes with HerdDogg. Axivity data was not complete for the 9-day study (days −4 to 4) for the majority of the five ewes due to loss of battery charge in the Axivity accelerometer. In addition, the HerdDogg accelerometer tags are commercially available and transmit to a website and smart phone app in real time. Axivity accelerometers are “store on board” and accelerometers must be removed from the ewe to download data and are not applicable to commercial livestock operations but serve as a valuable research tool.3. Results3.1. Evaluation without Considering Symptom PresenceNo differences in the Herddogg activity index among days (−4 to 4) were detected (p = 0.16). Activity varied (p < 0.001) among periods (Figure 2). Morning had greater (p < 0.001) activity than midday, evening, or night, and night had lower (p < 0.0001) activity than morning, midday, and evening. No differences (p = 0.18) were detected between evening and midday. Activity also varied (p < 0.001) among hours within period.There was an interaction (p < 0.001) between day and period (Figure 2). No differences in activity were detected (p > 0.01) in the morning between days −4 to 0. Activity in the morning was lower (p < 0.05) on days 1 to 4 than days −4 to −2. No differences in morning activity were detected (p > 0.05) between day −1 and days 0 and 1. Activity in the morning was lower (p ≤ 0.01) on days 2 to 4 than day −1. Activity in the morning on days −1 and 0 were higher (p < 0.01) than days 2 to 4. Morning activity on day 2 was lower (p = 0.04) than day 1. No differences (p > 0.05) were detected between morning activity on day 1 than days 3 and 4. No differences in morning activity were detected (p > 0.05) among days 2, 3, and 4.No differences among days in activity at midday were detected (p > 0.05). Activity in the evening was lower (p < 0.05) on day 1 than days −4, −3, −2, 0, 2, 3, and 4. No differences in activity were detected (p = 0.14) between day 1 and day −1 in the evening. No other differences in evening activity were detected (p > 0.05) among the other days. No differences in nighttime activity were detected (p > 0.05) among days.3.2. Evaluation Considering Symptom PresenceSimilar to the previous analysis, the Herddogg activity index varied among days (p = 0.03). As expected, no differences in activity were detected (p > 0.05) prior to feeding the moldy corn silage (days −4 to day −1). No differences in activity were detected (p > 0.05) between day 0 and days −4 to day −1. Activity was lower (p < 0.05) on days 1 and 2 than days −4 to 0. No differences in activity were detected (p > 0.05) between days 3 and 4 and days −4 to 0. No differences were detected (p > 0.05) between day 1 and 2 and days 3 and 4.Activity differed (p < 0.001) among periods with morning having greater (p < 0.001) activity than midday, evening or night, and night being lower (p < 0.0001) than morning, midday, and evening. No differences (p = 0.36) were detected between evening and midday. Activity also varied (p < 0.001) among hours within period.No differences in activity were detected (p = 0.47) between ewes displaying symptoms and ewes not displaying symptoms. However, there was an interaction (p < 0.001) between period and presence of symptoms (Figure 3). Ewes displaying symptoms had lower (p < 0.001) activity during the morning and higher (p < 0.001) activity at night than ewes that did not display symptoms. No differences in activity were detected (p > 0.05) between ewes displaying symptoms and ewes not displaying symptoms during midday and evening periods.Herddogg accelerometer data from the three ewes displaying symptoms identified a decrease in activity (p = 0.03) for 2 days after feed exposure compared to the 4 days before exposure (Figure 4). Three days after exposure and 2 days after treatment, no difference in activity was detected (p > 0.05) between pre- and post-mold exposure levels.3.3. Correlations between HerdDogg and Axivity MetricsCorrelations between the HerdDogg accelerometer index and Axivity accelerometer metrics were not consistent across the five ewes evaluated (Table 2). Across all ewes, the HerdDogg accelerometer index was most correlated to the standard deviation of MI (0.62) and the minimum of MI (−0.65). For all but one of the five ewes, there was a strong correlation (>0.60) between MI standard deviation and the HerdDogg index (Table 2). For ewes 601 and 742, mean MI was only weakly correlated to the HerdDogg accelerometer index. Stronger correlations were found between the minimum and standard deviation of MI for ewe 742 (Table 2). The correlations between the HerdDogg index and the Axivity metrics were weak for ewe 601. Mean SMA was weakly or moderately related to the HerdDogg index. The HerdDogg accelerometer index was weakly correlated to the means of the x, y, and z axes of the Axivity accelerometer (Table 2).4. DiscussionThe HerdDogg accelerometer index is designed to monitor livestock activity (https://herddogg.com accessed on 15 June 2021). Metrics from the Axivity accelerometers were more correlated to the mean and variation (minimum and standard deviation) of MI than the means of individual axes and SMA. Results from the associated study, Gurule et al. [12], showed that greater MI values and greater variation in MI were associated with active behavior. Similarly, Fogarty et al. [30] found that ewe activity was most associated with the variation in accelerometer metrics rather than the means of the metrics. Results from the present study, suggest that the HerdDogg accelerometer tags provide an index that reflects the variation in accelerometer movements Changes in the HerdDogg accelerometer index are related to the variation in the head movements of the ewe. More variation in head movements likely means higher HerdDogg index values and likely more ewe activity.When all ewes were evaluated without considering if the ewe displayed symptoms, no differences in activity when considering the entire day were detected. However, there was a clear identification of a reduction in activity during the morning after ewes were fed. Decrease in activity monitored by the HerdDogg ear tag showed a clear decrease from 0800 to 1100 h after the moldy corn silage was fed. There was also a decrease in activity in the evening (1700 to 2000 h), when ewes began displaying symptoms and the moldy feed was removed from the bunk. This HerdDogg monitoring data was transmitted and recorded using their website. This change in behavior could have alerted the caretakers, had an algorithm been developed to detect the change in morning behavior, most likely feeding behavior. Results from this study show the potential to detect a decrease in feeding behavior. The accelerometer index dropped to less than half of its previous values (days −4 to 0) on day 2 (Figure 4). This decrease in activity was also apparent in the HerdDogg accelerometer reading during the evening of day 1 when symptoms were first observed.When ewes that displayed symptoms were included in the statistical model, there was a clear change in activity among days. Activity on days 1 and 2 were lower (after feeding) than previous days (days −4 to 0). After treatment on days 3 and 4, no differences in activity could be detected from pre-mold exposure (days −4 to −1). Both ewes with symptoms and without systems displayed lower activity after feeding moldy feed, but the change in behavior of ewes without symptoms was less than ewes with symptoms and limited to the morning period (Figure 4).Ewes that displayed symptoms displayed different diurnal activity patterns from ewes that did not have symptoms. Ewes with symptoms had lower activity in the morning (normal feeding period) and greater activity at night than ewes that did not display symptoms. These behavioral differences may or may not be associated with the susceptibility of the ewes to mycotoxins. Accounting for the differences in activity of the ewes displaying symptoms did improve the precision of the statistical model (lower AIC value) and allow us to detect a difference among days. More research is needed to determine if diurnal activity patterns affect the susceptibility of ewes to mycotoxins.The decrease in activity that was detected after moldy forage was fed to the ewes could occur if ewes became ill from other sources. For examples, Tobin et al. [16] found that the activity of heifers diagnosed with bovine ephemeral fever declined. Much more research is needed before the exact cause of illness can be determined from accelerometers and other on-animal sensors. However, this does not negate the value remote monitoring with on-animal sensors. If the system identifies a potential problem as indicated by a deviation from an animal’s normal behavior and notifies the caretakers, the staff can locate and observe the animal in question and make diagnosis. Providing prompt notice of potential health concerns from a change from normal behavior should help caretakers quickly respond and provide treatment if needed.In this study, the majority of the HerdDogg accelerometer ear tags were not able to successfully transfer all the data from the tag to the Dogbone reader. However, the technology in this system continues to be developed and improved. The problems transferring data in this study likely would not occur with current versions of the HerdDogg tags and Dogbones. The distance that data can be transmitted has increased from approximately 10 m to 100 m [31].Mycotoxins may not always be present in moldy feed; nonetheless, mold itself can cause negative effects on health and production [32]. It has been suggested that ruminants are less susceptible to mycotoxins by conversion in the flora to biologically inactive metabolites; however, that does not apply to all mycotoxins that may contaminate feed [33]. In a study by Kiessling et al. [34], mycotoxins were incubated in rumens of sheep and cattle, mycotoxin concentration was measured and demonstrated that aflatoxin B1 and deoxynivalenol were not degraded by rumen microorganisms. A common practice in ill ruminants is transfaunating the rumen, by providing microorganisms from a healthy donor to re-establish the microbial population in a sick animal [35]. One ewe exposed to the moldy silage had persisting symptoms (six days post-exposure) even after treatment, so a rumen transfaunation was performed using a healthy donor ewe.There are over 300 known mycotoxins [25], which can make detection difficult, as only three to four of the most common mycotoxins are tested [32]. Not all mycotoxins can be detected and conjugated mycotoxins can be masked in routine testing by commercial laboratory analysis [19,36]. In addition, detection can be compromised by the variance in a representative sample, as molds can produce large quantities of mycotoxins in small areas and are not evenly distributed in the feedstuffs [36]. Although no mycotoxins were detected in our study, high mold count has been demonstrated to produce potent mycotoxins effecting animal health [36]. Table 1 of the present study demonstrates no detectable mycotoxin. However, a negative test of mycotoxin with animals showing symptoms suggests mycotoxin was present but not detectable [32].Stage of production may be an important factor in resulting symptoms of mycotoxin exposure. Animals experiencing stressful situations, such as parturition, may have more pronounced symptoms due to their already suppressed immune system [36]. Applebaum et al. [37] observed that cows that were treated with impure aflatoxin B1 demonstrated a significant decrease in milk production. During the lambing season following the exposure to mold-contaminated feed, ewes from our study displayed low milk production, as numerous lambs were being supplemented with synthetic milk.Animals may respond to mold and mycotoxin exposure differently based on duration and dose of exposure, stress, and age [38]. One of the most prominent symptoms of mycotoxicosis is reduced feed intake or feed refusal [39] and knuckling of the feet (Figure 1). Ewes in the present study that displayed acute toxicosis refused feed, which could be detected in accelerometer data when compared to normal diurnal activity patterns. These sensor technologies are fixed on a lightweight ear tag providing minimal to no obstruction, allowing natural movement of the animal. Therefore, these sensor technologies have potential to discriminate between subtle changes in normal behavior, indicating deviation in animal’s health status before observable health consequences are notable. This may be beneficial to the manager by minimizing their economic losses and indicating change in activity before complications arise.5. ConclusionsHerdDogg accelerometer ear tags were capable of detecting changes in activity and behavior of ewes that were exposed to moldy feed in this case study. Several ewes displayed symptoms, such as reduced or no feed intake and difficulty walking. This ear tag monitoring system is commercially available and transmits the data in real time in a pen setting. Continued developments in remote monitoring systems will facilitate reliable transmission of data from on-animal sensors to the manager. Development of algorithms that can detect changes in behavior and activity could be used to alert managers when ewes in a pen setting face a well-being concern such as consumption of mycotoxins in moldy feed. More research is needed for development of such algorithms and validation of this “real-time” technology for remotely monitoring sheep well-being in a pen setting.

animals : an open access journal from mdpi

10.3390/ani11072048

PMC8300261

Post-ovulatory and maternal oocyte aging impair female reproductive capacity through several mechanisms that are not fully understood. Urolithin A (UA) is a natural compound previously identified to exert an anti-aging effects in several cells, which has never been used in bovine germinal cells. Our goal was to study UA effect on the developmental potential of the female gamete and the surround cumulus cells obtained from young and adult cows. A model for in vitro aging of female gametes was implemented to study different problems associated with reproductive aging and fertility impairment. Results confirmed that aging exerts a harmful effect on oocyte quality measured by using different parameters and gene expression levels of cumulus cells. Moreover, UA supplementation was an effective way to prevent oocyte aging, improving the subsequent bovine embryonic development.

Oxidative stress and mitochondrial dysfunction have been associated with the age-related decline of oocyte quality and strategies for their prevention are currently quested. Urolithin A (UA) is a natural metabolite with pro-apoptotic and antioxidant effects, capable of preventing the accumulation of dysfunctional mitochondria in different aged cells. UA has never been tested in bovine oocytes. Our aim was to study the effect of UA on the developmental potential of cumulus-oocyte-complexes (COCs) and granulosa cells’ (GCs) expression of important genes related to reproductive competence. Nuclear maturation progression, mitochondrial membrane potential (MMP) and developmental competence of physiologically mature (22 h) and in vitro aged oocytes (30 h of IVM) obtained from prepubertal and adult females, either supplemented with UA or not were assessed. Additionally, the amount of mRNA of several genes (NFE2L2, NQO1, and mt-DN5) and the number of mt-ND5 DNA copies were quantified in cultured GCs from prepubertal and adult females, either supplemented with UA or not. Our study confirmed the harmful effect of oocyte aging on the nuclear maturation progression, MMP, developmental competence and gene expression levels. UA treatment during in vitro maturation enhanced (p < 0.05) the maturation rate and subsequent developmental capacity of aged oocytes. A positive effect (p < 0.05) of UA on physiological maturation, MMP and embryonic development was also identified. UA also interfered on the expression profile of NFE2L2 and NQO1 genes in GCs cultures. Our findings demonstrate that UA supplementation is an effective way to prevent oocyte aging and improves the subsequent bovine embryonic development.

1. IntroductionDecline of female reproductive ability is one of the first physiological functions adversely affected by aging and thus considered as an emerging health problem worldwide [1,2]. In addition to several pathological problems, the age-associated decrease in female fertility is largely attributed to a decline in the ovarian reserve of oocytes [1,3,4,5] allied to a time-dependent deterioration of their quality [2]. This deterioration process can occur due to the exposure of oocytes to an aged ovarian microenvironment before ovulation. In addition, the female gamete is often subjected to post-ovulatory aging when the fertilization process does not occur within the best optimal span period, and the unfertilized oocyte remains in the oviduct or in vitro prior to insemination for extended periods [6]. The impairment of oocyte quality is a critical factor associated to the failure of assisted reproductive technologies (ART), since its quality is the main determinant for the embryo’s developmental potential after fertilization [7,8]. Ovulation asynchrony and aged oocytes were often reported to impair the success of artificial insemination and embryo production programs in the mare, cattle and sheep implying important economic losses [1,9,10]. Therefore, it is of primordial importance to study the mechanisms underlying oocyte aging, in order to design better therapeutic approaches to rescue fertility in several species, including humans, and also as a tool for genetic improvement in livestock. Particular attention must be devoted to improving the developmental capacity of oocytes from prepubertal cattle, which are often used to accelerate genetic gain and shorten generation intervals.One of the major causes of impaired developmental competence in aged oocytes is the increase in oxidative stress, which induces mitochondrial dysfunction, DNA damage and spindle formation errors, influencing the oocyte quality [11]. It is well established that increased production of free radicals is a cause of cellular aging in several chronic diseases and also in reproductive biology, resulting in poor fertility outcomes [12]. The ovarian microenvironment, which includes oocytes and granulosa cells (GCs), provides an antioxidant defense mechanism able to regulate oxidative conditions and to maintain the oxidant/antioxidant balance [13]. However, during the aging process, the efficiency of antioxidant defenses to neutralize reactive oxygen species (ROS) is attenuated, thus increasing the level of oxidative stress. The Nuclear factor-E2-related factor 2 (Nrf2 or NFE2L2), also known as Nrf2/Kelch-like ECH-associated protein 1 (Keap1) pathway, is a dominant response cascade activated by oxidative stress [14]. This pathway is a cellular defense mechanism that cells have developed to cope with deleterious effects of oxidative stress. Under normal conditions, Nrf2 is negatively regulated by Keap1, held in the cytoplasm and maintained at low levels. When exposed to oxidants, Nrf2 is dissociated from Keap1, allowing its translocation in the nucleus where it binds to specific DNA sequences. These sequences, named antioxidant response elements (ARE), lead to the transcriptional activation of cytoprotective genes, such as NAD(P)H:quinone-oxidoreductase-1 (NQO1), heme oxygenase-1 (HMOX1), and glutamate-cysteine ligase catalytic subunit (GCLC) [15,16]. Previous studies showed that the activation of Nrf2-Keap1 signaling pathway decreases the oxidative stress damage by elevating antioxidant levels in human GCs and mouse ovaries [17,18]. However, its role in the female gamete aging process remains elusive, although it is clearly established that mitochondria are the main production site for ROS during this process [9,10].Conversely, mitochondria are involved in several critical cellular functions and are fundamental for meeting the demand of energy production required during oocyte maturation and subsequent embryonic development [19]. Competent mitochondrial activity has been highly associated with higher contents of mitochondrial DNA (mt-DNA) and ATP generation [20], higher mitochondrial membrane potential [21] and maintenance of mitochondria quality and quantity through mitophagy [22]. Moreover, the mitochondrial activity has been suggested to be directly correlated with embryo viability and better fertility outcomes [23]. Increasing evidence supports that age-related mitochondrial alterations drive to ovarian aging and subsequently reduced embryo viability and implantation potential. These alterations include decreased mt-DNA copy number, decreased ATP generation [20], alterations in mitochondrial gene expression [24], mt-DNA damage [25] and reduced mitochondrial membrane potential [26].Mitochondria-targeted therapeutic approaches prompted a huge interest for several pathologies associated with aging due to their great potential in enhancing mitochondrial function [27]. Within this framework, Urolithin A (UA)—a natural metabolite obtained after the ingestion of food such as pomegranates followed by its conversion by the gut microbiota—has been demonstrated to prevent the accumulation of dysfunctional mitochondria with age, inducing mitophagy and also maintaining mitochondrial biogenesis and respiratory capacity [28,29]. UA has been applied as a promising therapeutic drug to prevent some cancers, such as colorectal and prostate cancer [30,31]. Additionally, UA also has anti-inflammatory [32], anti-obesity [33], antioxidant [34] and anti-aging properties [35]. A recent study has highlighted the effects of UA supplementation to senescent human skin fibroblasts on the activation of Nrf2-Keap1 pathway enhancing their antioxidant capacity. This activation of the Nrf2-Keap1 pathway effectively mitigates the ROS level, through the upregulation of the expression of Nrf2 downstream ARE-response genes (SOD, NQO1, GCLC and HMOX1), indicating a promising anti-aging effect [35].Several studies have been performed with the goal of delaying ovarian aging, consequently improving oocytes quality and the fertility outcomes. Due to the contribution of oxidative stress to the ovarian aging process, as well as to mitochondrial dysfunction, supplementation with antioxidants has appeared as a promising therapy [36,37]. However, it is unknown whether Urolithin A supplementation may restore the damage that occurs during ovarian aging contributing to prevent infertility problems. Therefore, the aim of this study was: (1) to demonstrate that aging could alter cumulus-oocyte-complexes’ (COCs) developmental potential and GCs’ expression of important genes involved in the Nrf2 signaling pathway; (2) to determine whether UA can rescue female fertility demonstrating an anti-aging effect in aged COCs and GCs; and (3) to evaluate UA effect on the expression level of genes involved in the Nrf2 signaling pathway as well as on oocyte quality.2. Materials and Methods2.1. Experimental DesignThis study was approved by the Animal Care Committee of the National Veterinary Authority (N°08965DGAV), following European Union guidelines (no. 86/609/EEC). To investigate the effect of aging on the alteration of oocyte quality and the potential anti-aging effect of UA, a model using COCs collected from prepubertal and adult cows submitted to in vitro aging (30 h of maturation) or to the physiological maturation (22 h) processes were applied.2.1.1. Previous Assay—Dose-Response StudyA previous assay to determine the concentration of UA that should be used during the bovine COCs maturation process was performed based on a dose-response study in four sessions. Since UA has never been tested in bovine oocytes, previous doses successfully applied for prevention and mitigation of some cancers and to demonstrate the anti-aging effect of UA in different cell lines were used [28,38,39]. COCs obtained from prepubertal and mature adult cows (n = 978) were selected and then randomly divided into five groups to test different doses of UA: control, 1, 10, 25, and 50 μM during physiological in vitro maturation. After the maturation period, some oocytes (n = 154) were stained to determine the chromosomal configuration and maturation stages. The remaining matured oocytes were submitted to in vitro insemination with frozen/thawed semen. Presumptive zygotes were cultured, and cleavage and blastocyst rates were determined at day 2 and day 7 of culture, respectively. Based on the obtained results, namely the absence of harmful effects and the promotion of maturation and blastocyst development, the concentration of 1 μM of UA was selected. 2.1.2. Experiment 1In this experiment, carried out in six sessions, both COCs from prepubertal (mean age = 9 months, n = 660) and adult (mean age = 39 months, n = 674) cows were collected to assess the oocyte quality and developmental potential of aged and physiologically matured oocytes as well as UA effect to rescue female fertility. COCs were randomly divided into 8 groups: (1) control prepubertal group, COCs from prepubertal calves matured for 22 h (n = 148); (2) UA prepubertal group, COCs from prepubertal calves matured in medium supplemented with 1 μM of UA for 22 h (n = 155); (3) control aged 30 h prepubertal group, COCs from prepubertal calves aged through 30 h of in vitro maturation (n = 149); (4) UA aged 30 h prepubertal group, COCs from prepubertal calves aged in vitro for 30 h in maturation medium supplemented with 1 μM of UA (n = 144); (5) control adult group, COCs from adult cows matured for 22 h (n = 155); (6) UA adult group, COCs from adult cows matured in medium supplemented with 1 μM of UA for 22 h (n = 129); (7) control aged 30 h adult group, COCs from adult cows aged through 30 h of in vitro maturation (n = 148); and (8) UA aged 30 h adult group, COCs from adult cows aged in vitro for 30 h in maturation medium supplemented with 1 μM of UA (n = 138). After the respective in vitro maturation periods, oocytes were inseminated with thawed capacitated bull semen. Subsequently, embryonic development was assessed, evaluating both the rate of cleaved and produced embryos, as well as their quality.Additionally, in this experiment, COCs from each group were retrieved to assess their nuclear maturation stage (control prepubertal group, n = 7; UA prepubertal group, n = 7; control aged 30 h prepubertal group, n = 10; UA aged 30 h prepubertal group, n = 6; control adult group, n = 16; UA adult group, n = 21; control aged 30 h adult group, n = 16; UA aged 30 h adult group, n = 18). The mitochondrial membrane potential (MMP) of COCs was also evaluated (control prepubertal group, n = 10; UA prepubertal group, n = 9; control aged 30 h prepubertal group, n = 9; UA aged 30 h prepubertal group, n = 9; control adult group, n = 15; UA adult group, n = 13; control aged 30 h adult group, n = 10; UA aged 30 h adult group, n = 14). 2.1.3. Experiment 2As the GCs play an essential role in follicular growth and oocyte development, a second experiment was performed in five sessions to further study the effect of age and UA on the expression of NFE2L2, NQO1 and mt-ND5. The number of copies of mt-ND5 gene was also evaluated. GCs were obtained after centrifugation of the follicular fluid aspirated from ovaries of prepubertal (mean age = 10 months) and adult cows (mean age = 62 months). These cells were cultured in the following conditions: (1) prebubertal control, culture of GCs of prepubertal calves; (2) prepubertal UA, culture of GCs of prepubertal calves supplemented with 1 μM UA; (3) adult control, culture of GCs of adult cows; and (4) adult UA, culture of GCs of adult cows supplemented with 1 μM UA. After GCs’ confluence at the 5th day of culture, they were snap frozen in liquid nitrogen and later the DNA and RNA were extracted, allowing the subsequent quantification of NFE2L2, NQO1 and mt-ND5 mRNA transcripts and also mt-ND5 copies number.2.2. Oocyte Collection and In Vitro MaturationOvaries from adult and prepubertal cows (previous assay, n = 978 and exp. 1, n = 1334) were collected at a local slaughterhouse, and kept at 35–37 °C, in a phosphate-buffered saline (PBS) supplemented with 0.15% of bovine serum albumin (w/v, BSA) supplemented with 0.05 mg mL−1 of kanamycin. At the laboratory, ovarian follicles with 2–8 mm in diameter were aspirated with a 19-gauge needle. Only COCs with at least three layers of compact cumulus cells and a homogeneous ooplasm were washed and selected for maturation according to the experimental design. Maturation was accomplished in an incubator at 38.8 °C, 5% CO2 in humidified air for 22 or 30 h in a maturation medium composed of tissue culture medium 199 (TCM) with 10% of fetal bovine serum, 0.2 mM sodium pyruvate, 10 ng mL−1 of epidermal growth factor, and 10 μL mL−1 of gentamicin [40].2.3. Granulosa Cells Collection and CultureGranulosa cells were obtained from the recovered follicular fluid after centrifugation for 10 min at 200× g [41]. The pellet was suspended in 1 mL of culture medium (TCM199 + 10% serum) to perform another centrifugation for 5 min. The new pellet was resuspended in 1 mL of culture medium either supplemented with 1 μM of UA or not according to the experimental design and homogenized with a syringe attached to a 19G-needle, at least 30 times to detach the cells. After evaluation of GC viability (tripan blue dye, 0.4% w/v), cells were seeded at a concentration of 2 × 105 viable cells mL−1 and cultured for five days at 38.8 °C, 5% CO2 in a humidified atmosphere until confluence. At every 48 h, the culture medium was discharged and refreshed with a new one. For DNA and RNA extraction, GCs were collected and washed by centrifugation at 200× g for 10 min. Cell pellets were resuspended in 1 mL of PBS, immediately snap frozen in liquid nitrogen, and stored at −80 °C.2.4. Oocyte Nuclear MaturationNuclear maturation stages were assessed following the 22 h or 30 h period of in vitro maturation. Denuded oocytes were fixed in an acetic acid/ethanol (1:3, v/v) solution, and maintained at 4 °C for 48 h. Then oocytes were stained with 1% aceto-lacmoid solution, mounted in a Neubauer chamber and observed under a phase contrast microscope (Olympus BX41). Oocytes were classified as follows: Germinal Vesicle (GV), Condensing Chromosomes I (CCI), Condensing Chromosomes II (CCII), Diakinesis, Anaphase-I/Telophase-I (AI/TI), and MII (Metaphase-II). Only oocytes with visible chromatin staining were taken into account [42].2.5. Assessment of Mitochondrial Membrane PotentialTo measure the mitochondrial membrane potential (MMP), an indicator of mitochondrial activity, mitochondria were stained with 5, 5′, 6, 6′-tetrachloro-1, 1′, 3, 3′-tetraethylbenzimidazolcarbocyanine iodide (JC-1, Invitrogen, Waltham, MA, USA). Denuded oocytes were incubated with 5 μg mL−1 of JC-1 [37] in maturation medium for 30 min at 38.8 °C and 5% CO2 in humidified air in the dark. Oocytes were washed twice in PBS and immediately transferred to a pre-heated slide glass and observed under a fluorescence microscope (Olympus BX51) using the blue fluorescence filter (BP 470–490, objective UPlanFI 20×/0.50). Mitochondrial membrane potential was then calculated as the ratio of the measured red/green fluorescence using the ImageJ software (National Institute of Health, Bethesda, MD, USA).2.6. In Vitro Fertilization and Embryo CultureIn vitro fertilization was performed with frozen-thawed sperm of a Holstein-Frisian bull, previously submitted to capacitation using the Percoll gradient (45 and 90) method, at a concentration of 2 × 106 spermatozoa mL−1. COCs and sperm were co-incubated for 20 h at 38.8 °C and 5% CO2 in humidified air. Presumptive zygotes were then transferred to droplets of synthetic oviductal fluid (SOF) medium supplemented with BME and MEM amino acids, glutamine, glutathione, and BSA [40]. After 48 h of the insemination, the cleavage rate (cleaved embryos per total inseminated oocytes) was assessed, and cleaved embryos were maintained in SOF supplemented with BSA and 10% of fetal bovine serum (FBS). Embryos were cultured for 12 days [41,43] to assess the blastocyst development rate (at days 7, 9, and 12; D7 embryos per cleaved embryos) and hatched embryo rate (hatched embryos per D7 embryos) and their quality [43]. Day 7 embryos were classified as grade 1 (good quality), 2 (fair quality), and grade 3 (bad quality) [40,44].2.7. DNA and RNA Extraction and QuantificationTotal DNA and RNA were isolated from GCs using the High Pure PCR Template Preparation Kit (Roche, Basel, Switzerland) and PureLink™ RNA Mini Kit (Invitrogen™, Waltham, MA, USA), respectively, according to the manufacturer’s instructions. Those protocols included the use of spin columns used to isolate high-quality total DNA and RNA, and DNase as treatment to remove genomic DNA from RNA [40]. After extraction, the samples were stored at −80 °C. The concentration and quality of DNA and RNA were determined using a NanoDrop™ One/OneC Spectrophotometer (ThermoFisher Scientific™, Waltham, MA, USA).2.7.1. Complementary DNA SynthesisSynthesis of complementary DNA (cDNA) from RNA isolates were performed using the Xpert cDNA Synthesis Mastermix kit (GRiSP, Porto, Portugal, according to the manufacturer’s instructions. RNA was reverse transcribed using 500 ng of extracted RNA from each sample to perform cDNA synthesis, which was carried out using a thermocycler (T100 Thermal Cycler, Bio-Rad, Hercules, CA, USA). The resultant cDNAs were stored at −20 °C until use for further assays.2.7.2. Primer DesignFor this study, primers for the targeted (NFE2L2 and NQO1) and an endogenous control gene (β-actin) were designed using the Primer BLAST software of the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/tools/primer-blast/, accessed on 6 March 2020). Sequences of primers for the reference genes and the target genes are depicted in Table 1. Additionally, the mt-ND5 gene was used in this work and details about mt-ND5 primers were retrieved from a previous study [45].2.7.3. Quantitative Reverse-Transcription Polymerase Chain ReactionReal-time PCR analyses were performed using the Xpert Fast SYBR Green Mastermix 2X with ROX in a QuantStudio 3 thermocycler (ThermoFisher Scientific™, Waltham, MA, USA), using cDNA at a concentration of 25 ng uL−1. The assessment of the mitochondrial DNA (mt-DNA) copies number of the ND5 gene was carried out by qPCR through the previously extracted DNA, using the same equipment as for the quantification of genes. Each optimized reaction was performed, consisting of Xpert Fast SYBR Green Mastermix 2X with ROX, primer (Forward and Reverse) of each target gene, sample (cDNA/DNA) and RNase free water making up a total volume of 10 μL. The samples were analyzed in duplicate and reactions containing water instead of template were included as negative controls. The samples were subjected to an amplification protocol that consisted of an initial cycle at 95 °C for 2 min of denaturation phase, followed by 40 denaturation cycles at 95 °C for 5 s, 40 annealing cycles for 30 s at 60 °C (depending on the melting temperature of primer sequences), and extension phase at 72 °C for 30 s and, lastly, a final extension period at 72 °C for 10 min. For the gene expression quantification, the relative quantification method was used. This method of relative quantification of gene expression was carried out with the expression levels of the target genes under study, which were normalized with the housekeeping genes, by the CT comparative method. As the amplification by RT-qPCR was performed in duplicate, the mean CT values for each gene were determined and the expression levels were calculated using the following formula:2(−ΔCt),(1) where ΔCt = Ct target gene − Ct endogenous control gene.For the quantification of the mt-DNA number of copies, the relative quantification normalized against unit mass method was used. This method was carried out with the CT values of GCs treated with UA (named as test), which were normalized with control samples (currently designated as calibrator). As the mt-ND5 copy number was assessed by qPCR and performed in duplicate, the mean CT values for the tests and calibrators samples were determined and the ratios were calculated using the following formula:Ratio = E(ΔCt),(2) where ΔCt = Ct calibrator − Ct test, and E is the efficiency.2.8. Statistical AnalysisData from embryo production and quality sessions were analyzed by using Proc Glimmix from SAS (Statistical Analysis Systems, SAS Inst., Inc., Cary, NC, USA), using the binary distribution and the logit as link function. The generalized linear mixed model included treatment (UA doses, aging effect and female age, and their interaction) as fixed effect and replicates as random effect. In addition, the means for each treatment were calculated, and comparisons between groups were performed using the PDIFF test. The JC-1 data, mRNA transcript levels and mt-DNA number of copies were analyzed using the Proc Mixed of SAS with a model including treatment (UA effect, aging effect, and/or female age, and their interaction) as fixed effect. The session was considered a random effect. The data of chromosomal configurations were compared between groups using the exact Fisher test, in a 2 × 2 contingency table. The analysis of results was considered statistically significant when p < 0.05.3. Results3.1. Previous Assay—Dose-Response StudyThe oocyte maturation status observed during the dose-response study of UA supplemented to the maturation medium is represented in Figure 1. This supplementation significantly influenced the nuclear maturation stages of oocytes. The 50 UA group (38.9 ± 0.5%) revealed a higher rate of oocytes in the AI/TI phase status compared to the Control (10.8 ± 0.5%, P = 0.03) and 10 UA (0.0 ± 0.0%, p = 0.0006) groups. In this phase, a higher rate in the 25 UA (17.2 ± 0.5%, p = 0.05) group compared to the 10 UA group was also identified. Moreover, a trend was found between 50 UA group and 1 UA (14.3 ± 2.0%, p = 0.08). The highest dose also showed a harmful effect, reducing the number of oocytes classified at the Metaphase-II (MII) stage (50 UA, 44.4 ± 2.0%) when compared to control (86.5 ± 7.0%, p = 0.003), 1 UA (82.1 ± 2.5%, p = 0.01), 10 UA (89.3 ± 6.5%, p = 0.002) and 25 UA (82.8 ± 9.0%, p = 0.01) groups. No differences were observed between the remaining stages of oocyte maturation. The GV and CCI maturation phases were not shown in the Figure 1, as no oocytes were assessed in these categories.The dose-response test showed no significant effect of UA concentrations on cleavage and hatched embryo rates. However, the rate of embryos produced at day 7 was influenced by the different concentrations of UA. The 50 UA group (7.2 ± 2.7%) had lower rates of embryos at day 7 compared to the 1 UA (28.5 ± 4.8%, p = 0.004), 10 UA (20.4 ± 4.0%, p = 0.03) and 25 UA (18.5 ± 4.7%, p = 0.05) groups. Although no significant differences were found between the control and 50 UA groups, a trend (p = 0.06) was identified. Moreover, day 7 embryo rates tend (p = 0.07) to be higher after 1 UA dose supplementation compared to control. Moreover, the 1 UA group doubled the number of excellent/good embryo quality (68.9% of grade 1 embryos, n = 20) when compared to control (34.6%, n = 10) and 10 UA (26.8%, n = 10). A low number of grade 1 embryos was obtained with the higher doses: 25 UA group (15.4%, n = 4) and 50 UA (35.4%, n = 3). Based on the above-mentioned results, the 1 μM UA was chosen to proceed to the following studies.3.2. Experiment 1—Oocyte Quality and Developmental Potential of Aged and Physiologically Matured Oocytes3.2.1. Nuclear MaturationWe investigated the effect of UA supplementation on chromosomal maturation status of aged and physiologically matured oocytes collected from prepubertal and adult females. Independently of the UA effect, oocyte aging and female age induced a significant harmful effect on maturation. Higher rates of MII phase were identified in control oocytes compared to aged ones (22 h = 93.2 ± 5.4% vs. 30 h = 77.6 ± 4.1%, p = 0.02). Moreover, a delay on maturation progression was identified on aged gametes and prepubertal females with more oocytes at AI/TI phase (22 h = 3.4 ± 0.5% vs. 30 h = 17.9 ± 0.6%, p = 0.01 and prepubertal = 21.1 ± 1.1% vs. adult = 6.8 ± 0.7%, p = 0.03). Moreover, a higher number of degenerated oocytes after 30 h of IVM was identified compared to after 22 h (7.1 ± 1.39% vs. 4.1 ± 0.89%, p = 0.02). The female age and UA supplementation did not interfere (p > 0.05) with the number of degenerated oocytes.Both in vitro aging and UA supplementation to the maturation medium significantly influenced the chromosomal configuration of oocytes from prepubertal and adult females (p < 0.05, Figure 2). On the AI/TI phase of the maturation status, higher rates of oocytes were found in this stage in the control aged 30 h prepubertal group (46.7 ± 0.5%) when compared to control prepubertal (0.0 ± 0.0%, p = 0.05) and UA aged 30 h prepubertal (0.0 ± 0.0%, p = 0.02), and both adult control (5.6 ± 0.5%, p = 0.01 and UA adult, 0.0 ± 0.0%, p = 0.0002) and control aged 30 h adult (9.1 ± 1.0%, p = 0.02) groups. A trend was also identified between UA adult and UA aged 30 h adult (p = 0.0769).Regarding the MII phase corresponding to a complete nuclear maturation, significant differences were also found between groups (Figure 2). Lower rates of oocytes on the MII stage were found in the control aged 30 h prepubertal (53.3 ± 2.0%) compared to the control prepubertal (100.0 ± 1.5%, p = 0.05), UA aged 30 h prepubertal (100.0 ± 2.5%, p = 0.02) and UA adult (100.0 ± 10.5%, p = 0.0002) and control aged 30 h adult (86.0 ± 5.5%, p = 0.056) groups. Moreover, the UA adult group showed higher rates of matured oocytes (MII) compared to the UA aged 30 h adult (76.2 ± 11.3%, p = 0.01), control aged 30 h adult (86.0 ± 5.5%, p = 0.08) and control adult (83.3 ± 4.5%, p = 0.057).No differences were observed between the remaining stages of oocyte maturation.In summary, UA supplementation improved in vitro maturation progression in oocytes submitted to the aging process, especially in prepubertal females. An important anti-aging effect of UA was thus identified.3.2.2. Mitochondrial Membrane PotentialIn order to analyze the involvement of mitochondrial dysfunction in female aged oocytes and the effect of UA and maternal age, the mitochondrial membrane potential (MMP) was assessed in oocytes of different groups.Independently of the female age and UA effects, oocyte aging had a significant harmful effect on MMP assessed as the ratio of the measured red and green fluorescence (22 h ratio = 0.54 ± 1.8% vs. 30 h ratio = 0.47 ± 1.5%, p = 0.002).The combination of the female age, aging effect and the supplementation with UA during maturation significantly (p = 0.007) influenced the MMP (Figure 3). As observed in Figure 3, higher rates of MMP were obtained in the UA prepubertal (ratio = 0.55 ± 0.04) and UA adult (ratio = 0.63 ± 0.04) groups where oocytes were matured for 22 h, compared with the control aged 30 h adult (ratio = 0.45 ± 0.04, p = 0.04 and p = 0.0007, respectively) and UA aged 30 h adult (ratio = 0.45 ± 0.03, p = 0.03 and p = 0.0003, respectively) groups. Moreover, a significant increase in JC-1 aggregate/monomers ratio, which indicates a significant increase in MMP, was observed in the UA adult group when compared with control adult (ratio = 0.47 ± 0.04, p = 0.0031), control prepubertal (ratio = 0.52 ± 0.04, p = 0.03), control aged 30 h adult (ratio = 0.48 ± 0.03, p = 0.003) and UA aged 30 h adult (ratio = 0.47 ± 0.03, p = 0.001) groups. A trend was also identified between UA prepubertal, and UA aged 30 h prepubertal (p = 0.07).In summary, UA improved the MMP during the physiological maturation but was not capable of overcoming the negative effect of aging in both prepubertal and adult oocytes.3.2.3. Embryonic DevelopmentThe embryonic developmental potential was evaluated through cleavage, D7 and hatched embryo rates, obtained with the different treatments (aging and UA supplementation) applied to the prepubertal and adult females COCs during maturation. Independently of the female age and UA effects, oocyte aging had a significant harmful effect on cleavage (22 h = 79.5 ± 1.7% vs. 30 h = 68.6 ± 2.0%, p = 0.0001) and D7 embryo rates (22 h = 18.6 ± 1.9% vs. 30 h = 12.1 ± 1.7%, p = 0.01). The UA supplementation was also shown to improve the D7 embryo rates (control = 11.1 ± 1.5% vs. UA = 20.0 ± 2.0%, p = 0.0009).The combination of the female age, aging effect and the supplementation with UA during maturation significantly (p = 0.01) influenced the embryonic development, namely the cleavage and D7 embryo rates (Table 2). Higher rates of cleavage were achieved when the adult oocytes were matured for 22 h with or without UA supplementation (control adult = 1.4 ± 3.2% and UA adult = 80.9 ± 3.4%) and prepubertal oocytes matured for 22 h with UA supplementation (UA prepubertal = 80.6 ± 3.2%), compared to UA aged 30 h adult (66.5 ± 4.2%, p ≤ 0.01), control aged 30 h prepubertal (66.9 ± 3.9%, p ≤ 0.02) and UA aged 30 h prepubertal (67.2 ± 4.0%, p ≤ 0.01) groups (Table 2). A trend (p = 0.09) was also identified between the control adult and the control aged 30 h adult groups.Another parameter that was significantly (p = 0.01) influenced by the combination of the female age, the supplementation with UA and COCs aging was the rate of embryos produced at day 7 (Table 2). Exception made for the UA prepubertal and UA 30 h prepubertal groups, the UA adult group presented the highest rates of embryo at day 7 (26.8 ± 4.3%, p ≤ 0.05). Moreover, lower D7 embryo rates were produced from aged oocytes from both adult and prepubertal females without the supplementation of UA (control aged 30 h prepubertal = 7.7 ± 2.6% and control aged 30 h adult = 7.7 ± 2.6%) compared to those that were supplemented with UA, respectively (UA aged 30 h prepubertal = 18.9 ± 4.0%, p = 0.03 and UA aged 30 h adult = 15.9 ± 4.0%, p = 0.09). The supplementation of UA to the maturation medium did not significantly (p > 0.05) influence the quality rates of produced embryos of grade 1, 2 and 3, nor did the aging effect or all the studied effects together (Table 3). However, the number of embryos of excellent/good quality that were produced when the UA were supplemented to oocytes matured for 22 h doubled (Table 3).3.3. Experiment 23.3.1. Gene Expression LevelsIn order to study the UA effect and maternal age influence in the Nrf2 signaling pathway, analysis of NFE2L2 and NQO1 genes expression levels in GCs culture from prepubertal and adult cows was assessed by RT-qPCR. Independently of the other studied factors, female age significantly influenced the NFE2L2 gene expression, reflected in the higher level of NFE2L2 transcripts in GCs from prepubertal females (mRNA levels prepubertal = 0.00015 ± 0.0012% vs. adult = 0.00011 ± 0.0012, p = 0.048).The supplementation of UA to the culture medium of GCs, both independently and considering female age, did not significantly (p > 0.05) influence the gene expression levels of NFE2L2, NQO1 and mt-ND5 genes. Figure 4 represents the expression levels of these genes that were normalized with the respective control, in both UA treated prepubertal and adult GCs.3.3.2. mt-DNA Copy NumberAnalysis of mt-ND5 DNA content in GC culture from prepubertal and adult cows were assessed by qPCR. The supplementation of UA to the culture medium of GCs, both independently and considering female age, did not significantly (p > 0.05) influence the copy number of mt-ND5 gene. Figure 5 represents the copy number of this gene normalized with the respective controls, in both UA treated prepubertal and adult GCs.4. DiscussionOver the past decades, an increasing demand for ART application in livestock has been observed [8,46,47]. Additionally, delayed childbearing associated to advanced maternal age has now become the main factor leading women to resort to ART [5]. Oocyte quality is critical for the occurrence of a successful conception, and aging of the female gamete a major concern for ART success. Despite the increasing advances made in this field, the applied technologies are not yet able to restore fertility, reverting the biological clock [7]. In the present work, a model for the study of age-associated infertility in cattle oocytes was developed. Due to the similarities to human pregnancy, follicular and endocrine events [48,49], this model may also be useful for human research, avoiding the ethical and physical restrictions that hamper these studies in human oocytes. Our model for aging female gametes proved to be very efficient and suitable for studying different problems associated with reproductive aging and fertility impairment, currently one of most critical challenges in the world. Furthermore, the search for new upcoming antioxidant therapies with the potential to prevent infertility provoked by the female gamete aging, as studied in the present work, is equally of the utmost importance.Presented results reported for the first time the beneficial effect of UA in ART outcomes. UA is a food metabolite capable of preventing the accumulation of age-related dysfunctional mitochondria by inducing their mitophagy and an extended lifespan of cells [28]. Since UA has never been tested before in bovine reproduction, doses successfully applied in different cell lines showing its anti-cancer, anti-inflammatory and anti-aging effects were used [28,35,39]. In our study, a previous dose-response assay was carried out and the concentration of 1 µM UA was clearly identified as the most promising. Moreover, a deleterious effect was demonstrated at higher doses, especially at the 50 µM UA dose, harming the progression of nuclear maturation until MII. This effect was reflected on embryonic development impairment. Accordingly, Liu and colleagues (2019) noticed significant reduced cell viability and proliferation of human senescent skin fibroblasts, at a UA concentration of 50 µM. They referred that high UA doses lead to diminished cell viability increasing the number of cells arrested in the G/M cell-cycle [35]. On the contrary, higher rates of embryos were produced at day 7 after the supplementation of 1 µM UA to the maturation medium when compared to the other doses. To further explore the potential mechanisms of the action of UA in female reproduction, a study during oocyte aging and using females with different ages as oocyte donors was implemented. Yamamoto and co-workers (2010) reported an age-associated decline in the fertilization rate of old cows, showing that these oocytes were more prone to resume first meiotic division during maturation and often had already initiated the meiotic maturation at oocyte collection [50]. These data suggested that oocytes from older cows had a faster nuclear maturation progression and reached the MII phase faster than oocytes from young females due to a lower oocyte competence [51]. Accordingly, our study reveals that female donors’ age and the process of oocyte aging significantly influence the chromosomal configuration of oocytes during maturation progress. In fact, oocytes from prepubertal cows showed a higher rate of delayed stages, such as AI/TI phases (prepubertal = 21.1% vs. adult = 6.8%), revealing a slower progression of nuclear maturation, which may be due to lower oocyte competence as proposed by Soto-Heras and colleagues (2018). Aged oocytes also present lower rates of oocytes that have reached the MII phase during a 30 h period (77.6%) compared to the physiological period (93.2%). Moreover, a positive effect has been identified with UA treatment during the physiological maturation process, and also an anti-aging effect in both prepubertal and adult females.In agreement with our findings, a study has demonstrated that melatonin supplementation during the in vitro maturation could stimulate the meiosis resumption in bovine COCs, whereas control oocytes cultured without hormones had slower meiosis resumption rates [52]. On the contrary, the supplementation with other antioxidant agents, such as quercetin, vitamin C or resveratrol did not present any effect on nuclear maturation rates, even when a reduced ROS levels or increased antioxidant enzymatic levels were observed [53,54]. Conversely, our results show that UA could rescue oocytes from aging effects in both aged adult and prepubertal females, which present lower competence, improving maturation rates.One of the main contributors to poor fertility outcomes affecting oocyte quality is related to mitochondrial functions, which become compromised with advanced maternal age and post-ovulatory aging [21,55]. The MMP have been widely studied in different models, revealing that both aged gametes after ovulation and maternal aging, induce the loss of mitochondrial function [56]. Consequently, the loss of MMP was negatively reflected on oocyte and embryo development [26,57]. According to these findings, in our study, a reduced MMP ratio was found in aged oocytes from both prepubertal and adult cows. Moreover, UA supplementation to the culture medium induced an increase on MMP of prepubertal and adult oocytes matured for 22 h, reverberating in higher cleavage rates. These data are in agreement with the results found by Liang and co-workers (2017) which observed a MMP enhance when bovine oocytes were supplemented with melatonin for 22 h [26]. However, we also identified a reduction in MMP in aged oocytes supplemented with UA, denoting that UA may not be able to reverse the negative effect of aging in MMP. Discrepant results have been observed concerning the MMP after supplementation of aged oocytes with antioxidant compounds. Indeed, several authors reported greater levels of MMP on oocytes after supplementation with melatonin [26] and laminarin [21], whereas others have observed the opposite, a decreased MMP in aged oocytes treated with L-carnitine [58] and melatonin [38]. Additionally, Ryu and colleagues (2016) showed a reduction in MMP in mice myoblasts cultured with UA [28]. Regarding our results, further studies should be addressed to deepen the knowledge of the mechanism of action of UA in the oocyte mitochondria and explain the identified differential effects on aged and physiologically matured oocytes.Intrinsic quality of oocytes has been widely reported as the main determinant of subsequent embryonic development. Accumulated evidence revealed that several cellular and molecular abnormalities occur during extended in vitro maturation periods as well as during in vivo post-ovulatory aging [57]. Furthermore, these abnormalities can exert relevant impacts on oocyte quality reverberating on embryo production [52]. To demonstrate that UA can act as an anti-aging compound and improve oocyte quality, delaying oocyte aging, we investigated the developmental capacity of aged oocytes after in vitro fertilization. In our study, a significant (p ≤ 0.01) harmful effect of gamete aging on cleavage and day 7 embryo production rates were observed. Accordingly, several authors referred that old females have an age-associated decline in reproductive capacities, reflected in lower fertility rates and poor embryo quality [44,55]. On the other hand, previous studies testing other antioxidant molecules to rescue aged oocytes have reported the beneficial effect of a few of these compounds during in vitro maturation on the embryo development of aged females from different species, such as cattle [58], pig [38] and mice [6]. For instance, L-carnitine was tested on bovine aged oocytes and no significant differences were found on the obtained cleavage rates. However, a significant increase in the rate of zygotes developed to the blastocyst stage was identified, compared to aged oocytes without L-carnitine [59]. Our results also demonstrate that UA supplementation during physiological maturation improved the cleavage rate in prepubertal and adult females. Although the cleavage rates of UA aged oocytes were not significantly different from control aged oocytes, higher D7 embryo rates were identified in UA aged oocytes from both prepubertal and adult females. These results pointed out that the anti-aging effect of UA previously identified in different cell cultures [28,35] is valid for oocytes and embryos. Oocyte aging is a multifactorial process that impairs the development of the embryo, and restoring the developmental capacity of aged oocytes is an important objective that was attained in the present study.Additionally, we also assessed the preventing role of UA in the age-related deterioration on embryo quality. Although in this study no significant differences were found on embryo quality rates, the oocytes supplemented with UA increased the number of transferrable embryos, compared to the untreated ones. A previous study reported that L-carnitine treatment could improve the quality of embryos developed from aged bovine oocytes through the reduction in ROS levels and higher levels of glutathione as well as of others antioxidant enzymes [58]. The differences between these results and ours may be due to the different techniques applied to assess embryo quality. The morphological evaluation of embryo quality remains a subjective method that depends on the observer’s experience. In the future, more studies should be addressed using other techniques to accurately assess embryo quality.It is widely known that oxidative stress plays a crucial role in the age-associated decrease in fertility. Oxidative stress is one of the major contributors to low oocyte maturation efficiency, and oocyte quality deterioration, thereby impairing subsequent embryo development [11]. Intercellular communication between the gamete and somatic cells is crucial for the proper development of high-quality oocytes. Changes in the microenvironment of aged ovaries have been reported, such as decreased antioxidant enzymatic activity leading to an impaired ROS scavenging efficiency. Furthermore, GCs from young and older females were shown to have differentially expressed genes associated to antioxidant activities and maternal age [2,13]. Thus, it is of great importance that found mechanisms manage oxidative stress in order to rescue oocyte from aging and to overcome infertility issues. The Nrf2-Keap1 pathway has been extensively studied due to its capacity to cope with the deleterious effects of oxidative stress and exerting antioxidant proprieties [60]. To further study the activation of the Nrf2 signaling pathway in bovine GCs, we assessed the mRNA expression level of NFE2L2 and its downstream antioxidant (NQO1). Our results showed a significant influence of female age on the level of NFE2L2 transcripts, which decreases with age. Indeed, greater mRNA expression levels were observed in prepubertal cows when compared to adults. This is in agreement with previous studies that reported a highly expressed level of Nrf2 protein and mRNA on ovarian tissues of childbearing young mice and women from 22 and 49 years old, whereas in aged mice and women a lower expression was found. It was suggested that decreased expression of Nrf2 may be involved in the decline of reproductive capacity of older women and its control may have important implications in delaying ovarian aging [15,61]. Furthermore, Akino and co-workers (2019) showed that the activation of the Nrf2-Keap1 pathway through the administration of dimethylfumarate could reduce ROS levels and lead to delayed infertility [18]. Similar results regarding the effect of UA in reducing ROS in senescent human skin fibroblasts were reported. When these cells were treated with UA, a significant increase in the mRNA expression of Nrf2 targeted genes, such as SOD1, NQO1, GCLC and HMOX1 were reported. [35]. In our study, we observed that the expression level of NFE2L2 and NQO1 genes in GCs was not significantly affected by UA supplementation. However, further studies should be addressed to confirm the reduction in ROS and subsequent improvement of produced blastocyst, as well as the UA effect on mRNA expression of the aforementioned genes on aged oocytes.In addition to the damage to mitochondrial DNA (mt-DNA) [25,57], the disruption of mitochondrial gene expression has also been shown to contribute to mitochondrial dysfunction with age. Zhang and colleagues (2019) reported that some genes involved in the OXPHOS, namely mt-ND2, mt-ND3, mt-ND4, mt-ND4L and mt-ND5, were significantly downregulated in the GV stage of oocytes from aged mice, compared with those from young mice [62]. In our study, a no significant difference was found in the mt-ND5 mRNA expression, between the GCs retrieved from prepubertal and adult cows. Moreover, we also assessed the mt-DNA copy number, because the content of mt-DNA in GCs has been positively associated with oocyte quality and embryo development [63,64]. Although in our study no significant differences were identified between the mt-DNA copy number in GCs from prepubertal and adult cows, or when treated with UA, we observed a greater number of oocytes that developed to the blastocyst stage, compared with the untreated groups. Additionally, Ryu and co-workers (2016) reported that the mt-DNA content and protein level from the respiratory complexes did not change in mice myoblasts supplemented with UA [28]. Further studies should be performed to disclose the mechanism of action of UA in improving female fertility.5. ConclusionsThe results obtained in this study confirmed the harmful effect of oocyte aging on its developmental competence. Moreover, our model for aging female gametes proved to be very efficient and useful to study different problems associated with reproductive aging and consequent fertility impairment. Additionally, UA supplementation during the maturation process of aged oocytes improved maturation rates and produced embryos. Therefore, an anti-aging effect of UA in rescuing aged gametes was identified for the first time, improving the blastocyst development, which leads to an increased number of embryos for transfer to recipient females. A positive effect of UA on physiological maturation, MMP and embryonic development was also identified. In conclusion, UA treatment provides a new therapeutic approach to prevent or delay gamete aging, and improve the subsequent blastocyst formation and fertility outcomes in ART.

animals : an open access journal from mdpi

10.3390/ani11113073

PMC8614315

Reductions in the fertility, body weight, and growth rate of cattle across the world are associated with the global warming phenomenon. Developing optimal management strategies is an important aspect of breeding programs for different breeds. Blood tissue undergoes dramatic physiological and metabolic changes during heat stress conditions, which involves the expression and regulation of a great number of genes across species. Real-time quantitative PCR (qPCR) is a method for the rapid and reliable quantification of mRNA transcription. Reference genes are used to normalize mRNA levels between different samples. Thus, the selection of high-quality reference genes is necessary for the interpretation of data generated by real-time PCR.

Real-time PCR is widely used to study the relative abundance of mRNA due to its specificity, sensitivity, and repeatability quantification. However, relative quantification requires a reference gene, which should be stable in its expression, showing lower variation by experimental conditions or tissues. The aim of this study was to evaluate the stability of the expression of five commonly used reference genes (actb, ywhaz, b2m, sdha, and 18s rRNA) at different physiological stages (alert and emergency) in three different cattle breeds. In this study, five genes (actb, ywhaz, b2m, sdha, and 18s rRNA) were selected as candidate reference genes for expression studies in the whole blood from three cattle breeds (Romosinuano, Gyr, and Brahman) under heat stress conditions. The transcription stability of the candidate reference genes was evaluated using geNorm and NormFinder. The results showed that actb, 18SrRNA, and b2m expression were the most stable reference genes for whole blood of Gyr and Brahman breeds under two states of livestock weather safety (alert and emergency). Meanwhile, actb, b2m, and ywhaz were the most stable reference genes for the Romosinuano breed.

1. IntroductionHeat stress is a physiological condition that occurs when an animal cannot dissipate body heat, leading to an increase in body temperature [1]. In livestock production, the heat stress decreases body weight, average daily gain, growth rate, fat thickness, meat quality, and milk production [2]. Cattle exposed to high temperatures also exhibit alterations in folliculogenesis and oocyte viability [3]. Additionally, heat stress decreases pregnancy rates and embryonic development in embryos produced in vivo and in vitro [4]. Due to heat stress effects, humans have reevaluated management decisions regarding which animals to use for food production [5]. In this way, breeds that originated in warm climates such as African zebu (Bos primigenius indicus) and African taurus (Bos taurus africanus) show adaptive advantages to heat stress compared with breeds that originated in temperate areas such as European taurus (Bos taurus taurus) [5,6].The heat stress in cows can be evaluated through the change in behavior and physiological variables such as respiratory rate, heart rate, and vasodilation [7]. Furthermore, the quantification of gene expression for conserved proteins that increase their expression under heat stress conditions allows them to be used as a reference to evaluate the stress of an individual [8]. The qPCR technique allows the quantification of gene transcript expression [9]. Relative quantification requires a reference gene, which should be stable in its expression and show lower variation by experimental conditions or tissues [10]. Initially, highly conserved genes that code for proteins involved in functional processes and the structure of cells were chosen, which were previously called housekeeping genes [11]. The use of these reference genes for qPCR data normalization may have solved problems that could affect the quantification, such as the concentration variability of RNA and inhibitors from the extraction protocols [12]. Likewise, the use of reference genes as endogenous controls in the relative quantification can allow the correction of the sample variations [13]. However, it has shown that the gene expression can be variable in some experimental conditions, and it has been necessary to validate the stability of these genes in different conditions [14].In cattle, the expression stability of several references genes such as actin beta—actb, glyceraldehyde-3-phosphate dehydrogenase—gapdh, succinate dehydrogenase—sdha, tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta—ywhaz, TATA-box binding protein—TBP, beta-D-glucuronidase—GUSB, H2A clustered histone 14—H2AC14-, peptidylprolyl isomerase A—PPIA, ribosomal protein L15—RPL15, battenin—CLN3, eukaryotic translation initiation factor 3 subunit K—EIF3K in the bovine liver, kidney, pituitary gland, thyroid gland, muscle, and mammary gland have been reported [15,16]. In livestock production, despite the use of temperature and humidity index in the control of heat stress in cattle, the appropriate reference genes for cattle under heat stress conditions are still clear. Due to qualities such as accessible source of systemic information of the transcriptome that allows measuring changes in relevant biological processes and pathways, blood samples are considered good samples [17]. In the present paper, the expression stability of five reference genes (actb, ywhaz, b2m, sdha, and 18S rRNA) in whole blood from Romosinuano, Gyr, and Brahman cattle breeds collected under two states of livestock weather safety (alert and emergency) were evaluated.2. Materials and Methods2.1. Ethics StatementAll procedures involving animals were approved by the Ethics committee of the University of Tolima based on the Law 84/1989 and the Resolution 8430/1993 and complied with the guidelines for animal care and use in research and teaching [18,19].2.2. Study PopulationHealthy cows of Brahman (n = 10), Gyr (n = 10), and Romosinuano (n = 10) breeds (age between 48 and 96 months) were located on a farm near to Monteria city, Cordoba department at northern region of Colombia, (Latitude 8°45′36″ N and Longitude 75°53′08″ W), between April and November of 2020, with an average temperature of 29 °C and relative humidity between 70 and 85%.2.3. Weather DataAmbient temperature (°C) and relative humidity measured as a percentage for each hour throughout the study was measured using a PCE-FWS20N weather station (PCE Instruments™, Meschede, Germany). The temperature-humidity index (THI) was calculated for each hour applying the National Research Council (1971) formula as follows: (1)THI=(1.8×Tdb+32)−[(0.55–0.0055×RH)×(1.8×Tdb−26)THI data were used to identify two categories of livestock weather safety index (alert and emergency) [20]. In our study period, an alert condition period was identified from 21:00 to 08:00 h with THI values of 75 to 78, and an emergency state was identified from 13:00 to 14:00 h with THI values of 84 to 86.1. Therefore, the blood samples for the gene expression analysis were taken at 7:00 h with a THI value of 76.3 (alert state) and 14:00 h with a THI value of 86.1 (emergency state).2.4. Samples, RNA Extraction, and cDNA SynthesisBlood samples were obtained by venipuncture of the caudalis medium vein, transferred into 4 mL EDTA tubes (Becton Dickinson Vacutainer Systems, Franklin Lakes, NJ, USA), and collected twice daily at 7:00 h and 14:00 h. Immediately after sample collection, blood samples were divided into small volume aliquots of 2 mL in Graduated Safelock Microcentrifuge Tubes. Later, all blood samples were frozen in liquid nitrogen and stored at −20 °C until experimental analysis.RNA was extracted from blood samples using the RNA-solv reagent kit (OMEGA, Norcross, GA, USA) according to the manufacturer’s protocol with certain modifications. The modified RNA extraction protocol consisted of 1000 μL of RNA-Solv® reagent (OMEGA, Norcross, GA, USA), which was mixed with 200 μL of whole blood. The mixture (sample and RNA-Solv® reagent) was homogenized in a vortex (30 s); then, 200 μL of chloroform (J.T.Baker®, Radnor, PA, USA) at −20 °C were added, vortexed (30 s), and incubated at 4 °C for 5 min. The mixture was centrifuged at 12,000 rpm for 15 min at 4 °C, and the aqueous phase was transferred to a clean tube. For the precipitation stage, 2 volumes of isopropanol were added to the recovered aqueous phase and mixed by inversion (6 times) followed by incubation at 4 °C for 30 min. Later, centrifugation was performed at 12,000 rpm for 10 min at 4 °C to obtain a pellet, which was washed twice as follows: 1 mL of 75% ethanol (Merck, Darmstadt, Germany), centrifugation at 12,000 rpm during 10 min at 4 °C, and discarding the supernatant. Finally, the pellet was dried for 5 min at room temperature and dissolved in DEPC water (21 μL); afterwards, RNA quality was measured by spectrophotometry with the NanoDrop One (Thermo Scientific, Wilmington, DE, USA), and the pellet was stored at −20 °C.Prior to reverse transcription, all RNA samples were diluted to 200 ng/μL, and cDNA was synthesized using GoScriptTM Reverse Transcription System kit (Promega, Madison, WI, USA) following the manufacturer’s instructions. End-point PCR and agarose gel electrophoresis were conducted to determine the cDNA quality and the amplicon size.2.5. Gene Selection and Primer DesignFive reference genes, actb, 18S rRNA, b2m, ywhaz, and sdha were selected as candidate reference genes for this study based on previous reports [15,16]. Primers were designed based on sequences from Bos taurus and Bos indicus using Geneious Prime software v2021.1 [21] (Table 1).2.6. End-Point PCR and Quantitative Polymerase Chain Reaction (qPCR)All primers were examined for their target specificity by end-point PCR with a total volume of 25 µL, composed of 14.8 µL of distilled–deionized water, 5 µL of 5X green GoTaq® Flexi Buffer (Promega, Madison, WI, USA), 1 µL of dNTPs (1.5 mM) (Invitrogen, Carlsbad, CA, USA), 1 µL of each primer (forward and reverse) (10 pmol/µL), 1 µL MgCl2 (25 mM), 0.125 µL of GoTaq® Flexi DNA polymerase (Promega, Madison, WI, USA), and 1 µL of the cDNA as template. The amplification was carried out in a ProFlexTM PCR System (Applied Biosystems, Carlsbad, CA, USA) with an initial denaturation step at 95 °C for 3 min, which was followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at the specific annealing temperature for each set of primers (Table 1) for 30 s, extension at 72 °C for 30 s, and a last step of final extension at 72 °C for 5 min. Amplicons were revealed on 1% agarose gel by electrophoresis (PowerPac™ HC, Bio-Rad, Hercules, CA, USA) using a GeneRuler 100 bp DNA Ladder (Thermo Fisher Scientific, Waltham, MA, USA). The gel was stained with HydraGreen™ (ACTGene, Piscataway, NJ, USA) and visualized under UV light, using the ENDUROTM GDS gel documentation system (Labnet International, Inc., Woodbridge, NJ, USA).Relative gene expression of b2m, sdha, ywhaz, actb, and 18S rRNA genes was measured by qPCR using a Luna® Universal qPCR Master Mix (New England BioLabs Inc., Beverly, MA, USA) in a QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA), by the Fast ramp program. Thermal cycling conditions were initial denaturation 1 min at 95 °C; then, 40 cycles of denaturation for 15 s at 95 °C and annealing for 30 s at 60 °C. Subsequently, a melting step was performed at 95 °C for 1 s, 60 °C for 20 s, and a continuous rise in temperature to 95 °C at a rate of 0.15 °C per second. Each sample was run in triplicate.2.7. Analysis of Reference Gene Expression StabilityExpression levels of the tested reference genes were quantified by the quantification cycle (Cq) values obtained through qPCR from the three technical replicates, averaged, and used as input data on NormFinder and geNorm to evaluate the gene expression stability [14,23,24].3. Results3.1. Primer SpecificityFive reference genes for Bos species were chosen for this study based on previous reports [15,16] (Table 1). All the primers designed for the reference genes were specific through evaluation by end-point PCR and qPCR; as shown in Figure 1, the qPCR melting curves showed a single peak, suggesting that there was no formation of primer dimers or nonspecific PCR products.3.2. Expression Profiles of Reference GenesAs shown in Figure 2, the Cq values of the five reference genes from blood samples among Brahman, Gyr, and Romosinuano breeds ranged between 15.86 and 35.61. 18SrRNA was the most highly expressed gene, with Cq values ranging between 15.86 and 26.8, followed by b2m, sdha, and actb, which showed Cq values from 19.77 to 26.72, 19.68 to 30.03, and from 20.50 to 29.23, respectively. In addition, ywhaz exhibited Cq values from 25.50 to 35.61 (Figure 2).3.3. Reference Gene Stability: geNormThe expression stability of the reference genes in terms of M values was analyzed using geNorm software. As shown in Figure 3, the stability ranking of the five reference genes was different among bovine breeds. However, all reference genes had an M value below 1.5, which is the recommended geNorm (the most stable reference genes have the lowest M values), and the b2m gene was the most stable gene (Figure 3).3.4. Reference Gene Stability: NormFinderThe reference gene stability value was calculated for each gene using NormFinder software, indicating that those with the lowest stability values are the most stable genes. NormFinder identified actb and b2m as the two most stable genes with stability values of 0.016 and 0.021 respectively, in contrast with sdha gene (Table 2) with values of 0.029 to 0.043.4. DiscussionThe real-time PCR is a powerful tool for evaluating mRNA levels due to its specificity, sensitivity, and repeatability quantification [25,26,27]. However, when the expression of the target gene is analyzed by this method, there are unavoidable operational errors; e.g., in the absolute expression level, the same target gene can display significant errors between different biological groups or technical repetitions [28]. This is unlike relative quantification, where the RNA transcription level is normalized based on the expression level of the internal reference gene [29]. The ideal reference gene should be stably expressed, and its expression should not be affected by the experimental conditions [30]. Numerous studies have demonstrated that the expression of commonly used reference genes varies among different cell types, tissues, and experimental conditions; for example, actb and gapdh, which are largely accepted, can show large variations in expression [31,32]. Thus, the selection and validation of reliable reference genes for each particular condition are essential to quantitative accuracy [33].Several studies have been conducted to assess the reference genes in specific tissues in numerous species [34]. In cattle, De Ketelaere et al. (2006) selected sdha, ywhaz, and 18S rRNA as being the most stable genes for the accurate normalization of qPCR of bovine polymorphonuclear leukocytes [35]. Likewise, sdha has also been ranked greatest in terms of expression stability in bovine neutrophils [36]. However, the reference genes mentioned previously were described for different experimental conditions. In the present study, two statistical methods (geNorm and NormFinder) were used to evaluate the gene expression stability of five reference genes (actb, ywhaz, b2m, sdha, 18SrRNA) in the whole blood of three cattle breeds under two states of livestock weather safety. The Temperature–Humidity Index has been widely used to alert cattle producers of potential weather-based heat stress; for example, some recommendations for mitigating heat stress are based on estimating THI values [20,37,38]. In the present study, two states of heat stress (alert THI = 70–80 and emergency THI ≥ 84) in cattle were chosen for blood sample collection due to the expected cellular stress responses in these states [5,17,37,39].The geNorm and NormFinder software were used to evaluate the stability of the reference genes. The geNorm method calculates the gene stability value (M) by computing pairwise comparisons and geometric averaging of each reference gene under different experimental conditions, where genes with the smallest M values below 1.5 are considered excellent constitutive genes [14]. On the other hand, the NormFinder method assesses gene expression stability (Stability Value, SV) based on parameters of the estimates for both intragroup and intergroup variations of each gene [23]. Based on the geNorm program, the most stable reference genes were actb, 18SrRNA, and b2m for Brahman, and those for the Gyr breeds were b2m, 18SrRNA, and actb. Whereas for Romosinuano, b2m, actb, and ywhaz were the most stable genes (Figure 3). The stability ranking of the reference genes presented here is consistent with previous studies [40,41] According to the stability ranking, b2m is considered a good reference gene for emergency conditions, and these data agree with several studies that have suggested b2m be one of the reliable reference genes under different experimental conditions [42,43,44].NormFinder identified actb as the most stable gene for Brahman and Romosinuano breeds, while for Gyr, 18SrRNA was the most stable gene (Table 2). Genes such as actb and 18SrRNA have been successfully used as reference genes in other studies [45,46,47]. actb gene has been widely used as an internal control for different experimental assessments due to this gene encoding one of the six existing actin proteins, which are involved in cell motility, structure, and integrity, which is essential for all cellular physiological conditions [1,48]. Regarding the 18srRNA gene, it is widely used as an internal control gene for normalization in gene expression because it has a low turnover rate and is less prone to substantial changes due to physiological disturbances [49]. In this study, sdha was the least stable gene in all of the three cattle breeds using two statistical methods. Nevertheless, it has been used as a reference gene in other studies [35,50].The current Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) suggests the use of more than one reference gene in all qPCR studies [51]. Following MIQE and despite the discrepancy in the ranking orders of reference genes observed by different software (geNorm and NormFinder), actb, 18srRNA, and b2m were consistently identified as the most stable reference genes for the Brahman breed, and actb, b2m, and ywhaz were the most stables genes for the Romosinuano breed. Regarding the Gyr breed, the most stables genes were b2m, 18srRNA, actb, and ywhaz. The reference genes differences between breeds can be explained by the genetic diversity of the cattle breeds shaped by evolutionary forces such as genetic drift, migration, selection, and geographical separation [52]. Bos indicus of Indian origin and Bos taurus of European and African origin are the two main cattle subspecies [53]. In general, Bos indicus cattle breeds (Brahman and Gyr) have a greater adaptive capacity to stressful environments than Bos taurus breeds [54]. In tropical countries, Bos indicus breeds such as Gyr and Brahman are very important because of its tolerance of heat and parasites and because they are essential to the breeding of hybrids [55].Notwithstanding, some Bos taurus breeds adapted to tropical climates might be heat tolerant and exhibit a higher reproduction, growth, and carcass quality than Bos indicus breeds [53]. For example, the Romosinuano tropically adapted Bos taurus is a breed native to Colombia, South America, that is characterized by having a high reproductive efficiency [56,57]. In this way, differences in the reference genes can be linked to the genetic difference related to the subspecies and breed differences.5. ConclusionsIn conclusion, by using two statistical methods to determine the expression stability of five reference genes under heat stress conditions, our study suggests the use of the geometric mean of actb, 18srRNA, and b2m genes (for Gyr and Brahman) and actb, b2m, and ywhaz genes (for Romosinuano) as suitable reference genes for the normalization of gene expression.

animals : an open access journal from mdpi

10.3390/ani13061100

PMC10044356

Emotions play an important role in animal survival through better cohesion and coordination, and affect behavioral, physiological, and cognitive responses in animals. Improving positive emotions and reducing negative emotions has been advocated for better compliance with animal welfare and to improve the productivity of animals. The preslaughter handling of animals is a very crucial stage of meat production as it affects animal welfare and meat quality. The slaughter environment could lead to emotional stress in animals. There is a need to study the effect of exposure to the slaughter environment in goats.

Recent advances in emotions and cognitive science make it imperative to assess the emotional stress in goats at the time of slaughter. The present study was envisaged to study the electroencephalogram and physiological responses as affected by slaughter empathy in goats. A total of 12 goats were divided into two groups viz., E-group (goats exposed to slaughter environment, n = 6) and S-group (goat slaughtered in front of E-group, n = 6). The electroencephalogram and physiological responses in male Boer cross goats (E-group) were recorded in a slaughterhouse in two stages viz., control (C) without exposure to the slaughter of conspecifics and treatment (T) while visualizing the slaughter of conspecifics (S—slaughter group). The exposure of the goat to the slaughter of a conspecific resulted in a heightened emotional state. It caused significant alterations in neurobiological activity as recorded with the significant changes in the EEG spectrum (beta waves (p = 0.000491), theta waves (p = 0.017), and median frequency MF or F50 (p = 0.002)). Emotional stress was also observed to significantly increase blood glucose (p = 0.031) and a non-significant (p = 0.225) increase in heart rate in goats. Thus, slaughter empathy was observed to exert a significant effect on the electric activity of neurons in the cerebrocortical area of the brain and an increase in blood glucose content.

1. IntroductionEmotions are very intense, short-term positive or negative state responses to external or internal stimuli of specific importance for a living being. Emotions determine the behavioral decisions to approach or avoid stimuli [1]. Emotions allow animals to cope with situations with negative or positive meanings and involve certain neurophysiological responses [2]. Emotions play an important role in animal welfare and it has been advocated to improve positive emotions and reduce negative emotions for better compliance with animal welfare and to improve the productivity of animals [3]. This area is getting due recognition in the last two decades with a focus limited to pharmaceutical applications and animal welfare compliance by studying animal behavior (Ethology). Analyzing and comprehending the emotional experiences of an animal could provide information about its welfare status [2]. The present research focus in ethology is to apply innovative research frameworks such as studying the valence (activation) and arousal (excitation) aspects of emotions [4] with various indicators such as neurophysiological indicators (heart rate, brain activity, neuroendocrine response) [5,6], behavior indicators (facial expression, vocalization, tail, and ear postures) [5,7], facial expressions [2,8], and cognitive changes (judgment biases) [9]. The emotional changes (vocalization and facial expression) can be detected by conspecifics through olfactory, visual, or audible means, and an automatic trigger state matching between two individuals (emotional contagion) [10]. In the whole process, one animal is affected and shares the emotions of another conspecific via empathetic processes [11]. This emotional contagion helps in regulating social interactions and the fast exchange of information among group members. It facilitates better cohesion and coordination among group members in defense (in case of negative emotions such as fear due to the presence of a predator) or better group adhesion in positive emotions [12]. This emotion sharing leads to cognitive forms of empathy comprising sympathetic and empathetic concerns in turn helping the receiver to downregulate its own emotional response by effective sharing among conspecifics [13]. The sympathetic form of empathy could be widespread among animals but still lacks proper information in non-human animals due to a lack of a suitable methodology or experimental design [13,14]. Goats are small ruminants that contribute significantly to the socio-economic development of rural economies owing to their survival and productivity in a harsh climate, disease resistance, low neophobic responses, ability to cope with stressors, and inquisitiveness [15,16,17]. Goats have the capability of identifying and responding to calls with different emotional valences such as food frustration or reward and were observed to have head-orientation bias to the right side upon the vocalization of a conspecific in the context of frustration and dog barks indicating frustration [18]. Similarly, positive and negative emotional-linked vocalization was reported to affect the behavior and cardiac response in goats [19]. There are several reports (published reports, spy cameras, hidden videos, etc.) mentioning the improper handling of animals during slaughter [20,21,22,23,24,25]. However, physical mishandling is widely reported and studied in the slaughter of animals, but the emotional mishandling of animals during slaughter has been largely overlooked. Positive emotions could promote positive welfare among livestock [19]. During slaughter, animals undergo severe emotional stress and distress due to the slaughterhouse environment (novelty, noise, unfamiliar animals, objects, and persons), the odor of blood and animal waste, and animals’ vocalization. Such types of situations are widely prevalent in both developing and developed worlds [18,19,20,21,22,23]. Further, at the time of religious sacrifice of animals during festivals, animals are slaughtered in groups in front of conspecifics. Research on emotional stress as affected by slaughter empathy is scarce. With the advancement in cognitive science in non-human animals, it is becoming imperative to study this aspect of non-human animals with an appropriate research methodology [26]. Recently, electroencephalogram (EEG) has increasingly been used for assessing pain and stress during the slaughter of livestock [27]. It is a technology used to measure the electric activity of neurons in the cerebrocortical region of the brain by fixing electrodes on various positions of the brain [28,29]. The electric activity of neurons was used to assess pain and stress in animal welfare during the slaughter process in goats [30,31,32,33]. However, the high cost of equipment, experimental conditions, and analysis still remain issues in its popularization. To the best of our knowledge, there is no published study available on the application of EEG in assessing potential pain and stress during exposing animals to a slaughter environment. The neural oscillation/electric signals produced by the cortical pyramidal neurons upon various emotions or feelings could be measured by placing electrodes at different areas of the scalp and these signals could be analyzed by various EEG spectrum variables such as frequencies, timings, total energy, and amplitudes [34]. Thus, the present study was designed to evaluate the oscillation/electric signals produced by the cortical pyramidal neurons via the EEG recording during slaughter empathy in goats. The physiological parameters were also assessed to correlate these with the EEG power spectrum. 2. Materials and Methods2.1. Ethical ApprovalThe present study was conducted following the animal ethics guidelines of the Research Policy of Universiti Putra Malaysia as per Institutional Animals Care and Use Committee approval No.: UPM/IACUC/AUP-R003/2022, Dated 27 May 2022.2.2. AnimalsGoats (12 Boer cross, age 12 months, 25–30 kg live weight) were purchased from the local market (Global Field Trading, No 12, Jalan 9/6, Seksyen 9, 43650 Bandar Baru, Bangi, Malaysia). These animals were housed at a small ruminant housing facility at the Institute of Tropical Agriculture and Food Security (ITAFoS) in Universiti Putra Malaysia, located on latitude 259′06.5″ N and longitude 101043′40.7″ E (Jalan Maklumat) for 14 days adaptation period. Animals were housed separately with 0.3 m2/animal size individually in naturally ventilated pens. During the stay, animals were fed twice daily and accessed the ad libitum freshwater source. Animals had proper access to veterinary services, and various physiological parameters were recorded daily on the animal monitoring sheet (heart rate, rectal temperature, breathing rate, normal/abnormal movement, and normal/abnormal activity). Prior to the start of the experiment, the animals found suitable were transported (2.0 km) from the farm to the research slaughterhouse of the Department of Animal Science, Faculty of Agriculture, Universiti Putra Malaysia (258,059.000 N; 10,144,006.400 E). The animals were rested overnight in the lairage with ad libitum drinking water availability. A trained veterinarian conducted the ante-mortem and post-mortem inspection during the slaughtering process. 2.3. Experimental Conditions and DesignThis study was conducted in September and October 2022. Animals were assigned into two groups with the treatment group exposed to the slaughter environment and emotional stress (n = 6) (E group) while slaughtering the other animal (S-group). The goats were Halal slaughtered by transverse severance of the carotid arteries and jugular veins as per the standard protocols outlined in the MS 1500:2009 (Department of Standards Malaysia, 2009). Figure 1 presents the experimental design of the experiment. 2.4. Electroencephalogram RecordingThe EEG sampling of goats was performed at the slaughter hall. The goat (treatment group) was moved from the lairage to point of slaughter by using a race and EEG sampling was performed in the absence of another goat. After that, another animal (from S-group) was taken to the slaughter point and slaughtered. To mimic the normal practice following slaughter in developing countries and in religious sacrifice, the goats were exposed to slaughter environment by auditory, olfactory, and optic senses. The distance between the goat visualizing the slaughtering and undergoing the emotional stress of slaughtering environment and the slaughtered goat was kept at approx. 2 m throughout the study. Goats were standing and restrained minimally during the whole process of the EEG reading. The whole slaughter process was completed within 4–6 min. After each slaughter, the site was thoroughly washed with water before repeating the same process for another pair of animals. The EEG sampling was performed by using two conductive electrode patches attached to the zygomatic process of the frontal bone and the mastoid area by following the method as followed by Sabow et al. [30]. A fur area of 5–6 cm diameter was shaved (5–6 h prior to study) in between the mastoid process and the medial canthi of the eyes. The area was cleaned and gently rubbed with cotton rolls containing 70% ethanol to degrease the area, thereby improving the attachment of electrode gels hydrogel conductive adhesive sterile disposable electrodes (Covidien LLC, Mansfield, MA, USA) to the shaved skin. It was ensured to properly shave and clean the area to improve the quality of signals, thereby EEG quality. A negative (inverting) electrode was also placed on the zygomatic process of the frontal bone (on the right side, 1.5–2.0 cm below eye level). The positive (non-inverting) electrode was placed on the cleaned mastoid process [35]. The attachment of electrodes is depicted in Figure 2. The EEG recording was carried out by using Powerlab 4/20 data recording system (Powerlab data acquisition system, ADInstruments Ltd. Sydney, Australia) with the help of Chart 5.0 (PowerlabTM data acquisition system, Sydney, Australia) installed in a laptop. The EEG recording was started within 30 s upon the placement of the electrodes and recorded for 5–7 min on E-group goat till another goat undergoing slaughtering from S-group was dead. The determination of the state of death was confirmed by the absence of pupillary and corneal reflexes, flaccid tongue, absence of breathing, and fully dilated pupils as per Malaysian Protocol for the Halal Meat and Poultry Production, Department of Islamic Development, Malaysia (MS 1500:2009). Observations were made during the EEG reading to record the artifacts resulting from the physiological rhythmic movements such as eyelids or cardiovascular movements, electrical interferences, and physical movements from the goats themselves, such as ear flapping or rumination. EEG activities were analyzed later offline using the Chart 5.0 software (ADInstruments Ltd., Sydney, Australia). The EEG was recorded at a sampling rate of 1 kHz. The individual power spectrum of alpha, beta, delta, and theta waves was calculated based on the amplitude and frequency of the EEG signals [33]. Artifacts were removed from the overall activity, and individual waves were subjected to fast Fourier transformation (FFT) analysis. FFT is a mathematic tool that assists in quantifying information within the raw EEG signal by changing the raw EEG signal to the frequency domain from the time domain, thereby generating a power spectrum. Total power (Ptot, the total area under the curve), root mean square (RMS), and median Frequency (F50, frequency below which 50% of Ptot lies) were calculated repeatedly for non-overlapping of one-second epochs, yielding 60 epochs per minute [35]. A 60-s block EEG data were collected in E-goat at the control value (prior to exposure to the slaughtering process) and after 90 s of neck cut of S goats (treatment value). Each block was calculated for consecutive non-overlapping 1-s epochs. The EEG power spectrum is depicted in Figure 3a,b. Figure 3a,b represent the electroencephalogram’s electrical activity categorized as delta (4 Hz), theta (4–7 Hz), alpha (8–13 Hz), or beta (>13 Hz) waves. Figure 3a refers to the EEG power spectrum recorded in E-goat during the control state without exposure to slaughter environment. Figure 3b refers to the EEG power spectrum recorded in E-goat during the treatment phase by exposing to slaughter environment by slaughtering a goat from the S-group. A black box on the screen represents an epoch length.2.5. Physiological Responses The physiological responses to emotional stress during exposure to the slaughter process in goats (E-group) were assessed by measuring heart rate (by stethoscope), rectal temperature (by thermometer), and blood glucose (by portable blood glucometer by putting a drop of blood on test strip onto the device) before bringing the S-goat (control phase) and after 1.5–2 min of neck cut of S-goat (treatment phase). The animal was restrained minimally during the whole process and heart rate and blood collection were undertaken by experienced technical staff by gently placing knees behind the shoulder and 300 raising the animal head at lairage and immediately after exposure.2.6. Statistical AnalysisThe data were tested for normal distribution using a Shapiro–Wilk test using SPSS Statistics Version 20 software (IBM Corporation, New York, NY, USA). Paired T-test was used to determine the differences in values between pre-T and post-T (n = 6). A p-value of less than 0.05 was considered statistically significant.3. Results and Discussion3.1. EEG VariableThe emotional stress due to exposure to a slaughter environment in goats was observed to have a significant effect on the electrical activity of neurons as recorded by the EEG power spectrum (Table 1).An in-depth analysis of the EEG power spectrum could provide details about the changes in the electrical activity of cerebrocortical activity [33]. These neurons are widely acknowledged to play an important role in pain perception [36]. An increase in the brain activity of conscious animals during the preslaughter handling and slaughter was proposed to be associated with pain sensation [37].3.1.1. Alpha, Beta, Gamma, and Theta Waves PatternThe alpha waves of the goats were observed to have a non-significant increase (p > 0.05) during the treatment phase (exposure to the slaughter process/slaughter environment) as compared with the control phase. Alpha waves have a frequency width of 8–12 Hz and, in humans, these waves are correlated with auditory and visual stimulations with memory-related events [38]. Various slaughter environments could be attributed to the increase in the alpha waves. The alpha waves were also recorded as significantly higher in lairage compared with the baseline value after 6 h of transportation in goats [31]. An increase in alpha wave activity was also recorded in lambs and goats after head-only and head-to-back electrical stunning [39]. The beta waves significantly (p = 0.000491) increased due to treatment compared with the control value. The beta waves in the EEG power spectrum increased upon the increased brain activity [40]. The beta waves were reported to increase during brain activity in a panic condition [38,41]. In humans, a higher beta power was recorded under stressful conditions [42]. Several studies have recorded an increase in the beta waves during stress in animals such as during the transportation and slaughter of goats [31,32]. The delta waves followed a similar trend to that of the alpha waves with the treatment phase having higher but comparable values to the control phase. Higher delta waves were recorded in goats during slaughter compared with the corresponding value at the farm [31]. These waves are associated with a brain’s default mode network [43]. The theta waves in the treatment phase were recorded as significantly (p = 0.017) higher than their control values. In humans, these waves represent a heightened emotional state with increased alertness and arousal [38,41]. The theta waves were also recorded significantly higher under pre-slaughter stress but not under transport stress [31]. In horses, the increased theta wave was correlated with stereotypic behavioral performance/compromised animal welfare, and horses with a good welfare status had lower gamma waves in the right hemisphere [44]. The significant increase in the theta waves in the present study could be due to the heightened emotional status of goats arising due to exposure to the act of slaughter. Similarly in humans, Kim et al. [45] proposed the accurate and early detection of emotional stress and stages of stress by recording EEG using three-dimensional (3-D) convolutional neural networks by considering theta, alpha, beta, and gamma wave patterns. Further, a comparative value of theta/beta power was reported to be used in the stress monitoring system with more than 90% accuracy and classifying stress in a low level, a moderate level, and a high level [46]. In our present study, a significant change occurred in the value of beta and theta waves. Sabow et al. [33] observed an increase in brain activity of goats as reflected by the EEG power spectrum due to stress during slaughter. Similarly, Zulkifli et al. [47] observed the changes in alpha, beta, theta, and delta waves of the EEG power spectrum under different stunning and slaughter methods in cattle. 3.1.2. Ptot and F50The total power of the EEG spectrum (Ptot) (Cohen’s d value- −0.34) showed a non-significant increase during the treatment phase as compared with the control phase (Figure 4). The Ptot of the EEG spectrum correlates with the relaxed phase of animals, with animals in the relaxed phase usually having lower power [27]. The median frequency (MF or F50) of the EEG spectrum was recorded as significantly (p = 0.002) higher in the treatment phase compared with the control. An increase in the median frequency in the EEG spectrum typically indicates stress or painful conditions [35]. Under situations of pain, the EEG spectrum was noticed to have a significant (p < 0.01) effect on Ptot, F50, and F95 (95% spectral edge frequency) values in lamb during the castration process [48]. Under a conscious state, EEG could be used as a tool to assess pain and stress in animals by measuring F50 and F95 [48]. The increase in the F50 of the EEG power spectrum was related to noxious stimulation and pain during the neck cut [34]. During slaughter, ventral neck cuts in goats were observed to have a significant increase on the F50 as compared with the stay in lairage [31]. Reports regarding the changes in Ptot in association with F50 are inconsistent. Murrel and Johnson [49] observed a decrease in Ptot with an increase in F50. Imlan et al. [50,51] and Abubakar et al. [52] in cattle and Raghazli et al. [31] in goats also reported a positive correlation in Ptot and F50 under stress as well as noxious stimuli. On the other hand, Kaka et al. [29] and Karna et al. [53] reported no change in Ptot in association with F50 in response to noxious stimuli under anesthesia. Thus, the results of this study also show a trend similar to that reported by Kaka et al. [29] and Karna et al. [53] in dogs under anesthesia, however, the present study was conducted in conscious goats. It has been reported that changes in the Ptot were not directly associated with F50 in response to noxious stimuli [29] and that these changes could represent a different component of nociception than F50 [53]. These lines of evidence, including the present study, verifies that Ptot is not directly associated with MF and might have a connection with other components of pain and stress, which is yet to be explored. Thus, the higher F50 in the present study (Cohen’s d value- 0.79) could be correlated with emotional stress in goats. The higher responses of F50 and F95 were reported to associate with pain in calves [54,55,56]. The EEG spectrum was also noticed with increased F50 and Ptot [33]. Furthermore, in gregarious animals such as goats, the visual and physical separation from their herd and the novelty of the environment could be an additional factor that contributed to substantial fear and anxiety at the slaughter point in the present study [57]. Several psychological studies on human subjects have also confirmed the long-lasting effect on stress hormones due to negative emotions such as anger, disgust, and fear, whereas positive emotions help in the production of beneficial hormones [45]. The EEG has been used in humans for emotion recognition with remarkable results [58,59,60]. However, there is a lack of studies on emotional stress in animals by applying EEG recording. To the best of our knowledge, this is the first study recording emotional stress and slaughter empathy in goats upon exposure to the act of slaughter by measuring the electric activity of cerebrocortical neurons. During slaughter, the overall slaughter environment is the key determinant in affecting animals’ physiological and emotional states. The emotional response in animals comprises behavioral, physiological, cognitive, and subjective components [61]. This emotional stress resulted in changing the EEG variables and physiological responses in the goats. As emotional stress is of a very short duration in the context of slaughter, with varying degrees of intensity or threat levels, measuring these slaughter empathy reactions or various neurobiological responses warrants using an appropriate methodology that records these variations instantaneously, sensitively, and accurately. EEG could be used in non-human animals to recognize emotions up to the point of slaughter [62,63,64].3.2. Physiological ResponsesThe stimuli at the slaughterhouse differ from the farm, and this may affect the emotional status of animals, and their transport from the farm to the slaughterhouse may further aggravate it [65]. Goats, being a prey animal, have well-developed mechanisms to respond to any situation or potential state of threat or danger. The physiological responses in goats were recorded with higher values for heart rate (p = 0.225) and a significant increase in blood glucose (p = 0.012) during the stage of exposure to the slaughter of conspecifics (Table 2). However, the temperature (both rectal and auditory) was recorded as comparable in the present study. Semiochemicals released from the blood of the slaughtered animals were considered a major factor in causing distress in animals during slaughter. As per Grandin and Vogel [66], the vision or smell of blood is not thought to cause distress unless the animal whose blood is present had been distressed during slaughter (e.g., he or she struggled and vocalized). Preslaughter handling had been established to affect animal welfare and meat quality [67,68]. An increase in the glucose value in a conscious state is regarded as an indicator of stress in goats [69]. Various stress factors affect the heart rate and blood glucose levels in goats due to the increased release of catecholamines and glucocorticoids. It facilitates increasing glucose production from glycogenolysis and gluconeogenesis required for preparing animals for the response to a stressor (fight or flight response) [67,70]. An increase in blood glucose concentration was also reported during stressful conditions (pasture and slaughter) in deer [71]. Sim et al. [72] recorded increased blood glucose in mice upon emotional stress via the activation of adrenergic and glucocorticoid responses. In the present study, the blood glucose levels were recorded within the normal range of glucose in animals (4.4–6.6 mMol/L) [73,74]. Further, the increase in heart rate could be correlated with the various frequency bands of EEG in sheep during slaughter [75]. 4. ConclusionsBased on the present study, it can be concluded that the exposure of goats to the slaughter of conspecifics alters the emotional state of goats, consequently causing significant changes in neurobiological activity as recorded with the significant changes in the EEG spectrum (beta waves, theta waves, and MF 50). Emotional stress was also observed to significantly increase blood glucose levels with no difference (p = 0.225) in the heart rate in goats.5. Limitation and Future DirectionThis study highlighted the issue of emotional stress in goats upon exposure to the slaughter environment. There is a need to take further studies on evaluating its effect on other common and established stress biomarkers such as stress hormones. Its overall impact on meat proteomics and meat quality should be assessed. As in the present study, all three senses (auditory, olfactory, and optic) were used to mimic the common practice in some places. A further study on which sense has more/no effect could be useful, so to provide accurate and practical recommendations for such situations. There are some practical challenges while conducting such studies due to potential ethical issues as the present study correlated the higher electric activity cerebrocortical neurons and physiological parameters with the emotional stress during the slaughter of goats. There is a need to take more in-depth studies to confirm slaughter empathy, which if established could have wide implications in the goat meat industry such as the requirement of slaughter out of sight of a conspecific. Interestingly, the same is recommended for Halal slaughter management. Such data will be valuable in increasing awareness among common people and also sensitizing the person involved in the meat industry, thereby improving animal welfare standards. However, there is a need to take a study on emotional stress and slaughter empathy in goats with a higher sample size to get more insights into this aspect.

animals : an open access journal from mdpi

10.3390/ani11092666

PMC8472284

The age of horses can influence several properties of the obtained raw material. As the age of horses increases, the meat retains less water, and more fat (p < 0.05) and minerals. In general, horse meat from older animals exhibits undesirable stringiness and hardness, due to a large proportion of connective tissue (collagen). Currently, many methods are applied to improve the tenderness of meat. Of these, the most popular is marinating the meat with various substances, which enhances the functional and sensory properties of the meat. Freezing is a widely accepted method for extending the shelf life of meat. Both the technique used for freezing and further storage at negative temperatures have an impact on some of the properties of meat. Most importantly, the pH value, color, and water absorption of meat tend to change with freezing. In addition, the dry matter content and tenderness of meat increase. This study aimed to analyze the impact of horse age, marinating substances, and frozen storage on the quality of horse meat. As horses age, the values of meat cutting force increase (p < 0.05). For example, the cutting force increases by 4.57 N/cm2 during the first period of freezer storage, and by 3.28 N/cm2 after 3 months of freezer storage (p < 0.05).

The present study analyzed the influence of horse age, substances used for marinating, and frozen storage on the quality of horse meat. It was conducted on the samples of the longest thoracic muscle, obtained from 12 carcasses of horses (aged 4–7 and 8–12 years). Among the analyzed samples, a higher fat content (p < 0.05) was found in the meat obtained from the carcasses of older horses. The pH value of the meat samples was influenced by the treatment applied (p < 0.05). Of the substances used for marinating, malic acid caused a decrease in the pH of the meat obtained from young horses (p < 0.05). A similar effect was observed with the addition of phosphates to malic acid-marinated meat. On the other hand, the use of phosphates for marinating resulted in an increase in the pH of the meat obtained from older horses (p < 0.05). The substances used for marinating the horse meat did not significantly affect the reduction in cutting force values. Furthermore, the values of shear force, hardness, stiffness, gumminess, and chewiness of the meat increased with horse age (p < 0.05). An influence on the color parameters a* and b* of the meat was found for the interaction between age, storage period, and the type of treatment (p < 0.05). The use of lactic acid and malic acid for marinating the meat of young horses caused a decrease in the proportion of red color (4.67 and 3.43) and an increase in the proportion of yellow color (3.81 and 1.71), especially after 3 months of freezer storage. All the substances used for marinating (except for phosphates) were associated with higher (p < 0.05) thermal and forced drips of meat from the carcasses of both young and older horses during each storage period, in comparison to the control. The interaction between age and the type of treatment had an influence on the tenderness and juiciness of the horse meat (p < 0.05). In sensory evaluation, it was noted that the interaction between age and the treatment procedure influenced the tenderness and juiciness of the meat samples (p < 0.05). There is still a need for further research to increase knowledge regarding how to improve the quality of horse meat, and ultimately increase the demand from consumers and meat processing plants.

1. IntroductionHorse meat is a distinct food with a specific consumer base, and its production is very popular all over the world. Young, well-muscled animals, whose meat is highly valued and willingly bought, are targeted for slaughter. In Europe, horse meat is mainly consumed by Italians, followed by Belgians, with an annual consumption per capita of 0.88 and about 0.5 kg, respectively [1]. Horse meat is particularly popular in Western European countries, where it is treated as an equivalent to other meat types and is often valued higher than beef or pork. In Poland, horse meat is not consumed, due to various reasons, including emotional resistance, lack of skills in preparing horse meat dishes, and a traditional consumption model that prefers other types of meat. Popularizing horse meat in Poland also requires breaking traditions regarding its distribution. Horse meat is incorrectly regarded as unworthy of consumption and promotion, and it is worthwhile to make attempts to popularize its consumption, since horse meat constitutes a significant reserve of meat mass that can be utilized. Moreover, there are very few promotional activities promoting the nutritional value of horse meat and the products made from it, and due to their high price, almost all products are sold to richer EU countries.The low popularity of horse meat in Poland is also associated with its poor quality. In the past, it was mainly obtained from older animals, which are not suitable for export. Horse meat from older animals is characterized by low tenderness, considerable hardness and coarseness, a very dark color, and a high fat content [2]. In 2019, horse production in Poland was estimated at 27 thousand heads and horses, with a total weight of 13.3 thousand tons (industrial slaughtering of animals) [3]. Due to the fact that horse meat is rarely consumed in Poland, 80–95% of domestic production is exported. The main customers are Italy, where draft horse meat is highly valued, followed by France, Belgium, Austria, and Germany, which account for 70–72% of horse meat export from Poland [2]. Horse meat is a source of valuable nutrients. It is lean, has a low fat content [4,5,6,7,8], and is rich in proteins with a high biological value as well as desirable amino acids [9,10,11]. However, unlike the meat obtained from other animals, horse meat has a high amount of glycogen, which imparts a sweet taste [12,13]. Another unfavorable quality of horse meat is its dark red color, accompanied by a faint brown tinge, due to the high concentration of the muscle pigment myoglobin [9,14,15,16,17]. In addition, the meat darkens with the age of the animal, while the fat turns yellowish or even orange in color. Horse meat also has a large proportion of connective tissue (collagen), which is an additional distinguishing feature [16,18,19]. The age of animals is considered an important influencing factor of the obtained raw material, because as the animal gets older, several changes occur not only in the chemical composition and color of the meat, but also in the structure of proteins in the muscle and connective tissue. The mechanical stability of connective tissue increases with age, due to the cross-linking of collagen. As horse meat has a high content of collagen, it undergoes softening and physicochemical transformations for a long time in an acidified environment [20,21]. Maturation is a critical step in the production of culinary horse meat, which should be conducted with care and for a sufficiently long duration. However, it is one of the least complicated treatments, and improves the acceptance and suitability of horse meat. Significant changes also take place in meat during storage, and the lower the storage temperature is, the smaller the changes are. Freezing is a commonly used method that maintains the quality and durability of perishable meat [22,23,24]. The changes in meat quality caused by freezing are determined by the technique used and the subsequent frozen storage. The post-slaughter maturation processes of meat are slowed down or inhibited as a result of freezing. On the other hand, the processes responsible for water freezing and the formation of ice crystals within the muscle structures are intensified. Although freezing is important for the preservation of meat, both freezing and further storage at negative temperatures directly affect some of its properties. Most importantly, the pH, color, and water absorption of meat tend to change, while the dry matter content and tenderness of meat increase. Furthermore, the size of thermal drip and gel ability of muscle proteins change after the thermal treatment. Freezing also results in loosening of the capillary structures of muscle tissue, which, in turn, leads to a reduction in the tissue’s ability to retain its water during defrosting, as well as significant losses during thermal treatment, thus affecting the juiciness of the meat [25].Currently, the approach used to improve the functional and sensory properties of meat is marinating. This procedure involves soaking, injecting, or mixing the product with aqueous solutions containing various ingredients. A marinade is a water solution composed of salt and additional substances. The brine composition is selected individually for each product, taking into account the necessary additives [26,27].Due to the specific effect on muscle proteins, phosphates added for marinating increase their water absorption, and improve the binding and emulsifying properties. They also enhance the textural properties and consistency of the product, stabilize the color and fat emulsion, and increase the production efficiency. In meat processing, phosphates mainly help to dissociate the actomyosin complex, regulate the pH of the product, and increase the ionic strength of the environment and complex divalent cations [28,29]. Polyphosphate-induced changes in the water absorption of meat, together with the electrostatic effect of phosphate ions on proteins, not only alter the conformation of protein molecules, but also change the structure of the surrounding water molecules. This, in turn, improves the ability of proteins to bind water and emulsify fat. The increase in water absorption and the reduction in drip, caused by polyphosphates, can be partially explained by the increase in the pH of the cell fluid in relation to the isoelectric point of proteins. However, the pH increase is determined by the amount and type of phosphate added. A high water-binding capacity improves the binding of the plaster and keeps its surface dry, thus indirectly contributing to the stabilization of color. In addition, the sticky juice seals the pores in the muscle tissue and prevents the penetration of oxygen. This explains the antioxidant properties of phosphates and their ability to stabilize the color of cured products. These properties of phosphates are further enhanced by their metal ion-chelating activity. Moreover, phosphates limit the growth of spoilage microbes [29,30,31].Organic acids can influence the muscle fibers and connective tissue, and contribute to improving the tenderness of muscle tissue. For example, citric acid, a food acidifier, is commonly used in marinating, as it not only increases the water-holding capacity and tenderness of beef muscles, but also acts as a chelator and controls the pro-oxidative activity of metals [32,33,34]. However, this acid can lower the pH of the meat, resulting in an excessively acidic flavor, and hence can decrease consumer acceptance. For this reason, solutions with an acid concentration exceeding 0.15 M are not recommended for marinating [35]. Nonetheless, this issue can be overcome by initially reducing the pH of the muscles with citric acid, to change the texture, and then increasing the pH by adding, for example, sodium triphosphate to improve the organoleptic characteristics of the meat [36]. The acids commonly used for marinating in the meat industry are lactic acid, which acts as an antimicrobial agent [32], and acetic acid, which is known for its acidity, pH-lowering ability, and bacterial growth-inhibiting properties.The quality of horse meat determines its technological and culinary value. The meat industry currently aims at using methods that will allow the negative features of horse meat to be eliminated, especially related to the color and tenderness of this raw material [2]. The culinary use of horse meat justifies the need for further research on its properties. Knowledge of the specificity of horse meat, and especially the influence of various factors on its properties, will allow the selection of the appropriate technological procedure, and this thus will allow optimal processing of the raw material and the desired characteristics to be achieved. This would also provide an opportunity to disseminate and popularize horse meat-derived products.The broad range of horse age (from foals to horses over 20 years old) causes large differences in the obtained raw material, because as the animal ages, not only does the tissue composition of the carcass change, but also the functional and sensory properties of the meat change. Because meat obtained from the carcasses of young and old horses varies in its properties, it is advisable to investigate and identify which method used to improve the tenderness of horse meat will enhance its quality parameters. A few studies have comprehensively analyzed the influence of horse age, marinating substances, and period of frozen storage on the quality of horse meat. Those studies [17,37,38,39,40,41,42,43] contribute to our understanding of how to improve the quality of horse meat, and ultimately increase the demand from consumers and meat processing plants. Nevertheless, there is still a need for further research to increase knowledge regarding this topic.Therefore, the aim of this study was to determine the texture parameters, color, and sensory characteristics of horse meat, based on the animals′ age, marinating substances, and freezing storage time.2. Materials and Methods2.1. Raw MaterialThe study was conducted on Longissimus thoracis muscle isolated from 12 half-carcasses of younger horses (4–7 years) and 12 half-carcasses of older horses (8–12 years) from right side. The average age of the horses in the 4- to 7-year group was 5.5 ± 0.5, and that in the 8- to 12-year group was 10.0 ± 0.5. The age of horses was determined from the purchase documentation. Their weight antemortem was 500–560 kg (530 ± 30). Average carcass weight was 345 ± 30 kg. The horses used in the study were Malopolski and Silesian breeds and were obtained from farmers from Southeastern Poland. They were grouped according to their gender, as follows: 50% were geldings and 50% were mares. Each age group consisted of six half-carcasses of females and six half-carcasses of males. The horses were normal and maintained in an extensive system. After transport, the animals were kept in separate pens in livestock warehouses for about 24 h while maintaining animal welfare and under the supervision of the appropriate veterinary services. All horses were slaughtered the same day, by stunning with a captive bolt pistol according to the current methodology applied in the meat industry. From each carcass, and 24 h after slaughter, two samples (1000 g each half-carcass) from the M. longissimus thoracic, at the 13th–14th thoracic vertebrae level, were obtained to determine age, marinating substances, and frozen storage effect on meat quality. The medial portion of the samples collected was located at the height of the 13th–14th thoracic vertebrae. Each sample was cleaned to remove external fat, connective tissue, and tendons, and seven steaks (3 cm thick approximately) were obtained (from each age group, as follows: 12 half-carcasses × 2 M. longissimus thoracis samples × 7 steaks = 168 steak samples). One of the steaks was used as control, while the others were treated, after 48 h postmortem, with the corresponding compound in 1% solutions at an amount of 10% with reference to the sample weight. Muscle samples (steaks) were injected with the following reagents: (i) lactic acid (2-hydroxypropanoic acid, 80%); (ii) malic acid (hydroxysuccinic acid, 99%); (iii) phosphates (Hamina S containing emulsifiers)—E 451 (pentasodium triphosphate and pentapotassium triphosphate), E 450 (disodium diphosphate, trisodium diphosphate, tetrasodium diphosphate, tetrapotassium diphosphate, dihydrogen diphosphate, and calcium diphosphate), E 452 (sodium polyphosphate, potassium polyphosphate, sodium calcium polyphosphate, and calcium polyphosphate), and E 339 (monosodium phosphate, disodium phosphate, and trisodium phosphate (TSP)) with NaCl (salt); (iv) solution containing phosphates (Hamina S containing the abovementioned emulsifiers) and rosemary (0.1% rosemary oil); (v) solution containing lactic acid (2-hydroxypropanoic acid, 80%) and phosphates (Hamina S containing the abovementioned emulsifiers); and (vi) solution containing malic acid (hydroxysuccinic acid, 99%) and phosphates (Hamina S containing the abovementioned emulsifiers).Then, they were immediately marinated in various aqueous solutions of the compounds at 1% concentration in glass vessels (solution/sample ratio = 2:1). Later, the marinated meat samples as well as control samples were refrigerated (6 °C) for 72 h. After storage, the meat samples marinated with lactic acid and malic acid were treated with phosphates by injecting 1% solution (Hamina S containing the abovementioned emulsifiers) at an amount of 10% with reference to the sample weight. These samples were marinated again in aqueous solutions of 1% phosphates (Hamina S containing the abovementioned emulsifiers and refrigerated (6 °C) for 24 h) in glass vessels.After marinating, the batches of meat samples were subjected to flow freezing in a freezing cabinet (Budget Line type; Hendi, Warsaw, Poland), after vacuum packing them in PA/PE bags. At the beginning of freezing, the average temperature of meat was around 4 °C. Freezing at −28 °C was carried out for approximately 3 h. After freezing, the samples were stored for 1 and 3 months at −22 °C. Following predetermined periods of frozen storage, the samples were moved to the laboratory for analyses. Before quality testing, the packed samples were thawed at ambient temperature, approximately 10 °C. Once the temperature inside the meat sample reached 0 °C, defrosting was stopped and analyses were carried out.2.2. Analytical MethodsThe following parameters were measured in horse meat to determine its quality: chemical composition (amount of water, protein and fat content), pH, color, hydration properties (forced, thermal, and thawing drip), shear force of raw meat, texture (hardness 1 and 2, stiffness up to 5 and 8 mm, adhesiveness, resilience, cohesiveness, gumminess, chewiness, and springiness), and sensory quality (aroma intensity and desirability, tenderness, juiciness, taste intensity and desirability, general acceptability).The water content of the samples was determined in accordance with the PN-ISO, 1442:2000 standard [44].The protein content of the samples was determined using the Kjeldahl method. For this, the content of nitrogen calculated in the samples was converted into protein, according to the PN-75/A-04018 standard [45].The fat content of the samples was determined using the Soxhlet method in accordance with the PN-ISO, 1444:2000 standard [46].The active acidity (pH) was determined in cooled meat samples using an OSH 12-01 electrode and a CPC-411 pH meter (ELMETRON, Zabrze, Poland) with an accuracy of up to 0.01. Before measuring the pH, the device was calibrated with buffers of pH 4 and 7.The color of the meat samples was measured in their cross-section in the CIE L* a* b* system, using a HunterLab UltraScan PRO (HunterLab, Reston, VA, United States) electronic spectrophotometer (D65 light source, measuring head opening: 8 mm, white reference standard calibration: L*—99.18, a*—0.07, b*—0.05). Parameter L* denotes brightness (spatial vector), parameters a* and b* are the trichromaticity coordinates (positive a* values indicate red color, while negative values indicate green color; positive b* values indicate yellow color, while negative values indicate blue color).For measuring the physicochemical parameters, such as thermal and forced drips, the meat samples were first minced twice in a laboratory wolf device (Hendi, Warsaw, Poland) using sieves (4 mm diameter). Then, they were thoroughly mixed and homogenized, and subjected to further analyses.Thermal drip of the meat samples was determined as described by Janicki and Walczak [47]. Briefly, a finely ground meat sample (weighing 20 g) was transferred to a hygroscopic gauze and heated in a hot water bath (85 °C) for about 10 min. After heating, the sample was cooled down to 4 °C and reweighed. Thermal drip was estimated based on the change in the weight of the sample recorded before heat treatment and after cooling as follows:(1)Td (%)=WI−WIIWI × 100% where Td refers to the rate of thermal drip (%), WI refers to the sample weight before heat treatment (g), and WII refers to the sample weight after cooling (g).Forced drip of the meat samples was determined as described by Grau and Hamm [48]. Briefly, a minced meat sample (weighing about 300 mg) was placed on a Whatman paper No. 1. The paper was then placed between two glass plates and subjected to 5 kg pressure for 5 min. After squeezing, the boundaries of the surface occupied by the meat sample and the drip of meat juice were outlined on the paper and planimeterized. The forced drip size of meat juice was measured as the difference between both surfaces. Based on the obtained value, water absorption (cm2) of meat was interpreted (a higher value indicates lower water absorption).Thawing (free) drip was determined in the thawed meat samples. It was calculated based on the weight of plasma drip, in comparison to that of the sample (precision: 0.01 g). The value was expressed as drip percentage. (2)Wr (%)=MI−MIIMI × 100% where Wr is the size of thawing drip (%), MI is the sample weight before thawing (g), and MII is the sample weight after thawing (g).Shear force was measured in raw meat samples using a TA texture meter (XT plus; Stable Micro System Ltd., Surrey, UK). Briefly, the samples were cut into cylinders (along muscle fibers) of 1.0 cm diameter using a cork borer and sliced using a Warner–Bratzler blade with a triangular notch. The shear force required to cut them (N/cm2) was recorded, and the mean values of three successive replications (almost similar values) were determined.For textural analysis, samples from each batch of raw meat were cut into cubes with sides of 20 mm. Their texture parameters were determined by texture profile analysis using CT3-25 texture analyzer (Brookfield, WI, USA) equipped with a cylindrical attachment (diameter: 38.1 mm, length: 20 mm). Each sample was compressed twice and reduced to 50% of its height with a roll travel speed of 2 mm/s, with a 2-s interval between compressions. Using Texture Pro CT software (V.1.9 Build 39; Brookfield, WI, USA), the following texture parameters were determined in the samples: hardness 1 and 2, stiffness up to 5 and 8 mm, adhesiveness, resilience, cohesiveness, springiness, chewiness, and gumminess. All these parameters were counted automatically during serial measurements.2.3. Sensory EvaluationThe sensory properties of marinated horse meat samples were evaluated as described by Baryłko-Pikielna and Matuszewska [49]. Briefly, 100 g of samples was steamed at 95 °C until their internal temperature reached 80 °C ± 2 °C as determined using a digital thermometer with a needle probe (Sous Vide Thermapen; MERA, Warsaw, Poland). Before sensory evaluation, the samples were cooled down to 20 °C ± 2 °C and cut perpendicular to the fibers into 1.5-cm-thick slices. The slices were kept in disposable plastic boxes containing lids, individually coded, and offered in a random order to the evaluation panel consisting of six members (3 males and 3 females, aged 26–46 years). The members were experienced in evaluating meat and its products. Each sample was assessed in triplicate by the panel. Sensitivity and sensory fitness of the samples were tested in accordance with the ISO, 8586-2:2008 [50] and ISO, 8587:2006 standards [51]. Qualitative indices of the samples were assessed using a 5-point scale as follows: intensity of aroma (5 = very strong, 1 = negative and very poorly perceptible), intensity of taste (5 = very strong, 1 = negative and very poorly perceptible), desirability of aroma (5 = highly desirable, 1 = not desirable), desirability of taste (5 = highly desirable, 1 = not desirable), juiciness (5 = very juicy, 1 = very dry), and tenderness (5 = very tender, 1 = very hard). The evaluation was conducted in a specific laboratory that met the relevant standard requirements [52]. Before testing each sample, the evaluators took a 30-s break and washed their mouths with mineral water. The evaluation was conducted in 10 sessions, and 17 samples were assessed in each.2.4. Statistical AnalysisAll parameters were measured and sensory characteristics were assessed in triplicate. The results were statistically analyzed after grouping. All the observations (6 selected substances used for marination × 12 batches × 2 age groups × 2 storage times) were considered in the statistical analysis. Selected physical and chemical properties, texture, and sensory attributes of meat samples were analyzed by a three-way analysis of variance (ANOVA), using the GLM procedure in Statistica (STATISTICA v. 10; StatSoft, Krakow, Poland). The substances selected for marinating, age group of the animals, and period of storage were considered as a fixed effect and batch was considered as a random effect in the analyses. In the model, batch was included as a sensory variable (selected substances × age group × storage time) together with the main effects and their interaction, as well as the panelist included in the sensory evaluation. The significance of the main effects and their interaction was tested using batch as the error term (selected substances × age group × storage time). If the effects were found to be significant (p < 0.05), the means were compared using post hoc Tukey’s honestly significant difference test (ANOVA).3. Results and DiscussionThe results obtained from the chemical composition analysis of the horse meat samples are presented in Table 1. It was observed that the fat content of meat was statistically significantly influenced by the age of the horses. The meat samples obtained from the carcasses of older horses had a higher amount of fat (p < 0.05). A moderate amount of protein was also found in the meat of older horses, but the differences were statistically insignificant. This is in line with the study of Znamirowska [53], which reported that the fat content in horse meat increased with the age of animals. The authors observed that in foals (horses up to 2 years of age), the level of fat (2.37%) and protein (20.04%) was the lowest, while the content of water was the highest (76.42%) [53]. As the animals aged, the proportions of individual components changed, and in horses aged 2–7 years, the values were 3.46%, 21.41%, and 74.04%, respectively. In the case of horses from 7- to 12-year and 12- to 17-year age groups, a further increase in fat and protein content, and a decrease in water content in meat were noted. In the group of the oldest horses (over 17 years), the levels of fat, protein, and water were 5.36%, 22.38%, and 71.03%, respectively [53]. In addition, according to Korzeniowski et al. [54], as the age of horses increases, the meat retains less water, and more fat and minerals.The present study showed no significant effect of increasing the time of freezer storage on the basic chemical composition of horse meat. The results confirmed those of previous research works [37,38], which also did not show any significant effect of frozen storage on the chemical composition of horse meat.The results obtained from the analysis of the selected physical and chemical characteristics of horse meat samples are presented in Table 2. It was observed that the pH of the meat samples was statistically significantly influenced by the type of treatment applied. Marinating with malic acid decreased the pH of the meat obtained from the carcasses of young horses (p < 0.05). A similar effect was found when phosphates were added to malic acid- and lactic acid-marinated meat samples. The interaction between frozen storage and treatment type, in turn, influenced the acidity of the meat obtained from older horses. Marinating with phosphates with salt and phosphates with rosemary significantly increased the pH of the meat from older horses (p < 0.05). This is in line with the study of Bianchi et al. [55], which showed that the marination of turkey breast meat with a solution containing 2.0% sodium tripolyphosphate (STPP) and 1.4% sodium chloride caused a 0.20 unit increase in the pH of the material. Similarly, Mudalal et al. [29] observed that when chicken breast fillets were marinated with a solution of 0.3% STPP, their pH increased by 0.15 units. In turn, Garner et al. [56] marinated fresh meat from chicken breasts with 0.25% and 0.5% STPP solutions, and observed an increase in pH (p < 0.05) in both fresh meat and meat subjected to frozen storage at −20 °C for 6 days after STPP addition, which was attributed to marination with STPP solutions. Khan et al. [57] marinated duck breast meat samples with a 1.5% STPP solution and 3% salt, and noted an increase in pH in comparison to the control. When the marinating time was increased to 7 days, the pH of the meat samples increased further.It was observed that the interaction between age, storage period, and type of treatment had a statistically significant influence on the color parameter a* of the meat sample. A lower proportion of red color was found in the samples marinated with lactic and malic acids, while a higher proportion was found in phosphate with salt-marinated samples of meat from young horses, after 1 month of frozen storage (p < 0.05). In the case of meat from older horses, a statistically significant increase in red color was noted in the samples marinated with phosphates (with salt and rosemary). It should be emphasized that phosphates, and mainly polyphosphates, moderately protect myoglobin against oxidation by sequestering iron and copper ions, and thus contribute to preserving the red color of fresh meat, as well as the pink color of marinated meat. The results of our previous studies showed that the color parameter a* decreased during cold storage in horse meat marinated with acid solutions (p < 0.05) [39].Similarly to parameter a*, the interaction between age, storage period, and type of treatment also had an influence on parameter b* (p < 0.05). After 3 months of frozen storage, a higher (p < 0.05) proportion of yellow color was observed in the lactic acid-marinated samples of meat from young horses, in comparison to the control sample. However, in the case of meat samples from older horses, a decrease in parameter b* was observed after a 3-month freezer storage period if lactic acid, phosphates, and salt were applied to the marinade, and phosphates were added after marination with lactic acid, in comparison to the control sample (p < 0.05). In the same period of frozen storage, the proportion of yellow color was found to be increased in the control samples of meat obtained from older horses (p < 0.05).Mudalal et al. [29] found an increase in brightness, and the color parameters a* and b* in chicken breast meat samples marinated with 0.3% STPP, in comparison to the control samples. In turn, Bianchi et al. [55] observed an increase in brightness and color parameter b*, but a decrease in parameter a* in turkey breast meat samples marinated with a solution containing 2.0% STPP and 1.4% sodium chloride, in comparison to non-marinated meat samples. Garner et al. [56] marinated fresh meat from chicken breasts with 0.25% and 0.5% STPP solutions, and noted a decrease in brightness and a* as well as b* parameter in the meat samples. In the case of meat subjected to frozen storage for 6 days at a temperature of −20 °C after the addition of STPP, a decrease in brightness, as well as both a* and b* parameters, was observed [56].The results showed that the type of treatment had a statistically significant influence on the hydration properties of horse meat. Compared to the control sample, higher (p < 0.05) thermal and forced drips were observed in the meat samples from the carcasses of young as well as older horses in each frozen storage period with most of the marinating substances used (except for the samples marinated with phosphates with salt and those marinated with phosphates with rosemary, in which the differences in forced leakage were mostly statistically insignificant).Phosphates have the ability to buffer meat and change the pH of meat proteins from the isoelectric point, thus contributing to improving their water-holding properties [56]. Furthermore, these substances can open the structure of proteins. Such “open” muscle proteins exhibit a greater water-binding capacity, which explains the better water retention of meat observed during heat treatment. In addition, phosphates reduce muscle contractility during thermal treatment, and hence increase the efficiency of the process.The thermal drip values of the analyzed meat samples were found to be statistically significantly influenced by the interaction between the storage period and the treatment applied, as well as by the interaction between the age of the horses and the treatment.A higher thermal drip was observed in the samples of meat from young horses after 3 months of frozen storage (except for those marinated with phosphates). In general, the values of thermal and thawing drip were found to be higher in the meat of young horses in all the storage periods, regardless of the type of treatment. However, the level of thawing leakage was found to be statistically significantly influenced by the type of treatment, and the interaction effect between age and the treatment applied. In general, thawing leakage was higher (p < 0.05) in all the tested meat samples compared to the control. However, higher (p < 0.05) thawing leakage was noted in the meat of young and old horses marinated with acid solutions, and, after 3 months of freezer storage, in the meat from young horses marinated with phosphates with rosemary and both types of acids with phosphates. A similar relationship has been reported in our earlier studies [37].Marinating with different compounds is carried out to improve the hydration properties of meat. For instance, Pérez-Chabela et al. [41] analyzed the hydration parameters of horse meat marinated with CaCl2 and observed an increase in their values. By contrast, Pérez et al. [42] observed a decrease in water retention capacity in all the groups of meat samples marinated with CaCl2, and reported that this effect was related to the proteolysis of myofibrillar proteins and the decrease in the pH value. Similarly, Aktaş et al. [33] found a statistically significant increase in thermal drip values in beef samples marinated with CaCl2 and NaCl solutions.The results obtained from the texture analysis of the horse meat samples are presented in Table 3. It was observed that the shear force of meat was significantly influenced by the age of the horses. Regardless of the frozen storage period, the force needed to cut was lower (p < 0.05) in the samples of meat obtained from the carcasses of young horses. Taking into account the type of substance used for marinating the meat samples, it was noted that the phosphate–rosemary solution caused a statistically significant increase in shear force, and, as a result, a higher force was needed to cut the meat obtained from the carcasses of older horses after 3 months of frozen storage, in comparison to the control sample.Qin et al. [58] marinated beef samples with 5% disodium polyphosphate, 3% trisodium polyphosphate, 3% sodium hexametaphosphate, and 3% STPP, and investigated the effect of these substances on the shear force values of the material. They found that polyphosphates significantly reduced the shear force of the samples, as compared to the control. Similarly, Wang and Tang [59] marinated beef samples for 6 h with a 0.5% malic acid solution and found that the shear force values of the samples were significantly decreased in comparison to the control.The present study revealed that the age of horses had a statistically significant influence on the texture parameters of the meat samples. The results showed that the values of hardness, stiffness, gumminess, and chewiness of the meat increased with the animal’s age (p < 0.05). Additionally, the interaction between storage time and treatment statistically significantly influenced the values of hardness 1, while the interaction between age and storage time significantly influenced the values of stiffness up to 5 mm. The latter parameter was also found to be significantly influenced by the interaction between all the analyzed factors.Previous studies in different species [34,36,60,61] have indicated that marinating with organic acids reduced the hardness of meat. Hosseini and Esfahani Mehr [34] showed that organic acids decreased the pH value of beef meat samples, thereby leading to the solubilization of collagen tissue and causing an increase in the tenderness of the material.Different compounds have been used for marinating horse meat, to achieve improved texture parameters. Studies [43] on similar material showed that CaCl2 solution used for marinating contributed to lowering the hardness of meat samples in comparison to the control samples.The results obtained from the sensory evaluation of the horse meat samples are presented in Table 4. It was noted that the interaction between age and the treatment procedure statistically significantly influenced the tenderness and juiciness of the meat samples. In the case of meat obtained from older horses, a statistically significant difference was noted after 3 months of frozen storage in these parameters, between the control sample and samples marinated with phosphate solutions with salt. Similarly, a statistically significant difference in tenderness and juiciness was found between the control sample and the samples treated with lactic acid with phosphates, and malic acid with phosphates, after 1 month of storage in freezing conditions. An improvement in tenderness was observed during sensory evaluation in meat samples marinated with lactic acid with phosphates and malic acid with phosphates (p < 0.05). Another study of our research group [40] showed that substances such as citric acid, and 0.2 and 0.3 M CaCl2 caused a statistically significant improvement in meat tenderness.Phosphates are applied as functional additives in meat processing to improve the sensory quality of the raw material (tenderness, juiciness, color, aroma) [30]. Capita et al. [62] showed that chicken legs immersed for 15 min in 10% and 12% solutions of TSP showed improved sensory quality compared to the control sample. Dipping in a 10% TSP solution contributed to improving the smell and color of chicken legs, while dipping in a 12% TSP solution enhanced the color and overall acceptability. Sheard et al. [63] injected polyphosphate solutions into pork, and investigated their effect on the juiciness and tenderness of the meat. They analyzed the samples after cooking by grilling to achieve a temperature of 72.5 °C or 80.0 °C in the geometric center of the product. The effect of two doses of injection (5% and 10%) and three concentrations of STPP (0%, 3%, and 5%) were studied. The results of the sensory evaluation showed that pork steaks injected with a solution of 5% STPP at a dose of 10%, and cooked to a temperature of 80 °C exhibited better tenderness, but the juiciness of the meat remained unchanged.4. ConclusionsThe use of malic acid and malic acid with the subsequent addition of phosphates for marinating meat, in order to lower its pH value, is particularly recommended for the raw material obtained from carcasses of young horses. On the other hand, for horse meat obtained from carcasses of older animals, the use of phosphates is advisable, as it allows the pH value of the meat to be increased. With the age of horses, the values of cutting force, hardness, stiffness, gumminess, and chewiness of the meat increase (p < 0.05). This is most likely caused by the increase in the amount of collagen and its cross-linking in the meat of older horses. The present study showed that the application of lactic acid and malic acid for marinating the meat of young horses caused a decrease in the proportion of red color and an increase in the proportion of yellow color, especially after 3 months of frozen storage. In turn, an increase in the value of the a* parameter and most often a decrease in the b* parameter were observed with the use of phosphates for marinating meat from the carcasses of older horses. Each of the substances used for marination caused a decline in the hydration properties of horse meat. However, the lowest values of forced and thermal leakage from meat were achieved with the use of phosphates and salt.

animals : an open access journal from mdpi

10.3390/ani12020136

PMC8772551

Milk production is an important trait in the breeding and genetic improvement of Xinjiang Brown cattle. To obtain the best strategy for improving the reliability of the breeding value estimation for each trait, we used single-trait and multitrait models based on the A-array pedigree-based best linear unbiased prediction (PBLUP) and H-array single-step genomic best linear unbiased prediction (ssGBLUP) to perform the genetic evaluation of different strategies using the restricted maximum likelihood (REML) and Bayesian methods. Upon comparison, the ssGBLUP calculation results of the multitrait models obtained using the REML and Bayesian methods were better than those of other strategies. Considering the calculation time, the multitrait model REML method is recommended for ssGBLUP calculation to accurately predict the breeding value of young animals; thus, this strategy should be used for the early breeding selection of Xinjiang Brown cattle.

One-step genomic selection is a method for improving the reliability of the breeding value estimation. This study aimed to compare the reliability of pedigree-based best linear unbiased prediction (PBLUP) and single-step genomic best linear unbiased prediction (ssGBLUP), single-trait and multitrait models, and the restricted maximum likelihood (REML) and Bayesian methods. Data were collected from the production performance records of 2207 Xinjiang Brown cattle in Xinjiang from 1983 to 2018. A cross test was designed to calculate the genetic parameters and reliability of the breeding value of 305 daily milk yield (305 dMY), milk fat yield (MFY), milk protein yield (MPY), and somatic cell score (SCS) of Xinjiang Brown cattle. The heritability of 305 dMY, MFY, MPY, and SCS estimated using the REML and Bayesian multitrait models was approximately 0.39 (0.02), 0.40 (0.03), 0.49 (0.02), and 0.07 (0.02), respectively. The heritability and estimated breeding value (EBV) and the reliability of milk production traits of these cattle calculated based on PBLUP and ssGBLUP using the multitrait model REML and Bayesian methods were higher than those of the single-trait model REML method; the ssGBLUP method was significantly better than the PBLUP method. The reliability of the estimated breeding value can be improved from 0.9% to 3.6%, and the reliability of the genomic estimated breeding value (GEBV) for the genotyped population can reach 83%. Therefore, the genetic evaluation of the multitrait model is better than that of the single-trait model. Thus, genomic selection can be applied to small population varieties such as Xinjiang Brown cattle, in improving the reliability of the genomic estimated breeding value.

1. IntroductionXinjiang Brown cattle was the first breed of cattle used for milk and meat purposes after the founding of the People’s Republic of China [1]. The breeding industry of Xinjiang Brown cattle accounts for a large proportion of the local economic development as well as farmers’ and herders’ income. In 2018, the population of Xinjiang Brown cattle reached 1.5 million [2]. However, compared with the Holstein cattle, genetic improvement technologies for Xinjiang Brown cattle are relatively behind. Xinjiang Brown cattle are a unique species in Xinjiang, where for a long time, the breeding value for the milk production traits of this cattle breed was estimated using pedigree-based best linear unbiased prediction (PBLUP) through the construction of the additive genetic relationship matrix (A matrix) [1]. However, with the reduction in sequencing cost, cattle breeding has entered the genome era [3].The large-scale breeding of Xinjiang Brown cattle population is limited, and the number of the grazing or semihouse feeding and semi-grazing population is high. Although the production performance database of core breeding farms has been preliminarily established, the limitations of Xinjiang Brown cattle production performance data, coupled with the lack of pasture management and the presence of a significant amount of truncated data, have led to the genetic evaluation of Xinjiang Brown cattle in failing to achieve better reliability in terms of genomic estimated breeding value, thus influencing the accurate selection of excellent breeding animals. Compared with PBLUP, genome prediction can reduce the breeding cost and generation interval by at least 50% [4], leading to faster genetic progression. Genomic selection (GS) technology can also improve the reliability of genetic evaluation; therefore, it is imperative to apply it in genetically evaluating Xinjiang Brown cattle. A study on the GS of cattle mainly focuses on the reliability of models, methods, and genomic estimated breeding values [4]. On the basis of a previous calculation of genomic best linear unbiased prediction (GBLUP) using the genetic-pedigree joint relationship matrix (G matrix), single-step GBLUP (ssGBLUP) was proposed [5]. Furthermore, it is more comprehensive than GBLUP in genetically evaluating individuals [6]. ssGBLUP can simultaneously use the phenotype and pedigree information of genotype and non-genotype animals [7,8], and combine the A and G matrices to construct a genetic-pedigree relationship matrix (H matrix) [9,10,11] to perform a one-step genetic evaluation. For example, Li et al. [12] used a one-step method to estimate the milk production trait of Holstein cattle in China and increased the accuracy of estimation by 0.12 compared with the two-step GBLUP method. However, the study also found that adjusting the proportion of G and A matrices in the one-step method did not improve reliability [13,14]. Studies using different livestock species further confirmed that ssGBLUP was more reliable than BLUP and GBLUP in estimating GEBV [15,16,17,18].The multitrait BLUP (MBLUP) has been studied for decades. It is a method for the genetic evaluation of individuals for two or multiple traits using information such as the phenotypic and genetic correlations of traits. An advantage of MBLUP [19] is the increased accuracy of genetic evaluation. The multitrait genetic evaluation of the REML and Bayesian methods is widely used in cattle breeding [20]. Studies have shown that the accuracy of the breeding value estimated by the multitrait model is twice that of PBLUP [21]. The accuracy of estimation based on the G-matrix multitrait model was also 3% higher than that of the single-trait model [22], where many studies have reported the advantages of multitrait genetic evaluation or GS [23,24,25,26].Therefore, this study aimed to use the REML and Bayesian methods to estimate the genetic parameters of the milk production traits of Xinjiang Brown cattle using the A and H matrices, respectively, with a goal to ultimately obtain the variance components and genomic breeding values of each trait of the single- and multitrait models; this will improve the reliability of the estimated values and provide theoretical support for a more accurate estimation of the genetic parameters of the Xinjiang Brown cattle population.2. Materials and Methods2.1. Data Source and ProcessingData of Xinjiang Brown cattle were obtained from 7516 production performance measurement records and 16,795 pedigree records of the abovementioned four Xinjiang Brown cattle breeding pastures (The Xinjiang Tianshan Animal Husbandry Bio-Engineering Co.,Ltd, Xinjiang Uygur Autonomous Region local state-owned Urumqi cattle farm, Yili Xinjiang Brown cattle farm, Tacheng Agriculture and Animal Husbandry Technology Co. LTD, China) from 1983 to 2018 as well as DHI measurement records from 2010 to 2018. The pedigrees of Xinjiang Brown cattle breeding bulls, Xinjiang Brown cattle adult cows, and Swiss brown cattle breeding bulls were traced as well. This included 676 Xinjiang Brown cattle breeding bulls, among which 1 breeding bull had the most offsprings (619 offsprings), and 221 breeding bulls had only one offspring; 583 adult cows of Xinjiang Brown cattle had only one offspring, and 6199 adult cows of Xinjiang Brown cattle had ≥2 offsprings; the maximum number of offsprings was 12. The following milk production traits were obtained after sorting: 305 daily milk yield (305 dMY), milk fat yield (MFY), milk protein yield (MPY), and somatic cell score (SCS). SCC conversion to SCS is based on the formula determined by the American Dairy Cattle Improvement Program Board: SCS=log2SCC105+3.Effect Division: The field effect was divided into four levels, where the calving year was divided into seven levels according to phenotypic records: 1985–1995, 1996–2000, 2001–2005, 2006–2008, 2009–2011, 2012–2014, and 2015–2018; the calving season effect, according to the unique climatic conditions in Xinjiang, was divided into spring (April and May), summer (June, July, and August), autumn (September), and winter (January, February, March, October, November, and December) using the pentad mean temperature method; the birth order effect was divided into six levels: 1, 2, 3, 4, 5, and 6 (including >6 births).2.2. Genotyping DataThe chip data included 403 Xinjiang Brown cattle cows, 71 Xinjiang Brown cattle bulls, and 11 Swiss brown cattle bulls, where a total of 139,376 single nucleotide polymorphism (SNP) loci were detected using the GeneSeek GGP Bovine 150 K chip (Illumina). The Beagle v.4.1 software was used for imputation, which mainly inferred the presence of haplotypes in the population based on the principle of linkage disequilibrium. Therefore, the quality of the chip data needed to be controlled to ensure the accuracy of imputation. Quality control standard: the SNP loci with an individual genotyping detection rate of <90%, individual genotype deletion rate of <10%, minor allele frequency of >0.01, and Hardy Weinberg equilibrium p-value of >1 × 10−6 were excluded. Finally, 118,021 SNP loci of Xinjiang Brown cattle were retained for subsequent analysis.2.3. Statistical Analysis2.3.1. Estimation of Genetic Parameters of the Single-Trait ModelDifferent relationship matrices (A or H matrix) were constructed using single-trait models, PBLUP and ssGBLUP, and the genetic parameters of milk production traits of Xinjiang Brown cattle were estimated using the REML method.Matrix expression formula of each trait model is shown below:Y=Xb+Za+e where Y is the observed value vector of each trait (including 305 dMY, MFY, MPY, and SCS), b is the fixed effect vector (field, calving year, calving season, and birth order), a is the random additive genetic effect vector, e is the residual effect vector, X is the fixed effect coefficient matrix, and Z is the random additive genetic effect coefficient matrix.For PBLUP [27] or ssGBLUP [28], assume a~N0,A or Hσa2, e~N0,Iσe2, where A is the additive genetic relationship matrix based on the pedigree, H is the genomic-pedigree joint relationship matrix, σa2 represents additive genetic variance, I is the unit matrix, and σe2 represents the variance of random error effect. With given A and G values, H is calculated using the following formula:H=A11−A12A22−1A21+A12A22−1GA21A12A22−1GGA22−1A21G where subscripts 1 and 2 of A represent the populations of non-genotyped and genotyped animals, respectively; G is the genomic relationship matrix, and its calculation formula is G=MM′2∑k=1mpk1−pk; M is the incidence matrix of SNP effect, where elements 0−2pj, 1−2pj, and 2−2pj represent 11 homozygotes, 12 or 21 heterozygotes, and 22 homozygotes, respectively; pj is the minimum allele frequency of the jth SNP; m is the number of markers; and pk is the allele frequency of the Kth SNP. Therefore, the formula for H−1 is H−1=A−1+000G−1+A22−1, where A−1 is the inverse matrix of all pedigree relationships; G−1 is the inverse matrix of genomic relationship; and A22−1 is the inverse matrix of the pedigree relationship of genotyped individuals (the same letter in the multitrait model represents the same meaning).2.3.2. Estimation of Genetic Parameters of the Multitrait ModelThe REML and Bayesian methods were used on the multitrait animal model to perform genetic parameter estimation based on PBLUP and ssGBLUP, where the matrix form of the multitrait model [29] is as follows:y1y2y3y4=X10000X20000X30000X4b1b2b3b4+Z1000⋮Z20000Z30000Z4a1a2a3a4+e1e2e3e4 where yi is the observed value vector of all individuals, bi is the fixed effect vector of the ith trait, ai is the additive genetic effect vector of the ith trait, ei is the residual random effect vector of the ith trait, and Xi and Zi are the relationship matrices of β and ai, respectively. Assume a1⋱an~N(0,A or H⊗σa12⋯σa1an⋮⋱⋮Symmetric⋯σan2, e1⋱en~N(0,I⊗σe12⋯σe1en⋮⋱⋮Symmetric⋯σen2, where A is the additive genetic effect matrix; σai2 and σei2 are the additive genetic variance and random error effect variance of the ith trait, respectively; σaiaj and σeiej(i ≠ j) are the additive genetic covariance and random error effect covariance between the ith and jth traits, respectively. The formula for H−1 is H−1=A−1+000GW−1+A22−1.The REML method of the REMLF90 module of BLUPF90 software [30] was used to estimate the variance components. After the calculation results converged, AIREMLF90 was used to run for 0 times and the approximate standard errors of all calculation parameters were obtained based on the algorithm implemented by Meyer and Houle in the AIREMLF90 program. As an alternative of SE, calculate SD for function of (co)variances by repeated sampling of parameter estimates from their asymptotic multivariate normal distribution, following ideas presented by Meyer and Houle 2013 [31]. The Bayesian method was followed using the GIBBS1F90 module in the BLUPF90 software and the Bayesian Gibbs sampling method. In the Bayesian method, the total strand length of samples was 100,000 and the length of the preheating strand was 10,000, with a sparse interval of 50. The Geweke diagnostics method of POSTGIBBSF90 module of BLUPF90 software was used to check the convergence of the Gibbs chain. 2.3.3. Calculation of HeritabilityThe calculation formula of heritability is as follows:h2=σa2σa2+σpe2+σe2 where h2 is heritability, σa2 is the additive genetic variance, σpe2 is the permanent environmental variance of individual, and σe2 is the residual variance.The formula for the standard error of heritability is shown below:SE2(h2)=σa2σp2Varσa2σa22+Varσp2σp22−Covσa2,σp2σa2σp2 where SE2h2 is the standard error of heritability, σa2 is the additive genetic variance, and σp2 is the overall phenotypic variance, σp2=σa2+σpe2+σe2.2.3.4. Reliability of Breeding Value EstimationThe definition of reliability of GEBV was:RGEBV2=Cor2GEBV,a=Cov2GEBV,aVGEBVVaIf GEBV is unbiased, a=GEBV+ε, CovGEBV,ε=0, ε is prediction error. CovGEBV,a=CovGEBV,GEBV+ε=VGEBVRGEBV2=Cov2GEBV,aVGEBVVa=VGEBVVa where a and ε are independent, RGEBV2 is the reliability of genomic estimated breeding value, Cor2GEBV,a is the square of the correlation coefficient between genomic estimated breeding and actual values. The reliability of animal breeding values was calculated according to prediction error variance as Va=VGEBV+PEVRGEBV2=VGEBVVa=1−PEVVa where PEV is the prediction error variance and Va is the direct additive genetic variance.3. Results3.1. Descriptive Statistical Analysis of Each TraitTable 1 lists the sample number of observed values, minimum value, maximum value, mean value, standard deviation, and coefficient of variation of the milk production and reproductive traits of Xinjiang Brown cattle. Figure 1 shows that the data frequency distribution of the milk production traits of Xinjiang Brown cattle followed a normal distribution. The average milk production of Xinjiang Brown cattle in 305 days was 4216.49 kg, which was similar to that of Sanhe cattle and Chinese Simmental cattle but lower than that of dual-purpose cows such as European Fleckvieh cattle and Swiss brown cattle [32]. The mean MFY and MPY of the population were 168.53 kg and 143.79 kg, respectively, which were lower than those of Fleckvieh cattle and Swiss Brown cattle [33], and similar to those recently reported Italian Simmental cattle [34] SCS can be used as an indicator of breast health, where the lower the SCS value, the lower the risk for mastitis. These traits can reflect the production efficiency and health status of dairy cows and are also important goals for the breeding of Xinjiang Brown cattle.3.2. Estimation of the Genetic Parameters of Milk Production TraitsTable 2 shows the estimation results of the genetic parameters for the milk production traits of Xinjiang Brown cattle obtained using the single-and multitrait models based on PBLUP and ssGBLUP using the REML or Bayesian methods. The heritability of milk production traits of Xinjiang Brown cattle that was calculated on the basis of PBLUP and ssGBLUP using the multitrait models REML and Bayesian methods was higher than those of each trait calculated using the single-trait model REML method; moreover, the standard error of heritability of the former model was lower than that of the latter. However, the heritability results of the milk production traits of Xinjiang Brown cattle calculated using the REML and Bayesian methods based on PBLUP and ssGBLUP of the multitrait model were similar; Gibbs parameter chains of additive genetic variance of milk production traits under different relationship matrices were obtained using the Bayesian method. Geweke diagnostic test results show that the convergence trend of iteration curve indicates that the parameter chain converges. The heritability of 305 dMY, MFY, and MPY estimated by the two methods in the multitrait model was approximately 0.39 (0.02), 0.40 (0.03), and 0.49 (0.02), respectively, which indicated high heritability traits (h2 > 0.3). However, the heritability of SCS was approximately 0.07 (0.02), which suggested a low heritability trait (h2 < 0.1).3.3. Reliability of Breeding Value Estimation of Milk Production TraitsTable 3 compares the estimated breeding value (EBV), GEBV reliability, and increased reliability (Δreliability) of the overall and genotyped populations of Xinjiang Brown cattle that were calculated via different methods and models. Results showed that the reliability of GEBV calculated using ssGBLUP was higher than that of EBV calculated using PBLUP in the overall and genotyped populations regardless of whether it was based on the single- or multitrait model. Comparing with the multi- and single-trait models of the REML method, the reliability of EBV estimated using the multitrait model was higher than that estimated using the single-trait model; the reliability of GEBV estimated using the multitrait model was higher than that estimated using the single-trait model. However, comparing the REML and Bayesian methods of the multitrait model, the reliability of EBV and GEBV of SCS calculated using the Bayesian method was slightly higher than that calculated using the REML method in the overall and genotyped populations. In contrast, the EBV and GEBV reliability results of other traits were similar (Figure 2). In all results, the Δreliability of the genotyped population was higher than that of the overall population and the Δreliability of the Bayesian method was slightly higher than that of the REML method.4. Discussion4.1. Analysis of Genetic Parameters of Milk Production TraitsIn this study and the one by Zhou Jinghang, the heritability of 305 dMY and SCS estimated using the REML method of the DMU software based on the pedigree data of Xinjiang Brown cattle was 0.40 (0.017) and 0.08 (0.009), respectively, which were both higher than the heritability results of PBLUP calculated using the REML and Bayesian methods in this study; the standard error of the former was lower. However, the heritability results were similar to those of ssGBLUP calculated using the REML and Bayesian methods in this study. The estimated heritability of MFY and MPY was 0.30 (0.013) and 0.20 (0.011), respectively, which were significantly lower than the results estimated using the REML and Bayesian methods for different genetic relationship matrices of the multitrait model in this study [35]. In addition, the heritability of somatic scoring was lower than that of the American and Italian brown cattle populations (0.12 [36] and 0.14 [37], respectively).In this study, the heritability obtained using the REML and Bayesian methods in the multitrait model was higher than that in the single-trait model. Furthermore, in all model methods, except for the decrease in SCS, the heritability of other traits calculated based on ssGBLUP was slightly higher than that calculated based on PBLUP [29]; the standard errors of the heritability of the two were similar. The heritability in genetic parameter estimation based on the G matrix was lower than that based on the A matrix [38,39]. During ssGBLUP estimation, the slight increase in heritability was mainly reflected in the increase in additive genetic variance components [40,41]. The additive genetic variance difference between PBLUP and ssGBLUP estimations was primarily due to the construction of the A and H matrices being different, and the scale of diagonal elements was also different [42]. For example, the individual breeding value, ai=0.5as+ad+mi, where as is the breeding value of the father, ad is the breeding value of the mother, and mi is the Mendelian sampling deviation of individuals. For the A matrix, its diagonal is the expected value of the genetic relationship coefficient between individuals. In contrast, the diagonal in the G matrix is the real genetic relationship coefficient between individuals considering the Mendelian sampling deviation [43,44]. For example, when using paternal or maternal phenotypic information to estimate the breeding value, if the genetic relationship of the A matrix is 0.5, the genetic relationship of the G matrix will be ≥0.5 or ≤0.5; the H matrix combines the A and G matrices. Several studies have shown that GBLUP and ssGBLUP are better than PBLUP. The main reason is that the G matrix shows a more realistic genetic relationship between individuals than the A matrix [44,45,46]. Therefore, the parameter estimation of the A and H matrices may have deviations [47]. In addition, many studies conducted in China and overseas have confirmed that the REML method is relatively ideal for estimating genetic parameters of livestock breeding [48]. As the Bayesian method can consider prior information of unknown parameters and provide accurate posterior distribution for a limited sample size [49], it is favorable for the genetic evaluation of small breed populations when there is a large amount of historical data available [50]; however, its calculation time is relatively long.4.2. Predictive Analysis of EBV and GEBV ReliabilityWhen using the REML method of the single-trait model, only the reliability of the breeding value of 305 dMY was >0.4. The reliability of the breeding value of milk production traits estimated based on PBLUP and ssGBLUP of the REML and Bayesian methods in the multitrait model ranged from low to moderate. In general, the increased ratio of the reliability of genotyped population was 0.9%–3.6% higher than that of the overall population, and the GEBV reliability of the genotyped population was up to 83%. Compared with PBLUP, ssGBLUP method with chip information added significantly and simultaneously improved the reliability of breeding value estimation in both REML and Bayesian methods (Figure 2), which indicated that GS using ssGBLUP is feasible for the genetic evaluation of Xinjiang Brown cattle. Although the Bayesian method of the multitrait model is slightly better than the REML method, using the REML method to calculate ssGBLUP can save more calculation time.Relying on the linkage disequilibrium between SNP markers and quantitative trait locus of target traits, genomic prediction promotes the association between SNP markers and individual phenotypic values. Thus, the accurate estimation of SNP’s genetic effect based on the phenotypic value has become a key factor in genomic prediction; the accuracy of genomic prediction [51] can be improved by increasing the number and accuracy of phenotypic values (by expanding the reference population) [52]. Genotyping candidate animals with different SNP loci and chip densities can also influence the accuracy of GS [53]. In the ssGBLUP method, the genomic breeding value estimation accuracy is higher for traits with a larger number of animal phenotypes and genotypes and those with high heritability [54].With increased reference population size and improved pedigree data integrity, the accuracy of EBV or genomic breeding estimation increases [55]. However, there are relatively few genotyped populations of Xinjiang Brown cattle (genotyped individuals account for only 2.89%), which is the key factor restricting the reliability of ssGBLUP estimation of breeding value. Furthermore, some studies have reported that the reliability of low heritability traits is greatly improved when GBLUP is used [56]. In studies using different livestock [15,17,57,58,59], the same conclusion that ssGBLUP can more accurately predict the genetic value of animals than the classical PBLUP method was drawn. In addition to increasing the number of reference groups of this variety, some studies have shown that the combined reference population significantly improves the reliability of GEBV for the populations of four countries. Compared with the use of the domestic population alone as a reference, the reliability of GEBV is increased by an average of 10% when different reference population sizes are used for different countries and traits. In addition, expanding the reference population size can significantly improve the reliability of GEBV [50] by 2–19% [52,60]. Therefore, in the future, the cross variety GS of Xinjiang Brown cattle can also be used as a strategy to expand the reference population and improve the reliability of the breeding value estimation.5. ConclusionsIt is feasible to use ssGBLUP to perform the genomic evaluation of the milk production traits of Xinjiang Brown cattle. However, the single-trait model does not consider the covariance between traits, and thus has a large error. The use of the multitrait model, regardless of the use of PBLUP or ssGBLUP, can greatly improve the reliability of the breeding value estimation. Moreover, the estimated genetic parameters provide a basis for calculating a more accurate breeding value of Xinjiang Brown cattle. However, if it is needed to obtain a higher level of breeding value reliability, the number of genotypes needs to be simultaneously expanded when using high-density chips (150 k). Nevertheless, applying the ssGBLUP method to the genetic evaluation of Xinjiang Brown cattle necessitates further study and analysis. The next challenge is to build a joint reference group through cross variety GS to expand the number of genotyped animals with the objective to improve the reliability of the breeding value estimation of various traits of Xinjiang Brown cattle.

animals : an open access journal from mdpi

10.3390/ani11102816

PMC8532781

Elephant Endotheliotropic Herpesvirus hemorrhagic disease (EEHV-HD) is considered the primary cause of calf mortality in the global captive Asian elephant population. Once thought to be exclusively a zoo problem, EEHV-HD is currently acknowledged as a disease also present in wild populations, although the extension of this threat in some free-range countries is still poorly understood. The disease is characterized by an acute hemorrhagic syndrome due to vast endothelial destruction combined with disseminated intravascular coagulation, leading to the sudden death of mainly young elephants. In this study, we aimed at understanding the impact of EEHV-HD in the European captive Asian elephant population and acquiring a better understanding if hereditary or environmental factors could be linked to the manifestation of this disease. The findings of this investigation suggest the involvement of zoo-associated factors with possible sire or dam (or a combination of both) influence on the onset of the disease. This knowledge points us to the importance of continuous retrospective epidemiological studies and stresses the great importance of finding further underlying factors for the development of this disease if we wish to halt the high number of deaths caused by this hemorrhagic disease.

EEHV is a ubiquitous virus, which most likely has co-evolved with elephants and is shed by healthy individuals and maintained in the herds. Yet, the factors determining calf susceptibility to the virus remain unknown. Here, we explored the impact of EEHV-HD in the European captive Asian elephant population in a retrospective statistical study spanning the last 35 years. We show that EEHV-HD was implicated in more than half of all deaths recorded in calves older than one months old. Moreover, the median age across EEHV-HD fatalities was significantly lower compared to other death causes. Finally, we investigated if heredity and zoo-associated factors could be linked to a higher susceptibility of calves to this disease. We used a univariable logistic regression model to evaluate if either fathers, mothers, or zoos could, separately, be considered as risk factors to the development of the disease. Afterwards, we used a two multivariable model, combining: (1) fathers and zoos, and (2) mothers and zoos. Overall, we found that two fathers, one mother, and four zoos had three or more times higher risk of their calves becoming sick when compared to all others, pointing us to the presence of a management or environmental element, which can have paternal and maternal influence and leads to calf susceptibility or resistance to EEHV-HD.

1. IntroductionElephant Endotheliotropic Herpesvirus (EEHV) was initially reported in the captive Asian elephant population in 1990 after a three-year-old elephant calf died from an acute hemorrhagic disease (HD) [1]. At necropsy, a severe generalized hemorrhagic condition due to vascular endothelial lesions was observed [1]. Diseased elephants experience a rapid and systemic spread of the virus, followed by vascular endothelial cell damage associated with an uncontrolled virus replication [2,3]. This fulminant disease affects mainly very young calves, often leaving little or no time to provide adequate veterinary treatment [4,5]. Multiple EEHV genotypes and strains have been reported, with EEHV 1 being the most impactful [6,7,8,9]. In the European population, 80% of the calves’ EEHV-related deaths were reportedly caused by subtype EEHV1a [10].EEVH-HD is considered to be an ancient infection among Asian elephants {Formatting Citation} and is not a disease exclusive of this species as it may also affect African elephants. However, the recorded mortality rate in African elephants is lower, and the animals seem to present symptoms at an older age [11,12,13]. Currently, the most used antiviral treatment is a human anti-herpetic drug, despite its high costs and reported as presenting unproven efficacy, so far [4,14,15].Once thought to be an exclusive zoo disease, fatal cases due to EEHV-HD have been reported in several range countries, such as India [9,16], Thailand [2,6,17], Cambodia [18], Laos [19], Myanmar [8], Nepal, and Sumatra [7]. The prevalence of EEHV-HD in wild populations is expected to be high, since the medical veterinary teams working in close association with these populations have found evidence of this disease, during necropsies. However, due to a lack of logistic capacities, further investigations have been hampered [20]. In North American zoos, reports show that 53% of deaths since 1980 in their Asian elephant population were caused by EEHV-HD, while in Europe this accounts for 60% of the total deaths since 1995 [20]. Additionally, North American institutions reported that the virus presents a mortality rate of 68% [20]. In 2016, 40% of elephants’ deaths in the UK and Ireland were caused by EEHV-HD with an overall population mortality of 21.6% [4], making this the major mortality cause in both continents [20]. In range countries, such as India, a prevalence study showed that at least one of the EEHV variants is present in 35% of their captive Asian elephants [21]. Moreover, in Thailand, a seroprevalence of 42% was found (in private, touristic, and logging elephant camps [22], showing that EEHV is also maintained within the captive population. Most infectious diseases run a subclinical course and only part of the population will present clinical disease, where the mutual interactions between environment, host, and pathogen genetic factors, influence this ratio [23]. To similarity, EEHV-HD must also be influenced by the elephant host genetics and environmental pressures, being the presence and pathogeny of the virus alone, not the only determinant factor.Even though this disease has been under study for the past three decades, and a significant number of discoveries were recently made on its pathophysiology [2,24], the adequate treatment, and the epidemiological impact of it in the overall world elephant population is still not fully understood. Therefore, having a deeper understanding of the virus’ mechanism of action is yet of the highest priority. Moreover, there is an urgent need to identify what risk factors are involved in the onset of the disease, in order to establish proper actions to protect the calves.This study aims to assess the impact of EEHV-HD in the European captive Asian elephant population and to explore risk factors linked to a higher prevalence of the disease, such as gender, age, genetic lineage, and location. To address these, we used historical and current data from all captive calves born in Europe from January 1985 to June 2020, conducting the longest, retrospective, and longitudinal observational study so far. The disease seems to affect calves from different genetic backgrounds and breeding facilities at a different rate: while some are profoundly impacted by this hemorrhagic disease, others are minimally or not affected. Therefore, we hypothesize that hereditary (host genetics) and different zoo-associated factors (e.g., management protocols and growing environment) may protect calves against the potentially fatal outcome of the disease.2. Materials and Methods2.1. Data CollectionTo be able to identify the impact of EEHV-HD, regardless of the virus genotype, in the captive-born Asian elephant population in Europe and investigate the risk factors associated with high mortality, we compiled a dataset of all animals kept in captivity at European zoos, spanning the last 35 years, from January 1985 to June 2020 (n = 330, supplementary materials, Table S1—Study population database). This dataset comprises exact birth and death dates, maternal and paternal information, location, and the present status of the elephants (alive, dead by other causes, or dead by EEHV-HD), as well as EEHV infection reports. We collected information from the Asian elephant European Association of Zoo and Aquaria ex situ Programme (EEP, formerly European Endangered Species Programme) Studbook yearly reports, from Zoological Information Management system (ZIMS), from personal contacts with the zoological institutions, from up-to-date registers documented on zoo websites, and from information compiled at elephant large online databases.2.2. Data Cleaning, Selection and AnalysisThe starting year of the analysis (1985) was chosen to match the year when the first reported EEHV-HD fatal case was born—Lohimi, a female calf born in a circus, that presented a hemorrhagic syndrome in 1988, which led to her death, at the age of three years [1]. Since the population in the study were captive European Asian elephants, only calves born in captivity were kept in the data set, and all wild-born animals were removed from the study. Thus, non-European captive calves that were translocated to Europe afterwards were also not considered for analysis.Neonatal mortalities and early life deaths accounted for 24.8% of the total deaths due to several causes (e.g., miscarriages, abortions of twinning, stillbirths, surgically removed fetuses, infanticide, rejected by the mother). On this account, a subset of our initial population was created, including only records of successful births and minimal management to ensure a correct adaptation to the first months of life (e.g., proper feeding and non-life-threatening congenital defects). Animals that did not survive to reach two months of age (n = 83; n = 77 under one week and n = 6 dying in their first month of life) were excluded from this dataset. Under this threshold, three animals were mentioned as possible EEHV-HD deaths, presenting low titters of the virus, being stillborn, or having succumbed under 24 h after parturition. These deaths could not be clearly attributed to EEHV-HD and were removed.The frequencies of births, deaths due to EEHV-HD, and deaths due to other causes per year of study are shown and their distributions were evaluated. We investigated the trends of distribution according to age for each status (status 0 = alive, status 1 = death by EEHV-HD, and status 2 = death by other causes) for all captive-born elephants. Standardized residuals were visually assessed and were not fully normally distributed, therefore, a non-parametric Kruskal–Wallis test was used to compare median ages between groups.The association of gender with the overall survival time for the entire population in the study and within the EEHV-HD reported cases was investigated using the log-rank (Mantel–Cox) test. Afterwards, a survival analysis (Kaplan–Meier curve) was performed to compare the survival time between the animals that presented EEHV-HD disease (that survived or died) and all others that never presented symptoms.Finally, to test if hereditary lineage and/or the environment could be potential risk factors to the survival of the elephants in captive populations, we categorized all calves by fathers, mothers, and location during calfhood. The identities of the bulls, dams, and zoos will remain anonymous in our study.An explorative univariable logistic regression model was performed to separately assess the odds ratio (OR) of EEHV symptomatic calves for individual (i) fathers, (ii) mothers, and (iii) zoos. Parents and zoological institutions included in the analysis were grouped according to the number of calves produced. Bulls that sired more than ten calves were kept individually while all bulls that sired fewer calves were collapsed into a single group (bulls that sired less than ten offspring). For dams and zoos, the cut-off for keeping them individually was five calves. Fathers, mothers, and zoos with a lower number of calves were considered the baseline for comparison. Afterwards, two multivariable models estimated simultaneously the OR of (i) fathers and zoos as well as (ii) mothers and zoos combination. Results were screened for OR greater than 6.0 which indicates a sixfold higher chance of presenting EEVH sick calves when compared to the baseline category.All statistical analyses were performed considering an alpha level for significance and tendency of 0.05 and 0.10, respectively. Analyses were conducted using IBM SPSS (IBM SPSS Statistics for Windows, version 24.0, Armonk, New York, NY, USA) predictive analytics software and graphs were produced using GraphPad Prism (version 9, GraphPad Software, San Diego, CA, USA).3. Results3.1. Descriptive AnalysisA total of 247 captive-born Asian elephants (females = 116, males = 131) were born between January 1985 and June 2020 and survived more than one month of life. These births occurred in a total of 48 European zoological institutions and animals are now distributed in 68 zoological locations, due to transfers between zoos. A total of 72.1% of the population monitored since 1985 never presented the disease and are still thriving at the moment of writing. We found that 15.8% (n = 39) of the calves were infected and symptomatic for EEHV-HD. Of this percentile, 13.4% were lost to the hemorrhagic disease, and therefore, so far, only 2.4% (n = 6) of the affected calves managed to resist and survive this disease.A total of 25.5% (n = 63) of the population died within the study period due to several different causes (e.g., foot disease, infectious diseases—including EEHV-HD, tumors, etc.). Accordingly, EEHV-HD is the primary cause of death above one month of age in the European Asian elephant population, producing 52.5% of all reported deaths (n = 33).We found that only in 1988 no births were registered, and one death was reported, presenting, therefore, a negative balance for that specific year. Moreover, except for 1987, 2015, and 2018 where the number of births and deaths was the same, the number of offspring per year exceeds the number of deceased animals (Figure 1).3.2. Survival Age and Gender RelationThere was no impact of gender in the survival time of Asian elephants born after 1985 in Europe (p = 0.813) and EEHV-HD fatalities were also not gender related, with an almost 1:1 relationship (females n = 17, males n = 16). Moreover, males (n = 131) were found to be younger than females (n = 116); the overall male median age was around 24 years of age while the female average rounded 30 years.The animals which died from various non-EEHV-HD-related causes (n = 30), lived between two months and 23 years (median = 8.6 years). For the EEHV-HD fatal cases (n = 33), the earliest related death occurred at 9 months old, and the oldest animal died at 7.6 years of age resulting in a very narrow age range. Additionally, deaths due to this virus occurred at a significantly lower median age (2.7 years old) when compared to the median age of elephants that died due to other causes (8.6 years old) (Figure 2).Pairwise comparisons between EEHV fatal cases and animals that are alive revealed a significantly lower age of life for the diseased animals (p < 0.001). The same results were found for the comparison between animals dead due to other causes and those that succumb to EEHV (p = 0.007). The median age did not differ between the living animals and those that died due to other causes (p = 0.057).Kaplan–Meier analysis revealed that the survival curve of the animals that presented EEHV-HD and the survival curve of the other individuals that never presented symptoms are significantly different (p < 0.001, Figure 3). The median survival age of EEHV-HD symptomatic animals was 35 months, while animals with no reported EEHV-HD presented a median age of 122 months.3.3. Father and Mother Distribution of EEHV-HD Fatal CasesWhen investigating the distribution of the fatal EEHV cases per high breeders, we found that some fathers presented no loss of their offspring due to EEHV-HD (e.g., fathers F2, F3, F4; Figure 4) or minimal loss (e.g., father F8, Figure 4), while others, with nearly the same number of calves, have lost a high percentage of their calves (e.g., fathers F9 with 42% and F7 with 38% of calf loss due to EEHV-HD; Figure 4).From all the fathers analyzed (n = 45), 11 bulls had ten or more calves each. These animals have produced nearly 60% (n = 144) of the entire population present in the study population and were the ones used for the subsequent analysis of parental risk. Calves born to two specific fathers with a high frequency of offspring presented a significant increase associated risk to present EEHV-HD (F7, OR = 3.8, p = 0.03; F9, OR 4.4, p = 0.02) when compared with other sires.Maternal contribution (n = 97) to the overall deaths of the calves was also investigated; however, there is a very low frequency of births registered per dam when compared with the high offspring number presented by the fathers. One mother presented an increased tendency for her calves to have the disease when compared to all other mothers in the study (OR = 3.8, p < 0.1).3.4. Zoo Distribution of EEHV-HD Fatal CasesOur survival comparisons based on the living location (n = 68 zoos) of the calf showed that high breeding zoos (n = 18) that produced five or more calves conceived a total of 140 calves. The remaining 50 locations presented a lower breeding rate and produced 107 offspring, with the majority of the zoos having produced one or two calves.Similar to the distribution found for the fathers, we observed that some institutions have suffered high losses. When investigating only the zoos that bred five or more times, we found that some of these locations were not affected at all, while others present an overall offspring loss due to EEHV as high as 50% of the total offspring born at a particular zoo (e.g., zoos Z11 and Z6; Figure 5).We found that three institutions presented a significantly increased odds ratio, between 8 to 12 times higher, for their calves to present EEHV-HD (Z17, OR 11.8, p = 0.01; Z2, OR 10.6, p < 0.001; Z6, OR 7.9, p = 0.007), than the other zoos in the study.In the multivariable model with fathers and zoos, we found that F9 and Z17 presented a significant increased OR for presenting calves with the disease (OR 6.2, p = 0.04 and OR 19.1, p = 0.016, respectively) and Z9 had an OR > 6.0 (p = 0.086). When combining with mothers, we find that zoos Z2, Z6, and Z17 present a significantly higher probability of reporting calves with EEHV-HD (OR > 6.0, p = 0.012, p = 0.002, and p = 0.013, respectively).On another analysis, a cross-tabulation of all fatal cases caused by EEHV-HD by the respective fathers (n = 18) and locations (n = 18) showed deaths attributed to different sires at the same zoo. Likewise, different calves fathered by the same sire but living in different institutions were also lost (Supplementary Materials, Table S2—Crosstabulation of the distribution of EEHV-HD fatal events per Father and Zoo, for the captive European Asian elephant). At the end of the study, there were 18 calves reaching, or near the age of 2.7 years, the statistical age risk to succumb to EEHV-HD.4. DiscussionIn the present study, we compiled all data available since the first detection of a captive Asian elephant with EEHV-HD was detected, making it the most extensive study on the impact of EEHV on the European Asian elephant population to date. Our data showed that EEHV-HD affected calves at around 2.7 years old, which is significantly lower than the median age for other causes of death (8.6 years). These results are in accordance with recent reports from Europe, Thailand, and North American risk ages [6,10,20]. The European Endangered Species Programme’s latest report states that birth rates will not replace the loss of the high number of aged females (35–55 years old), of which the majority is considered unable to further reproduce. This will possibly lead to a decrease in female captive elephants in the future. The report also suggested that female elephants should become pregnant for the first time at 8 years of age and that ideally, there is an interbirth interval of 7 years [25]. Since EEHV-HD deaths occur at a significantly lower and narrower age range than other causes, killing mainly youngsters before sexual maturity is reached will, therefore, reduce the possibility of these calves substituting the elder ones, as well as reducing the overall number of possible future breeders. Consequently, this affects the breeding efforts made by the zoos on keeping a reproductive group to maintain a healthy and sustainable captive population.After removing all premature deaths, EEHV-HD alone was responsible for 52% of fatalities, nearly the same amount as reported for North American institutions (53% [20]). This mortality rate for EEHV deaths in Europe is slightly lower than the previous study published for the continent (57%—for calves surviving the first day of life, data until 2017; [10]) and is most likely a reflection on the increased number of survivors and the outstanding birth rate that year (20 new births in 2017). Despite the similarities between different countries, when we compare mortalities, we found that EEHV-HD presented a higher mortality rate in the European population (85%), compared to the one reported for North America (68%; [20]) or Thailand (nearly 69%; [6]). There are no indications that a more virulent serotype of the virus is present in Europe, therefore it is most likely that this mortality rate difference is related to the management of the disease. Due to their tradition of elephant training under the guidance of the mahouts, Thailand has facilitated veterinary access to these animals, to perform medical check-ups, and to treat very young calves. In the North American captive population, although in a protective contact system (where elephant keepers must not share the same unrestricted space with elephants), their management allows for direct training of young calves up to 24 months of age [26]. This allows for regular monitoring, as well as prompt and more effective prevention or veterinary treatment of calves once symptoms are present. In Europe, all EAZA members must also comply with the protected contact handling policy as it will become effective from 2030 on [27]. European zoos are also encouraged to start training their calves from the age of 4 months, and several behaviors facilitating medical support are expected to be achieved by the age of one year, for all breeding European institutions [14]. Hence, the lower mortality rates presented by Thailand and North America most probably reflect the substantial amount of survival cases due to effective treatment when compared to the population of this study. Nevertheless, the numbers of survivors in Europe have risen in the past few years, and it is expected to improve due to a higher awareness of the disease and the positive outcome that early monitoring and fast medical intervention can have.In this study, Europe presented lower EEHV-HD morbidity (15.8%) than North America, where one in every four calves (25%) has presented the disease [20]. This finding suggests that the European captive-born calves, although also exposed to the virus, become ill with EEHV-HD less often. However, it is most likely that this is related to under-detected or subclinical cases during the past years in Europe.Bennet [28] has performed a genogram on the Asian elephant captive populations living in Europe and North America to assess the possibility of a family link and found that EEHV-HD-related deaths appeared to be grouped into clusters. However, since the elephants in that study were originally located at the same institution, it remained unclear whether the clustering was due to genetic or environmental pressure [28]. Our study supports this indication of clustering of cases in certain zoos and accepts potential effect modification by either mothers or fathers. Combining a multivariable model, fathers and zoos revealed a higher risk for two specific zoos and one father to have their calves developing the disease. In the model using mothers and zoos, we also find a significantly higher risk for three specific zoos. Together, these findings indicate the possibility of a multifactorial disease, where a zoo-associated component must be assumed to be involved and a hereditary predisposition might be expressed under the influence of certain environmental pressures. This highlights the importance of collecting relevant risk factor information for all calves (retrospectively and prospectively) for more detailed analyses on risk factors. As an example, a hereditary coagulation disorder has been reported in an Asian elephant herd, where a breeding bull, although asymptomatic, presented a prolonged prothrombin time (one of the tests used to assess coagulation capability). This coagulopathy was caused by a specific mutation, leading to a lack of activity of one important clotting factor (coagulation factor VII) which led to an increase in bleeding time. Three of his five offspring were reported to be carriers of this mutation [29]. How the body of a calf, carrier of this hereditary coagulopathy, would react to the vascular endothelial damage caused by EEHV is unknown.One can debate whether the initial year of the study (1985) may be considered premature since, in the ’80s, diagnostic techniques for EEHV were not sufficiently developed or accurate. Part of the diagnostic gaps in the early years of the study period have been addressed by performing retrospective analyses with qPCR in frozen samples. These samples were tested to detect and quantify the virus, giving us a better idea of possible past cases that might have been overlooked [9,18,30].The narrow age range of EEHV-HD deaths found also implies that there might be an essential element that debilitates the Asian elephant calves at this specific age of their life. Therefore, a stressful element may play a part in triggering the virus, but also, there might be protective factors helping the calves that survived this risk age to overcome this period and thrive. Therefore, another worthwhile line of research would be to focus on finding what are the protective risk factors, especially at this sensitive young age.EEHV should not stop breeding programs at zoological institutions due to several reasons, including the continuous decrease of the global population of Asian elephants and its endangered status to face extinction. Lethal cases of this disease are found worldwide, and reports show that EEHV is ubiquitous and that elephants are the natural host and co-evolved with EEHV [5,9]. It is essential to gain a better knowledge of the disease’s pathophysiology and risk factors, to support the development of vaccination, and to improve treatment. All these research efforts to deepen this virus’ investigations can only be undertaken at a global scale and they are of extreme importance to halt EEHV-HD.At the end of this study, 80 Asian elephant calves were at the age of the previously reported fatal cases. Therefore, routine monitoring of these young calves and preparedness to tackle this disease is crucial to favor a positive outcome of the disease, while efforts to find more epidemiological risk elements of this hemorrhagic disease should be under investigation.Finally, this is an observational study, and therefore it is not possible to prove causality. Nevertheless, it guides us on the importance of follow-up studies to assess management conditions and to find the factors that protect or place the calf at a higher mortality risk. It is important to mention that the most meaningful and novel findings of this statistical study come from the updating and continuous analysis of a long-life living being, with a very long gestation time and inter-generational gap, enlightening the importance of longitudinal studies in elephants. Therefore, we suspect that more fathers, mothers, and institutions will be considered as related risk factors in the future and suggest that the starting period of this study should be used as a “milestone” for further studies.5. ConclusionsThis longitudinal epidemiological study investigates the elephant endotheliotropic herpesvirus impact in European zoological institutions, using the largest up-to-date dataset on captive Asian elephants.Our findings support previous studies, showing that EEHV is the primary cause of death among Asian elephants, besides neonatal mortality. Calves with EEHV-HD died at a very young age, around 2.7 years old (median age), which is a significantly younger age at death than that for other causes. Nevertheless, it is important to keep monitoring for EEHV until a later age of at least 8 years old (the oldest animal died with EEHV at 7.6 years of age).The results of this study suggest the involvement of zoo-associated factors, which might in part be related to management, and which can be influenced by either father or mother (or a combination of both), on the onset of EEHV-HD. Indeed, in total, two fathers, one mother, and four zoos presented a higher risk for their calves to develop the disease, when compared to all others in the study, hinting at the involvement of one or more environmental and triggering elements, with possible genetic associations.More focus needs to be placed on the underlying factors of this disease, in particular, the study of management differences between zoos with a higher risk of fatal outcomes due to EEHV-HD and low-risk zoos could inform Studbook breeding decisions.

animals : an open access journal from mdpi

10.3390/ani11030842

PMC8002471

Calving is a difficult moment in a cow’s life that causes stress, and the ease of calving determines the course of further lactation. The hypothesis of our study was to investigate how the difficulty of calving may influence changes in lactose concentration and other milk components and how well these two factors correlate between each other. We found a statistically significant (p < 0.001) negative correlation of calving ease score with milk lactose % (r = −0.376) and positive correlation coefficients with milk lactose yield (kg) (r = 0.277) as well as milk fat/lactose % ratio (r = 0.191). The analysis showed a regular increase (p < 0.001) with decreasing calving ease scores for milk electric conductivity and milk somatic cell count.

The aim of our study was to determine how the ease of calving of cows may influence changes in lactose concentration and other milk components and whether these two factors correlate with each other. To achieve this, we compared data of calving ease scores and average percentage of in-line registered milk lactose and other milk components. A total of 4723 dairy cows from nine dairy farms were studied. The cows were from the second to the fourth lactation. All cows were classified according to the calving ease: group 1 (score 1)—no problems; group 2 (score 2)—slight problems; group 3 (score 3)—needed assistance; group 4 (score 4)—considerable force or extreme difficulty. Based on the data from the milking robots, during complete lactation we recorded milk indicators: milk yield MY (kg/day), milk fat (MF), milk protein (MP), lactose (ML), milk fat/lactose ratio (MF/ML), milk protein/lactose ratio (MP/ML), milk urea (MU), and milk electrical conductivity (EC) of all quarters of the udder. According to the results, we found that cows that had no calving difficulties, also had higher milk lactose concentration. ML > 4.7% was found in 58.8% of cows without calving problems. Cows with more severe calving problems had higher risk of mastitis (SCC and EC). Our data indicates that more productive cows have more calving problems compared to less productive ones.

1. IntroductionCalving is a hazardous moment in a cow’s life that causes stress, not only that, the ease of calving determines the course of further lactation. Dystocia is a prolonged and difficult calving where a veterinary intervention is required. However, the time from the second stage of calf birth (amniotic rupture) to the moment when help should be provided, varies on average from 2 to 3 h [1]. The assessment of the severity of dystocia is not well defined, but most sources provide a dystocia scoring system from 0 to 4 or from 0 to 5 score, where 0 is easy calving and no intervention is needed, and 4–5 is very difficult calving when veterinary help is needed [2]. The ease/difficulty of calving depends on a variety of factors [2]. The main risk factors for difficult calving (dystocia) are associated with proximal or immediate causes such as feto-pelvic disproportion; uterine inertia; fetal malposition; vulval or cervical stenosis; uterine torsion [3,4,5,6,7]. Cows that suffered from dystocia, the risk for various diseases after calving, such as retained placenta, ketosis, metritis, displacement of abomasum, or mastitis increases [2,8]. Data from several studies indicate that dystocia may cause a number of adverse consequences, such as loss of production related to decreased milk yield [9], decreased milk protein, fat and lactose as well as increased somatic cell counts in milk [10]. Problematic calving and subsequent consequences have also been found to affect the fertility of cows negatively [10].A difficult calving may negatively influence the productivity of cows [11]. Moore et al. [12] found moderately positive correlations between calving and protein, milk and fat yield. In the first part of lactation, a veterinary-assisted dam showed a clear loss in milk yield compared to a nonassisted dam. An interesting finding has been established, that a difficult birth has long-term effects on the production of a calf in later life. The physiological causes or effects influencing a troublesome birth appear to be long-lived. This problem needs acknowledged, and more studies must be undertaken [11].The ability to record milk lactose concentration using automatic milking systems (AMS) allows its changes to be monitored in real time, several times a day. Thus, it is possible to monitor its changes in various physiological conditions and throughout the period of diseases of cows. Monitoring of lactose concentrations allows for analysis of its relationship with other parameters. An association between milk lactose percentage and cow fertility has been established [13]. Changes in lactose levels in milk have been shown to be a useful indicator for predicting first and second post-calving ovulation [14]. Higher lactose levels are associated with an increased probability of pregnancy [15] as well as a close association of lactose levels with postpartum recovery of luteal function has been reported [16]. Milk electrical conductivity and lactose concentration have been found to be one of the most useful parameters for monitoring and identification of subclinical and clinical mastitis [17]. Since milk lactose concentration decreases during inflammation, it could be used as an indicator for mastitis [17]. The adoption of precision farming technologies on a large scale allows for the daily registration of individual milk and changes of certain milk components that could be used to detect health disorders and start early treatment.Monitoring of lactose concentrations has become widely used as a parameter for early diagnosis and herd management. An association between milk lactose percentage and cow fertility has been established [13]. Changes in lactose levels in milk have been shown to be a useful indicator for predicting first and second post-calving ovulation [14]. Higher lactose levels are associated with an increased probability of pregnancy [15] as well as a close association of lactose levels with postpartum recovery of luteal function has been reported [16]. Milk electrical conductivity and lactose concentration have been found to be one of the most useful parameters for monitoring and identification of subclinical and clinical mastitis [17]. Since milk lactose concentration decreases during inflammation, it could be used as an indicator for mastitis [17]. The ability to record milk lactose concentration using automatic milking systems (AMS) allows its changes to be monitored in real time, several times a day. Thus, it is possible to monitor its changes in various physiological conditions and throughout the period of diseases of cows. The adoption of precision farming technologies on a large scale allows for the daily registration of individual milk and changes of certain milk components that could be used to detect health disorders and start early treatment.While researching literature, we found a lack of information on how calving ease affects in-line milk lactose levels and other milk components, them being important tools in cow health and productivity assessment. Our hypothesis was that ease of calving has an influence on the lactose concentrations of milk and that they are closely related. Therefore, we set an aim to evaluate the relationship of calving ease (determined by scoring) and the average percentage of in-line registered milk lactose concentration and other milk components.2. Materials and Methods2.1. Location and AnimalsThe study was carried out in nine Lithuania dairy farms (from 1 February 2018–7 April 2020) with more than 500 milking cows. A total of 4723 dairy cows from nine dairy farms were studied. The selection of the cows was according to the following criteria: cows had to be first 30 days after calving, having two or more lactations. The cows were kept in a free-stall barn and were fed a total mixed ration (TMR) routinely throughout the year, which was balanced according to their physiological and production needs. Feeding time for the cows took place every day at 06:00 and 18:00 with a total mixed ration for high-producing, multiparous cows. Diets were formulated accordingly to meet or exceed the requirements of a 550 kg Holstein cow producing 35 kg of milk per day according to NRC [18]. The average chemical composition of rations of the nine farms is as follows: dry matter (DM) (%) 48.8; neutral detergent fiber (% of DM) 28.2; acid detergent fiber (% of DM) 19.8; non-fiber carbohydrates (% of DM) 38.7; crude protein (% of DM) 15.8; net energy for lactation (Mcal/kg) 1.6. On average 2 kg per day of concentrate were provided for the cows at the milking robot during milking.2.2. MeasurementsMilk traits of all 4.723 cows were studied during the early lactation (from calving to 30 days after calving). Cows were milked with Delaval milking robots with free traffic and combined with the fully automated real-time milk analyzer Herd Navigator (DeLaval Inc. Tumba. Sweden). The milking robot automatically takes a representative several milliliters sample of milk from a cow during the milking process. Calving score records were collected during 2018–2020 by trained farms technicians at the same methodology according to Jensen [19]. For the evaluation of the calving score, a 4-point scale was used accordingly: 1—easy unassisted (n = 2.264, 47.94%); 2—easy, assisted (n = 1.479, 31.31%); 3—difficult, assisted (n = 490, 10.37%); 4—difficult, requiring veterinary assistance (n = 490, 10.37%). A total of 4723 calvings were assessed. All cows were identified when they start calving according Saint-Dizier and Chastant-Maillard (2015) three stage methodology [20].The cows were from the second to the fourth lactation. Based on the data from the milking robots, during first 30 days after calving every day milk indicators of the cows were recorded: milk yield MY (kg/day), milk fat (MF), milk protein (MP), lactose (ML), milk fat %/lactose % ratio (MF/ML), milk protein %/lactose % ratio (MP/ML), milk urea (MU), and milk electrical conductivity (EC) of all quarters of the udder (according to Televičius et al. [21], milk lactose level during first 30 days after calving (21.2 ± 11.04 days). All cows (n = 4.723) were divided into two groups: group 1—lactose < 4.70% (n = 2.732, 57.84%); group 2—lactose ≥ 4.70% (n = 1.991, 42.16%). Based on the average milk EC level, the cows were divided into four classes: EC < 4.5 mS/cm (n = 1.294, 27.11%), EC = > 4.5–5.5 mS/cm (n = 2.940, 61.60%), EC > 5.5–6.5 mS/cm (n = 477, 9.99%), EC > 6.5 mS/cm (n = 62, 1.30%). 2.3. Data Analysis and StatisticsThe statistical analysis of data was performed using the SPSS 25.0 programme package (SPSS Inc., Chicago, IL, USA). The normality of data was assessed for all variables using the Kolmogorov–Smirnov test. To obtain a normal distribution, milk somatic cell (SCC) data were transformed into somatic cell score (SCS = log2(SCC/100) + 3). Thus, parametric methods of statistical analysis were applied to all milk indicators. Variables of the descriptive statistics are presented as mean ± standard error of mean (M ± SEM) and 95% confidence interval (CI). When comparing the analyzed milk indicators in terms of milk lactose level, the T-test of independent samples was used. Multiple comparisons of observed mean values with Bonferroni criterion were used to analyze milk characteristics according to the calving ease (CE) scale. The relationships between the studied indicators were examined according to the linear Pearson correlation coefficient (correlation of milk lactose and other milk components), the Spearman correlation (between CE and milk traits) and the χ2 test (relation between milk lactose level, other milk components and CE group). For all tests, a probability less than 0.05 was considered significant (p-Value < 0.05).3. Results and Discussion During the study there was no apparent increase in the number of somatic cells (SCC) in the milk, which averaged 100.02 ± 5.126 thousand/mL milk. The average milk urea concentration (24.16 ± 0.125 mg/dL) and milk fat to protein ratio (1.20 ± 0.003) of the cows corresponded to balanced norms. The average milk yield of all cows during lactation was 28.27 ± 0.118 kg, milk fat 4.08 ± 0.012%, milk protein 3.48 ± 0.006%, milk lactose 4.65 ± 0.004%, average milk EC of all quarters 4.94 ± 0.008 mS/cm, MF/ML 0.88 ± 0.003, MP/ML 0.75 ± 0.001.3.1. Relationship of Calving Ease with Milk Lactose and OtherMilk Components.In this experiment, a steady linear decrease in milk lactose % with increasing CE scores was found (Table 1). The average milk lactose percentage of cows in the second CE group was lower than in the first group (0.16 %, p < 0.001) and higher than in the third (0.04%, p < 0.001) and fourth (0.05%, p < 0.001) groups. On the other hand, the average lactose content in kg per day increased with increasing CE scores (from 1.21 ± 0.008 to 1.47 ± 0.016 kg, p < 0.01). The analysis showed that the MF to ML ratio increased (from 0.85 ± 0.004 to 0.95 ± 0.009, p < 0.001) as the cows’ CE scores deteriorated and calving problems increased. The highest mean of the MP/ML ratio was evaluated in the second CE group (0.77 ± 0.002) and the lowest in the fourth group of cows (0.37 ± 0.004, p < 0.001).According to Costa et al., lactose is suggested as a potential health indicator in cows [3]. Lactose is synthesized in the udder from blood glucose absorbed by the basal membrane of mammary epithelial cells [22]. Glycemia and energy balance in cows have a positive correlation with milk lactose [23], especially in high-producing breeds [24]. Lemosquet et al. [24] suggested that post-hepatic blood glucose availability could be a vital indirect regulator of milk yield. Lactose shows potential to be a valid health indicator in cows [25]. Based on the data presented in Table 1, the observed trends are confirmed by the calculated correlation coefficients in Figure 1.We found a statistically significant (p < 0.001) negative correlation of CE with cows ML% (r = 0.376) and positive correlation coefficients with ML yield (kg; r = 0.277) and MF/ML % ratio (r = 0.191). Furthermore, we registered the correlations between milk lactose and risk of clinical mastitis (based of EC) in early lactation or across the whole lactation, respectively. Costa et al. [26] reported negative correlation between milk lactose and mastitis in the first 150 days in milk. Lactose, EC and their combination together were the most accurate parameters for detection of mastitis in dairy farms equipped with in-line sensors [19]. During mammary tissue inflammation, the osmotic balance is maintained by an increase of Na+ and Cl−; in particular, Na+ derived from the highly Na+-concentrated extracellular environment is the main ion responsible for the increase of the electrical conductivity [3]. Phenotypic correlations between milk lactose and SCC range from −0.15 to −0.66 [27]; in this sense, ML has been widely reported to be one of the most informative parameters used in mastitis diagnosis, other than SCC and milk electrical conductivity [27]. The combined information from ML, SCC and electrical conductivity can be used to provide a precise diagnosis of mastitis at the individual level, the potential of alternative or derived traits (or both) as predictors of udder inflammation [26]. EC and ML were the most reliable indicators of subclinical mastitis (in combination with SCC) and were used to establish predictive patterns of subclinical mastitis in Holstein cows [26]. The decrease of lactose in milk during mastitis can be caused by partly compromised milk lactose synthesis, since the secretory cells are damaged by inflammation, part of the lactose is lost in urine, and a disruption of tight junctions and altered permeability of the basal membrane of the mammary cells that separates blood and milk, mastitis pathogens use available milk lactose to reduce milk lactose and increase lactic acid in milk [3]Therefore, considering the existing relations between ML and the traits above, lactose and perhaps its ratios with fat, protein or both may be used well as potential biomarkers for metabolic disease in early lactation, as reported by Ederer et al. [28]. The experiment showed that ML > 4.7% was found in 58.8% of cows with a calving score of 1, in 27.9% of cows with a calving score of 2, in 26.9% of cows with a calving score of 3, and 23.5% with a score of four points (χ2 = 498,970, df = 3, p < 0.001).We detected that milk from cows with higher lactose levels had higher milk protein contents (0.04%) and milk fat content (0.22%) of such cows was lower (p < 0.01; Table 2). The increase in lactose % was closely related to the decrease in the electrical conductivity of milk according to the linear regression equation obtained in Figure 2. Milk electrical conductivity, lactose concentration and their combination together proved to be the most accurate parameters for detection of mastitis in dairy farms equipped with in-line sensors [17]. This is also confirmed by the correlation analysis data summarized in Figure 3. Both milk lactose percentage and lactose content were negatively correlated with milk somatic cells, as was milk EC and fat content. We also found that milk lactose yield (kg) was positively correlated with cow productivity (p < 0.001). The dependence of milk yield on lactose concentration and the uptake of glucose from the blood to produce lactose is a metabolic priority [29]. 3.2. Relation of Calving Score with Milk Traits of CowsOur data indicates that higher yielding cows had more problems while calving than cows with lower milk yield. As calving scores increased from one to four, the average productivity of cows increased by 6.77 kg (p < 0.001). Not only that, with increasing CE scores, a regular increase (p < 0.001) in milk EC and milk somatic cell count was seen (Table 3).CE score positively correlated (p < 0.001) with MY (r = 0.342), EC (r = 0.365), MF and SCS (r = 0.191) and negatively with MP (p < 0.001) and MU (p = 0.006; Figure 4). A positive association of ML with fertility in the subsequent lactation has been reported by Bastin et al. [30], highlighting increased success in fertility in cows yielding milk with higher ML. Both milk ML and fertility greatly depend on cow energy balance [25], meaning that the relationship between them is likely indirect. On the other hand, Costa et al. [26] reported that genetic correlations were close to zero between ML and a few fertility disorders, namely ovarian cysts and retained placenta. In the last decade, authors have combined ML and milk urea together as an indicator of metabolic health [31]. Ganter et al. [32] stated that due to a negative energy balance at the beginning of lactation, the milk fat content increases, followed by a decrease in milk protein contents, which makes the F/P ratio a good indicator in metabolic disorder identification. Their results state that the amount of lactose in milk, which correlates with the amount of glucose in the blood, can be used to assess the energy balance of cows According to Harjanti and Sambodho [33] the capability of ruminant mammary gland to produce milk is determined by the activity of milk secreting cells and their numbers. Therefore, the milk yield and the concentrations of lactose, protein and fat in milk might be affected by the inflammation level of the mammary gland [34]. There were studies where milk lactose concentrations decreased, and somatic cell counts increased during subclinical and clinical mastitis [22]. Thus, the monitoring of lactose concentrations in milk could be used for identifying cows with mastitis, as with inflammation a clear decrease is seen [22]. Schneeberger and Lynch [35] reported an increase in blood lactose concentration, which is an indicator of mammary epithelium health. According our previous study, we found that cows with a higher lactose concentration (≥4.70%) were registered as more active and were at less risk of mastitis (as indicated by lower milk EC and SCC) and metabolic disorders. Low levels of lactose can indicate mastitis (milk SCC ≥ 100 thousand/mL) and metabolic disorders (subclinical ketosis, subclinical acidosis), described by milk F/P. Coulon et al. [34] performed a study to estimate the influence of walking activity on milk production and energy status in dairy farms that housed cows in tie-stall. Dairy cows present signs of sickness in their behavior during mastitis—changes in activity, lying time and feeding behavior have received the most scientific attention [36]. Changes in behavior are suspected to be induced by pain or different unpleasant experiences [37]. SCC increases and milk lactose concentrations decrease during clinical and subclinical mastitis. Milk electrical conductivity, milk lactose and SCC have been widely reported to be some of the most informative traits for mastitis diagnosis [38]. According to a study carried out by Costa et al. [26], there was a genetic correlation between mastitis and milk lactose yield. 4. ConclusionsWe found that cows without calving difficulties had higher milk lactose concentrations. ML ≥ 4.7% was evaluated in 58.8% of cows without calving problems and we can suspect that they have a more positive energy balance. Cows with higher calving problems were at higher risk of mastitis (indicated by SCC and EC). Higher yielding cows have more calving problems compared to less productive ones.

animals : an open access journal from mdpi

10.3390/ani13050823

PMC10000092

The transition from milk to solid feed in commercial pig-production systems negatively affects gut health, particularly the composition of the residing microbial community. This can subsequently impair pig growth and long-term health. Natural dietary supplements including seaweed extracts have the capacity to reduce pathogen load (antibacterial activity) and/or increase beneficial microbes (prebiotic activity). This study evaluated the antibacterial and prebiotic potential of two seaweed species, Laminaria hyperborea and Laminaria digitata, and their extracts using laboratory-based simulations of the gut microbial community of weaned pigs. Our investigation identified seaweed extracts that could decrease the numbers of pig- and food-related pathogens or increase the number of beneficial microbes, albeit to a different extent. These findings indicate that seaweeds are a promising source of antibacterial and prebiotic dietary supplements for use in pigs during the weaning period.

Laminaria spp. and their extracts have preventative potential as dietary supplements during weaning in pigs. The first objective of this study was to evaluate increasing concentrations of four whole seaweed biomass samples from two different Laminaria species harvested in two different months in a weaned pig faecal batch fermentation assay. Particularly, February and November whole seaweed biomass samples of L. hyperborea (LHWB-F and LHWB-N) and L. digitata (LDWB-F and LDWB-N) were used. In the next part of the study, the increasing concentrations of four extracts produced from L. hyperborea (LHE1–4) and L. digitata (LDE1–4) were evaluated in individual pure-culture growth assays using a panel of beneficial and pathogenic bacterial strains (second objective). The LHE1–4 and LDE1–4 were obtained using different combinations of temperature, incubation time and volume of solvent within a hydrothermal-assisted extraction methodology (E1–4). In the batch fermentation assay, the L. hyperborea biomass samples, LHWB-F and LHWB-N, lowered Bifidobacterium spp. counts compared to the L. digitata biomass samples, LDWB-F and LDWB-N (p < 0.05). LHWB-F and LDWB-N reduced Enterobacteriaceae counts (p < 0.05). LHWB-F and LDWB-F were selected as the most and least promising sources of antibacterial extracts from which to produce LHE1–4 and LDE1–4. In the pure-culture growth assays, E1- and E4-produced extracts were predominantly associated with antibacterial and bifidogenic activities, respectively. LHE1 reduced both Salmonella Typhimurium and Enterotoxigenic Escherichia coli with LDE1 having a similar effect on both of these pathogenic strains, albeit to a lesser extent (p < 0.05). Both LHE1 and LDE1 reduced B. thermophilum counts (p < 0.05). LDE4 exhibited strong bifidogenic activity (p < 0.05), whereas LHE4 increased Bifidobacterium thermophilum and Lactiplantibacillus plantarum counts (p < 0.05). In conclusion, antibacterial and bifidogenic extracts of Laminaria spp. were identified in vitro with the potential to alleviate gastrointestinal dysbiosis in newly weaned pigs.

1. IntroductionA healthy gut microbiota which is compositionally and functionally diverse and stable is essential to support host health and growth [1,2,3,4]. Contrarily, dysbiosis represents a state of imbalance in the composition and function of this microbial community, characterised by decreases in beneficial microorganisms and/or overgrowth of pathogens and/or a loss of overall diversity, with a subsequent negative impact on gastrointestinal health [5]. Commercial weaning in pigs is a typical example of dysbiosis, whereby the transition from milk to solid feed, coupled with emotional, social and environmental stressors leads to gastrointestinal dysfunction characterised by dysbiosis that predisposes them to intestinal infection and disease [6,7]. In a recent review, the potential of marine macroalgae or seaweeds were considered as natural dietary supplements with which to promote gastrointestinal health and subsequently growth in weaned pigs [8]. Brown seaweeds are rich in nondigestible polysaccharides, minerals, polyphenols and vitamins [9,10]. Wide-ranging biological activities [11] have been attributed to seaweed components, particularly fucoidan and laminarin, including prebiotic [12,13] and antibacterial [14,15] potential. However, various factors influence the concentration, structure and biological activity of seaweed-derived polysaccharides, such as seaweed species, harvest season, environmental conditions, and extraction methodologies [14,16]. Recently, a multivariate statistic technique, response surface methodology, has been utilised to improve the extraction efficiency by optimising the extraction conditions for a selected seaweed polysaccharide and/or bioactivity [17]. In that study, a novel hydrothermal-assisted extraction (HAE) methodology with combinations of temperature, time and solvent to seaweed ratio optimised for the best concentration of laminarin and/or fucoidan and/or antioxidant activity was developed using the response surface methodology. Seaweed extracts of Ascophyllum nodosum produced using this HAE methodology exhibited enhanced antibacterial and prebiotic activity compared to the conventional extraction methods [18].The brown seaweed Laminaria spp. is a rich source of biologically active nondigestible polysaccharides. Previous in vitro investigation has associated this seaweed species with various biological activities including anti-inflammatory, immunomodulatory, antioxidant, antitumor and antihypertensive [9,11] effects, several of which have also been observed in in vivo studies with pigs [19,20,21,22]. Concerning its effect on the gastrointestinal microbiota, dietary supplementation of pigs with crude Laminaria spp. extracts consistently led to a reduction in the numbers of the Enterobacteriaceae family [22,23,24,25,26,27] which include several animal and human pathogens such as Salmonella enterica subsp. enterica serotype Typhimurium and pathogenic Escherichia coli [28,29]. Furthermore, an increase in Enterobacteriaceae family is considered to be an indication of dysbiosis and a risk factor for post-weaning diarrhoea in pigs [6,30]. Dietary supplementation of pigs with crude Laminaria spp. extracts led to a more variable response in the intestinal lactobacilli and Bifidobacterium spp. populations, as both increases [22] and decreases [23,24] in their counts have been reported. Lactobacilli are dominant members of the gastrointestinal microbiota in pigs and have an important role in growth and health due to their contributions to nutrient bioavailability, inhibition of pathogen colonisation and immunomodulation [31,32]. Bifidobacterium spp. are considered a beneficial bacterial population due to their probiotic status [33], but are present in low abundance in the gastrointestinal tract of pigs [34]. The current study focused on the antibacterial and prebiotic potential of two polysaccharide-rich members of the Laminaria spp., L. digitata and L. hyperborea, with an average total carbohydrate content of 70.7% and 65.5% of dry weight, respectively, as described in previous reports [35]. Batch fermentation and pure-culture growth assays were useful screening tools when assessing the direct effects of whole biomass seaweed samples and their extracts on key bacterial populations and species in the porcine gastrointestinal tract [18]. Thus, the first objective of this study was to assess the influence of seaweed species and harvest season on the effect of whole biomass samples of L. digitata and L. hyperborea with respect to selected faecal bacterial populations in a batch fermentation assay inoculated with pig faeces. The second objective was to investigate whether the different extraction conditions of the HAE methodology led to L. digitata and L. hyperborea extracts with improved antibacterial and prebiotic activities using a panel of pure-culture growth assays.2. Materials and Methods2.1. Laminaria spp.: Whole Biomass Samples and ExtractsThe whole biomass samples (WB) of L. digitata (LD) and L. hyperborea (LH) were harvested in February (LDWB-F and LHWB-F) and November (LDWB-N and LHWB-N) by Quality Sea Veg Ltd., Co. (Donegal, Ireland). For each seaweed species, whole biomass samples were collected at a single time point and from the same collection site. The preparation (oven-dying, milling) and compositional analysis (dry matter, ash, protein, crude lipids, polysaccharide content, total phenols) of the dried whole seaweed biomass samples was performed as previously described [36]. LDWB-F, LDWB-N, LHWB-F and LHWB-N were stored at room temperature until their evaluation in the batch fermentation assay. A HAE methodology with optimised extraction conditions (temperature, incubation time and volume of solvent) was used to produce the extracts of LDWB-F and LHWB-F, as described previously by Garcia-Vaquero et al. [17] and presented in Table 1. The parameters for each extraction condition were optimised towards the concentration of fucoidan for E1, laminarin for E2, antioxidant activity for E3 and all the above for E4. The produced extracts of L. digitata (LDE1–4) and L. hyperborea (LHE1–4) were freeze-dried and their laminarin and fucoidan content was determined as previously described [18]. All extracts were analysed on two independent occasions (two biological replicates) with three readings each time. LDE1–4 and LHE1–4 were stored at −20 °C until their evaluation in the pure-culture growth assays.2.2. Batch Fermentation AssayThe preparation of the faecal inoculum and the batch fermentation assay were carried out as described previously [37]. Briefly, faeces from 29 healthy newly weaned crossbred pigs (Large White × Landrace) fed a cereal- and milk-based diet were pooled, aliquoted and stored at −20 °C. One day prior to the batch fermentation assay, the pooled faeces were diluted (1:5 w/v) in phosphate-buffered saline (Sigma-Aldrich, St. Louis, MO, USA) after oxygen removal using oxyrase (Sigma-Aldrich, St. Louis, MO, USA) to prepare the faecal inoculum (FI) that was stored at 4 °C anaerobically. The FI was added to the fermentation medium at a 1:10 v/v ratio (21 mL final volume). The inclusion levels of LDWB-F, LHWB-F, LDWB-N and LHWB-N in the FI/fermentation medium were 0 (control tubes), 1, 2.5 and 5 mg/mL. The batch fermentation was carried out under anaerobic conditions using oxyrase and CO2 flushing at 39 °C for 24 h with gentle stirring (100 rpm). Sampling (5 mL fermentation broth) was performed at 0 and 24 h in duplicate. After centrifuging at 12,000× g for 5 min, the resultant pellets were stored in −20 °C until further analysis. All experiments were repeated on three independent occasions (biological replicates n = 3). 2.3. Quantification of Bacterial Groups Using Quantitative Real Time Polymerase Chain Reaction (QPCR)DNA extraction: Bacterial DNA was extracted using QIAamp Fast DNA stool mini kit (Qiagen, West Sussex, UK) according to the manufacturer’s instructions, and its quantity and quality was evaluated spectrophotometrically (Nanodrop, Thermo Fisher Scientific, Waltham, MA, USA).Bacterial primers: The primers targeting the 16S rRNA gene of selected bacterial groups (total bacteria, lactobacilli and Bifidobacterium spp.) or the rplP gene (Enterobacteriaceae) are provided in Table 2. Primer design software, Primer3 (https://primer3.org/ (accessed on 26 June 2018)) and Primer Express™ (Applied Biosystems, Foster City, CA, USA) were used for larger amplicons (>150 bp) and smaller amplicons (<125 bp), respectively. Primer specificity was verified using Primer Basic Local Alignment Search Tool (Primer-BLAST), https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi, accessed on 26 June 2018.Bacterial enumeration by QPCR: The quantification of the above-mentioned bacterial groups was carried out using QPCR with plasmid-based standard curves as described previously [37]. Briefly, Competent E. coli was transformed with a pCR4-TOPO™ TA vector containing each fragment of the targeted 16S rRNA genes for total bacteria, lactobacilli and Bifidobacterium spp. or the rplP gene for Enterobacteriaceae and the resistance to ampicillin gene using a TOPO™ TA Cloning™ Kit for Sequencing (Invitrogen, Thermo Fisher Scientific, Carlsbad, CA, USA) and stored in cryoprotective beads (TS/71-MX, Protect Multi-purpose, Technical Service Consultants Ltd., Lancashire, UK). Transformed E. coli was recultured in 200 mL LB Broth Base (Invitrogen, Thermo Fisher Scientific, Carlsbad, CA, USA) containing ampicillin (100 µg/mL) at 37 °C for 18 h at 150 rpm. Plasmids were extracted on a large scale using the GenElute™ HP Plasmid Maxiprep kit, (Sigma-Aldrich, St. Louis, MO, USA), linearised using APA1 restriction enzyme (Promega, Madison, WI, USA) and purified using GenElute™ PCR Clean-Up kit (Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturers’ instructions. Following quantification, the plasmid copy number/μL was determined using the URI Genomics & Sequencing Centre online tool (http://cels.uri.edu/gsc/cndna.htmL, accessed on 14 May 2019). For the QPCR, the final reaction volume (20 μL) included 3 μL template DNA, 1 μL of forward primer (10 μM), 1 μL of reverse primer (10 μM), 5 μL nuclease-free water and 10 μL of Fast SYBR® Green Master Mix (Applied Biosystems, Foster City, CA, USA) for the lactobacilli or GoTaq® qPCR Master Mix (Promega, Madison, WI, USA) for the remaining bacterial groups. All QPCR reactions were performed in duplicate on the ABI 7500 Fast PCR System (Applied Biosystems, Foster City, CA, USA) using the following cycling conditions: a denaturation step (95 °C/10 min), 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Dissociation curve analysis and visualisation on a 2% agarose gel stained with ethidium bromide were used to confirm the production of single and specific PCR products. All PCR reactions used in this study exhibited 90–110% efficiency established by plotting the threshold cycles (Ct) derived from the 5-fold serial dilutions of each plasmid against their arbitrary quantities. Bacterial counts were determined using the standard curve derived from the mean Ct value and the log-transformed gene copy number of the respective plasmid and expressed as log-transformed gene copy number per gram of digesta (logGCN/g digesta). animals-13-00823-t002_Table 2Table 2List of forward and reverse primers used for the bacterial quantification by QPCR.Target Bacterial GroupForward Primer (5′–3′)Reverse Primer (5′–3′)Amplicon Length (bp) Tm (°C)ReferencesTotal bacteriaF: GTGCCAGCMGCCGCGGTAAR: GACTACCAGGGTATCTAAT29164.252.4[38]LactobacilliF: AGCAGTAGGGAATCTTCCAR: CACCGCTACACATGGAG34154.555.2[39]Bifidobacterium spp.F: GCGTGCTTAACACATGCAAGTCR: CACCCGTTTCCAGGAGCTATT12560.359.8[40]EnterobacteriaceaeF: ATGTTACAACCAAAGCGTACAR: TTACCYTGACGCTTAACTGC18554.056.3[41]bp, base pairs; Tm, melting temperature.2.4. Bacterial Strains and Pure-Culture Growth AssaysPure-culture growth assays using a panel of commensal strains Lactiplantibacillus plantarum subsp. plantarum (formerly Lactobacillus plantarum, DSMZ 20174), Limosilactobacillus reuteri (formerly Lactobacillus reuteri, DSMZ 20016) and Bifidobacterium thermophilum (DSMZ 20210) and pathogens S. typhimurium PT12 and enterotoxigenic E. coli (ETEC) O149A+ selected for their beneficial roles and negative impacts on pig and human health, respectively, were carried out as described in our previous work [37]. Briefly, 24 h cultures of all bacterial strains were prepared using standard procedures and diluted in 10% medium: 10% de Man, Rogosa and Sharpe broth (MRS, Oxoid Ltd., Hampshire, UK) for L. plantarum, L. reuteri and B. thermophilum cultures; or 10% Tryptone Soya Broth (TSB, Oxoid Ltd., Hampshire, UK) for S. typhimurium and ETEC cultures, to obtain an inoculum of 106–107 CFU (colony-forming unit)/mL (verified for each assay). Two-fold dilutions (2–0.125 mg/mL) of LDE1–4 and LHE1–4 were performed in 10% MRS and 10% TSB prior to each assay from a working concentration of 4 mg/mL. 100 μL of each extract and each dilution and 100 μL of inoculum were added to duplicate wells of 96-well microtiter plates (CELLSTAR, Greiner Bio-One, Kremsmünster, Austria). Control wells were also included (bacterial inoculum only). Assay sterility was assessed using blank wells (no bacterial inoculum) for each dilution of each extract. After gentle agitation to ensure thorough mixing, plates were incubated aerobically at 37 °C for 18 h, apart from B. thermophilum, which was incubated anaerobically. Afterwards, bacterial enumeration was carried out by 10-fold serial dilution (10−1–10−8), spread plating onto MRS agar (Oxoid Ltd., Hampshire, UK) for L. plantarum, L. reuteri and B. thermophilum, and Tryptone Soya Agar (Oxoid Ltd., Hampshire, UK) for ETEC and S. typhimurium, and incubation aerobically at 37 °C for 24 h or anaerobically at 37 °C for 48 h for B. thermophilum. The dilution resulting in 5–50 colonies was selected for the calculation of CFU/mL using the formula, CFU/mL = average colony number × 50 × dilution factor. The bacterial counts were logarithmically transformed (logCFU/mL) for the subsequent statistical analysis. All experiments were carried out with technical replicates on three independent occasions (3 biological replicates n).2.5. Statistical AnalysisAll data were statistically analysed using Statistical Analysis Software (SAS) 9.4 (SAS Institute, Cary, NC, USA). Normality tests were initially carried out using PROC UNIVARIATE procedure for each data set. Batch fermentation assay: For each bacterial group and tested compound, the bacterial counts (n = 12, 3 flasks/each compound concentration) were analysed using PROC GLM procedure (Tukey’s test). The statistical model included the fixed effects of seaweed species (L. digitata, L. hyperborea), the season of collection (February, November), the concentration of whole biomass (0, 1, 2.5 and 5 mg/mL) and assay replicates (3 biological replicates) and their associated two- and three-way interactions with the bacterial counts at 0 h as a covariate. Pure-culture growth assay: To control for the natural variability in bacterial growth in the pure-culture assays, bacterial counts were expressed as the difference between the counts of each bacterial strain for each extract concentration and their respective control (0 mg/mL). The resulting positive or negative values representing the difference in bacterial counts were analysed using PROC GLM procedure (Tukey’s test). The statistical model assessed the effects of seaweed species (L. digitata, L. hyperborea), extraction conditions (E1–4) and concentration of extracts (0.125, 0.25, 0.5, 1 and 2 mg/mL) and their associated two- and three-way interactions. The biological replicate was the experimental unit.Probability values of < 0.05 denote statistical significance. Results are presented as least-square mean values ± standard error of the means (SEM).3. Results3.1. Proximate Composition of L. digitata and L. hyperborea, and Laminarin and Fucoidan Content of Their ExtractsThe proximate composition of the whole seaweed biomass samples LDWB-F, LHWB-F, LDWB-N and LHWB-N is presented in Table 3, as reported previously [36]. The laminarin and fucoidan contents in the L. digitata (LDE1–4) and L. hyperborea (LHE1–4) extracts are presented in Table 4.3.2. Effects of the Whole Biomass Samples of L. digitata and L. hyperborea on Selected Bacterial PopulationsThe effects of the whole biomass samples of L. digitata and L. hyperborea collected in February (LDWB-F and LHWB-F) and November (LDWB-N and LHWB-N) were evaluated on selected faecal bacterial populations in a batch fermentation assay. The effects of species, season and concentration and their interactions are presented in Table 5 and Table 6 and are described below. The species × concentration interaction and the season × concentration interaction were only significant for Bifidobacterium spp. (p < 0.05) and, as a result, were excluded from the statistical analysis of the other bacterial groups.Enterobacteriaceae: There was a species × season interaction whereby LHWB-F led to lower Enterobacteriaceae counts compared to LHWB-N, while the opposite was true for L. digitata (p < 0.05, Table 5). There was also a concentration effect on Enterobacteriaceae counts, whereby the 5 mg/mL reduced Enterobacteriaceae counts compared to the control (7.74 logGCN/g digesta (5 mg/mL) vs. 7.99 logGCN/g digesta (0 mg/mL) ± 0.056, p < 0.05).Bifidobacterium spp.: There was a species × season interaction, whereby LHWB-N led to lower Bifidobacterium spp. counts compared to LHWB-F, while harvest season had no effect on Bifidobacterium spp. counts with regard to L. digitata (p > 0.05, Table 5). There was a species × concentration interaction, whereby the concentrations of 2.5 and 5 mg/mL of L. hyperborea led to lower Bifidobacterium spp. counts compared to the control and 1 mg/mL (2.73 logGCN/g digesta (2.5 mg/mL) and below the limits of detection (5 mg/mL) vs. 6.72 (0 mg/mL) and 6.60 (1 mg/mL) logGCN/g digesta ± 0.056, p < 0.05)), while this effect was not as potent with the corresponding L. digitata concentrations (5.95 logGCN/g digesta (5 mg/mL) vs. 6.46 (0 mg/mL), 6.55 (1 mg/mL) and 6.37 (2.5 mg/mL) logGCN/g digesta ± 0.056, p < 0.05)). There was a season × concentration interaction, whereby the 5 mg/mL of both February and November suppressed Bifidobacterium spp. counts, while at 2.5 mg/mL, there was a greater reduction in the counts in November relative to February (p < 0.05, Table 6). Total bacteria: There was a species effect and a concentration effect on total bacterial counts. L. hyperborea increased total bacteria compared to L. digitata (9.83 logGCN/g digesta (L. hyperborea) vs. 9.72 logGCN/g digesta (L. digitata) ± 0.034, p < 0.05). The 1 and 2.5 mg/mL gave higher counts compared to the control (9.87 (1 mg/mL) and 9.87 (2.5 mg/mL) logGCN/g digesta vs. 9.64 logGCN/g digesta (0 mg/ mL) ± 0.048, p < 0.05).Lactobacilli: There was a species effect and a concentration effect on lactobacilli counts. L. hyperborea increased lactobacilli counts compared to L. digitata (8.79 logGCN/g digesta (L. hyperborea) vs. 8.45 logGCN/g digesta (L. digitata) ± 0.030, p < 0.05). The concentrations of 1 and 2.5 mg/mL were associated with higher counts compared to the control (8.66 logGCN/g digesta (1 mg/mL) and 8.69 logGCN/g digesta (2.5 mg/mL) vs. 8.56 logGCN/g digesta (0 mg/mL) ± 0.035, p < 0.05).In summary, whole seaweed biomass samples from L. hyperborea and L. digitata collected in February had the least negative impact on Bifidobacterium spp. counts. Furthermore, LHWB-F reduced Enterobacteriaceae counts to the greatest degree, while LDWB-F had no effect. Both LHWB-F and LDWB-F were selected to generate the extracts evaluated in the next part of the screening process to determine whether the extraction methodology could improve the bioactivity of two whole seaweed biomass samples with varying effects. 3.3. Identifying L. digitata and L. hyperborea Extracts with the Highest Antibacterial and Prebiotic Potential in Pure Bacterial CulturesThe HAE methodology with four different extraction conditions (E1–4) was employed for producing the extracts from the L. digitata (LD) and L. hyperborea (LH) samples, collected in February, to investigate whether the extraction method could improve their biological properties. LDE1–4 and LHE1–4 were evaluated for their antibacterial and prebiotic activities in pure-culture growth assays with selected beneficial (L. plantarum, L. reuteri, B. thermophilum) and pathogenic (ETEC, S. typhimurium) bacterial strains. Bacterial counts were expressed as the difference between the counts of each bacterial strain for each extract concentration and their respective control (0 mg/mL). The effects of species, extraction condition and concentration and their interactions are presented in Table 7 and Table 8 and are described below. The species × concentration interaction was significant only for S. typhimurium (p < 0.05) and was excluded from the statistical analysis of the other bacterial species.3.3.1. The Effect of the Different Extraction Conditions on the Antibacterial and Prebiotic Effects of L. hyperborea and L. digitata ExtractsETEC and S. typhimurium: There was a species × extraction condition interaction, whereby LHE1 had more potent antibacterial activity than LHE2, LHE3 and LHE4, whereas the effect of the E1 extraction condition was not as potent with L. digitata, despite being significant (p < 0.05, Table 7). B. thermophilum: There was a species × extraction condition interaction, whereby LDE4 was more bifidogenic than LDE1, LDE2 and LDE3, whereas the effect of E4 extraction condition was not as evident with L. hyperborea, despite being significant compared with E1 and E2 (p < 0.05, Table 7). L. plantarum: There was a species × extraction condition interaction, whereby LHE4 was more stimulating on L. plantarum growth than LHE1, LHE2 and LHE3 (p < 0.05) and there was no effect of the extraction condition on L. digitata (p > 0.05, Table 7). L. reuteri: There was a species effect and an extraction condition effect on L. reuteri counts. LH extracts led to higher L. reuteri counts compared to LD extracts (0.24 logCFU/mL (LH extracts) vs. 0.16 logCFU/mL (LD extracts) ± 0.029, p < 0.05). The extraction conditions E1 and E2 increased L. reuteri counts compared to E3 (0.30 logCFU/mL (E1) and 0.27 logCFU/mL (E2) vs. 0.07 logCFU/mL (E3) ± 0.041, p < 0.05).3.3.2. The Effect of Concentration on the Antibacterial and Prebiotic Activity of the Different Extraction ConditionsETEC and S. typhimurium: There was a concentration × extraction condition interaction, whereby 2 mg/mL was more potent than 1, 0.5, 0.25 and 0.125 mg/mL for the E1 extraction condition (p < 0.05), whereas the effect of concentration was not as evident with E2, E3 and E4 extraction conditions (p > 0.05, Table 8). There was also a species × concentration interaction for S. typhimurium (p < 0.05), whereby the 2 mg/mL LH extracts led to lower counts compared to the 2 mg/mL LD extracts (−1.56 logCFU/mL (LH extracts) vs. −0.66 logCFU/mL (LD extracts) ± 0.058, p < 0.05), however, species had no effect at any of the other concentrations.B. thermophilum: There was a concentration × extraction condition interaction, whereby all concentrations of E4 were more bifidogenic than the equivalent concentrations in E1, E2 and E3, where some of the concentrations had no effect, while some were antibacterial (p < 0.05, Table 8). L. plantarum: There was a concentration effect on L. plantarum counts. The concentration of 1 and 2 mg/mL of all extracts increased L. plantarum counts compared to the 0.125 mg/mL (0.15 (1 mg/mL) and 0.15 (2 mg/mL) logCFU/mL vs. 0.02 logCFU/mL (0.125 mg/mL) ± 0.032, p < 0.05). L. reuteri: There was a concentration effect on L. reuteri counts. The concentration of 2 mg/mL of all extracts increased L. reuteri counts compared to 0.125 mg/mL (0.34 logCFU/mL (2 mg/mL) vs. 0.10 logCFU/mL (0.125 mg/mL) ± 0.045, p < 0.05). 4. DiscussionThe influence of seaweed species and harvest season on the effects of the whole biomass samples of L. hyperborea and L. digitata on selected bacterial markers of the porcine faecal microbiota were evaluated in a batch fermentation assay. In this study, seaweed species was the predominant factor affecting the growth of Bifidobacterium spp., Enterobacteriaceae, lactobacilli and total bacteria. Bifidobacterium spp. counts were also influenced by the harvest season. The February-harvested L. hyperborea biomass sample, LHWB-F, led to the lowest Enterobacteriaceae counts among all tested samples whilst also having a reduced negative impact on Bifidobacterium spp. compared to the November-harvested counterpart, LHWB-N. Contrarily, the February-harvested L. digitata biomass sample, LDWB-F, was the least promising in terms of its antibacterial properties, having no major effects on the tested bacterial groups. These two whole biomass seaweed samples were used to produce LHE1–4 and LDE1–4 using four extraction conditions (E1–4) of the HAE methodology. The extracts were assessed in a panel of pure-culture growth assays with selected beneficial and pathogenic bacterial strains, to evaluate whether the optimised extraction conditions could enhance their antibacterial and prebiotic activities. Regardless of the seaweed species, the extraction condition E1 was predominantly associated with improved antibacterial activity against S. typhimurium, ETEC and to a lesser extent B. thermophilum, while the E4 extraction condition was predominantly associated with bifidogenic activity. Total bacteria, lactobacilli, Bifidobacterium spp. and Enterobacteriaceae were monitored in the batch fermentation assay as part of the evaluation of the whole biomass of L. digitata and L. hyperborea collected in February (LDWB-F and LHWB-F) and November (LDWB-N and LHWB-N). The whole biomass of L. hyperborea, LHWB-N and LHWB-F, reduced the Bifidobacterium spp. counts in a concentration-dependent manner with LHWB-F having a lesser impact. The whole biomass of L. digitata, LDWB-F and LDWB-N, also showed evidence of minor reductions in this bacterial population. In addition, LHWB-F and LDWB-N were associated with reduced Enterobacteriaceae counts. Reductions in Bifidobacterium spp. and Enterobacteriaceae counts have been previously observed in the faeces and colonic and caecal digesta in pigs supplemented with crude extracts of L. hyperborea, L. digitata or Laminaria spp. [23,24,25,26]. In this study, whole biomass samples of L. hyperborea were associated with minor increases in lactobacilli and total bacterial counts compared to whole biomass samples of L. digitata. Thus, bacterial growth was predominantly influenced by the seaweed species rather than harvest season, which only had a significant effect on the Bifidobacterium spp. population. It is interesting to hypothesise what the bioactive components within the whole seaweed biomass samples could be based on their proximate composition analysis and the results of the batch fermentation assay. The whole biomass samples of L. hyperborea had higher total polysaccharide content compared to L. digitata for both months. The main polysaccharides present in L. digitata and L. hyperborea are laminarin, mannitol, alginate and cellulose, of which laminarin and mannitol have been reported to exhibit significant seasonal variation in their concentration [35]. This, along with the increase in total glucans (laminarin and cellulose combined) observed in November for both seaweed species in the current study suggests that the variation in the total carbohydrate content was due to laminarin. Fucoidan was confirmed to be a relatively minor polysaccharide in the whole biomass samples of L. digitata and L. hyperborea as expected for the Laminaria spp. [42] and increased in November in both seaweed species. Previous research has reported that laminarin reduced the Enterobacteriaceae counts in the caecum and increased lactobacilli counts in the faeces and colon of weaned pigs [22,27,43], while fucoidan from whole A. nodosum biomass samples was considered to be the bioactive reducing Bifidobacterium spp. and Enterobacteriaceae counts in a batch fermentation assay inoculated with faeces from weaned pigs [18]. The reduction in Bifidobacterium spp. counts could additionally be attributed to inhibitory effects due to the wide-ranging components within the extracts, including phenols, alginate, cellulose and fucoidan, on the activity of bacterial carbohydrate-degrading enzymes [44,45,46]. As whole seaweed biomass samples are inherently complex, it is not possible to attribute the observed effects on the faecal microbiota to specific bioactive components within the whole biomass samples of L. hyperborea and L. digitata with certainty.For the second part of the study, LHE1–4 and LDE1–4 were produced from LHWB-F and LDWB-F, respectively, using the HAE methodology with four extraction conditions (E1–4). Of these, LHWB-F was identified as the most promising antibacterial sample in the batch fermentation assay and was selected for further analysis. In parallel, LDWB-F was included to investigate whether the extraction protocol could improve its limited bioactivity, an effect that was demonstrated in a previous study [18]. LHE1–4 and LDE1–4 were evaluated for their antibacterial and prebiotic potential in a panel of pure-culture growth assays. The pathogens S. typhimurium and ETEC were selected as representatives of the Enterobacteriaceae family. While S. typhimurium infection in pigs is mostly asymptomatic, it is associated with intestinal inflammation and compositional changes in the gastrointestinal microbiota that can have a negative impact on animal health and performance [47,48,49]. Furthermore, pigs and their meat products can become a reservoir for S. typhimurium, which can impact on human health [50]. ETEC infection in newly weaned pigs contributes to the development of post-weaning diarrhoea, an economically significant disease characterised by diarrhoea, dehydration, stunted growth and significant mortality [51]. The effects of the L. hyperborea and L. digitata extracts on representative beneficial bacterial strains, B. thermophilum, L. plantarum and L. reuteri, were also evaluated. These bacterial species commonly colonise the porcine gastrointestinal tract and exert a range of beneficial roles such as inhibition of intestinal pathogens, immunomodulation, improved composition in the gastrointestinal microbiota and enhanced health and growth [52,53,54,55,56,57,58]. In the pure-culture growth assays, the E1 and E4 extraction conditions were predominantly associated with antibacterial and bifidogenic activities, respectively. LHE1 was the most potent extract in reducing S. typhimurium and ETEC counts. LDE1 also inhibited the growth of both pathogenic strains to a lesser extent. Additionally, LHE1 and LDE1 reduced B. thermophilum counts, whereas both extracts were also associated with a slight increase in L. reuteri counts. Interestingly, LDE4 followed by LHE4 increased B. thermophilum counts in a concentration-dependent manner, with LHE4 additionally stimulating the growth of L. plantarum. Based on the above, the use of the E1 and E4 extraction conditions of the HAE methodology produced antibacterial and bifidogenic extracts with the potential to promote a healthy composition in the gastrointestinal microbiota of pigs. The laminarin and fucoidan contents of LHE1–4 and LDE1–4 were determined to establish the concentrations of these polysaccharides achieved by each combination of extraction conditions of the HAE methodology [17]. LHE1–4 extracts had higher concentrations of laminarin and fucoidan compared to LDE1–4, an expected outcome based on the proximate composition of the respective whole seaweed biomass. Interestingly, both sets of extracts had higher fucoidan content (12.76–14.68% for LHE1–4 and 3.84–5.80% for LDE1–4) than laminarin content (4.94–7.59% for LHE1–4 and ≤0.70% for LDE1–4). The presence of laminarin is reported to be at lower concentrations during the winter months in these seaweed species, in agreement with our observation [35,59]. Apart from laminarin and fucoidan, alginate is a polysaccharide which is present in high and relative stable concentrations throughout the year in both L. hyperborea and L. digitata [35], and could also be a significant component of the LHE1–4 and LDE1–4. While the alginate content of the tested extracts was not determined in the current study, this assumption is supported by the findings of a recent study evaluating an L. hyperborea extract produced using the E2 extraction conditions of HAE methodology [60]. Furthermore, the different extraction conditions (Table 1) could affect not only the content but also the structure of these seaweed polysaccharides in the produced extracts, and hence, their bioactivity. For instance, the use of HCl and increasing temperatures in the extraction protocol was previously associated with changes in the chemical composition (monosaccharide content, sulphation level) and lower molecular weight due to partial hydrolysis of fucoidan and partial depolymerisation of alginate [61,62,63]. Although we did not determine the antibacterial and bifidogenic components of the L. hyperborea and L. digitata E1 and E4 extracts, we hypothesise that fucoidan was likely the main bioactive, with the variation in bioactivities attributed to structural alterations due to the different extraction conditions (Table 1). Regarding the antibacterial activity, this assumption is supported by the following three facts: (1) LHE1 had both higher fucoidan content and stronger antibacterial activity against S. typhimurium and ETEC compared to LDE1, suggesting a connection between this bioactivity and fucoidan; (2) The fucoidan-rich A. nodosum extracts produced using the same E1 extraction protocol also led to significant reductions in S. typhimurium and ETEC counts in our previous studies [18,64]; (3) Depolymerised fucoidans from Laminaria spp., Sargassum spp. and Undaria spp. were reported to have improved antibacterial activity against various pathogenic strains including E. coli and S. typhimurium compared to the parent polysaccharide [65,66,67]. The antibacterial activity of LHE1 and LDE1 against B. thermophilum indicate that bioactives other than fucoidan are involved. The bifidogenic effect of LHE4 and LDE4 may also be attributed to the depolymerised fucoidan fraction due to the similar effects on Bifidobacterium spp. growth of the fucoidan-rich A. nodosum extract produced using the same E4 extraction protocol and depolymerised fucoidans of Laminaria spp. and Sargassum spp. in previous in vitro studies [18,68,69]. Alginate oligosaccharides have also exhibited a bifidogenic effect in pure-culture growth assays [70,71]. Therefore, depolymerised alginate may have contributed to the increases in B. thermophilum, particularly in the case of LDE4. The slight increases in L. plantarum and L. reuteri counts with LHE4 counts and E1-produced extracts, respectively, indicate limited ability of these bacterial strains to utilise seaweed polysaccharides, most likely laminarin [72] and alginate oligosaccharides [71]. Taken together, all of the above results suggest a strong indication that fucoidan is the candidate bioactive responsible for the antibacterial and bifidogenic activities, although other seaweed constituents such as alginate may also contribute to the latter in the E4-produced extracts, particularly LDE4. In future studies, investigation into the chemical composition of LHE1, LDE1, LHE4 and LDE4 would provide better insight into the prebiotic and antibacterial bioactive components of these extracts, which was not possible at the laboratory-scale production of the extracts during the development of the HAE methodology.5. ConclusionsThe species of seaweed was the main determinant of the growth of Bifidobacterium spp., Enterobacteriaceae and lactobacilli when whole seaweed biomass samples were tested in a porcine batch fermentation assay. Whole biomass samples of L. hyperborea (LHWB-F) and L. digitata (LDWB-F) harvested in February were then selected as the most and least promising sources, respectively, for the generation of antibacterial extracts, based on their effects on the Enterobacteriaceae counts in the batch fermentation assay. E1- and E4-produced extracts from both seaweed species were associated with antibacterial and bifidogenic activities, respectively, indicating that the extraction conditions were a more important determinant of bioactivity than seaweed species. Of these extracts, LHE1 was the most potent extract against S. typhimurium and ETEC, whereas LDE4 stimulated the growth of B. thermophilum to the greatest extent. Further compositional characterisation of these extracts is required to facilitate the identification and purification of the bioactive components involved in the observed bioactivities. Nevertheless. these crude extracts, particularly LHE1, merit further exploration in terms of their ability to promote a more beneficial microbiota and, thus, overall health and growth in weaned pigs, as a means of minimising the costs associated with the purification of the responsible bioactives from these extracts.

animals : an open access journal from mdpi

10.3390/ani13061077

PMC10044293

The dairy cattle industry relies heavily on artificial insemination facilitated by cryopreservation of bull semen. Cryopreservation causes a number of injuries to sperm reducing fertility and decreasing economic value. Protective extenders are media made up of a number of compounds. Most studies evaluate the impact of changing the concentration of one or two compounds at the same time. Here we use multivariate statistical methods on a large experimental dataset over twelve different compounds to identify key classes of compounds, key components of those classes, and the sensitivities of post-thaw viability on concentrations of these compounds. These statistical models and this methodology point towards improved cryopreservation protocols an improved understanding of the complex interactions among extender components.

Cryo-injury reduces post-thaw semen quality. Extender components play a protective role, but existing experimental approaches do not elucidate interactions among extender components, semen samples, and post-thaw quality. To identify optimal concentrations for 12 extender ingredients, we ran 122 experiments with an I-optimal completely random design using a large dataset from our previous study. We obtained a maximum predicted total motility of 70.56% from an I-optimal design and 73.75% from a Monte Carlo simulation. Individual bull variations were significant and interacted with extenders independently. 67% of bulls reliably preferred extender formulations to reach maximum motility. Multifactor analysis suggests that some antioxidants may offer superior protection over others. Partial least squares path modeling (PLS-PM) found the highest positive loadings for glutathione in the antioxidant class, glycerol in the CPA class, and fructose in the basic compounds class. The optimal ranges for milk, water, and ethylene glycol were extremely narrow. Egg yolk, cholesterol-loaded cyclodextrin, and nerve growth factor had medium-loading impacts. PLS-PM showed that CPA, osmoregulators, and basic components were the most efficient contributors to motility, while the antioxidant and extracellular protectant classes had less efficiency. Thus, ingredients, concentrations, and interactions of extender compounds are critical to extender formulation, especially when using multiple compounds with the same function.

1. IntroductionArtificial insemination using frozen-thawed sperm is a critical technology in the dairy cattle industry [1,2]. Cryopreserved semen also has a variety of applications for assisted reproductive technology as well as endangered animal preservation [3]. Cryopreservation entails different chemical and temperature steps that place sperm under many biochemical, mechanical, and ultrastructural stresses, resulting in detrimental impacts on post-thaw parameters and fertility potential [3]. The supporting media used for cryopreservation (known as extenders) play a critical role in the protection of sperm in the freeze-thaw process, and classical approaches mainly focus on improving one or two components of extenders at the same time. Understanding the effects of various compounds and their interactions within the media and between bulls requires sophisticated multiple models and statistical analysis that are rarely used [4,5]. Response surface methodology, optimization algorithms, and machine learning approaches have been applied to discover the interactions between media ingredients [4,6,7,8,9]. These methods are quick and cost-effective. For instance, Ramesh Pathy et al. [4] used the statistical tool “Plackett-Burman design” to find that egg yolk, vitamin C, and glucose were very significant for maintaining the motility of human sperm cells. Expanding beyond sperm, Lawitts and Biggers [6] applied simplex optimization for developing mouse embryo culture media and determining requirements for development, and Pi et al. [7] used four types of differential evolution (DE) algorithms to optimize the formulation of multicomponent dimethyl sulfoxide-free cryoprotectants in Jurkat cells.I-optimal design is one of the most flexible ways to plan different types of experiments with many different continuous, categorical, and mixture variables, and it can be used with any number of runs [10]. This design is associated with better predictions, as well as improved factor effect estimate accuracy [10].Alternatively, PLS regression has been widely used in estimating some dependent variables with different independent variables [11,12,13,14]. This analysis takes into account each individual in the population and gets around the problems of collinearity and overfitting [13,14]. Simulations based on the interactions between inputs and outputs are developed with the assistance of the Monte Carlo methodology. Morrell et al. [15] used PLS regression to identify variables related to bull fertility. Argiris et al. [11] used a structural equation model with a partial least squares method to assess the superiority of Holstein bulls as frozen semen producers. Structured equation modeling (SEM) is a good way to determine how active sets of variables interact with each other, and it may be used to assess complicated systems.One of the main types of SEM is the PLS-PM method, which uses simple regressions to shed light on latent variables. As mentioned in Sharafi et al. [3], classical media components are used to make sperm resistant to damage during the cryopreservation process. The most common extender constituents, like cryoprotectants, antioxidants, fatty acids, sugars, and membrane stabilizers, can be thought of as latent variables or global levels. Two or more manifest variables can be related to each latent variable. Each latent variable and the corresponding manifest variables are presented as a block or table. As examples of manifest or single variables, glycerol and DMSO may be related to “cryoprotectants”; SOD and CAT may be related to “antioxidants”; fructose and glucose may be related to “sugars”; and cholesterol-loaded cyclodextrins and docosahexaenoic acid may be related to “membrane stabilizers”. Thus, integrating data from many factors may result in more accurate information and stronger inferences (classifications with lower error rates and predictions with less uncertainty) than a single property [16].This study is a continuation of our earlier work [9], which also used some parts of these data. The objective of this study was to evaluate the single and global interactions of extender components as well as how each individual component can influence post-thaw sperm motility, both directly and indirectly. To do so, we used multiple factor analysis (MFA) and PLS path modeling (PLS-PM) as tools for evaluating post-thaw motilities at the level of a single variable and at the global level. The information obtained from this study is important for establishing the use of these new tools in future studies that aim to find the best ways to use cryopreservation protocols to keep fertility levels as high as possible.2. Materials and Methods2.1. ChemicalsAll chemicals used for media preparation in this study were provided by Sigma (St. Louis, MO, USA) and Merck (Darmstadt, Germany) unless otherwise indicated. 2.2. Animal Management and Semen Collection, Extender Preparation and CryopreservationAs described in Tu et al. [9], semen samples were collected using an artificial vagina from 43 Holstein bulls, aged between 20 and 40 months, regularly used for breeding purposes at Semex (Quebec, QC, Canada). All animal experimental procedures were approved by the University of Saskatchewan Animal Care Committee (UAP 002CatA2018). Samples with standard quality metrics that include a concentration ≥ 1 × 109 sperm/mL, motility ≥70%, and ≤15% abnormal morphology were selected for the experiments. After semen collection, samples were individually processed and divided into four equal aliquots to be diluted with four extenders each day. Semen samples were kept in a water bath at 33 °C while a preliminary analysis of fresh semen was performed to measure the concentration, motility, and morphology. Samples that did not meet minimum quality assessments (namely greater than 109 sperm/mL, motility greater than 70%, and less than 15% abnormal morphology) were excluded from the study.The standard control extender consists of tris base (hydroxymethyl-aminomethane; 2.4 g, w/v), citric acid (1 g, w/v), fructose (1 g, w/v), glycerol (7 mL v/v), (25 mg) gentamicin, 50,000 IU penicillin, and streptomycin 300 µg/mL in 100 mL distilled water. A total of 488 tris-based extenders (122 runs × 4 replicates) were prepared using different ranges of water (0–4 mL), basic tris (3–8 mL, composed of 3.02 gr tris and 1.74 gr citric acid), egg yolk (1–2 mL), milk (0–0.4 mL), fructose (0–125 mg), trehalose (0–165 mg), cholesterol loaded cyclodextrin (CLC) (2–10 mg), glutathione (0–10 uL), melatonin (5–12 uL), nerve growth factor (NGF) (0–500 ng), glycerol (0–0.75 mL), and ethylene glycol (0–0.85 mL).To freeze semen, samples were diluted with the corresponding extenders in one step dilution protocol and then were cooled to 4 °C for a 4 h equilibration time in each medium with one-step processing, packaged in 0.25 mL French straws, and then frozen in a controlled rate freezer (Digitcool 007262, IMV, L’Aigle, France) as follows: from 4 °C to −12 °C at −4 °C/min, from −12 °C to −40 °C at −40 °C/min, and from −40 °C to −140 °C at −50 °C/min before plunging into liquid N2. After at least 24 h of storage, one frozen straw per individual bull was analyzed. To do that, frozen straws were thawed in a 37 °C water bath for 45 s and were immediately analyzed for total motility and progressive motility.2.3. Measurement of Post-Thaw RecoveryWe used total and progressive motility post-thaw motility as the main metrics in models to analyze and interpret our multiple statistics. We used a Sperm Class Analyzer (Microptic, Barcelona, Spain) that captures video at 50 frames per second. The diluted semen (2.5 µL) was placed on a pre-warmed chamber slide (33 °C, Life optic slide, 20 µM depth chamber), and motility metrics were determined using a phase-contrast microscope (Nikon, Mississauga, ON, Canada) with a 10× objective at 33 °C. The following cut-offs were applied for all CASA analyses: the range for particle size was defined as 8 to 150 µm2; VCL was used to characterize immotile cells (VCL < 40 µm/sec), slow cells (VCL between 40 and 80 µm/s), medium cells (VCL between 80 and 150 µm/s) and fast cells (VCL > 150µm/s); and progressive motility was defined as >85% STR. A minimum of 300 spermatozoa from 8 fields in 50 frames of video were acquired for each analysis.2.4. Experimental Design and Multiple Statistics AnalysisAn I-optimal, completely randomized design with a quadratic model and 122 runs were used to develop the best set of factor levels and evaluate main effects and two-way interactions among the 12 main components of the extender. A coordinate-exchange algorithm was used to construct the I-optimal design.We used Design Expert software (version 13, 2021) to build runs and do analyses of variance for total and progressive motilities. The partial least squares (PLS) regression, cluster analysis, and Monte Carlo simulation were done by using the Microsoft Excel XLSTAT program (Version 2019.2.2.59614). Multiple factor analysis (MFA) and PLS path modeling (PLS-PM) were performed with R (version 4.2.2, R Core Team, 2022), utilizing the “FactoMiner” [17] and “plspm” [18] packages, respectively.The Statistical Analysis System (SAS) version 9.4 (SAS Institute, Cary, NC, USA) was used to find outliers in a set of data that included all the observations from four experiments (replications). To figure out which observations were outliers, residuals were calculated for each one, and box plots were used to get rid of the 88 observations with the highest residuals. First, a random number function in Excel was used to mix up 400 data points. Then, the data were split into two groups: the calibration data subset (75%, 300 records) and the validation data subset (25%, 100 records). Data subsets have been used to determine the efficiency of PLS models [12]. The coefficient of determination (R2), root mean square error (RMSE), mean absolute percentage error (MAPE), and relative percent difference (RPD) were used to measure the effectiveness of models (Equations (1)–(4)) and were defined as:(1)R2=1−∑in(Mi−Pi)2∑in(Ai−Pi)2, (2)RMSE=1n∑i=1n(Pi−Mi)2, (3)MAPE=100n∑i=1n1Mi|Mi−Pi|, (4)RPD=SDRMSE, where Pi, Mi, Ai, and SD are the values of the predicted, measured, average, and standard deviation of total motility or progressive motility, and n is the number of measuring points.For PLS regression, the data included two dependent (response) variables: total and progressive motilities, as well as 12 independent (explanatory) variables: water, tris, egg yolk, milk, fructose, trehalose, CLC, glutathione, melatonin, NGF, glycerol, and ethylene glycol. The analysis was done in steps, such as making a correlation matrix between the variables, testing the model’s quality based on the number of components, and estimating the model’s parameters so that prediction equations for both total and progressive mobilities could be made. Jackknife cross-validation was performed, and bulls were taken into account in the PLS. The usual loading plot (w/*c map) of PLS regression is presented (Figure 1). Using the agglomeration method of the unweighted pair-group average (UPGMA) and the Euclidean distance, a cluster analysis was done on three PLS components (t1, t2, and t3) to evaluate variations and similarities among the bulls.In order to estimate trends and the expected magnitude of differences in motility for different media components and their variation, a Monte Carlo simulation model was made using 12 independent variables with normal distributions and one result for either total motility or progressive motility (multiple linear models) from the PLS. The component concentrations of the extenders were determined by 10,000 simulations.An MFA can be used to account for the variable distribution in different subspaces that they generate [17]. On the MFA, the analysis of the groups of variables on manifolds or complementary (global analysis) is balanced, and each group’s structure is respected. In fact, the MFA analysis can be used to show how the sets of variables in a common space relate to each other. Multiple-factor analysis takes the complex data of the correlation matrix, which is often hard to understand and reduces it to a smaller number of factors (dimensions) [19].Fourteen different parameters were put together into new active continuous sets of variables called “basic compound”, “extracellular protectant”, “CPA”, “osmoregulator”, “antioxidant class”, and “motility” (cf. Figures 6 and 7). This was done to group parameters that work well together and represent a complementary or global analysis. One supplementary categorical group, “bulls”, was added to the groupings to help with the analysis’s interpretation. MFA analysis can be supplemented by the PLS-PM method, which uses simple regressions to find out more about latent variables. PLS-PM is a combination of two models: a measurement model (also called an “outer model”) and a structural model (also called an “inner model”). The measurement model shows how manifested variables relate to latent variables in blocks or tables. Each block is made up of manifest variables and represents a latent variable. Using linear regression, the structural model explores the connections among latent variables (Sanchez, 2013).In the present study, the measurement model had five blocks: “basic compound”, “extracellular protectant”, “CPA and osmoregulator”, “antioxidant class”, and “motility”. Each block included a latent variable and the corresponding manifested variable. According to Sanchez’s (2013) method, correlations between variables defined the construct type (reflective or formative). Positive and significant correlations are required for reflective constructs, while they should be avoided for formative constructs. Examining the unidimensionality of these blocks using Dillon-Goldstein’s rho confirmed this criterion. A block is unidimensional if its rho value exceeds 0.70 [18]. The structural model specified the block-to-block linkages based on our prior review [3]. The other blocks don’t explain the exogenous latent variables, but they do explain the endogenous latent variables. To validate that the blocks were correct, the PLS-PM model was used to figure out the R2 coefficient of each exogenous block, which showed how much variation each block could explain. When the R2 coefficients are high, other latent factors better explain the block. The link between blocks was clarified via path coefficients. The robustness of the model was finally assessed using the goodness-of-fit test [18].3. Results and DiscussionIn this study, we used different multivariate statistical approaches to find the best combination of media components for bull semen cryopreservation media. Each tool looked at relationships from a different angle and had different strengths and challenges.With I-optimal design, PLS, and Monte Carlo simulation, less than 500 different experimental datasets were used to find the best media composition. These methods were necessary for many experiments to find the best way to make the above compositions. These methods have been used in several fields, including agriculture, the food industry, remote sensing, physics, etc. [10,12,13,14]. In our previous study, Tu et al. [9] used artificial neural networks and Gaussian process regression, supervised learning models, to optimize the media components of bull sperm cryopreservation.We analyzed how the MFA and PLS-PM methods show the direct, indirect, synergistic, and opposing effects. Our MFA suggested that some antioxidants may have more beneficial effects on motility. We also found that the immediate effect of some antioxidants on motility may have been negative.A fit summary suggested a linear regression between motilities and components (Table 1). The models were both significant (p < 0.01) and reliable (R2 > 0.8). The main effects of milk, fructose, and trehalose, and the two-way interactions of water × melatonin, tris × melatonin, milk × fructose, milk × trehalose, and trehalose × glutathione on progressive motility were significant (Table 1). Analysis of variance showed only a two-way interaction between CLC and melatonin (Table 1). Higher-level interactions may be significant when the main effects and two-way interactions are insignificant. This means some components can neutralize the effects of other ingredients, and some compounds, especially those with the same role, should be screened for future experiments. Similar main and interaction effects were found by Ramesh Pathy et al. [4], who reported that the effects of egg yolk, vitamin C, and glucose on the motility of human sperm were significant, and by Dorado et al. [20], who observed that a higher concentration of egg yolk made cold storage of dog sperm more effective for preservation. We are unaware of other studies that have examined more than two factors simultaneously.Various quantiles of total motility measured for extenders obtained from the I-optimal design are presented in Table 2. Extenders 1 and 7 had the minimum and maximum total motility and progressive motility, respectively. The critical message of Table 2 is that we expect to observe higher total and progressive motility with more fructose, trehalose, and glycerol and less NGF and ethylene glycol.Relationships between motilities as dependent variables and 12 components as independent variables, as well as coefficients of the models for total motility and progressive motility projection, are shown in Table 3. These equations were then used for Monte Carlo simulations. Water, egg yolk, glutathione, melatonin, NGF, and ethylene glycol have negative coefficients. This means that using these compounds in large quantities has harmed motilities. Considering these parts’ direct and indirect effects simultaneously is essential [21]. The PLS models had high precision (R2 ranged from 0.64 to 0.70) and accuracy (RMSE ranged from 7.35 to 7.76) (Table 3).The independent variables closest to the dependent variables (total and progressive motilities) are assigned as positive effect components. In contrast, those far from the dependent variables are given as negative effect components (Figure 1). When PLS loadings (component 1 vs. component 2) were analyzed, it was found that the variables trehalose, fructose, CLC, glycerol, milk, and tris were all positively linked to total and progressive motilities.Figure 2 shows three PLS components (t1, t2, and t3) used for cluster analysis, showing four main groups of bulls. The first group consisted of 28 bulls; the second, 10; the third, three; and the fourth group, only one bull. This shows that the viability after thawing varies from one sample to the next, and according to the most significant group, only 67% of bulls are nearly the same. These differences between these cohorts of bulls can be explained by the inherent inter-individual genetic variations and ejaculates within each bull. To understand whether the ejaculates within each bull are significantly different, samples at other time points must be taken from the same bulls. Salinas et al. [22] found that the response of swamp buffalo sperm to freeze-thaw procedures was not always the same among bulls. In contrast, however, Cardoso et al. [23] found that, in dogs, there was no significant within-male or among-male difference in post-thaw semen motility. This factor is vital because Argiris et al. [11] found that the quality of frozen semen production was the highest determinant of the superiority of any individual bull.The associated values of R2 suggest that the PLS models could predict 74% and 67% of total and progressive motility variations, respectively (Figure 3). These were excellent quantitative models based on RPD. RPD values of about 2 are ideal for evaluating models [12].The typical deviation is the standard deviation of a random variable with a Gaussian distribution. In the natural sciences, normal distributions are often employed to describe random variables with actual values whose distributions are unknown. We expect a decrease in marginal productivity as levels of the limiting factor rise. In Figure 4, there is a degree of distortion (skewness) in the symmetrical bell curve for components that indicates whether higher or lower quantities are required. In other words, when some parts are more asymmetrical than others, less or more of those parts are needed. For example, ethylene glycol has a right (positive) skewness, which means a lower concentration is necessary. We observed that glycerol has a left (negative) skewness, which means we can use higher concentrations in future experiments. In addition, we can limit the range of components. For example, the optimal content for egg yolk is 1.8 to 2.0 mL.Ethylene glycol and NGF had the most unfavorable impacts on total and progressive motility, respectively (Table 4), while glycerol and trehalose had the most favorable results. The cryoprotective effects of glycerol and trehalose on the motility parameters of bull sperm have been shown in different studies [24,25,26,27,28]. In contrast, Foote et al. [29] discovered that trehalose did not increase the fertility of bull sperm frozen in whole milk. As a mechanism, Xu et al. found that adding trehalose can change sperm amino acid synthesis and the glycerol metabolism pathway [30].The components and concentrations of top performing extenders are shown in Table 5 and Table 6. The total motility of top extenders ranged from 71.3% to 73.3%, and progressive motility varied from 58.4% to 65.7%.The MFA model was used to determine which parameters are more important for increasing motility. This evaluation is shown by the “bulls” supplementary variable. According to the MFA scatter plots, the first two dimensions explain 45.42% of the total variance (Figure 5). In the MFA contribution plot, the “motility”, “CPA”, and “osmoregulator” groups have a strong correlation with dimension 1, while the “antioxidant class”, “basic compound”, and “extracellular protectant” groups have a strong correlation with dimension 2. The supplementary variable, “bulls”, does not alter the MFA dimensions, but supports the explanation of the findings. In fact, in this figure, the “bulls” variable is immediately next to motility and shows that motility isn’t always the same among bulls and there is significant bull-to-bull variation. Reports show that bull breeds have a significant effect on sperm quality and that different breeds need different measures of sperm quality [15].The current study proposed that CPA and osmoregulator parameters had a significant favorable effect on motility. This is supported by the fact that they have a higher correlation with dimension 1. The correlation between dimension two and “antioxidant class” was more robust, suggesting that some antioxidants may have helped increase motility. Considering the biplot at the single parameter level informs how the different variables are connected (Figure 6). The strongest correlation with dimension one was found between total motility, glycerol, progressive motility, trehalose, fructose, milk, and melatonin, in that order (Figure 6). However,, glutathione, melatonin, tris, NGF, CLC, and egg yolk, in that order, had the most vital positive relationship with dimension 2. Most parameters were linked to dimension 1; both dimensions had melatonin in the joint.According to the MFA data, determining which factors have the most vital relationships with motility remains to be determined. With this analysis, we cannot decide on and measure the relationship between the factors and the post-thaw viability of the sperm at the levels of a single variable and a global construct. Therefore, a PLS-PM model was used to study how well single and group variables fit together. It must be noted that the primary objective of using different methods to judge the quality of post-thawed sperm is to improve the synergy between parameters so that better predictions can be made. The overall goodness-of-fit for the PLS-PM was 0.427. In other words, about 57% of the total difference can’t be explained. This could be because the bulls being considered are different or because low or high amounts of some components are insufficient for motility.The findings obtained in the MFA are supported by Figure 7. The parameter with the greatest loading value among antioxidant markers was glutathione. Glycerol had the highest loading of all the CPA and osmoregulator parameters. The largest loading among essential compounds was associated with fructose. The loadings of milk as an extracellular protectant, water as a primary compound, and ethylene glycol as CPA were negative. Although egg yolk had a favorable loading, it showed a low loading value. The correlation coefficients between the manifest and corresponding latent variables were lower than 0.5 for the melatonin variable in the antioxidant block, the tris variable in the basic compounds block, and the egg yolk variable in the extracellular protectant block (Figure 7). Öztürk [27] pointed out that freezing extenders containing modest concentrations of cryoprotectant combined with trehalose reduced the sensitivity of ram spermatozoa to cryopreservation and osmotic damage.Except for the extracellular protectant block, the R2 value for all endogenous blocks was greater than 0.2 (Figure 8). The R2 coefficients for the latent endogenous variables were ordered as follows: motility > CPA and osmoregulator > antioxidant class > extracellular protectant (Figure 8). The “motility” block was positively linked to the “CPA and osmoregulator” block (0.64) and the “basic compound” block (0.16), but it was negatively related to the “antioxidant class” block (−0.19) and the “extracellular protectant” block (−0.13). The “CPA and osmoregulator” block was positively correlated with all other blocks. The “antioxidant class” block had a positive correlation with the “CPA and osmoregulator” block (0.42) and the “extracellular protectant” block (0.16). There was a negative relationship between the “basic compound” block and the “extracellular protectant” block (−0.31). We expected to see positive direct effects of antioxidants on motility but did not. Thus, removing antioxidants with lower loadings like melatonin and NGF may be advantageous, with the possibility of substituting more powerful antioxidants like proline [31] and humanin [1].In this study, we aimed to develop the best-performing extender while exploring the interactions between various ingredients. Extender formulations play a significant role in the standard cryopreservation protocol, and each ingredient has a specific function such as cryoprotectant, antioxidant, energy metabolism function, and membrane function during freeze-thaw [3]. We observed multiple interactions between each individual component, and this is not possible to discover with conventional experimentation such as comparing two components at a time [9]. With our multivariate statistical approaches, we observed high within- and among-bull post-thaw total motility and progressive motility from sample to sample.This is not surprising given that each ejaculate was from a different bull and given that each ejaculate has different biochemical signatures and physiological characteristics [32]. The fact that each individual bull showed a different reaction to a standard freezing protocol has been previously reported in the boars [33], horses [34] sheep [35], and dogs [36]. This phenomenon is independent of breed, genetic background, or diet [37]. However, differences in genomes, proteome, and epigenomes have been identified as markers of variation. Our multivariate statistical approach can be used as a robust tool to predict widely varying post-thaw fertility within and among bulls..Our multivariate approach predicted a small variation in the post-thaw motility even though the optimal ranges for some of the components, such as NGF, were large. This may indicate that the interaction effects between NGF and other compounds are larger than other mutual compounds. This is aligned with a study in which NGF was reported as a detrimental factor for semen cryopreservation [38]. Therefore, many other factors, such as types and concentrations of ingredients as well as processing, can have an impact on NGF competency. It must be noted that semen samples are extremely heterogeneous, with sampling day, breed difference, and seasonal variation producing samples with large variability [39].Additionally, in one of our previous studies, we showed that antioxidants and other ingredients of the extender can only provide better protection to semen with poor pre-freeze quality [40]. These pre-freeze quality indicators include a wide range of parameters such as motility, viability, membrane functionality, and, importantly, the antioxidant level of seminal plasma. It has been shown that semen with sufficient antioxidant levels is better able to resist freeze-thaw stress and therefore might not need additional antioxidants or other protectants [41].We selected total motility and progressive motility parameters to compare different extenders because they are the main quality control parameters for selecting semen with the highest fertility potential [42]. There is a positive significant correlation between motility and fertility. In our study, post-thaw total motility and progressive motility were quite comparable to our standard control extender that is regularly applied for artificial insemination. The post-thaw motility obtained by this approach was comparable to commercial and industry standards; however, only a field fertility trial will provide complete evidence of the true fertility of each sample as a function of the preservation medium.4. ConclusionsEven though there was a positive relationship between total motility and some media components, some of these components had a negative direct effect on motility. To design protocols for future experiments, the types and amounts of antioxidants and extracellular protectants must be changed and improved. In addition, milk, ethylene glycol, and water, in that order, had direct and indirect negative effects on total motility, which should be assessed to determine whether the toxic effects are due to type, amount, or both. The correlation between motility and milk was positive, but the direct effect of extracellular protectants on motility was negative. This negative effect can be explained by the fact that milk has a high negative loading and egg yolk has a very low positive loading. The direct effect of basic compounds on motility was positive. This can be attributed to the positive loadings of fructose and tris, but not to the negative loading of water. Glycerol and trehalose showed positive correlation coefficients with motility and had the highest positive path coefficient as an “CPA and osmoregulator” block with them. In contrast, ethylene glycol had a significant negative effect on motility.

animals : an open access journal from mdpi

10.3390/ani12020145

PMC8772550

The gastrointestinal tract is a complex organization of various types of epithelial cells forming a single layer of the mucosal barrier, the host mucosal immune system, and microorganisms termed as gut microbiota inhabiting this area. The mucosal barrier, including physical and chemical factors, spatially segregates gut microbiota and the host immune system preventing the development of immune response directed towards non-pathogenic commensals and dietary antigens. However, for the maintenance of the integrity of the mucosal surfaces, cross-talk between epithelial cells and microbiota is required. The microbiome and the intestinal epithelium developed a complex dependence necessary for sustaining intestinal homeostasis. In this review, we highlight the role of specific epithelial cell subtypes and their role in barrier arrangement, the mechanisms employed by them to control intestinal microbiota as well as the mechanisms utilized by the microbiome to regulate intestinal epithelial function. This review will provide information regarding the development of inflammatory disorders dependent on the loss of intestinal barrier function and composition of the intestinal microbiota.

The gastrointestinal tract, which is constantly exposed to a multitude of stimuli, is considered responsible for maintaining the homeostasis of the host. It is inhabited by billions of microorganisms, the gut microbiota, which form a mutualistic relationship with the host. Although the microbiota is generally recognized as beneficial, at the same time, together with pathogens, they are a permanent threat to the host. Various populations of epithelial cells provide the first line of chemical and physical defense against external factors acting as the interface between luminal microorganisms and immunocompetent cells in lamina propria. In this review, we focus on some essential, innate mechanisms protecting mucosal integrity, thus responsible for maintaining intestine homeostasis. The characteristics of decisive cell populations involved in maintaining the barrier arrangement, based on mucus secretion, formation of intercellular junctions as well as production of antimicrobial peptides, responsible for shaping the gut microbiota, are presented. We emphasize the importance of cross-talk between gut microbiota and epithelial cells as a factor vital for the maintenance of the homeostasis of the GI tract. Finally, we discuss how the imbalance of these regulations leads to the compromised barrier integrity and dysbiosis considered to contribute to inflammatory disorders and metabolic diseases.

1. IntroductionThe mucosal surfaces form one of the major barriers protecting the host against invasion and systemic dissemination of both pathogens and local microbiota. In humans, mucosal surfaces cover over 400 m2 and can be classified into two types based on their distinct characteristics: type I mucosal surfaces that include the gastrointestinal (GI) tract, respiratory and female upper reproductive systems, and type II mucosal surfaces that are found in the visual, mouth alimentary, and female lower reproductive systems [1]. The basic difference is that type I mucosal surfaces are covered with the simple columnar epithelium, major secretory antibodies are immunoglobulins A (sIgA), and polymeric immunoglobulin receptor (pIgR) is present, while stratified squamous (non-keratinized), epithelium occurs on type II mucosal surfaces, the dominant isotype is IgG instead of IgA, and pIgR is absent [1]. The GI tract, together with gut-associated lymphoid tissue (GALT), is a specific system that is constantly challenged with contradictory stimuli. This is a region where the organism encounters more antigens than any other site of the body. In addition to its functions in food intake, digestion, absorption of food-derived nutrients, water and electrolytes exchange, endocrine and paracrine hormones production, it has to discriminate immediately between invasive pathogens and harmless food antigens as well as microorganisms that form gut microbiota. For pathogens, the major portals of entry are skin and mucosal surfaces. Enteroinvasive pathogens trigger a strong protective immune response that shields the GI tract, and by that means the host, against the development of the disease and further dissemination of the pathogens. However, such immune response towards non-pathogenic microorganisms that inhabit the GI tract and delivered dietary antigens will be wasteful; moreover, it can provoke hypersensitivity reactions and inflammation, accountable for barrier damage, and may lead to inflammatory disorders. As a result, specific mechanisms responsible for the tolerance induction had to develop to maintain the integrity of mucosal surfaces. The GI tract evolved together with intestinal microorganisms, resulting in the development of immune tolerance, dependent on the lack of responsiveness to the gut microbiota, rather than immune ignorance. Currently, the gut microbiota is considered a biological barrier, protecting against colonization of the GI tract with pathogens. Accordingly, this mutual GI tract-microbiota partnership does not lead to the stimulation of the immune response, however, at the same time, the physical barrier provides the sensing and defense mechanisms ready to protect against invading infectious agents. This recognition requires microbial sensing by host cells, which is carried out by pattern recognition receptors (PRRs), including Toll-like receptors (TLRs), nucleotide-binding oligomerization domain [NOD]-like receptors (NLRs), C-type lectin receptors (CLR), retinoic acid-inducible gene I [RIG-I] receptors (RLRs), absence in melanoma 2 [AIM]-like receptors (ALRs), which recognize microbial-associated molecular patterns (MAMPs) that are molecular structures essential for microbial survival or damage-associated molecular patterns (DAMPs), released from host cells facing injury or molecular stress [2,3]. Activation of PRRs induces several intracellular signaling pathways resulting in the cytokines and chemokines secretion as well as the transcription of other genes important for initiating and controlling the immune response. Since bacteria are the major component of microbiota, the intracellular signaling pathway dependent on TLRs most often is triggered, however, the involvement of other PRRs also occurs. TLR signaling can be organized based on the adaptor proteins involved in the intracellular pathway: myeloid differentiation primary response gene-88 (MyD88), engaged by TLR1, TLR2, TLR4, TLR5, TLR6 and TLR10, and TIR-domain–containing adapter-inducing IFN-β (TRIF), employed by TLR3, TLR4, and TLR9. Studies conducted on mice with disrupted either specific TLR or MyD88 indicated that sensing and recognition of commensal and/or pathogenic bacteria, as well as DAMPs via these receptors, is critical, on one hand, to the enforcement of protective barrier, and on the other hand, for signaling induction that, subsequently, will lead to the stimulation of protective immune response [3,4]. Inflammatory bowel diseases (IBD) are progressive disorders of the GI tract, broadly classified as either ulcerative colitis (UC), Crohn’s disease (CD), or are IBD-unclassified (IBDU) when they share features of both UC and CD. The incidence and prevalence of IBD are increasing with time and in different regions around the world and the elucidation of their pathogenesis is paramount [5,6]. They are characterized by chronic inflammation of the intestinal barrier. CD can affect all parts of the GI tract, with the primary manifestation site of the terminal ileum, and is characterized by transmural inflammation and epithelioid granulomas. In turn, UC always begins in the rectum, and major manifestation occurs within the colon, with inflammation restricted to the mucosal and submucosal parts of the intestinal wall [7]. The definite etiology and pathogenesis of IBD remain unclear, although genetic predispositions, defects, and alterations of local and systemic immune responses, environmental factors, are considered as triggers for IBD development. Other factors, like the loss of intestinal barrier function and composition of the intestinal microbiota, contribute to the onset of intestinal inflammation that can lead to the development of IBD. However, it still has to be resolved, if dysbiosis can induce IBD or if IBD accounts for gut dysbiosis [5,8].In this review, we will focus on some crucial, innate mechanisms protecting mucosal integrity and being responsible for maintaining intestine homeostasis. Additionally, the involvement and characteristics of some epithelial cell populations that are bricks, building the first line of defense, participate in transferring signals between the luminal environment and immunocompetent cells in lamina propria, will be presented. To our knowledge, the interplay between epithelial cells and the GI tract microbiota is one of the most important interactions required for maintaining the integrity of the intestinal barrier and has recently highly improved. We will concentrate on the significance of this cross-talk in the prevention of intestinal inflammation.2. The Structure and Function of the Intestinal EpitheliumThe physical barrier in the GI tract is formed by epithelial cells. Small and large intestines form a major part of the GI tract and the epithelial cells in these areas differ in their structure and composition. The small intestine consists of three sections: duodenum, jejunum, and ileum. Specific, finger-like projections, seen on the mucosal surface that protrude into the lumen, form villi, increase mucosal area and adsorption rate of the nutrients enzymatically digested in the duodenum. Additionally, invaginations of the small intestine mucosa occur, forming crypts of Lieberkühn, which are comprised, predominantly, of proliferating stem cells localized at the base of the crypts (Figure 1).The large intestine (colon) that connects to the small intestine from one end and the anus at the other, comprises four parts: the cecum, colon, rectum, and anal canal. It is an important part of the digestive system and participates in reabsorbing water, absorbing some vitamins, and processing undigested food material (e.g., fiber). Colonic mucosa is devoid of villi, however, it contains deep tubular crypts that increase in depth towards the rectum. Absorptive, columnar cells, endocrine cells, goblet cells, and stem cells form colon epithelium. There are no Paneth cells. Epithelial cells are renewed every 6 days [10].The gut epithelium is one of the most rapidly proliferating cells in mammals. To maintain epithelial integrity, those cells are continuously replaced by proliferating progenitors, derived from multipotent intestinal stem cells (ISCs), localized in the base of the crypts of Lieberkühn and colon crypts. Constantly dividing stem cells give rise to progenitors that differentiate into mature intestinal epithelial cells (IECs) that can be grouped by the functions, with newly formed epithelial cells migrating upwards toward the villus tip, with one exception. Newly formed Paneth cells, present only in the small intestine, move downwards further into the crypt where they mature and start to play their crucial role in maintaining homeostasis of the gut [11,12]. Epithelial cells can be distinguished by their ability of proliferation, the renewing rate, and age (Figure 1) [11]. The aged IECs undergo apoptosis and are shed off into the intestinal lumen. Meanwhile, Paneth cells escape from the crypt bottom by cellular fragmentation and are phagocytosed by macrophages infiltrating from lamina propria [13]. It is estimated that under homeostatic conditions the entire ileal crypt is replaced every 4–5 d [10,11,14].Various differentiated cell populations in both, small intestine and colon, comprise (i) the most numerous population of enterocytes, named colonocytes in the large intestine, (ii) secretory lineages, such as goblet cells—source of the mucus, (iii) enteroendocrine cells, which secrete peptides and hormones (cholecystokinin, serotonin), which stimulate intestinal peristaltic movements that renew the mucus layer, (iv) tuft cells, participating in the clearance of parasites from intestinal lumen by synthesizing IL-25 and, subsequently, polarizing Th2 immune response, (v) Paneth cells in the small intestine or deep crypt secretory cells in the colon (vi) and microfold (M) cells, located within the follicle-associated epithelium (FAE), overlying Peyer’s patches and binding luminal antigens to transport them to the subepithelial regions, where they are captured and processed by dendritic cells (DCs), that afterwards migrate to mesenteric lymph nodes (MLNs), and stimulate immune response (Figure 1) [14,15,16,17]. It has to be noticed that M cells mainly located in the epithelium overlying Payer’s patches can also be found in isolated lymphoid follicles, appendix, colonic patches, and nasopharyngeal associated lymphoid tissue. They act as the entrance gate and deliver luminal antigen to the lamina propria localized immunocompetent cells in the small intestine. However, in the colon under inflammatory or infectious conditions, M cells are responsible for increased bacterial translocation and they enhance inflammatory response [18,19]. All these intestinal epithelial cell populations collaborate to form a protective barrier that confines gut microbiota in the intestinal lumen. At the same time, epithelial cells serve as a link to immunocompetent cells localized in lamina propria, the effector site of the GALT, sending information through direct cell-cell contact or mediators (cytokines and chemokines), thus priming immune response or immune tolerance. GALT includes multi-follicular lymphoid tissues, such as Peyer’s patches of the small intestine and numerous isolated lymphoid follicles that are distributed along the length of the small and large intestines. The specific structure of Peyer’s patches, located just below FAE, allows for the direct contact of antigens delivered to the basolateral area, with cells localized in the subepithelial dome (SED). SED is dominated by professional antigen presenting cells (APCs) such as DCs and macrophages, as well as diverse populations of T cells and B cells of IgA, IgG, and IgM isotypes. Bacteria that penetrated the epithelial barrier without triggering a stronger immune response are removed by resident macrophages, and upon inflammation other macrophage populations are recruited to this area. The interplay between DCs and epithelial cells is very important. DCs as the professional APCs, can interact with transported antigens or send their dendrites through the intercellular junctions of epithelial cells to the lumen, for sampling luminal antigens, and they also respond to secreted cytokines and chemokines. Subsequently, antigen-primed DCs interact with T cells residing in SED, and this interplay results in antigen-specific immune response stimulation, or T cell polarization towards tolerance induction. Of special importance are B cells that are activated in SED via their BCRs by M cells-delivered antigens, secrete IgA that transferred by transcytosis through enterocytes into the lumen, control intestinal microbiota and invading pathogens. These interactions occurring in lamina propria are bidirectional, and immunocompetent cells, through released mediators, influence the physiology of epithelial cells. The important factors are IL-10 and TGF-β that are produced by many cells, including CD4+ T cells, some populations of macrophages, and other cells located in lamina propria. These two cytokines act as anti-inflammatory factors limiting the expansion of effector cells, inducing proliferation of regulatory T cells, and maintaining immune tolerance [20,21,22,23]. Thanks to such an enormous complexity of IECs, their cooperation with GALT and with microbiota, the regulation and maintenance of the GI tract homeostasis, and immediate mobilization of defense against hazardous signals is possible.3. Epithelial Cells—Their Involvement in the Formation of the GI Tract Protective BarrierThe major functions of epithelial cells in the GI tract are ‘segregation’ and ‘mediation’ to maintain the state of unresponsiveness towards microbiota. The term ‘segregation’ is simply the separation of microorganisms and/or their components in the gut lumen from the sterile, deep tissue. This separation is achieved, partially, by forming two types of barriers: physical and chemical, to spatially isolate luminal gut microbiota and immune cells localized in lamina propria and intercalate epithelial cells to prevent the development of inflammation. The physical barrier, commonly regarded as a monolithic wall-like structure, in fact, consists of numerous specialized components, heterogeneous cell types, and intercellular junctions. On the other hand, the term ‘mediation’ represents the delivery of signals from gut microbiota or pathogens to the host immune cells. Epithelial cells are able to react upon signals delivered by microorganisms or their metabolites, as well as signals delivered by gut immune cells through direct cell-cell contact or chemokines and cytokines. Secretion of such mediators results in mobilization of protective immune mechanisms or inducing oral immune tolerance. The physical barrier comprises of the mucus layer that covers the surface of the GI wall, the glycocalyx on the microvilli of absorptive enterocytes, and the cell junctions responsible for tight sealing epithelial cells, forming an obstacle impenetrable for microbiota and pathogens. The chemical barrier is made up of many constituents, like digestive acids secreted by the GI tract, digestive enzymes, mucopolysaccharides, glycoproteins, glycolipids, and other compounds defined as antimicrobial peptides (AMPs). The biological, physical as well as chemical barriers, form mutually dependent microecosystem that, together with GALT, are responsible for maintaining the GI tract homeostasis.The first barrier in the GI tract is mucus. It coats the surface of the intestine forming an obstacle that protects against binding and infiltration of the epithelium not only with the intestinal microbiota but also with enteric infectious agents as well as digested food antigens and food-associated toxins. Continuously secreted mucus forms a protective layer, expresses a very important feature–viscosity, which is balanced to ensure entrapment of microbiota, pathogens, but at the same time enables the flow along the mucosal surfaces. This protective layer presents different properties in the stomach, small intestine, and colon. Mucus, a viscous fluid, is secreted by the goblet cells, named by their shape, featuring the ability to store and secrete granules into the lumen of the intestine. In the large intestine, where the density of microorganisms is much greater than in the small intestine, the number of goblet cells is much higher. 3.1. Mucus Barrier FormationMucus is formed primarily of mucins, highly glycosylated large proteins characterized by the presence of 3 domains: the N’ terminal domain, central domain–composed of protein core (called PTS sequences), containing the amino acids residues: proline, threonine and serine, and C’ terminal domain. The protein core of mucin is protected from endogenous protease degradation, due to glycan formation dependent on O-glycosylation within the Golgi apparatus [20,24,25]. There are more than 20 subtypes of mucins identified in humans. They are either secreted or membrane bound-mucins, and their presence varies throughout the GI tract, for example, MUC6 is found in the duodenum in Brunner’s gland, MUC5B and MUC7 are produced by salivary glands, however, the major, best known and studied mucin found in the small and large intestine, is MUC2, which is built by monomers of about 2.5 MDa and 20% comprise the protein core while the rest is glycan [15,24,26,27,28,29].Before secretion, MUC2 is stored in the goblet cell granules, where low pH, high Ca2+ concentration, and the absence of water, promote organized storage conditions. Upon release, it is necessary to expose mucins to an increased pH and to decreased calcium concentration. Therefore, the Ca2+ is chelated with bicarbonate ions (HCO3−), provided by the cystic fibrosis transmembrane conductance regulated channel, leading to alkalizing conditions and decreasing in Ca2+ concentration, compared to the environment in the mucus granules. In such conditions, the packed mucins unfold to organize spontaneously a sieve-like structure, which allows the passage of small molecules and impedes the migration of microorganisms from the gut lumen [24,30,31,32].Whereas mucus forms the first line of defense coating epithelial cells and contains the same biological components along the whole GI tract, the properties of mucus vary with regional differences. The thickness of the mucus layer differs between the small and large intestine. The major function of the small intestine is digestion of the food and uptake of the nutrients, moreover, the exposure to the microbiota is much lower than in the colon. Therefore, the small intestine expresses a loose, unattached, discontinuous mucus layer that, under experimental conditions, can be easily removed. The detachment of the mucus blanket at a steady state, an essential step for maintaining small intestine homeostasis, is regulated by metalloprotease meprin β that requires the presence of microbiota for its activation [33]. This mucus is also porous and penetrable for different components, including bacteria (Figure 1) [15,34,35,36]. On the other hand, in the large intestine, the thickness of the mucus cover is determined by the number and composition of inhabiting microorganisms. It is organized as two layers: the inner, firm mucus layer, and the outer, loose one. Both layers have almost similar peptide compositions, but there are significant physical differences between them. The inner mucus layer, that remains anchored to the epithelial cells, is highly organized into the flat, lamellar structure, and it does not allow bacteria to penetrate, therefore, at the steady state, it is free from microorganisms. The relative demarcation line divides the inner layer from the outer layer. The outer layer is formed from the inner layer by proteolytic processing of polymerized MUC2, by host bacteria. It thus contains the same components, but is loose, unattached to the epithelial cells, and is inhabited by a large number of intestinal microorganisms (Figure 2) [34,35,37,38]. In addition, the transformation from the inner to outer layer is regulated, at least partially, by CLCa1, a metalloprotease with a high abundance in colonic mucus. Nevertheless, studies on a mouse model with Clca1 deleted gene revealed that, so far, an unidentified cysteine protease can compensate for the lack of the Clca1 gene. Further elucidation of this mechanism is required [39]. 3.2. Role of Goblet CellsAs it was previously indicated, the major source of mucus is goblet cells. They are present throughout the entire length of the intestine, although they occur in larger quantities in the crypts than on the villi in the small intestine, and in the upper crypts in the colon. Like with other epithelial cells, their differentiation starts within the crypts and is promoted by several transcription factors, including ATOH1 and SPDEF, and also by Notch and canonical Wnt signaling pathways, whereas is negatively regulated by HES1 [28,40,41,42,43].The goblet cells, migrating from the bottom of the crypts upward the villus, fill their secretory granules, located below the apical membrane, with the main component, MUC2, while the nucleus and other cellular organelles, concentrated in narrow stem-like subcellular regions, are located at the base of the cells [44]. They also synthesize and secrete bioactive molecules, such as secretory and membrane-bound mucins, trefoil factor family peptide 3 (Tff3), resistin-like molecule β (RELMβ), and Fc-globulin binding protein (FCGBP), which all are components of mucus. Tff3 is a tissue-protective factor that promotes epithelial restoration and mucosal repair by apoptosis inhibition, cell migration, and angiogenesis assistance [14,45]. Other proteins, largely originating from the epithelial cells, shed to the lumen, are trapped and present within the mucus biofilm, including: calcium-activated chloride channel 1 (ClCa1), zymogen granule membrane protein 16 (ZG16), anterior gradient 2 (AGR2), and antibodies, especially IgA.Recent studies focused on the analysis of transcriptomic and proteomic profiles, revealed functional heterogeneity among goblet cell population, defining new functional subpopulations: canonical goblet cells expressing known characteristics for goblet cells genes (e.g., Clca1, Fcgbp) and non-canonical goblet cells expressing genes connected with enterocytes (e.g., Dmbt1, Gsdmc4) [32]. A new population, named sentinel goblet cells (senGCs), localized at the entrance of the colonic crypt, expressing higher levels of Il18 and Nlrp6, a feature characteristic for non-canonical population, was defined. Studies performed on intestinal explants clearly presented that senGCs, featuring non-specific endocytosis, were able to respond to TLR2/1, TLR4, and TLR5 ligands, leading to NLRP6 inflammasome formation and production of reactive oxygen species, which in turn triggered the release of Ca2+ that passed through gap junctions. This resulted in a double-type response: the senGCs secreted their granules and simultaneously sent signals to the neighboring cells. Interesting was that degranulation and MUC2 secretion was sequential. Occurring at first in senGCs, the intracellular signal was then sent through the gap junctions to the closest goblet cells inducing their granules secretion. Consequently, stimulation of goblet cells is progressed around the crypt. Massive MUC2 secretion, initiated by senGCs, was responsible for the physical removal of bacteria from the crypt opening, protecting the lower crypt and ISCs from bacterial invasion. This mechanism was accompanied by the expulsion of activated senGCs to the lumen of the intestine along with remaining endocytosed microbial products [32,46]. Mucus secreted by epithelial cells forms a layer separating microbiota as well as potential pathogens, and protects epithelial cells against colonization or even breaching this thin physical barrier. In the colon, it is achieved by the presence of the inner mucus layer that creates the outer mucus layer, which is more accessible for microbiota, while the small intestine is protected by a mucus layer that is penetrable for microorganisms but saturated with AMPs. Any deficiency in mucus formation may lead to colitis. In the colon, goblet cells localized between crypts, defined as intercrypt goblet cells (icGCs), are the most differentiated cells, and present expression profiles distinct from the crypt-resident goblet cells. Mucus secreted by icGCs fills the area between colonic crypts and is distinguishable from crypt mucus plume. It is not penetrable for bacteria-size beads but it is permeable for smaller size molecules. This aspect is important in the absorption of ions and other low molecular weight compounds, while denser, impermeable mucus within crypts is responsible for the protection of stem cells area. Both types of mucus form the colon inner mucus layer (Figure 2). However, the significance of inGCs in the development of UC, spontaneous (age-dependent), or chemically induced, was proved in the Spdef−/− mouse model, when the formation of functional mucus barrier was altered. Differences in the mucus thickness, composition (crypt goblet cells were the major source of mucus), and the host susceptibility to the development of colitis, due to the alteration of inGCs, were found in comparison to the wild type control. These results correlated with data obtained from patients with diagnosed UC and those in remission. Their intestinal count number of inGCs was reduced, due to the increased cell shedding to the lumen. As a consequence, in those patients, the structural defects in the mucus layer, including gaps in the intercrypt mucus were observed, exposing epithelium to microbial intrusion. These results clearly indicate the role of different subpopulations of goblet cells in providing a tight mucus barrier [32]. In addition to their secretory functions, goblet cells play a key role as the luminal sensors of the antigens for the immune system. The question posed by scientists was how the epithelial barrier simultaneously forms a physical obstacle that distances microbiota from the epithelial cells and, at the same time, allows immunocompetent cells that are present in lamina propria to selectively sample luminal antigens (microbiota and food antigens) to promote the homeostasis. Beside M cells that can sense and deliver luminal antigens to APCs localized in lamina propria, the involvement of goblet cells was elucidated. It was shown that goblet cells after granules secretion, form goblet cells-associated antigen passages (GAPs) that can deliver small, soluble antigens to DCs localized in the lamina propria. Since GAPs formation was increased after treatment of goblet cells with cholinergic agonists (secretion stimulators), it suggested that GAPs arrangement was strongly connected with goblet cells mucus secretion. It seems that GAPs formation has an educational impact on the gut immune system by delivering innocuous luminal antigens detected during homeostasis. Recent studies presented that GAPs not only deliver luminal antigens to the lamina propria APCs, but also are responsible for maintenance of pre-existing regulatory T cells, imprinting of DCs with tolerogenic features, and promoting lamina propria macrophages to produce IL-10. As a result, the formation of the immune tolerance to the dietary antigens is supported. Moreover, in the GAPs absence, this tolerance was impaired, consequently, goblet cells can be considered as the gate-keepers for the orally delivered antigens to the gut immune system [20,47]. 3.3. Role of EnterocytesThe most numerous populations among intestinal epithelial cells are enterocytes in the small intestine and colonocytes in the large intestine. Enterocytes are structurally polarized with the apical surface directed to the lumen and the basolateral surface facing the lamina propria. From the apical site, enterocytes form microvilli, actin-based protrusions that increase tremendously the absorptive surface of nutrients, and are described as a brush border [20,25]. Absorptive enterocytes present the pIgR that mediate transportation of dimeric IgA and polymeric IgM from the lamina propria across the epithelial barrier to mucosal surfaces [48]. Microvilli facilitate the transport of many molecules (nutrients, electrolytes, vitamins, water, bile salts), however, they are not only in constant contact with microbiota and possible enteropathogens, but are also exposed to mechanical stresses associated with peristalsis. Over the microvillar surface, a carbohydrate-rich glycocalyx is formed, which acts as a protective layer between the mucus and the underlying epithelial cells. Recently, it was shown that the glycocalyx of the brush border is formed of the mucins that belong to the transmembrane mucins family [20,49]. These mucins are synthetized by enterocytes and are characterized by a single domain, passing the plasma membrane, and a large glycosylated extracellular mucin domain, with a similar PTS core as gel-forming mucins (e.g., MUC2). All these mucins also have a cytoplasmic tail that interacts with the cell cytoskeleton. The major components of the enterocyte glycocalyx are transmembrane MUC3, MUC12, and MUC17. MUC1 is more abundant in the stomach and only small amounts of this transmembrane mucin are found in the intestine. The glycocalyx layer comprises epithelial membrane glycolipids and glycoproteins that many act as adhesion receptors for microorganisms [20]. Thus, glycocalyx operates as an attachment site for microbiota and limits the potential colonization area for pathogens. Since the mucus layer in the small intestine, predominantly formed by MUC2 secreted by goblet cells, is a non-attached, relatively porous structure penetrable for bacteria, glycocalyx forms the line of defense. Moreover, the glycocalyx helps in the lubrication and hydrophobicity of mucosal surfaces [25,49].4. Intercellular Junctions—A Protective BarrierThe intestinal epithelium forms a protective lining to maintain the barrier integrity and restrict the entry of microorganisms and toxins, and at the same time, allow penetration of essential ions, nutrients, and water. Even if microorganisms penetrate the mucus layer in the small intestine, the inner layer of mucus in the colon or the glycocalyx, they meet a barrier on the level of epithelial cells formed by a structure composed of three junctions, from apical to basal: tight junctions (zonula occludens) (TJs), adherens junctions (zonula adherens) (AJs) and desmosomes (macula adherens). AJs and desmosomes provide adhesive and mechanical features that contribute to barrier function while TJs form not only tight bonds, constituting links between epithelial cells that physically block microbial invasion, but also seal the paracellular space between the cells and tightly restrict the transport of hydrophilic molecules [50,51]. AJs are formed between cells by proteins such as catenins, cadherins, and integrins that are involved in cell surface adhesion, providing mechanical strength between cells to facilitate cell-cell adhesion and polarization. Furthermore, they are responsible for binding with the cytoskeleton and thus participate in cellular signaling pathways. One of the catenins, β-catenin, which is involved in cellular adhesion and cell differentiation, also induces Wnt signaling pathway, which is important in gene expression regulation. Cadherins are present on the membranes of adjacent cells binding each other. E-cadherin, the major constituent of AJs, forms homophilic cell-cell interactions and binds intracellularly to catenins (α-catenin, β-catenin, p120-catenin), which links E-cadherin to the cytoskeleton (Figure 3) [52,53].E-cadherin expression is relatively constant in the small and large intestines, and it contributes to proper cells differentiation and migration from the crypts. Impaired expression of E-cadherin in the small intestine and colon is connected to disturbed gut homeostasis and barrier functions. Overexpression of E-cadherin in the intestine results in retarding cells migration from the crypt to the villus, suppressing their proliferation and induction of apoptosis within the crypt, also delaying the enterocytes differentiation. It has to be said that complete E-cadherin knockout (KO) in mice is embryonically lethal [56], therefore, studies conducted on a mouse model with targeted loss of E-cadherin, somehow elucidated the role of E-cadherin in the GI tract. Loss of E-cadherin causes, among others, accelerated cells migration and aberrant differentiation, as well as villus blunting [57,58,59]. Moreover, the lack of E-cadherin disrupted the paracellular transport pathway, since intestinal epithelium displayed loss of claudin-1 and increased claudin-4 expression, proteins crucial for proper TJs arrangement. Deletion of E-cadherin in adult mice resulted in bloody diarrhea, cell shedding, and deficient cell maturation [5,57,60]. It is clear that E-cadherin is required for maintaining the architecture and function of epithelium since the lack of this protein affects Paneth cells and goblet cells maturation [60]. Moreover, it resulted in reduced AMPs production that led to deficiencies in clearing enteropathogens and predisposed to the development of IBD [61]. Desmosomes, similarly to AJs, provide mechanical strength for the cell junctions by comprising transmembrane proteins termed desmosomal cadherins, divided into the desmoglein and desmocollin types (gene names DSGs/DSCs; protein names: Dsgs/Dscs, respectively). They are abundant also in tissues of high mechanical stress such as the heart and skin. However, in the intestinal epithelium, only Dsg2 and Dsc2 are present [62]. Both types of desmosomal cadherins are required for cellular adhesion, forming homo- and heterophilic interactions. Dsg2/Dsc2 complex interacts with armadillo proteins plakoglobin and plakophilin, via the cytoplasmatic domain. These, in turn, are connected to the intermediate filaments by desmoplakin (Figure 3) [5,63,64]. In vitro studies presented the importance of Dsg2, since DSG2 knockdown resulted in compensation with Dsc2, but deletion of DSC2 did not cause changes in Dsg2 protein level [65]. Gross et al. confirmed in an animal model that the desmosomal cadherin Dsg2 is required for the integrity of the GI tract epithelial barrier in vivo. Furthermore, these results have relevance in human diagnostics, since dysregulation of desmosomes was observed in patients with diagnosed CD [63].Apically located TJ complex is generated by transmembrane proteins and plays a crucial role in maintaining gut homeostasis, as well as controlling the permeability of the paracellular transport pathway. TJs form a continuous network of proteins between membranes of neighboring cells completely closing the apical intercellular space. Additionally, some components of TJs constitute a boundary inside the membrane by itself, restricting the migration of transmembrane proteins and lipids from the apical to the basolateral side and participating in enterocytes polarization [66,67]. The involvement of TJs in the modulation of gene expression, required for cell proliferation and differentiation, was proved to be critical [66,68,69]. The arrangement of the TJ complex is based on the transmembrane proteins that mainly belong to three groups: the claudins family, the Marvel domain-containing proteins (occludin, tricellulin [known as MarvelD2 and MarvelD3 proteins]), immunoglobulin superfamily members (Junctional Adhesion Molecules [JAMs], Coxsackie and Adenovirus Receptor proteins [CAR]). These transmembrane structures are bridged to the actin cytoskeleton and other signaling proteins through a cytoplasmic TJ plaque. TJs plaque is formed by peripheral membrane adaptor proteins, zonula occludens-1 (ZO-1), ZO-2, ZO-3, as well as the cingulin, cingulin-like proteins, and the afadine (Figure 3). TJs establish the backbone of the first line of the host defense by sealing the portal of entry for the pathogens. Various exogenous and endogenous factors regulate the integrity of TJs and, as a consequence, the permeability of the intestine wall. It depends on the involvement of kinase pathways including, protein kinase C (PKC), A (PKA), G (PKG), and MAPK (ERK, p38, JNK) signaling; the calcium/calmodulin-dependent kinase 2 (CaMKK2)-AMP-activated protein kinase (AMPK), as well as Rho and NF-κB pathways [50,67,70]. The regulation and permeability of TJs also depend on the involvement of proinflammatory cytokines, and participation of TNF-α, IL-1β, IL-13, and IFN-γ in the increased gut permeability was highlighted [57,70,71,72]. The structural proteins of TJs play important role in maintaining barrier integrity, although not all proteins functions, so far, were elucidated. Occludin is highly expressed at cell-cell contact and provides structural integrity through the interaction with ZO-1, and phosphorylation of occludin regulates TJs stability as well as permeability [73]. Occludin KO mice expressed morphologically intact TJs structure but exhibited elevated inflammation and a defective gut barrier [74]. Claudins, the family that comprises 23 isoforms, are responsible for the regulation of paracellular space. The extracellular loops of claudin participate in hetero- or homophilic interactions with adjacent cells, which generate barriers or pores for the passage of selective molecules in the paracellular pathway [75,76]. Changes in claudin profiles within TJs, affect the intestinal barrier integrity, depending on the claudin isoform. Therefore, a deficit of claudin-1 in mice leads to the abnormal TJs formation, which induces cancer development and metastasis. Higher expression of claudin-2, a protein required for arrangement of paracellular water channels, together with downregulation of sealing claudin-5 and -8, results in the altered TJs and barrier dysfunction in CD patients and contributes to the inflammation [77,78]. ZO-1, ZO-2, ZO-3 connect claudin and occludin to the actin cytoskeleton, and these proteins interplay maintains TJs function and arrangement. TJs in the intestine seal epithelial cells, forming an effective blockade that protects against pathogen intrusion and any alterations in the TJs structure can be detrimental to the host. Some of the examples were already discussed and certain pathological conditions are strongly correlated with a defective intestinal TJs structure such as IBD, obesity (the impairment of intestinal barrier function, alteration in microbiota), nonalcoholic steatohepatitis, and nonalcoholic fatty liver disease (abnormal morphologies of crypts and villi in duodenal mucosa, alteration in microbiota) [70]. However, it cannot be forgotten that this impermeable intestinal barrier creates the target for pathogenic microorganisms, and they developed strategies to disorganize TJs and translocate through the mucosa to invade the host. Paradis et al., thoroughly reviewed the strategies of enteropathogenic bacteria, fungi, and viruses, aiming for the specific proteins, participating in TJs formation. Pathogens modulation of TJs structure can occur through defined molecules, present on the surface of the bacterial or fungal cell walls, bound to the viral capsids, or secreted by microorganisms. In many cases, pathogens employ virulence factors for the disengagement of TJs proteins from the junctional complex. Furthermore, pathogens can induce signaling pathways responsible for TJs and cytoskeleton down- or up-regulation of gene expression, causing disorder of those arrangements. Breaching this barrier by pathogens stimulates immunocompetent cells localized in the lamina propria, inducing proinflammatory response as well as oxidative stresses of IECs, leading to the enhancement of TJs dysregulation and this can contribute to the development of IBD [67].5. Paneth Cells—Stem Cell Curators and Microbiota ControllersOne of the major populations, responsible for maintaining homeostasis of the GI tract are Paneth cells, specialized epithelial lineage that resides at the base of crypts of Lieberkühn, and is characterized by the presence of eosinophilic granules in the cytoplasm. Paneth cells were described in the late 19th century, first in 1872 by Gustav Schwalbe, and later in 1888, characterized more thoroughly by Joseph Paneth [79,80]. Paneth cells, found only in the small intestine are derived from ISCs, like other intestinal epithelial cells [11,12]. While other, newly formed epithelial cell populations migrate upwards, newly differentiated Paneth cells move downwards further into the crypt, where they mature and start to play their crucial functions in preserving homeostasis of the gut. Paneth cells are considered the major regulators of microbial density in the small intestine as well as protectors of stem cells. There are 5–12 Paneth cells per one intestinal crypt with a span life of 3–6 weeks [11,13,81]. These pyramidal in shape cells, with basally located nuclei, are filled with apically localized numerous eosinophilic granules that, upon exposure to Gram-negative or Gram-positive bacteria or their products (lipopolysaccharide, lipoteichoic acids, lipid A, muramyl peptide), release AMPs and enzymes, which are important host-defense factors in the communication between microbiome and the host. It has been proved that bacteria orchestrate secretion by Paneth cells, several proteins and peptides and that include: (i) lysozyme, enzyme active predominantly against Gram-positive bacteria, conducting disruption of glycosidic bonds in peptidoglycan, resulting in bacterial lysis, (ii) secretory phospholipase A2 (sPLA2), showing bactericidal activity against Salmonella enterica serotype Typhimurium (Salmonella Typhimurium) and Listeria monocytogenes, (iii) regenerating islet-derived protein 3 (RegIII) family, that includes RegIIIγ and RegIIIβ, active against Gram-positive bacteria by binding peptidoglycan, component of the cell wall, and critical in the spatial separation of the intestinal bacteria from epithelium in the small intestine, (iv) CRP-ductin responsible for agglutination of Gram-positive and Gram-negative bacteria, (v) enteric α defensins (cryptdins), (vi) cryptdin related sequence (CRS), peptides binding and reducing immunomodulatory activity of LPS, (vii) cathelicidin family (in human the only representative is LL-37) that work similarly to defensins, forming channels in the cytoplasmic membranes, and many other factors like IL-1β, CRIP, IgA [14,79,82,83,84,85,86,87,88]. Enteric α defensins are a subfamily of defensins peptide family that presents a broad spectrum of peptide antibiotic activity, primarily by disrupting microbial cell membranes. Unlike neutrophils, Paneth cells do not store defensins as processed, mature peptides, but maintain them as pro-peptides requiring processing after secretion. In mice, maturation of pro-defensins depends on metalloproteinase (MMP) matrilysin (MMP-7) processing, and mice deficient in this enzyme lack functional cryptdins and are susceptible to oral challenge with Salmonella spp. Moreover, the animals had a significantly greater percentage of Firmicutes and fewer Bacteroidetes in the gut microbiome, compared to wild-type control mice. In humans, instead of MMP-7 that is not present in the small intestine, the involvement of trypsin, as a processor of pro-peptides, was indicated. Moreover, Ghosh et al., showed that trypsin, as a zymogen, was co-localized with human α-defensin5 (HD-5) in Paneth cells granules. These data suggest that trypsin activation occurs either during or after the secretion of granules [79,89,90,91]. Nevertheless, HD-5, the most abundant antimicrobial peptide produced by humans, is shown to have direct bactericidal activity towards distinct members of the human gut microbiota, and thereby it can alter the human microbiome in vivo. More interestingly, HD-5 exposed in vitro to the natural human duodenal fluid underwent proteolytic degradation, resulting in the formation of several active defending fragments, capable of affecting the growth of both, commensal and pathogenic organisms. In vivo studies confirmed that HD-5 fragments were able to shift microbiota composition without altering its diversity [92].Since the epithelial cells are replaced every 2–5 days, the antimicrobial protection of the crypts, the source of the newly formed epithelium, is cardinal, as the damage of ISCs or bacterial outgrowth of the crypts could lead to serious consequences for the host. These granular components of Paneth cells are assembled and packed into dense granules within the endoplasmic reticulum (ER) and Golgi apparatus. It is possible that some granule components are formed elsewhere and are then collected and put together in a granule. IgA are such components that are synthesized by plasma cells in lamina propria, accumulated, and then loaded to the Paneth cells granules [79]. Paneth cells activity and health are crucial for microbiome modulation and mediation of the inflammatory response. This secretory cells population is characterized by extensive ER and Golgi network. Endoplasmic reticulum membrane complex subunit 3 (Emc3), encoded in mice by Tmem111 gene, is involved in proteins folding required for their maturation. Deletion of Emc3 in the intestinal epithelium affects the differentiation of secretory lineages, goblet cells as well as Paneth cells, resulting in a reduction in the number of the latter ones. The increased level of apoptosis was observed, through which undifferentiated Paneth cells were cleared in Emc3-deficient mice, and that led to a lowering of the final cargo of AMPs, resulting in the growing susceptibility to dextran-sulfate sodium (DSS)-induced colitis and Salmonella Typhimurium infection. These results indicate the role of Emc3 in maintaining secretory lineages in the gastrointestinal tract and thereby protecting against the development of inflammation [93].Nevertheless, the Paneth cells secretory responses remain debatable and the mechanisms that regulate the secretion still are not fully understood. Most of the data, concerning granular contents and secretion by Paneth cells, are based on immunochemistry, and enteroids models, which are three-dimensional cultures of small intestinal epithelial cells [91,94]. It was demonstrated that detection of enteric bacteria by Paneth cells occurs through the MyD88-dependent pathway, but TLR4 independent way [95,96,97,98]. Moreover, the presence of microbiota was required for Paneth cells activation. Studies, performed by Vaishnava et al. on mice, in which Paneth cells were specifically ablated by expressing a diphtheria toxin fragment under the control of cryptdin-2 promoter (CR2-tox176 mice), provided information concerning the involvement of Paneth cells in limiting mucosal penetration by microbiota and pathogenic bacteria. Although in both wild type and Paneth cell-deficient mice the luminal bacterial load was comparable, the number of bacteria recovered from MLNs was higher in CR2-tox176 mice. Similar results were obtained when germ-free (GF) CR2-tox176 mice and GF wild type mice were orally introduced with Bacteroides thetaiotaomicron, and in a parallel experiment, when mice were challenged with Salmonella Typhimurium. These results emphasized the importance of Paneth cells as a mandatory population for limiting pathogens translocation from gut lumen and their dissemination to the tissues [95]. Studies, performed on the three dimensional enteroid models, shed new light on Paneth cells activation and their functions. Since enteroid is a closed structure, with Paneth cells apical sites opened to the lumen of the enteroid, the microinjections with bacterial stimuli (LPS, or Salmonella Typhimurium phoP), were used to evaluate their activation. This experimental model clarified the ability of Paneth cells for a rapid response with AMPs secretion (within 2–14 s), upon only apical stimulation. Moreover, the potential of quick refilling of granule store (21 h post secretion), with newly generated granules and ability to respond to secondary stimuli with comparable strength, was demonstrated. However, basolateral site stimulation is also responsible for Paneth cells induction, leading to the AMPs secretion, but in this pathway, stimulation with, among others, muscarinic nerve stimuli, cytokines (IL-13 and IL-22) are involved [91,99,100]. Various bacterial components, such as LPS, non-methylated CpG oligodeoxynucleotides, and muramyl dipeptides, induce Paneth cells secretion that is followed by regulation of the number of bacteria in the intestine. Although the involvement of microbiota and pathogens in Paneth cells stimulation and subsequent granule release is unquestionable, the other stimuli, controlling their ability to AMPs secretion, should be considered. IFN-γ, a strong modulator of the immune response, was a candidate for Paneth cells activation. However, in vitro studies revealed that exposure of enteroids to this cytokine caused Paneth cells death through caspase 3/7 pathways shortly post exposure. These results are in accordance with other studies, both in vivo and in vitro. They indicate that IFN-γ caused a rapid and complete Paneth cells degranulation, followed by their luminal extrusion and apoptosis. Such a strong degranulation and loss of Paneth cells leads to dysregulation of microbiota control and results in dysbiosis. Moreover, this feature is shared with goblet cells, since depletion of both cell populations (Paneth cells and goblet cells), during infection is induced by IFN-γ. Some mechanisms, regulating granule releasing were defined, however, the exact stimuli that induce Paneth cells to secrete AMPs have to be elucidated [14,91,101,102,103,104]. One of the proposed hypotheses suggests that at first there is an impact of AMPs on bacteria that are apposed to the mucosal surface since bacteria must be mucosa-associated before the uptake by DCs for MLNs translocation or before the direct invasion of epithelium. This is consistent with the fact that AMPs are retained by the mucus layer that overlies the epithelium. Moreover, secreted lysozyme and cryptdins can be recovered as intact, functional form from the lumen of the small and large intestines as well as from feces, suggesting that these factors affect the microbiota composition of both the small intestine and colon, and contribute to the intestinal homeostasis. By specific controlling of the bacteria-mucosal surface interactions, Paneth cells contribute to maintaining the proper composition of luminal microbiota that is essential for the host metabolic health [91,95,105,106]. As discussed, Paneth cells are the major source of AMPs participating in shaping host microbiota. On the other hand, they can be the origin of intestinal inflammation in the host. This point of view has some relevance since the recent studies link the reduced number of Paneth cells or their abnormal function with microbial dysbiosis in the ileum, detected in patients diagnosed with CD. Paneth cells are identified as the site of susceptibility gene defects that impair the secretion of Paneth cell granules. The mutations of CD susceptibility genes Atg16/1 and Xbp1 are linked with defective granule exocytosis from Paneth cells with diminished levels of defensins, due to the abnormal autophagy and ER stress that affects the maintenance of secretory cells [13,91,107,108]. NOD2, a cytoplasmic PRR belonging to NLRs, present in Paneth cells cytoplasm, is critical for regulation of the bacterial microbiome in the ileum through secretion of AMPs. Mutations in the NOD2 gene affect the C-terminal leucine-rich repeat receptor domain that recognizes and binds muramyl dipeptide leading to deficiency in sensing bacteria in the ileum. As a result, the abnormal microbe colonization of the small intestine occurs, which may provoke chronic inflammation, since Paneth cells with NOD2 mutation show reduced secretion of α-defensins. Tcf4 gene mutation of Wnt signaling pathway transcription factor that orchestrates Paneth cells differentiation, is directly linked with impaired cell maturation and diminished α-defensins production, and its corresponding antimicrobial function [13,14,91,108,109]. Microbial dysbiosis is one of the factors of obesity pathogenesis. A strong connection between the pathology of such disease and Paneth cells was found since HD-5 and lysozyme expression was significantly decreased in Paneth cells in obese patients comparing to healthy subjects of normal weight [110]. Furthermore, Paneth cells dysfunction is associated with enteropathy in graft-versus-host disease (GVHD). In GVHD, the intestinal tract is frequently affected, as a result, Paneth cells loss is a crucial point in shifting to microbiota dysbiosis. The GVHD mouse model developed dysbiosis due to depletion of Paneth cells that resulted in the loss of secreted α-defensins, but can be partially reversed after oral administration of α-defensins and that improves GVHD survival [111,112]. Paneth cells develop in the middle of human gestation, however, they do not rich their density and immunocompetence until closer to the term of gestation. Preterm neonates lack fully differentiated Paneth cells and they show great susceptibility for the development of intestinal pathology, such as necrotizing enterocolitis (NEC) [13,79]. Although Paneth cells are localized in the base of the crypts and are confined to the small intestine, in distinct intestinal disorders they can appear in various areas, in which they are not normally found. Paneth cell metaplasia is seen throughout the gastrointestinal tract but can be also found in extra-gastrointestinal sites, like the lungs, tracheobronchial system, urogenital tract, pancreaticobiliary tract, and rarely in the nasopharyngeal system [13,113]. So far, the mechanism of Paneth cells metaplasia is not fully comprehended. The Wnt/β catenin canonical signaling pathway is required for full differentiation of Paneth cells. There is some evidence that this signaling is involved in metaplasia development [41]. Moreover, additional components required for Paneth cells differentiation that differ between tissues can be involved in this mechanism, where inflammation that can induce metaplasia/hyperplasia, seems to be one of them. Indeed, chronic inflammation usually precedes the formation of intestinal metaplasia, like in the case of Barrett’s esophagus, or in the stomach, following Helicobacter pylori infection. Paneth cells metaplasia occurs in the chronic inflammatory state within the colon (UC, colonic CD, diverticulitis), and it was noticed that metaplastic Paneth cells express their regular AMPs. It was suggested that the occurrence of metaplastic Paneth cells at mucosal sites, either intestinal or extra-intestinal, probably represents a protective, antibacterial and inflammatory response evoked by an altered microbial activity [2,13,113,114,115]. Studies on the relationship between intestinal cells and ISCs are based on the application of organoid technique and animal model. Thanks to these studies it was possible, at least partially, to elucidate the mechanisms responsible for maintaining intestinal homeostasis. The base of the crypt is an intrinsic pattern of crypt base columnar cells expressing Wnt target gene, leucine-rich repeat containing G protein-coupled receptor 5 (Lgr5+), in close contact with Paneth cells that are the source of secreted proteins, such as Wnt3, epidermal growth factor (EGF), and Notch ligand Delta-like (Dll) 4 and Dll1 that are crucial for stem-cell support [11,14,87,116]. Due to their proximity to Lgr5+, Paneth cells affect the function of ISCs activating the canonical Wnt/β-catenin signaling pathway by delivering Wnt3 that is bound by Frizzled receptors to improve the function of Lgr5 stem cells [101,116,117]. Studies conducted on the organoid model indicate the role of EGF in ISCs proliferation. Paneth cells and mesenchymal cells are the major source of EGF with EGFR expressed on ISCs. Lack of EGF in cytokine environment or blocking EGFR signaling pathway by EGFR Gefitinib inhibitor subdues ISCs proliferation. However, the proliferation process of ISCs is more than that regulated by Lrig1, expressed thereby cell population, which serves as a cell surface negative regulator of EGFR/ErbB pathway [12,116,117,118]. Notch signaling is dependent on cell-to-cell contact of membrane bound Notch ligands on one cell, and Notch receptors (Notch 1 and Notch 2) in neighboring ISCs. Dll4 and Dll1, produced by Paneth cells, maintain ISCs proliferative ability. Abrogation of both Notch ligands leads to the complete conversion of proliferative progenitors into post-mitotic goblet cells [79,119,120]The relationship between Paneth cells and ISCs was much harder to prove in vivo. Nevertheless, the conditional deletion of Sox9, a protein required for Paneth cells differentiation dependent on the Wnt signaling pathway, in the animal model, allowed to confirm the role of Paneth cells as a source of supplying signals for stem cells as well as proved that the cell-cell contact facilitated cell homeostasis. Subsequent reduction of Paneth cells in the crypts resulted in a withdrawal of ISCs, implying that their maintenance depends to a large extent on the presence of Paneth cells [94,121,122,123].Paneth cells are able to participate in the regeneration of the epithelial layer. Mature, terminally differentiated Paneth cells are featured by the presence of CD24, lysozyme, and MMP7. Upon epithelial injury, followed by an inflammation, when the quick rebuilding of barrier is required, Paneth cells are able to acquire stem cells characteristics through de-differentiation via Notch and SCF/c kit signaling, contributing to the epithelium reconstitution [12,116,124,125,126]. The plasticity of Paneth cells was proved in studies performed on irradiated mice. Intestinal Lyz+ cells collected from irradiated mice, seeded and maintained under in vitro conditions were able to enter the proliferative state followed by differentiation into villus epithelial cells. Moreover, these Paneth cells-derived epithelial cells did not express Paneth cells markers, such as MMP7 and Lyz+. Inflammation of the intestine wall often is accompanied by the loss of Paneth cells [125,127,128]. Studies conducted by Yu et al. suggest a high possibility that the diminished number of Paneth cells, noticed in various diseases, may be the consequence of Paneth cells de-differentiation. It is worth mentioning that, in addition to Paneth cells, other differentiated intestinal epithelial cells can obtain stem cell features upon specific conditions and participate in epithelial damage repair [125]. The Paneth cells participation in aging cannot be excluded. The aging in the GI tract is represented by decreasing balance between stem cell reserve and differentiation. Animal models, as well as organoid technique, clearly present that during physiological aging the reduction of crypt number along with the increase in the crypt length and width is observed. Additionally, the aging of ISCs is driven by mTORC1 and canonical Wnt signaling in ISCs is decreased. However, the expression of genes encoding Wnt3 or EGF in old Paneth cells is not significantly altered, so their influence on the ISCs is not achieved through the Wnt signaling pathway [129,130,131]. However, at the same time, a significant upregulation of Notum, the extracellular Wnt inhibitor responsible for negative-feedback loop formation, occurred in evaluated Paneth cells in aging mice. It shows that by producing Notum, aging Paneth cells can affect the regenerative capacity of stem cells by silencing the Wnt signaling pathway [132,133]. Paneth cells are a group of terminally differentiated epithelial cells, localized at the base of the crypts, and contain apical secretory granules filled with AMPs that are mandatory for controlling intestinal microbiota. Moreover, their localization and proximity of ISCs regulate the function of ISCs throughout paracrine-specific proteins secretion. In the pathological state, Paneth cells express plasticity and undergo de-differentiation into stem cells supplying the pool of ISCs. By shaping the host microbiota and controlling ISCs, they prevent the development of intestinal inflammation and maintain gut homeostasis.6. Cross-Talk between Microbiota and Epithelial Cells, a Step Required for the Maintenance of the Intestinal BarrierThe GI tract is inhabited by an enormous number of microorganisms, termed gut microbiota, and they form a mutualistic relationship with the host. First, the presence of microbiota in the GI tract is beneficial, but at the same time those microorganisms create a permanent threat to the host, therefore preventing their translocation into the underlying tissue is of paramount significance. Thus, the integrity of the GI tract barrier is vital for maintaining the health of the host. The critical role in barrier arrangement is played by epithelial cells, simultaneously participating in the segregation of intestine microbiota and possible pathogens by forming physical and chemical obstacles. At the same time, epithelial cells perform mediation by sensing the microorganisms and, as a result, secreting mediators, including cytokines and chemokines, that will stimulate or inhibit the immunocompetent cells located in lamina propria, leading to inflammation or immune tolerance induction. However, to form physical and chemical barriers, epithelial cells require continuous cross-talk with the GI tract microbiota. Any impairment in barrier functions can be associated with the uncontrollable immune reaction in the intestine and/or promotion of the unrestricted growth of microbiota, which leads to various diseases, including inflammatory and metabolic disorders, such as IBD, obesity, or even cancer [24,32,36,134].6.1. Sensing and Recognition of MicroorganismsAs it was discussed previously, a physical barrier is formed by mucus, secreted by goblet cells, with major participant MUC2, and by intercellular junctions. Mucus that covers the GI tract is a part of the innate mucosal barrier and prevents inflammation by limiting the antigen exposure as well as bacteria to the immune cells underlying the epithelial layer. It creates a diffusion barrier, through which small molecules such as ions, nutrients, gases, water can penetrate, and reach the epithelial cells. Moreover, it protects against mechanical, chemical, and biological attacks, also it works as a lubricator to facilitate the passage of cellular debris, bacteria, and fecal material, flushing them away through the intestinal channel [15,24,31,135,136,137]. The physical barrier can be increased by defense mechanisms induced as a result of microbial sensing and recognition by IECs that carry out PRRs, especially TLRs and NLRs, which recognize MAMPs or DAMPs released from the host cells. TLRs are expressed in most IECs lineages, including stem cells, enterocytes, goblet cells, enteroendocrine cells, Paneth cells, M cells, and they play an important role in the detection of microorganisms under pathologic as well as homeostatic conditions [3]. The distribution of TLRs varies and changes along the intestine, with small intestinal IECs expressing lower levels than colonocytes, which correlates with a much lower number of microorganisms than in the large intestine [96]. To maintain the GI tract homeostasis, IECs had to develop a tight TLR-regulating system to avoid disproportionate reaction towards gut microbiota. These cells can control and modulate TLR-signaling at different levels, such as by restricting access to the ligands or by inactivating downstream cascade.Enterocytes are structurally polarized with the apical surface facing the intestinal lumen, and the basolateral site connecting with the lamina propria. Enterocytes express TLR2, TLR3, TLR4, TLR5, and TLR9, with the majority of TLRs present at the basolateral membrane, while TLR2, TLR3, and TLR9 are also expressed at the apical surface [138,139,140,141]. As a consequence, TLRs activation can lead to different cellular responses, depending on the localization of the sensing receptors. Basolateral TLR stimulation results in a signaling cascade that leads to the NF-κB nuclear translocation, subsequently, expression and secretion of cytokines and chemokines, including TNF-α, IL-6, IL-12, IL-18, CXCL8, CCL20, which activate immunocompetent cells localized in lamina propria. Such activation and induction of inflammatory response are the results of breaching the epithelial barrier by pathogenic microorganisms [14]. This mechanism was explained in the elegant experiments presented by Gewirtz et al., indicating the requirement of crossing the epithelial barrier by flagellated bacteria to be recognized by TLR5. TLR5 recognizes flagellin, the structural component of bacterial flagella, and is critical for the detection of invasive flagellated bacteria at the mucosal surfaces. Only flagellated bacteria, that were able to cross the mucosal membrane, induced inflammation, while flagellin of luminal, commensal Escherichia coli strains did not evoke a comparable effect [142]. TLR9 is presented on the apical and basolateral surface of IECs and recognizes unmethylated CpG sequences expressed at high levels in prokaryotic DNA found within the commensal microbiome as well as within prokaryotic pathogens, invading the GI tract. Detection of luminal unmethylated CpG of microbiota via apically located TLR9 results in stabilization of IκB (inhibitor of NF-κB), and making IECs hyporesponsive to apical interaction to TLR9 ligands. However, once pathogenic bacteria break through the epithelial barrier and unmethylated CpG sequences are detected via basolateral TLR9, NF-κB activation occurs, subsequently leading to the proinflammatory cytokines and chemokines production [143]. The maintaining of the physical barrier relies on mucus production and secretion. Moreover, the density of mucus largely depends on the presence of bacteria. In MUC2−/− mice there is no inner mucus layer physically segregating microbiota and colonocytes, and as a result of their direct contact with microorganisms, the inflammatory response is provoked and spontaneous colitis is induced [36,37,144].The correlation between the sensing of microorganisms and mucus secretion was discussed earlier, in the context of senGCs that respond through TLR1/2, and TLR4 and TLR5 ligands leading to the increased MUC2 secretion, and enforcement of mucus barrier [46]. Probably disruption of this mechanism is the explanation of the altered microbiota composition and increased susceptibility to the development of spontaneous colitis observed in TLR5−/− mice. However, mice without TLR5 express mosaic phenotype, with a subset of mice that shows increased susceptibility to the spontaneous colitis development and lack of the normal, double structure of colon mucus layer. Only a disorganized mucus layer that is penetrable for microorganisms is present. Non-colitic TLR5−/− mice had normal, though slightly thinner mucus layer. Studies of another group indicated that deficiency of TLR5 altered the composition of the intestinal microbiota in comparison to the wild type mice, and low-grade inflammation and susceptibility to colitis were observed [145,146]. Furthermore, TLR1 is involved in mucus synthesis, since a large area of the colon with a patchy and significantly depleted mucus layer is observed in TLR1−/− mice, as a result of a defective production and/or secretion of MUC2 [147]. Additionally, increased intestinal inflammation and overgrowth of Candida albicans and E. coli in the colitis model were observed in mice with TLR1 deficiency [148]. Furthermore, commensal bacteria, upon interaction with TLRs, are responsible for the enforcement of the epithelial barrier. One of the species is Akkermansia muciniphila, the Gram-negative bacteria, inhabiting the human GI tract, producing mucin-degrading enzyme, and utilizing mucins as a source of energy. Within the outer membrane, it presents a protein Amuc_1100, interacting with TLR2, and thereby interplay is able to modulate the gut barrier and the intestinal permeability by increasing mucus thickness and TJs proteins, such as occludin, claudin 3, and cannabinoid receptor 1 [149,150]. These data clearly show that TLRs involvement is important in the development of a healthy mucus layer in the GI tract.6.2. Factors Influencing the Colonization of the GI Tract with MicrobiotaMicroorganisms are spatially organized along the length of the intestine and distributed in the lumen according to the oxygen levels and nutrients availability. The gradient of microorganisms increases from the proximal to the distal GI tract, and from the epithelial layer towards the lumen [24,28,151]. Moreover, the viscosity of the mucus increases toward the distal regions of the GI tract. In addition, since mucus, at a steady state, is constantly secreted by goblet cells, it creates a continuous flow towards the lumen that, together with AMPs and IgA present in mucus, keeps microorganisms apart from the epithelium. Mucus determines the distribution and organization of the microbiota in the intestine and protects against the bacterial colonization of epithelial cells and crypts. However, for proper arrangement of the intestinal mucus, the presence of microbiota is absolutely required. Studies performed on the GF mice and antibiotic-treated mice highlighted the importance of interaction between goblet cells and microorganisms. The early studies indicated the thinner layer or even local lack of mucus in the colon of the GF animals compared to the conventionally raised (Convr) rats. The mucus in the colon was penetrable to bacteria. Moreover, the replication of epithelial stem cells was disrupted and antibiotic-treated mice were more susceptible to colitis, induced physically or chemically [24,152,153]. Since comparison of the GF rodent model with Convr animals gave some insights to the formation of mucus protective layer, in nature we are facing a natural model that can be, at least initially, compared with GF animals. The colonization of the intestine with individual microbial populations in neonate animals can also serve as a model for analysis of the microbial influence on the development of mucus. Indeed, the increased expression of genes encoding Muc1, 3, 4, membrane-bound mucins was observed even in the absence of microbiota, between 1–6 postnatal days. However, the increased expression of the gene, encoding secreted MUC2, required the presence of microorganisms. Moreover, mice monocolonized with the probiotic bacteria (Lactobacillus acidophilus or E. coli Nissle 1917), exhibited a similar gene expression profile to the neonate GF mice, indicating that the complex microbial population is required to stimulate the Muc2 gene [154]. Moreover, bacterial products (LPS, peptidoglycans), can be involved in the restoration process of mucus secretion, since the GF mice treated with those compounds obtained a functional mucus protective layer, comparable to the Convr mice [155]. It can be said that the protective mucus layer in the small and large intestines depends on the presence of microorganisms and can be rebuilt since the impenetrable layer of mucus that served as a barrier against microbiota translocation occurred within 5 weeks in GF mice after colonization with a complex microbial community [156]. In the small intestine, where mucus is thinner and penetrable for microorganisms, or there are intended loopholes within the mucus layer, the microbiota is controlled and shaped due to the presence of AMPs, the major source of which are Paneth cells. This chemical barrier formed by the defensin family of proteins, cathelicidins, RegIII family, lysozyme, RELMβ, and IgA has a critical role in the segregation of intestinal bacteria and epithelial cells, suppressing colonization and overgrowth of the microorganisms. Using the enteroid culture technique Schoenborn et al., indicated a diminished number of Paneth cells in the GF mice in comparison to the Convr animals. Moreover, the alterations in RegIIIγ transcript levels in Paneth cells were significantly reduced in the GF small intestine crypts, relative to those in the Convr animals. Meanwhile, the enteric microbiota did not influence the number of ISCs [157]. Therefore, the presence of intestinal microorganisms affects the number of Paneth cells and hence, the integrity of the epithelial barrier, since Paneth cells regulate stem cells homeostasis. In addition, the dependence of defensin cryptdin 2 and RegIIIβ and RegIIIγ production was demonstrated by Vaishnava et al., indicating a reduction of AMPs in Myd88−/− mice [95].Furthermore, sIgA, the hallmark of mucosal immunity, are secreted in response to luminal antigens delivered to immunocompetent cells located in lamina propria. The major entrance gates for such antigens are areas devoid of mucus in the small intestine that are localized above Peyer’s patches and are characterized by the presence of M cells (Figure 1). M cells are not strict APCs in the context of DCs or macrophages, but they are antigen delivering cells that transfer luminal particles and antigens to the lamina propria DCs for antigen presentation, leading to the immune response or tolerance induction. Constant contact of M cells with microbiota results in B cells stimulation for IgA production, their transport to the intestine lumen using pIgR, and, as a result, control of microbiota by sIgA through the aggregation of bacteria and prevention of mucosal barrier crossing by size exclusion [3,18,158]. Glycoprotein 2 (GP2), a transcytolytic receptor present on M cells surface, is involved in the antigen uptake. Antigen-specific IgA response is suppressed in mice lacking GP2 [159,160]. It is worth noticing that the expression of pIgR, involved in IgA transcytosis, is MyD88-dependent since in vitro studies revealed upregulation of pIgR and increased transcytosis of IgA upon stimulation of epithelial cells with LPS or heat-killed E. coli, indicating TLR-mediated recognition dependence [161,162]. Although AMPs secreted by Paneth cells can be identified in the colon or even in feces in their native form, their expression level is not high compared with a small intestine, and it is not clear what compounds participate in the segregation of microorganisms in this area. The inner mucus layer is involved in this mechanism since at the steady state, this barrier is free of microbiota and not penetrable for microorganisms. However, the mechanisms that separate bacteria and colonocytes are not fully elucidated. Early events of colonization often depend on the flagella-mediated motility of bacteria. Ly6/Plaur domain-containing 8 (Lypd8) protein, highly expressed on colonocytes and constitutively secreted to the lumen is a novel molecule, contributing to the segregation of microorganisms in the colon. This molecule participates during the early-phase protection by preventing colonization of the large intestine by flagellated bacteria, such as Proteus mirabilis and E. coli via flagella binding, thereby suppressing colonic epithelium invasion [163]. The parallel mechanism, engaged thereby molecule, was recently elucidated by Okumura et al. using a Citrobacter rodentium model, frequently used for clarification the mechanisms of pathogenesis of human infections with enteropathogenic E. coli (EPEC) and enterohaemorrhagic E. coli (EHEC). The invasion of the intestinal mucosa with enteropathogenic bacteria routinely depends on the presence of virulence factors, among which, type III secretion system (T3SS) is involved, forming a link between bacterial cytoplasm and a target host cell cytoplasm. Following the maturation of the T3SS translocon, translocated intimin receptor (Tir) is exported by the bacteria and integrated into the host cell plasma membrane. Intimin, a bacterial adhesion molecule involved in an intimate attachment of enteropathogens, interacts with Tir, which plays a central role in actin condensation beneath the adherent bacterium, required for characteristic, pedestal-like structures formation. Those structures are known as the ‘attaching and effacing’ (A/E) lesions on the IECs and promote tighter binding of bacteria to epithelia [164,165]. Lypd8 shed by colonocytes, suppressed the attachment of C. rodentium to colonocytes by inhibition of the interaction between intimin and Tir. It is competitively bound to intimin, effectively blocking Tir-intimin interplay, and is required for the generation of A/E lesions. More rapid intestine colonization with C. rodentium and more severe colitis in Lypd8−/− mice were observed. Interestingly, human Lypd8 bound to EHEC intimin indicates that the ability of Lypd8 proteins to connect with intimin of A/E bacteria is conserved. These data emphasized the importance of the enterocytes in the protection of the integrity of the intestinal barrier [36,166]. In addition, a lectin-like protein ZG16 that specifically binds peptidoglycan of Gram-positive bacteria and thereby inhibits their penetration into the inner colonic layer is present in the colonic mucus [167].6.3. Mutual Dependency of IECs and Intestinal MicrobiotaMucus in the small and large intestines, engaging different, yet effective protective mechanisms, maintains segregation of microbiota from epithelial cells and thereby prevents inflammation. On the other hand, the presence of gut microbiota in the mucus is required, since microorganisms are the source of many metabolites, beneficial for the host physiology. At the same time, mucus is the source of the energy, and thereby it can shape the gut microbiota supporting the GI tract homeostasis. The composition of the gut microbiota can undergo changes from mucosal to the luminal site, forming different ecosystems. Their complexity can be affected by many factors, such as hygiene, diet (especially “Western diet” low in fiber and high in sugar and fat), oxygen concentration, mucus, microbial adherence, antimicrobial compounds, host stress, and immune response, leading, in many cases, to dysbiosis [24,28,31]. The GI tract microbiota degrades dietary substrates that are not used and absorbed in the small intestines and thus reach the colonic lumen. They are usually plant-derived polysaccharides, for which the host presents a rather limited enzymatic profile of approximately 17 carbohydrate-active enzymes (CAZymes) as opposed to, at least, human microbiota-encoded 89 CAZymes families, suggesting the ability to digest a huge range of carbohydrates [168]. Though gut microbiota feeds on dietary fibers non-digestible for the host and produces compounds that exert positive effects on the intestinal mucosa, the mucus layer is an alternative source of host-derived glycans. Mucus creates the selective niche, which serves as an attachment site for microorganisms, broadly described as “mucus-associated microorganisms”. Additionally, mucin glycans provide nutrients for microorganisms called “mucolytic bacteria”, supporting their growth and colonization as well as offering sources of carbon and energy. These bacteria digest glycan using exoglycosidases that allow removing one sugar residue per time, and when all glycans are removed, the protein core of the mucin is degraded [34,151]. This process leads to MUC2, and, finally, mucus degradation. Not all bacteria are able to remove glycan residues, therefore removed glycans can be used by bacteria and other members of gut microbiota that digest them. Among specific enzymes essential for mucin degradation are sialidases, fucosidases, sulfatases, proteases, belonging to the category of carbohydrate-active enzymes. Such metabolic plasticity is evident and beneficial for microbiota when complex dietary carbohydrates are missed in the diet. This ability of switching from dietary glycans to mucus glycans determines which gut microbiota can survive when supplementation of dietary fibers is reduced. However, such diet deprivation can have serious consequences for the host. It was proved in the mouse model that animals fed a fiber-free diet presented decreased thickness of mucus layer that increased their susceptibility towards infection with C. rodentium and that effect was similar to the MUC2−/− mouse model, where the colon inner mucus layer was thinner, the proximity of gut microbiota to the epithelial cells was smaller, and as a consequence, the development of spontaneous colitis was observed [31,37,144,169]. Hence, the strong correlation between diet-dependent loss of mucus layer resulting from the lack of MUC2, and protection against infection by enteric bacteria is based on the involvement of gut microbiota in forming a protective, impervious, and functional barrier.Mucus degrading bacteria present within mucus include Akkermansia muciniphila, Bacteroides thetaiotaomicron, Bifidobacterium bifidum, Bacteroides fragilis, Ruminococcus gnavus, and Ruminococcus torques. Those species generate short-chain fatty acids (SCFAs) through a fermentation process, using glycans as an energy source [24,28]. SCFAs are carboxylic acids, produced by anaerobic fermentation of dietary fibers in the intestine, of which acetate and propionate are produced by Bacteroidetes (Gram-negative bacteria) while butyrate is produced by Firmicutes (Gram-positive bacteria) in the human gut [169]. Those compounds are then absorbed and used by colonocytes to recover part of the energy spent for MUC2 synthesis. Furthermore, in the GF or antibiotic-treated mice fed a diet supplemented with a mix of SCFAs, the proliferative activity of small intestine IECs was restored. It indicates the role of the GI tract microbiota in maintaining homeostasis of the intestine [170]. However, this activating influence depends on the type of the cells and concentration of SCFAs, since butyrate had an inhibitory effect on ISCs. It seems that this is a kind of safety mechanism, protecting against uncontrollable cell divisions, as long as sodium butyrate is used in colorectal cancer cell lines, suppressed cancer cell line proliferation, and induced apoptotic cell death [171]. There is a close relationship between the maintenance of anaerobic conditions necessary for SCFAs-producing bacteria and the strengthening of the epithelial protective barrier. Butyrate is the main energy source of colonocytes, which consume more than 70% of oxygen due to butyrate oxidation and thereby, support maintaining anaerobic niche. Depletion of anaerobic bacteria, source of SCFAs (due to antibiotic treatment or fiber-low diet) induces colonocytes to undergo anaerobic respiration, which causes a release of oxygen and nitrates to the lumen. This mechanism results in promoting anaerobic environment and inducing overgrowth of facultative anaerobic pathogenic bacteria e.g., E. coli and Salmonella spp. [172,173,174]. Moreover, it is possible that butyrate can enforce the epithelial barrier by inducing gene expression encoding proteins of TJs, as in vitro studies revealed these effects [175]. Acetate produced by Bifidobacterium longum protects IECs against apoptosis induced by the O157 toxin. In addition, acetate induces goblet cells differentiation, mucin secretion, and their sialylation [153]. Dysbiosis, which is frequently observed in patients diagnosed with UC or CD, leads to a diminished number of bacteria producing SCFAs. Deficiency in those metabolites within the intestine results in the weakening of the protective barrier, the functionality of epithelial cells as well as mucus production. As a result, unwanted overgrowth of microorganisms occurs that can lead to an unrequired inflammatory response. In addition, other microbe-derived metabolites are involved in maintaining epithelial cells functionality. Lactate is a potent inducer of small intestine Lgr5+ cells, causing their hyperproliferation. Lgr5+ cells are supported by Paneth cells with various factors (Wnt3, Dll1, Dll4, EGF) required for maintaining proper cell functions. Using the organoid model it was revealed that Paneth cells support Lgr5+ cells functions by providing lactate to aid enhanced mitochondrial oxidative phosphorylation, which is required for establishing mature crypt phenotype through signaling [176]. However, even though the importance of this metabolite in maintaining crypt homeostasis was emphasized, it is still controversial whether lactate producing bacteria, e.g., Lactobacillus spp., affect epithelial stem cell homeostasis in vivo. Nevertheless, intestinal microbiota and microbiota-derived metabolites are important for maintaining the epithelial barrier integrity and homeostasis. Moreover, this cross-talk between epithelial cells and commensal microorganisms prevents the development of host dysfunctions.7. ConclusionsThe GI tract is a place that is constantly exposed to a multitude of stimuli, however, homeostasis of this area, and by that means homeostasis of the host, is maintained. In this review, we focused on the role of specific epithelial cell populations engaged in the formation of the first line of defense against mucosa colonization with pathogens as well as their mutualistic relationship with the GI tract microbiota. It summarizes the data of many investigators, working on all aspects concerning mechanisms involved in preventing the development of inflammation in the GI tract. IECs, due to their localization between the GI tract microbiota and immunocompetent cells found in lamina propria, play important role in transferring signals responsible for silencing immune response and inducing tolerance directed towards commensal microorganisms and food antigens. Maintaining the integrity of the barrier, based on the mucus secretion by goblet cells and enterocytes as well as formations of the intercellular junctions between epithelial cells, is crucial for the prevention of inflammatory response due to the spatial segregation of microbiota. However, as a major producer of AMPs, Paneth cells have a strong influence over microbial composition in the small intestine. Moreover, through their localization at the base of the crypts of Lieberkühn and proximity of ISCs, Paneth cells regulate the function and differentiation of crypt stem cells, subsequently affecting the physiology of newly formed epithelial cells. Nevertheless, cross-talk between the IECs and gut microbiota is required to maintain the homeostasis of the intestine. It is comprehensively demonstrated that members of gut microbiota are not the passive bystanders in the GI tract, but instead are active participants in establishing homeostasis in the GI tract. TLR interactions regulate the barrier formation, inducing/silencing immune response, controlling the permeability of the barrier, and stimulating the secretion of AMPs. The formed mucus layer serves not only as a barrier but also as an attachment site and the source of energy for microbiota, creating a microecosystem that influences the physiology of the host through SCFAs secretion by commensal bacteria. The GI tract microbiome is a very complex community that can be altered by many factors, including antibiotic treatment as well as a diet. Recently, the importance of a fiber-rich diet as a vital factor affecting the profile of the gut microbiome and the interplay of microorganisms with epithelial cells was emphasized. Dysbiosis results in disorder of SCFAs production, and can subsequently affect the integrity of the epithelial barrier. In this review, the mechanisms involved in the loss of the stability of the intestinal barrier, and dysbiosis, which are the reasons for unrequired inflammatory response that can promote the development of IBD in humans, were discussed. However, epithelial cells also coordinate the development and maturation of downstream innate and adaptive immune responses, induced in lamina propria and MLNs residing immune cells, resulting in local and/or systemic protective immunity against invading infectious agents.

animals : an open access journal from mdpi

10.3390/ani13091431

PMC10177277

No previous studies have been identified that have investigated the impact of neutering before or after known puberty on growth and physical development in a large number of bitches. This study was designed to examine data on the physical development, vulval size, and conformation of 306 bitches neutered before (n = 155) or after (n = 151) puberty. Data were gathered for bitches at six- and 17-months of age using bespoke physical assessment forms and digital photographs of the vulva. Bitches neutered before puberty had significantly greater changes in height and smaller changes in measurements of vulval length and width between six- and 17-months of age than those neutered after puberty. Although not significant, bitches neutered before puberty were taller and heavier with smaller vulval size measurements at 17-months of age. At 17-months of age, significantly more bitches neutered before puberty had vulvas that appeared juvenile and recessed at the physical assessment, and significantly more bitches neutered before puberty had vulvas that appeared ‘recessed/inverted’ on the examination of digital images. The results from this study could suggest that neutering before puberty may be a suitable option for large breed bitches. However, any longer-term health consequences of the differences in physical development seen need to be investigated and better understood before recommendations can be made.

No previous large prospective cohort studies have been identified that have investigated the impact of the surgical neutering of bitches before or after known puberty on their growth and physical development. This study was designed to examine the data on physical development, vulval size, and conformation for bitches neutered by ovariohysterectomy before puberty (PPN, n = 155) or after puberty (control, n = 151) using a prospective cohort study design. Data were gathered at six- and 17-months of age using bespoke physical assessment forms and digital images of the vulva. PPN bitches had greater changes in height measurements (mean difference = 2.039, SEM = 0.334, 91% CI = 1.471 to 2.608, p < 0.001) and smaller changes in the measurements of vulval length (mean difference = −0.377, SEM = 0.079, 91% CI = −0.511 to −0.243, p < 0.001) and width (mean difference = −0.221, SEM = 0.063, 91% CI = −0.328 to −0.113, p < 0.001) between six- and 17-months of age than for the control bitches. Although not significant, the PPN bitches were taller (mean 58.5 vs. 56.6 cm) and heavier (mean 28.3 vs. 27.3 kg) with smaller vulval size measurements (mean vulval length 2.8 vs. 3.2 cm, mean vulval width 1.7 vs. 2.1 cm) at 17-months of age. At 17-months of age, significantly more PPN bitches had vulvas that appeared juvenile (Yates’ Chi-square = 14.834, D.F. = 1, p < 0.001) and recessed (Yates’ Chi-square = 7.792, D.F. = 1, p = 0.005) at the physical assessment, and significantly more PPN bitches had vulvas that appeared ‘recessed/inverted’ on the examination of digital images (Chi-square = 9.902, D.F. = 1, p = 0.002). The results from this study suggest no contraindications to prepubertal ovariohysterectomy for large breed bitches. However, any longer-term health implications of these differences in physical development need to be investigated and better understood prior to recommendations being made.

1. IntroductionSexual maturity in dogs is reached as early as six to seven months of age [1,2,3,4]. While neutering prior to sexual maturity, or ‘early’ neutering, considered to be between 6 and 14 weeks of age [5], is sometimes recommended [6], the literature is inconsistent, making it challenging to decide on the best time to neuter female dogs. Despite many studies investigating the impact of neutering on health [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24], and to a lesser extent, behaviour [9,12,19,25,26,27,28,29,30,31,32], no cohort studies have been identified that investigated the impact of the surgical neutering of bitches before or after known puberty on their growth and physical development.In dogs, longitudinal skeletal growth is regulated by a combination of genetic, hormonal, environmental, and biomechanical factors [33,34,35]. Oestrogen is associated with inhibition of the growth hormone–insulin-like growth factor axis and closure of the physes at skeletal maturity (potentially by directly influencing growth plate chondrocytes) [34,36]. Early removal of the gonads has been associated with delayed closure of the radial and ulna physes, extended time to reach a growth plateau, and longer radial and ulna bone length in dogs and cats of both sexes [26,37,38,39,40]. The work that is commonly referenced in the dog [26] compared neutered versus entire bitches and bitches neutered at different ages, and identified effects on long bone growth and development of the external genitalia. However, the study included only small numbers and did not consider the timing of neutering in relation to puberty. In contrast, when radial lengths were compared between 11 mixed-breed bitches ovariohysterectomised at 10-weeks of age and 10 bitches that underwent sham surgery [41], no significant differences were identified, and the authors concluded that skeletal development was not impacted by prepubertal gonadectomy. However, final measurements in this study were taken at 24-weeks of age, at which time, their growth may not have reached a stable level.Oestrogen is also essential for the development of the external genitalia in female dogs [42,43,44]. Oestrogen production prior to the first oestrus causes development of the external reproductive tract and the increase in vulval size [2]; an effect that is maintained by subsequent oestrus cycles. Reduction in oestrogen concentration due to gonadectomy may fail to stimulate normal vulval development, particularly if ovary removal occurs before puberty [42,45]. Salmeri et al. [26] measured the dorsoventral length of the vulvar commissure and reported that subjectively, vulvas in bitches neutered at seven weeks (n = 7) or seven months of age (n = 4) were smaller than those of entire bitches (n = 6), although the numbers of bitches were small and statistical analysis was not performed. This raises the question of whether the length of exposure to ovarian hormones has an effect on vulval development.Suboptimal vulva conformation is described as a juvenile/infantile or recessed vulva and is associated with increased skin surrounding the vulva, perivulvar skin folds that can trap urine and bacteria leading to bacterial overgrowth, dermatitis and an increased incidence of urogenital disease such as urinary tract infection (UTI), cystitis, and chronic or juvenile vaginitis [4,45,46,47,48,49]. Verstegen-Onclin and Verstegen [42] presented data for 27 bitches with a history of recurrent UTI, vaginitis, and/or perivulvar dermatitis, and on examination, 85% of the bitches had a recessed or hypoplastic vulval appearance and 40% had a perivulvar skin fold. Twenty-five of the bitches were neutered; 21 before or around the time of puberty (mean age at neuter = 4.7 ± 0.3 months) and four after puberty (mean age at neuter = 2.4 ± 0.9 years). The authors proposed that neutering prior to puberty decreases the release of oestrogen and prevents the normal development of external genitalia. However, each of these cases was from a specialist referral hospital with the reason for referral unclear, therefore, the sample was potentially biased due to the study population comprising bitches with known urogenital disease. In contrast, Salmeri et al. [26] and Root-Kustritz [50] suggested that immature vulval development may not cause a clinical problem in otherwise healthy bitches.Studies on development including general physical characteristics such as height and weight and specific characteristics (e.g., vulval development and abnormalities), and the effect of puberty and surgical neutering on these characteristics in bitches are not well documented in the literature. Such information would be useful to assist with decision making on the optimum time for neutering bitches. Thus, the aim of this study was to compare the measurements of height, weight, body condition score, vulval size, and vulval appearance in bitches neutered either before or after puberty.2. Materials and Methods2.1. Study DesignUsing a prospective cohort study design, 306 bitches born in an assistance dog programme between 22 February 2012 and 9 August 2015 were neutered either prepubertally or post-pubertally as previously described [51]. Data were gathered from bitches at six- and 17-months of age to examine their physical development.2.2. Study SettingWithin the assistance dog programme, puppies are placed into volunteer homes with puppy raisers between seven and eight weeks of age. During the puppy raising stage, dogs are neutered before entering formal assistance dog training at approximately 14-months of age. Dogs in the programme are managed under similar conditions and all are fed the same commercially available extruded dry diet from weaning. The study bitches had a physical assessment performed by experienced veterinarians at one of four veterinary practices in the United Kingdom (identified as VP1, VP2, VP3, or VP4) at six-months of age and again at 17-months of age. Physical assessments for bitches that had been withdrawn and rehomed by 17-months of age were performed by a qualified member of the assistance dog organisation’s Health and Well-being team. All assessments were completed within two weeks of the bitches reaching (a) six-months and (b) 17-months of age. For bitches neutered before puberty, the six-month physical assessment was conducted the day prior to neutering. Additionally, the organisation’s electronic health records were examined for bitches with missing assessment forms and missing data on completed assessment forms.2.3. Study AnimalsBitches were allocated into two groups: neutered prepubertally (at six-months of age: PPN, n = 155) or post-pubertally (after their first oestrus: control, n = 151). Bitches were from five different Labrador/Golden Retriever cross breeds [51]. All 306 bitches were available for physical assessments at six-months of age. By 17-months of age, 56 bitches had been withdrawn from the assistance dog programme. The dog rehoming team were contacted to gather the data from these bitches, with the exception of those rehomed to other working dog organisations. The number of bitches for which data were collected for each assessment was reported.2.4. Variables and Data Sources2.4.1. Physical AssessmentsData collection forms in Microsoft Word (See Supplementary Materials S1) were emailed to the Dog Health and Well-being staff member responsible for the bitch when the bitch reached five- or 16-months of age. Forms were completed by the veterinarian or Health and Well-being Specialist assessing the bitch either electronically or in hard copy. Data collected at the six- and 17-month assessments included height at the withers (cm) measured using an equine measuring stick, weight (kg) measured on veterinary practice scales, body condition score (BCS) measured using a 9-point scale (where a score of ‘4’ or ‘5’ represented optimum body condition at six- and 17-months, respectively), vulval measurements (length and width, cm; indicated on Figure 1) measured using a ruler, and descriptive appearance of the vulva. The following six features were assessed and recorded as either ‘0’ [not observed] or ‘1’ [observed]: vulval swelling, perivulvar skin folds, perivulvar dermatitis, vulval discharge, recessed appearance, and juvenile appearance. These scores were later combined to give a cumulative vulva score out of six; bitches with a score of ‘0’ were deemed to have a normal vulval conformation.At the six-month assessment, the veterinarian also noted any relevant history and clinical signs including evidence of previous or current conditions such as vaginitis or endocrine disease (specific diseases were not specified on the form). Each bitch’s suitability to be neutered at six-months of age (PPN group only) was considered by the veterinarian based on current health and clinical history.For bitches with missing or incomplete forms for either assessment, electronic health records were examined to determine whether the physical assessment had been completed. Any available height, weight, or BCS data were extracted.The outcome variables examined at six- and 17-months of age were measurements of bitch height, weight and vulval size, the BCS, responses to questions relating to vulval appearance, and cumulative vulval score. Additionally, the change in measurements between six- and 17-months of age were examined. Confounding variables included breed and the VP performing the assessment as fixed factors, the age at assessment, and the number of days from neutering surgery (six- and 17-month assessments) or the number of days between assessments (for the change in measurements between repeated assessments) as covariates. Confounding variable data were extracted from the organisation’s electronic database or physical assessment forms.2.4.2. Digital Images of the VulvaA digital photograph of the vulva was taken at the time of the six- and 17-month assessments from a convenient (not fixed) distance (Figure 1). Following a period of training by a senior experienced veterinary reproduction specialist, all vulval images were examined by the same veterinary student who was blinded to bitch age and trial group. A vulval scoring system was devised for four aspects of vulval appearance: presence of vulval discharge, dorsal skin folds, recessed/inverted appearance, and perivulval changes were denoted using ‘0’ (absent) or ‘1’ (present) for each bitch. Vulval discharge (if present) was then categorised as: 1—mucoid, 2—mucoid purulent, 3—purulent, 4—haemorrhagic. Where present, the percentage cover of the vulva by the dorsal skin fold was estimated using 10% intervals (Figure 2). The nature of any perivulval changes were categorised as 1—skin discolouration, 2—hair discolouration, 3—hair and skin discolouration. If any image was of too poor a quality to make a confident assessment, it was recorded as non-diagnostic and excluded from analysis for the corresponding variable. Vulval images were used to describe appearance and assess development at six- and 17-months and were categorised as ‘normal’ or ‘recessed/inverted’ according to Figure 3.2.5. Bias and Study SizeConfounding factors were included in the statistical analysis for bitch height, weight, vulval length, and width. Height and weight can vary with breed, and the effect of different crosses and back crosses on the adult size of Labrador and Golden Retrievers cross bitches is unknown, therefore breed was included in the models. Similarly, age at assessment varied by up to two weeks, which can impact the height and weight. The VP that completed the physical assessment, or whether the assessment was completed by a Health and Well-being Specialist was included due to potential bias in measurement or differences in the calibration of weighing scales. The number of days between neutering surgery and the assessment was also included in the models investigating differences in height, weight, and vulval size at the 17-month assessments.Study size was determined by the available cohort of bitches that were born during the recruitment period and that were placed with puppy raisers within travelling distance to one of the four national VP for assessments and neutering. G*Power (http://www.gpower.hhu.de/ (accessed on 15 February 2023)) was used to determine the values of alpha for each analysis (see Supplementary Materials S2). The mean (±SEM) and confidence intervals (CI) were reported where appropriate.2.6. Quantitative Variables and Statistical MethodsData were checked manually for obvious recording errors. Any incorrect dates (identified by assessments appearing to be completed outside of the 2-week period allowed) or weights (unexpectedly low or high for the age) were checked in electronic health records and were corrected where errors were confirmed. Vulval length and width measurements were checked manually, and any cases where length < width were crosschecked with the bitches’ digital images, and corrected if they had been inadvertently transposed (n = 4). For statistical analysis, categorical covariates that had less than five bitches were grouped, and one breed group was created that contained all second-generation backcross bitches.Change in measurements in height at the wither (cm), weight (kg), vulval length (cm), and vulval width (cm) between the six- and 17-month assessments were calculated. For bitches with a calculated change in height that was zero or negative (representing a bitch not growing between six- and 17-months of age, n = 19), the data were checked by examining the completed assessment forms to ensure that recording errors were not present. Data errors were excluded from all height analyses and from analyses for other dependent variables where height was included as a covariate. The remaining data were used to examine differences between PPN and the control bitches at six- and 17-months of age, and in the change in measurements between 6-and 17-months of age using univariate general linear models. The dependent factors were bitch height (cm), weight (kg), vulval length (cm), and vulval width (cm) for the six- and 17-month assessments, and were the calculated values for change in height (cm), change in weight (kg), change in vulval length (cm), and change in vulval width (cm). For all models, trial group, breed, and VP were included as fixed factors. For models examining the change in measurements between six- and 17-months of age, VP had five categories: VP1–4 for bitches with physical assessments completed by the same VP at each age, and ‘Different’ for bitches that had assessments completed by different VPs at each age. For the six- and 17-month models, age in days, and for the 17-month models, the number of days between neutering surgery and the assessment, were included as covariates. For models examining the change in measurements between six- and 17-months of age, the number of days between measurements was included as a covariate. Height at the wither was included as a covariate in the general linear models for weight to account for any variation in weight attributable to bitches being taller. Height and weight were included as an interaction term in the general linear models used to examine vulval length and width measurements to control for any variation attributable to bitches being taller and heavier. Standardised residuals were saved and checked for normality (assumptions were met for all general linear models). Pairwise comparisons within the models were used to describe significant differences for the fixed effect variables. General linear models were performed using IBM SPSS Statistics for Windows, version 22 (IBM Corp., Armonk, NY, USA).A binary logistic regression model using stepwise backward elimination was used to examine the impact of the trial group on BCS while controlling for confounding variables of breed, VP, and body weight. For the six-month assessments, the BCS was grouped as ideal (BCS = 4) and overweight (BCS greater than 4). For the 17-month assessments, the BCS was grouped as ideal (BCS = 5) or under/overweight (BCS less or greater than 5).Data for visual appearance of the vulva from the assessment forms and digital vulval images and cumulative vulval scores were examined using Chi-square tests. Chi-square with Yates’ continuity correction was used for the analysis of 2 × 2 contingency tables where frequencies were <5. Where frequencies were too small for Chi-square analysis, data were grouped and reported. Sequential Bonferroni correction was applied where multiple testing was conducted to minimise the risk of Type 1 error. Chi-square analyses and binary logistic regression models were conducted using XLStat2016 (Addinsoft, New York, NY, USA).3. Results3.1. ParticipantsThree hundred and three bitches (152 PPN, 151 control) had completed six-month physical assessment forms. Three PPN bitches did not have completed six-month assessment forms returned; from the examination of electronic health records, all three were neutered at six-months of age at a VP and had a physical assessment noted as completed. All physical assessments at six-months of age were completed by one of the four VPs. All bitches were entire at the time of the six-month assessments. Two hundred and seventy-eight bitches (140 PPN, 138 control) had completed 17-month assessment forms. Twenty-four of these were completed outside of the two-week window and one bitch was still entire at 17 months of age; these were excluded. A total of 253 bitches remained for analysis at the 17-month time point (125 PPN, 128 control). Of these 253 bitches, 251 (123 PPN, 128 control) also had completed six-month physical assessment forms. At the time of the 17-month assessment, 36 of the 253 bitches were withdrawn from the assistance dog programme and 19 bitches had assessments completed by either a Health and Well-being Specialist or a veterinary practice other than one of the four VPs; 234 were completed by one of the four VPs. Nineteen bitches were excluded from all analysis that included height data due to potential errors represented by measurements, suggesting a lack of growth between six- and 17-months of age.3.2. Height3.2.1. Six-MonthData were available for 282 bitches (146 PPN, 136 control). Bitches measured between 42.0 and 60.0 cm in height (mean 52.2 cm ± 0.2 cm). There was no significant difference in height between the PPN and control bitches at six-months of age (F = 1.941, D.F. = 1, p = 0.165; mean height of PPN bitches 52.0 ± 0.2 cm and control bitches 52.3 ± 0.2 cm, Figure 4). Age at assessment (F = 7.396, D.F. = 1, p = 0.007) and VP (F = 4.653, D.F. = 3, p = 0.003) were significantly associated with bitch height. A one day increase in age was associated with a 0.09 unit increase in height (β = 0.086, SEM = 0.032, 92% CI = 0.030 to 0.142). VP3 had bitches with significantly lower measured heights than VP1 (mean difference = −0.806, SEM = 0.402, 92% CI = −1.512 to −0.100, p = 0.046) and VP4 (mean difference = −1.501, SEM = 0.409, 92% CI = −2.220 to −0.782, p < 0.001).3.2.2. 17-MonthData were available for 229 bitches (119 PPN, 110 control). Bitches measured between 51.0 and 66.0 cm in height (mean 57.6 ± 0.2 cm). There was no significant difference between the PPN and control bitches in height at the 17-month assessment (F = 1.067, D.F. = 1, p = 0.303; mean height of PPN bitches 58.5 ± 0.2 cm and control bitches 56.6 ± 0.3 cm; Figure 4). VP was the only factor significantly associated with the height measurement (F = 7.872, D.F. = 4, p < 0.001). VP3 had bitches with significantly lower measured heights than all other VPs. Mean differences between VP3 and VP1 was −1.539 (SEM = 0.460, 93% CI = −2.376 to −0.702, p < 0.001); VP2 was −1.666 (SEM = 0.497, 93% CI = −2.571 to −0.762, p < 0.001); VP4 was –2.311 (SEM = 0.475, 93% CI = −3.175 to −1.447, p < 0.001); and ‘Other’ VP was −2.480 (SEM = 0.667, 93% CI = −3.694 to −1.266, p < 0.001).3.2.3. Change in Height between Six- and 17-Month Physical AssessmentsData were available for 225 bitches (115 PPN, 110 control): the mean change in height measurement for PPN bitches was 6.53 ± 0.22 cm and for the control bitches was 4.36 ± 0.25 cm (F = 37.329, D.F. = 1, p < 0.001; Figure 5). PPN bitches had a significantly greater change in height measurements than the controls (mean difference = 2.039, SEM = 0.334, 91% CI = 1.471 to 2.608). The number of days between assessments (F = 5.816, D.F. = 1, p = 0.017) and VP (F = 2.258, D.F. = 4, p = 0.064) also impacted the change in height measurements. A one day increase in days between assessments was associated with a 0.067 unit increase in the change in height measurement (β = 0.067, SEM = 0.028, 91% CI = 0.020 to 0.115). VP3 had a smaller change in height measurements than for all the other VPs. These were significantly smaller than VP2 (mean difference = −1.002, SEM = 0.509, 91% CI = −1.869 to −0.135, p = 0.050), VP4 (mean difference = −0.985, SEM = 0.484, 91% CI = −1.810 to −0.160, p = 0.043), and for bitches that were measured by ‘Different’ VPs at six- and 17-months of age (mean difference = −1.569, SEM = 0.604, 91% CI = −2.597 to −0.540, p = 0.010).3.3. Body Weight3.3.1. Six-MonthData were available for 300 bitches. Bitches weighed between 15.0 and 27.2 kg (mean 21.0 ± 0.2 kg). Data were analysed for 279 bitches (145 PPN, 134 control); two PPN bitches that did not have height data for inclusion as a covariate and 19 bitches with potential errors with height measurement were excluded. There was no significant difference in weight between the PPN and control bitches at six-months of age (F = 0.952, D.F. = 1, p = 0.330; mean weight of PPN bitches 21.0 ± 0.2 kg and control bitches 21.0 ± 0.2 kg; Figure 6). Age at assessment (F = 17.221, D.F. = 1, p < 0.001), height (F = 52.834, D.F. = 1, p < 0.001), and VP (F = 2.688, D.F. = 3, p = 0.047) significantly impacted the weight measurements. A one day increase in age was associated with a 0.11 unit increase in weight (β = 0.106, SEM = 0.026, 90% CI = 0.064 to 0.148). A one unit increase in height was associated with a 0.35 unit increase in weight (β = 0.349, SEM = 0.048, 90% CI = 0.270 to 0.429). VP2 had bitches that were significantly heavier at six-months of age than all other VPs (compared to: VP1 mean difference = 0.795, SEM = 0.387, 90% CI = 0.155 to 1.434, p = 0.041, VP3 mean difference = 0.808, SEM = 0.352, 90% CI = 0.227 to 1.388, p = 0.022, VP4 mean difference = 1.055, SEM = 0.392, 90% CI = 0.407 to 1.702, p = 0.008).3.3.2. 17-MonthAll 253 bitches with completed 17-month assessment forms had their body weight data recorded. Bitches weighed between 21.7 and 33.5 kg (mean 27.80 ± 0.1 kg). Five bitches with no height measurement and 19 bitches with potentially incorrect height measurements were excluded from analysis. Data were analysed for 229 bitches (119 PPN, 110 control). There was no significant difference in weight between the PPN and control bitches at 17-months of age (F = 1.289, D.F. = 1, p = 0.258; mean weight of PPN bitches 28.3 ± 0.2 kg and control bitches 27.3 ± 0.2 kg; Figure 6). Height (F = 39.050, D.F. = 1, p < 0.001) and VP (F = 2.747, D.F. = 4, p = 0.029) were significantly associated with body weight at 17-months of age. A one unit increase in height was associated with a 0.35 unit increase in weight (β = 0.345, SEM = 0.055, 91% CI = 0.251 to 0.440). Body weight measurements by ‘Other’ VP were significantly lower than the measurements by all four VPs. Mean differences between ‘Other’ and: VP1 were −1.758 (SEM = 0.566, 91% CI = −2.721 to −0.795, p = 0.002); VP2 were −1.328 (SEM = 0.579, 91% CI = −2.314 to −0.342, p = 0.023); VP3 were –1.733 (SEM = 0.562, 91% CI = −2.691 to −0.775, p = 0.002); VP4 were −1.399 (SEM = 0.571, 91% CI = −2.370 to −0.427, p < 0.015).3.3.3. Change in Body Weight between Six- and 17-Month Physical AssessmentsData were available for 248 bitches (122 PPN, 126 control): the mean change in weight for the PPN bitches was 7.2 ± 0.2 kg and for the control bitches was 6.2 ± 0.2 kg. Bitches with potentially incorrect (n = 19) or missing height measurements (n = 7) were excluded from the analysis. Changes in weight data were examined for 114 PPN (mean change = 7.2 ± 0.2 kg) and 108 control bitches (mean change = 6.3 ± 0.2 kg), and the trial group had no impact (F = 2.450, D.F. = 1, p = 0.119). Breed was not included in the general linear model due to poor model fit.The number of days between assessments (F = 9.126, D.F. = 1, p = 0.003), change in height (F = 8.262, D.F. = 1, p = 0.04), and VP (F = 2.640, D.F. = 4, p = 0.035) significantly impacted the change in body weight. A one day increase in days between assessments was associated with a 0.077 unit increase in change in weight (β = 0.077, SEM = 0.025, 91% CI = 0.034 to 0.120). A 1-cm increase in change in height was associated with a 0.177 unit increase in the change in weight (β = 0.177, SEM = 0.062, 91% CI = 0.072 to 0.282). VP2 had a smaller change in weight measurements than for all the other VPs, and these were significantly smaller than VP1 (mean difference = −1.350, SEM = 0.479, 91% CI = −2.166 to −0.534, p = 0.005), VP3 (mean difference = −1.078, SEM = 0.454, 91% CI = −1.851 to −0.304, p = 0.019), and VP4 (mean difference = −1.228, SEM = 0.492, 91% CI = −2.067 to −0.390, p = 0.013).3.4. Body Condition Score3.4.1. Six-MonthTwo control bitches did not have a BCS reported at their six-month physical assessment. For the remaining 301 bitches (152 PPN, 149 control), the BCS ranged from 3 to 6 (median = 4; Table 1).Examination of the BCS data used a binary logistic regression including the data for 297 bitches; two bitches (1 PPN, 1 control) without weight data for inclusion as a covariate, and the two bitches that were BCS 3 (1 PPN, 1 control) were excluded. The model was significant (D.F. = 4, Chi-square = 134.135, p < 0.001). Body weight (OR = 1.453, 96% CI = 1.229 to 1.718, p < 0.001) and VP influenced whether BCS were ideal (BCS 4) or overweight (BCS > 4) and were retained in the final model. The BCS reported by VP2 were significantly more likely to be reported as higher than ideal than for all the other VPs. The BCS were also significantly more likely to be reported as higher than ideal for VP1 than VP3 and VP4 (see Table S1). The AUC was 0.873. The equation for the best fit model was:Pred(BCS)=1/(1+exp(−(−9.045+0.374×Weight+3.167×VP2−0.892×VP3−1.276×VP4))) where BCS = body condition score, VP = veterinary practice (VP1 being the reference practice).3.4.2. 17-MonthOne control bitch did not have a BCS reported at the 17-month physical assessment. For the remaining 252 bitches, BCS at 17-months of age ranged from 4 to 6 (median = 5; Table 2).Examination of the BCS data used a binary logistic regression including the data for 252 bitches. The model was significant (D.F. = 5, Chi-square = 30.113, p < 0.001). Body weight (OR = 0.807, 96% CI = 0.680 to 0.957) and VP influenced whether the BCS were ideal (BCS 5) or under/overweight (BCS 4 or 6) and were retained in the final model. The BCS reported by VP2 were significantly less likely to be ideal (BCS5) than those for VP1, VP3, and VP4. BCS reported by VP4 were significantly more likely to be ideal than those for VP1, VP2, and ‘Other’ (see Table S1). The AUC was 0.771. The equation for the best fit model was:Pred(BCS)=1/(1+exp(−(7.685−0.215×Weight−1.030×VP2+0.475×VP3+2.346×VP4−0.773−Other))) where BCS = body condition score, VP = veterinary practice (VP1 being the reference practice).3.5. Vulval Size3.5.1. Six-MonthVulval measurements were not reported for one PPN bitch. Six bitches (1 PPN, 5 control) were reported to have a ‘swollen’ vulva appearance at their six-month assessment and were excluded from analysis for vulval size. For the remaining 296 bitches (150 PPN, 146 control), the vulval length ranged from 1.5 to 4.2 cm (mean 2.8 ± 0.03 cm, Figure 7A) and vulval width ranged from 0.8 to 3.0 cm (mean 1.8 ± 0.02 cm Figure 7B). Bitches with no measurements for height (n = 1) and weight (n = 3) as well as bitches with potential errors in height measurements (n = 19) were excluded from the univariate models.For the remaining 273 bitches (144 PPN, 129 control) mean vulval lengths were 2.8 ± 0.04 cm for the PPN and control bitches and the mean vulval widths were 1.8 ± 0.03 cm for both groups. In the model controlling for the height * weight interaction, there was no significant effect of trial group on the vulval length (F = 0.394, D.F. = 1, p = 0.531) or width (F = 2.769, D.F. = 1, p = 0.097) measurements. Vulval length was affected by breed (F = 3.808, D.F. = 2, p = 0.023), with backcross bitches having significantly smaller vulval lengths than Labrador cross Golden Retriever bitches (mean difference = −0.261, SEM = 0.097, 94% CI = −0.488 to −0.035, p = 0.022). The vulval width was affected by age at assessment (F = 7.794, D.F. = 1, p = 0.006). A one day increase in age was associated with a 0.01 unit increase in vulval width (β = 0.012, SEM 0.004, 94% CI = 0.004 to 0.020).3.5.2. 17-MonthVulval measurements were not reported for six bitches (3 PPN,3 control) at their 17-month physical assessment. For the remaining 247 bitches (122 PPN, 125 control), the vulval length ranged from 1.0 to 4.4 cm (mean 3.0 ± 0.03 cm, Figure 7A) and the vulval width ranged from 0.6 to 3.5 cm (mean 1.9 ± 0.03 cm, Figure 7B). Bitches with no measurements for height (n = 1) and with potential errors in height measurements (n = 19) were excluded from the univariate models.For the remaining 227 bitches (118 PPN, 109 control), the mean vulval lengths were 2.8 ± 0.05 cm for PPN and 3.2 ± 0.04 cm for the control bitches and the mean vulval widths were 1.7 ± 0.03 cm for PPN and 2.1 ± 0.03 cm for the control bitches.In the model controlling for the height * weight interaction, there was no significant effect of trial group on the vulval length (F = 0.011, D.F. = 1, p = 0.916) or width (F = 0.175, D.F. = 1, p = 0.676) measurements. Vulval length (F = 5.216, D.F. = 4, p < 0.001) and width (F = 8.271, D.F. = 4, p < 0.001) were affected by VP. Measurements of vulval length were larger from VP1 (mean difference = 0.291, SEM = 0.093, 92% CI = 0.041 to 0.541, p = 0.021) and VP2 (mean difference = 0.398, SEM = 0.099, 92% CI = 0.132 to 0.664, p < 0.001) than for VP4. Measurements of the vulval length were also larger from VP2 than VP3 (mean difference = 0.255, SEM = 0.092, 92% CI = 0.010 to 0.500, p = 0.058). Measurements of the vulval width were larger from VP2 than VP1 (mean difference = 0.338, SEM = 0.069, 92% CI = 0.154 to 0.522, p < 0.001), VP4 (mean difference = 0.268, SEM = 0.070, 92% CI = 0.081 to 0.456, p = 0.002), and ‘Other’ VP (mean difference = 0.289, SEM = 0.093, 92% CI = 0.040 to 0.537, p = 0.021). Measurements from VP3 were larger than from VP1 (mean difference = 0.232, SEM = 0.060, 92% CI = 0.071 to 0.393, p = 0.001).3.5.3. Change in Vulval Size between Six- and 17-Month Physical AssessmentsTwo hundred and forty-four bitches (119 PPN, 125 control) had vulval length and width measurement data available at six- and 17-months of age. The mean change in vulval length and width for the PPN bitches was 0.08 ± 0.05 cm and −0.05 ± 0.04 cm, respectively. The mean change in vulval length and width for the control bitches was 0.39 ± 0.05 cm and 0.19 ± 0.04 cm, respectively (Table 3). Significantly more PPN than control bitch vulvas changed to be smaller and fewer changed to be larger in length (Chi-square = 22.334, D.F. = 2, p < 0.001) and width (Chi-square = 20.131, D.F. = 2, p < 0.001) between six- and 17-months of age (Figure 8).Bitches with potentially incorrect or missing height (n = 19) or weight (n = 5) measurements were excluded from the univariate models. Changes in vulval length and width data were examined for 113 PPN (mean change vulval length = 0.08 ± 0.06 cm, mean change vulval width = −0.05 ± 0.04 cm) and 107 control bitches (mean change vulval length = 0.41 ± 0.05 cm, mean change vulval width = 0.20 ± 0.04 cm). In the model controlling for a height * weight interaction, the mean change in vulval length (F = 22.901, D.F. = 1, p < 0.001) and width (F = 12.239, D.F. = 1, p < 0.001) was significantly affected by the trial group (Table 4). Changes in both measurements were also affected by VP (vulval length F = 8.005, D.F = 4, p < 0.001; vulval width F = 6.932, D.F. = 4, p < 0.001; Table 4, Figure 9).3.6. Vulval Appearance3.6.1. Vulval Appearance at the Physical AssessmentsData on vulval appearance were available for 303 (152 PPN, 151 control) bitches at the six-month stage. The number of bitches with each vulval anomaly was not significantly different between the PPN and control groups following sequential Bonferroni correction (Table 5). Proportionally, more control bitches were reported to have a swollen vulva, vaginal discharge, a recessed or inverted vulva, prominent perivulval skin folds, and perivulval dermatitis. There was no difference in the number of PPN and control bitches with a cumulative vulva score of 0 (106 PPN, 89 control), 1 (28 PPN, 36 control), or 2 and 3 (18 PPN, 26 control) at the six-month assessment (Chi-square = 3.933, D.F. = 2, p = 0.140). No bitch had more than three vulval anomalies noted.Data on vulval appearance were available for 253 bitches (125 PPN, 128 control) at the 17-month stage. Following sequential Bonferroni correction, significantly more PPN than control bitches were reported to have vulvas that appeared juvenile (Yates’ Chi-square = 14.834, D.F. = 1, p < 0.001) and recessed (Yates’ Chi-square = 7.792, D.F. = 1, p = 0.005; Table 5). No bitches from either trial group had swollen vulvas, and there was no significant difference in the number of bitches reported to have discharge or perivulval folds. Analysis for perivulval dermatitis was not possible due to the small numbers of bitches affected. Significantly more PPN than control bitches had a cumulative vulva score of 1 (18 PPN, 7 control), or 2 and 3 (13 PPN, 3 control) at the 17-month assessment and fewer had a score of 0 (94 PPN, 118 control) (Chi-square = 13.773, D.F. = 2, p = 0.001). No bitch had more than three vulval anomalies noted.3.6.2. Vulval Appearance from Examination of Digital ImagesDigital vulval images were captured for 274 bitches (134 PPN, 140 control) at six-months and for 270 bitches (137 PPN, 133 control) at 17-months, although some images were unusable due to poor image quality. The numbers examined for each anomaly are shown in Table 6. At six-months of age, there were no significant differences between the PPN and control bitches in the number of images showing vulval discharge, each category of vulva discharge (excluding ‘haemorrhagic’ due to small n), estimated % dorsal fold coverage grouped as less than 20%, 30%, 40% or greater than 50%, perivulval skin changes, the nature of perivulval skin changes (excluding ‘hair changes only’ due to small n), or ‘recessed/inverted’ appearance (Table 6). Statistical analysis for the presence/absence of dorsal skin folds was not possible due to small numbers of bitches (1 PPN, 0 control) with no dorsal fold present.Seventy-two PPN (82.8%) and 71 control (78.0%) bitches had vulvas that were classified as abnormal based on being recessed/inverted at six-months of age. At 17-months of age, significantly more PPN (71.1%, n = 54) than control bitches (46.3%, n = 38) had vulvas that were ‘recessed/inverted’ in appearance (Chi-square = 9.902, D.F. = 1, p = 0.002; Figure 10). Ninety-six bitches (46 PPN, 50 Control) had vulvas at both the six- and 17-month assessments that were classified as being ‘recessed/inverted’ or ‘normal’. Ten PPN and 10 control vulvas were ‘normal’ at six months. Of these, three PPN and eight control vulvas remained ‘normal’ at 17-months of age (Yates’ Chi-square = 3.232, D.F. = 1, p = 0.072). Thirty-six PPN and 40 control bitch vulvas were ‘recessed/inverted’ at six-months of age. Of these, 10 PPN and 17 control changed to ‘normal’ by 17-months of age (Chi-square = 1.793, D.F. = 1, p = 0.181). These differences were not statistically significant.There was a significant difference in the numbers of PPN and control bitches that had skin discolouration and skin and hair discolouration noted on digital images at 17-months of age (Chi-square = 5.563, D.F. = 1, p = 0.018). There were no significant differences in the number of PPN and control bitches that had vulval discharge, each category of vulva discharge (excluding ‘haemorrhagic’ due to small n), estimated % dorsal fold coverage grouped as less than 20%, 30%, 40%, or greater than 50%, or perivulval skin changes (Table 6).4. DiscussionThe impact on the physical development of neutering bitches before or after known puberty has not been well-studied. Physical development is an important consideration when deciding when to neuter due to potential consequences on future disease risk. This study presents the first prospective cohort study investigating the impact of neutering before or after puberty on physical development in a large number of bitches. In isolation, the results do not identify any significant contraindications to neutering large-breed bitches prepubertally. However, it is advised that any future consequences for health based on the differences presented here are considered prior to recommendations being made.Bitches neutered before puberty had significantly greater changes in height and vulval size between six- and 17-months of age than those neutered post-pubertally. Although not significant, bitches neutered prepubertally were also taller and heavier, with vulvas that were smaller at 17 months of age compared to bitches neutered post-pubertally. Interestingly, a dorsal vulval skin fold was apparent in almost all vulval images from bitches at both ages, and percentage cover of the skin fold was reported, which to the authors knowledge is the first time that this feature has been documented in such a large study population. Similar to findings relating to peri and postoperative outcomes [51], veterinary practice was one of the biggest and most consistent influencing factors for height, weight, BCS, and vulva size measurements.The present study included only female Labrador and Golden Retriever crossbreeds and the findings may not be applicable to other breeds of dog. Similarly, no entire bitches were included for comparison, and therefore no conclusions can be made about whether the physical development observed in bitches neutered pre/post-pubertally differs from that of entire bitches. However, previous studies reporting vulval size [52] suggest that vulval size in bitches neutered after puberty is similar to entire bitches, indicating similar maturation of the external genitalia. Whether vulval size in post-pubertally neutered bitches would subsequently decrease with age is unknown; no later life vulva measurements were obtained. The accuracy of vulval size measurements could have been improved if standardised callipers had been provided to the veterinary practices. Such tools have been used by others [53] and are recommended for future studies.In the present study, height was measured once, at the time of physical examination, and was subject to measurement variability between the observers and potential inaccuracies. For example, 19 bitches were recorded as having zero or negative change in height between six- and 17-months, and these were subsequently excluded from the analysis. Despite the potential errors, height data were reported due to the importance of considering growth differences between dogs neutered before or after puberty, and the results indicate an impact on growth in a direction that was expected: greater change in height for bitches neutered before puberty with lower levels of oestrogen. Other studies [26,37,38,39,41] have used radiographs to measure the long bone length and identify age at growth plate closure, which would likely provide more accurate ‘height’ data. However, this was not possible within the present study due to ethical constraints.Two methods of determining vulval appearance were used, with assessments made by veterinarians when examining the bitches, and from digital images of the vulva. While in-person assessment by an experienced veterinarian is useful, the ability to use the digital images enabled one person trained in assessment and blinded to the study group to examine all images, significantly reducing the risk of bias.There were also effects of VP in almost all analyses. The effects of VP on height were relatively consistent, but were less consistent for the vulva measurements. This highlights the importance of limiting the individuals making measurements, and otherwise controlling for this confounder in analysis. While this does not compromise the findings related to the trial group as VP was included as a covariate in the models, the reasons for the differences warrant further consideration, especially when using multiple veterinary practices to manage a large population of dogs or for future research.Many authors seem to suggest an impact of early neutering on bitch growth [12,16,18,40,50,54,55,56], and the assertion that this is due to an absence of oestrogen makes sense due to the association of oestrogen with the inhibition of the growth hormone–insulin-like growth factor axis and closure of the physes at skeletal maturity [34,36]. However, most of these are review papers and they reference the one study in dogs by Salmeri et al. [26], which included dogs neutered at seven-weeks of age (seven male, seven female), seven-months of age (four male, four female), and entire dogs (four male, six female) as well as the literature relating to cats [37,39]. Of the two studies that examined long bone length in surgically neutered bitches, both reported increased bone length following early neutering, but the differences were only significant in one of the studies for bitches neutered at seven-weeks of age [26]. For the other, no significant differences were identified, and the authors suggested that there were no skeletal developmental implications of prepubertal gonadectomy [41]. Both of these studies included small numbers of bitches with no consideration of power analysis or adjustment of alpha for the sample size. Only Salmeri et al. [26] considered bitches neutered at different ages; the study by Sontas and Ekici [41] included bitches that were all neutered or underwent sham surgeries at 10 weeks of age and measured the final radial length at six-months of age. Therefore, there are issues with the study design and methodology that should be considered when interpreting these findings.Our study showed that bitches neutered before puberty were taller (mean 58.5 vs. 56.6 cm) and heavier (mean 28.3 vs. 27.3 kg) by 17-months of age than bitches neutered after puberty, although the differences were not significant. However, the change in height between timepoints did differ significantly by trial group (PPN = 6.5 cm, control = 4.4 cm). Bitches neutered prior to puberty may have an extended growth period, reduced physeal closure, and consequently longer bones and increased height, in agreement with Salmeri et al. [26]. However, Salmeri et al. [26] failed to include whether the bitches had undergone puberty, making direct comparison to the present study difficult. Radiography and measurement of long bone length was not undertaken in the present study. A more accurate assessment of growth and the determination of age at growth plate closure would have been possible from radiographs of the long bones, as described by Salmeri et al. [26]. However, by 17-months of age physeal growth plates would be expected to be closed in all but giant dog breeds [3,57,58]. Therefore, our measurements at 17-months, in contrast to the findings of Sontas and Ekici [41], are likely to be representative of adult height.The lack of significant findings when comparing height measurements at 17-months for bitches neutered before and after puberty in the present study may be explained by the timing of neutering. The major growth in dogs occurs between three- and six-months of age [58], therefore bitches in both groups were neutered after this period. It is possible that growth plate closure was delayed in bitches neutered before puberty and that this influenced the greater change in height measurements, but that neutering at six-months of age, after the period of major growth had ended, prevented greater differences in height being observed. Salmeri et al. [26] neutered bitches at seven-weeks and seven-months of age, and therefore compared bitches neutered before and after the major growth phase, which could have caused greater differences in the length of the long bones. Indeed, Salmeri et al. [26] only reported significantly greater radial lengths in bitches neutered at seven-weeks (18.5 cm) compared to bitches neutered at seven-months of age and entire bitches (both 16.6 cm). However, comparisons between the studies are impossible due to the different methods of measurement. Age at assessment significantly impacted the height and weight measurements at six- but not 17-months of age in the present study, as would be expected due to more rapid growth around six-months of age compared to at 17-months of age. When considering growth, time before puberty as well as pubertal status at neutering are important factors to consider.There are few studies examining the impact of neutering on dog growth and related health complications. Studies have demonstrated an increased risk of musculoskeletal diseases such as hip dysplasia and cranial cruciate ligament rupture with neutering [11,12,16]. Spain et al. [12] suggested that the increased bone length caused by early neutering had secondary effects on joint conformation. Bitches neutered prepubertally in the present study were taller and perhaps had increased bone length compared to post-pubertally neutered bitches and could therefore be at increased risk of musculoskeletal disease. There is no research directly investigating the effect of neutering before compared to after known puberty on hip dysplasia or other musculoskeletal disease; studies that examine the impact of neutering at different ages commonly do not consider or define pubertal status at the time of neutering. Examination of health data for bitches in the present study in later life could provide useful information.While literature relating to the impact of neutering on vulval conformation and development are rare, some authors have suggested that normal vulva development may be impacted by neutering due to the removal of oestrogen, which is essential for the development of the reproductive tract and external genitalia [2,42,45]. In agreement, our study showed differences in vulval size and appearance at 17-months of age for bitches neutered before or after puberty and the change in vulval size between six- and 17-months was significantly greater for post-pubertally neutered bitches; vulval size in prepubertally neutered bitches was similar to that reported for bitches at six-months of age. In a previous preliminary analysis of the data from the present study presented as an abstract, a significant difference was identified; bitches neutered before puberty were found to have significantly smaller vulvas (length 2.9 cm, width 1.8 cm) than bitches neutered post-pubertally (length 3.2 cm, width 2.1 cm) at 17-months of age [52]. In that study, data from entire bitches were also included and vulval sizes for bitches neutered after puberty were not significantly different to entire pure-bred Labradors and Golden Retrievers and their crosses. The reasons why differences were no longer significant in the present study, despite smaller measurements in vulval length and width for prepubertally neutered bitches, are likely due to the more complex method of statistical analysis including confounding variables, and the removal of bitches with missing confounding variable data and incorrect height measurements for inclusion as covariates from the models. Therefore, the authors suggest that the sample size and statistical methodology of the current study provide a robust analysis.Vulval size has been measured by other authors and compared for bitches neutered at different ages. Salmeri et al. [26] reported that vulvas were smaller and appeared infantile for bitches neutered at seven-weeks (17.8 mm, n = 7) and seven-months of age (16.8 mm, n = 4) than in entire bitches (19.8 mm, n = 6), although the number of bitches was small and statistical analysis was not performed. Marino et al. [59] reported vulval appearance as juvenile and “sunken” for eight bitches that received repeated GnRH implants to delay puberty starting at 4.5 months of age, although actual measurements of the vulva were not performed. In contrast, Kaya et al. [53] reported no difference in vulval size for 13 bitches that received either a deslorelin acetate or a placebo implant between four- and five-months of age for 40 weeks after insertion. The findings of the present study support the suggestion that vulval development may be impeded by neutering prior to puberty.Studies investigating the effect of GnRH agonists implanted in small numbers of prepubertal bitches with the aim of delaying puberty present conflicting results regarding the impact on height and vulval development that may relate to the age at first implantation, or the number of repeated implants administered [53,59,60]. Marino et al. [59] recruited 24 prepubertal Sicilian hounds to a study and treated eight with GnRH implants at 4.5-, 9-, and 13.5-months of age and 16 bitches did not receive an implant; eight of these were ovariohysterectomised at 18-months of age and eight at 4.5-months of age. The authors reported that the vulvas of the eight bitches that received repeated GnRH agonist implants remained juvenile in appearance at 18-months of age. They also identified no significant difference in weight measurements between bitches in different groups and they noted that all growth plates were closed on radiographs taken at 15-months of age. However, no radiographs were taken at timepoints throughout the study to identify the specific age at growth plate closure, therefore differences could not be compared. Kaya et al. [53] studied the impact of a GnRH agonist implant in 13 medium-sized crossbreed bitches; five bitches that received a 9.4 mg deslorelin acetate implant, four bitches that received a 4.7 mg deslorelin acetate implant, and four bitches that received a placebo implant. Results suggested that there were no significant differences between the groups in body weight, height at the withers, vulval size, and humeral length. However, growth plate closure determined radiographically was delayed for longer in bitches that received the higher dose 9.4 mg implant (83.5 ± 8.5 weeks of age) compared to bitches with the 4.7 mg implant (73.4 ± 4.5 weeks of age) and those with the placebo implant (60.9 ± 9.9 weeks of age). The authors suggest that incomplete suppression of hormones due to a single treatment may have resulted in the lack of observable differences in most measures, although small groups sizes (n ≤5 for each) may have affected the reliability of the results.Results relating to veterinarian assessment of vulval appearance at the six- and 17-month assessments and examination of digital images of the vulva in the present study confirmed that vulva size and development are impacted by neutering before puberty. By 17-months of age, significantly more pre than post-pubertally neutered bitches had vulvas that were juvenile and recessed based on veterinary examination and more classified as recessed/inverted based on digital image examination. These findings are to be expected due to the earlier removal of oestrogen for bitches neutered before puberty, which may prevent oestrogenic influence on the maturation of the external genitalia [2,42,45]. Such anomalies in appearance are suggested to predispose to urogenital disease [4,45,46]. However, seemingly in contrast to the findings related to vulval development and the potential for urogenital disease later in life, fewer bitches neutered before than after puberty had skin discolouration noted when digital images from 17-months of age were examined (23% vs. 39%) and more had skin and hair discolouration (72% vs. 60%). The reasons for this are not clear, but could relate to the presence of disease such as dermatitis or vaginitis; this could result in greater time spent licking or cleaning the perivulval area [61]. Further investigation of urogenital disease incidence for the bitches in this study later in life would be useful to examine this in more detail.Dorsal occlusion of the vulva is suggested to be an abnormality in vulval conformation which, if a large percentage of the vulva is covered, can lead to mating difficulties and predispose to disease such as vaginitis and cystitis [61]. In the present study, a dorsal vulva skin fold was observed in almost all bitches on digital images of the vulva at six- and 17-months of age, although by 17-months of age, the percentage coverage reduced to 40% or less for most bitches. The authors propose that some coverage of a dorsal fold is a normal anatomical variant in bitches of these breeds based on the majority of the study population having no other clinical disease. However, the physiological purpose of such a fold is unknown.5. ConclusionsOur results suggest that for Labrador/Golden Retriever crossbreed bitches, neutering before puberty does impact growth and physical development. Changes in height and vulval size were significantly impacted by neutering before puberty, and bitches neutered before puberty were more likely to have vulvas that appeared juvenile, recessed, or inverted at 17-months of age. Although not significantly different, bitches neutered before puberty were taller, heavier, and had smaller vulvas at 17-months of age than those neutered after puberty. While these findings suggest no contraindications for prepubertal ovariohysterectomy, the longer-term health implications of these differences in physical development need to be further investigated and better understood before recommendations can be made.These findings improve our understanding of the impact on physical development of neutering bitches at specific timepoints in relation to known puberty. As such, they will be of interest to veterinarians, assistance and working dog organisations, and pet owners who have to make decisions about whether or not, and when, to neuter female dogs.

animals : an open access journal from mdpi

10.3390/ani12070913

PMC8996943

Although many dogs with blindness diagnosis can reach a similar age compared to those not affected, often the owners require euthanasia of their animals. This choice leads to conflicting moral principles relating to what is better for the animal and the owner. This article discusses the suitability of euthanasia in blind dogs. To better assess factors influencing the choice of euthanasia, four different scenarios were constructed that described various situations regarding the animal’s aptitude, pet owner, and veterinarian relations.

In dogs, several primary or secondary diseases affecting the ocular structures may cause blindness. In cases where the visual impairment is not associated with severe systemic involvement and the animal can still have, predictably, a good “long-term” quality of life, the veterinarian should inform the owner about the differences between humans and animals, concerning the type of visual perception. In the light of the daily findings in veterinary clinic practice, the Authors report four different scenarios with conflicting views between veterinarians and owners about the euthanasia request for a blind dog. They underline how the diagnosis of incipient or already established blindness in dogs can sometimes lead to an inappropriate request for euthanasia.

1. IntroductionDogs’ blindness often is due to a variety of acute and progressive diseases or trauma or can be consequent to Brachycephalic Ocular Syndrome [1]. Some of the most common causes of vision loss in dogs include cataracts, glaucoma, sudden acquired retinal degeneration syndrome (SARDS), progressive retinal atrophy (PRA), and retinal detachment (RD). Blindness can be also caused by certain cancers (brain and orbital tumors), neurologic diseases, and drug overdoses that affect the retina (e.g., an overdose with ivermectin) [2].The main causes of blindness are reported in Table 1.Loss of vision can occur suddenly or develop gradually over time. It may be complete (involving both eyes) or partial, involving only one eye or even certain parts of the visual field. To best treat patients and maintain their good “long-term” quality of life, a complete clinical examination of the dog considering potential clinical manifestations of the diseases involving the eyes is required. Indeed, many diseases such as diabetes, Cushing’s syndrome or hypertension show serious ocular signs that may compromise the quality of life.Identifying the specific etiology of blindness and/or ocular structures involved is fundamental to the veterinarian advising owners on the prognosis for the animal’s vision and general health [4]. Then, it is necessary to carry out a diagnostic approach that includes:(i)an ophthalmic history identifying the onset and duration of blindness, degree of blindness (as perceived by the owner that could often report disorientation, clumsiness, and/or anxiety of the dog) [5], if there have been any changes in the appearance of the eyes and any behavior changes, other signs of disease, and received treatments.(ii)vision assessment, using the menace response test, the visual placing reaction, and the maze test [6,7];(iii)causative lesion localization by Pupillary Light Reflex (PLR) examination of the eye, potentially with ocular ultrasound, electroretinogram (ERG), and/or functional magnetic resonance imaging (MRI) [8].Therefore, the veterinarian must help the owner to comprehend why his/her dog has developed a loss of vision (whether the blindness is complete or partial), if systemic signs of disease are present or systemic diseases have been previously diagnosed (because often several systemic diseases, such as infectious diseases, hypertension, etc., may initially be recognized by their ocular manifestations), and how to help the animal to adapt to ocular impairment. Furthermore, modern progress in veterinary ophthalmic surgery, such as cataract removal by phacoemulsification, retinal detachment surgery, and laser endoscopic cyclophotocoagulation, has increased the treatment options for the patients that previously would be permanently blind [9,10,11,12], thereby improving their life quality. Although many dogs with blindness diagnosis can live a good life [13], the owners require euthanasia of the animal but with a treatable and/or non-fatal condition (“convenience euthanasia”), either for financial reasons or simply because they do not want he/she anymore. This choice raises ethical questions because of veterinarians’ duties towards both animals and owners, particularly if owners’ wants are against the animal’s interests or welfare. This is because the animals, by biocentric viewpoint, are viewed as moral patients and their interests are considered to be a priority; veterinarians act as advocates to protect them, tending to act similar to pediatricians [14]. This objective is in contrast with convenience euthanasia [15]. Furthermore, lacking a clearly defined set of rules when approaching a canine blindness case could result in the development of conflicting moral principles. Thus, decisions over dog euthanasia must be made considering the severity of the condition, the long-term prognosis, the animal’s rights, and the interests of the owners. The veterinarian should weigh up all of these considerations before deciding whether euthanasia is the best course of action.Although there are a variety of potential scenarios that involve the euthanasia dilemma in daily clinical practice, in this study the focus is to discuss if putting down a dog with complete blindness is right.To better discuss the ethical dilemmas inherent in the decision-making around euthanasia of a blind canine patient, four different hypothetical scenarios that describe realistic clinical cases relating to dogs with the diagnosis of blindness were constructed.2. Blindness from the Animal Welfare Point of View2.1. Animal Welfare and EthicsAll animals, regardless of the species to which it belongs, are considered sentient beings capable of feeling emotions such as pleasure, pain, stress, and suffering, both at physical and psychological levels. Animal welfare has been defined in several different ways [16,17,18,19,20,21,22,23,24,25]; it can be considered a positive mental and physical condition related to the satisfaction of its physiological and behavioral needs and its expectations [26]. This state varies according to the animal’s perception of its situation. Animal welfare includes three concepts: (i) an animal’s feelings; (ii) an animal’s ability to perform natural behavior, and (iii) an animal’s health and biological functioning [27,28]. To evaluate animal welfare, all of these concepts (behavior, affective state, and health and biological functioning) should be considered.Good animal welfare requires disease prevention and veterinary treatment, appropriate management, nutrition, responsible care, and, when necessary, a humane death [28].Although welfare scientists have not always agreed on the animal welfare definition [29], our view is that good health and the absence of stress is not enough to ensure welfare. It is also necessary to consider what the animal feels, not only unpleasant subjective perceptions, as mentioned above, but also to look for the signs of expression of positive emotions [30,31,32,33]. The analysis of behaviors and health status of the animal provides an integrated vision of its welfare [34].There are several pro-social and behavior signals of welfare in dogs indicating a positive emotional state such as relaxed body posture and facial expression (ears to the side or forward if anticipating a positive event), consuming food when offered, playing, soliciting attention, etc. Instead, some medical problems can modify animal behavior. A medical condition such as blindness can, however, modify animal behavior as it may modify or prevent the animal’s perception of the environment. Partial blindness could cause fears or phobias [35]. If the ocular problem causes pain, the animal changes behavior (decreased interaction with other animals or humans, reduction of appetite, aggression, fear reaction, vocalization, altered facial expression, and/or posture, restlessness, etc.) [36].Sharing the thought of Rollin [37], in our opinion, a significant part of animal welfare is ethics, given that it influences an individual’s concept of welfare, even if other Authors believe that animal welfare science is not interchangeable with or a synonym of animal ethics [38]. In this context, ethics would prescribe the veterinarian’s course of action in terms of the blind animal’s welfare as well as the rights/obligations/welfare of the owner. In fact, the veterinarian has an ethical obligation to provide good care for his/her patient. Ethics may also be considered when addressing value-based questions such as the acceptability of animal quality of life. Therefore, the veterinarian should take into consideration the animal’s needs and discuss potential ethical concerns with the owner, taking the role of patient advocate [39]. In this way, his/her actions would primarily look after the well-being of the animal following the paediatrician model based on the moral value of the animal [15].2.2. Blindness and Animal WelfareGiven that poor quality of life as sensed by the owner implies a reason for euthanizing the companion animals [40,41], management of any disease that is blinding is important both from an animal welfare point of view (concerning, e.g., the pain) and from an owner’s viewpoint (e.g., respect of the treatment).Given that the perception of pain decreases the quality of life in all animals, respect for animal welfare and dignity must be viewed as the main principle in the veterinary practice, with the consequent need for a better understanding of animal pain management from an ethical viewpoint. Alleviating pain is necessary for both ethical and clinical reasons [42]. Animals are a valued part of society and must be protected from needless suffering [43]. From a clinical standpoint, the pain has detrimental effects on animal well-being [42], and untreated pain can lead to weight loss, aggression, self-mutilation, and heightened pain sensitivity. As above mentioned, the issue of persistent pain in any blind eye is a serious animal welfare concern, though not all blindness conditions are painful. In estimating the deleterious effect of blindness in a dog, it is important to pay attention to the potential pain because the animal does not always seem to demonstrate visible signs of discomfort.Severe ocular pain is present in conditions such as endophthalmitis, glaucoma, and penetrating ocular trauma. A study reported that the chronicity of the pain associated with glaucoma exacerbated lethargy [44]. This implies that it may be a reliable indicator of the quality of life [45]. For these and other ocular diseases with painful eyes, to avoid dog euthanasia, enucleation surgery may be strongly recommended as an end-stage solution to irreversible pain. However, enucleation, especially if bilateral, might be seen as a discouraging and reluctant solution for many owners, as highlighted by owners of horses by Wright et al. (2018) [46]. A recent study investigating owners’ perceptions of their dogs’ quality of life following bilateral enucleation assessed their satisfaction with the procedure [47]. The owners showed good satisfaction and perception of improvement in their dog’s quality of life, considering this surgical procedure a viable option for managing dogs with irreversible pain. Hamzianpour et al. (2019) [47] found that a greater part of patients appeared able to return to their activity and playfulness and showed a lack of aversion to facial/ocular palpation.In this perspective, the veterinarian must educate the owners, for example, to quickly recognize ocular pain, explain in more detail when enucleation may be necessary, and better manage the dog [44] because the management that optimizes the blind dog’s welfare depends on multiple interventions aimed at creating a safe and comfortable home environment. Sudden onset blindness could be significantly harder for both the dog and owner than a gradual loss of vision. To reduce the stress because of the progressive blindness, it could be utilized to adapt to social and physical environments and new relational practices by increasing the predictability and controllability of the animal’s situation [48]. In this regard, it might be useful to remove potentially hazardous objects that the blind dog may bump into; to leave food and water bowls in the same place; to place scent or tactile location signals to aid orientation around the house; to consider the levels of lighting, e.g., dogs with PRA function better in bright light. Welfare management strategies therefore should be considered part of a more comprehensive medical approach. Because animal welfare considers psychological as well as physical aspects, blindness can change the behavior of the animal, i.e., modifying or preventing the perception of the environment [35]. If the loss of sight is gradual, the dog compensates gradually, and behavior changes may be slight and not noticeable until the animal is completely blind. Sudden blindness can result in more dramatic behavior changes: the dog may be disoriented and hesitant when walking, bumps into things, and may vocalize more often. Dogs who become blind suddenly may also develop uncharacteristic behaviors until they learn to adapt. They may be unwilling to leave their sleeping area and may also be reluctant to explore, may appear withdrawn, and may bark when disoriented or in need of reassurance.Therefore, to adapt living conditions to the dog’s impaired vision means guaranteeing his/her welfare.2.3. Is Euthanasia in the Dog’s Interest?Determining whether euthanasia is in a dog’s interest remains a complex and difficult matter. There is no prescriptive list evaluating when to resort to euthanasia. However, this evaluation could be done by weighing up the benefits and harms in terms of suffering and longevity of the animal, i.e., in the sense that life can continue only if beneficial—that the resultant life is worth living. Even if it is not clear whether nonhuman animals might be able to develop a desire to die in certain situations, death can be in their best interest [49]. Regarding the above evaluation, some Authors [50,51] have proposed a checklist that includes the degree of pain the animal can experience such as changes in walking, presence or absence of appetite, hydration, breathing without or with difficulty, normal urination and defecation, etc., in other words, conditions that could lead to suffering. Therefore, the incapacity to prevent any suffering could legitimize the euthanasia of any dog. Therefore, euthanasia could be justified when an animal cannot avoid having a life of overall negative welfare except by dying, and any medical over-treatment would be unethical. Indeed, owners considering their dogs as members of the family might want to utilize every possible treatment to ensure their animals remain alive if possible and probably against the interest of the animal in the case of a poor prognosis. A conflicting situation could arise between moral arguments in favor of the animal’s life and situations such as time, effort, and money that finally determine the euthanasia choice despite what is in the best interest of the animal. For this reason, as will be mentioned in the “Final Thoughts” section, it could be considered useful in euthanasia decision an algorithm that includes questions such as “Is the benefit to the owner of euthanasia greater than the harm to the animal?” [52]. Yates (2010) [52] affirmed that a relevant evaluation is not of the animal’s state in the present time, but its expected welfare over time until the next evaluation. This makes the assessment even more complicated because it involves a balance between the welfare and each of the numerous potential outcomes. This should be annealed by knowledge of human–animal differences. For example, it is not well defined whether animals have long-term objectives that they hope to realize, death, anxiety, or can predict that their suffering will finish. 3. Different Scenarios Concerning the Euthanasia RequestIn clinical practice, there is no single and applied framework to determine the correct approach of a request for euthanasia in blind dogs because each must evaluate the animal needs and the context in which the case is referred. To understand what ethical approach should be taken into consideration to establish when euthanasia is appropriate, we report four different realistic scenarios in which the owner bringing his/her dog with blindness (complete or partial) to the veterinarian requires euthanasia although it seemed to be not appropriate by veterinarian’s viewpoint. Scenario 1: Irreversible blindness and dogs’ adaption A nine-year-old female pug affected with Cushing’s syndrome was presented for evaluation of rapid loss of vision developed in the last weeks. The absence of menace response and a decrease of the pupillary light response was revealed in both eyes. At the ophthalmological examination, the tapetal fundus appeared hyper-reflective, with reduction of the retinal vasculature and pallor of the optic nerve head. SARDS was suspected and confirmed from the electroretinogram test. The owner was informed of the irreversible blindness and required the euthanasia of his dog believing that she did not have a good quality of life. He also considered that the presence of frequent urination due to the syndrome was frustrating to himself. Irreversible blindness is a complex scenario for the emotional involvement of the owner because appropriate management of blind animals is always challenging, requiring several skills in addition to clinical knowledge. Though some conditions are irreversible, blindness could be generally well-tolerated, and dogs could adapt well with minor adjustments to their lifestyle, maintaining a good quality of life [13]. Relating to the frequent urination as reported in the above case, it can be pharmacologically treated. In a survey carried out by Stuckey et al. (2013) [53], it was reported that most owners of dogs affected with SARDS perceived the quality of their animal’s life as good, therefore discouraging euthanasia of dogs with SARDS. Though the literature suggests that blind companion animals may have an excellent capacity for adaptation, it often happens that the owners do not appreciate this, and they were distressed because their dogs presented abruptly with irreversible blindness, requiring euthanasia.Comprehensibly, owners often do feel empathy for their animal [54] and imagine themselves being blind and the consequences it would have on their lives.Consequently, given that dogs lack visual acuity, they adapt more easily to blindness than humans. Dogs’ vision is based more on shape, movement, and contrast than on detail [55]. Moreover, their other senses, such as hearing, touch and smell, and taste, become more highly attuned to permit them to successfully navigate to seek food, water, etc. [56,57]. Svensson et al. (2016) [58] reported that dogs with PRA learned to adapt to their decreased visual function so that the owner rarely recognized visual impairment until a late stage of the disease. Menotti-Raymond et al. (2010) [59] described the ability of cats with PRA to adapt to the progressive visual impairment using their other well-developed senses. Therefore, in our opinion, vision loss alone is not enough justification for euthanasia. For this reason, the veterinarian may refuse to euthanize, believing that dogs typically have a good quality of life despite blindness. Despite the adaptability, deciding to euthanize a blind dog may become permissible if the animal is geriatric with concomitant debilitating diseases and little hope of full recovery. Scenario 2: Blindness in working dogs A 7-year-old male English Setter was presented to a referral hospital for a history of conjunctivitis and hesitance navigating around objects (trees, branches, and bushes.) in the last days. Menace response, dazzle reflex, and pupillary light reflex were negative, indicating blindness of both eyes. A complete ophthalmic examination, including slit-lamp biomicroscopy, tonometry, and indirect ophthalmoscopy, was performed. Conjunctivital petechiae and corneal edema in the right eye and hyphema in the left eye were observed. Ocular ultrasonography showed a complete retinal detachment and vitreous hemorrhage. Canine monocytic ehrlichiosis was diagnosed. The owner did not correctly perform the pharmacological treatment for ehrlichiosis, and the condition of the eyes worsened. The veterinarian suggested bilateral enucleation. The owner did not agree to the veterinarian’s proposal because the dog was no longer suitable for the purpose for which he was bought (hunting). Consequently, he required euthanasia. In other situations when owners adopt companion animals for a very specific purpose, such as guarding or hunting dogs, and the animal becomes unable to perform the desired purpose because is blind, the owner may require his/her euthanasia. This happens because companion animals are considered tangible personal property in the strict legal sense [43] and can be euthanized for any rationale their owners think up. In fact, in a recent study relating to the origin of the convenience euthanasia dilemma, it has been reported that the request of convenience euthanasia from owners is normal because animals are also seen as objects in their utility role in relation with owners, contributing to minimize needs and interests of the animal [60]. In this scenario, the owner, being disappointed in his/her animal’s expectations because of the animal being unsuitable for the purpose for which it was intended (i.e., hunting) due to the blindness, had an illusionary vision of what his/her dog should have been and done. The owner rejected his/her animal easily (that is, required euthanasia) knowing that it would be simple to find another one that perhaps met his/her expectations [60].In these cases, when death is not in a dog’s best interest, the veterinarian, unconvinced that the burden on the owner is great enough to justify the animal’s death, must consider other reasons for euthanizing the animal. Therefore, the veterinarian must decide whether euthanizing an animal is justifiable and, whether he/she can ethically decline an owner’s request to perform euthanasia, offering to the owner alternatives and suggestions for cases in which the animal is blind, i.e., avoiding changes in the domestic environment, use auditory stimulation, etc. Conversely, if the owner is not able to correctly perform a treatment, he would not be able to support his blind dog. In this context, the choice of the veterinarian to euthanize the dog may be difficult. However, if the dog presents a good prognosis, as a medical professional, the veterinarian must act as an advocate to protect his/her patients. This is because all animals have preferences, interests, needs, and can feel pleasure and pain [61,62]. In other words, they are seen morally as having an inherent value similar to humans [62] and consequently, are entitled to have respectful treatment.Those who possess an inherent value (humans and animals) must be treated as ends in themselves, and whenever they are treated as a means to an end (in the case of dogs used for the hunting)—as if they had a value—then they are treated with a lack of respect.Since there is a connection between rights and duties, moral rights also generate duties not to kill physically healthy animals (blind dogs).Nevertheless, some veterinarians prioritize the owner’s interest and proceed with euthanasia [60]. Scenario 3. Financial constraints of the pet owner A 9-year-old Labrador Retriever suffered from a mature cataract in both eyes. The veterinarian determined that the animal may be a candidate for surgery, but cataract extraction surgery by phacoemulsification required a cost of about 2500.00 €. Besides, before the surgery, the animal needed to have specific tests (i.e., electroretinogram, ocular ultrasound, comprehensive blood panel) to evaluate retinal and overall health to provide a prognosis for return of vision following the surgery, with additional costs. The owner did not want to spend the money either on the surgical treatment or the tests, but required to euthanize the animal. In this scenario, where a cataract impairs the dog’s vision, the owner is not willing to pay for the animal’s treatment, even if the surgical treatment such as cataract extraction surgery by phacoemulsification can improve his/her vision. Consequently, he/she decides to euthanize his/her animal. Therefore, in this economic decision making, euthanasia becomes a viable and economic alternative, although, as reported by some Authors [63], the request of euthanasia for economic limitations is one of the more frequent stressful scenarios involving the veterinarian.Financial reasons are considered by Herfen et al. [64] as the ultima ratio for euthanasia. In this case, premature euthanizing a dog who became blind is done for instrumental rather than compassionate reasons.In the case in which the owner is unable to afford treatment, it may be suitable “to make known the options and eligibility for charitable assistance or referral for charitable treatment”, as stated by the RCVS (Royal College of Veterinary Surgeons) Code of Professional Conduct for Veterinary Surgeons [65]. Another financial instrument to reduce the burden of veterinary medical costs on the owner could be pet health insurance that would consequently reduce or abolish the risk for euthanasia [66]. In our opinion, animal welfare should be the veterinarian’s primary concern. Even if animals are considered not able to participate in the decision making [67], the veterinarian does not treat them as objects (the client’s property) but as subjects, patients who deserve quality medical care. The veterinarian, where there is no legal framework about euthanasia in dogs and cats, is morally obligated to attempt influencing client decisions, although he/she also has a duty to the owner and should consider his/her autonomy in decision making [52]. It must be the veterinarian’s responsibility to use his/her judgment on different options beneficial to the patient and present all available options to the owner ranging from the best practice (gold standard) [68,69] to euthanasia (even if in some cases, e.g., in intracranial neoplasia, the best practice may be opting for euthanasia). Indeed, we believe that client education is essential, and the exchange of knowledge between veterinarian and owner is an important factor in the decision-making process [70]. Autonomy in decision making requires the owner to adequately comprehend what the situation involves. Consequently, the respect for autonomy is strictly associated with informed consent. Informed consent, in turn, may be comprehended as given only if the owner receives and understands full information on the options available and voluntarily chooses one of those options. The veterinarian must also inform the owner of uncertainties about the treatment and about what a treatment option involves. Together, respecting owner autonomy and obtaining informed consent, aims to reduce the risk of unduly influencing the patient’s decision making and provide guidance in veterinarian–client relationships. Scenario 4. Central nervous system (CNS) blindness A ten-year-old female English cocker spaniel was presented to a referral hospital for a 24-h history of blindness associated with lethargy occurring in the last weeks. During the neurological examination, the animal showed disorientation and tended to bump into objects. The menace response was bilaterally absent, as well as the response to the cotton ball test; pupillary light reflexes were reduced, pupils appeared mydriatic, and dazzle reflex was normal. A complete blood count and biochemical tests were unremarkable. The results of neurological examination suggested the presence of a lesion localizing to bilateral retinas, optic nerves, optic tracts, or optic chiasm. The veterinarian proposed diagnostic investigations to formulate a more accurate diagnosis (Magnetic Resonance Imaging of the brain, thoracic radiographs, and abdominal ultrasound), but the owner required euthanasia. This scenario takes into consideration central nervous system (CNS) blindness. Several infectious and conditions causing inflammation of the CNS can affect the visual cortex and cause cortical blindness. Given that the optic nerve is an extension of the CNS, some diseases, such as granulomatous meningo–encephalomyelitis (GME) and infectious optic neuropathies due to distemper (paramyxovirus) and tick-borne encephalitis (flavivirus), can lead to central blindness. Intracranial neoplasia, of which meningioma is the most common primary tumor found in the dog [71], can also lead to sudden blindness. Although the disease process is progressive, the gradual onset of more subtle vision deficits is not recorded by owners. To increase survival time in dogs with brain tumors, it might be worthwhile for surgical resection and fractionated radiotherapy [72], although these tumors are associated with significant morbidity [73]. Rossmeisl et al. (2013) [74] reported that the prognosis for dogs with palliatively treated primary brain tumors is poor. Given that, some Authors [75] highlight how it is difficult for veterinarians confidently to advise to owners treatment decisions because there have been no robustly designed formal prospective clinical trials comparing various treatment options for intracranial tumors. They suggest that owners would know the data for all-cause mortality because they ignore that other lesions might contribute to their animal’s death within a specified period. Unfortunately, in these situations, it is not easy to be able to opt for one (treatment) or the other solution (euthanasia), given the influence of tumor-associated variables such as lesion burden or histopathologic type and grade and the severity of neurologic dysfunction [74]. For example, euthanasia could be performed due to refractory seizures or a sudden decompensation to the clinical signs existing before the treatment. Therefore, the decision to perform euthanasia or not should consider some factors of the patient animal (e.g., quality of life and advanced-stage disease), the owner (e.g., previous experience of dogs with tumors, emotional bond with the animal, and perceptions concerning tumor treatment), and veterinarian (e.g., experience and training in oncology treatment, knowledge of tumor behavior and cancer treatment options) and be taken based on the expected biological behavior of the tumor and expected benefit to the animal. Without any doubt, in our opinion, owners should be advised of the survival times based on the neuroanatomic location of the tumor when discussing outcome expectations with palliative treatments or when attempting to compare the relative efficacy of a brain tumor treatment or when choosing to euthanize the animal. The most important thing is the dog’s quality of life. A treatment that could prolong his/her life may also prolong his/her suffering.4. Final ThoughtsScenarios make us realize how under a relational logic of care, moral decisions cannot be made in the abstract sense.To assist the pet owner in the euthanasia decision-making process, it could be useful to apply the algorithm for euthanasia decisions produced by the BVA’s (British Veterinary Association) Ethics and Welfare Group [76].The owner often sees premature euthanasia as the only option by believing that it is the only way to show care or compassion for his/her dog. In this case, in our opinion, euthanasia is not a form of compassionate care. In other words, empathic atonement to the blind animal may sometimes, but not always, oblige the owner to conduct compassionate euthanasia. Among other things, the vision’s loss produces emotional distress in owners. The inability of many of them to cope with the circumstances surrounding a blind dog’s visual status needs to be considered when communicating medical information to these individuals. Moreover, when advice concerning treatment is given, it would be hopeful that the owner understands all those factors that he/she perceives as important to improve the life quality of his/her blind dog. Although the main objective of medical or surgical treatment should be to improve visual function, this should be understood given those factors important to each owner. To appreciate the severity of the condition that a dog is experiencing and how this is affecting his/her ability to enjoy life, health-related quality of life assessment tools should be carried out as shown by several authors [77,78,79,80]. The scale of Villalbos [81] represents a straightforward means for veterinarians and owners to assess dogs’ quality of life. This tool would help to monitor objectively the progression of the animal’s condition and agree whether euthanasia is necessary and ethically appropriate.The owner could assist his/her animal gone blind by developing in himself/herself qualities of character such as patience and compassion.In other words, euthanizing a companion animal occurs because humans cannot bear the costs, financial or emotional, of continued medical care or because the death might save them from more suffering. In some cases, some options may prolong and improve their lives. Several aspects must be considered concerning the good health of the animal, his/her age, medical prospects, and possible alternatives. Given the strong bond between the pet owner and his/her animal, the veterinarian must also attend to the emotional needs of the owner as a part of the provision of a professional business service [82]. The veterinarian should educate the owners on the welfare needs of their animal, including euthanasia as a valid treatment choice by engaging “in a process of ethical reasoning when considering treatment choice” [65,83]. To assist in the euthanasia decision, it could be appropriate to apply the checklist of questions proposed by Edney (1989) [50].Is the animal in pain, distress, or discomfort?-Can the animal walk and balance?-Can the animal eat and drink?-Does the animal have inoperable tumors that can cause pain, distress, and discomfort?-Can the animal breathe without difficulty?-Can the animal urinate and defecate normally?-Can the owner cope physically and emotionally with nursing?In conclusion, owner education with appropriate information and support is required to help him/her better manage his/her animal’s condition [84] (Box 1) because many dogs can live a fulfilled and happy life without their sight.Box 1Suggestions for the owner.KEY POINTS: Creating a safe space to retreat might help the dog feel more confident.Assigning a room or a corner and filling it with his/her water and food bowls, favorite toys, and his/her bed. For example, utilizing noisy toys, which can be especially rewarding for the dog.Removing any hazards or sharp objects that might injure him/her.Avoid shift things to prevent confusion and accidents.Talking to the dog can help him/her feel at ease and help locate the owner.Use the voice to get the dog’s attention before stroking him/her to avoid scaring.Installing baby gates at the top and bottom of stairs to keep the dog safe.Using different textured rugs, carpets and flooring might help him/her build a mental map of his/her surroundings.Putting bells on other pets’ collars or shoes might also help the dog locate his/her owner around the house. It is also important to educate the pet owner of his/her duties to his/her dog and ensure he/she clearly understands any legal implications if he fails to fulfill his/her responsibilities. Therefore, ophthalmologist veterinarians have a crucial role in improving the welfare of both blind dogs and their owners. Requests of killing without a “reasonable reason” are considered morally undesirable and an offense in animal welfare law [63] in European Union where the legislation recognizes animals as sentient beings.

animals : an open access journal from mdpi

10.3390/ani13071160

PMC10093630

The objective of this study was to measure the evolution with age of the antioxidant defense ability of two strains of chicken with fast or slow growth rates. Several parameters related to metabolic status, redox balance, and antioxidant defense activity were measured from hatching to 42 days of age, in liver, muscle (breast and thigh), and in plasma. Our results confirm the high level of oxidative stress just after hatching in chicks and, consequently, a high need for antioxidant defense during this period. Several enzymes or molecules, whose levels fall during the first week or two after hatching, would help to ensure this need. The antioxidant pathways used to maintain redox balance change with age and may differ between slow- and fast-growing chickens. SOD activity appears to be a key player in the antioxidant response of slow-growing chickens, while uric acid is thought to play a more important role in fast-growing chickens, particularly at the end of rearing.

The evolution of parameters known to be relevant indicators of energy status, oxidative stress, and antioxidant defense in chickens was followed. These parameters were measured weekly from 1 to 42 days in plasma and/or muscles and liver of two strains differing in growth rate. At 1-day old, in plasma, slow-growing (SG) chicks were characterized by a high total antioxidant status (TAS), probably related to higher superoxide dismutase (SOD) activity and uric acid levels compared to fast-growing (FG) chicks whereas the lipid peroxidation levels were higher in the liver and muscles of SG day-old chicks. Irrespective of the genotype, the plasma glutathione reductase (GR) and peroxidase (GPx) activities and levels of hydroperoxides and α- and γ-tocopherols decreased rapidly post-hatch. In the muscles, lipid peroxidation also decreased rapidly after hatching as well as catalase, GR, and GPx activities, while the SOD activity increased. In the liver, the TAS was relatively stable the first week after hatching while the value of thio-barbituric acid reactive substances (TBARS) and GR activity increased and GPx and catalase activities decreased. Our study revealed the strain specificities regarding the antioxidant systems used to maintain their redox balance over the life course. Nevertheless, the age had a much higher impact than strain on the antioxidant ability of the chickens.

1. IntroductionIn poultry farms, birds face different types of stress, whether they are environmental (cold, heat, ventilation), technological (prolonged storage of eggs before hatching, delay for placing chicks after hatching, high rearing density), nutritional (dietary transitions, mycotoxins, acid oil, toxic metals, insufficient trace elements), or related to their health or immune status (pathologies, vaccination process) [1]. Modern strains of broilers that have been selected for high growth rate and breast meat yield are more susceptible to oxidative stress [2,3]. In comparison, the strains used for free-range Label Rouge production are characterized by slower growth rates, lower breast muscle yield [4], and are known to be more robust [5]. As mammals, birds are equipped with antioxidant defense systems (enzymes, glutathione, vitamins, mineral, etc.) that allow them to maintain the redox balance at the cellular and tissue levels [1]. In poultry, the development of these systems during embryo development has been well-described [6,7,8], but their evolution after hatching is not so clear, even though several studies have focused on this period. Indeed, Yang et al. [9] measured different markers related to the redox balance in the plasma of chickens fasted for 12 h at the ages of 14, 21, and 28 days but did not evidence any evolution of these markers with age. However, it has been shown that the antioxidant potential decreased significantly over the first 10 days post-hatching and was offset by an increase in glutathione peroxidase (GPx) activity in the liver [10]. Surai et al. [11] also observed a sharp decrease in hepatic vitamin E concentration in different avian species (chicken, turkey, duck and goose) during the first two weeks of post-hatch growth. Between 21 and 42 days, Del Vesco et al. [12] showed that in the liver, the glutathione (GSH) concentration decreased and catalase activity increased while the TBARS value (thio-barbituric acid reactive substances), a lipid peroxidation marker, and the GPx activity remained unchanged. Mahmoud and Edens [13] measured various forms of glutathione (oxidized and reduced, GSSG and GSH) in chicken blood and the activities of GPx and glutathione reductase (GR) at 2, 3, and 4 weeks of age. The GSH concentration and GPx activity decreased with age while the GR activity increased. Mizuno [14] assayed in the breast muscle of chickens Cu–Zn superoxide dismutase, Mn superoxide dismutase, catalase, glutathione peroxidase, and glutathione reductase activities and thio-barbituric acid-reactive products 1, 2, and 4 weeks, and 4 months after hatching. All of these enzyme activities declined as the chickens grew older. The aim of our study was to provide a general overview of the evolution during rearing of metabolic and redox status as well as antioxidant defense in chickens used for meat production. The evolution of several plasma and tissue markers was then monitored on a weekly basis during the first weeks of growth in two types of strains: a fast-growing (FG) strain used for standard production and a slow-growing (SG) strain used for free-range French Label Rouge or organic production. In this study, a large set of 15 metabolites, vitamins, or enzyme activities known as good indicators of the energy status, oxidative stress, and antioxidant defense of chickens were measured in plasma or/and in muscles and liver.2. Materials and Methods2.1. Animals and Experimental DesignAll experimental procedures were performed in accordance with the French National Guidelines for the care and use of animals for research purposes (Certificate of Authorization to Experiment on Living Animals no. 7740, Ministry of Agriculture and Fish Products). The animal experiment was ethically approved by the French authorities under number APAFIS #22060-2019091916239724v2. Fast-growing (FG) Ross 308 and slow-growing (SG) JA 657 male chicks were furnished by Boyé Accouvage (La Boissière-en-Gâtine, France) and reared on wood straw under controlled conditions at the poultry experimental unit (PEAT) of INRAE Nouzilly (France). One hundred and twenty chicks per strain were distributed into two contiguous rooms (one room per strain) in a closed building to limit environmental disturbance and presenting the same environmental conditions (temperature, humidity, light duration, and intensity). The rearing density was 10 birds/m2 at the beginning of the experiment. All birds were individually identified by wing-banded and ad libitum fed diets under pelleted form. There were three feeding periods: starting (1–14 days), growing (15–28 days), and finishing (29–42 days). The composition and main characteristics of the diets are presented in Table 1. Diets were formulated according to the breeders’ recommendations. The vitamin E content of diets was 20 ppm (amount sufficient to cover growth needs) to avoid interfering with the antioxidant defenses of the chicks. To evaluate the growth performance, all chickens were weighed at D1, D14, D28, and D42 and the feed consumption per strain was registered for each period. To evaluate redox status, twenty-one-day-old chicks of each strain, chosen at random, were weighed and then sacrificed by cervical dislocation, alternating strains between each chick. Before sacrifice, blood samples were taken from the occipital sinus to recover plasma after centrifugation and stored at −80 °C. Samples of Pectoralis major muscle, a mix of thigh muscles, and liver were also collected, immediately frozen in liquid nitrogen, and stored at −80 °C. At 7, 14, 21, 28, 35, and 42 days of age, 10 birds per strain chosen at random were weighed, sacrificed, and sampled as at hatching. Eight hours before sacrifice, the animals were fasted but continued to have ad libitum access to water.2.2. Determination of Plasma Metabolites and Redox Status of ChickensThe activities of superoxide dismutase (SOD), glutathione peroxidase (GPx), glutathione reductase (GR) and peroxidase, the total antioxidant status (TAS) and the concentrations of hydroperoxides (HPO), glucose, β-hydroxybutyrate (β-OH), free fatty acids (FFA), uric acid, and triglycerides were determined on plasma using a biological analysis robot (Arena 20XT, Thermo Scientific, Ilkirch, France) and commercial kits (Table 2). The TBARS index was determined according to Lynch and Frei [15] to estimate the lipid peroxidation. The concentrations of α- and γ-tocopherol were determined using a HPLC Thermo U3000 (Thermo Fisher Scientific, Illkirch, France) after the extraction of the organic phase with hexane (Faculty of Medicine, Marseille, France). On tissues, only the activities of SOD, GPx, GR, and catalase, and the TAS and the TBARS values were measured using the same methodologies as for plasma. All measurements of enzymatic activities were realized after the solubilization of around 100 mg tissue with 800 µL of a buffer containing 0.05 M Tris-HCl, 1 mM EDTA, and 0.25 M sucrose (pH 7.4). The tissue was first ground with a ball mill (Retsch MM400, Grosseron, Coueron, France) for 1 min, then centrifuged (10,000× g, 30 min, 4 °C), and the supernatant was collected and stored at −80 °C. For each tissue sample, the protein content was determined by UV–Visible spectrometry using a BCA Protein Assay Kit (Sigma-Aldrich, Saint-Quentin Fallavier, France) to express results relative to the protein contents.2.3. Statistical AnalysisFor all of the measured parameters, the replicate was the chicken. For the whole rearing period, the effects of age, strain, and their interaction were tested by analysis of variance using the Statview software (version 5.0) and a significance level at p ≤ 0.05.3. Results3.1. Growth Performance of ChickensThe body weight of chickens increased regularly with age (p < 0.001) from 51.6 g at D1 to 3134 g at D42 for the FG strain, and from 33.9 g at D1 to 1090 g at D42 for the SG strain (Figure 1). The effect of strain was significant (SG < FG, p < 0.001) for the whole experimental period and the difference between strains strongly increased with age. The growth performance of each group is presented in Table 3.3.2. Plasma MetabolitesThe glucose level remained quite stable between D1 and D42 in the FG strain. In contrast, it increased regularly from 2026 mg/L at D1 to 2577 mg/L at D28 and then decreased to 2362 mg/L at D42 in the SG strain (p < 0.001; Figure 1). The SG strain had a lower glycemia at D1 (2026 vs. 2342 mg/L, p < 0.05) but a higher one at D21 (2506 vs. 2186 mg/L, p < 0.05) than the FG strain.The uric acid level of the SG strain decreased regularly from 87.44 mg/L at D1 to 36.66 mg/L at D21 (p < 0.001) and then remained stable (Figure 1). Interestingly, the opposite evolution was observed in the FG strain with an increase from 29.41 mg/L at D1 to 91.78 at D21 followed by a decrease to 31.31 mg/L at D35 and a further increase to reach 52.66 mg/L at D42 (p < 0.001). Consequently, the SG chickens were characterized by a higher uric acid content at D1 but lower contents at D14, D21, and D28 than the FG chickens (p < 0.05).The triglyceride level decreased from 1061 or 462 mg/L at D1 to 266 or 135 mg/L at D7 in the FG and SG strains, respectively (p < 0.001; Figure 1) and then remained stable until D42. The FG strain had higher values at D1, D7, and D14 than the SG strain (p < 0.05), with the difference between the strains being particularly high at D1 when the FG chickens had 3-fold higher plasma triglyceride levels than the SG chickens.The FFA level increased regularly from 217 µmol/L at D1 to 877 µmol/L at D21 and then remained stable in the FG strain (p < 0.01; Figure 1). In the SG strain, the FFA content increased from 96 µmol/L at D1 to 643 µmol/L at D7, followed by a decrease to 437 µmol/L at D14 and a further increase to 664 µmol/L at D42 (p < 0.01). Higher values in the FFA levels were observed in the plasma of FG chickens at all ages except for D7 and D14 (p < 0.05).Finally, the β-OH level increased from 450 or 368 µmol/L at D1 to 737 or 821 µmol/L at D14 in the FG and SG strains, respectively (p < 0.05; Figure 1). Then, it remained stable except in the FG chickens for which we observed a strong decrease between D28 and D35. The SG strain had a higher β-OH content than the FG strain at this age (817 vs. 314 µmol/L, p < 0.001).3.3. Plasma Redox Status and Enzyme ActivitiesThe TBARS value was not affected by age and strain, except at D28 when the FG strain showed a higher value than the SG strain (p < 0.05; Figure 2).The TAS value was differently affected by age according to the strain. For the FG strain, it increased from 1.04 mmol/L at D1 to 1.31 mmol/L at D28, then decreased to 0.93 mmol/L at D35 and increased again to 1.25 mmol/L at D42 (p < 0.05; Figure 2). In contrast, the TAS value decreased from 1.48 mmol/L at D1 to 0.88 mmol/L at D21, then remained stable until D42 in the SG strain. Regarding the strain effect, a lower TAS value was observed at D1 but higher values were observed at D21 and D28 in the FG chickens compared to SG chickens (p < 0.05).The SOD activity increased regularly from 16.48 U/mL at D1 to 33.14 U/mL at D35, then remained stable in the FG strain (p < 0.01; Figure 2). However, it remained stable around 29–34 U/mL during the whole rearing period in the SG strain. The SOD activity remained lower from D1 to D21 in the FG strain before reaching similar levels than the SG strain (p < 0.01).The evolution of the GPx activity with age is quite complex. It decreased between D1 and D7 from 8932 or 9198 U/L to 8226 or 8577 U/L in the FG and SG strains, respectively (p < 0.01; Figure 2). Then, it increased to 9417 or 9237 U/L at D14 in the FG and SG strains, respectively (p < 0.01), and remained stable in the SG strain while it decreased to 8205 U/L at D42 in the FG strain. There was no main effect of strain on this parameter (p > 0.05), even though at D42, the FG strain had a lower activity than the SG strain.The GR activity decreased between D1 and D7 from 0.042 or 0.043 U/L to 0.028 or 0.025 U/L in the FG and SG strains, respectively (p < 0.001; Figure 2). Then, it remained stable in the SG strain while it increased to 0.048 U/L between D28 and D42 in the FG strain. As a consequence, GR activity was higher in the FG strain than in the SG strain at D35 and D42 (p < 0.05).The peroxidase activity decreased regularly during the first two weeks of age from 3.68 or 4.689 mU/mL at D1 to 2.22 or 2.56 mU/mL at D14 in the FG and SG strains, respectively (p < 0.001; Figure 2). Then, it remained stable until D28 before increasing to 2.25 or 2.24 mU/mL at D35 and decreased to 1.14 or 1.37 mU/mL at D42 in the FG and SG strains, respectively (p < 0.001). The strain had no effect on this parameter (p > 0.05).The hydroperoxide level decreased sharply from 11.67 or 16.48 µmol/mL at D1 to 2.22 or 2.56 µmol/mL at D14 for the FG and SG strains (p < 0.001; Figure 3). Then, it slightly increased at D21 to 5.19 or 4.62 µmol/mL and remained stable until D42 in the FG and SG strains, respectively (p < 0.001). The strain had no effect on this parameter (p > 0.05).The α-tocopherol level sharply decreased during the two first weeks of age, from 53.99 or 34.15 ng/µL at D1 to 6.98 or 6.12 ng/µL at D14 in the FG and SG strains, respectively (p < 0.001). Then, it remained stable until D42 in the two strains. The α-tocopherol content at D1 was higher in the FG than in the SG strain (p < 0.05).For the FG strain, the γ-tocopherol level decreased from 6.41 ng/µL at D1 to 2.06 ng/µL at D7 (p < 0.05; Figure 3). Then, it increased to 4.31 ng/µL at D14 and remained stable until D42. In the SG strain, the γ-tocopherol content decreased from 3.27 ng/µL at D1 to 1.69 ng/µL at D21 and then remained stable until D42 (p < 0.05). Differences between strains were observed at some ages, with higher values of γ-tocopherol observed in the FG strain than in the SG strain at D1, D14, and D35 (p < 0.05).3.4. Muscles and Liver Lipid Peroxidation StatusIn the liver, the TBARS value increased during the first week of age from 1.17 or 1.88 mg equivalent MDA/kg at D1 to 2.77 or 3.08 mg equivalent MDA/kg at D7 in the FG and SG strains, respectively (p < 0.001; Figure 4). Then, it decreased with age until reaching 1.53 or 1.45 mg equivalent MDA/kg at D42 in the FG and SG strains, respectively. Lower and higher values were observed in the FG chickens at D1 and D14, respectively, compared with the SG chickens.In the breast muscle, the TBARS value sharply decreased during the first week, from 0.68 or 1.03 mg equivalent MDA/kg at D1 to 0.20 or 0.21 mg equivalent MDA/kg at D7 in the FG and SG strains, respectively (p < 0.001; Figure 4). Then, it remained stable in the SG strain until D42. In the FG strain, we observed a slight increase to 0.36 mg equivalent MDA/kg at D14 followed by a decrease between D21 and D28 and a stabilization thereafter. The FG strain had a lower value at D1 and D35 and a higher value at D21 than the SG strain (p < 0.05).In the thigh muscles, the TBARS value decreased from 0.51 mg equivalent MDA/kg at D1 to 0.34 mg equivalent MDA/kg at D7 only in the FG strain (p < 0.05; Figure 4). Then, it remained stable until D42. In the SG strain, the TBARS value decreased later (between D7 and D14) from 1.02 mg equivalent MDA/kg to 0.65 mg equivalent MDA/kg (p < 0.001), then remained stable until D42. TBARS was much greater in the FG strain than in the SG strain between D1 and D21 (p < 0.01) before reaching similar values.3.5. Muscles and Liver Antioxidant Status and Enzyme ActivitiesIn the liver, the TAS value increased between D14 and D35, from 10.04 mmol/µg proteins to 18.56 mmol/µg proteins in the SG strain (p < 0.001; Figure 4). In the FG strain, the TAS value also increased between D14 and D28, from 8.99 mmol/µg proteins to 16.79 mmol/µg proteins, but decreased to 14.35 mmol/µg proteins at D35 and increased again at 16.54 mmol/µg proteins at D42 (p < 0.001). The FG strain had lower TAS values than the SG strain at D1, D7, and D35 (p < 0.05).In the breast muscle, the TAS value increased from 0.296 mmol/µg proteins at D1 to 0.814 mmol/µg proteins at D42 in the SG strain (p < 0.001). In the FG strain, the TAS value increased from 0.457 mmol/µg proteins at D1 to 0.649 mmol/µg proteins at D21, but like in the liver, decreased to 0.487 mmol/µg proteins at D35 and increased again to 0.533 mmol/µg proteins at D42 (p < 0.001; Figure 4). The FG strain had higher values at D1 and D21 and lower values at D35 and D42 than the SG strain (p < 0.05).In the thigh muscles, the TAS value decreased regularly with age from 0.550 mmol/µg proteins at D1 to 0.390 mmol/µg proteins at D42 in the SG strain (p < 0.001; Figure 4). In the FG strain, it increased from 0.482 mmol/µg proteins to 0.738 mmol/µg proteins at D7, then remained stable until D35 and decreased to 0.655 mmol/µg proteins at D42 (p < 0.001). From D7 to D42, the FG strain had much higher TAS values than the SG strain (p < 0.001).In the liver, the GPx activity decreased from 4748 U/µg proteins at D1 to 4139 U/µg proteins at D7 and then remained stable until D42 in the FG strain (p < 0.01; Figure 5). It strongly decreased from 4612 U/µg proteins at D1 to 1712 U/µg proteins at D14, then increased to 4796 U/µg proteins until D28 and remained stable thereafter (p < 0.01) in the SG strain. The FG strain had higher values at D7 and D14, but a lower one at D28 than the SG strain (p < 0.05).In the breast muscle, the GPx activity decreased from 1630 U/µg proteins at D1 to 1257 U/µg proteins at D14 in the FG strain (p < 0.01; Figure 5). It then remained quite stable until D35 before increasing at 1467 U/µg proteins at D42 (p < 0.01). In the SG strain, the GPx activity remained stable between D1 and D7, then decreased from 1438 U/µg proteins at D7 to 1268 U/µg proteins at D14. It increased to 1421 U/µg proteins at D28 and remained stable until D42 (p < 0.01). The FG strain had a higher value at D1 (p < 0.0001) and a lower value at D28 (p < 0.05) than the SG strain.In the thigh muscles, the GPx activity decreased from 1557 or 1457 U/µg proteins at D1 to 1185 or 1206 U/µg proteins at D14 in the FG and SG strains, respectively (p < 0.001; Figure 5). Then, it increased regularly with age until 1357 or 1341 U/µg proteins at D42 for the FG and SG strains, respectively (p < 0.001). Like in the breast muscle, the FG strain had a higher value of GPx activity at D1 and a lower value at D28 (p < 0.05) than the SG strain.In the liver, the evolution of the GR activity was similar between the two strains (Figure 5). It gradually increased from 0.796 or 0.717 U/µg proteins at D1 to 1.133 or 1.143 U/µg proteins for FG and SG, respectively, then remained quite stable until D42. The GR activity was higher at D28 in the SG strain compared with the FG strain (p < 0.05).In the breast muscle, the GR activity decreased from 0.216 U/µg proteins at D1 to 0.127 U/µg proteins at D7 in the FG strain and from 0.157 U/µg proteins at D1 to 0.100 U/µg proteins at D14 in the SG strain, then remained quite stable until D42 for both strains (Figure 5). There was a strain effect on this parameter, with higher GR activity values observed in the FG strain at D1 and between D21 and D42 (p < 0.05).In the thigh muscles, the GR activity evolved similarly than in the breast muscle (Figure 5). It decreased from 0.156 U/µg proteins at D1 to 0.108 U/µg proteins at D7 in the FG strain and from 0.140 U/µg proteins at D1 to 0.090 U/µg proteins at D14 in the SG strain, then remained quite stable until D42 in both strains. For the whole rearing period, the FG strain had higher values for GR activity than the SG strain (p < 0.01).In the liver, the SOD activity was stable from D1 to D28 in the FG strain (p > 0.05, Figure 6), then increased from 126 to 268 U/µg proteins at D35 and decreased again to 159 U/µg proteins at D42 (p < 0.05). In the SG strain, the SOD activity increased from 679 U/µg proteins at D1 to 961 U/µg proteins at D14 (p < 0.001), then sharply decreased to 372 U/µg proteins at D21 and to 158 U/µg proteins at D42 (p < 0.001). The liver SOD activity was much lower in the FG strain than in the SG strain between D1 and D14 (p < 0.05).In the breast muscle, the SOD activity increased from 20.41 or 19.30 U/µg proteins at D1 to 46.24 or 44.74 U/µg proteins at D7 in the FG and SG strains, respectively (p < 0.01, Figure 6). It then remained stable until D42 in the SG strain and decreased regularly with age to 27.25 U/µg proteins at D42 in the FG strain (p < 0.01). The FG strain had lower values than the SG strain from D14 to D42 (p < 0.01).In the thigh muscles, the SOD activity increased from 4.15 U/µg proteins at D1 to 10.48 U/µg proteins at D7 in the FG strain (p < 0.001, Figure 6). Then, it remained stable until D21, decreased to 6.07 U/µg proteins at D28, and remained stable until D42 (p < 0.001). In the SG strain, the SOD activity strongly increased from 6.54 U/µg proteins at D1 to 28.30 U/µg proteins at D21 and then remained stable until D42 (p < 0.001). The FG strain had much lower values than the SG strain from D7 to D42 (p < 0.001).In the liver, the catalase activity was not affected by age in the SG strain (p > 0.05). In the FG strain, decreased from 71.06 U/µg proteins at D1 to 44.23 U/µg proteins at D7 (p < 0.05) and then remained stable until D42. The FG strain had a higher catalase activity between D1 and D14 and a lower value at D21 than the SG strain (p < 0.05).In the breast muscle, the catalase activity strongly decreased from 9.50 or 9.55 U/µg proteins at D1 to 0.929 or 1.257 U/µg proteins at D7 in the FG and SG strains, respectively (p < 0.001, Figure 6). Then, it remained stable until D42 in the two strains. The strain had no effect on this parameter (p > 0.05).In the thigh muscles, the catalase activity decreased from 4.24 U/µg proteins at D1 to 2.79 U/µg proteins at D28 and then remained stable until D42 in the SG strain (p < 0.001, Figure 6). In the FG strain, it decreased from 7.81 U/µg proteins at D1 to 1.27 U/µg proteins at D28 and then remained stable until D42 (p < 0.001). The FG strain had higher values at D1 and D7 and lower values at D21 and D28 than the SG strain (p < 0.05).4. Discussion4.1. Effect of Age on the Markers of Redox Status and Plasma Metabolites of ChickensFor each sampling date, the chicks had been fasted for 8 h before sacrifice. The metabolites measured in plasma therefore reflected the basal metabolism of the animals. At D1, the plasma concentration in triglycerides was high but decreased rapidly during the first week where an increase in plasma FFA levels was observed. This can reflect the use by embryo and hatched chicks of lipids stored in the egg yolk and then in their liver as a major source of energy [16]. Moreover, β-hydroxybutyrate (β-OH), one of the main ketone bodies produced during the breakdown of fats in the body [17], strongly increased between D1 and D14. Moran et al. [18] and Uni and Ferket [19] have reported a high level of lipid peroxidation in the early stage after chick hatch. Lipid peroxidation, assessed through the TBARS index, was indeed maximal at D1 in both the breast and thigh muscles and D7 in the liver. The plasma concentration of hydroperoxides was also maximal at D1, then decreased until D14, confirming the high level of oxidative metabolism just after hatch. Consistently, the plasma levels of several enzymes or molecules involved in the antioxidant response declined within one to two weeks after hatching. This is the case for peroxidase, whose role is to break down toxic peroxides produced by oxidative stress [20], but also glutathione peroxidase (GPx) and reductase (GR) and α- and γ-tocopherol.There are few data in the literature on the post-hatching evolution of poultry defenses against oxidative stress, and they are sometimes inconsistent. Yang et al. [9] measured the serum glutathione content, glutathione peroxidase activity, and total antioxidant in the serum of fasting chickens for 12 h at 14, 21, and 28 days of age but failed to show any significant variations with age. However, between 3 and 6 weeks of age, the hepatic glutathione (GSH) concentration decreased and the catalase activity increased while the TBARS value and glutathione peroxidase activity were unchanged [12]. In our study, the catalase activity remained quite stable in the liver after 7 days. By measuring the various forms of glutathione (oxidized and reduced, GSSG and GSH) and activities of the glutathione peroxidase and reductase in chicken blood, Mahmoud and Edens [13] showed that GSH concentration and GPx activity decreased between 2 and 4 weeks while GR activity increased. Again, these last results differ from ours, since we observed stable GR and GPx activities between 2 and 4 weeks in both strains. Such discrepancies between studies highlight the fact that the evolution of antioxidant status in chickens is probably influenced by many factors related to animal genetics, nutrition, or the management of farming conditions.From our results, it is interesting to note that the high oxidative metabolism measured in chicks the first week after hatching does not imply a decrease in the total antioxidant status either in plasma but also in the liver and muscles. This could be explained by the use of vitamin E (tocopherol), a strong antioxidant molecule whose concentration in plasma strongly decreased during the first week post-hatching. This is consistent with previous results that showed a sharp decrease in hepatic vitamin E concentration in different avian species (chicken, turkey, duck, and goose) during the first two weeks of post-hatch growth [11].Just after hatch, the main enzymes acting as antioxidants seem to be the peroxidase and glutathione reductase, whose activities decreased thereafter in plasma but also in the liver and muscles. Catalase activity was also maximal one day after hatch in liver and muscles, also suggesting a major role of this enzyme in maintaining the redox balance of chicks at this stage. In the breast muscle, the catalase activity decreased 10-fold during the first week and remained stable at low-levels thereafter, suggesting a minor role of this enzyme in the antioxidant response of this tissue in the later stages. SOD then appears to take over the antioxidant defense of chickens at the plasma and tissue levels in both strains.4.2. Effect of Strain on the Markers of Redox Status and Plasma Metabolites of ChickensFrom D1 to D14, the FG strain had a higher plasma triglyceride content than the SG strain. At D1 and after D21, the plasma concentration in FFA was lower for the SG strain compared to the HG strain, suggesting a lower lipolysis activity or a higher lipid oxidation. Indeed, at D1 in the liver and breast muscle, and from D1 until D21 in the thigh muscles, the lipid peroxidation evaluated with the TBARS value was higher for the SG chicks compared to the FG chicks. This would be more related to the energetic metabolism of SG chicken muscles, which have a higher oxidative activity and a higher iron content (myoglobin and hemoglobin) in muscles than FG chickens because of their greater physical activity [21,22]. Castellini et al. [23] also reported higher lipid peroxidation estimated with the TBARS value in the leg muscles of chickens with a slow growth rate than in the leg muscles of chickens with a high growth rate. However, in the plasma, the strain had no effect on the lipid peroxidation, the hydroperoxide concentration, and the peroxidase activity. Moreover, at D1, the plasma concentration in uric acid was higher in SG chicks compared to FG chicks, suggesting a higher catabolism of proteins. The chicks of the two strains were received on the same day, but it is possible that those of the SG strain hatched earlier than those of the FG strain, hence a greater exhaustion of their energy reserves and the obligation to increase their protein catabolism. This could also explain the difference at D1 between strains on the α- and γ-tocopherol contents in plasma. Another possibility is a different supplementation with tocopherol in the diets of breeders producing SG and FG chicks. At D14, D21, and D28, the FG strain had a higher plasma concentration in uric acid than the SG strain, suggesting a higher protein turnover, probably to allow for the high development of breast muscles. The strain had few effects on the GPx activity in the plasma and muscles. In the liver, this activity was lower at D7 and D14 and higher at D28 for the SG strain compared to the FG strain [24]. The strain had few effects on the GR activity in the liver and thigh muscles. However, the FG strain had a higher GR activity than the SG strain in the breast muscle and plasma at D35 and D42. The FG strain had a lower SOD activity in the plasma from D1 to D21, in the liver from D1 to D14, in the thigh muscle from D7 to D42, and in the breast muscle from D14 to D42 than the SG strain, which had a higher locomotor activity and higher oxidative metabolism in muscles. Whatever the tissue considered, the SOD activity seemed to differentiate the two strains. For the catalase activity, the FG strain had higher values than the SG strain in the liver from D1 to D14 and in the thigh muscles at D1 and D7 whereas the strain had no effect on the catalase activity in the breast muscle. The SG strain had a higher TAS value than the FG strain at D1 in the plasma, at D1, D7, and D35 in the liver, and at D35 and D42 in the breast muscle, whereas the FG strain had a higher TAS value than the SG strain at D1 and D21 in the breast muscle, and from D7 to D42 in the thigh muscles, which may be in relation to a higher locomotor activity and a higher oxidative metabolism in SG chicks. In the plasma, the evolution with age of TAS value for each strain seemed to be related to that of uric acid content, this metabolite also having antioxidant properties [25]. The activity of enzymes implicated in the antioxidant defense is different according to the strain and the tissue considered. Finally, the strain had few effects on the plasma concentration in glucose and β-OH.Differences observed between strains suggest that fast-growing broilers might use different pathways than slow-growing birds to maintain their redox balance during growth. Catalase, GPx, and GR would be preferentially mobilized just after hatching in the liver or muscles in both strains, whereas GR in the pectoral muscle and uric acid in plasma would take over later in development to provide antioxidant defense in FG birds, as also suggested by Machin et al. [26]. In contrast, the SOD pathway, whose activity remained at high levels over time in the blood and tissues, could play a major role in the SG chickens. Finally, the increase in plasma uric acid content specifically observed between hatch and three weeks of age in the FG broilers may also reflect their high protein turnover, allowing them to maintain their high growth rate. By exploring oxidative phosphorylation in breast muscle, Hubert and Athrey [27] found that it was significantly reduced in FG broilers compared to SG chickens.5. ConclusionsOur results confirm the high level of oxidative stress just after hatching in chicks, and consequently, a high need for antioxidant defense during this period. Several enzymes or molecules, among which vitamin E, catalase, peroxidase, glutathione peroxidase, and reductase, whose levels drop significantly in the first one to two weeks post-hatching, would ensure this. Interestingly, our study revealed that during the first stage of growth, slow-growing chickens had a lower energy status linked to stronger oxidative metabolism and antioxidant response compared to fast-growing chickens. During growth, the antioxidant pathways used to maintain redox balance in chickens change, and may be different between slow and fast-growing genotypes. SOD activity seems to be a key player in the antioxidant response of slow-growing chickens, solicited first at the blood and hepatic levels and then at the muscle level, while uric acid could play a more important role in the antioxidant response of fast-growing chickens, especially in the late stages of development. The detailed description of the evolution of metabolic status and antioxidant defenses during the growth of two genotypes of chicken with slow and fast growth can be useful for the research of solutions adapted to the various contexts of production to maintain a good redox balance, and thus good health throughout the life of the animal.

animals : an open access journal from mdpi

10.3390/ani11061822

PMC8234949

Supplemental nutrition for cattle is the greatest operating cost for cow-calf producers, accounting for 65% of the annual expenses. In addition, residual feed intake (RFI) is being used as a selection tool for purchasing and retaining heifers, as well as selecting bulls with the goal of improving feed efficiency and/or reducing supplemental inputs. However, the use and relevance of RFI as a selection tool for the cow-calf industry needs additional research. In our studies, heifer post-weaning RFI did not influence mature cow dry matter intake and intake behavior for both lactating and non-lactating beef cows. In contrast, cow age did correspond to increases of intake and intake rates of mature cows. However, when intake was expressed as g/kg body weight−1, no differences were observed with respect to cow age for lactating and non-lactating cows. Milk production was influenced by heifer post-weaning RFI in 5–6- and 8–9-year-old cows, however, did not influence 9–10-year-old cows. Therefore, our research suggests that cow age has greater impacts on dry matter intake than RFI, however, the relationship between RFI of heifers and subsequent mature cow milk production warrants further investigation.

We evaluated heifer post-weaning residual feed intake (RFI) classification and cow age on dry matter intake (DMI) at two stages of production. Fifty-nine non-lactating, pregnant, (Study 1) and fifty-four lactating, non-pregnant (Study 2) commercial black Angus beef cows were grouped by age and RFI. Free-choice, hay pellets were fed in a GrowSafe feeding system. In Study 1, cow DMI (kg/d) and intake rate (g/min) displayed a cow age effect (p < 0.01) with an increase in DMI and intake rate with increasing cow age. In Study 2, cow DMI (kg/d) and intake rate (g/min) displayed a cow age effect (p < 0.02) with an increase in DMI and intake rate with increasing cow age. Milk production displayed a cow age × RFI interaction (p < 0.01) where both 5–6-year-old and 8–9-year-old low RFI cows produced more milk than high RFI cows. For both studies, intake and intake behavior were not influenced by RFI (p ≥ 0.16) or cow age × RFI interaction (p ≥ 0.21). In summary, heifer’s post-weaning RFI had minimal effects on beef cattle DMI or intake behavior, however, some differences were observed in milk production.

1. IntroductionSupplemental nutrition for cattle is the greatest operating cost for cow-calf producers, accounting for 65% of the annual expenses [1,2,3]. Traditionally, selection pressure has been placed on production traits associated with increasing outputs (average daily gain, weaning weight, yearling weight, etc.), which can also result in increased inputs to achieve animal production potential. Since feed costs constitute the greatest proportion of total inputs, selection pressure for efficient animals that have lower feed intake but maintain production, or average intake with higher production, could have a great impact on cow-calf profitability [3]. Thus, improving feed efficiency through genetic selection holds significant opportunity for the beef industry.Residual feed intake (RFI) is currently being used as a selection tool for purchasing and retaining heifers and for selecting bulls. Most studies have used steers and terminal heifers when evaluating RFI impact on various aspects of beef cattle production [4,5,6]. However, the use and relevance of RFI as a selection tool for the cow-calf industry needs additional research [7,8,9]. Most RFI studies have included energy-dense diets and rations focusing on feedlot performance [4,6,10]. Research pertaining to RFI of cattle offered forage-based diets is limited [11], with even less data available related to beef cattle forage-based production systems [3,7,12,13]. As a result, more research is needed to evaluate the utility of RFI estimates on beef cattle production systems in extensive forage-based environments [8,12,14].The use of RFI for beef cattle selection in rangeland environments is based on the assumption that heifer post-weaning RFI will be expressed throughout the lifetime productivity of that heifer [9]. Past research suggests that RFI is a moderately heritable trait [15], however, few published studies exist comparing RFI of individuals of two different physiological states [5,6,16]. Multiple studies have investigated feed efficiency in young, pre-mature cattle, but few have quantified how RFI relates to feed efficiency at different ages and stages of production [16,17,18]. The repeatability of RFI between growing and finishing phases of beef cattle production has recently been examined [6,19], but there are few reports demonstrating the relationship of RFI of growing females compared with those same females as mature, lactating [17,20], and non-pregnant, non-lactating beef cows [18].Research evaluating heifer post-weaning RFI on subsequent cow intake and intake efficiency is limited. Therefore, the objectives of these research studies were to: 1) evaluate the effects of heifer post-weaning RFI on non-lactating, pregnant, beef cows dry matter intake (DMI) at different cow ages; and 2) evaluate the effects of heifer post-weaning RFI on lactating, non-pregnant, beef cow DMI at different cow ages, as well as peak milk production. We hypothesized that heifers identified as low RFI eat less than high RFI cows and the influence of RFI may interact with cow age.2. Materials and MethodsThe use of animals in this study was approved by the Agricultural Animal Care and Use Committee of Montana State University AACUC #2018-AA12. These studies were conducted at the Northern Agricultural Research Center (NARC), located in Havre, Montana. All cattle were synchronized and time artificially inseminated (AI) corresponding to initial calving date of March 15, with exposure to cleanup bulls occurring approximately 7 days post AI for an additional 45 days. Calves were strategically weaned mid-September to mid-October each year with the timing based on fall forage conditions and/or cow body condition.All cows were managed as one contemporary group post-weaning in the fall to calving in the spring of each year. All heifers/cows were exposed to electronic feed bunks as heifers and again during winter supplement studies.2.1. Heifer RFI TrialsStarting in 2011, all Northern Agricultural Research Center cattle have been utilized in a heifer RFI trial for a minimum of 77 days on a forage-based ration provided in a GrowSafe system (GrowSafe DAQ 4000E; GrowSafe System Ltd., Airdrie, AB, Canada). Calves were weaned on pasture mid-September to early October each year and entered an RFI trial 60 to 75 days post weaning. On average, 45–95 AI sired replacement heifers were retained annually. Upon RFI trial initiation, and completion, all heifers were weighed post feeding, on two consecutive days to record beginning and ending body weights (BW), then again, every 28 days to record BW gain. A 7-d acclimation period preceded a 70-d feeding trial, while the GrowSafe system recorded individual daily feed intakes. All heifers had free access to 16–32 GrowSafe feed bunks (depending on year and number of heifers) and ad libitum access to water and forage-based diets, consisting of 30.4% corn silage, 41.1% grass hay, and 28.5% alfalfa on a dry mater basis, formulated to meet requirements for growing moderate frame beef heifers (10.5% CP and 66.0% TDN [21]). Individual heifer post-weaning RFI was calculated following parameters set forth by Archer et al. [22] and Arthur et al. [4]:RFIPhe = FI-β w(phe) × MWT-βg(phe) × DG,(1) where RFIphe = phenotypic residual feed intake, FI = daily feed intake, MWT = metabolic body weight at mid-test, DG = average daily gain, and βw(phe) and βg(phe) = partial regression coefficients of animal’s FI on MWT and DG, respectively. Heifers were classified as either low (>−0.50 SD from mean), or high (<+0.50 SD from the mean) within a year [9].Two feed intake studies were conducted to evaluate the impacts of heifer post-weaning RFI at differing ages and stages of beef cattle production. Individual BW and body condition scores (BCS) were recorded and cattle were placed in a GrowSafe feeding system for DMI analysis over a 21-d period. For each study, treatments were replicated in two pens, each containing 16 GrowSafe feeding units. Individual animal intake was continuously recorded. The system was monitored daily for unaccounted feed balance. When greater than 5% of the feed disappearance was unaccounted for, the GrowSafe system automatically deemed the 24-h period as failed. Therefore, we selected the last 7 days of the 21-d period that met the criteria to calculate average DMI per individual animal for each DMI study. Variation in dry matter intake (kg/d), measured as the coefficient of variation (% CV), was based on daily intake estimates for individual animals.2.2. Study 1: Non-Lactating, Pregnant CowFifty-nine non-lactating, pregnant, black Angus females were utilized to evaluate the impacts of heifer post-weaning RFI on DMI post weaning. All cows were fed commercially available free-choice alfalfa/straw pellet formulated to meet nutrient requirements for non-lactating, pregnant cows (Table 1; CHS Nutrition, Sioux Falls, SD, USA; [21]). At the initiation of the trial (14-d post weaning), cows were classified by age, (1–2, 4–5, and 7–8 years old;) and within age class represented both low and high RFI, and dry-lotted for 16 h to obtain uniform shrunk BW and BCS.2.3. Study 2: Lactating, Non-Pregnant CowFifty-four lactating, non-pregnant, black Angus females were utilized to evaluate the impacts of heifer post-weaning RFI (low and high) and cow age (2–3-, 5–6-, and 8–9-year-old cows) and dry-lotted for 16 h to obtain uniform shrunk BW and BCS. Cows were selected by the same age and RFI criteria as described for Study 1. However, for this study, we only utilized cows that calved within the first 42 days of the calving period. Calf BW and Julian birth date were measured to characterize the influence of cow age and RFI. All cows were fed commercially available free-choice alfalfa/grass base pellet formulated to meet NRC requirements for lactating cows for the DMI study (Table 1; CHS Nutrition, Sioux Falls, SD, USA; [21]). At the conclusion of the 7-d DMI period, approximately day 60 post calving, a weigh-suckle-weigh trial was conducted to evaluate the impacts of heifer post-weaning RFI and cow age on milk production following methods detailed by Williams et al. [23]. For our study, we utilized an 8-h calf removal protocol.2.4. Statistical AnalysisFor both Studies 1 (Supplementary Materials Table S1) and 2 (Supplementary Materials Table S2), the influence of RFI and cow age on initial cow BCS and BW were analyzed using ANOVA with a generalized linear model including RFI, cow age and the interactions of RFI and cow age as fixed effects. Additionally, the influence of RFI and cow age on intake and intake behavior were analyzed using ANOVA with a generalized linear mixed model including RFI, cow age, and the interactions of RFI and cow age as fixed effects, and individual cow and pen as random effects. When RFI × cow age interactions were observed means were separated within age groups. Individual animal was considered the experimental unit and an alpha ≤ 0.05 was considered significant. Orthogonal polynomial contrasts were used to determine linear and quadratic effects for cow age. Means were separated using the Tukey method when p < 0.05. Tendencies were reported when significance was p ≤ 0.10. All statistical analyses were performed in R [24].3. Results3.1. Study 1: Non-Lactating, Pregnant CowCow BW displayed a cow age × RFI interaction (p < 0.01) where 4–5-year-old low RFI cows had a lighter BW than high RFI cows, however, low RFI 7–8-year-old cows tended (p = 0.06) to have greater BW than high RFI cows (Table 2). No differences in BW were observed between RFI classifications in 1–2-year-old cows (p = 0.17). Cow BCS also displayed a cow age × RFI interaction (p = 0.02) where low RFI 4–5-year-old cows had lower BCS than high RFI 4–5-year-old cows with no differences observed between RFI in other cow ages (p ≥ 0.24). Both DMI (kg/d) and intake rate (g/min) displayed a cow age effect (p < 0.01) with a quadratic increase with increasing cow age (p ≤ 0.02; Table 2). Specifically, young cows (1–2-year-old) ate less and had lower intake rates than older cows (p < 0.01). Neither DMI (g/kg of BW), intake variation (% CV), or time spent at the feeder (min/d) were affected by cow age or RFI (p ≥ 0.16), averaging 28.8 g/kg of BW, 19.0% CV, and 107.5 min/d, respectively.3.2. Study 2: Lactating, Non-Pregnant CowCow BW displayed a cow age × RFI interaction (p < 0.01), with low RFI 5–6- and 8–9-year-old cows having a greater BW than high RFI 5–6- and 8–9-year-old cows (p ≤ 0.01; Table 3). Similar to cow BW, cow BCS displayed a cow age × RFI interaction (p < 0.01) where low RFI 5–6-year-old cows had lower BCS than high RFI 5–6-year-old cows (p < 0.01), whereas, low RFI 8–9-year-old cows had higher BCS than low RFI 8–9-year-old cows (p < 0.01).Calf BW, measured at weigh-suckle-weigh, displayed a cow age × RFI interaction (p <0.03), where calf BW from 2–3-year-old cows were lower in high RFI 2–3-year-old cows compared to low RFI cows (p = 0.02; Table 3). Calf Julian birth date was affected by cow age (p < 0.01) with calf birth date increasing linearly with increasing cow age (p < 0.01; Table 3). There was also a tendency for an RFI effect on calf Julian birth date (p = 0.09) with low RFI cows tending to calve later in the calving season than high RFI cows, averaging 72.2 and 73.9 for high and low RFI, respectively.Cow DMI (kg/d) displayed a cow age effect (p < 0.01) with a quadratic increase in DMI with increasing cow age (p < 0.01; Table 3). Similarly, intake rate (g/min) displayed a cow age effect (p < 0.02) with a linear increase (p < 0.01) in intake rate with increasing cow age. In contrast, neither DMI intake (g/kg of BW), % CV, nor time spent at the feeder (min/d) were affected by cow age, or RFI, averaging 21.7 g/kg of BW, 11.7% CV, and 149.7 min/d, respectively.Cow milk production (kg) displayed a cow age × RFI interaction (p < 0.01; Table 3), with 2–3-year-old and 5–6-year-old low RFI cows producing more milk than high RFI cows of the same age (p < 0.01). In addition, cow milk production (g/kg of BW), displayed a cow age × RFI interaction (p < 0.01) with 2–3-year-old and 5–6-year old low RFI cows producing more milk per kg of BW than high RFI cows of the same age (p < 0.01), however, no differences were observed in 8–9-year-old cows (p = 0.48).4. DiscussionIt has been reported that as cattle grow and mature, composition of their gain changes from protein accretion to fat deposition [25]. Since the expense of protein accrual is less than for fat deposition, the efficiency that cattle convert feed into BW gain is reduced as they mature [26]. Multiple research papers have reported on the changes in feed efficiency (RFI) of cattle at different stages of physiological growth. Durunna et al. [19] reported that following two consecutive 70-d RFI periods, 49% of the heifers maintained their original RFI classification, whereas 51% had a different RFI classification, indicating re-ranking exists in heifers despite receiving the same basal diet. Loyd et al. [16] suggested that RFI determined during the pre-pubertal period may only be a moderate predictor of post-pubertal RFI.Archer et al. [22] reported a moderate correlation of 0.40 between RFI measured in heifers post-weaning and later as non-gestating, non-lactating 3-year-old cows. Freetly et al. [18] compared the RFI classification of yearling heifers following an 84-d RFI intake trial with subsequent RFI classification of 5-year-old non-pregnant, non-lactating cows, and reported that feed intake and ADG are heritable and genetically correlated between heifers and cows. Black et al. [17] compared the RFI classification of growing heifers following a 70-d RFI trial and subsequently as 3-year-old lactating beef cows and reported heifers that were the most feed efficient consumed less feed as lactating cows while maintaining similar performance. However, they reported that correlations between heifer and mature cow RFI values were not significant indicating that within-animal feed efficiency was not maintained as the calves developed from growing heifers to mature, lactating cows.Results from our research are in general agreement with previous research where intake tends to change between ages and physiological stages of production of female beef cows, indicating that heifer post-weaning RFI may not be a reliable predictor of mature cow feed intake. In contrast, cow age had a substantial impact on intake, and intake behavior. Previous research reported that cow age significantly impacted grazing behavior and terrain use when comparing older cows to younger cows [9,27,28]. Furthermore, although limited, previous studies evaluating supplement intake of mixed-age beef herds have reported that younger cows spent less time at feeders and consumed less feed than older cows [29,30]. In contrast, Wyffels et al. [28] reported younger cattle consumed more and visited the feeders more often than older cows. Parsons et al. [13] observed a quadratic effect in intake related to cow age where 1-year-old cows consumed more supplement and had a larger CV of intake than older cows.Broleze et al. [20] states that measuring RFI in lactating animals is important, however, RFI does not accurately reflect production efficiency. This is because the RFI models that are used to calculate residual traits for lactating cows do not account for the energy partitioning into the various components, some of which are more financially important (e.g., milk fat and protein yield) than others (metabolic BW; [31]). Previous research found no relationship between milk yield (obtained by weigh-suckle-weigh technique) and RFI, with low RFI and high RFI cows having similar milk yield [10,17,32]. Rutledge et al. [33] reported that dams nursing female calves produced significantly more milk than those nursing male calves. In contrast, Melton et al. [34] and Christian et al. [35] found no significant difference in dam’s milk yield attributable to sex of calf while Pope et al. [36] reported an advantage for cows nursing male calves.In this study, cow milk production (g/kg of BW−1) was impacted by both cow age and RFI with 3–4- and 6–7-year-old low RFI cows producing more milk than high RFI cows of the same age. Regardless of unit of measurement, the difference in milk production (kg vs. g/kg of BW) were consistent across young and middle-age cows. This suggests that heifer post-weaning RFI may be related to milk production of young and middle-aged cows. Few research trials have compared low RFI to high RFI lactating, non-pregnant beef cows [10,17,37]. Previous research report that low RFI (more efficient) and high RFI (less efficient) cows produce similar quantities of milk, but the former consumed less dry matter per day [17,32,38]. To our knowledge, this is the first study comparing post-weaning heifer RFI to subsequent milk production of beef cows across multiple age classes.5. ConclusionsHeifer post-weaning RFI did not influence mature cow dry matter intake, and this was consistent for both lactating and non-lactating beef cows. In contrast, cow age did correspond to quadratic increases of DMI and intake rates of mature cows. However, when DMI was expressed as g/kg of BW no differences were observed with respect to cow age in lactating and non-lactating cows. Milk production was influenced by heifer post-weaning RFI for 3–4- and 6–7-year-old cows, however, did not influence 9–10-year-old cows. Therefore, our research suggests that cow age has greater impacts on dry matter intake than RFI, however, the relationship between RFI of heifers and subsequent mature cow milk production warrants further investigation.

animals : an open access journal from mdpi

10.3390/ani11061544

PMC8230083

Enriched cages for laying hens must contain litter so that pecking and scratching are possible. This is typically provided using layer’s feed dispensed onto a scratch mat, however, there are no regulations on the size or materials of the mat. This study examined how different scratch mat designs and bird age affected behaviours on the mat at three times of day, and their influence on where eggs were laid and shell quality. The proportion of hens at the scratch mats did not increase during or shortly after the application of scratch feed, however, they were more likely to be foraging then. Most eggs collected were clean and laid in the nest. Of the small proportion of eggs that were cracked or dirty, the mat type did not affect dirty eggs, but eggs laid opposite the Big Dutchman mats were more likely to be cracked at 79 weeks of age than at any other mat type or age. There appeared to be no optimal scratch mat design (of those studied) and their use (during observations) was low, suggesting that mat designs were not major influencers on bird behaviour.

Laying hens in the UK and EU must be provided with litter for pecking and scratching. In enriched cages, this is commonly provided by dispensing layer’s feed onto a scratch mat. Mats vary in design and size, which might affect hen behaviour and egg quality, since eggs are sometimes laid at the mats. We investigated if four different scratch mats (BD, K, V, Z) provided to hens in enriched cages resulted in differences in behaviour on the mats and external egg quality. Twenty-four 60-bird cages (6 cages/bank × 4 banks) with 2 mats/cage at one tier of a commercial enriched cage unit were used. Mats were allocated to cages in a balanced design prior to the flock arriving. Hens and eggs were studied at 30, 50 and 79 weeks of age, with three behaviour observations (before, during or after scratch feed application). The data were analysed by GLMMs or LMMs. The vast proportions of birds on the mats were standing (0.720) or sitting (0.250). Bird proportions on the mats were low overall and declined from 0.028 (30 weeks) and 0.030 (50 weeks) to 0.020 (79 weeks) (p < 0.001). The greatest proportion of hens were observed on Z (p < 0.001), which had the largest area, but relative to the available area least birds were on Z and most were on K (p < 0.001). Foraging was not affected by bird age or mat type but was greater at the second observation (p < 0.001). Most eggs were laid in the nest box and were clean. Clean eggs declined, and dirty eggs increased, significantly with age, particularly at the scratch mat (p < 0.001). Dirty eggs were not affected by mat design. Cracked eggs were highest at 79 weeks of age, particularly with BD mats (p < 0.001). Overall, scratch mat designs had minimal effects on behaviour (but few hens were seen there) and egg quality.

1. IntroductionEnriched cages (sometimes referred to as furnished cages) are the only permitted method of housing laying hens in cages in the EU and UK [1]. Although they are falling out of favour in some countries, in others they are still in common use: in 2020, enriched cage eggs made up 40% of eggs through UK packing stations [2]. The benefits of enriched cages over conventional (battery) cages are that they provide more space per hen and enrichments such as a nesting area, perches, and litter so that pecking and scratching are possible.There are two main methods used to provide litter. One is to put litter into a dustbath, used in research cage prototypes [3] and some commercial cage designs (e.g., Specht, [4,5]; Victorrson, [6]; Big Dutchman, [7]). However, this method can result in eggs being laid there [8] and the litter (woodshavings, sand) can interfere with the auger mechanisms. The other design is to put a mat on top of the wire floor, onto which layer’s feed is dispensed (e.g., Zucami, [9]; Big Dutchman, [10]). Eggs may still be laid there, but they can roll freely onto the egg belt. In either method of litter provision, hens may use the litter to express dustbathing or foraging behaviours [9,11,12].Foraging is an important natural behaviour in laying hens. It includes food-seeking behaviours such as ground pecking and scratching [13]. In a study of captive-bred jungle fowl (the ancestors of domesticated chickens), hens were seen ground pecking for 60.6% and ground scratching for 34.1% of observations, respectively (although behaviours were not mutually exclusive) [14]. In a group of feral bantam chickens observed over 8 months, the proportion of time observed feeding was 47.9%, with an average of 50.4 pecks/min [15]. Laying hens that do not have opportunities to forage may develop feather pecking, a serious welfare concern that can lead to severe feather loss, cannibalism and even death [16]. Feather pecking is thought to be related (in part) to the inappropriate redirection of foraging behaviour onto the feathers of other hens, usually where suitable substrates for pecking and scratching are not available [17,18]. In commercial environments, laying hens have various opportunities to show pecking and scratching behaviour, depending on the resources available. In alternative systems such as those used in barn, free-range and organic egg production, where at least one third of the floor is litter [1] which is continuously present, there are arguably more opportunities to show pecking and scratching than in enriched cages, where the scratch mat is small, and litter is not always present due to its being dispersed by the birds. This is compounded because the legislation does not stipulate how to present the litter in cages. As a result, scratch mats in enriched cages come in a variety of designs and sizes, which may affect hens’ abilities to express foraging behaviour there. Furthermore, although nests are provided in enriched cages, scratch mats can be attractive for egg laying, particularly if competition for nest space is high [19]. Laying eggs on the scratch mats (or at least, outside of the nest box) can affect the proportion of dirty eggs produced, particularly if mats are dirty with excreta [20,21]. It may be that different designs of scratch mats will affect external egg quality in different ways, because they can be made of different materials that might be more likely to damage eggs or hold excreta. The aim of this study was to examine the effects of four different scratch mat designs on the behaviour of commercial hens housed in an enriched cage system. We investigated behaviours on the mats, where eggs were laid and the external shell quality.2. Materials and Methods2.1. Hens and CagesA flock of Hy-Line Brown laying hen pullets was sourced from a single rearing farm and placed into enriched cages at 15–16 weeks of age. The laying hen farm was located in Scotland and consisted of 1674 60-bird enriched cages (Big Dutchman, Vechta, Germany), over six banks and nine tiers, with 31 cages/tier. Each cage spanned the entire bank width, and contained two nest boxes, perches (15 cm/hen) and two scratch mats, one on either side of the bank (north and south) (Figure 1).For this study, 24 cages were used in banks 2–5 at tier 5 (6 cages/bank) (Figure 2). Every other cage was used from cage 11–21, so that there was no cross-contamination of eggs with neighbouring cages for egg quality data. Prior to pullets arriving, some of the original mats were replaced with other mat types (K = Kovobel®, Domazlice, Czech Republic; V = Valli®, Italy; or Z = Zucami® Poultry Equipment, Beriain, Spain) in a balanced design, so that each mat was equally represented across banks, cage locations and cage sides.Where the original mats (which had claw shorteners on them) were removed, claw shorteners were stuck to the feed baffle on that cage side, to stay within the regulations. Mats varied in size and colour but were all located in the original mat position so that feed from the litter auger tube would fall onto the mat when activated (Figure 3). Lights came on at 04:00, with first feed at 04:35 until 72 weeks of age, when an extra hour of light was added (lights on 03:00, first feed 03:35). Layer’s mash was applied as litter (scratch feed) through the litter auger tube at 04:35 (03:35 from 72 weeks), 09:20 and 14:20. Lights went off at 18:00. As a result of their various designs, the mats provided different amounts of scratch mat space per hen and different relative areas (Table 1).2.2. BehaviourHens were observed at three ages (30, 50 and 79 weeks of age). At each age, hen behaviour at the mats was observed three times. The first (1st) observation was at 11:00 (30 weeks), 11:30 (50 weeks) or 08:20 (79 weeks) which equated to 1 h 40 min, 2 h 10 min, or 4 h 45 min since last scratch feed provision respectively, and was conducted before the observed scratch feed application; the second (2nd) observation was during and immediately after scratch feed application (14:20 for both 30 and 50 weeks, 09:20 for 79 weeks) but only observed half of the cages (balanced for mat types) to capture behaviour when scratch feed was most likely to be present; the third (3rd) observation was at 15:00 (30 and 50 weeks) or 10:20 (79 weeks) which equated to 40 min or 1 h after the observed scratch feed application. A total of 360 observations (24 cages × 2 sides/cage × 3 visits × 2.5 observations/visit) were carried out, using scan sampling methods to count the number of hens performing behaviours on the mats. The proportions of hens on the scratch mat engaged in behaviours were calculated (number of hens seen on scratch mat/total hens in the cage, or number of hens engaged in particular behaviour on the mat/total hens seen on the mat) (Table 2).2.3. EggsIn enriched cages, to preserve egg quality, eggs roll forward from the nest box and rest against an ‘egg saver’ wire under the feed trough. This lifts at regular intervals set by the farmer to allow the eggs to roll further forward onto the egg belt, which is stationary. At intervals set by the farmer, the egg belt then nudges forward by approximately 1 m towards the egg elevators at the end of the bank, to prevent the build-up of eggs opposite the nest box. On the day before egg and behaviour observations, farm staff cleared the egg belts of all remaining eggs laid by that afternoon (most eggs are laid in the first few hours after lights on). On the day of egg assessment, in order to keep treatment cage eggs distinct from neighbouring cage eggs, neighbouring cage eggs that were on the egg belts since the previous day’s last egg saver wire lift (17:00) were removed. Immediately after the first egg saver wire lift at 05:30, treatment cage eggs were collected (from both sides of the cage) and placed on trays labelled according to the cage furniture area from whence they came (i.e., opposite the nest box, scratch mat, or any other area), while neighbouring cages’ eggs were simply removed. This was repeated a further three times when the egg saver wires lifted at 06:15, 07:00, and 07:45.Treatment cage eggs were counted by location laid and by egg quality factors likely to be affected by cage furniture (dirty (with faeces), cracked, clean). The proportions of eggs by location and by egg quality out of the total eggs collected per cage were calculated.2.4. StatisticsAll data were compiled in Excel. Genstat 18 was used for data processing and all statistical analyses. To analyse proportions, Generalised Linear Mixed models (GLMMs) were fitted to binomial counts with appropriate binomial totals, logit link function, binomially distributed errors and dispersion fixed at 1. To analyse counts, Generalised Linear Mixed models (GLMMs) were fit to the counts with log link function, Poisson distributed errors and dispersion fixed at 1. Where data was sparse and GLMMs with all effects included would not converge random and fixed effects, these models were simplified. Linear Mixed models (LMMs) with all effects included were used as approximations in addition to simplified GLMMs for binomial data. With LMMs, proportion data were first angular transformed to degrees scale (see (1) below) to normalise the distribution of residuals, i.e., for proportion p:(180/π)sin−1(√p)(1)In the results, statistical analyses for the following measurements are reported: counts of birds observed on the mat (all behaviours) on each cage side at each behaviour observation with and without offset log relative area of the scratch mats.counts of birds on the mat on each cage side at each behaviour observation exhibiting different behaviours (standing, sitting, foraging) out of number of birds observed on the same mat at the time of the behaviour observation (thus all reported estimates in the results are on the scale of the proportion (or proportion transformed) of each behaviour on the mat, out of total birds on the mat).counts of egg types (clean, dirty, cracked) out of eggs of all types collected from the belt opposite each location in the cage side (nest, scratch, other) on egg assessment days (thus all reported estimates in the results are on the scale of the proportion (or proportion transformed) of each egg type, out of total eggs at each location).In the LMMs for proportions of birds on the mat or behaviours, fixed effects were bird age (30, 50, 79), observation (1st, 2nd, 3rd), and mat type (BD, K, V, Z) and all 2- and 3-way interactions; random effects were bank, cage, cage.age, cage.age.observation and cage side within cage. In the LMMs for proportions of egg types, fixed effects were bird age (30, 50, 79), location (nest, scratch, other) and mat type (BD, K, V, Z) and all 2 and 3 way interactions; random effects were bank, cage position, side, cage, cage.age, cage.age.location and cage side within cage. Corresponding GLMMs were similar to these LMMs but with both fixed and random effects simplified as required depending on how sparse the data counts were for each response variable in order to achieve model convergence whilst retaining the most important effects. In GLMMs dispersion was fixed at 1. P values are based on approximate F tests when available but otherwise are based on Wald tests; statistics are given in the results as Waldndf and Fndf,ddf, where ndf is the numerator degrees of freedom (the number of effects to be estimated, which is the number of levels for a categorical factor less 1) and ddf is the denominator degrees of freedom. Model estimates ± standard errors (SE) obtained from the LMMs and GLMMs are reported as well as estimates back transformed onto the original scale (proportion) to aid interpretation.The data relating to this study have been deposited in the repository Zenodo (https://zenodo.org/, Geneva, Switzerland, accessed on 23 March 2021), access number md5:2bbc31c13aaf30f8e56ae2c3b5691b01.This study was ethically approved by SRUC’s Animal Experiments Committee, number POU AE 11-2019.3. Results3.1. BehaviourA total of 805 hens were observed on the mats over 360 observations (thus on average 2.2 hens/mat/observation). Of the total hens seen on the mat, the greatest proportion was seen standing followed by sitting, with 0.03 of hens seen engaging in forage, preen, or walk (Table 3). Hens were not observed dustbathing or in other behaviour.3.1.1. Counts of Birds on the MatsThe mean number of birds on the mats engaged in any behaviour was significantly affected by bird age: the count of birds was significantly lower at 79 weeks of age (0.181) than at 30 (0.518) or 50 (0.571) weeks of age (mean SE 0.097, p < 0.001 by GLMM, Wald2 = 24.18, back transformed counts (proportions of 60 birds shown in parentheses): 30 weeks 1.68 (0.028), 50 weeks 1.77 (0.030), 79 weeks 1.20 (0.020)). There was no significant effect of observation (i.e., 1st 2nd or 3rd) (p = 0.587).The mat type significantly affected the proportion of birds on the mats, with a greater proportion of hens seen on the Z mats (0.798) than any other type (BD 0.338, K 0.273, V 0.284, mean SE 0.106, p < 0.001 by GLMM, Wald3 = 42.78, back transformed counts (proportions) BD 1.40 (0.023), K 1.31 (0.022), V 1.33 (0.022), Z 2.22 (0.037)). In the model which adjusts for mat areas, however, although mat type is still significant, more birds were on K and less on Z relative to the available area (BD 1.879, K 2.284, V 1.964, Z 1.562, SE 0.106, p < 0.001 by GLMM, Wald3 = 31.89, back transformed counts (proportions) BD 1.64 (0.027), K 2.46 (0.041), V 1.78 (0.030), Z 1.19 (0.020)). There were no significant interactions (age × observation, age × mat type, or observation × mat type) on the number of hens seen on the mats.3.1.2. StandingFor the proportion of hens seen standing on the mats (number of hens seen standing/total birds on the mat), there was a significant age effect (p < 0.001), with the number of hens standing declining with age (30 weeks 2.578, 50 weeks 1.362, 79 weeks 0.212, mean SE 0.283, by GLMM, Wald2 = 63.37, back transformed proportions 30 weeks 0.929, 50 weeks 0.796, 79 weeks 0.553). There was a weak effect of observation (p = 0.021), with a greater proportion of hens seen standing at the second observation (1.913) than at the first (1.066) or third (1.173) (mean SE 0.283, by GLMM, Wald2 = 7.75, back transformed proportions 1st 0.744, 2nd 0.871, 3rd 0.764). There was no significant effect of mat type on the proportion of hens seen standing on the mat (p = 0.589), nor any significant interactions from the GLMM or LMM.3.1.3. SittingFor the proportion of hens seen sitting on the mats, there was a significant age effect (p < 0.001), with the proportion of hens seen sitting increasing with bird age (30 weeks −2.998, 50 weeks −1.741, 79 weeks −0.568, mean SE 0.266, by GLMM, Wald2 = 57.34, back transformed proportions 30 weeks 0.048, 50 weeks 0.149, 79 weeks 0.362). There was a significant effect of observation (p < 0.001), with a smaller proportion of hens observed sitting at the 2nd observation (−2.912) compared to the 1st (−1.169) and 3rd (−1.225) observations (mean SE 0.273, by GLMM, Wald2 = 20.85, back transformed proportions 1st 0.237, 2nd 0.052, 3rd 0.227). The mat type effect was not significant (p = 0.685) and there were no significant interactions for proportion of hens sitting, although the GLMM model failed for age × observation due to sparse data. The LMM analysis indicates that there was a significant, but weak, age × observation interaction (mean SE 4.840, p = 0.024, by LMM, F4,154 = 2.90) (Figure 4) with the proportion of hens sitting slightly greater for hens at 79 weeks of age at the third observation than at other ages, but fairly similar at other ages and observations. (In fact, generally when birds were observed on the mat, if they were not sitting, they were mostly standing, although the related age × observation for standing was not significant.)3.1.4. ForagingThe proportions of birds foraging on the mats were very low (0.021 overall), however, this was an area of particular interest, to see if different mats stimulated more foraging than others. There were no significant effects of bird age (p = 0.512) or mat type (p = 0.892) on the proportion of birds foraging, by GLMM, however GLMM models with any other effects or interactions included failed to converge. When LMM was used, there was a significant effect of observation on the proportion of foraging seen (p < 0.001), with most at the 2nd observation (1st 0.557, 2nd 7.046, 3rd −0.001, mean SE 1.122, by LMM, F2,133 = 11.49, back transformed proportions 1st 0.000, 2nd 0.015, 3rd 0.000). The LMM indicated a significant age × observation interaction also (p = 0.002, mean SE 1.808, by LMM, F2,146 = 4.51), where the proportion of birds foraging was similarly low across ages at both the 1st and 3rd observation, but at the 2nd observation, more birds were seen foraging at 79, then 50, then 30 weeks (Figure 5).Statistical results for other behaviours are not reported, due to their rare occurrence.3.2. EggsA total of 3564 eggs were assessed from all studied cages over 3 ages. Of those, almost 89% were laid in the nest box, and only 3.9% were laid at the scratch mats, with just over 7% laid in other areas of the cage. The majority of eggs (96.5%) were clean, 1.9% were cracked and 1.6% were dirty (overall eggs from all cages studied).The proportion of clean eggs was significantly affected by bird age (Table 4), with the proportion of clean eggs per location declining with age (p < 0.001 by GLMM, Wald2 = 71.13, 30 weeks 4.871, 50 weeks 3.538, 79 weeks 1.791, mean SE 0.386, back transformed proportions 30 weeks 0.992, 50 weeks 0.972, 79 weeks 0.857). When comparing egg types per location between cage locations, a significantly greater proportion of clean eggs came from the nest, followed by other areas of the cage, and least proportion of clean eggs came from the scratch mat (p < 0.001, by GLMM, Wald2 = 28.89 Nest 4.223, Scratch 2.677, Other 3.300, Mean SE 0.377, back transformed proportions Nest 0.986, Scratch 0.936, Other 0.964) (Table 5). There was no effect of mat type (p = 0.699) or any interactions, although the GLMM model failed for bird age × location due to sparse data. However, bird age × location was significant (p < 0.001 by LMM) where the proportion of clean eggs per location was seen to be significantly lower at the scratch mat at 79 weeks than all other ages by location (F4,252 = 5.05, mean SE 3.566, Figure 6).Statistics on the proportion of dirty eggs and cracked eggs should be treated with caution, as these are based on less than 2% each of all eggs assessed. For the proportion of dirty eggs per location, most GLMM analysis failed for this reason, however, from the GLMM including fixed effects of mat type and location only, there was a significant effect of location (p < 0.001), with the greatest proportion of dirty eggs per location laid at the scratch mat, and least at the nest (Nest-4.843, Scratch-2.711, Other-3.187, Mean SE 0.330, by GLMM, Wald2 = 40.83, back transformed Nest 0.008, Scratch 0.062, Other 0.040) (Table 5). There was no effect of mat type (p = 0.827). From the LMM, there was a significant effect of age (p < 0.001), with the proportion of dirty eggs per location increasing with age (30 weeks 0.047, 50 weeks 3.415, 79 weeks 11.218, Mean SE 2.586, F2,59 = 12.93, back transformed 30 weeks 0.000, 50 weeks 0.004, 79 weeks 0.038). There was also a significant effect of bird age × location on dirty eggs (p = 0.001, by LMM, mean SE 3.219, F4,101 = 5.0, Figure 7) with the proportion of dirty eggs per location similar at all cage locations at 30 and 50 weeks of age, but with significantly greater dirty eggs laid in the scratch at 79 weeks compared to the nest and other cage locations. All other interactions were not significant in the LMM.With the proportion of cracked eggs per location, there was a significant effect of age (p < 0.001 by GLMM, 30 weeks −5.256, 50 weeks −4.905, 79 weeks −2.942, Mean SE 0.3983, Wald2 = 49.07, back transformed 30 weeks 0.005, 50 weeks 0.007, 79 weeks 0.050) with a higher proportion of cracked eggs per location seen at 79 weeks. There was no significant effect of location or mat type. With interactions, all GLMMs failed, but LMMs indicated some significant interactions. There was a significant bird age × location effect (p = 0.016, by LMM, F4,267 = 3.11, mean SE 1.465) with a greater proportion of cracked eggs per location seen at the nest and scratch mat than at other areas of the cage at 79 weeks (Figure 8). There was a significant age × mat type effect (p = 0.002, by LMM, F6264 = 3.49, mean SE 1.691) with the proportion of cracked eggs per location significantly higher at 79 weeks opposite Big Dutchman (BD) mats, as opposed to any other mat type (Figure 9).There was no significant effect of location × mat type (p = 0.152), but there was a significant 3-way interaction of age × location × mat type (p < 0.001, by LMM, F12,268 = 3.42, mean SE 2.819), which indicates that the biggest influencer on the proportion of cracked eggs per location at the scratch mat at 79 weeks of age (as seen in Figure 8. The mean proportions (angular transformed) of cracked eggs per location by bird age (weeks) and location in the cage (nest, scratch mat, all other areas) estimated from LMM. Mean SE 1.465) is from Big Dutchman scratch mats (Figure 10).4. DiscussionIn this study, only about 2 hens were observed on the mats per observation, and as a result the proportions of birds out of populations of 60 hens/cage on the mats was very low. Although the proportions of hens on the mats was significantly lowest at 79 weeks of age, the differences are small (between 0.02 and 0.03). The lack of an effect of observation on the counts of hens on the mats was surprising, given that more hens were expected on the mats during or shortly after scratch feed application (2nd observation), when the scratch mat is assumed to be at its most attractive, but mat areas (apart from Z) are similar in area to an A4 sheet of paper, which would not be able to accommodate many hens.High proportions of inactive (stand, sit) behaviours were recorded at the scratch mats, accounting for 0.97 of observations. This may be because hens are genuinely inactive at the mats, or that observer presence disturbed hens from more active behaviours. By contrast, in a study of laying hens in enriched cages with various keel bone fracture severity, stand and sit behaviours accounted for only 23.6–30.0% of behaviours observed, but that was not restricted to the scratch mat area, and hens were in a research environment [22]. Commercial hens may see people less frequently than those in a research facility, so may be less habituated to their presence. Also, in commercial enriched caged systems, aisle widths restrict the distance the observer can be from the area of interest: aisle widths must be at least 90 cm between tiers [1], and to make best use of space, may typically be no more than this. Using remote or automated equipment is one way to avoid bird disturbance [23], however it was not possible to install video equipment to record behaviour at the commercial farm. The larger mat (Zucami) saw a higher proportion of hens on it than any other mat, but the difference in bird proportions is in reality small (i.e., equates to 0.037 of hens in a cage on Z mats versus 0.022–0.023 of hens in a cage on other mats), and in fact, relative to the area, half as many birds were on Z compared to K.As hens got older, they tended to stand less, and sit more, on the mats. Sitting behaviour was lower, and foraging behaviour was higher (particularly at 50 and 79 weeks of age), at the 2nd observation during or shortly after scratch feed application. The amount of foraging behaviour observed on the mats was generally very low, although it was higher at 50 and 79 weeks of age compared to 30 weeks of age, and dustbathing (which might also be expected to be elicited by litter on the mat) was not observed at all. As with other work [24], it is logical that foraging behaviour is most likely to occur during the presence of litter, however litter is quickly eaten or depleted [25], and thus the positive feedback from foraging also ceases then. Therefore, if foraging behaviour is to be stimulated for longer durations or more frequently, a greater quantity of litter or a higher frequency of provision might need to be provided. The use of layer’s feed as litter should be attractive to hens as a foraging substrate: in a study comparing wood shavings, pelleted lignocellulose, no substrate or layer’s feed, the feed was preferred for foraging [26], whereas a bare mat is not attractive for foraging [26,27]Most eggs were laid in the nest, which suggests that hens found the nest design suitable for egg laying behaviour over most other areas of the cage [28]. This agrees with previous work where Big Dutchman enriched cages were studied, and those nests were favoured by Hy-Line brown hens for the majority of egg laying [29]. Eggs that were not laid in the nest box were more likely to be found in other areas of the cage rather than opposite the scratch mat. This is in contrast to Hunniford et al. [19], who found that in both small (28 or 40 hens/cage) and large (55 or 80 hens/cage) enriched cages, most eggs laid outside of the nest were laid at the scratch mat. The vast majority of eggs were clean, which is highly desirable in commercial egg production. The proportion of clean eggs declined, and the proportion of dirty eggs increased, particularly at the scratch mat, at 79 weeks of age. Onbaşilar et al. [21] found that a higher percentage of dirty eggs in enriched cages were laid outside of the nest box, but they did not distinguish between eggs laid at the mat and other areas. In cage designs that offered litter either on scratch mats on the wire or in litter boxes above the nest, both egg laying in the litter facility and dirty eggs tended to be higher in cages providing scratch mats [8], however the cage design was also confounded with group size. Here, no mat type was more likely to have dirty eggs than another. It is notable in this study that no mats were made of Astoturf-type material, but instead were hard plastic which may stay clean more easily.As expected, cracked eggs were greatest when hens were oldest, because eggshell quality declines with hen age as egg size increases and eggshell thickness decreases [30,31]. Of the small proportion of cracked eggs seen, at age 79 weeks they were less likely to be cracked if they were laid in other areas than the nest or scratch mat. This is similar to Onbasilar et al. [21] who found significantly fewer cracked eggs outside rather than in the nest box. This may be because hens on these other areas are closer to the food trough (so eggs have less far to roll) than the nest or scratch mat, which are positioned furthest from the egg belt. Another explanation for this may simply be that since fewer eggs are laid outside the nest box, there are less likely to be collisions between eggs, which can crack one another. There were more likely to be cracked eggs if laid at the BD mat than any other mat type, at 79 weeks of age, but there is no clear reason for this, and it should be kept in mind that cracked eggs overall were low.5. ConclusionsOverall, scratch mat designs studied here did not appear to be major influencers on hen behaviour or egg characteristics. Although the application of litter onto the mats did not increase proportion of hens found there (compared to other observation times), hens at that observation were more likely to be foraging then. Most eggs laid were clean and laid in the nest. It may be that these scratch mat designs are equally adequate (or inadequate) at eliciting behaviours there, or that the study design disturbed behaviour too much to get a true record of what happens at the mats. Further work would benefit from studying behaviour at mats remotely or with human-habituated hens.

animals : an open access journal from mdpi

10.3390/ani11123562

PMC8698172

Feral cats are difficult to manage and harder to monitor. We report on the efficacy of Eradicat® baiting and the cost and the efficacy of monitoring the activty of feral cats via camera-traps or track counts. Pre-baiting surveys for 2020 and 2021 suggested that the population of feral cats on Matuwa was very low, at 5.5 and 4.4 cats/100 km respectively, which is well below our target threshold of 10 cats/100 km. Post-baiting surveys then recorded 3.6 and 3.0 cats/100 km respectively, which still equates to a 35% and 32% reduction in cat activity despite initial low cat detection rate. Track counts recorded more feral cats than camera traps and were cheaper to implement.

Feral cats are difficult to manage and harder to monitor. We analysed the cost and the efficacy of monitoring the pre- and post-bait abundance of feral cats via camera-traps or track counts using four years of data from the Matuwa Indigenous Protected Area. Additionally, we report on the recovery of the feral cat population and the efficacy of subsequent Eradicat® aerial baiting programs following 12 months of intensive feral cat control in 2019. Significantly fewer cats were captured in 2020 (n = 8) compared to 2019 (n = 126). Pre-baiting surveys for 2020 and 2021 suggested that the population of feral cats on Matuwa was very low, at 5.5 and 4.4 cats/100 km, respectively, which is well below our target threshold of 10 cats/100 km. Post-baiting surveys then recorded 3.6 and 3.0 cats/100 km, respectively, which still equates to a 35% and 32% reduction in cat activity. Track counts recorded significantly more feral cats than camera traps and were cheaper to implement. We recommend that at least two methods of monitoring cats be implemented to prevent erroneous conclusions.

1. IntroductionFeral cats (Felis catus), cats that live in the wild and can survive without human reliance or contact, are recognised as a key threatening process to native species in Australia [1,2,3] and around the world [4,5]. Predation by feral cats has been demonstrated to threaten the persistence of many native species [6,7], and causes billions of dollars damage to the natural and agricultural environment [8] alongside disease transmission [9]. Predation by feral cats has been identified as one of the major obstacles to the successful reintroduction of extirpated native fauna [10,11,12,13]. Therefore, the suppression of feral cat populations is a critical component to the successful conservation of small to medium-sized native fauna [14,15].While implementing methods of feral cat control is difficult, measuring the outcomes of feral cat control, the pre- and post-management abundance of cats, is harder. Detecting feral cats is difficult because they are a cryptic species that avoids interactions with humans [16] and in some environments cannot be readily detected via remote sensing technologies, such as camera-traps, despite the use of lures [17,18]. Estimating the abundance of an animal species typically requires capturing or identifying individual animals on multiple occasions [19]. Capturing feral cats on multiple occasions is extremely difficult, requiring the use of multiple labour intensive techniques [20] and feral cats frequently lack the unique markings required to identify individuals for mark-resight analysis [21]. New analytical techniques such as N-mixture models and Royle–Nichols abundance models that may alleviate some of these issues require further investigation [22,23].Feral cat control is primarily a task performed or funded by government agencies [1]. Between 1998 and 2003, $4.7 million was spent by Australian conservation organisations on labour associated with cat control with operational costs (materials, vehicles, equipment) requiring additional funding [24]. From an economic perspective, it is essential that we measure the outcomes of the pre- and post-management abundance of cats. Cats can breed rapidly, potentially doubling in abundance each year [25], and disperse over long distances [26,27]. This allows populations of feral cats to potentially recover quickly post-management. Moseby et al. [28] demonstrated that some fauna can survive and increase in abundance in the presence of a low density of cats (~0.5 km2). From a biological perspective, it is essential that consistent and ongoing monitoring of the pre- and post-management abundance of cats occurs to ensure we maintain a low density of cats. The challenge is to develop a cost-effective method of monitoring cats at a landscape-scale. Several recent studies have tested the efficacy of camera-traps with slight modifications to camera placement, lures, and survey duration with minor improvements in the efficacy of camera-traps [17,29,30,31]. Edwards et al. [32] compared track counts with spotlighting and concluded that track counts were a more reliable method of monitoring mammalian carnivores. These studies did not consider the cost of implementing these techniques. Lohr and Algar [33] used both camera-traps and track counts to monitor feral cats and concluded that camera-traps provide more reliable data, but are more expensive and time-consuming to implement than track counts. Track counts provide a cheaper, rapid survey technique, but are susceptible to error from inexperienced observers, and weather conditions erasing tracks. Lohr and Algar [33] did not formally analyse the cost or efficacy of these two techniques.The purpose of this manuscript is to analyse the cost and the efficacy of monitoring the pre- and post-bait abundance of feral cats via camera-traps or track counts. Additionally, we report on the recovery of the feral cat population and the efficacy of subsequent Eradicat® aerial baiting programs on the Matuwa Indigenous Protected Area (IPA) following 12 months of intensive feral cat control [33] that consisted of aerial baiting and landscape-scale leg-hold trapping.2. Materials and Methods2.1. Study SiteThe Rangelands Restoration program at the Matuwa Indigenous Protected Area (2440 km2; ex-Lorna Glen pastoral lease) in central Western Australia (26°13′ S, 121°33′ E; Figure 1) aims to achieve the successful reconstruction of an Australian arid zone native species assemblage. To date, five species have been successfully reintroduced to Matuwa; the bilby (Macrotis lagortis), common brushtail possum (Trichosurus vulpecula hypoleucus), Barrow Island golden bandicoot (Isoodon auratus barrowensis), burrowing bettong (Bettongia leseuer), and mala (Lagorchestes hirsutus), of which the final two are still confined to a predator-free fenced area [34,35,36,37]. The successful reintroduction of native species to the open landscape can only be maintained if an effective and sustained feral cat control program can be achieved [10,38,39,40].Matuwa consists of two main land systems: (1) Bullimore—sand plains and dunes dominated by spinifex (Triodia spp.); and (2) Sherwood—breakaways and stony plains dominated by mulga and other acacia shrublands with the most common vegetation unit being mulga (Acacia aneura) and Eucalyptus kingsmillii over hummock grasslands (Triodia basedowii) [41,42]. Matuwa, being in the arid zone, is characterized by extreme temperatures and low and erratic rainfall with an annual average of 261.7 mm (Bureau of Meteorology, records 1898–2018; Wiluna weather station No. 13012 located 137 km WSW of Matuwa). Average maximum daily temperatures range from 19 °C in winter to 38 °C in summer, and average minimum temperatures range from 5 °C in winter to 23 °C in summer.2.2. Feral Cat ManagementSince 2003, we have been using the poison bait known as Eradicat® on Matuwa to control feral cats [38,43,44]. Eradicat® baits contain 4.5 mg of directly injected toxin ‘1080′ (sodium monofluoroacetate). Prior to being laid, feral cat baits are thawed in direct sunlight and sprayed, with an ant deterrent compound (Coopex®) at a concentration of 12.5 g/L. This process is aimed at reducing ant attack and maintaining the palatability of the bait to cats. Most years, baits are deployed from a fixed wing aircraft at a rate of 50 baits/km2 during the cool, dry winter periods when the abundance and activity of all prey types is at its lowest [44]. Since Matuwa is a site subject to adaptive management [33] there has been some variation in the portion of Matuwa subject to aerial baiting and/or ground based baiting. Hence, for this analysis, we have selected data from feral cat survey points that occurred on portions of Matuwa subject to aerial baiting only between 2018 and 2021 (minimum 1353 km2).Three feral cat trapping programs were conducted at Matuwa between 2018–2020. The first, a small-scale exercise with 1600 trap-nights, was conducted immediately following the baiting program in August 2018 to provide a snapshot of the population demographic of resident cats that had survived the baiting program [33]. The second, a more comprehensive trapping program with the goal of reducing the abundance of feral cats on Matuwa and a total of 5398 trap-nights were conducted across the site prior to the baiting program in March and April 2019 [33]. The third trapping program, which is first published here, was conducted from 5 August to 6 September 2020, with each trap in commission for 10 consecutive days as per previous years (Figure 1). The whole trapping circuit, comprised of a linear track length of 280 km. The three trapping programs used the same personnel and trapping methodology. Trapping was conducted using pairs of padded leg-hold traps, Victor ‘Soft Catch’® traps No. 1.5 (Woodstream Corp., Lititz, PA, USA), using a mixture of cat urine and faeces as the attractant. Trap pan tension was maintained at manufacturer standard to ensure that smaller cats were not excluded from the study thereby reducing the risk of biased demographic data. Trap locations recorded in 2019, using a Garmin GPS Rhino 650 (Garmin Ltd., Olathe, KS, USA), were used to re-position trap-sets in 2020. Of the 573 trap-sets deployed in 2019, 558 trap-sets were recommissioned in 2020. Fifteen trap sites were discontinued because of access difficulty along one track in the southeast corner of the property (Figure 1). All traps were checked each morning within three hours of sunrise, and any trapped cats were euthanised using a 0.22 calibre rifle shot to the head at point blank range. Chi-squared tests were used to test whether there were significant differences in captures between the two trapping programs.2.3. Feral Cat DemographicsAll animals captured were weighed and sexed; a broad estimation of age (as either kitten, juvenile or adult) was recorded using weight as a proxy for age. The yearling weight and age classes adopted in the previous study (see Table 1; adapted from Jones and Coman [45]) were used to define the population age structure. The pregnancy status of females was determined by examining the uterine tissue for embryos. 2.4. Camera-Trap MonitoringTwo camera-trap arrays were used to monitor feral cat activity on Matuwa pre- and post-management. In 2018–2020, 120 camera-traps (Reconyx Hyperfire PC900 Professional camera; Reconyx, WI, USA) were installed using a stratified-random design based on the 20 most common geological types in the Wiluna region [46]. The cameras were placed between 30 m and 200 m away from an ungazetted track (Figure 1). Camera-traps were, on average, 2.80 km from their nearest neighbour (min = 1.0 km, max = 5.9 km). Spatial autocorrelation is unlikely at this scale [47] despite the potentially large home-ranges of feral cats [48]. Data from 70 of these cameras that were placed in an aerial baited site in 2018 and 2019 were used in subsequent analysis [33]. In 2020 and 2021, the entirety of Matuwa was aerial baited. Data from all 120 cameras in 2020 were used. In 2020–2021, a grid of 130 camera-traps with cameras spaced 1 km apart was installed on Bullimore sandplain in the south of Matuwa (Figure 1). Previous research has shown that feral cats are most active on sandplains [33].All cameras were mounted on a 30 cm high plastic sand peg, in a horizontal position, facing south, in a space with at least 3 m of open ground in front of the camera. Two olfactory lures (Canines-a-plenty and Catastrophic from Outfoxed Pest Control, Victoria, Australia) were placed on two natural sticks approximately 30 cm tall and 1 m apart, 3 m from the front of the camera and refreshed at least 10 days before and after management. Herbaceous vegetation immediately in front of the camera was removed. Camera-traps captured three photos per trigger, with no quiet period. Timed photos were taken at 23:00 h to monitor the operation of the camera. The efficacy of the lures is thought to fade over time. Therefore, we used data from the first 10 days after camera-traps and lures were set, pre- and post-management. Photos were stored in the Colorado Parks and Wildlife Photo Warehouse database (CPW) [49] and viewed by at least two observers to confirm species identification. A histogram of time intervals between consecutive photos revealed that 99.6% of photos were captured either <5 min apart or >60 min apart (Supplementary Materials, Figure S1). To minimise temporal autocorrelation, we grouped consecutive photos that were <5 min apart to create independent records for subsequent analysis [50].2.5. Track CountsTrack counts collect data that reflects that activity of feral cats on unsealed roads referred to as the track activity index (TAI). Approximately two weeks pre- and post-baiting, two teams of experienced observers ran a single TAI transect at least 50 km in length each day [33,38] for four consecutive days (Figure 1). Teams alternated transects each day to reduce observer bias. Since Matuwa is a site subject to adaptive management [33] there has been some variation in the placement of TAI transects between years (Figure 1), but the placement of TAI transects is consistent within years pre- and post-cat management. TAI-transects occur on sandy 4WD tracks, which are initially cleared by towing a heavy iron drag behind a 4WD vehicle. Observers, driving all-terrain vehicles (ATVs) at a speed of 10–15 km/h then inspect the transect for cat tracks, and clear new signs of animal activity by towing a chain iron drag. Cat tracks that occur within 1 km radius of one another on a daily survey are aggregated into one cat detection to minimise spatial autocorrelation. A histogram of distances between consecutive cat tracks revealed that 92.3% of tracks were <1 km apart and usually caused by cats that travelled along roads leaving a continuous set of prints, which were documented as having left one discrete set of prints every 100 m for the purposes of this analysis only (Supplemental Information, Figure S1). The single 50 km transect is split at disused wells and intersections with any observations recorded within 1 km of the well or intersection being discarded. The number of cats observed on each TAI-transect is scaled against the total length of the TAI-transect within each day and then averaged across sequential survey days. Only TAI-transects that occurred on areas that were aerially baited with Eradicat® were analysed.2.6. Analysis of Monitoring DataCount data from camera-traps and continuous TAI data were analysed via negative binomial mixed-effects models with a parameter for zero-inflation in the R (V4.0.2, [51]) package glmmTMB [52], with monitoring method (dispersed camera-traps, grid camera-traps, or track count), survey (pre- or post-management and post-trapping), and year as factorial fixed effects, while year and TAI-transect name or camera ID were used as random effects. Models were generated through a backward stepwise refinement process after fitting a global model and reviewing the significance of individual fixed effects. We used the DHARMa package [53] to review model residuals and fit. Models were ultimately compared via Akaike’s Information Criterion (AICc) in the package AICcmodavg 2.3-1 [54].3. Results3.1. Feral Cat DemographicsThe 2020 trapping program, which was conducted over 5539 trap-nights, resulted in the capture of eight cats (5 males, 3 females). A number of the 558 traps were decommissioned early when they captured non-target species, or when red kangaroos (Osphranter rufus) and euros (O. robustus erubescens) destroyed the trap-set. Percentage trap success was 0.14 cats/trap-night; capture locations are presented in Figure 1. No kittens or juvenile cats were captured. There was no significant difference between adult male and female captures (Chi2 = 0.5, df = 1, p > 0.50). Two of the adult males were 2+ years of age (weights 4.3 and 4.6 kg) and the remaining were adults 1–2 years of age with a mean weight of 3.6 (±S.E. 0.3) kg, range 3.1–3.9 kg. The three females captured were all adults 1–2 years of age with a mean weight of 2.8 (±S.E. 0.1) kg, range 2.8–2.9 kg. None of the trapped females were pregnant. Assuming equal trappability between years, there was a significant difference in the numbers of cats trapped with greater numbers in 2019 (n = 126) compared to 2020 (n = 8, Chi2 = 103.9, df = 1, p < 0.001), despite 141 fewer trap-nights (Figure 2). Similarly, more cats were trapped along the same trap-lines used during the pilot study of 2018 (n = 33) than at the same time of year in 2020 (n = 1). This was a significant difference (Chi2 = 20.17, df = 1, p < 0.001), with a percentage trap success of 1.87 cats/trap-night in 2018 compared to 0.08 cats/trap-night in 2020.3.2. Monitoring Feral CatsThe full suite of negative binomial mixed effects models revealed that the various arrangements of the cameras did not collect significantly different numbers of cat detections (p = 0.92; Model 1; Online Supplementary Materials). Therefore, through the iterative model refinement process we pooled data from the two camera arrays resulting in 120 cameras being deployed in 2018 and 2019, 250 cameras in 2020, and 130 cameras in 2021. We assessed the goodness-of-fit of the remaining models via R package DHARMa [53] and selected the Model 4 as the best model because the residuals did not significantly deviate from the expected distribution (Kolmogorov–Smirnov (KS) test = 0.58) nor were there significant outliers (p = 0.92; Table 2). The 130 cameras used in a grid array in 2021 were placed in spinifex sandplain habitat, which consistently recorded the greatest number of cats in prior studies [33]. From our best model (Table 2; Model 4) the TAI did record significantly more cat detections than cameras (p = 4.59−15; Figure 3). Pairwise comparisons revealed that the average number of feral cat detections during the pre-baiting survey was significantly higher than post-baiting surveys (p = 1.60−9) and post-trapping surveys (P = 2.02−6). The difference between post-baiting and post-trapping surveys was not significantly different (p = 0.93). In our second best model (Model 5, Table 2, KS = 0.72) year significantly affected the number of cats detected revealing that 2018 recorded significantly more cats than 2019 (p = 6.40−4), 2020 (p = 2.14−13) or 2021 (p = 1.35−15). This result confirms that the second comprehensive trapping program for feral cats in March and April 2019 (prior to pre-baiting survey of 2019) significantly reduced the on-going detection of feral cats on Matuwa. The proportion of camera-traps or TAI-transects that recorded zero cats was considerable with 65% of TAI transects, 99% of cameras spread across the landscape, and 99.8% of cameras in the grid recording zero cats each day.In 2020 and 2021, pre-baiting surveys suggested that the population of feral cats on Matuwa was very low, at 5.5 (SE ± 1.2) and 4.4 (SE ± 1.3) cats/100 km, respectively, which is well below our target threshold of 10 cats/100 km [33]. Post-baiting surveys then recorded 3.6 (SE ± 1.1) and 3.0 (SE ± 0.9) cats/100 km, respectively, which still equates to a 35% and 32% reduction in cat activity in 2020 and 2021. Wide error margins around the average number of feral cats detected by 100 km of TAI transects is to be expected as the activity of feral cats varies with habitat type [33].Considerably more cats were detected in 2018 prior to the second comprehensive trapping program for feral cats in March and April 2019 potentially inflating the statistical significance of year and survey. Re-analysing camera-trap and TAI data from 2020 and 2021 only, using model formulation 5 (KS = 0.41; Outlier test = 0.59) reveals ongoing significant difference between the pre-baiting and post-baiting surveys (p = 1.08−4; Figure 3), despite the initial low abundance of cats.The camera-trap data was more severely zero-inflated resulting an average feral cat detection rate for track counts that was 5 to 25 times higher than the average detection rate for camera-traps. This difference may be masking the value of camera-trap data in statistical analysis. Re-analysing data from camera-traps and TAI-transects separately reveals that camera-trap data did not detect a significant difference between survey periods (p > 0.72) or years (p > 0.51), whereas TAI-transects did detect a significant difference between 2018 and subsequent years (p > 3.90−5), no difference between the pre-baiting and post-baiting surveys (p = 0.80) and a significant difference between pre-baiting and post-trapping surveys (p = 2.74−3; Figure 3; Online Supplemental Information). The 10-day camera-trap survey in the post-baiting period of 2021 detected zero cats, as did the post-trapping survey of 2020, which may lead to erroneous conclusions. The camera-trap data also suggest that the number of cat detections increased in 2018 following trapping, whereas track-counts suggested the opposite trend (Figure 3).3.3. CostFeral cat monitoring on Matuwa is a long-established project. Table 3 illustrates the operational costs associated with initiating either TAI-transects or camera-trap monitoring on a new site for four years in 2021 with salary calculated as the average hourly rate of the people involved in monitoring on Matuwa. We do not include travel costs, analysis, or overtime benefits as these are site, project, and employee specific and hence are not costs that are transferable to other budgetary frameworks. We assume that a feral cat management project would only use one camera array. Ultimately, implementing camera-traps is twice as expensive as track counts for detecting feral cats (Table 3). 4. DiscussionTrack counts proved to be cheaper to implement and more effective at detecting feral cats, especially when cat density was very low. Feral cats typically have a low probability of detection per night by cameras [31], possibly because each camera has a field of view of 40° by 30.5 m [55] and hence surveys a maximum 320 m2. Reconyx cameras are designed to capture larger species such as deer (Cervidae) [55]. Feral cats are smaller and less likely to be detected when 30 m from the camera [56]. An array of 130 camera-traps used over 10 nights, surveys approximately 417,846 m2. In contrast, 100 km of track counts on 4WD tracks approximately 3 m wide over four nights, surveys approximately 1,200,000 m2.On Matuwa, the low probability of detecting feral cats via cameras meant that during two surveys (post-trapping survey of 2020 and post-baiting period of 2021) zero cats were detected and therefore no estimate of cat activity or population density could be derived. Given that both surveys occurred after the implementation of feral cat management actions we could make a type 1 error by falsely concluding that we had removed all feral cats from the property. Use of a second method of detecting cats prevented that erroneous conclusion.In 2018, data from camera-traps suggested that leg-hold trapping increased the number of feral cat detections and hence may have increased the abundance of feral cats on the property. Similarly, other studies have detected a net positive effect on cat detections following the use of toxic baits [57]. In eastern Australia, this nonsensical result was attributed to a significant reduction in cat activity at a paired unbaited site, which biased model outputs [57]. We attribute our increased number of feral cat detections to increased activity in feral cats, which were not removed by our management actions and may have been expanding their territories and seeking sign of conspecifics, as a result of neighbouring individuals being removed and the use of scent-based lures at camera-traps. The use of a second method of detecting cats prevented us from drawing a flawed conclusion from our 2018 data. No lures are used during track counts. The use of multiple monitoring systems for feral cats is advisable, particularly in eradication programs either on islands or within fenced areas, especially as the population declines to low numbers. Multiple monitoring systems were crucial during the eradication of feral cats from Dirk Hartog Island off the coast of Western Australia where the last cat was not recorded on camera, but its sign was observed on a track transect [58]. Evidence of this cat sign resulted in deployment of traps in that area and the subsequent removal of the last cat. Where the soil substrate is not sand, and track counts are difficult to implement, alternative monitoring techniques, such as importing sand to make sandpads [24,59] and/or deployment of hair snares [60], could be used to complement and verify data from camera-traps. During trapping in 2020, no cats in the 0 to 1 age class were captured suggesting that the baiting program in 2020 had a significant impact on this age class. Prior studies have concluded that juvenile and female cats are more susceptible to toxic baits as they have higher energy requirements [33]. Our data seem to support this hypothesis. Future research should continue to monitor the demographics of the feral cat population because consistent loss of juveniles to baiting would confirm that any recovery in the feral cat population at Matuwa is a result of immigration from neighbouring properties.Ecosystems are spatially and temporally dynamic. The abundance of fauna is a response to dynamic abiotic (e.g., rainfall) and biotic (e.g., abundance of food) conditions. It is difficult to ensure that an experimental design, with sufficient power, replicates and control treatments, is applied to research that occurs at a landscape-scale [61]. We used a study design that consists of repeated before/after sampling at a single site [62]. Two potential control sites for Matuwa were discarded: Kurrara Kurrara, a property to the north, has considerably more salt lake country, which feral cats rarely use [33], and considerably more feral herbivores [63]; Jundee, a property to the west, has a similar habitat assemblage, but is managed as a pastoral lease and mine site and hence offers potentially confounding variables at a landscape-scale. Likewise, other neighbouring properties that should experience sufficiently similar climatic variables are functioning cattle stations with potentially confounding variables. Sampling designs that use only a single control site to contrast against a single potentially impacted site (BACI) may be confounded with any pre-existing cause of variability between the two locations [62]. Comparing a single year of Eradicat® baiting in a National Park to a cattle station with only 30 replicates for calculating a relative abundance of cats is inappropriate, as it is unlikely to be statistically robust given the variation in management regimes [64]. If typical climatic conditions are maintained throughout the experiment, a large number of survey replicates are implemented, multiple monitoring techniques are used, and researchers carefully consider the ecological relationships among species, then reasonable inferences can be drawn from repeated before/after sampling at a single site with a pulse disturbance [62] such as baiting. We caution researchers against the use of single-season before/after sampling experiments without careful consideration of relevant ecological relationships and wider temporal and spatial context or the use of alternative methods of monitoring feral cats that may corroborate results. Wysong et al. [30], for example, concluded that annual aerial baiting for feral cats was more effective at reducing the wild dog population than the feral cat population, and that the reduction in wild dogs had allowed an increase in the abundance of kangaroos [30], but did not consider the variation in timing of breeding behaviour and response to environmental conditions that these species may exhibit when interpreting their results. They could have used data presented in their other work [65] to assess the validity of their results. Two survey methods, track counts [66] and mortality of cats with GPS collars [65], concluded that aerial baiting reduced relative cat abundance by 61% [66] and 66% [65], whereas camera-traps and occupancy analysis found an approximately 15% reduction in relative cat abundance [30].5. ConclusionsAfter many years of research, we conclude that feral cats are difficult to manage and harder to monitor. We recommend that at least two methods of monitoring cats be implemented to prevent inaccurate conclusions. Ideally, BACI designs are likely to be more informative where multiple control sites with a similar management history are available. We recommend an integrated pest management framework in designing feral cat management programs using both toxic baits and trapping/control methods.

animals : an open access journal from mdpi

10.3390/ani11102834

PMC8532723

Postfracture treatment with nonsteroidal anti-inflammatory drugs might result in delayed healing. This study was aimed to evaluate and compare the effects of flunixin meglumine (FM) and ketoprofen on bone fracture healing in rabbits. A simple unilateral diaphyseal fracture was made and followed by fixation by K-wire. Healing was evaluated with radiography, histopathology, and immunohistochemistry. Interestingly, the results revealed that FM enhanced bone fracture healing combined with the activation of early collagen deposition, marked angiogenesis, and enhanced vascular endothelial growth factor. However, ketoprofen delayed bone fracture healing. These findings provide novel baseline information about the potential beneficial effects of FM on bone fracture healing cases.

Nonsteroidal anti-inflammatory drugs (NSAIDs) are among the most commonly used postoperative analgesics, antipyretics, and anti-inflammatories, and they help prevent blood clotting. However, most NSAIDs delay bone healing. This study was aimed to investigate bone healing in a rabbit animal model by assessing the ability of flunixin meglumine (FM) and ketoprofen to induce fracture healing by examining histology, radiological changes, and vascular endothelial growth factor (VEGF) immunostaining during bone healing. For this purpose, 24 New Zealand rabbits were assigned to three groups: the control group, the FM group, and the ketoprofen group. Our results revealed that there were no intraoperative complications, and all surviving rabbits achieved full-weight bearing. Significant periosteal reaction and callus formation were confirmed at 2 postoperative weeks. Interestingly, FM enhanced callus formation, bone union, and remodeling in the FM group compared to the control and ketoprofen groups. FM enhanced bone healing through early collagen deposition and marked angiogenesis process activation by increasing the expression of VEGF. Our findings demonstrated, for the first time, the potential imperative action of FM in the bone healing process rather than other NSAIDs in animals.

1. IntroductionThe bone healing process involves multi-sequential events associated with numerous cellular functions that promote mineralization of the fracture site followed by remodeling to return the affected bone to its original structure [1]. In the literature, many postoperative drugs, including antibiotics [2] and anti-inflammatory drugs [3,4,5,6], have been used in different stages of the healing process. Among them, anti-inflammatory drugs have been extensively used as a painkiller to ensure fracture site immobility. However, numerous studies have demonstrated that steroidal and nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit healing in soft and hard tissues [3,4]. Furthermore, various reports have suggested that NSAIDs interfere with bone healing, though others have contradicted these findings [7]. Celecoxib [8,9] and aspirin [10] inhibit bone healing [11,12], while ibuprofen [13] and parecoxib [14] do not affect bone healing.The inhibitory action of NSAIDs might result from the inhibition of prostaglandin production from arachidonic acid via either the cyclooxygenase 1 (COX-1) or cyclooxygenase 2 (COX-2) enzyme pathway [15,16]. COX-1 is initially involved in physiological functions such as the preservation of hemostasis and gastric protection [17]. Meanwhile, COX-2 is initially found and involved in the pathophysiological pathways such as inflammation, fever, and pain [3,18]. However, prostaglandins play an important role in the bone healing process, especially through the regulation of osteoblast and osteoclast functions [19]. The marked inhibition of prostaglandin production decreases bone formation rate [20].Ketoprofen is a potent nonselective COX-1 inhibitor that is extensively used in small animal medicine [15]. Ketoprofen is a nonselective NSAID used to decrease inflammation, pain, and fever, and it has analgesic effects. Data regarding the effect of ketoprofen on bone healing have been variable and controversial. Ketoprofen may not affect bone healing, especially in small doses for a short time [21]. In a study of the effects of ketoprofen on lumbar spinal fusion healing, It was revealed that using a single dose after intertransverse spinal fusion has no effect on bone healing [22]. Flunixin meglumine (FM) is a nonselective NSAID used as a painkiller, antifever, anti-inflammatory, and antitoxemic [23,24,25]. Few studies have explored the effect of FM on the soft tissue healing process. A previous study reported that FM promotes tendon healing by increasing fibroblast and blood supply to the wound site [23]. To the best of our knowledge, there has been no study on the effect of FM on bone healing. Growth factors play pivotal roles in the bone healing process [26]. The preservation of bone blood flow is considered one of the main mechanisms that promotes bone healing [26]. Vascular endothelial growth factor (VEGF) expression is essential for neurovascular development, i.e., vasculogenesis or the de novo production of blood vessels from mesenchymal precursor cells [26,27,28]. Given this information, the authors of this study aimed to determine the potential effects of ketoprofen and FM on bone healing in a rabbit femoral fracture model. The authors assessed these effects through a series of methods, including routine X-ray, histopathology, collagen deposition, and VEGF immunohistochemistry.2. Materials and Methods2.1. Ethical Committee ApprovalThe Research Ethics Committee of the Faculty of Veterinary Medicine, Kafrelsheikh University, reviewed and approved the procedures and protocols involved in both the animal welfare and surgical techniques of this study under approval number KFS-2020/7. The ethical approval code number is KFS-2020/7.2.2. AnimalsA total of 24 New Zealand rabbits at ten weeks of age and with bodyweight ranging from 2000 to 2200 g were used in this study. Rabbits were individually housed in a battery cage system with a natural day/night cycle at 21–28 °C. Rabbits had free access to complete commercial rabbit diets (Ibex International Co., Ltd., Nubaria, Egypt) and clean water throughout the study period. Two weeks before the start of the study, the rabbits were acclimatized.2.3. Surgical ProceduresFigure 1 illustrates the surgical procedures. Rabbits were anesthetized via the intramuscular injection of 5 mg/kg−1 bwt of xylazine hydrochloride (Xylaject, Adwia, New Cairo, Egypt) and 40 mg/kg−1 bwt of ketamine hydrochloride (Ketamine, Sigma, Cairo, Egypt) [23,29]. The right hind limb of each rabbit was prepared for aseptic surgical intervention. After a craniolateral skin incision, the biceps femoris was caudally retracted to expose the vastus lateralis muscle. The fascial septum of the vastus lateralis was incised, followed by a reflection of the vastus lateralis from the surface of the femur to expose the femoral diaphysis [30,31]. The fracture was made via the exposure of the femoral diaphysis using the surgical Saw Bojin BJ4101, and K-wire (size 1.2 mm) was introduced into the bone for fixation. The suturing of the muscle layer was performed with simple continuous absorbable polyglycolic acid sutures that were braided and coated with violet 4/0 (SURGICRYL-PGA, SMI AG, Vith., Belgium), and then the skin was sutured using simple interrupted absorbable polyglycolic acid sutures that were braided and coated with violet 2/0 (Surgicryl PGA, SMI AG, Vith., Belgium) [30]. The bandage was applied directly after surgery and kept for one week using a soft cotton pad, tongue depressor, gauze, and self-adhesive tape. All groups were injected with 10000 IU/kg bwt of penicillin and 10 mg/kg bwt of streptomycin (Pen & Strep Norbrook, West Sussex, UK) once daily Subcutaneous (S/C) for 5 successive days.2.4. Animal GroupingAnimals were randomly divided into 3 groups as follows: a control group in which animals received saline solution as a placebo for 5 successive days, an FM group in which animals were intramuscularly injected with 2 mg/kg bwt of FMne (Finadyne®, MSD Animal Health, Walton, UK) once daily for 5 successive days, and a ketoprofen group in which animals were intramuscularly injected 2 with mg/kg bwt of ketoprofen (Ketolgin, Amoun, Al Obour, Egypt) once daily for 5 successive days [1,21,23].2.5. Clinical EvaluationAnimals were clinically evaluated for body temperature, food intake, and daily activity during the week after surgery until external bandage removal. After this period, the animals were evaluated twice a week throughout the experimental period. The evaluation technique consisted of inspecting the external wound for dehiscence, edema, or pus; palpating the limbs by estimating the pain and muscular atrophy of the quadriceps muscle group of the operated limb; and comparing the operated limb with the contralateral limb [29].2.6. RadiographyRadiography was performed using GXR 52 S, an X-ray apparatus (DRGEM Co., Ltd., Seoul, Korea). The X-ray beam was applied using the following factors: 40 KV, 25 MA, and 1.25 MAS. Two radiographic views (ventrodorsal and lateral) were taken. The X-ray was performed before and after surgical procedures at 0, 2, 4, 6, and 8 weeks after operation, as mentioned in [1].2.7. Radiographic ScoringThe scoring system of this study was based on the point system described in [30]. It involved the quantitative assessment of periosteal reaction, bone union, and remodeling. Periosteal callus formation was given 4 when it was full, 3 for a moderate reaction, 2 for mild reaction, 1 for a minimum reaction, and 0 for no periosteal callus formation. The bone union was designated as 3 for complete bridging, 2 for moderate bridging more than 50% of the gap, 1 for slight union less than 50% of the gap, and 0 for nonunion. Bone remodeling was given 3 for advanced remodeling, 2 for moderate remodeling, 1 for early remodeling, and 0 for no remodeling. Images were scored by an independent radiologist who was blinded to the animal’s treatment.2.8. HistopathologyAnimals were euthanized by overdose anesthesia 4 and 8 weeks after the operation. At a rate of 4 specimens per each group (1 specimen/animal) during two time points of treatment, the specimens were subjected to the histopathological examination of osteogenesis at the fracture site using hematoxylin and eosin (H&E) stain and Masson’s trichrome stain [32]. Histopathological grading was performed according to the Lane and Sandhu histopathological scoring system, modified by Heiple et al. [33] and Bigham et al. [34], who used 20-point grading score to assess proximal union, distal union, cancellous bone, cortical bone, and marrow (4 points for each category).2.9. Immunohistochemical ProceduresThe immunohistochemical staining procedures followed that of Saber et al., 2019 [35]. Sections were dewaxed and immersed in a solution of a 0.05 M citrate buffer (pH 6.8) for antigen retrieval. These sections were treated with 0.3% H2O2 and protein block, and they were incubated with polyclonal rabbit anti-VEGF antibody (cat. no. PA5-16754, dilution 1/100, Thermo Fisher Scientific, Waltham, MA, USA). After being rinsed with phosphate-buffered saline, they were incubated with a goat anti-rabbit secondary antibody (cat. no. K4003, EnVision+™ System Horseradish Peroxidase Labelled Pomer, Dako, Thermo Fisher Scientific, Waltham, MA, USA) for 30 minutes at room temperature. Slides were then visualized with a DAB kit and eventually stained with Mayer’s hematoxylin as a counterstain. The staining intensity was assessed on a threshold basis using ImageJ Analysis software (NIH, Bethesda, MD, USA) and is presented as the percentage of positive area per mm2 in about 8 high power fields [35].2.10. Statistical AnalysisData were analyzed using GraphPad Prism 6 statistical software (GraphPrism Software, La Jolla, CA, USA). The statistical significance among groups was tested by a one-way analysis of variance (ANOVA) using the nonparametric Kruskal–Wallis H test followed by Dunn’s multiple comparison test. The p value was considered significant at less than 0.05.3. Results3.1. Radiographic FindingsTable 1 presents the scores of radiographic findings according to the timeline. The radiographic findings of the control group at 2 postoperative weeks showed mild soft callus formation, with the noncomplete bridging bone union increasing over time. Meanwhile, the ketoprofen group revealed weak callus formation without a bone union. The FM group exhibited moderate soft callus formation with a noncomplete bridging bone union that increased over time (Figure 2). The callus formation score in the FM group (3.14 ± 0.51) was significantly higher (p < 0.05) than those in the ketoprofen group (2 ± 0.84) and the control group (2.5 ± 0.54). In stark contrast, callus formation was significantly lower (p < 0.05) in the ketoprofen group (2 ± 0.84) than that in the control group (2.5 ± 0.54). Importantly, the bone union score in the FM group (2.40 ± 0.40) was significantly higher (p < 0.05) than those in the ketoprofen group (0.83 ± 0.40) and the control group (1.66 ± 0.51). However, the bone union was significantly lower (p < 0.05) in the ketoprofen group (0.83 ± 0.40) than that in the control group (1.66 ± 0.51). At 4 postoperative weeks, the control group showed mild callus formation and bone union, with mild to moderate bridging between the bone edges. The ketoprofen group showed less bone callus formation than the other 2 groups and nonunion between the bone edges. The FM group showed greater mild soft callus formation than that in the control group and complete bone union. The callus formation score in the FM group (3.13 ± 0.51) was significantly higher (p < 0.05) than those in the ketoprofen group (1.66 ± 0.82) and the control group (2.67 ± 0.52). Callus formation was significantly lower (p < 0.05) in the ketoprofen group (1.66 ± 0.82) than in the control group (2.67 ± 0.52). The bone union score in the FM group (2.42 ± 0.41) was significantly higher (p < 0.05) than those in the ketoprofen group (1.17 ± 0.75) and control group (1.83 ± 0.75). In contrast, the bone union was significantly lower (p < 0.05) in the ketoprofen group (1.17 ± 0.75) than that in the control group (1.83 ± 0.75).Based on the radiographic changes at 6 postoperative weeks, moderate callus formation was still present in the control group. The remodeling started in the few animals with moderate to complete bone union. Meanwhile, in the ketoprofen group, the minimal callus continued with a nonunion and no remodeling. However, the good callus started to decrease in size due to the early bone remodeling happening in approximately all FM rabbits with a complete bone union. The callus formation score in the FM group (2.29 ± 0.52) was significantly lower (p < 0.05) than that in the control group (3.17 ± 0.75). Moreover, the callus formation was significantly lower (p < 0.05) in the ketoprofen group (1.83 ± 0.75) than those in the control group (3.17 ± 0.75) and the FM group (2.29 ± 0.52). The bone union score in the FM group (2.46 ± 0.40) was significantly higher (p < 0.05) than those in the ketoprofen group (1.17 ± 0.75) and the control group (2.33 ± 0.52). Moreover, the bone union was significantly lower (p < 0.05) in the ketoprofen group (1.17 ± 0.75) compared to that in the control group (2.33 ± 0.52). Furthermore, the bone remodeling score in the FM group (1.33 ± 0.51) was significantly higher (p < 0.05) than those in the ketoprofen group (0 ± 0) and the control group (0.33 ± 0.52).The callus decreased in most animals in the control group due to remodeling, with a complete bone union at 8 postoperative weeks. However, the fractured bone was close to its normal shape in the FM group because of the great remodeling and good bone union. In the ketoprofen group, there was bad remodeling with nonunion. The callus formation score in the FM group (0.59 ± 0.52) was significantly lower (p < 0.05) than that in the control group (2.67 ± 0.52). Moreover, callus formation was significantly lower (p < 0.05) in the ketoprofen group (0.33 ± 0.52) than those in the control group (2.67 ± 0.52) and the FM group (0.59 ± 0.52). Additionally, the bone union score in the FM group (2.75 ± 0.41) was significantly higher (p < 0.05) than that in the ketoprofen group (0.33 ± 0.52) and the control group (2.66 ± 0.52). Moreover, the bone union was significantly lower (p < 0.05) in the ketoprofen group (0.33 ± 0.52) than that in the control group (2.66 ± 0.52). Likewise, the bone remodeling score in the FM group (2.33 ± 0.41) was significantly higher (p < 0.05) than those in the ketoprofen group (0 ± 0) and the control group (1.33 ± 0.52). Figure 3 depiects the X-ray statistics at different times (2, 4, 6, and 8 postoperative weeks). 3.2. Histological Findings3.2.1. Fourth Week StudyAs illustrated in Figure 4, control animals showed large hematoma with the still-defined junction of the two bone edges. The granulation tissues on both fracture sides showed entangled immature fibroblasts within an edematous matrix, with few collagens, and revealed the presence of few osteoblastic cells. Femur fracture in animals treated with ketoprofen demonstrated a marked decrease in inflammation, with the presence of cisterns of blood enclosed by fibrous connective tissues. Fractured animals treated with FM showed a marked decrease in inflammation signs, with a marked increase in the granulation tissue, revealing marked maturation of the fibrous connective tissues with a marked increase in osteoblastic activity. The newly osseous tissues were well-connected with the spongy bone.3.2.2. Eighth Week StudyAs shown in Figure 5, animals in different groups showed various healing attempts. Control animals showed a decrease in hematomas and inflammation, with increased remodeling and osteogenesis. The cortical bones of both sides are related to immature osseous and cartilaginous callus within a fibro–hyaline matrix. The ketoprofen-treated animals showed a marked decrease in inflammation, with the presence of a gap filled with soft callus that consisted of granulation tissues with a small amount of osteoblastic cell activity. The FM-treated animals revealed mature, organized, thick bands of spongy osseous forming well-connected cortical bones covered with a thick periosteal layer. Figure 6 illustrates the quantitative scoring of bone healing within the four and eight weeks of the study. Marked early union, callus formation, and remodeling occurred in fractured animals treated with FM.3.3. Masson’s Trichrome StainingResults of Masson’s Trichrome Staining are shown in Figure 7. In the 4th week, the animals showed a variable amount of collagen. Control animals showed collagen deposition around the hematomas. Meanwhile, collagen deposition decreased on the margin of both sides of the fracture within the ketoprofen group. Interestingly, marked collagen deposition was visible within the FM-treated group (mostly within the early stage of bone healing), though it decreased in the late stage of bone healing. In the 8th week, the control and FM groups showed decreased collagen deposition with increased osseous tissues, while animals in the ketoprofen group showed late collagen deposition at the fracture sites. The quantitative scoring of collagen deposition of the 4th week’s study revealed a significant increase in collagen deposition within the FM group compared to the control group (p < 0.05). The ketoprofen group showed a significant decrease in collagen compared to the control group (p < 0.05). The 8th week’s study showed later collagen deposition in the ketoprofen group than that in the control and FM groups (p < 0.05).3.4. Immunohistochemical FindingsVEGF ImmunostainingAs illustrated in Figure 8, the FM-treated group showed a marked expression of VEGF antibody within the soft callus tissue within the formed osseous cells and granulation tissues. Meanwhile, the ketoprofen-treated group showed a significant decrease in VEGF immunostaining within the granulation tissues in the 4th week study. In the 8th week, the expression was also increased within the bone marrow tissues in the FM group, whereas the ketoprofen group showed a mild expression of the VEGF within the hard callus tissues. The quantitative scoring of positive area expression of the VEGF antibody demonstrated a significant increase in VEGF immunostaining within the FM group at the 4th and 8th week studies compared to the control group (p < 0.05). There was also a significant decrease in VEGF expression within the soft and hard callus tissues of the ketoprofen group upon both sacrifice times (p < 0.05).4. DiscussionBone healing could be affected by multiple factors, such as postoperative treatment [36,37,38,39]. NSAIDs are commonly used postoperatively, but they might have diverse effects on bone healing [3,4,7]. Some delay the bone healing process [11,40,41], while others do not affect bone healing [12]. For the first time, this study has revealed the potential effect of one of the NSAIDs, FM, on bone fracture healing by enhancing angiogenesis at the fracture site, as reflected by VEGF expression and increasing osseous tissue formation. However, ketoprofen, like other NSAIDs, causes a delay in bone fracture.Importantly, serial radiographs have been used to monitor the progress of bone fracture healing [42], including aspects of bone callus formation, bone union, and bone remodeling [30,43,44]. In this study, the radiographic findings revealed that FM improved fracture healing by improving callus formation, bone union, and bone remodeling. In stark contrast, ketoprofen delayed bone fracture healing by decreasing bone callus formation, bone union, and bone remodeling. The study’s findings are in harmony with previous studies that reported the delayed effect of ketoprofen on bone fracture healing in rabbits and cats [40,41].Furthermore, good bone healing requires suitable stability, blood, and oxygen supply [44]. Several previous reports on different species (rabbits and rats) have shown the effect of FM on skin healing [24,45]. These studies proved that FM has adverse effects on the inflammatory phase of wound repair but not on the proliferative phase when fibroplasia is a major factor in wound strength [24]. Moreover, FM was not found to adversely influence breaking strength and wound contraction [45]. A previous study induced fibroblast proliferation in tendon injury, which fastened the tendon’s healing process [23]. In this study at 4 postoperative weeks, the histological results showed that FM had the same effect as that of the control and ketoprofen groups on osteoblast by increasing osseous tissues’ expression at the fracture site.A bone matrix consists of nonorganic and organic components [46,47,48,49]. Collagen is the most copious protein in bone and represents approximately 90% of its organic matrix [49,50]. The collagen expression at a fracture site indicates osteogenesis of the mesenchymal stem cells (MSCs) [48,49]. In this study, the FM group showed a higher expression of collagen fibers at 4 weeks than those of the ketoprofen and the control groups, which were decreased by the effect of remodeling at 8 weeks.In this study, ketoprofen stimulated fibroblast aggregation at the fracture site. The mechanism by which ketoprofen causes this fibroblast aggregation is still unknown and needs further research. It seems that it could be beneficial to soft tissue. However, in bone, it might hinder healing, as evidenced by the higher expression of fibroblast at the fracture site in the ketoprofen-treated group compared to the control and FM groups at the same stage (4 postoperative weeks). However, the study results contrasted with some previous reports that revealed that ketoprofen induced apoptosis in fibroblasts [51]. Furthermore, FM was found to enhance the healing process in contrast to other NSAIDs such as ketoprofen, which delayed this process through its influence on COX-1 and COX-2 enzymes. Some previous reports have revealed that prostaglandin E2 (PGE2) might regulate osteoblast behavior via the relative expression of the receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, which is regulated through COX-1 and COX-2 enzymes. The inhibition of COX isozymes and the consequent decrease in PGE2 may be the mechanisms by which NSAIDs delays bone healing [4].The available literature suggests that FM does not affect neovascularization, while ketoprofen reduces neovascularization [39,52,53]. More importantly, vascular supply at the fracture site is an essential step in the healing process since a good blood supply means plenty of the oxygen and nutrients required for fast healing [54]. Furthermore, the key physiological regulator of angiogenesis during embryogenesis and skeletal growth is VEGF [54,55]. A previous study revealed that FM stimulates angiogenesis in the wound site, enhancing tendon healing [23]. Similarly, in this study, FM was found to stimulate angiogenesis at the fracture site, which was obvious from the expression of VEGF at the fracture site in the FM group. However, no expression was reported in the other groups, suggesting that the FM group enhanced bone healing and the control and ketoprofen groups did not.5. ConclusionsGiven the present findings and their direct clinical applications, the use of FM after orthopedic surgery as a pain killer and anti-inflammatory is recommended. This study has proven for the first time that FM can enhance the bone healing process, callus formation, bone union, and remodeling in rabbits. Meanwhile, ketoprofen has an adverse influence on the bone healing process. The main limitations of the present study included its experimental fracture model that created osteotomy, significant violation of natural biology to cause the fracture, limited number of samples, short-term doses of ketoprofen, and lack of information regarding the detailed mechanisms by which ketoprofen delays healing. It would be interesting to investigate the effect of the short-term and long-term administration of ketoprofen on fracture healing to explain the main mechanisms by which ketoprofen may influence non-bone union and delay bone fracture healing. Future research to explore the mechanistic pathways involved in the underlying effects of FM on bone healing is also warranted.

animals : an open access journal from mdpi

10.3390/ani11071969

PMC8300195

Biosecurity advice is an important way veterinarians can help farmers to reduce disease burdens on their farms. Many different factors are at play when delivering this advice, one being veterinary competence and their communication skills. This study looked at the private veterinary practitioners’ perceptions of their own competence to deliver biosecurity advice as part of a longitudinal biosecurity project spanning two years. Their responses were collected in the form of a telephone questionnaire. The results showed that as the project progressed the private veterinary practitioners felt more comfortable, better capable, and more consistent in giving their advice. In addition, they felt the uptake of their advice by the famers had improved throughout the study period. The mean average time spent delivering biosecurity advice increased and dropped subsequently, suggesting an initially more thorough process, and later a more efficient process. The results suggest development of the participating veterinarians following the conscious-competence learning model, showing a need to improve the knowledge and training of future generations of vets in the area of biosecurity with an increased focus on the importance of the veterinarian-farmer relationship in particular.

Biosecurity advice is an important way veterinarians can help farmers to reduce disease burdens on their farms. Many different factors are at play when delivering this advice, one being veterinary competence and their communication skills. This study looked at the private veterinary practitioners’ perceptions of their own competence to deliver biosecurity advice as part of a longitudinal biosecurity project. Their responses were collected in the form of a telephone questionnaire. The results showed significant increases in private veterinary practitioners’ responses to comfort (p = 0.022), capability (p = 0.002), and consistency (p = 0.006) as well as an increase of uptake of advice (p = 0.015) as the project progressed. The mean time spent delivering biosecurity advice increased and dropped subsequently, suggesting an initially more thorough and later on a more efficient process. The overall perceptions of the veterinarians of the study were also assessed. The results suggest development of the participating veterinarians following the conscious-competence learning model showing a need to improve the knowledge and training of future generations of private veterinary practitioners in the area of biosecurity with, in particular, an increased focus on the importance of the veterinarian–farmer relationship.

1. IntroductionBiosecurity is the practice aimed at keeping infectious diseases from populations and a key part of safe and efficient farming. It allows farmers to increase welfare of their animals and reduce production losses due to disease. Some farming sectors, such as the commercial pig industry, have imposed stringent biosecurity measures and shown the importance of keeping disease at bay in these large scale production systems and reducing reliance on antimicrobials [1,2,3]. In the cattle industry, however, there are marked differences. On dairy farms, private veterinary practitioners (PVP) are more involved on a regular basis with the management process of larger groups, ranging from fertility to mastitis and infectious disease control [4]. In commercial beef production more extensive systems are employed with less intense PVP involvement with their efforts mainly focusing on reproductive management.Using PVPs as a means to spread awareness of the importance of disease prevention on farm enterprise has been deemed a key way for governments to spread policy regarding biosecurity [5,6]. Government agencies in the United Kingdom (UK), such as Department for Environment, Food & Rural Affairs (DEFRA) and the Animal and Plant Health Agency (APHA), use government appointed veterinarians to ensure PVP and farmers are following regulations, for example with auditing bovine tuberculosis (bTB) tests [7]. There is some difference of opinion as to whether this is the best way to increase farmer uptake of biosecurity measures, in particular related to non-notifiable, endemic diseases. Indeed, farmers may resent the use of government veterinarians to regulate [8]. These professionals may be perceived as being detached from the issue and not having the farmer’s best interests at heart, which would suggest that it is not the most effective way to increase uptake of biosecurity measures. PVPs, who have a relationship with the farmers, are far more likely to persuade farmers and increase preventative measures on farms [8].PVPs, however, may face challenges when trying to give biosecurity advice to clients. Traditional farming methods and farmer-to-farmer communication can play an important role in forming opinions towards disease control. For example, the UK Randomised Badger Culling Trial (1998–2005) aimed to control the spread of Mycobacterium bovis from badger populations to cattle herds [9,10]. The relative success of the proactive culling program in high-risk areas showed a decrease in bTB cases in cattle and therefore demonstrated to farmers the success of this approach [9,10]. In high risk areas, such as Wales and the South West of England, a proactive badger cull gained support from local farmers as bTB is a constant battle for them [11]. The success of this culling trial has led to them deciding as a community that this is the way forward in control of this disease. A PVP may be faced with farmers who as a community want to adopt a single solution approach to biosecurity issues. Although it will help with building trust in the PVP and the advice given, it is not a realistic option in many endemic diseases as there are farm specific risks that need to be considered.Because of their relationship with the farmer, a PVP may be able to play a key role in providing a greater variety of biosecurity advice. Farmers like to see results and the PVPs working closely with them can use this by explaining how more stringent biosecurity advice can bring benefits to the farmer. By showing these capabilities of the PVP the best possible outcomes for the farmer can be achieved [12,13]. This outcome is more likely when there is sufficient self-confidence in the PVPs’ own competence. In addition, an understanding of how to discuss disease risks with the farmer is essential. Being able to communicate concise, clear, and practical advice regarding these measures is important for uptake [14]. PVPs in the agricultural sector need to be given the training and support needed so that they understand the impact they can have. In return, they themselves will get a vested interest in the eradication process of disease [14]. Veterinary tact and technique with the individual farm clients is required, combined with well thought out biosecurity measures that can be tailored to individual circumstances. Compared to standard advice, this farm specific approach could provide a far more effective in increasing biosecurity on farms and therefore decreasing risk of disease [13].The professional development of PVPs providing biosecurity advice could follow the learning cycle of the conscious-competence model [15,16]. This learning model is used in the development of clinical reasoning. At the first stage a naïve (unconscious incompetent) outlook is proposed, followed by a conscious incompetence and conscious competence and ultimately after more exposure unconscious competence would follow [15,16]. In order to gain insights to the perceived competence of PVPs advising on biosecurity, PVPs in a biosecurity project on beef suckler farms were interviewed in this longitudinal trial [17]. Aspects related to their perceptions as well as their performance (biosecurity score and visit time spent) were collected by structured questionnaire with the aim to explore the attitudes of PVPs working on the project and to suggest some answers to the challenges mentioned above.2. Materials and Methods2.1. Participant RecruitmentThe data being analysed in this study was collected as part of a larger biosecurity project carried out by the Royal Veterinary College [17]. The biosecurity project focused on five endemic cattle diseases in the United Kingdom, and had 10 different veterinary practices as agents in the field; the practices recruited on average 12 beef suckler farms across Wales and the South West of England.2.2. Biosecurity ScoringAn initial meeting to explain what the project was about, followed by an expert opinion workshop style of discussion to go through all aspects of risks associated with disease spread onto and within a cattle farm: biosecurity. The ten participating PVPs were briefed on previous work in which a farm-specific computer-based risk scoring tool had been developed. Evidence on generic risk factors for disease introduction on cattle farms was used to create generic risk factor categories: cattle purchasing, direct and indirect contact with other cattle, ruminants and other animals, use of shared equipment and types of visitors to the farm. These broad categories were divided into sub-factors to provide more detail. To elucidate risk factor weightings, the PVPs took part in two expert opinion workshops. During the workshops, PVPs were asked to allocate weights to reflect he relative importance of specific sub-factors, such that the total weight of all sub-factors with each broad risk factor category would be 100%. This participatory approach resulted in a reasonable agreement amongst the PVPs involved.A semi-Delphi approach was used in order to achieve near-consensus, to achieve this the median scores were reported to the PVPs and discussed in detail, after which they scored the (sub)factors once again. The subsequent scores were used to construct a scoring tool in MS Excel™. The underlying algorithm generated an overall biosercurity score, with higher risk for disease introductions or spread and a lower scores for more biosecure units. The overall biosecurity score is the sum of factors contributing to the overall risk and the spreadsheet identifies the main risk contributor. This allowed farmers and PVPs to identify specific factors that could be targeted for change during the following year, and by altering these factors an aspirational score could be generated. Before the scoring tool was used on the farms, training was provided for participating PVPs, to familiarise them with the spreadsheet and to address any concerns.The PVPs visited the farms annually in winter to take blood samples to identify which of the five endemic diseases are currently active on the farms. A risk assessment visit was then booked in spring where the farms were scored, the blood results were discussed and a plan of risk reduction was set out and agreed upon. The tool and the scores can be found in the supplementary materials as well as the technical results of the biosecurity project are reported previously [17].2.3. Participant PerceptionsAlongside the biosecurity scores, specific data for this study was collected in the form of a questionnaire. The questions related to the individual PVP’s confidence and perceptions about delivering biosecurity advice at different stages of the study. The stages were, before the study commenced (before), one year into the study (Year 1), and two years into the study (Year 2). The answers to the questionnaire were collected by the author during phone interviews with the PVPs who took part in the study. Before the data collection took place, the questionnaire was trialed on a PVP not involved in the biosecurity project, ensuring the questions were worded appropriately.Based on aspects of service management and the conscious-competence model a questionnaire was developed [15,16]. The questions relating to the PVPs’ perceptions that were repeated for each stage were marked on a scale of 1–10, 10 being fully able to identify with specific aspects of giving biosecurity advice. Aspects covered were: how comfortable, capable, and consistent the PVPs felt when giving advice. Also, the level of discretion the PVP felt when tailoring farm specific advice on biosecurity as well as the perceived uptake by the farmers of their biosecurity advice.The PVPs were additionally asked to rate the overall impact on themselves as a result of the project participation: improved or increased knowledge, interest, involvement, likely role model or likelihood of charging for biosecurity advice. The project impact was graded using a five-point Likert scale (1 = ‘Strongly Disagree’ to 5 = ‘Strongly Agree’).The amount of time spent on providing biosecurity advice by the PVP to the farmer was also recorded (in minutes) throughout the study. Time spent was captured in three measures: the shortest time, the mean time and the longest time taken to score the farms and give biosecurity advice. Whether the participating PVPs would be charging for biosecurity advice was captured in ‘No’, ‘Occasionally’, and ‘Always’ prior to the project as well as for the subsequent years.Finally, themes were identified from the free text that the participating PVPs provided on the benefits and the disadvantages of taking part with the biosecurity project. Due to the limited number of participating PVPs, the thematic analysis of the free comments was performed manually. The questionnaire is attached in the Appendix A.2.4. Statistical AnalysisThe responses to the questionnaire were entered in a Microsoft Excel spreadsheet and transferred to SPSS (Version 26, IBM) for further statistical analysis. Normally distributed data were reported in mean and standard deviation (SD) and non-parametric data reported as median and interquartile rage (IQR). Spearman rank correlations were calculated between the responses in the different respective years. A Friedman test was used to evaluate the change in perception over time. A post hoc Wilcoxon Signed Rank Test was carried out to compare the respective years. A difference was considered significant if the p-value was lower than 0.05, after Bonferroni correction.Evaluating the impact of the perceptions of the PVPs, as well as their time spent doing the biosecurity visits on the PVP’s specific biosecurity score, measured as mean score, the standard deviation (SD) of their scores and the coefficient of variation (COV). This allows exploring whether the biosecurity scores were getting more uniform (lower SD and COV) or whether PVPs felt more able to use the full breath of the scoring tool. These biosecurity measures were the dependent variable and tested by running a mixed linear model with PVP as mixed effect with year and the questionnaire responses as fixed effects. A multivariate approach was taken with a backwards stepwise approach, removing the least non-significant fixed effect variables until all variables were significant. Finally, the overall project participation was evaluated (improved or increased knowledge on, interest in, and involvement with biosecurity of cattle farms, likely to perceive themselves as a role model with respect to biosecurity for the farmer or likelihood, and charging for biosecurity advice) using the participant’s mean, SD and COV biosecurity score comparing ‘strongly agree’ responses versus lower Likert scores with a t-test.3. ResultsOf the ten biosecurity project PVPs, one veterinarian provided their perceptions before the study commenced and one year into the study, but not for the second year. We received complete data from eight participants, each from a different practice, giving a response rate of 80 percent. The eight PVPs represented 90 farms visits where biosecurity was scored and advice was given. The mean scores on these biosecurity visits was 150 (SD = 42) and it took them on mean 83 min (SD = 42). The shortest reported visits lasted on mean 56 (SD = 31) minutes and the longest 119 (SD = 54) minutes in duration. The mean time spend on farm scoring for biosecurity and giving advice on project farms on the two rounds of the project compared to their initial reported time spent giving biosecurity advice. The time spent on giving biosecurity advice went from 21 +/− 7 (mean +/− SD) minutes before the biosecurity project, to 93 +/− 42 min in year one (p < 0.001) and 69 +/− 39 min (p = 0.017) in year two of the project.3.1. Participant PerceptionsThe overall reported perceptions of the participating PVPs before (baseline), year one and year two reported values for ‘Comfortable’, ‘Capable’, ‘Consistent’, ‘Discretion’ and the level of advice ‘Uptake’ was (median, IQR): 8 (7.25–9), 8 (7.125–8), 8.25 (7.25–9), 8 (6.25–8.5), and 5 (5–7) respectively. The perceptions of the PVPs over the span of the biosecurity project are presented in Table 1 below. In all bar the level of discretion, the perception of these PVPs was graded higher as the project progressed (p < 0.05). The participating veterinarians reported charging for biosecurity advice prior to the project as ‘Occasionally’, (IQR: No-Occasionally), this was moved towards ‘Always’ (IQR: Always-Always). This shift was significant when evaluating this with the Wilcoxon Signed Ranks test (p = 0.030). Although the sense of being a role model for biosecurity moved from ‘Agree’ (IQR: Agree-Strongly agree) to ‘Strongly agree’ (IQR: Agree-Strongly agree), this was not significant (p = 0.102). The perceived role model prior to the project had a significant correlation with the level of comfort in giving biosecurity advice, r = 0.735 (p = 0.024), and again at Year 2 with the post role model feel: r = 0.716 (p = 0.046). There was, however, no correlation with comfortability of giving biosecurity advice during Year 1 of the project.There were a few correlations (Spearman Rank) identified, which were not the same throughout the progress of the project. Before the project, ‘Consistency’ of the advice and the ‘Capability’ of the PVPs to give the advice showed a positive correlation coefficient of (r = 0.761) which was significant (p = 0.017). In year one, there was a positive correlation between the ‘Uptake’ of the advice by the farmer and the ‘Comfort’ of the PVPs delivering the advice (r = 0.772, p = 0.015). Furthermore, ‘Capability’ and ‘Comfort’ of the PVPs also showed a moderate positive correlation with (r = 0.670, p = 0.048). Finally, ‘Capability’ delivering information and ‘Consistency’ of advice revealed a positive correlation, (r = 0.767, p = 0.016). For year two, no significant correlations were found amongst the reported parameters.3.2. Biosecurity ScoreThe overall mean biosecurity score of years 1 and 2 was 146.4, with a standard deviation of 86.9 and a coefficient of variation of 0.568. The linear mixed effects models revealed that for the mean biosecurity score, the second year had an mean 34.5 lower score and longer visits resulted in higher scores (p = 0.001). In addition, the mean time spent on the biosecurity consult visit resulted in a higher score (0.7/minute, p = 0.007), however, when combined in the multivariate approach, year was the only variable remaining in the model. The standard deviation of the biosecurity score was lower in PVPs reporting feeling more comfortable in giving biosecurity advice (−11.7 per score point, p = 0.042). Taking into account both the standard deviation and the mean, the coefficient of variation (SD/mean) increased with 0.128 per score of sense of capability (p = 0.003) and 0.059 per score of sense of discretion (0.024). The multivariate approach was left with capability of being the only significant parameter in the model.The mean, SD and COV biosecurity score in year 1 was 163.4, 91.5, and 0.540 and in year 2129.4, 82.4, and 0.596 respectively. The mean score changed significantly (p = 0.001), the other measures of the biosecurity score did not (p = 0.131 and p = 0.100 respectively). Comparing the responses of the overall project participation (improved or increased knowledge, interest, involvement, likely to perceive themselves as a role model with respect to biosecurity for the farmer or likelihood of charging for biosecurity advice), responses of ‘strongly agree’ versus not, there are a few significant differences in the biosecurity scores. The mean score in year 1 was higher in PVPs that felt strongly that they had an increased interest in biosecurity: 177.1 vs. 135.9 (p = 0.037) and this was also the case in year 2: 147.5 vs. 93.1 (p = 0.019). Similarly, the SD and COV of their scores was higher in both years: SD year 1: 109.8 vs. 54.8 (p = 0.012), SD year 2: 102.7 vs. 41.6 (p = 0.023), COV year 1: 0.608 vs. 0.402 (p = 0.030) and COV year 2: 0.669 vs. 0.450 (p = 0.052). For the other overall project scores only improved knowledge had higher mean biosecurity scores in year 2 in PVPs that strongly agreed and the ones who did not: 165.1 vs. 111.5 (p = 0.045). The main themes on the benefits in taking part with the biosecurity project are an improved level of awareness, knowledge and interaction in both PVPs and farmers, testing for disease allowed opening up the discussion. The involvement in the biosecurity project allowed identification of disease presence on the farm and through that helped the case for vaccine sales or disease eradication. More specifically, five of the participating PVPs reported that they “increased their involvement on farm” and “farmers improved understanding of risks”. However, one PVP mentioned that “uptake wasn’t great” and three other PVPs mentioned they felt they had to “hassle farmers”, suggesting that individual experiences did vary. Overall, there were fewer disadvantages reported than advantages, and they mainly pivoted around the feeling of the need to persuade the farmers to take part.4. DiscussionThis study describes the changes in perceptions of PVPs that have taken part in a biosecurity project where they scored the level of biosecurity and gave biosecurity advice on beef cattle farms. The success of the use of the actual tool supporting the advice has been reported earlier [17] and has been echoed by other research groups as well [18]. The use of a structured questionnaire or tool to assess the level of biosecurity reduces observation bias. There are inevitable biases in the perception questionnaire described in this paper. Recall, conforming, and reporting bias during the collection of the PVPs’ perception, which necessitate caution with interpreting the results.The current findings show an association between the training during the project and their ability to deliver advice effectively and eloquently: over the duration of the longitudinal biosecurity project, the time spent initially increased, dropping again the next year. The initial increase suggests that more thorough visits were taking place and more comprehensive advice was being given later on. This would explain the finding that visits that lasted longer had higher scores on biosecurity. Time spent on farms was, however, not significant in the multivariate model, with the number of years into the project staying in the model.The improved biosecurity in Year 2 was possibly due to an increased knowledge and interest in the topic of biosecurity, as shown by their questionnaire responses and that they then refined their delivery over time leading to a reduction in time and more concise advice. The more time spent delivering the advice and the more experience gained in the field of biosecurity has led to the increases in comfort and capability [19]. It is also suggestive of greater confidence and willingness to make a change on the farm by the PVPs where organisation and communication skills also come into play: communication needs to be concise and effective to transmit the point across to the clients [14]. The importance of listening is also key in this aspect; poor communication skills lead to misinformation, confusion, and errors [20]. It is reasonable to suggest that the use of the biosecurity scoring tool has helped with the effective listening, recording, and advising. The decrease in time that followed, suggests that the PVPs may have honed their communication skills to provide advice more consistently and more efficiently to farmers as the project progressed, as they gained in confidence and improved their skills when discussing biosecurity [21].As biosecurity visits were becoming time efficient, there was progress made in the overall biosecurity scores that the advisors reported back: a drop in the score represents a better level of biosecurity by increasing the farm’s barriers to disease introduction. The coefficient of variation, that shows the variability of the biosecurity score relative to the mean score, tended to increase at the same time. This was particularly noticeable in PVPs that felt more capable with giving said advice. This suggests that although the overall biosecurity increased, the advice remained variable and therefore specific for the individual project farms, which in turn was reflected in the level of discretion that the PVPs felt when using the scoring tool. This is in line with previous findings, where the ability to provide farm-specific biosecurity advice related with changes in farmer behaviour [22].The PVPs that reported to ‘Strongly agree’ with an increased interest in biosecurity also reported higher biosecurity scores, suggesting a poorer level of biosecurity. Additionally, their variation of the scores across their farms is higher. These findings could be explained by an increased level of engagement and scrutiny during their visit on the farm, making sure all risks are identified and weighted in the scoring tool. This is echoed by the finding that PVPs reporting to have an increased knowledge on biosecurity also returned higher mean scores on their farms. By scoring the farms more thoroughly, the risks are better identified allowing fuller and therefore better communication between of farmers and PVPs. This allows better knowledge exchange and understanding with the farmer, which has been shown to have influenced farming practices [13,23,24].Improved biosecurity resulting in reduced presence of diseases is fundamental to producing better welfare and safer consumables for people [2,25]. In this study, the median scores on the questionnaire obtained from the PVPs before they began the study compared to at the end show increases on all aspects of providing a competent advisory service. All but the level of discretion increased significantly. The perceived discretion could be affected by the use of a biosecurity scoring tool in the project. The structured biosecurity tool may have led to more targeted advice towards areas of biosecurity that needed the most improvement, rather than allowing a large level of discretion in their advice. In terms of biosecurity and disease prevention on farms, poor communication could lead to increased production losses, loss of trust in the PVPs and could harm the relationship between farmer and PVP, making them less likely to listen to the PVP’s advice [26]. This compromise between farmers and PVP seemed to be one of the key aspects to improve. The data showed that the PVPs’ perceptions of farmer advice uptake has increased throughout the project, suggesting that the project or the use of the tool may have reduced the compromise between the farmers and their PVPs. The repeated nature of the project and the shared interest of both parties has possibly contributed to this as literature shows how the relationship between PVP and client is fundamental, especially when it comes to farm clients [8].The increased level of competence through feeling more comfortable, capable and consistent when giving biosecurity advice suggests that the PVPs at the end of the study felt they had improved their abilities. There was also a PVP reported sense of increased uptake of advice by the farmer, which was confirmed by the drop in numerical score on biosecurity. The scores for all categories showed changes, which would suggest that the extra training, advice, and support provided by the project to the PVPs allowed them to develop to be competent in biosecurity out in the field. Of course, many other factors could have an effect here, such as experience, years qualified, and previous training. It is interesting, however, to see that PVPs who perceive themselves to be role models for farmers felt comfortable at the start of the project, as well as in Year 2, but this was not the case in the first year. This could reflect the learning cycle of the conscious-competence model [15,16]. This learning model is used in the development of clinical reasoning. At the first stage a naïve (unconscious incompetent, prior) outlook is proposed, followed by a conscious incompetence (Year 1) and conscious competence (Year 2) and ultimately after more exposure unconscious competence would follow [15,16]. PVPs may be in different phases of their development. This could explain why the effectiveness of utilising PVPs to spread information and educate farmers on biosecurity is well founded but not always necessarily successful [22,27]. This could be alleviated by a problem based practical training of PVPs, to bring them to the unconscious competent phase effectively. This allows PVPs to give high quality biosecurity advice—and charge for it.Overall, the increase of knowledge and interest in biosecurity of the PVPs show that these individuals will feel better prepared to improve biosecurity on farms in the future. Better communication and education of farmers and their advisors has been shown to have influenced farming practices in the past and there is no reason to believe that this is not also the case when it comes to biosecurity [13,23,24]. Furthermore, from the open questions asked to the PVPs at the end of the study, despite feeling the need to hassle farmers, the overall response was positive, appreciating the opportunity to engage with disease control on their beef suckler farms, as on beef farms in the UK the veterinary involvement is relatively low. On dairy farms, there is a programme in the UK that focusses on the control of Johne’s disease (Mycobacterium avium ssp. paratuberculosis) that is using trained and certified practicing PVPs to set out control plans as part of their farm assurance [28]. A similar model could be used to improve the biosecurity training for PVPs. An increased veterinary involvement with disease control on farm, could result in increased welfare and production on farms and an improved veterinarian–farmer relationship.5. ConclusionsThe reported research shows that there is still room for improvement when it comes to delivering biosecurity advice to farmers by PVPs. The professional relationship has been highlighted as a key factor in the promotion of good biosecurity practice. In addition, the increased competence in PVPs allows for an increased knowledge exchange. The study has shown how exposure to enhanced biosecurity training has led to an increase in competence in the PVPs when providing advice. The increased veterinary competence in disease control on farms followed the conscious-competence learning model. The training on biosecurity, the disease testing and advising on farm, and the use of the biosecurity scoring tool facilitated this professional development.

animals : an open access journal from mdpi

10.3390/ani12040531

PMC8868240

The Chinese giant salamander (Andrias davidianus) is one of the largest extant amphibian species, and it is considered critically endangered by the IUCN Red List. Previous studies have demonstrated that future climate change could strongly affect this species. However, how to conduct the related conservation activities are still unclear. Understanding the thermal physiology of A. davidianus is meaningful in guiding its conservation, e.g., habitat selection and preadaptation before population translocation. In this study, the influences of temperature and diet on the metabolic capacity and thermal limits were studied for A. davidianus larvae based on laboratory experiments. Our results indicated prominent physiological plasticity in the thermal tolerance of A. davidianus in response to temperature and diet changes. This thermal plasticity likely, to some extent, buffers the effects of climate change on the Chinese giant salamander. In addition, the potential mechanisms underlying this plasticity were discussed. Our results provide insights for the formulation of conservation strategies for this species.

The Chinese giant salamander (Andrias davidianus), one of the largest extant amphibian species, has dramatically declined in the wild. As an ectotherm, it may be further threatened by climate change. Therefore, understanding the thermal physiology of this species should be the priority to formulate related conservation strategies. In this study, the plasticity in metabolic rate and thermal tolerance limits of A. davidianus larvae were studied. Specifically, the larvae were acclimated to three temperature levels (7 °C, cold stress; 15 °C, optimum; and 25 °C, heat stress) and two diet items (red worm or fish fray) for 20 days. Our results indicated that cold-acclimated larvae showed increased metabolic capacity, while warm-acclimated larvae showed a decrease in metabolic capacity. These results suggested the existence of thermal compensation. Moreover, the thermal tolerance windows of cold-acclimated and warm-acclimated larvae shifted to cooler and hotter ranges, respectively. Metabolic capacity is not affected by diet but fish-fed larvae showed superiority in both cold and heat tolerance, potentially due to the input of greater nutrient loads. Overall, our results suggested a plastic thermal tolerance of A. davidianus in response to temperature and diet variations. These results are meaningful in guiding the conservation of this species.

1. IntroductionTemperature is one of the most important climatic factors for ectotherms. Variations of environmental temperatures can not only impact the fitness of ectotherms directly by reducing the behavior and physiological performance [1], but they also disrupt the homeostasis of ecosystems and cause the spread of pathogens or shrink of habitat [2,3]. The critical thermal limit (Tc, including upper Tc and lower Tc) is the most widely used index representing the thermal tolerance of animals [4,5]. It is defined as the thermal point at which locomotory activity becomes disorganized and the animal loses its ability to escape from conditions in gradually heating or cooling thermal conditions [6]. Many studies suggested that the Tc are linked to species geographical distributions by matching with the local climate [7,8,9,10], making it an important index in predicting the population dynamics of animals under climate change [11,12]. However, the biological processes that determine the thermal tolerances of animals are still controversial. One of the best-known hypotheses is the “Oxygen Limitation of Thermal Tolerance” [13], which claims that a mismatch between the demand for oxygen and the capacity of oxygen supply to tissues is the critical mechanism restricting the whole animal’s tolerance to thermal extremes [14]. The association between aerobic metabolism and upper Tc has been supported by a large body of evidence from fish [15] and arthropods [16]. Moreover, a higher metabolic rate is always associated with better locomotory performance under cold stress [17,18]. These results suggest that the metabolic architecture of animals can provide mechanistic insight into their thermal tolerance.The thermal tolerances of many ectotherms are plastic with the change of the environment. These animals can remodel their cellular processes and structure to reduce the variation degree of physiological activities in response to environmental changes [19]. For example, ant foragers in late summer had higher average upper Tc compared to those in March and December [20]. More evidence is provided from laboratory studies [21,22]. This physiological plasticity is called acclimation or acclimatization capacity. From a metabolic perspective, thermal acclimation induces metabolic compensation in individuals to offset the thermodynamic variations in metabolic reactions [23]. Plasticity enabled by thermal acclimation is expected to broaden the range of temperatures at which animals are active [24] and is often highlighted as a powerful mechanism to buffer the impact of climate change [25,26]. In addition to environmental temperature, variations in other factors (e.g., oxygen level, water level, and nutrient conditions) can also affect the thermal tolerance (called the cross-talk effect) [27,28]. Therefore, it is necessary to consider the plasticity in thermal tolerance at multiple conditions to better understand species’ thermal physiology.Global decline in the amphibian populations is an urgent environmental and ecological problem [29]. The Chinese giant salamander (Andrias davidianus) is one of the largest extant amphibian species, and it is often referred to as a living fossil [30]. This species was once widely distributed in central and southern China [31,32]. In the past 60 years, the wild populations of A. davidianus have declined dramatically due to habitat degradation, pollution, and overexploitation [33,34]. Recently, this species was evaluated as critically endangered by the International Union for Conservation of Nature Red list [33] and Chinese specialists [35]. Its evolutionary and ecological significance makes it a flagship species for biodiversity conservation in China. In particular, the effect of climate change (e.g., rise of temperature) has been considered a severe challenge to its survival in the wild [34,36]. In addition to the climatic factors, empirical evidence from farmers and our laboratory indicate that the diet types (e.g., fish, red worm, and pork liver in the farming industry) and temperature can jointly affect the growth rate, a primary fitness index, of the A. davidianus larvae. It is interesting to know whether the diet type can shape the thermal physiology of these animals. This knowledge can be implicative in their conservation in the context of climate change.In this study, the influences of acclimation temperature and diet on the thermal performance (i.e., metabolic capacity and thermal limits) were studied for the Chinese giant salamander larvae. Our main target was to study the plasticity in the thermal performance of this species. This knowledge may extend our understanding of the physiology of this important species and provide useful information for the conservation of the wild A. davidianus population under future climate change.2. Materials and Methods2.1. Animals and AcclimationLarvae of A. davidianus from the same clutch were purchased. They were cultured under the same environmental conditions in a farm located at Hongya, Sichuan Province, China (103°10′05″ E, 29°52′36″ N). Specifically, the average water temperature was 15 (±1.1 SD) °C, and their main food throughout the first year after hatching was red worm. Once collected from the farm, the larvae were cultured in artificial rearing tanks (length × width × height = 29 cm × 20 cm × 9.7 cm; with 3000 mL water) in laboratory conditions (light: dark = 12: 12; 15 ± 0.5 °C; dissolved oxygen level >90%) for two weeks before treatments. The larvae were fed with sufficient red worm twice a day at 09:00 a.m. and 18:00 p.m., respectively. Water was replaced daily. Animal procedures were approved by the Animal Care and Use Committee of the Chengdu Institute of Biology, Chinese Academy of Sciences.According to laboratory studies, A. davidianus larvae are sensitive to temperature variations [37,38]. The optimal growth occurs in water with a temperature range of 15–21 °C [39]. Their feeding behavior was inhibited by water temperature higher than 25 °C or lower than 8 °C [40,41]. Therefore, three discrete temperature levels (7 °C/cold stress, 15 °C/optimum, and 25 °C/heat stress) were selected for thermal acclimation. The red worm (Limnodrilus sp.) was an empirical diet for captive giant salamanders at the first year after hatching, while wild individuals likely have more diverse prey, including fish fray, frogs, and arthropods. In this study, the red-worm and fish fray were selected to feed the experimental larvae.Two independent acclimation programs were conducted successively to measure the respiration rate and thermal limits, respectively. For the measurement of respiration rate, 210 larvae were collected on their 60th day after hatching. After two weeks of laboratory acclimation, these larvae (1.65 ± 0.3 g, mean ± SD) were randomly divided into six groups: worm diet at 7 °C, worm diet at 15 °C, worm diet at 25 °C, fish diet at 7 °C, fish diet at 15 °C, and fish diet at 25 °C (Figure 1A). Each group included two tanks, and each tank kept 15–20 individuals. The nutrient composition of red worm and fish fray is presented in Table 1. The whole acclimation duration lasted for 20 days. The daily culture followed the conditions described above. The body weight of the larvae was measured at the nearest 0.01 g at the end of treatment, as well as before the measurement of respirate rate. To measure thermal limits, 150 larvae were collected from the farm on their 120th day after hatching. After two weeks of laboratory acclimation, these larvae (3.26 ± 0.5 g, mean ± SD) were randomly divided into five groups (30 individuals for each group): worm diet at 7 °C, worm diet at 15 °C, worm diet at 25 °C, worm diet at room temperature (15–25 °C), and fish diet at room temperature (15–25 °C) (Figure 2A). The whole acclimation duration lasted 30 days. For worm diet at 7 °C, worm diet at 15 °C, worm diet at room temperature (15–25 °C), and fish diet at room temperature (15–25 °C) acclimation groups, 15 individuals were tested for either upper or lower thermal limits. However, only 10 individuals were tested at the lower limit for worm diet in the 25 °C-acclimated group as 5 individuals died during the acclimation.2.2. Respiration RateOxygen consumption of A. davidianus individuals was measured by intermittent respirometer. For each round of tests, the respiration rate of one larva was measured. All individuals were fasted for 24 h before measurement, and the animal could swim freely in the chamber. For any acclimation groups, 11–18 individuals were randomly selected and measured at each test temperature (i.e., 7, 15, or 25 °C). For each individual, the respiration rates at 7, 15, and 25 °C were not measured continuously. Instead, these measurements were separated by recovery intervals (24 h) at the acclimation temperatures to make sure the larvae returned to their initial status. During the recovery intervals, the larvae were placed back in their respective tanks. It should be noted that the larvae from the same tank (more than one individual in each tank) could not be distinguished from each other, as there was no suitable method to label them. This meant that the same larva might be selected to measure the respiration rate at two or three different test temperatures. Thus, to avoid pseudo-replication in the statistical models, the average value of the respiration measurements, which shared the same test temperature, acclimation temperature, and diet type (n = 11–18 individuals or measurements for each condition), was treated as the only replication for each condition. See the raw data in Supplementary files.The measurement began when the larvae became quiet. During measurement, their movements, if any, were transient and limited in small ranges as the chamber is small, and most of the time, these larvae were stationary. The respirometer consisted of one chamber (0.15 L) and one dissolved oxygen analyzer (HQ30d, HACH company, New York, USA). In addition, the intermittent respirometer was equipped with an internal circulation pump to fully mix the dissolved oxygen in the internal water. One chamber in each group without A. davidianus individual was used as a blank control chamber to calculate the background oxygen consumption. Then, oxygen consumption rate (VO2) was measured one time at 1 min intervals for 30 min. The following formula was used to calculate VO2 (mg O2kg−1h−1) [42]:VO2=ΔO2×v/m where ΔO2 is the difference in the oxygen concentration level (mg O2 L−1) between the experimental and blank control chamber, v is the water flow rate in the chamber (Lh−1), and m is the body weight of the A. davidianus individual (kg).2.3. Measurement of Thermal LimitsThe thermal limits of the larvae were measured after acclimation. The larvae were placed in a digital water bath, and a square space (height × width × depth = 23 × 11 × 18 cm) was delimited by a plastic net to avoid the direct touch between animals and the bath. The water temperature was initially set at 20 °C, and it increased or decreased by ~1 °C/min. The larvae exhibited free swimming at the beginning and then exhibited behavioral agitation (struggled to escape the confinement). The water temperature for the onset of escaping behavior was recorded, and it was defined as the escaping temperature (Tesp). Once the larvae exhibited escaping behavior, they were turned over by a glass hook every 1 min. The water temperature was considered to have reached the critical temperature (Tc) once the larvae exhibited a loss of righting response (five seconds). The loss of righting response was characterized by the animal’s inability to correct their orientation after being flipped upside down using a glass rod, while submerged in water [43]. Note that the absolute Tc values might be overestimated, as there was likely a lag in the variation of the body temperature following the variation of the water temperature. Each individual was tested for either lower thermal limits or upper thermal limits.2.4. Statistical AnalysesStatistical analyses were conducted on SPSS v25.0 (SPSS Inc., Chicago, IL, USA). The differences in growth rate between groups were analyzed by ANOVA, with acclimation temperature and diet as two independent factors. The variations in respiration rates were analyzed by ANCOVA and LSD post hoc test, with test temperature as a covariant. The interactive effects between the independent factors were considered in the ANOVA and ANCOVA models. The differences in upper or lower Tc and Tesp were analyzed by Kruskal–Wallis or Mann–Whitney U tests. Graphs were generated by Graphpad Prism 5 or ggplot2, an R package [44].3. ResultsThe interaction between environmental temperature and diet affects the larvae growth rate (F2,87 = 20.356, p < 0.01, two-way ANOVA; Table S1). In worm-fed groups, 7 °C larvae had decreased somatic growth compared to their 15 °C counterparts. However, no significant acceleration in growth was detected between larvae from the 7 °C group and those from the 25 °C group (simple effect analysis; Figure S1). In fish-fed groups, the growth rate tended to increase with the rise of environmental temperature, but the inter-group variations were not significant (simple effect analysis; Figure S1).3.1. Influences of Temperature and Diet Acclimation on Larvae MetabolismAcclimation temperature affect the larvae’s metabolic rate (presented as oxygen consumption rate) independently (F2,11 = 10.987, p = 0.002; Table 2). Cold acclimation enhanced the metabolic capacity of larvae, while warm acclimation reduced their metabolic capacity (p < 0.05, LSD post hoc test; Figure 1B). The larvae metabolism was not significantly affected by diet (F1,11 = 0.468, p = 0.508; Figure 1C).3.2. Influences of Temperature and Diet Acclimation on the Acute Thermal Tolerance WindowThe influences of acclimation temperature and diet on acute thermal tolerance were studied (Figure 2A). Temperature acclimation caused a shift of the acute thermal tolerance windows of giant salamanders (Figure 2B,C). The upper Tesp and Tc of warm-acclimated larvae increased from 28.57 ± 1.2 °C and 32.52 ± 0.8 °C to 32.39 ± 0.5 °C and 38.72 ± 0.6 °C, respectively (mean ± SD, and similarly hereinafter). This was accompanied by an increment in their lower Tesp and Tc from 5.51 ± 1.5 °C and 0.48 ± 0.4 °C to 6.63 ± 1.3 °C and 3.72 ± 0.7 °C, respectively. Cold-acclimation caused opposite changes, with their lower Tesp and Tc decreasing to 4.64 ± 2.1 °C and 0 ± 0.02 °C, respectively, which was accompanied by a decrease in the upper Tesp and Tc. Fish-fed larvae exhibited a wider thermal tolerance window than worm-fed individuals, characterized by increased upper Tesp and Tc and decreased lower Tc (Figure 2D,E).4. Discussion4.1. Temperature-Induced Shift of Thermal LimitsOur results indicated that acclimating A. davidianus larvae to higher and lower environmental temperatures decreased and increased their metabolic capacity, respectively, suggesting metabolic compensation and adaptive plasticity in response to thermal fluctuation. Thermal compensation enables the maintenance of physiological rates across environmental conditions. This was frequently observed in animals facing mildly cold environments [23,45,46]. This biological phenomenon is believed to partly offset the reduced physiological performance under cold and thus allows better exploitation of the environment by maintaining physiological activities such as locomotion, feeding, and development at low temperatures [47]. For animals in warming conditions, their increased metabolic activity due to thermodynamic effects resulted in accelerated resource consumption. Downregulation of metabolism can benefit their long-term survival from the perspective of resource-saving [48,49,50]. Given that amphibians’ larval stage is devoted to somatic growth and energy storage [51], either an increased metabolic rate at a lower temperature or decreased metabolic rate at a higher temperature are aligned with their life strategy.This study measured the thermal limits of A. davidianus larvae in the early developmental stage. It demonstrated their plasticity in thermal tolerance. The ability to acclimatize to changing thermal conditions is expected to be a primary factor that dictates the vulnerability of taxa to climate change. Generally speaking, the plasticity in thermal tolerance means better outcomes for A. davidianus larvae in response to global warming or extreme weather than expected. Currently, the existence and distribution of wild A. davidianus in the future has become a topic that is in great need of research and discernment. The ecological niche model raised by Zhao, et al. [34] suggested that climate change can potentially affect the distribution of these animals. To have a more reliable and refined prediction of the fate of the wild populations, physiological models considering their thermal limits and plasticity should be considered. Our results may provide a foundation for these studies. However, two factors should be still be considered, the variation of thermal tolerance with life stages [4] and the potential differences in thermal traits between different phylogenic clades of A. davidianus [52]. Further studies should focus on these questions, particularly in clarifying which life stage or phylogenic clades exhibit the lowest thermal tolerance and plasticity. This knowledge could be meaningful in guiding the conservation of wild A. davidianus, e.g., making more reliable predictions of the individual survival from different geographical populations and optimizing the trophic structure of their habitats to enhance their tolerance to extreme weather conditions.Making clear the mechanisms underlying the thermal tolerance may guide the conservation of Chinese giant salamander in the context of climate change. Despite the complexity in the mechanisms of acute thermal intolerance, metabolic capacity is always an important contributor to the thermal tolerance of aquatic ectotherms [14,53]. This is particularly true for cold conditions, where the ectotherms have difficulty maintaining their metabolic activities. For example, the atu mutant Drosophila melanogaster has increased metabolic capacity, which improves its cold tolerance [54]. Accordingly, variations in metabolic capacity explained the improved and weakened cold tolerance in cold- and warm-acclimated individuals, respectively. For heat tolerance, as the activities of most biological reactions increase with ambient temperatures within certain thermal scopes, the metabolic capacity may no longer be a major limiting factor for heat tolerance. Instead, the overactive metabolic capacity can limit heat tolerance by overburdening the oxygen supply [13,14]. Additionally, it was reported that downregulation of metabolism could improve the acute heat tolerance of animals by reducing their requirement for oxygen [55]. Accordingly, decreased metabolic capacity after heat acclimation is likely associated with higher upper limits and vice versa. Taken together, the metabolic rate may be an important factor for predicting the tolerance of A. davidianus to extreme temperature. In conservation practice, to improve the survival rate of reintroduced populations, we might screen individuals whose thermal tolerance matches the climatic characteristics of the translocation habitats, e.g., introducing cold-tolerant individuals to regions with severe winters or introducing warm-tolerant individuals to hot environments. As direct measurement of the thermal tolerance is detrimental or even lethal to animals, the metabolic rate can be an applicable indicator for their thermal-tolerance properties.4.2. Diet-Induced Shift of Thermal LimitsUnlike the temperature acclimation, which induced a unidirectional shift of thermal limits, diet acclimation broadened the thermal-tolerance window of A. davidianus larvae in both directions. Diet did not change the metabolic capacity of giant salamander. It implied different mechanisms between diet and temperature acclimations in affecting thermal tolerance. Nutrition was reported to modify the critical thermal limits of ectotherms [56,57]. Despite the effects of nutrition depending on the animal size or thermal conditions [58], most studies supported that nutrient supplements (e.g., carbohydrates and amino acids) can improve thermal tolerance in animals [56,59,60,61,62,63,64]. This is reasonable, as rich nutrient storage can improve cellular metabolic maintenance and benefit the synthesis of protectants, which are necessary for survival in stressful conditions. In our study, the most prominent difference between fish and red-worm diets was that the former was richer in total energy, protein, and lipid levels (Table 1). This might be a reason for the superior of fish-fed larvae in thermal tolerance. This finding suggests that the productivity or prey abundance of the natural habitats could be a potential determinator for the thermal tolerance of wild A. davidianus. Introducing suitable prey with great nutrient loads to the natural or artificial habitats of A. davidianus may be an alternative approach to reduce the impact of climate change and extreme weather on these animals.Some mechanisms by which the nutrients modify the thermal physiology of animals were revealed. For example, it was reported that the storage of carbohydrates, the anaerobic substrate, is associated with the tolerance of animals to extreme heat stress [16,65] when the oxygen supply cannot match the metabolic requirement [14]. Although the fish diet contains less sugar than the worm diet, the protein level in the former is much higher. Amino acids derived from proteolysis can be easily converted into carbohydrates. For cold tolerance, the abundance of lipids should be important, as these compounds can be major metabolic substrates in cold conditions [47,66]. More importantly, lipids are required for cellular membrane remodeling, which is an important mechanism underlying cold tolerance [67]. Accordingly, the rich lipid in the fish diet might contribute to the cold tolerance of fish-fed larvae. Overall, these findings shed light on the new thoughts referring to the conservation of this species under climate change. Currently, the reintroduction of captive-bred individuals to the historical natural habitats has been an important measure for the recovery of wild giant salamander populations [68]. Since the diet is a determinant of thermal performance, the prey abundance and diversity in the habitat should be considered before reintroduction. Moreover, to improve the survival rate of the released giant salamander, a preadaptation procedure should be conducted before reintroduction.5. ConclusionsIn this study, we demonstrated the plasticity in thermal physiology of A. davidianus larvae in response to environmental temperature and diet changes. These larvae exhibited apparent thermal compensation in metabolic capacity after thermal acclimation. Specifically, cold- or heat-acclimation improved their tolerance to more extreme thermal stress but compromised their tolerance to the opposite thermal extremes. Diet did not affect the metabolic rate; however, the fish diet, which was richer in protein, lipid, and total energy, broadened the thermal-tolerance window of A. davidianus larvae. This knowledge provides some implications in the conservation of this endangered animal.

animals : an open access journal from mdpi

10.3390/ani12030393

PMC8833368

The functional SNPs discovered in this work will give helpful information on the crucial molecular markers that may be employed in breeding efforts to improve the heart development of broiler chickens.

This study aims to identify molecular marker loci that could be applied in broiler breeding programs. In this study, we used public databases to locate the Transcription factor 21 (TCF21) gene that affected the economically important traits in broilers. Ten single nucleotide polymorphisms were detected in the TCF21 gene by monoclonal sequencing. The polymorphisms of these 10 SNPs in the TCF21 gene were significantly associated (p < 0.05) with multiple growth and body composition traits. Furthermore, the TT genotype of g.-911T>G was identified to significantly increase the heart weight trait without affecting the negative traits, such as abdominal fat and reproduction by multiple methods. Thus, it was speculated that the g.-911T>G identified in the TCF21 gene might be used in marker-assisted selection in the broiler breeding program.

1. IntroductionChicken (Gallus gallus) is the most commonly raised poultry by humans. Since the 1950s, the growth rates and meat yield of broilers have significantly improved. However, with the rapid growth of broilers, some problems inducing huge economic losses have emerged, such as obesity, ascites syndrome, leg diseases, broiler immunity decline, and sudden death [1]. The growth rate of broilers is positively correlated with these unfavorable traits. As a result, it is difficult to simultaneously increase the growth rate and decrease these unfavorable traits in broilers by the traditional phenotypic selection method alone. Notably, molecular marker-assisted selection (MAS) can provide new ideas for overcoming such problems [2]. The combination of molecular genetic marker breeding with traditional phenotypic selection helps to greatly improve the breeding efficiency and accelerate the breeding process [3]. In recent years, with the rapid development of molecular genetics, genetic markers have been gradually applied in the MAS of livestock and poultry [4]. This technology contributes to substantially improving the breeding efficiency and shortening the generation intervals [5]. Among the numerous molecular markers, single nucleotide polymorphisms (SNPs) have been the most extensively studied [6,7,8,9]. Therefore, SNPs are of practical significance to identify genes or markers related to the economically important traits of broilers.Transcription factor 21 (TCF21) plays an important role in a variety of economically important traits such as heart development [10], testis formation [11], and adipogenesis [12] in chickens. This study aims to identify the SNPs of the TCF21 gene that are significantly associated with the growth and body composition traits of broilers. The results of this study can provide useful information for the molecular genetic marker-assisted breeding of the economically important traits of broilers.2. Material and Methods2.1. ChIP-Seq AnalysisThe ChIP-seq dataset for histone modification marks (H3K4me3, H3K27ac, H3K4me1, H3K27me3) and CTCF data in the seven tissues of chickens used in this work were downloaded from the GEO Datasets: GSE158430 [13]. The Bedtools software 2.29.1 version was used to separate the ChIP-seq datasets of all tissues within the same merged [14]. The reference genome and annotation file for galGal6 (Gallus gallus) were downloaded from the UCSC Genome Browser (http://hgdownload.soe.ucsc.edu/goldenPath/galGal6/bigZips/, accessed on 16 June 2021). These combined data were genetically annotated using the ChIPseeker software 1.30.2 version [15] and visualized using the IGV browser (http://software.broadinstitute.org/software/igv/, accessed on 16 June 2021).2.2. Experimental Populations and Phenotype MeasurementsDetails on the Northeast Agricultural University broiler lines divergently selected for abdominal fat content (NEAUHLF) were described by Zhang et al. [16]. In this study, altogether, 675 male birds from the generations 21 (G21) populations of NEAUHLF were used for an association study. All birds were raised and managed in accordance with the routine commercial broiler feeding procedures.For the G21 populations, the body weight (BW) of all male birds was measured at 1, 3, 5, and 7 weeks of age (assigned as BW1, BW3, BW5, and BW7, respectively). At the age of 7 weeks, the above birds were slaughtered, and the body composition traits were recorded. Before slaughter, the chest angle (ChA), keel length (KeL), body oblique length (BoL), chest width (ChW), metatarsus length (MeL), and metatarsus circumference (MeC) of all birds were measured. After slaughter, the carcass weight (CW), abdominal fat weight (AFW), liver weight (LW), muscular stomach weight (MSW), glandular stomach weight (GSW), heart weight (HW), spleen weight (SW), and testicle weight (TeW) were measured. For the reporting of results, we complied with the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines [17].2.3. Genotyping of SNPsIn total, 20 individuals from NEAUHLF were randomly selected (with 10 birds from each line). Then, the whole gene of TCF21, including the gene body region, the 5 ‘flanking region (2000 bp), and the 3’ flanking region (2000 bp), was sequenced, which resulted in 10 SNPs, referred to as g.-1243C>T, g.-1171T>C, g.-911T>G, g.-891C>T, g.691C>T, g.897T>C, g.1033G>A, g.1892A>G, g.2091C>T, and g.2155C>T, according to their respective positions in the TCF21 gene.Using Primer Premier 5.0 (Premier, Canada), a series of PCR primers were designed to amplify the various portions of the chicken TCF21 genomic DNA sequence based on the chicken gene sequence (NCBI Reference Sequence: NC_006090.5), and all PCR primer sequences were synthesized and purified by Invitrogen (Camarillo, CA, USA). The primer sequences are shown in Table 1.Furthermore, the total genomic DNA was extracted from 675 male birds of the G21 of NEAUHLF for PCR analysis, according to previous depiction [18]. These SNPs were genotyped with the PCR-restriction fragment length polymorphism (PCR-RFLP) method. The PCR amplification system included: 50 μg/μL genomic DNA 1 μL, 10 mmol/L dNTP 0.8 μL, 10 × PCR Buffer 1 μL, 10 mol/L upstream and downstream primers each 0.2 μL, 5 U/μL Taq DNA polymerase 0.1 μL, and deionized water 6.7 μL. The PCR amplification conditions were as follows: pre-denaturation conditions were all 94 °C for 5 min, denaturation conditions were all 94 °C for 30 s, extension conditions were all 72 °C for 30 s, and ultimate extension conditions were all 72 °C for 7 min. The annealing conditions and cycle number are listed in Table 1. After the PCR reaction was finished, 1.2% of the agarose gel was configured, the PCR amplification products were added, and the electrophoresis time was set for 20 min at 100 V. This agarose gel was removed from the electrophoresis solution and placed in the gel imaging analysis system to take pictures for identification. All the PCR reagents and electrophoresis reagents were obtained from Dalian Treasure Biological Engineering Co., Ltd. (Dalian, China).The PCR amplification product was detected by the agarose electrophoresis of the target band single bright sample, and carried out by an enzymatic reaction test and enzymatic reaction system (2 μL of PCR product, 1 μL of Cutsmart Buffer, 6.8 μL of deionized water, and 0.2 μL of endonuclease, which were digested overnight at 37 °C). The SnapGene 5.0 Viewer (https://www.snapgene.com/snapgene-viewer/, accessed on 20 October 2021) was used to select the restriction enzymes, and the endonuclease for each SNP are displayed in Table 1. New England Biolabs provided all of the restriction enzymes (New England Biolabs, Ipswich, MA, USA). The digested products were detected by 3.0% agarose gel electrophoresis at 110 V for 50 min, and three genotypes were acquired for each of the 10 SNPs (Figure S1).2.4. Transcription Factor Binding Site AnalysisTo explore the potential molecular mechanisms underlying the association of SNPs loci in the TCF21 gene with the economically important traits in broiler chickens, bioinformatic analysis was performed using three transcription factor binding site software, including JASPAR (http://jaspar.binf.ku.dk/, accessed on 12 December 2021), TFBIND (http://tfbind.hgc.jp/, accessed on 12 December 2021), and Mulan (http://mulan.dcode.org/, accessed on 12 December 2021). These three bioinformatics software predicted overlapping transcription factors that were considered to possibly bind to the DNA sequence of SNPs in the TCF21 gene.2.5. Statistical AnalysesThe difference in allele frequencies between the lean and fat lines was determined and examined using the Chisquare test, with p < 0.05 as a significant difference between the lean and fat lines.In this study, the JMP 7.0 software (SAS Inst. Inc., Cary, NC, USA) was employed for establishing a generalized linear mixed model to analyze the associations of SNP polymorphisms with the growth and body composition traits, with p < 0.05 being adopted as a threshold. In addition, the significant differences between the least-square means of different genotypes were calculated by the contrast test (p < 0.05). Then, the statistical model for analyzing the associations of genotypes with the growth and body composition traits was constructed based on the population characteristics [19]. The following model was utilized:Y = μ + G +L + F (L) + D (F, L) + BW7 + e I where Y is the observed value of traits, μ stands for the population mean, G indicates the genotype fixed effect, and L suggests the line fixed effect. In addition, F (L) indicates the random effect of the family within the line, whereas D (F, L) represents the random effect of dams in the family of the line, and e is the random effect. Model I was adopted to analyze the associations of SNP polymorphisms with the growth and body composition traits in 675 male birds (335 individuals from the lean line and 340 individuals from the fat line) from the G21 population of NEAUHLF, in which each line consisted of 40 family lines (one sire and four dams, respectively).The statistical analysis model for genetic parameter estimation is shown below:y = Xβ + Zu + e II where y stands for the n-dimensional vector of the broiler growth and body composition traits, X represents the n × p matrix of fixed effects, β indicates the p-dimensional vector of fixed effects, Z suggests the n × q matrix of random effects, while u is the q-dimensional vector of random genetic effects, and e denotes the n-dimensional vector of random residual effects. Moreover, model II was applied in estimating the genetic parameters of the growth and body composition traits of the lean and fat lines in the G21 population of NEAUHLF.3. Results3.1. Identification of Genes Associated with Growth and Body Composition Traits in BroilersGenome-wide searches for genes affecting the important economic traits in broilers were conducted using the ChIP-seq data for histone modifications. The results revealed that the TCF21 gene plays an important role in the adipose, liver, lung, and spleen tissues of broilers (Figure 1). Then, the entire gene of TCF21, as well as 2000 bp upstream and downstream of the TCF21 gene, was sequenced, and 10 SNPs were identified (Figure S1).The genotype frequencies and allele frequencies of those 10 SNPs in the TCF21 gene in NEAUHLF were analyzed. Meanwhile, the chi-square independence test was conducted to calculate the differences in allele frequencies between the lean and fat lines. As discovered from the results, differences in the allele frequencies of these 10 SNPs were statistically significant between the lean and fat lines (p < 0.05; Table 2).3.2. NEAUHLF Is an Ideal Test Material for Studying the Correlation between Growth and Body Composition Traits in BroilersThe phenotypic information of the growth and body composition traits is displayed in Figure 2. As observed from Figure 2, differences in most of these traits (except for HW and BW5) were significant (p < 0.05) between the lean and fat lines in the NEAUHLF population.Furthermore, the heritability (h2) values of the growth and body composition traits were estimated. The results indicated that AFW, BW1, BW5, GSW, MSW, and TeW showed high heritability values (h2 > 0.3), whereas BW3, BW7, CW, and MeC had moderate values (0.2 < h2 < 0.3), and BoL, ChA, ChW, HW, KeL, LW, and MeL had low values (h2 < 0.2; Table 3). In addition, this study also estimated the genetic correlation (rg) between AFW and the other growth and body composition traits. As a result, at the genetic level, AFW was highly positively correlated (rg = 0.696 ± 0.223) with LW, but highly negatively correlated (−0.8 < rg < −0.3) with BoL, BW1, 3, 5, 7, GSW, KeL, and MeL. In addition, AFW showed low genetic correlations with ChW, CW, HW, MeC, MSW, and TeW (−0.3 < rg < 0.3; Table 3).3.3. Associations of TCF21 Gene Polymorphisms with Growth and Body Composition TraitsThe positions of these 10 SNPs in the TCF21 gene are shown in Figure 3A. Furthermore, this study analyzed the associations of the polymorphisms of those 10 SNPs in the TCF21 gene with the growth and body composition traits in NEAUHLF. According to the results, the polymorphisms of g.-1243C>T, g.-1171T>C, g.-911T>G, and g.-891C>T were significantly related (p < 0.05) to HW. In addition, the polymorphisms of g.2091C>T and g.2155C>T were significantly correlated (p < 0.05) with BW and TeW (Figure 3B). Linkage disequilibrium (LD) analysis revealed the existence of 2 different LD blocks, with 4 SNPs from block 1 (g.-1243C>T, g.-1171T>C, g.-911T>G and g.-891C>T) in a strong linkage disequilibrium and 2 SNPs from block 2 (g.2091C>T and g.2155C>T) were also in a strong linkage disequilibrium state (Figure 3C). All these results suggest that SNPs within Block 1 may have important effects on the HW trait, and SNPs within Block 2 may have important effects on the TeW and BW traits.Subsequently, this study further compared the least squares means of SNPs within these two blocks for different genotypes and traits. The results showed that the CC genotype of g.-1243C>T, TT genotype of g.-1171T>C, TT genotype of g.-911T>G, and CC genotype of g.-891C>T had higher heart weight than the heterozygous genotype (p < 0.05, Figure 4). Furthermore, the TT genotype of g.2091 C>T and g.2155C>T had higher body weight and lower testicle weight in broilers (p < 0.05; Figure 4).In order to investigate the potential molecular mechanism underlying the association of the HW trait with four SNPs from Block 1 (g.-1243C>T, g.-1171T>C, g.-911T>G, and g.-891C>T), we carried out an in silico analysis of the transcription factor binding site using three bioinformatic tools. The results showed that g.-911T>G was located in multiple potential transcription factor binding regions (Table 4).4. DiscussionIn this study, two broiler lines were divergently selected for abdominal fat content for over twenty generations. The results revealed significantly different AFW values between the lean and fat lines. In addition to AFW, some other growth and body composition traits (except for HW and BW5) also showed significant differences (p < 0.05) between the lean and fat lines. The above results indicated that when AFW was selected, the other traits associated with AFW were also under selection. Therefore, the genetic correlations between AFW and other growth and body composition traits were estimated. The results indicated that AFW showed high genetic correlations with most of the other growth and body composition traits, including LW, GSW, BW1, 3, 5, 7, MeL, KeL, and BoL. Some studies also analyzed the correlations of AFW with the growth and body composition traits in chickens and reported that AFW exhibits high genetic correlations with BW5, BW7, LW, CW, and skin weight [26,27], which are consistent with our results. Therefore, the lean and fat lines were the ideal experimental materials used to study the growth and body composition traits.It was discovered in this study that the polymorphisms of g.2091C>T and g.2155C>T in the TCF21 gene were significantly associated (p < 0.05) with the TeW and BW traits. As revealed by studies on mammals, the TCF21 gene plays an important role in the functions of testicles [11]. In addition, the testis growth and development of chickens are controlled by genetic factors [28,29,30,31,32], and cocks with lower TeW are usually less fertile [33]. Furthermore, it is found that male mice with the TCF21 gene knockout have sex differentiation phenotypes [34]. The male sex determining factor SRY affects TeW through regulating TCF21 [35,36]. Regrettably, the least squares mean analysis revealed that the TT genotype of g.2091C>T and g.2155C>T had higher body weight and lower testicle weight. It also indicated that selection for these two SNPs did not result in neither fast growth rate (BW) nor high reproductive performance (TeW) in broilers.Heart hypertrophy increases the risk of sudden death in broilers, especially those with higher BW and AFW traits [37]. This research study found that the polymorphisms of g.-1243C>T, g.-1171T>C, g.-911T>G, and g.-891C>T were significantly related (p < 0.05) to the HW trait. Some literature reports 13 susceptible sites are detected in a GWAS on human coronary heart disease, among which rs12190287 is located at the 3’UTR of TCF21 [38,39]. Generally, TCF21 is expressed in mesoderm cells in the epicardial organ and then in mesenchymal cells that form the pericardium [40]. The loss of TCF21 in chickens leads to epicardial blistering, increased smooth muscle differentiation on the heart surface, a paucity of interstitial fibroblasts, along with neonatal lethality [10]. It is encouraging that the least squares mean analysis revealed that the CC genotype of g.-1243C>T, TT genotype of g.-1171T>C, TT genotype of g.-911T>G, and CC genotype of g.-891C>T had higher heart weight (p < 0.05; Figure 4). It also indicated that selection for these four SNPs could improve the HeW trait without affecting other unfavorable traits at the same time in broilers. The non-coding regions of genes have a large number of regulatory elements, including enhancers, promoters, and silencers. Studies have shown that SNP located within these regulatory elements can affect traits by influencing the activity of regulatory elements [41,42,43]. In silico analysis suggested that the g.-1243C>T was located in the regions of potential binding of BACH1 [20], and the g.-911T>G was located in multiple transcription factor binding regions (GATA4, SMAD1, and SOX17). Studies have shown that these transcription factors all play important roles in animal heart development [21,22,23,24,25]. It is hypothesized that the TCF21 gene g.-911T>G regulates the HW trait probably through binding to transcription factors (GATA4, SMAD1, and SOX17) to influence the activity of regulatory elements in this region.5. ConclusionsIn this study, the associations of TCF21 gene polymorphisms with the growth and body composition traits in broilers were analyzed. The results indicate that the g.-911T>G in the TCF21 gene may be important molecular markers that affect the HW trait, and could be used in breeding programs to improve the heart development of broilers.

animals : an open access journal from mdpi

10.3390/ani13050893

PMC10000101

The European pig industry needs to adapt to growing social interest regarding animal welfare. One of these concerns is the surgical castration of male piglets—a common practice primarily performed to avoid the risk of boar taint released from the meat of uncastrated males, especially during heat treatment. Several EU countries are trying to stop the surgical castration of pigs. The European Commission (EC) strongly supports these activities. One of the two currently feasible alternatives to the production of castrates is the fattening of entire male pigs. It is well known that skatole, one of two main compounds responsible for boar taint, can be eliminated or reduced by feeding additives. Recently, some promising results have been achieved using hydrolysable tannins in the diet of entire males. However, it should be mentioned that these studies focused on the influence of tannins on fattening, carcass value, meat quality and the deposition of androstenone and skatole in adipose tissue but not their influence on sensory characteristics. Therefore, the objective of this study was, in addition to determining the effects of tannins on skatole and androstenone accumulation in fatty tissue, to assess the sensory attributes of pork from entire males after supplementation of the diet with 1–4% tannins. The results showed that 2–4% supplementation of tannins in the feed reduced the accumulation of skatole in fatty tissue. The odour and flavour of pork were not influenced by tannin supplementation, but higher doses of tannins decreased the juiciness and tenderness of pork from entire males but only in men’s evaluation. The effect of the sex of the panellists on both of these sensory traits was observed in both the control and tannin-supplemented groups.

Boar taint is an unpleasant odour and flavour released during heat treatment of pork from uncastrated male pigs. The two main compounds responsible for boar taint are androstenone and skatole. Androstenone is a steroid hormone formed in the testis during sexual maturity. Skatole is a product of microbial degradation of the amino acid tryptophan in the hindgut of pigs. Both of these compounds are lipophilic, which means that they can be deposited in adipose tissue. Several studies have reported heritability estimates for their deposition from medium (skatole) to high magnitudes (androstenone). In addition to efforts to influence boar taint through genetic selection, much attention has also been paid to reducing its incidence using various feeding strategies. From this point of view, research has focused especially on the reduction in skatole content by supplementation of feed additives into the nutrition of entire male pigs. Promising results have been achieved using hydrolysable tannins in the diet. To date, most studies have investigated the effects of tannins on the production and accumulation of skatole in adipose tissue, intestinal microbiota, growth rate, carcasses and pork quality. Thus, the objective of this study was, in addition to determining the effects of tannins on androstenone and skatole accumulation, to assess the effects of tannins on the sensory traits of meat from entire males. The experiment was performed on 80 young boars—progeny of several hybrid sire lines. Animals were randomly assigned to one control and four experimental groups (each numbering 16). The control group (T0) received a standard diet without any tannin supplementation. Experimental groups were supplemented with 1% (T1), 2% (T2), 3% (T3) or 4% (T4) SCWE (sweet chestnut wood extract) rich in hydrolysable tannins (Farmatan). Pigs received this supplement for 40 days prior to slaughter. Subsequently, the pigs were slaughtered, and sensory analysis was applied to evaluate the odour, flavour, tenderness and juiciness of the pork. The results showed a significant effect of tannins on skatole accumulation in adipose tissue (p = 0.052–0.055). The odour and flavour of the pork were not affected by tannins. However, juiciness and tenderness were reduced by higher tannin supplementation (T3–T4) compared to the controls (p < 0.05), but these results were sex-dependent (in favour of men compared to women). Generally, women rated tenderness and juiciness worse than men regardless of the type of diet.

1. IntroductionAnimal welfare has become a very important factor in livestock breeding, including that of pigs. A general ban on surgical castration of entire male pigs was expected in all EU member states by the end of 2018. Due to various circumstances, it was postponed until after 2021. One feasible alternative to painful surgical castration is to fatten all males. It is well known that boars grow faster and more efficiently, and they have higher lean meat content in carcass than surgical castrates. On the other hand, fattening boars increases the risk of and results in a higher incidence of boar taint and thus the dissatisfaction of consumers with such meat.The incidence of boar taint is mainly attributed to two substances, androstenone [1] and skatole [2], especially after heat treatment of pork. Androstenone (α-androst-16-en-3-one) is a steroid hormone synthesized in the Leydig cells of the testis of uncastrated boars. This compound has an odour similar to urine or sweat, is perceived by approximately two-thirds of the human population and has been shown to be different according to country/region. Deposition of androstenone in fat tissue has high heritability estimates (0.55–0.88), which indicates that it is affected mainly by genetics [3,4]. Skatole (3-methyl-indole) is a metabolite derived from microbial catabolism of the amino acid tryptophan in the hindgut of pigs. Its deposition is influenced mainly by environmental factors (h2 = 0.23–0.55) [5], especially nutrition, the system of feeding and the housing conditions [6,7,8,9,10]. Since both of these compounds are lipophilic, they can accumulate in adipose tissue and therefore may have a negative effect on sensory attributes and result in the rejection of such meat by consumers.Since the odour of skatole is negatively perceived by practically the entire human population, reduction efforts have focused mainly on various feeding strategies in fattening all males. Promising results have been achieved by supplementation of diets with a variety of feed additives, such as chicory root or inulin [11,12,13,14,15,16,17,18,19,20], raw potato starch [21,22,23,24], sugar beet pulp [25], Jerusalem artichoke [26], oligofructose and fructooligosaccharides [15,27] and tannins [28,29,30].Tannins are plant metabolites with great diversity, resulting in different physiological effects according to their form, animal species and amount of supplementation [29,31]. Sweet chestnut (Castanea sativa Mill.) wood extract, consisting mainly of hydrolysable tannins, has antimicrobial and antiviral properties. Therefore, products containing this substance are used in animal nutrition, especially in piglets, as supportive treatment for diarrhoea after weaning [32,33,34,35,36,37,38].Several studies have demonstrated a reduction in total protein digestibility, as well as inhibition of microbial activity, in the colons of pigs after supplementation of the pig diet with hydrolysable tannins [39,40,41]. Lower disponibility of tryptophan and cell debris in the hindgut may lead to reduced intestinal production of skatole [28,42]. These findings are interesting from the entire male production point of view.Apart from a few studies dealing with Iberian pigs fed natural sources [43,44,45,46], other research on the effects of hydrolysable tannins (HTs) has been mainly aimed at growth and carcass parameters, meat quality and oxidative stability or the intestinal skatole production and microbiota composition in the large intestines of boars [28,29,30,31,32,37,41,47], but almost no research has addressed the effect of HTs on the sensory traits of entire male meat.The main objective of the present study was to assess the impact of hydrolysable tannins on parameters of eating quality, considering the possible effect of the sex of consumers, as well as to investigate the relationships between sensory evaluation and the content of skatole and androstenone in adipose tissue.2. Materials and Methods2.1. Animals and DietEighty young boars were used in the experiment. They were the progeny of Landrace sows and Yorkshire × Pietrain boars. Two weeks before the experiment, the pigs were housed in pairs/pens at the experimental test station of the Research Institute for Animal Production (RIAP) Nitra. Subsequently, the boars were randomly distributed within litters to one control and four experimental groups (each containing 16 animals). Control pigs (T0) received a diet without any supplementation. Experimental groups received the same diet as the control group but supplemented with 10 (T1), 20 (T2), 30 (T3) or 40 (T4) g/kg sweet chestnut wood extract (SCWE) rich in hydrolysable tannins (Table 1). The producer of the Farmatan product is Tanin Sevnica d.d., Sevnica, Slovenia, and the supplier was Product Feed a.s., Luzianky, Slovakia. The content of tannins in this product is 73 ± 2% (the value declared by the producer). The analysis of feed was performed according to the Folin–Ciocalteu method [48]. The total phenolic content is expressed as gallic acid equivalents and is 45.1%.Supplementation of the diet with tannins started when the boars reached an average live weight of 80 kg (after a 2-week transitional period) and lasted for 40 days prior to slaughter. Access of the animals to drinking water and feed (automatic feeders—Schauer s.r.o., Nitra, Slovakia) was ad libitum.2.2. Slaughter and Sample CollectionEntire males were slaughtered in one batch at the experimental slaughterhouse of RIAP Nitra. The average live weight of the pigs was 122.28 kg ± 5.63 kg. Standard slaughter conditions were used: electrical stunning (90–100 V, 0.9–1.0 A, 50 Hz) followed by exsanguination. Evisceration was completed approximately 20 min post mortem. Chilling of the carcasses (air temperature 2–4 °C, velocity 0.5–1.0 m.s−) started approximately 60 min after slaughter and was continued overnight. After 24 h of chilling of the carcasses, musculus longissimus thoracis (LT) samples (1.0 cm thick) with subcutaneous fat were removed from the right side of the carcass (at the level of the last rib) and stored at −20 °C until sensory evaluation.2.3. Sample PreparationOne day before sensory evaluation, the raw LT sample was thawed overnight at 4 °C. Subsequently, each LT slice was trimmed to have 5 mm of subcutaneous fat. Each LT slice was placed in a grill and cooked for 4 min at 180 °C. The average measured internal temperature of the samples was 80 °C. After grilling, each steak was cut into four strips and immediately served to different panellists.2.4. Sensory EvaluationPanellists were recruited from the staff of RIAP Nitra. All of them were consumers who liked and ate pork regularly (2–3 times weekly). Before the sensory evaluation of samples, consumers were tested for their sensitivity to androstenone according to the modified method of Weiler et al. [49]. On the basis of this method, 20 panellists were selected (12 men and 8 women aged 32 to 60 years old).In total, 320 samples were evaluated in eight sessions (each of 40). In each session, eight panellists evaluated five samples (one from each dietary group). Each panellist participated in several sessions. Meat samples were randomly served to the consumers, and the attributes classified were odour, flavour, tenderness and juiciness. The scale applied in the sensory test was structured into 5 points, with 1 being the worst and 5 the best evaluation (Table 2). The panellists were asked to score for odour first, followed in order by flavour, tenderness and juiciness.2.5. Skatole and Androstenone DeterminationFat samples (100 g from a part of the belly) from entire males were removed 24 h after slaughter. One part of each sample was individually packed in a microtene bag, marked and frozen (−20 °C) until panel testing. The second part of each fat sample was transported to the authorized private laboratory of EKOLAB, s.r.o. (Košice, Slovakia), to analyse the androstenone and skatole concentrations according to the methods of Ampuero Kragten et al. [50] and Bekaert et al. [51]. Briefly, adipose tissue samples were melted in a microwave for 4 min. Liquid fat was transferred to centrifuge tubes (2 mL), and 1.75 mL of extraction solvent methanol:hexane (9:1) were added. The extract was ultrasonically cleaned in a bath at 50 °C for 5 min and centrifuged for 5 min at 10,000× g. After cooling, approximately 2.0 mL of extract were then injected into a gas chromatograph equipped with a mass spectrometry (MS) detector (Shimadzu GCMS-TQ8030, Shimadzu Corp., Kyoto, Japan) at an injection temperature of 260 °C.The limits for detection were 0.02 µg/g for androstenone and 0.01 µg/g for skatole.2.6. Statistical AnalysisThe observed data were evaluated by 1-way analysis of variance (ANOVA) with fixed effects [52] using the following model:yij= μ + αi + eij with N (0,σ2) for androstenone and skatoleFor multiple comparisons of treatment means, Bonferroni’s test was used [53]. Dunnett’s test was not used to compare the control and treatment groups since, from the analyses of variance of both traits, it could be concluded that differences between all groups by the F test were nonsignificant. The observations of sensory traits were evaluated by 2-way ANOVA with fixed effects using the following statistical model:yijk = μ + αi + (αβ)jj + eijk with N (0,σ2)Pearson’s correlation coefficients of androstenone and skatole with sensory traits were calculated.3. Results3.1. Effect of Tannins on Androstenone and SkatoleLevels of boar taint compounds in adipose tissue with regard to tannin supplementation are shown in Figure 1. The samples that were high in both androstenone and skatole levels (greater than thresholds of 1.0 and 0.2 µg/kg, respectively) in the control group numbered 5, whereas in the supplemented groups, they numbered 1, 2, 3 and 0 for the T1, T2, T3 and T4 groups, representing 31.25%, 6.25%, 12.5%, 18.75% and 0%, respectively. The total number of samples high in androstenone, skatole or both compounds was 10 (62.5%). After tannin supplementation, these numbers were reduced to 7%, 6%, 8% and/or 7% (43.75, 37.5, 50.0 and/or 43.75%).Deposition of androstenone in adipose tissues was not affected by supplementation of the EM diet with tannins (Table 3). A numerical reduction, however nonsignificant (p = 0.052–0.055), in skatole concentration in fat tissue was found after administration of 2%, 3% and 4% tannins compared to the control group.Correlations between observed traits are presented in Table 4. Generally, all the relationships of concentrations of androstenone or skatole in adipose tissue with sensory traits were small and negative. Higher correlations were observed between skatole and eating quality parameters, with the exception of juiciness, which was more correlated with androstenone concentration.3.2. Effect of Tannins on Eating Quality of PorkSupplementation of the pig diet with tannins did not show any impact on the panellists’ scores for meat odour and flavour (Table 5 and Table 6). However, significant differences among the treatment groups were found for tenderness and juiciness (Table 7 and Table 8). Men scored both traits significantly better in meat from the control EM group than in meat from the T3–T4 or T4 group (3.05 vs. 2.74 and 2.81; 3.28 vs. 2.75, p < 0.05). Moreover, significant differences in the evaluation of these two traits were observed between the sexes of the panellists. Women were more critical than men regarding both tenderness and juiciness traits regardless of diet group (2.70 vs. 3.00, p < 0.05, and 2.65 vs. 2.99, p < 0.01, respectively).4. DiscussionGenerally, tannins are considered to be antinutrients, as they create compounds with other nutrients, such as proteins, minerals, or digestive enzymes, and therefore are capable of reducing feed palatability, feed intake and nitrogen digestibility [39,54,55]. However, pigs, wild as well as domestic, seem to be relatively resilient to the intake of feedstuffs with a high content of tannins without any negative consequences on performance or health status [56]. This resistance is associated with elevated synthesis of proline-rich proteins (PRPs) in the saliva, which bind tannins from feedstuffs and prevent intoxication of organisms with diets rich in hydrolysable tannins [57,58]. Moreover, recent studies have suggested that tannins have antimicrobial, anticancer, and antioxidant properties and can improve feed efficiency and reduce bacterial proteolytic reactions in piglets, thus protecting them from severe diarrhoea during weaning [32,33,34,35,36,37,59]. It is well known that wild pigs, as well as some special native breeds of domestic pigs (Iberian, Cinta, etc.), can eat foodstuffs rich in tannins (content: 4–7%); therefore, a dose of 40 g per kg of feed mixture was selected as the highest dosage in the present study.It is well known that the level of skatole in fatty tissue is influenced by many processes, such as formation, absorption, metabolism and deposition. The main role is associated with the activities of digestive enzymes in the CYP450 family, such as 2E1, 2A19, 1A1, and 1A2, in the liver [6,28].Generally, data relating the effect of tannins on skatole production and accumulation in pig adipose tissue are limited. Some studies have shown that tannins can inhibit proliferation and apoptosis in the caecum. This inhibition results in decreased skatole production in the large intestine due to the lower availability of cell debris from lower apoptosis and tryptophan [41]. In the present study, skatole accumulation in adipose tissue tended to decrease with increasing tannin supplementation. Similar numerical decreases were observed in other studies after 2–3% [28] or 3% [29] tannin supplementation, but surprisingly, lower supplementation (1.0 or 1.5%) resulted in higher skatole accumulation than in the control group [28,30]. Čandek-Potokar et al. [28] suggest that this result could be associated with lower activity of CYP2E1 and CYP2A19 enzymes, two major enzymes of the phase 1 metabolism of skatole in the liver. It should be mentioned that some of the above studies [29,30] used products with lower contents (only 50%) of hydrolysable tannins than our study.Regarding the effect of tannins on androstenone, the present study showed that supplementation with these additives did not have any impact, even though the two highest doses had higher (although nonsignificant) androstenone concentrations in fat tissue than the control group. This result is partially in contrast with that of other studies [28,29,30] in which numerical reductions were found after supplementation of diet with 1–3, 3, and 3% hydrolysable tannin extracts. However, the pigs in the latter study [30] had low levels of androstenone overall.Generally, the results regarding the effects of tannin supplementation on skatole and androstenone levels in back fat were not consistent and were highly dependent on the dose of tannin supplementation, often reporting curvilinear dependence. Therefore, research to determine the optimal dose of tannin supplementation for reducing boar taint is still needed. Moreover, the effects of hydrolysable tannins on androstenone are still unclear, and if any are confirmed, further studies will be needed to clarify the mechanism of action.As previously mentioned, very few studies have been published thus far relating the effects of hydrolysable tannins on pigs. Those studies focused mainly on intestinal skatole production, growth rate, carcass and meat quality, and the intestinal microbiota [30,32,37,47,59], but almost none focused on the sensory properties of pork. Thus, our results related to the effects of tannins on eating quality are difficult to compare with those of other studies. Only one study in pigs [30] and one in sheep [60] investigated the impact of tannin supplementation on the sensory traits of meat. In the present study, the odour and flavour of boar meat were not affected by tannin supplementation. Panellists scored these two parameters in the control and supplemented groups at almost the same level, and the differences were not significant. Bee et al. [30] reported different findings. Panel members detected a stronger boar taint odour but not flavour in the meat of entire males supplemented with 1.5% tannins compared to the controls and group with 3% tannins. In lambs, Priolo et al. [60] observed lower sheep meat odour in meat from animals supplemented with tannins compared to those not supplemented. Contrary to odour and flavour, the other two sensory traits—tenderness and juiciness—in the present study showed significant differences. Panellists, but only men, scored (p < 0.05) tenderness better in the control group than in the two supplemented groups (T3–T4), and juiciness in the control group was evaluated better than in T4. This outcome is contrary to the results of a study [30] that reported no significant effect of tannin supplementation on juiciness and/or tenderness. Regarding differences between the sexes of panellists in the present study, women scored tenderness and juiciness worse than men in both the control and supplemented groups.It seems that higher supplementary tannin doses (3–4%) in our study significantly reduced juiciness (T4) and tenderness (T3–T4) compared with the control group. However, this result may be sex-dependent.5. ConclusionsSupplementation with tannins had no effect on androstenone accumulation in adipose tissue. In contrast, the concentration of skatole in fat tissue tended to decrease with increasing tannin supplementation. The odour and flavour of pork were not affected by tannin supplementation, but its juiciness and tenderness were lower after supplementation of the entire male diet with higher concentrations of tannins. The effect of the sex of consumers on sensory evaluation showed significant differences for the tenderness and juiciness of pork in favour of men. It seems that tannin concentrations greater than 3 in the diet negatively affect some sensory characteristics. These findings might be useful in solving appropriate feeding strategies for entire males to reduce boar taint, as well as to maintain good eating quality of their meat.

animals : an open access journal from mdpi

10.3390/ani11041068

PMC8070412

Since the introduction of the hole–board test, its validity and applicability have been repeatedly re-examined. The hole–board protocol remains one of the standard procedures applied in psychopharmacology and behavioral studies. Some authors advocate the use of the hole–board procedure in studies on various aspects of behavior regulation, such as exploration and anxiety, habituation to a novel environment, spatial learning and memory (working and reference memory), spatial pattern learning, and food search strategies. In this study, we focused on rats’ activity in the hole–board test that we considered to be a type of exploratory activity. Based on our results and our previous studies of rats’ exploratory behavior in the free-exploration box, we suggest that the hole–board apparatus might not be the best tool for measuring exploratory behavior in laboratory rodents.

This study focuses on the rat activity in a hole–board setting that we considered a type of exploratory behavior. The general hypothesis is based on the claim that a motivational mechanism is central to both the response to novelty in a highly familiarized environment and the activity in the hole–board apparatus. Our sample consisted of 80 experimentally naive Lister Hooded rats. All rats were tested in the hole–board apparatus. Twenty individuals with the highest hole-board scores and twenty subjects with the lowest hole–board scores subsequently underwent an established free-exploration test. In our study, the scores obtained in the hole–board test had little predictive value for the rats’ activity in the free-exploration test. Based on our previous experience in studying exploratory behavior in the free-exploration test and the data presented in this paper, we suggest that the hole–board test is not an appropriate tool for measuring exploratory behavior in laboratory rodents.

1. IntroductionThe hole–board test is a behavioral test that has been used to assess different aspects of cognitive abilities and emotions in small mammals. It originated from an open field test, which was designed to evaluate exploratory behavior and anxiety [1]. The hole–board apparatus has small cylindrical holes at the bottom of the experimental arena that allows experimenters to conduct more complex behavioral observations. A head-dipping in the holes is considered a main feature of the behavior in this experimental arena. However, since the introduction of the hole–board test [2], its validity and applicability have been repeatedly re-examined. Head-dipping in the hole was considered by File and Wardill [3] to be a valid measure of exploratory activity. The discussion continued, and R. Hughes [4] (p. 449) stated: “There is also the possibility that head-dipping could involve attempts to find an escape route rather than reflect a genuine interest in objects underneath the holes”. This view was substantiated by an experimental study on Lister-Hooded rats [5] (p. 442) in which the authors concluded: “Rather than being a measure of neophilia, these results support the hypothesis that head-dipping represents an escape response, which declines as the subject becomes less fearful”. Despite this controversy, the hole–board protocol remains one of the standard procedures applied in psychopharmacology and behavioral studies. Some authors [6] advocate the use of the hole–board procedure in studies on various aspects of behavior regulation (such as exploration and anxiety, habituation to a novel environment, spatial learning and memory (working and reference memory), spatial pattern learning, and food search strategies) [7,8]. More recently, the hole–board procedure has also been used to assess behavioral characteristics that are supposed to reflect an animal model of Autism spectrum disorder (ASD) [9].Neophillia (the tendency to approach unfamiliar objects or environments—[10,11]) in animal behavior has long been a subject of scientific discussion. We share the view expressed in earlier papers that any behavior in a novel environment would be regulated by both neophilia and neophobia [5,6,12,13]. Therefore, rather than being at the two extremes of a continuum, neophilia and neophobia should be thought of as two independent mechanisms that can come into play simultaneously. Over the past three decades, research on curiosity and exploration in animals has established some basic methodological guidelines for further studies. The experimental stimulus of novelty should be controlled by ensuring a sufficiently long habituation period. Moreover, the test environment should provide the subjects with a comfortable low-stress setting, which does not trigger neophobic or defense responses. A validated protocol of this methodological approach has already been established [14], and a detailed description of the free-exploration test was published [15], enabling readers to fully replicate all of the procedural details. Since the ecological validity of the protocol for neophillia and the exploratory responses has been clearly demonstrated, it suggests that the free-exploration test may serve as a tool for evaluating other methods of behavioral assessments in terms of their validity for measuring exploratory behavior. Since the hole–board procedure is often used as a tool for measuring exploratory behavior in rodents, our study sought to validate the two protocols in question, namely the hole–board protocol and the free-exploration test (low-stress), as previously described [15].To date, there are many variants of the hole–board devices. We chose the most standard and widely used apparatus. While the hole–board procedure is used for numerous purposes (such as anxiety or general activity assessment or behavioral profiling), we have focused on the exploratory aspect of the behavior in the rats’ activity on the hole–board. The general hypothesis was based on the claim that there is a motivational mechanism central to both the response to novelty in a highly familiarized environment and the activity in the hole–board apparatus. If it is true, there should be a strong positive correlation between both kinds of activity, namely the response to novelty in the free-exploration box and the activity in the hole–board apparatus.2. Materials and Methods2.1. AnimalsThe sample consisted of 80 experimentally naive Lister Hooded rats. The rats were bred and kept in the vivarium of the Institute of Psychology, Polish Academy of Sciences, Warsaw, Poland. The rats were approximately 90 days old and weighed approximately 350 g at the start of the experiment.The rats were housed in groups of 3–4 in Tecniplast Eurostandard Type IV cages (610 mm × 435 mm × 215 mm) with dust-free softwood granules (Tierwohl Super, Rosenberg, Germany) as bedding and with ad libitum access to water and standard laboratory fodder (Labofeed H, WP Morawski, Kcynia, Poland). The day/night cycle was set at 12/12 h (with the lights on at 8 a.m.) and the temperature was maintained at a constant 21–23 °C. The cages and pens were cleaned once a week, on the same day and at a fixed time, in the late afternoon (5 p.m.). However, in order to ensure that the experimental procedure was not disturbed, the cages in which the test animals were kept were cleaned just before the onset of the experiment and again after the experiment was finished. All of the rats kept in our laboratory were housed, bred and taken care of in accordance with the Regulation of the Polish Minister for Agriculture and Rural Development of 14 December 2016 on laboratory animal care. The experimental procedures were approved by the First Local Ethics Committee on Animal Experimentation in Warsaw, Poland (No. 1116/2020).The sample size was estimated using a commonly used formula for calculating sample size for repeated measures [16]:N = 1 + C(s/d)2 where:s—standard deviation of the population meansd—size of difference in means (the effect)C—constant dependent on the value of α (significant level) and 1-ß (power)For the purpose of our study, we employed the following parameters: α = 0.05; β = 0.20 (C = 10.51).Group size calculations were carried out on the basis of a previous study [14] in which the average time spent on exploring changed objects was M = 96.51, with standard deviation s = 36.63, and on the assumption that the detectable difference between the variables should be d = 41.Therefore, the total sample size for the free-exploration test was estimated at 10.2.2. ProcedureThe experiment was conducted in two phases. In Phase I, a hole–board test was carried out. Then, the selected animals were subjected to the exploration test in Phase II.2.2.1. Phase I—The Hole–board TestThe aim of the test was to measure the rats’ activity (i.e., head-dipping behavior) in the experimental area and select individuals with the lowest and the highest levels of analyzed behavior. The scores were assigned based on the number of instances of head-dipping in the holes during a 5-min session. Twenty subjects (10 males and 10 females) with the highest scores and twenty individuals (10 males and 10 females) with the lowest scores on the hole–board test were admitted to Phase II.We used a square hole–board arena measuring 600 × 600 × 450 mm (Hole–board for rats, manufactured by MazeEngineers, Skokie, IL, USA) (Figure 1). The transparent Plexiglas walls were covered to prevent the animals from being distracted by visual stimuli. There were 16 round-shaped holes in the bottom of the arena, each 50 mm in diameter, distributed evenly at equal distances from each other. The experimental arena was illuminated by fluorescent ceiling lamps, set at approximately 75–100 lx. The video camera used for recording the rats’ behavior was placed 1.2 m above the measurement apparatus.After being taken out of the housing cage, each rat was immediately placed in a transporter (a small cylindrical cage 60 mm in diameter with doors 120 mm high and 100 mm wide) and moved to another room where the experiment took place. The transporter with the animal was placed in the corner of the experimental arena in such a way that the rats could leave the transporter through the exit into the interior of the arena. After the transporter exit was opened, each animal was left in the arena for five minutes. Each rat was free to explore the experimental arena and leave and enter the transporter without impediment. After completing the session, the animal was removed from the experimental arena and returned to its home cage. The hole–board test was carried out once. The testing arena was cleaned after each rat using Virkon S (Bayer, Leverkusen, Germany).2.2.2. Phase II—The Free-Exploration TestOne day after the hole–board test, the selected (high-activity and low-activity) rats participated in the free-exploration test. The high-activity group included 20 individuals (10 males and 10 females) that had the highest scores in the hole–board test whereas the low-activity group included another 20 individuals (10 males and 10 females) that had the lowest scores in the test.The aim of the free-exploration test was to measure the differences between the process of exploring the new environment, the rate of habituation to it, and the response to the introduction of harmless novelty into the familiar context. The apparatus and the measurement methods were similar to those used in our previous studies (e.g., [14,15,17,18,19]).The experimental chamber was a box with dimensions of 800 mm × 600 mm × 800 mm. The space inside the chamber was divided into three zones (A–C), separated by two walls diametrically perpendicular to its longer side (Figure 2). The front part of the chamber was a transparent wall that could be elevated in order to gain full access to the testing arena. The wooden partition walls between the zones had triangular passages (120 mm × 140 mm) at the bottom, which allowed animals to move freely between parts of the chamber. The entire chamber was coated with a layer of removable varnish. Wooden tunnels (200 mm × 120 mm × 80 mm) covered with washable paint were positioned in zones B and C. Unlike the most commonly used two-dimensional experimental settings, these tunnels provided a complex three-dimensional environment. The middle zone (A) was empty. There was a hole in the back wall of the chamber that functioned as an entrance for animals moving from the animal transporter to the chamber. The transporter functioned as a starter box (60 mm in diameter with doors 120 mm high and 100 mm wide).At the beginning of each trial, the transporter with the animal inside was placed at the entrance to Zone A. Then, the rat was left unhindered in the starting box for 15 s, after which the entrance door was opened. The animal was allowed to stay in the starting box or leave it to examine the chamber. The first seven trials were habituation trials, during which the apparatus was adjusted in exactly the same way (Figure 3). The implementation of the novelty took place on trial 8. The novelty took the form of adding additional tunnels, as it is shown in the right panel of Figure 3. Three consecutive tests trials were carried out with the chamber in this new configuration.Each trial lasted 7 min and was carried out for each animal once a day. Each experimental session was followed by a cleaning of the experimental arena, the tunnels, and the transporter using Virkon S (Bayer) in order to eliminate odors left by the previous animal.A video camera was placed at a distance of approximately 1.5 m from the transparent front wall of the experimental chamber. The camera was set up in night-time shooting mode to ensure the possibility of filming in the dark.To prevent the effects of lateralization or visual/auditory cues, novel objects were introduced in the left zone (as described above) for half of the examined subjects and in the right zone for the other half (mirror image of Figure 3).2.3. Data Processing and Statistical AnalysesTo code the behaviors on the basis of the recorded material, we used BORIS software (http://www.boris.unito.it), which made it possible to define selected behaviors and to assess their duration and frequency. We scored the behaviors the animals engaged in during the entire experimental trial. Consequently, we were able to assign specific scores to the times of separate bouts of behaviors, their frequency, and the total time an animal spent engaging in a given behavior. The following variables were measured: (1) time spent in the transporter (excluding the latency required to leave the transporter); (2) time spent in the central zone; (3) time spent in the unchanged zone of the chamber; (4) time spent in the changed zone of the chamber; (5) frequency of moving between the zones (left/right/transporter) of the chamber; (6) time spent in contact with the tunnels in the unchanged zone of the chamber; (7) frequency of contact with the tunnels in the unchanged zone of the chamber; (8) time spent in contact with the tunnels in the changed zone of the chamber; and (9) frequency of contact with the tunnels in the changed zone of the chamber. The time spent in the experimental chamber was measured in seconds.To enhance the legibility of the results and tables, the habituation phase has been indicated as H (the mean score from habituation trials 5 through 7 that presents the outcome of the process of habituation to the experimental environment and served as a reference value for further analyses), while the test trials have been indicated as T1, T2, and T3, respectively. Novelty (i.e., the addition of tunnels in zone C) was introduced in the first test trial (T1). Groups selected on the basis of the activity in the hole–board test were named as follows: the high-activity group was known as HB_H and the low-activity group was known as HB_L.The data were analyzed using a general linear model procedure (GLM), with repeated measurements (H, T1, T2, T3) as within-subject factors, as well as sex and hole–board group assignment as between-subject factors. This was followed by post-hoc t-tests with Bonferroni correction for multiple comparisons. Differences were considered significant for p ≤ 0.05.3. ResultsThe first-step analysis was designed to extract the individuals of the high and low levels of activity (i.e., head-dipping) in the Hole–board apparatus. Table 1 shows the descriptive statistics of this measurement. Individuals falling below the 25th percentile and above the 75th percentile were qualified for further tests.The complete set of descriptive statistics of all the free-exploration test variables analyzed is shown in Table 2.3.1. Time Spent in the TransporterThe analysis showed significant trial by sex interaction (Wilks’ Lambda; F(3,105) = 7.684; p ≤ 0.001; eta2 = 0.180), trial by group interaction (F(3,105) = 8.875; p ≤ 0.001; eta2 = 0.202) and a main effect of the trial (Wilks’ Lambda; F(3,105) = 41.646; p ≤ 0.001; eta2 = 0.543).A post-hoc analysis of sex interaction showed a significant decrease in the time spent in the transporter in T1 compared to the habituation phase in females (p ≤ 0.001, MH = 65.79, SDH = 22.87, MT1 = 29.47, SDT1 = 20.30, Cohen’s d = 1.120) and in males (p ≤ 0.001, MH = 64.60, SDH = 25.77, MT1 = 24.79, SDT1 = 10.88′ Cohen’s d = 1.275). In females during the next trials, the amount of time spent in the transporter remained at the same low level. In males, there was an increase in time spent in the transporter between T1 and T3 (p ≤ 0.001, MT3 = 51.66, SDT3 = 21.61, Cohen’s d = 0.861).A post-hoc analysis of group interaction showed a significant decrease in the time spent in the transporter in T1 compared to the habituation phase in the HB_high group (p ≤ 0.001, MH = 55.72, SDH = 17.60, MT1 = 23.30, SDT1 = 12.64, Cohen’s d = 0.998) and in the HB_low group (p ≤ 0.001, MH = 74.67, SDH = 26.36, MT1 = 30.96, SDT1 = 17.77, Cohen’s d = 1.400).3.2. Time Spent in the Central Zone of the ChamberThe analysis showed a significant main effect of the trial (Wilks’ Lambda; F(3,108) = 20.910; p ≤ 0.001; eta2 = 0.367), and a main effect of the group (Wilks’ Lambda; F(1,36) = 37.157; p ≤ 0.001; eta2 = 0.508).A post-hoc analysis showed a significant decrease in time spent in the central zone in T1 compared to that in the habituation phase (p ≤ 0.001, Cohen’s d = 1.102). Subjects from the HB_high group spent less time in that zone than did subjects from the HB_low group (p ≤ 0.001, Cohen’s d = 0.964).3.3. Time Spent in the Unchanged Zone of the ChamberThe analysis showed significant sex by group interaction (Wilks’ Lambda; F(3,108) = 4.891; p = 0.003; eta2 = 0.120), trial by group interaction (F(3,108) = 4.60; p = 0.004; eta2 = 0.114), trial by sex interaction (F(3,108) = 9.183; p ≤ 0.001; eta2 = 0.203), and a main effect of the trial (Wilks’ Lambda; F(3,108) = 27.449; p ≤ 0.001; eta2 = 0.433).A post-hoc analysis showed a significant decrease in the time spent in the unchanged zone in trial T1 compared to that spent during the habituation in females from the HB_high group (p = 0.003, MH = 118.48, SDH = 15.68, MT1 = 69.95, SDT1 = 29.52, Cohen’s d = 0.696), in males from the HB_high group (p ≤ 0.001, MH = 142.64, SDH = 28.92, MT1 = 77.35, SDT1 = 20.07, Cohen’s d = 0.936), and in males from the HB_low group (p ≤ 0.001, MH = 128.81, SDH = 36.47, MT1 = 77.28, SDT1 = 36.89, Cohen’s d = 0.739), but not in females from the HB_low group. However, there was an increase in time spent in the unchanged zone between T1 and T2 (p = 0.008, MT1 = 52.20, SDT1 = 22.47, MT2 = 97.82, SDT2 = 36.72, Cohen’s d = 0.654) and between T1 and T3 in that group (p ≤ 0.001, MT3 = 107.42, SDT3 = 38.00, Cohen’s d = 0.792) in females from the HB_low group.3.4. Time Spent in the Changed Zone of the ChamberMauchly’s test indicated that the assumption of sphericity had been violated (χ2(5) = 12.43, p = 0.029), so the degrees of freedom were corrected using Greenhouse–Geisser estimates of sphericity (ε = 0.85). The analysis showed significant trial by sex by group interaction (F(2.547,108) = 2.893; p = 0.048; eta2 = 0.074) and a main effect of the trial (F(2.547,108) = 95.496; p ≤ 0.001; eta2 = 0.726).A post-hoc analysis showed a significant increase in the time spent in the changed zone in trial T1 compared to the habituation in females from the HB_high group (p ≤ 0.001, MH = 158.25, SDH = 28.27, MT1 = 279.09, SDT1 = 43.54, Cohen’s d = 1.176) and from the HB_low group (p ≤ 0.001, MH = 127.66, SDH = 22.00, MT1 = 244.77, SDT1 = 43.71, Cohen’s d = 1.139), in males from the HB_high group (p ≤ 0.001, MH = 113.63, SDH = 22.63, MT1 = 254.28, SDT1 = 37.00, Cohen’s d = 1.386), and in males from the HB_low group (p ≤ 0.001, MH = 85.13, SDH = 17.59, MT1 = 223.81, SDT1 = 33.10, Cohen’s d = 1.349). There was a decrease in time spent in the changed zone between T1 and T3 in females from the HB_high group (p = 0.045, MT3 = 219.43, SDT3 = 46.72, Cohen’s d = 0.580) and in males from the HB_high group (p = 0.035, MT3 = 193.41, SDT3 = 28.94, Cohen’s d = 0.592). There were also differences between females from the HB_low and the HB_high group in T2 (p ≤ 0.001, MHB_low = 186.35, SDHB_low = 33.69, MHB_high = 268.75, SDHB_high = 26.18, Cohen’s d = 0.765).3.5. Frequency of Moving between the Zones (Left/Right/Transporter) of the ChamberThe analysis showed a significant main effect of the trial (Wilks’ Lambda; F(3,108) = 8.930; p ≤ 0.001; eta2 = 0.199).A post-hoc analysis showed a significant decrease in frequency of moving between the zones in trial T1 as compared to that during the habituation phase (p ≤ 0.001, Cohen’s d = 0.670).The analysis of between subject effects revealed an interaction effect of sex and group (F(1,36) = 4.146; p = 0.049; eta2 = 0.103). However, post-hoc comparisons did not reveal any specific differences between the groups.3.6. Time Spent in Contact with the Tunnels in the Unchanged Zone of the ChamberThe analysis showed significant trial by sex by group interaction (Wilks’ Lambda; F(3,108) = 4.193; p = 0.008; eta2 = 0.104), trial by group interaction (F(3,108) = 4.210; p = 0.007; eta2 = 0.105), trial by sex interaction (F(3,108) = 9.013; p ≤ 0.001; eta2 = 0.200), and a main effect of the trial (Wilks’ Lambda; F(3,108) = 15.663; p ≤ 0.001; eta2 = 0.303).A post-hoc analysis showed a significant decrease the time spent in contact with the tunnels in the unchanged zone in trial T1 compared to the habituation trials in males from the HB_high group (p ≤ 0.001, MH = 98.23, SDH = 23.20, MT1 = 51.70, SDT1 = 14.57, Cohen’s d = 0.775), but not in males from the HB_low group or in females from either experimental group. An increase in females from the HB-low group in trial T2 (p ≤ 0.001, MT1 = 30.00, SDT1 = 14.16, MT2 = 74.88, SDT2 = 37.18, Cohen’s d = 0.428) and T3 (p ≤ 0.001, MT3 = 80.03, SDT3 = 32.70, Cohen’s d = 0.833) compared to trial T1 was also shown.3.7. Frequency of Contact with the Tunnels in the Unchanged Zone of the ChamberThe analysis showed a significant main effect of the trial (Wilks’ Lambda; F(3,108) = 15.463; p ≤ 0.001; eta2 = 0.300), a main effect of sex (F(1,36) = 14.877; p ≤ 0.001; eta2 = 0.292) and a main effect of the group (Wilks’ Lambda; F(1,36) = 4.724; p = 0.036; eta2 = 0.116).A post-hoc analysis showed a significant decrease in frequency of contact with the tunnels in the unchanged zone in trial T1 compared to that in the habituation phase (p ≤ 0.001, Cohen’s d = 0.908). Additionally, males more frequently interacted with the tunnels in the unchanged zone than did females (p ≤ 0.001, Cohen’s d = 0.610), and subjects from the HB_high group more frequently interacted with the tunnels than did subjects from the HB_low group (p = 0.036, Cohen’s d = 0.344).3.8. Time Spent in Contact with the Tunnels in the Changed Zone of the ChamberMauchly’s test indicated that the assumption of sphericity had been violated (χ2(5) = 12.085, p = 0.034), so the degrees of freedom were corrected using Greenhouse–Geisser estimates of sphericity (ε = 0.85). The analysis showed a significant main effect of the trial (F(2.554,108) = 111.948; p ≤ 0.001; eta2 = 0.757).A post-hoc analysis showed a significant increase in time spent in contact with tunnels in the change zone between H and T1 trials (p ≤ 0.001, Cohen’s d = 2.664) and between T1 and T3 trials (p ≤ 0.001, Cohen’s d = 0.807).The analysis of between subject effects revealed an interaction effect of sex and group (F(1,36) = 5.164; p = 0.029; eta2 = 0.125). On the basis of the post-hoc comparisons, it was found that females from the HB_high group spent more time on interaction with the tunnels than did females from the HB_high group (p ≤ 0.001, Cohen’s d = 0.882), and males from the HB_high (p = 0.002, Cohen’s d = 0.637) and the HB_low (p ≤ 0.001, Cohen’s d = 1.011) groups.3.9. Frequency of Contact with the Tunnels in the Changed Zone of the ChamberThe analysis showed significant trial by sex interaction (Wilks’ Lambda; F(3,108) = 5.386; p = 0.002; eta2 = 0.130), and a main effect of the trial (Wilks’ Lambda; F(3,108) = 6.838; p ≤ 0.001; eta2 = 0.160).A post-hoc analysis showed a significant decrease in frequency of contact with the tunnels in the changed zone in females between the T1 and T2 trials (p = 0.022, MT1 = 9.25, SDT1 = 2.42, MT2 = 7.25, SDT2 = 1.21, Cohen’s d = 0.547).3.10. Effect Size AnalysisTo allow the reader to compare the powers of the effects found in our study, Table 3 shows effect size estimations.A descriptive summary of the results is shown in Table 4.4. DiscussionIn this study, we focused on rats’ activity on the hole–board, which we considered to be a type of exploratory activity. This approach is based on a long-standing theoretical tradition [20], which is still being developed nowadays [21]. The general hypothesis was based on the claim that motivational mechanisms are similar in both the response to novelty in a highly familiarized environment and the activity in the hole–board apparatus. If this is true, there should be a strong positive correlation between both of these kinds of activity (namely the response to novelty in the free-exploration box and the activity in the hole–board apparatus). Although the validity of the hole–board protocol for novelty-seeking measurement is often assumed implicitly [22], sufficiently robust evidence to support the above claim is still lacking.In our study, the scores obtained in the hole–board test allowed us to predict the level of rats’ activity in the free-exploration box only to a very limited extent. The main factor explaining exploratory responses in the free-exploration box was the environmental change that occurred over the course of the experiment. The factors of sex and HB group designation (which indicated high vs. low hole–board activity scores) were of lower predictive value. The direction of the relation between hole–board activity and exploratory scores in the free-exploration box is similar (e.g., individuals that scored high on the hole–board test manifested a high level of exploratory responses, such as time spent in contact with the tunnels). Moreover, HB-high individuals demonstrated a stronger tendency to spend more time in the modified zone of the experimental chamber. As observed by Žampachová et al. [23], behavioral characteristics measured in an open field and the hole–board test share several common properties. An important characteristic of that study was a repeated measure scheme adopted by the authors. The authors drew conclusions about animal personality rather than any specific mechanisms of exploratory behavior. Moreover, their arguments overlapped with exploration understood as a mediator of the adaptation process derived from the theoretical framework of behavioral ecology. However, it has little to do with the theoretical speculation about the mechanism of behavior regulation at an individual level. A similar approach was adopted earlier when researchers tested the temperament concept in rats based on the hole–board procedure. J. Ray and S. Hansen [24] tested rats in the hole–board test and the canopy test six times in a 3-week period. They found the hole–board test to be relevant for assessing the rats’ temperament; the dimension responsible for exploratory behavior, however, was found to be of secondary importance, after harm avoidance. In a further study [25], they found that the relative role of the two aforementioned dimensions changes with ontogenesis. It seems that the prevalent nature of the dimension directly linked to animals’ emotionality, compared to the exploratory (or stimulus seeking) axis, is strictly related to the procedural details, namely the way the animal is placed in the apparatus. The standard procedure involves a human placing the animal in the apparatus’s central zone using their hands. This, in turn, may be recognized as a crucial element of the animal’s situation, meeting the conditions for what is called a “forced exploration” paradigm. As Márquez, Nadal, and Armario [26] have shown, the hole–board procedure involves some level of stress response in tested animals, especially during the first minutes of measurement. This, in turn, would support the view that the hole–board procedure allows researchers to analyze behavioral measures relevant for “active coping” or reactivity rather than exploration per se. To avoid this obstacle, we decided to use a free-exploration procedural variant, which involved a transportation container which was comfortable for the animal being carried, allowing the animal to stay inside or leave at any moment of the trial into the open area of the hole–board. This was intended to address the main legitimate objection voiced by Hughes [4] and Brown and Nemes [5], who suggested that animal activity may be seen rather as an expression of the tendency to escape/leave the apparatus than to explore it. Indeed, our data do not support the view that the exploratory component of behavior repertoire in the hole–board is predominant. On the contrary, the data seem to support the view expressed by Hughes [4] and Brown and Nemes [5]. The reason for this lies in the ecological validity of the two tests: the standard hole–board apparatus vs. the free-exploration chamber [15]. First, the standard hole–board procedure involves testing under daylight conditions, while our free-exploration box is mainly used in darkness. Studying rats under dark conditions is more ecologically valid, as rat are nocturnal animals and typically avoid brightly light places. Secondly, the hole–board procedure is very rarely combined with several habituation trials, which seem unnecessary and unjustified in the light of the simplicity of the environment that the hole–board offers to the animal. However, novelty results from the discrepancy between the previous experience and actual sensory input. The prolonged habituation allows for control of the selective effect of modified parts of the environment. On the contrary, placing the animal in a completely novel test arena does not allow for the attribution of the behavioral activity manifested by an individual tested to a particular stimulation source. Therefore, it may be easily put forward that novel stimulation’s intensity drives an individual to leave the area rather than to explore it. Thirdly, the general structure of the hole–board environment does not offer many affordances. Rather, it offers just one affordance, albeit multiplied.There is no doubt, however, that the hole–board procedure does measure behavioral responses to a stimulus-rich environment, that it is widely used, and that its reliability is considered sufficient [27]. Nevertheless, one should be cautious when suggesting a theoretical interpretative tool for measuring an animal’s activity in the hole–board apparatus. Based on our experience in studying the exploratory behavior of rats in the free-exploration box [15], we conclude that the hole–board apparatus is not an appropriate tool for measuring exploratory behavior in laboratory rodents. However, we believe that the hole–board procedure may be an attractive tool for behavior analysis in many other fields of study (e.g., [28,29]). What is more, additional data obtained from studies with various variants of hole–board and test conditions (e.g., [30]) are needed to propose more conclusive statements. Nonetheless, this study provides cues for the rethinking of the role of the hole–board procedure as a tool for exploratory activity measurements.5. ConclusionsBased on the results of our study and observations from previous experiments on exploratory behavior in low-stress environments, we must conclude that the hole–board apparatus is not an appropriate tool for measuring exploratory behavior in laboratory rodents. Other behavior regulation mechanisms (e.g., risk assessment, emotional reactivity, active coping) might play a greater role in shaping an animal’s activity in the hole–board apparatus. Our results stress the need for cautious reflection on behavioral tests’ ecological validity when it comes to the studies on animal behavior.

animals : an open access journal from mdpi

10.3390/ani11030711

PMC8001471

Swan geese are becoming a very popular economic wildfowl in China, however, there lacks nutrient requirement guidelines for farmed Swan geese. This study evaluated the effects of supplemental feeding of dietary metabolizable energy (ME) level [9.5, 11.5, and 13.5 MJ/kg of dry matter (DM)] on growth and carcass characteristics of grazing Swan geese. The results showed that growth performance and body-size measurements (except shank length) were greatly improved by offering supplemental feeds to grazing geese, with no difference among supplemental feeding treatments. As DM intake (DMI) decreased with increasing dietary ME from day 29 to 56; meanwhile, slaughter, semi-eviscerated, eviscerated, and thigh muscle yield decreased with increasing dietary ME on day 56, and a well-balanced supplemental diet with ME of 9.5 MJ/kg of DM was recommended to improve growth and meat production of grazing Swan geese.

Grazing Swan geese (Anser cygnoides) have good meat quality but grow slowly. This study aimed to study whether supplemental feeding could improve growth performance of grazing Swan geese and investigate a suitable dietary metabolizable energy (ME) level of supplemental diet for grazing Swan geese. Naturalized healthy male Swan geese (n = 144; 42 ± 2.0 days and 1.21 ± 0.17 kg) were randomly allocated into 4 groups and grazed on pasture alone (control, CON) or offered supplemental diets with ME of 9.5, 11.5, or 13.5 MJ/kg of DM after grazing. Growth performance and body-size measurements (including bone development) were lower (p < 0.05) in CON versus supplemented geese, as well as slaughter measurements on days 28 and 56. The DM intake linearly decreased (p < 0.01) with increasing dietary ME from day 29 to 56. Slaughter, semi-eviscerated, eviscerated, and thigh muscle yield linearly (p < 0.01) decreased with increasing dietary ME on day 56. Lightness (L*) and yellowness (b*) for breast and thigh muscle on days 28 and 56, and breast muscle shear force on day 56, were lower (p < 0.01) in supplemented versus CON geese. In conclusion, supplemental feeding improved growth performance and carcass characteristics of grazing Swan geese, and supplemental feed with ME of 9.5 MJ/kg of DM could be offered to improve growth and meat quality of grazing Swan geese.

1. IntroductionSwan geese (Anser cygnoides) are migratory birds that are widely distributed in Mongolia, northeastern China, and southeastern Russia. Compared to intensively produced domestic geese, Swan geese meat is considered healthier due to its high protein but low fat and cholesterol [1]. Thus, making it one of the most popular and economical wildfowl in several provinces of China. Consequently, there is increasing interest in raising naturalized Swan geese. However, there is no published nutrient requirements for farmed Swan geese, and its feed formulation is usually based on requirements for domestic geese or ducks.Farmed Swan geese are commonly reared in artificial environments like domestic geese, eating forages and/or formulated diets until they reach a standard market body weight (BW) of 3.0–3.5 kg. There are some common concerns, such as (1) will geese reach market BW on time if fed only forage? (2) Will grazing pasture alone have negative effects on bone development and will it affect fattening performance? (3) If geese are fed only formulated diets, will they be over-fattened and have lower meat quality? These concerns are also relevant for domestic goose production. Domestic geese fed only forages cannot reach standard market BW by 70 days of age [2]. Furthermore, early nutrition affects growth rate and physical development of birds, especially adult skeletal size [3,4]. Excessive fat accumulation, especially abdominal fat, is a major problem facing modern commercial poultry production, leading to reduced growth efficiency and meat quality [5]. Therefore, determining a feeding regime that optimizes both growth efficiency and meat quality of Swan geese is needed.Dietary composition and feeding regimes have potential to improve growth performance, promote physical development, and reduce body fat deposition in poultry [6]. Regulating dietary metabolizable energy (ME) is one of the most common strategies, and its effects on growth performance and carcass characteristics have been studied in domestic ducks and geese [7,8,9], with ME of 10.87–11.29 MJ/kg and 11.29–12.13 MJ/kg recommended for geese of 0–4 and 5–10 weeks of age [10], respectively. However, relevance of these recommendations for farmed Swan geese are unknown. Moreover, effects of feeding management, e.g., intensive feeding versus free range feeding on growth performance, carcass traits, and meat quality, were studied for domestic geese [2,11,12,13], but apparently not for Swan geese.We hypothesized that offering grazing Swan geese with supplemental feed containing suitable dietary ME would promote growth performance and physical development and improve carcass traits and meat quality without adversely affecting health. Therefore, our objective was to investigate effects of offering supplementary feeding diets with various levels of ME on growth performance, physical and bone development, carcass characteristics, and blood biochemistry of grazing naturalized Swan geese (Anser cygnoides).2. Materials and Methods2.1. Animals and Experiment DesignThis study was conducted on a pasture at the Changling Ecological Research Station for Grassland Farming at the Songnen Plain in the Northeast of China (44°33′ N, 123°31′ E; 145 m above sea level), where annual average temperature is 4.9 to 6.4 °C and annual average precipitation ranges from 250 to 500 mm, with 70–80% of total precipitation from June to September. Healthy, naturalized Swan geese (Anser cygnoides; n = 144, 42 ± 2.1 days and 1.21 ± 0.17 kg) grown under artificial feeding conditions were provided by Benjun Geese Specialized Farming Cooperative Society (Jilin, China). All geese were given 1 week to adapt to the new environment and forages before experiment. During adaptation, each goose was labeled with a leg ring and vaccinated following standard operating procedures. At the end of adaptation, all geese were blocked by BW and randomly allocated into 4 groups (6 replicates per group and 6 geese per replicate), which were then randomly allocated to 1 of 4 treatments: grazed on pasture alone (control, CON) or offered supplemental diets with low, medium, or high ME (9.5, 11.5, and 13.5 MJ/kg of dry matter (DM); LMED, MMED, and HMED, respectively) after grazing. The diets were formulated to meet nutrient requirements of geese [10,14], with minor modifications (Table 1). The grazing pasture was composed of perennial Leymus chinensis, Chloris virgate, and Puccinellia tenuiflora (approximately 70.5%, 18.2%, and 8.4%, respectively), with a small amount of leguminous grass.2.2. Feeding Management, Sampling, and MeasurementsDuring the study, geese were let out for grazing pasture at 06:00 am and returned to the pen at 17:00 pm, with 6 geese in each replicate housed in the same pen at night. For the 3 supplemental feeding treatments, supplemental diets were offered ad libitum by pen at 18:00 pm, with refusals weighed before the next feeding. All geese had free access to clean water. The daily supplemental DM intake (DMI) was calculated by pen as (daily DM offered–daily DM refused)/6. Geese were weighed on days 0, 28, and 56 and average daily gain (ADG) was calculated for each 28-day period and the overall experiment period. Samples of diets were collected weekly and pooled over the entire experimental period, mixed, subsampled, dried in an oven at 55 °C for 48 h, then ground through a 1 mm screen and stored. All chemical analyses including ME, DM, crude protein (CP), and crude fiber were done as described previously [15]. Calcium was determined by atomic absorption spectrophotometry and phosphorus was estimated colorimetrically, as described previously [16].2.3. Blood Sampling and SlaughterTwo geese of each replicate (with average BW) were selected for blood sampling and slaughter at 09:00 am on days 0, 28, and 56. Blood sample (~5 mL) was collected via wing root venipuncture using a 7 mL vacuum tube without additive (Vacutainer, Becton Dickinson, Franklin Lakes, NJ) at each sampling time. Blood samples were kept at 4 °C for 30 min, then centrifuged at 2000× g for 15 min at 4 °C and frozen at −20 °C until analyzed. Serum biochemical parameters, including albumin (ALB), total protein (TP), triglyceride (TG), total cholesterol (TC), blood urea N (BUN), and alkaline phosphatase (ALP), were determined using commercial kits (Jiancheng Biology Co., Nanjing, China) and an automated biochemical analyzer (7020 Model, Hitachi, Beijing, China). Serum globulin (GLO) content was calculated as the difference between TP and ALB.After blood sampling, geese were exsanguinated by severing the jugular vein and carotid artery on one side of the neck. Weight of each goose was recorded after bleeding and then measurements of body-size (body length, chest width, chest depth, keel length, tibia, hip width, and neck length) were determined. Thereafter, geese were plucked and eviscerated. Measurements of carcass traits (including weights of semi-eviscerated carcass, eviscerated carcass, and subcutaneous fat, and meat quality of thigh muscle and breast muscle), length of various sections of digestive tract (including duodenum, jejunoileal, cecum, and rectum), and weights of viscera (including heart, liver, kidney, lung, spleen, pancreas, gizzard, and proventriculus) were recorded. Slaughter yield, semi-eviscerated carcass yield, eviscerated yield, and relative weights of viscera were calculated as percentage of live BW, whereas yield of subcutaneous fat, breast muscle, and thigh muscle were calculated as percentage of eviscerated carcass weight. Meat quality was assessed based on breast muscle color and shear force. The meat color, where L* represents lightness, a* redness, and b* yellowness, was measured using a Minolta CR-410 color-meter (Minolta Camera Co., Ltd., Osaka, Japan). Shear force analysis was conducted according to guidelines of the American Meat Association [17], using a Warner-Bratzler Shear Device (C-LM3B, Beijing, China).2.4. Statistical AnalysesAll data were analyzed using the Mixed procedure of SAS (version 9.1, SAS Institute Inc., Cary, NC, USA) as a completely randomized design to account for time (days on feed) effects, with treatment, period (weigh day), and geese within treatment as random effect. For repeated measures, various covariance structures were tested, and Autoregressive (1) was selected based on the lowest value for Akaike’s information criteria. To test the effects of increasing dietary ME level on measured variables, orthogonal contrasts were conducted with unequal treatment spacing orthogonal polynomial contrast coefficients calculated to determine linear and quadratic responses. Least square means were compared using the Tukey correction for multiple comparisons and treatment effects were declared significant at p ≤ 0.05 and trends at 0.05 < p ≤ 0.10.3. Results3.1. Growth Performance and Body-Size MeasurementsThe BW (days 28 and 56) and ADG (days 0–28, days 29–56, and overall) of supplemented geese were greater (p < 0.05) than CON geese, but there were no significant differences in BW and ADG among the 3 supplemented treatments (Table 2). Notably, ADG of grazing geese increased (p < 0.01) from 9.40 g/day to 20.55 g/day, whereas ADG of supplemental feeding decreased (p < 0.01) from 40.65 g/day (averaged) to 26.12 g/day during days 29 to 56. The DMI was not different among LMED, MMED, and HMED treatments from day 0 to 28, but linearly decreased (p < 0.01) as levels of dietary ME increased from day 29 to 56; therefore, overall DMI linearly decreased (p < 0.01) with increasing dietary ME. There were effects of days on-feed for BW, DMI, and ADG (p < 0.01; Table 2).Most body-size measurements were greatly enhanced (p < 0.05) by supplemental feeding as compared to grazing pasture only (Table 3), with minimal differences among supplemented groups. There was an effect of days on-feed (p < 0.05) for all body-size measurements (Table 3).3.2. Carcass Traits and Meat QualityOn both days 28 and 56, greater (p < 0.01) slaughter, semi-eviscerated, and eviscerated weights were observed in supplemented versus grazed geese, with no differences among supplemented groups (Table 4). Slaughter yield was not different among treatments, except lower (p < 0.03) slaughter yield for geese of HMED on day 28 and slaughter yield linearly (p < 0.01) decreased with increasing dietary ME on day 56. Similarly, the semi-eviscerated yield did not differ among treatments, except greater semi-eviscerated yield for geese of LMED on day 28; furthermore, there was a linear (p < 0.01) decrease in semi-eviscerated yield with increasing dietary ME on both days 28 and 56. However, there was greater (p < 0.05) eviscerated yield in MMED and HMED than CON geese on day 28 and in LMED and MMED than CON geese on day 56, with a linear (p < 0.01) decrease of eviscerated yield with increasing dietary ME. Abdominal fat yield did not differ among treatments on day 28, but it was greatly (p < 0.01) increased in supplemented versus control geese on day 56. Thigh muscle yield was greatly (p < 0.01) enhanced by supplemental feeding as compared to the grazing geese on both days 28 and 56, and it increased linearly (p < 0.02) with increasing dietary ME. Breast muscle yield was not affected by treatment on day 28, but CON geese had greater (p < 0.01) breast muscle yield on day 56 than supplemented geese. There were effects (p < 0.05) of days on-feed for all carcass trait measurements (Table 4).Values of L*, a*, and b*, and shear-force of breast muscle were affected (p < 0.01) by treatment, except a* value on day 56 (Table 5). The L* and b* values of breast muscle were lower (p < 0.01) in supplemented versus CON geese on both days 28 and 56, with no difference among LMED, MMED, and HMED. The a* of breast muscle was higher (p < 0.01) in supplemented versus CON geese on day 28, however, the difference was eliminated on day 56. Shear force of breast muscle of geese of LMED treatment was significantly lower (p < 0.01) than the other three treatments on day 28, whereas shear force of breast muscle of CON geese was significantly greater (p < 0.01) than that of supplemented geese on day 56. There were effects (p < 0.01) of days on-feed for L*, a*, b*, and shear force (Table 5).3.3. Length of Gut Section and Relative Weights of VisceraThe only effect (p < 0.05) of treatment on length of digestive tract segments was jejunoileal segment of CON versus supplemented geese on day 56 (Table 6). There were effects of days on-feed (p < 0.05) for all sections.Relative weights of heart and spleen were not different among groups on both days 28 and 56 (Table 7). Relative weights of liver were affected (p < 0.01) by feeding management, with greater relative weight of liver in LMED and MMED geese than CON and HMED geese on day 28, and with greater value in grazing geese than in supplemental feeding geese on day 56. Furthermore, relative weight of liver linearly decreased (p < 0.04) with increasing levels of dietary ME on both days 28 and 56. Relative weights of kidney and spleen were not different among treatments on day 28, whereas it was lower (p < 0.01) for supplemented geese on day 56. Supplemented geese had lower (p < 0.01) relative weights of pancreas, gizzard, and proventriculus than CON geese on both days 28 and 56, and relative weight of proventriculus linearly (p < 0.02) decreased with increasing levels of dietary ME on day 28, as well as for relative weights of pancreas (p < 0.01) and gizzard (p < 0.05) on day 56.3.4. Serum Biochemical ParametersSerum biochemical parameters were differentially affected by treatments on days 28 and 56, with all serum biochemical parameters except glucose affected by treatments on day 28, whereas TG, TC, and ALP were not affected by treatments on day 56 (Table 8). On day 28, serum ALB, GLO, and TP were not different among CON, LMED, and MMED treatments, which were greater (p < 0.05) than that of HMED geese. Serum TC was greater (p < 0.01) in geese that grazed pasture only than in geese that received supplemental feeding, serum ALP was greater (p < 0.01) in LMED and MMED geese than in CON and HMED geese, and serum ALB, GLO, TP, TG, TC, and ALP linearly (p < 0.05) decreased with increasing levels of dietary ME. However, serum ALB, GLO, TP, BUN, and glucose in CON geese were lower (p < 0.01) than that in geese of MMED and HMED on day 56. Furthermore, all these measurements linearly (p < 0.02) increased with increasing levels of dietary ME.4. DiscussionGeese can consume green forages and fibrous byproducts of crops to derive a considerable amount of their nutrient requirements [18]. Dietary fiber is essential for geese to maintain normal performance; consequently, a low-fiber diet had negative effects on nutrient utilization and health, with poor growth and slaughter performance [19]. Although wild geese can be grown entirely on pasture and domestic geese also can subsist almost entirely on forage, duration of feeding prior to reaching market weight was prolonged by 2–3 weeks [14]. Domestic geese that consumed forage failed to reach standard market weight (3.0–3.5 kg) on time (usually 120 days of age) due to inadequate CP and energy in forage [2]. In the current study, Swan geese that grazed on pasture alone for 56 days had a final BW of 1932.1 g, which were well below standard market weight. In contrast, grazed Swan geese that received supplemental feeding had a final BW of ranging from 3008.3 to 3258.3 g, meeting the target for standard market weight. The lower final BW of CON geese was attributed to their lower feed intake as compared to supplemented geese, which was expected, as nutrients derived from pasture may not fulfill nutrient requirements. Current results were consistent with previous studies [2,11,12], in which grazing geese had significantly lower BW than those fed supplemental feeds. Furthermore, all body-size measurements, except shank length, were significantly lower in CON versus supplemented geese, suggesting that bone development was slowed due to inadequate macronutrients. There were significantly lower body-size parameters in grazing geese and grazing geese that received only 1 week of supplemental feeding at the end of fattening compared to those with supplemental feeding throughout the experimental period [2]. Therefore, we recommended that supplementing grazing Swan geese is a practical method to obtain ideal growth performance by providing well-balanced nutrition. This is not only a feasible method to ensure geese reach standard market weight by 100 days, but also to promote bone development.It was noteworthy that BW of geese supplemented with diets with ME ranging from 9.5 to 13.5 MJ/kg were not different throughout the study. Similarly, weight gain of Italian Legarth goslings on starter and grower diets with ME ranging from 11 to 13 MJ/kg did not differ significantly [16]. In addition, similar results were also reported in ducks with dietary ME ranging from 11.8 to 13.8 MJ/kg [9]. This was attributed to decreased feed intake as energy concentration increased, thereby achieving consistent energy intake [9,16]. Furthermore, excess energy consumption does not further increase body weight gain but makes apparent digestibility of nutrients decline [20]. Notably, ADG of supplemented geese decreased dramatically from day 29 to 56, accompanied by decreased DMI as compared to days 0 to 28. Similarly, there was a very rapid live weight gain of goslings during the starter period, followed by a gradually reduced live weight gain after 10 weeks of age, with poor feed conversion [21]. Although geese are fast growing, their efficiency to convert feed to weight gain diminished rapidly with age [16]. Despite compensatory growth for grazing geese during days 29 to 56, their bone development and growth was greatly reduced due to low nutrient quality and quantity. Offering supplemental feeding to grazing geese at the end of the fattening period for only 1 week (days 64 to 70) did not enable geese to reach standard market weight, as it was too late to support optimal bone development [2]. Geese have reached complete bone development after 8 weeks, whereas development of breast muscle continues until 9 to 10 weeks [22]. Thus, it is critical to offer supplemental feeding at the proper time to avoid irreversibly retarded bone development.Geese have a stronger gizzard than other waterfowl, making them successful grazers, able to break down and digest plant cell walls [14,23]. As degradation in the gizzard is largely mechanical, geese that accessed grazing pasture or were supplemented with roughage always had higher relative weights of gizzard and a longer digestive tract. In the current study, there were significantly greater relative weights of gizzard and proventriculus in CON geese on days 28 and 56, as well as a longer jejunoileal segment on day 56. Due to lower nutrient contents and poorer feed efficiency of pasture, grazing geese had to ingest more pasture than supplemented geese to meet nutrient requirements. Greater roughage intake increased bulk of digesta, which triggers increasing contraction frequency and weight of gizzard and proventriculus to promote digestion. Geese fed >20% defatted rice bran had greatly increased proventriculus weights [24]. Increased jejunoileal length of grazing geese implies increased intestinal surface area for nutrient absorption. It is accepted that geese modify volume and weight of digestive tract and digesta passage rate to adapt to fiber-rich diets [25]. However, there are also some contradictory reports that fiber-rich diets shortened relative lengths of duodenum, jejunum, and ileum [26,27], or had no effects on length and weight of gastrointestinal tract sections [21]. Apparent discrepancies in relative lengths of the digestive tract were possibly caused by breed and age and by sources and levels of fiber [26,27], and perhaps other factors, e.g., physical form of diets [28,29]. Similar relative weights of gizzard and proventriculus and length of various sections of intestine of geese that received diets with varying dietary ME, implies similar digestion and absorption ability, consistent with their similar growth performance and body-size measurements.Carcass traits, reflecting differential deposition of nutrients in various tissues or different parts of the same tissue, are important indicators in evaluating growth performance of meat animals [24]. In the current study, supplemented geese had significantly greater slaughter weight, semi-eviscerated weight, and eviscerated weight than CON, which was consistent with several previous studies. For example, geese under intensive or semi-intensive feeding had significantly greater carcass weight and edible meat production than those under a pasture system [12]. Supplemented geese had significantly greater eviscerated carcass yield than grazed geese [2], whereas intensively reared geese had higher slaughter and carcass part weights than free-range geese [30]. In the present study, geese that grazed pasture alone had significantly lower thigh muscle yield than supplemented geese, in direct contrast to a previous report [2], which was perhaps due to differences in animal management that altered animal movements and exercise. In that regard, supplemented geese were grazed during the day in the present study but kept indoors in the previous study [2]. Perhaps, when geese are fed a well-balanced diet, more exercise may promote thigh yield. Poorer carcass quality of grazing geese was mainly due to a higher dietary fiber intake and lower feed digestibility, plus unbalanced and inadequate available nutrients [12]. Notwithstanding, based on previous studies [2,12,30], grazing geese have some advantages to intensively fed geese, e.g., less subcutaneous and abdominal fat, lower percentage of skin, and greater meat quality. Similarly, in the present study, there was significantly lower abdominal fat yield in CON versus supplemented geese, which agrees with Liu and Zhou [31] that pasture intake reduced subcutaneous fat thickness and abdominal fat yield of geese compared to control. As the main organ involved in fat metabolism, relative liver weights were significantly lower in grazing versus supplemented geese on day 28, consistent with El-Hanoun et al. [12]. However, greater relative weights of liver in grazing geese were probably due to a compensatory response to low dietary fat level, as reported [7]. Furthermore, breast and thigh muscles of grazing geese had significantly higher protein content and muscle collagen than those of intensively fed geese [2]. These advantages not only increase consumer demand, but also stimulate producers to seek ways to improve both meat quality and growth rate. Either supplemental feeding of grazing geese or adding an appropriate portion of forage to intensively fed geese were effective to modulate meat quality and growth rate. Liu and Zhou reported improved carcass characteristics, meat quality, and enhanced polyunsaturated fatty acid ratios in geese with ad libitum access to a corn-based ration and an alfalfa-based pasture [31]. Janicki et al. also reported that semi-intensive feeding resulted in lesser monounsaturated and higher polyunsaturated fatty acids in abdominal fat [11]. Greater meat L* and b* in CON geese was in line with previous studies of higher breast meat b* and thigh meat L* values in poultry reared in a free-range system [23,30,32]. Shear force is used for evaluating tenderness of meat (samples with lower shear force are more tender). In the current study, supplemented geese had significantly lower shear force than CON geese on day 56, suggesting that supplemental feeding will increase meat tenderness of grazing Swan geese, presumably due to increased intermuscular fat.Although slaughter, semi-eviscerated, and eviscerated weights did not differ among geese offered diets with ME ranging from 9.5 to 13.5 MJ/kg, slaughter, semi-eviscerated, and eviscerated yields linearly decreased with increasing dietary ME, as well as a tendency for increased abdominal fat yield. Similarly, in a previous study, dietary ME significant affected eviscerated yield, which peaked at 11.87 MJ of ME/kg of diet [7]. Current results concerning abdominal fat yield also agreed with previous reports in both geese and ducks that increasing dietary ME was associated with increased abdominal fat percentage [7,8,9], whereas relative abdominal fat weight was reduced significantly by decreasing dietary ME in broiler chickens, ducks, and geese [5].Health status can be reflected by serum biochemical parameters. Serum cholesterol, triglyceride, and total protein concentrations were lowered by feeding fibrous diets [21]. In the current study, serum ALB, GLO, and TP did not differ between CON geese and those supplemented with LMED or MMED diets on day 28, whereas they were significantly lower in grazing Swan geese than all the supplemented geese on day 56. This was likely attributed to an insufficient CP intake of geese that grazed pasture alone as compared to geese supplemented with well-balanced diets. Similarly, replacing a basal diet with 20% grass meal or dried sugar beet pulp meal during the grower period significantly decreased serum TP and ALB concentrations [21]. Greater BUN and glucose in geese supplementally fed MMED and HMED diets than the grazing geese, suggested better N and energy metabolism status in supplemental fed geese. Numerically greater TC in supplemental fed geese than the grazing geese on day 56 agreed with results of abdominal fat yield. However, although the value fluctuated from day 0 to 56 and among treatments, that all serum biochemical parameters were within a normal range, based on a previous report [24], suggests that all geese were under good health during the study.5. ConclusionsIn conclusion, offering a supplemental diet of well-balanced nutrients to grazing geese at the grower stage not only significantly promoted growth and bone development of Swan geese, but also produced more edible meat without negative effects on meat quality. Thus, it is a feasible and economical practice for farmed Swan geese production. Most end points of growth performance, body-size, carcass trait, and meat quality, did not differ among geese fed any of the three supplemental diets. However, considering that slaughter, semi-eviscerated, eviscerated, and thigh muscle yields linearly decreased while abdominal fat yield linearly increased with increasing dietary ME, we concluded that ME of supplemental diet should be <9.5 MJ/kg of DM to optimize growth performance and meat quality.

animals : an open access journal from mdpi

10.3390/ani11061783

PMC8232102

In this study, the density and diversity of relevant groups of bacteria at a broiler farm have been studied, in the inside and outside air and in litter samples. A high number of bacteria was detected in the litter and in the inside air, but a low emission of bacteria was found in the outside air. Moreover, the bacteria detected in the outside air decreased with the distance to the farm. A total of 544 isolates were identified from all the samples (146 from the litter, 142 from inside air and 256 from outside air). From these, 162 staphylococci, 176 Enterobacteriaceae, and 190 enterococci were detected. E. hirae was the predominant species and the detection of identical DNA profiles in E. hirae isolates from inside and outside samples suggests the role of the air in bacterial dissemination from the inside of the broiler farm to the immediate environment. It is necessary to consider the relevance of air as a vehicle of disseminating bacteria at the farm level, which can involve potentially pathogenic bacteria and bacteria carrying antimicrobial resistance genes.

The role of the air as a vehicle of bacteria dissemination in the farming environment has been previously reported, but still scarcely studied. This study investigated the bacteria density/diversity of the inside and outside air and of litter samples at a broiler farm. Samples were collected considering two seasons, three outside air distances (50/100/150 m) and the four cardinal directions. Selective media was used for staphylococci, enterococci, and Enterobacteriaceae recovery. A high number of bacteria was detected in the litter (2.9 × 105–5.8 × 107 cfu/g) and in the inside air (>105 cfu/m3), but a low emission of bacteria was evidenced in the outside air (<6 cfu/m3). Moreover, the bacteria detected in the farm’s outside air decreased the further from the farm the sample was taken. A total of 544 isolates were identified by MALDI-TOF (146 from the litter, 142 from inside air and 256 from outside air). From these, 162 staphylococci (14 species; S. saprophyticus 40.7%), 176 Enterobacteriaceae (4 species; E. coli 66%) and 190 enterococci (4 species; E. hirae 83%) were detected. E. hirae was the predominant species, and identical PFGE clones were detected in inside and outside samples. The detection of identical DNA profiles in E. hirae isolates from inside and outside samples suggests the role of the air in bacterial dissemination from the inside of the broiler farm to the immediate environment.

1. IntroductionAir plays a key role in the dissemination of some microorganisms, especially molds and viruses, but data about its involvement in the spread of bacteria are still scarce [1]. Airborne particles consist of a mixture of biological material from a range of sources. These particles are generally between 0.3 and 100 µm in diameter, in which microorganisms can appear either as liquid droplets or as dry particles. Smaller particles (ranging in size from 1.0 to 5.0 µm) generally remain in the air and can spread to surrounding fields, while larger particles are deposited on surfaces [2].Numerous outbreaks associated with the consumption of raw fruits and vegetables have been reported in industrialized countries [3,4,5,6,7]. Vegetable crops can be contaminated with pathogenic/toxigenic bacteria from animal sources through different routes, such as irrigation or manure [7], leading to food poisoning outbreaks when they are ingested either raw or with minimal processing [8,9,10,11,12]. However, these routes do not explain all cases and there is some evidence to support other routes of propagation, in which the air appears to be an additional vehicle of dissemination. This hypothesis is supported by previous works that point towards airborne dissemination of bacteria from farms to neighboring areas, as in the case of cattle farms [13,14], chicken farms [15,16,17,18], and pork farms [19,20].Intensive poultry production means large densities of animals in small areas, which appears to be a significant source of air pollution [21,22]. This pollution consists of a variety of airborne particles of biological origin in which bacteria are present. These bacteria come from soil, dust feed, litter and from the birds themselves, and may include potential pathogenic bacteria such as enterococci, staphylococci and Enterobacteriaceae, among others. The abundance of airborne bacteria varies with the season and location [23].The majority of studies are focused on the concentration of microorganisms in the air inside poultry houses, but much less is known about the spread of bacteria from fresh litter or inside air to the outside air in surroundings areas [23].The aim of this study was to assess the bacterial air contamination (especially of potentially foodborne pathogenic enterococci, staphylococci and Enterobacteriaceae) in intensive broiler breeding, both inside and outside of the farm, and during two seasons. This work aims to study whether microorganisms from the farm can reach surrounding fields and enter the food chain.2. Materials and Methods2.1. Characteristics of the Broiler FarmThe study was conducted in a modern broiler farm, built in 2015, and located in La Rioja (Spain). The farm covers a total area of 12,000 m2 (bounded by a fence), is surrounded by agricultural fields (olive and almond trees to the south, cereal to the west, vineyards to the north and vineyards/non-cultivated area to the east), and is situated 2 km away from the nearest urban area. It consists of two identical and parallel production buildings of 1800 m2 area and 3 m height (5400 m3) with a north–south orientation. Both buildings are equipped with automatic systems for feeding and watering. The microclimate in the buildings (heating, ventilation, cooling and relative humidity) is managed by an automated computer system: from 33 °C and 65% of relative humidity (RH) at the beginning of the production cycle (2-day flocks) to 19 °C and 85% of RH at the end of the cycle (47-day broilers). Lighting is provided by luminescent lamps for 18 h per day. The buildings are ventilated through a mechanical longitudinal ventilation system with two fans fitted in the south side of the building, five chimneys on the roofs and 80 inlets located in the longitudinal walls (air enters by lateral inlets and is extracted by the fans in the south side). This mechanism expels polluted air from the buildings and draws air into the buildings. Each building has capacity for 31,000 broiler chickens (density 17 broilers/m2). The broilers are reared on deep litter (chopped straw 0.10 m thick) and have free access to food and water. The production cycle is 45 days. After the end of each production cycle, manure and litter are cleaned by mobile machinery in both buildings simultaneously. Before loading new flocks of broilers, the buildings are disinfected with a solution of chlorine, NaOH and broad-spectrum insecticide. The disinfectant solution is applied by automatic nebulization and the buildings remain empty for one to two weeks before introducing a new batch.2.2. Sample CollectionBroiler farm sampling was conducted during two different seasons (summer (July 2019) and winter (February 2020)), between 8:00 a.m. and 1:00 p.m. In both cases, it was conducted at the same time in the broiler cycle (15–16 days after chick entry) and temperature and relative humidity inside the building were 27 ± 1 °C and 70%, respectively.The air samples were taken inside the buildings (inside air) and in the farm surroundings (outside air), at distances of 50, 100 and 150 m in four directions (north, south, east and west) (Figure S1).Two sampling methods were used at each sampling point: the stationary and mechanical method. In the stationary method, three sticks were firmly set in the middle of each building and in the established sampling points outside. Two culture plates were placed at one-meter height in each stick (six culture plates in each sampling point). The plates were exposed to the air for 4 h. In the mechanical method, a volume of air (100 L of inside air and 1000 L in each outside sampling point) was collected with an Air Ideal air sampler (Biomerieux, Craponne, France). This device allowed the passage of a specific volume of air through a grid with direct impact onto agar plates to facilitate the detection and the count of viable microorganisms. This device takes the air with a constant flow (100 L per minute); thus, it lasted 1 min for the inside sampling and ten minutes for outside sampling.In addition, samples of litter bed were collected aseptically at the same time and place as the air samples. The free and random movement of the broilers in the enclosure guarantees the homogeneity of the litter bed. The microbiological analysis was performed on 10 g of each litter sample homogenized with 90 mL of sterile peptone water. The homogenized litter samples were subjected to a serial dilution with a dilution factor of 1/10 using a sterile saline (0.9% NaCl) diluent. A volume of 0.1 mL from each decimal dilution was spread onto the surface of agar plates.Mannitol salt agar (MSA) (Scharlau, Barcelona, Spain), Slanetz-Bartley Agar (SB) (Scharlau, Barcelona, Spain), and Chromocult coliform agar (CCA) (Merck, Darmstadt, Germany) were used for the isolation and enumeration of staphylococci, enterococci and Enterobacteriacie, respectively, both from the air and litter samples. Thus, 12 plates were employed for each air sampling point (two of each culture media and two sampling methods). To summarize and taking into account the two sampling seasons: 48 plates for inside air and 288 plates for outside air were used. For litter samples, 6 plates per dilution (two for each culture media) were employed (three dilutions seeded and two sampling seasons made a total of 36 plates).2.3. Bacterial IdentificationUp to 10 colonies per plate were randomly taken and grown in brain heart infusion agar (BHIA) (Difco) after 24 h of incubation at 37 °C on selective media. Bacterial morphology was determined by Gram staining and isolates with filamentous morphology and large bacilli (potential Bacillus) were excluded from the study. Colonies grown on BHIA were processed for species identification by MALDI-TOF system (matrix-assisted laser desorption/ionization-time of flight) (Bruker Daltonik GmbH, Bremen, Germany), using either direct colony testing or the standard protein extraction protocol, according to manufacturer instructions.2.4. Characterization of Enterococcus hirae IsolatesE. hirae was selected for further characterization in order to track the potential dissemination of specific isolates from the inside environment of the farm to the outside air. The antimicrobial resistance phenotype was determined by agar disk diffusion [24] for the following antimicrobial agents (µg/disk): penicillin (10), erythromycin (15), gentamicin (120), tetracycline (30), chloramphenicol (30), linezolid (30), and vancomycin (30). As the genus Enterococcus shows an intrinsic low-level resistance for aminoglycosides, in this study, we used disks with a high charge of gentamicin (120 µg/disk) to detect acquired high-level resistance for this antimicrobial [25]. The breakpoints recommended by the Clinical and Laboratory Standards Institute [26] were followed for all antimicrobials. The clonal relatedness of selected E. hirae isolates (isolates with similar resistance phenotypes obtained in all three sampling points) was determined by pulsed-field gel electrophoresis (PFGE) of the genomic DNA, after digestion with the endonuclease SmaI [27] and PFGE patterns were compared as previously recommended using the GelJ Program [28] and following a previously indicated strategy [29].3. Results3.1. Bacterial CountsBacterial counts were calculated with the samples of litter (as cfu/g), and with the samples of inside and outside air obtained with the mechanical method (Air Ideal device) (as cfu/m3). Table 1 shows the counts obtained for all the types of samples analyzed, with the observation that in the case of outside air these correspond to the range of all three distances studied (50, 100 and 150 from the farm) and the four cardinal directions (north, south, east and west).The air inside the farm showed a high bacterial count. In fact, sampling performed with the Air Ideal device (both in summer and winter sampling) programmed to take the smallest possible volume of air (100 L) clogged the plates of all the culture media used (>104 cfu per plate), which indicates counts of over 105 cfu/m3.However, despite the high bacterial load present inside the farm facilities (litter and air), the counts obtained in the outside air were very low: less than 6 cfu/m3 in the summer sampling and 4 cfu/m3 in the winter sampling. Figure 1 shows the distribution of total isolates obtained in the outside air (n = 820) by the two sampling methods used. Of them, 176 isolates were obtained by the stationary sampling method (21%) and 644 by the mechanical sampling method (79%). On the other hand, given that the plates used with the mechanical method were clogged, all the isolates from the inside air were obtained from the stationary method plates.3.2. Diversity of Bacteria ObtainedA total of 1255 isolates were initially obtained (225 from litter, 210 from inside air, and 820 from outside air). Of these, 460 (37%) were eliminated due to morphological reasons (yeasts, micelle structures or presumptive Bacillus colonies). Most of these (79%), were isolated from the MSA medium. The remaining 795 isolates were analyzed using MALDI-TOF and 544 of these were identified at the species level: 146 from litter (65% of the 225 isolates initially obtained), 142 from inside air (67%) and 256 from outside air (30%) (Table 2). The presence of non-identified microorganisms was especially evident in the samples taken from the air outside the farm in the summer sampling. In this sampling, only 137 (53 from MSA, 52 from Chromocult, 32 from SB) of the 575 isolates obtained were identified by MALDI-TOF (23.8%). In the winter sampling, 119 (91 from MSA, 24 from Chromocult, 4 from SB) of the 245 isolates obtained were identified (48.6%).Table 2 shows the isolates identified from the litter and from the inside and outside air samples. Microorganisms identified in the air outside the farm (n = 256) in relation to the sampling point and season are shown in Table 3.In litter and inside-air samples, E. hirae and E. coli were the predominant species, accounting for 81% and 70% of the total isolates identified, respectively (43% and 39%, in litter and 38% and 32% in inside air). Regarding the microorganisms identified in the air outside the farm, the predominant species were S. saprophyticus, Pantoea agglomerans and E. hirae, accounting for 60% of total of isolates (20% each). However, E. coli was not detected in the outside air.3.3. Characterization of E. hirae Isolates Recovered from Different Sampling Points Showing Similar Resistance PhenotypeE. hirae was the predominant specie recovered in this study (29% of total identified isolates), representing 43%, 32% and 19% of those isolates recovered of litter, inside air and outside air, respectively.The antimicrobial resistance phenotype allowed the classification of the 158 E. hirae isolates into six different phenotypic groups: (A) susceptible to all tested antimicrobials (n = 94, 59.5%); (B) resistant only to penicillin (n = 23, 14.6%); (C) resistant only to tetracycline (n = 12, 7.6%); (D) resistant only to erythromycin (n = 15, 9.5%); (E) resistant to tetracycline and intermediate resistance to chloramphenicol and susceptible to the remaining antibiotics (n = 7, 4.4%); (F) resistant to tetracycline and erythromycin (n = 7, 4.4%).The PFGE profiles showed that two E. hirae isolates (phenotypic group E) obtained from litter samples had indistinguishable PFGE patterns with isolates of inside and outside air (obtained at 50 and 150 m away from the farm) (Figure S2a). Identical PFGE profiles were also observed for E. hirae isolates of phenotypic group F that were detected in inside air as well as in outside air (at 100 m away from the farm) (Figure S2b).4. DiscussionThe strict control of temperature and relative humidity inside the farm, as well as the identical number of animals in the farm, could explain the uniformity of the counts obtained for litter and inside air at both seasons of the year investigated in this study (Table 1). These data correlate with those previously obtained in poultry farms with similar characteristics [19,30,31,32].The high bacterial load detected in the farm’s inside air could be explained by the fact that it is a closed room in which the environmental conditions (temperature, humidity, daylight absence) remain constant and suitable for the proliferation of microorganisms. The concentration of airborne microorganisms in poultry buildings in other studies varies significantly (between 103 and 107 cfu/m3), which could be explained by differences in sampling methods, poultry species (broilers, hens and turkeys), building capacity, density of birds rearing, age of birds and microclimate conditions, among others [16,32].The farm’s design and management system could explain the very low counts obtained in the outside air, both in summer and winter sampling (Table 1). These low bacterial counts should be attributed to the diluting effect of the air and the stressful conditions of the environment, such as light exposure, temperature changes and dehydration [33]. Previous studies in which the outside air of broiler farms was analyzed (with similar characteristics and sampling methods to our study), also found low densities of bacteria in the air. Kostadinova et al. [32] detected bacterial counts at 2 m away from the farm, only 1 log unit per m3 higher than the counts at the control sampling point, located 500 m away from the farm. However, Chinivasagam et al. [16] found values of 103–106 cfu/m3, but at 10 m from the ventilation fans. Moreover, the sampling method used affects recovery rates. Schulz et al. [34] and Friese et al. [17,35], using a liquid medium (glass impingers) for the recovery of staphylococci in the air, achieved higher counts (106 cfu/m3 and 105 cfu/m3, respectively).In our study, the mechanical sampling method was more effective for the recovery of isolates from the air at low bacterial load conditions (outside air). However, if the bacterial density was very high (inside air), the stationary sampling method was the only one that allowed one to obtain isolates. The use of both simple methods may be adequate to study the microorganisms present in the environment.It can be clearly observed that the number of isolates was higher in the summer (575) than in the winter (245), and also that the number of isolates decreased with distance from the farm in all cardinal directions. In this sense, the counts were especially high in summer at 50 m south, coinciding with the position of the fans (south side of the buildings), which are often activated at this time. However, this effect was not observed in the counts conducted in winter, which is attributable to the lower activity of the fans in that season. The farm is located on a flat terrain, with no slopes or nearby forest masses. The wind direction changed during sampling periods, and it was not very strong on the days when the sampling was carried out (maximum 10 km/h, both in summer and winter sampling).The influence of different factors in the distribution of microorganisms in the outside air has already been highlighted by other authors, indicating the importance of the wind in bacterial dissemination, but also highlighting aspects such as the type of animal, the orography and the environmental conditions in the area where the farm is located, as well as the distribution of spaces and the daily activities [16,35,36,37].Of all the isolates obtained in the samplings carried out, the percentage of isolates identified was very similar for litter and inside air (around 65%). However, only 30% of isolates from the outside air were identified. This percentage is especially high in the summer samplings (more than 75%); this could be explained by a high presence of environmental microorganisms not considered in our study. In fact, identification using MALDI-TOF could not be achieved for 32% of isolates. In most of these unsuccessful cases, the protein spectra obtained did not correspond to any of those listed in the reference library (Biotyper, Bruker), which contains almost all of the bacteria of the relevant groups in this study [38]. It is therefore possible that the unidentified isolates correspond to environmental bacteria that might not have any clinical relevance.Regarding the microorganisms identified the composition of the microbiota in litter and inside air (Table 2), the results were very similar and in agreement with those reported by other authors [30,39,40], with enterococci and Enterobacteriaceae being the major groups. However, in the air outside the farm, in both summer and winter sampling (Table 2 and Table 3), most isolates corresponded to staphylococci and enterococci, although their distribution differed according to the season. Among enterococci and staphylococci, E. hirae (76.6%) and S. saprophyticus (39.4%) were the predominant species, respectively.Several authors have already pointed out that both groups of bacteria, but especially staphylococci, have great environmental resistance and are, therefore, those most found in the air. The airborne dissemination of staphylococci has been widely referenced and has even been suggested as an indicator for airborne bacterial emission from animal houses [15,34]. Several studies have shown that enterococci can disseminate from the organic exudates present in the farm to the inside and outside air, leading to the conclusion that air could be an important vehicle for the dissemination of enterococci among different ecosystems [41,42,43].It is important to note the absence of airborne E. coli (both in summer and winter sampling), which contrasts with its high presence in litter and inside air. It has been stated that the survival of E. coli is significantly reduced outside, particularly when exposed to direct daylight and high temperatures [18]. Environmental factors such as temperature, relative air humidity, ultraviolet radiation and sampling stress are suggested as factors that cause the low frequencies of the detection of this microorganism in air samples collected outside [44]. However, Laube et al. [18] managed to recover E. coli outside a poultry farm using liquid culture media. In a previous study carried out in pork farms, E. coli was not found in air samples [20]. The absence of E. coli in outside air samples from both pork and broiler farms does not match the results obtained in outside air from cattle farms, where E. coli was isolated from many of the samples analyzed [14]. This difference may be due to the type of livestock, since ruminants continuously expel this microorganism (intestinal gases, solid and liquid fecal matter), unlike other animals that produce fewer emissions.Regarding the characterization of E. hirae isolates, and although this work did not aim to study the antibiotic resistance of the isolates obtained, it is important to highlight that 40.5% of E. hirae isolates were resistant to at least one of the antimicrobial agents evaluated. The presence of antimicrobial-resistant enterococci and staphylococci is a growing problem, and their airborne dissemination has been analyzed in several works. Recent studies revealed the airborne exchange of antimicrobial-resistant bacteria from livestock farms to the environmental microbial community [18,35,37,45].The detection of identical genomic DNA profiles in E. hirae isolates recovered from litter, inside and outside samples leads to the conclusion that the air was involved in bacterial dissemination from the inside of the broiler farm to the immediate environment.5. ConclusionsIn modern broiler farms, the emissions of bacteria to distances greater than 50 m is very low, despite the large number of bacteria in the inside air.Although the bacterial exchange does not seem quantitatively significant, the finding of bacteria with the same genetic profiles in the interior and in the air 150 m away shows that this exchange exists.Thus, it is necessary to consider the relevance of air as a vehicle of bacterial dissemination at the farm level, which can involve potentially pathogenic bacteria. Similarly, the possible airborne dissemination of bacteria carrying antimicrobial resistance genes should be monitored.

animals : an open access journal from mdpi

10.3390/ani11061750

PMC8230789

Animal welfare is an important aspect that affects the health and productivity of dairy animals. This study reports the knowledge and opinion of dairy cattle farmers regarding dairy cattle welfare (DCW) in Keningau, Sabah. A total of 30 dairy farmers participated in the survey and the information collected includes their socio-demographic characteristics, knowledge, and opinions regarding DCW. Seventy per cent of the farmers (n = 21) had satisfactory-to-good knowledge of the DCW criteria, but their opinions differed regarding indicators of poor animal welfare. The understanding of DCW differed among farmers depending on the production level, educational status, herd size, and cattle breeds kept on the farm.

This study aimed to assess the knowledge and opinions about DCW among dairy cattle farmers in Keningau, Sabah. A questionnaire was developed, validated, and administered by hand to 30 farmers. The data collected include farmers’ and farm demographics, and opinions regarding the criteria and indicators of DCW. Only 17 respondents (57%) had heard of “dairy cattle welfare” before this study. Nine farmers (30.0%) had poor knowledge about DCW criteria, whereas 13 (43.7%) and 8 (26.7%) farmers had satisfactory and good knowledge, respectively. Farmers with higher education, larger herd size, high production level, and exotic cattle breeds showed a better understanding. Farmers understood most of the indicators; however, opinions regarding cattle behavior during milking, their physical appearance and their lying down behaviour need to be improved. Nevertheless, 28 respondents ranked their animals’ welfare as either good or satisfactory, which further reflects a poor implementation of DCW measures. The main factors suggested by farmers to influence DCW in their herds were facilities, worker issues, management practices, and animal well-being. In conclusion, guidance from veterinarians and animal welfare specialists may be needed to improve the farmers’ understanding and practices of DCW.

1. IntroductionOne of the major public concerns regarding sustainable livestock farming is animal welfare [1], and societies have been pressuring farmers and shareholders in the livestock industry towards improved animal welfare [2]. The dairy industry is no exception in this regard, as the welfare of dairy animals remains an important aspect for preserving health and attaining better productivity [3,4].Robust assessment methods are necessary to ensure animal welfare and to disseminate such information to farmers and consumers [5]. Several welfare assessment methods are available in the livestock industry, which are tailored toward specific management and farming systems. For instance, the European Food Safety Authority (EFSA) employed scientific opinion on the feasibility of existing welfare assessment methods in small-scale dairy farms (<75 lactating cows, family-run farms) and came up with a modified procedure of the Welfare Quality Protocol (WQ) [4]. Recently, the assessment of dairy cattle welfare (DCW) has been channelled towards the application of animal-based measures (ABMs), as welfare outcomes may vary in different management systems [5]. ABMs provide a direct indicator of animal welfare and how the animal copes under a specific farming system. Examples of ABMs include a body condition score, hock injuries, lying down behaviour, tick infestation load, California Mastitis Test score, physical injuries, locomotion scores and the animal flight zone [6,7].Another vital aspect influencing animal welfare is the farmers’ knowledge and opinion about the subject. Several factors have been identified as drivers for dairy producers to improve their animals’ welfare [3,8]. This includes the desire to meet not only the demands of consumers and food retailers, but also public expectations of the proper treatment of dairy cattle [9]. In North America and Europe, most dairy farmers appreciate the need to ensure good comfort and welfare for their animals, and it was perceived as a priority in the industry [3]. Furthermore, management practices that are thought to influence animal welfare directly include calving management, nutritional management and housing environments [10,11]. For studies conducted among consumers, appropriate feeding, good stockmanship, and environmental cleanliness were stated as indicators of good dairy animal welfare [12,13]. Clark et al. [13] also found that the public was concerned about farm animal welfare in modern production systems, especially on naturalness and humane handling, in particular regarding the use of antibiotics as a prophylactic treatment.In Malaysia, farm animal welfare is yet to attain the level of attention seen in other developed countries. Although the new Animal Welfare Act [14] was officially implemented in 2017, it was aimed at fostering more responsible pet ownership among Malaysians. Malaysia dairy farms are mostly characterized as indoor housing systems, with only a few farms providing pasture access for short intervals. However, confined housing and intensive dairy systems come with a lot of challenges that may predispose cows to poor welfare. This includes, but is not limited to prolong standing times, abrasive flooring, inappropriate stall designs, weather-related factors, poor hygiene, and lack of grazing area. Although there are published studies relating to DCW such as lameness prevalence and assessment of ABM on dairies [15,16], it is not known whether Malaysian dairy farmers are aware of, or have an understanding of, how modern farming practices may affect the productivity and welfare of farm animals. Thus, this study was conducted as a preliminary report assessing the knowledge, awareness and opinion regarding DCW among dairy farmers in Keningau, Sabah, Malaysia.2. Materials and Methods2.1. Study Design and Enrollment of Study PopulationThe study entailed a cross-sectional design and was carried out in Keningau, 95.2 km from Kota Kinabalu, Sabah, Malaysia. Farms were recruited from the list and contacts of dairy farms registered with the Department of Veterinary Services (DVS), Sabah, Malaysia. The inclusion criteria comprised location of the farms within the study state, a 5 km radius of the milk collection centre (MCC) of Stesen Pembiakan Ternakan (SPT) Sebrang, Keningau, which is presently producing dairy cattle, records on animal health and production, and where dairy cattle farmers were willing to participate in the survey. All farmers located within the specified region around the MCC were considered to be the target population. The location was selected for easy assessment and questionnaire administration since it served as the milk delivery point for most farmers in Keningau, Sabah.2.2. Development of Instrument and ContentsA structured questionnaire was used in this study. The instrument was developed after reviewing related literature and published papers regarding DCW. Specifically, the Terrestrial Animal Health Code developed by the World Organization of Animal Health (OiE) and the Animal Welfare Enactment [14] were used as the main template in developing the instrument.The questionnaire was structured into five sections. Section 1 was dedicated to obtaining respondents’ demographic information including gender, years of working experience, educational qualification, nationality, and age. Section 2 focused on the farm characteristics which included herd size, cattle breed, management system, availability of veterinary services, and average milk yield and production level. Management systems were categorized based on the provision of external grazing and pasture access (semi-intensive) or completely confined (intensive). The milk yield (production level) was used as the indicator of farm performance as described by Boniface et al. [17]. Farms producing an average of 10 L/cow/day were categorized as low producers, while those producing between 11 to 15 L and > 15 L/cow/day were considered medium and high producers, respectively.Section 3 consisted of 18 items designed to assess farmers’ awareness and their opinion about the indicators of DCW. Farmers were asked if they had been exposed to the term “dairy cattle welfare”, whereas the other items were based on specific ABMs and indicators of DCW such as the presence of physical injuries, alterations in feed intake and body condition, isolation from herd mates, and lying down behaviour. Responses to items in Section 3 were presented using a dichotomous approach (yes or no).Farmers’ opinions about DCW criteria were evaluated in Section 4. The items (n = 8) were selected from the common criteria employed on Canadian and American farms to assess the welfare of dairy cattle [11]. The items included mortality and morbidity rate, body weight, body condition, milk yield, cows’ response to human handling procedures, and complications from common procedures such as dehorning and treatment. A 5-point Likert scale ranging from strongly agree (score 5) to strongly disagree (score 1) was used to present the responses. In Section 5, farmers were asked to rank the welfare of their animals on a 10-point scale from 1 (very poor) to 10 (perfect) and to state the factors influencing animal welfare on their herds. The research team checked the items in each section of the questionnaire for validity and consistency. For ease of comprehension and to minimize challenges during administration, the questionnaires were first made available in English and later translated into Malay.2.3. Administration of QuestionnaireThe questionnaires were administered by a single researcher (S.S.L.) to each farmer in a paper format. The administration was conducted over 5 days (4–8 August 2019). Each respondent was allowed to select the preferred language (English or Malay), and those that experienced challenges either when attempting to answer the questions or had difficulty comprehending them were allowed to seek an unbiased explanation. Each participant responded to the questionnaire independently without any communication with the others.2.4. Statistical AnalysisData collected from the questionnaires were transferred into Microsoft Excel sheets for accurate and efficient data tabulation. IBM SPSS Statistics® Version 25 was used to carry out the statistical analysis. Descriptive statistics were used to summarize the farmers’ and herd characteristics, and all the variables were categorized and presented as frequency distribution and percentages. However, the average milk yield was a continuous variable and it was categorized into 3 groups as described by Boniface et al. [17] with little modification. Descriptive statistics were also applied to present the number of responses to each item and the frequency distribution for dichotomous questions. Responses to Section 3 (indicators of DCW) were scored according to the provision of either correct (score 1) or incorrect (score 0) answers, and summated for each respondent. Thereafter, the cumulative knowledge score was computed for each respondent and checked for normality using the Kolmogorov–Smirnov test. The mean score was used to categorize the respondents into those having poor, satisfactory, or good knowledge about the indicators of DCW. A similar approach was used to compute and categorize the responses in Section 5. The association between farmers’ demographics and DCW knowledge was investigated using independent t-tests and one-way analysis of variance (ANOVA) depending on the number of levels or categories in each factor. A p value < 0.05 was considered for significant associations between the knowledge score and independent variables. Responses to items in Section 4 (DCW criteria) were presented using stacked bar charts. 3. Results3.1. Descriptive ResultsThe characteristics of the respondents and farm features are provided in Table 1. A higher proportion of the farmers were male (73%, 22/30), but only 6 (20%) were below 50 years old. As expected, most were Malaysians (83.3%, 25/30) and 77% (23/30) had secondary education. For the farms, most were managed intensively (90%), while similar proportions were classified as having a small (37%; 11/30) or medium (43%; 11/30) herd size. Fifty-three per cent (n = 16) of the farms were considered to have low-producing herds.3.2. Farmers’ Awareness about the Term “Dairy Cattle Welfare” and Opinions Regarding DCW CriteriaRespondents were asked whether or not they had heard of the term “dairy cattle welfare”. Only 17 (56.7%) had heard of the term before this study (Figure 1). Regarding the DCW criteria, the majority of the farmers selected morbidity rate (80%, 24/30), followed by changes in body weight, body condition and milk yield (73.3%, 22/30), and responses of animals to human handling (66.7%, 20/30). However, mortality and culling rate (36.7%, 11/30) and complications from common procedures (50%, 15/30) were the least considered criteria.Figure 2 shows the farmers’ responses to the other 10 items considered as DCW criteria. Poor indicators of DCW that were the most recognized by the farmers were the presence of injuries (93.3%, 28/30) and reduction in feed intake (93.3%, 28/30) followed by sudden change in body condition (90%, 27/30). The least recognized indicators were isolation of cattle (43.3%, 13/30), carrying out veterinary procedures in the milking parlour (53.3%, 16/30), and reduction in lying down time (56.7%, 17/30).3.3. Farmers’ Knowledge about DCW Criteria and Associated FactorsThe farmers obtained a mean (±SD) score of 12.1 (±3.47) out of 18 possible marks (Table 2). Nine farmers (30.0%) were considered to have poor knowledge about the DCW, whereas 13 (43.7%) and 8 (26.7%) farmers had satisfactory and good knowledge, respectively.The mean knowledge score of respondents regarding DCW criteria was closely associated with level of education, breeds of cows kept, herd size, and production level (Table 3). Farmers with tertiary education had a significantly higher knowledge score (p = 0.029) compared to those with secondary education. Farmers with a large herd size (>100 cows) had significantly higher knowledge score relative to those with medium and small farms. Likewise, the knowledge score about DCW criteria was significantly higher among those considered as high producers compared to medium and low producing farmers. Farmers that had crossbreeds (Jersey x Holstein–Friesian and Holstein–Friesian x Sahiwal) had a significantly higher knowledge score (p = 0.03) than those who had local breeds (Sabah Friesian Sahiwal).3.4. Farmers’ Opinion on the Indicators of DCWThe majority of the respondents (60%, 18/30) disagreed with the statements about inappropriate practices and cattle behavior: 30% (9/30) gave a neutral response, and only 10% (3/30) agreed with the statements. Specifically, most of the respondents agreed with the statement, “Fat cows are a sign of good farm practice” and “Reluctance and kicking behaviour by the cow during milking is normal”. Meanwhile, the farmers slightly disagreed with the rest of the statements, but strongly disagreed with the statement, “Milk the cow as long as it produces milk”. Figure 3 shows the responses of the farmers arranged according to the proportions that disagreed with each statement. 3.5. Farmers’ Ranking of on-Farm Animal Welfare Status and Factors Influencing DCWFarmers were asked to rate the welfare of their animals on a 10-point scale and to state the factors responsible for the score selected. The mean (± SD) score was 8.0 (±1.83) (Table 4). A higher proportion (56.7%, 17/30) considered their animal welfare to be good, whereas 36.7% (11/30) and 6.7% (2/30) regarded theirs as satisfactory or poor, respectively. The majority of the farmers (33.3%, 10/30) mentioned facilities on the farm as the most important factor influencing animal welfare. Five farmers (16.7%) mentioned workers’ handling of animals, and three farmers (10%) stated management systems, whereas other factors presented by six farmers (20%) were categorized as miscellaneous.4. DiscussionThe proportion of respondents who either knew of or were ignorant of “Dairy Cattle Welfare” before this study was 56.7% and 43.3%, respectively. Most of the farmers had only secondary education, and this might explain their low exposure to DCW since the term is more likely to be introduced in tertiary schools. Among the categories employed in this survey to understand farmers’ current knowledge of criteria of DCW were morbidity rate, changes in body weight, body condition, and milk production level. Most of the farmers (80%) understood “morbidity rate”, whereas 50–60% of them selected either one or more of the latter indicators. Mortality and culling rate were least understood by the farmers as only 37% related the factors to DCW. By referring to the welfare code, mortality and culling can directly or indirectly indicate welfare status as it affects the productive lifespan of an animal. Mortality and a lower productive lifespan often result from the failure of animals to cope in an environment [18]. This finding might be related to the farmers’ unwillingness to reveal information about the present mortality and culling rate to researchers. The second-least understood statement was “complications from common procedures”. Complications could result in the reduction of feed intake or prolonged pain to the animal, which may precipitate poor welfare. Farmers showed little understanding of the statement and were probably unsure about its relationship to DCW. Moreover, pain detection in farm animals requires expertise and training, and such information may not have been available to the respondents. In contrast, farmers showed a better understanding of DCW when presented with a criterion such as physical appearance and cows’ response to handling procedures. These items are more situational and obvious manifestations of poor animal welfare [5]. For instance, physical injuries are easily detected during routine farm operations. Moreover, these items are in line with factors listed in Sabah’s Animal Welfare Enactment [14].In this study, 9 (30.0%), 13 (43.7%), and 8 (26.7%) respondents were considered to have poor, satisfactory, or good knowledge about DCW, respectively, and showed that the knowledge differed among the farmers. A similar study conducted in the United States reported that dairy farmers showed significant variation in their understanding of cattle welfare [10]. Factors such as educational qualification, level of training, and farming system variables influenced the perception of DCW among cattle farmers in the United States [10], Brazil [9], and Bangladesh [19].The factors associated with farmers’ knowledge about DCW criteria included educational qualification, animals’ production level, cattle breeds, and herd size. Nizam and Rahman [20] reported that a low level of literacy in Asia contributed significantly to the poor understanding of animal welfare. Clark et al. [13] also found that better-educated societies appeared to be more concerned and aware of farm animal welfare. The concept of animal welfare is commonly introduced in tertiary institutions or during specific training programs. Thus, farmers with better educational qualifications are more likely to have more access to information regarding animal welfare and understand the concepts. We also found that respondents with high producing cows and a large herd had higher knowledge scores compared to those with low-yield cows and small herds, respectively. This finding is in agreement with that of Kumar et al. [21] whereby farmers’ knowledge of DCW increased with their farm’s milk production. Accordingly, farmers or personnel involved in the management of large and intensive herds may be more exposed to welfare-related diseases or disorders such as mastitis and lameness. These conditions are commonly associated with large herds and high production levels [22,23]. In addition, farmers who were well trained and knowledgeable in DCW were able to detect clinical cases of mastitis and reduce on-farm prevalence [14]. Dairy cows use most of their energy for milk production, reproduction, and growth, which increases their susceptibility to diseases and disorders. An understanding of these events and challenges may prompt knowledgeable farmers to provide more care for their cows.Farmers who kept crossbreeds had a higher knowledge about DCW compared to those that had only local breeds. This finding could be linked to the fact that most large dairy herds in Sabah have mainly exotic and crossbred cattle because they give more milk. Another reason could be the cost implications of poor animal welfare in exotic breeds because they are more likely to experience lameness, mastitis, or reproduction inefficiency compared to local breeds. Hence, the farmers who keep exotic breeds may prioritize DCW and identify the indicators of poor animal welfare better than those who have local breeds.On the other hand, the frequency of veterinarian visits did not affect the farmers’ level of knowledge. This finding is in contrast to the reports by Wolf et al. [10] where local veterinarians had the second most influential role in influencing farmers’ knowledge of DCW. Other studies [24,25,26] reaffirmed that cooperation and communication between farmers and veterinarians could assist in improving animal welfare. This communication can take the form of education or herd-health programs, which allows veterinarians to be in contact with the farmers. Likewise, years of experience in farming was not associated with the farmers’ level of knowledge of animal welfare in this study. The lack of knowledge transfer between veterinarians and farmers regarding the topic may play a role in the finding.The majority of the farmers also had a good opinion about DCW based on the criteria applied. They recognized that lower milk yield may be an indicator of poor DCW as they understood that reduction in milk production may indicate the presence of animal discomfort or stress. This assertion by the respondents is appropriate because common conditions such as mastitis and hoof lesions remain major components of poor welfare and may affect milk production. Furthermore, most of the respondents affirmed that the dry period is vital for ensuring optimal production in dairy cows. This was reflected as most of them disagreed with the statement, “cows should be milked during the last or two months of gestation”. The dry period allows the mammary gland to prepare for the next lactation and serves as an important period for the treatment and prevention of mastitis [27]. Furthermore, the vast majority of farmers, (93%, 28/30), understood that reduction in feed intake can be an indicator of poor animal welfare as it could ultimately lead to the reduction in a cow’s body condition score (BCS).The important findings in this study regarding the farmers’ response to indicators of DCW were their position on the role of cows’ body condition, behaviour during milking, and hygiene. When presented with the DCW criteria, most of the respondents agreed that sudden changes in body condition indicated poor welfare. However, they did not recognize that cattle with a high body condition score may also show signs of poor farming practices. Dairy herds with a high body condition score can be predisposed to lameness [28], and reduced dry matter intake may further lead to periparturient metabolic disorders [29]. Furthermore, animals may respond to human handling in different ways, and it is considered to be an indicator of animal welfare. Dairy cows may respond positively to human handling, which is expressed as a reduction in flight distance [30]. On the other hand, a negative response is characterized by an increase in flight distance such as reluctance to enter a chute [30]. Farmers may not have an appropriate understanding of the topic as most failed to recognize that cows that express reluctance to enter the milking parlour or kicking during milking display potential signs of poor DCW. However, kicking, lifting and stepping during milking was associated with mastitis [27]. Specifically, stepping is related to the cow’s discomfort during milking and kicking indicates pain felt by the cow, probably from a teat lesion [27]. A high proportion of respondents (>70%) agreed that the hydration status of dairy cows was an indicator of DCW. Almost all of them (>90%) agreed that the presence of injuries was an indicator of poor DCW. In Brazil, deficiency in the provision of drinking water and the occurrence of hock injuries was found to affect local cattle [6]. Injuries to the limbs especially in the hock region were significantly correlated to the prevalence of lameness and could be an indicator of inadequate standards of comfort [8]. Likewise, most of the surveyed farmers agreed that the presence of excessive feces on the cows’ body indicated of poor DCW. The degree of manure contamination on the body, upper flank and limbs are used to assess hygiene [31]. Few studies reported that poor hygiene, characterized by excessive manure contamination of the body parts and limbs, was associated with a high prevalence of subclinical mastitis [32] and lameness [16], respectively. However, only 56% (13/30) agreed that reduced lying time is a sign of poor animal welfare. The lying down duration remains one of the most important ABMs in dairy cattle, and alterations in the behaviour have been associated with an increased risk of hoof injury [33] and reduced perfusion of the mammary gland [34].Other items considered as indicators of DCW in this study included grooming and licking behaviour. Farmers seem to recognize them and the role they play in DCW because they classified behaviour into favourable and unfavourable groups. Excessive licking and grooming can be a sign of a mineral deficiency as cattle with a salt deficiency often show cravings or abnormal appetite for it [35]. They tend to lick various objects such as rock, wood, soil and even the sweat of other animals in the event of a salt deficiency. Calves without supplementary minerals would spend more time grooming, licking pen structures and ear sucking [36,37]. Napolitano [38] described the positive indications of mutual and self-grooming in dairy cattle and the negative indications associated with over-grooming, which includes high social tension. Mutual and self-grooming have been associated with hygiene, a comfortable living space and the reinforcement and stabilization of social relationships, which are integral aspects of behaviour in mammalian species [39,40].Most of the farmers ranked their on-farm animal welfare status as good (57%), 37% considered theirs as satisfactory, while only 2 farmers described theirs as poor. This is in contrast with the results obtained regarding their knowledge about the criteria of DCW, where only 27% were considered to have a good understanding of the concept. Moreover, farmers had little understanding of the important indicators of DCW. Therefore, this reflects an underestimation of the animal welfare status on the respondents’ farm. To test such a hypothesis, future studies might consider comparing farmers’ and researchers’ welfare assessment of dairy cows.This is the first study in the region to report the level of knowledge and attitude on dairy cattle welfare DCW among dairy farmers in Keningau, Sabah. In conclusion, the level of awareness about the topic needs to be improved. The majority of the farmers (70%) had satisfactory to good knowledge of DCW criteria; however, they need to be educated about the indicators of DCW, especially cows’ responses to human handling, behaviour during milking, and physical appearance. The factors observed that improved the farmers’ knowledge about DCW criteria included high education, large herds, high producing cows, and exotic breeds.Limitations inherent in this study are well identified. The findings in this study are specific to dairy farmers located within the study region and the data are too small for any form of extrapolation to Malaysia dairy farms. The surveyed farmers were not randomly selected, as participation was based on their willingness and signing a consent form. Moreover, DCW is a broad aspect of the dairy industry, and the topics covered in the present survey require more elaboration for a better understanding of farmers’ opinions and knowledge about it. Nevertheless, this study serves as a preliminary work and an addition to our existing knowledge on the subject area in Malaysia dairies.5. ConclusionsThis is the first study in the region to report the level of knowledge and opinion on DCW among dairy farmers in Keningau, Sabah. In conclusion, the knowledge and opinion of DCW differed among the farmers in this study. This was evident in the disparity in understanding the basic criteria and indicators of DCW. Farmers with high education, large herd size, and high production levels showed a better understanding of DCW. Specifically, the farmers had a good understanding of it concerning changes in body weight, body condition and milk yield. However, their understanding and opinion on cattle behaviour and physical appearance needs to be improved.

animals : an open access journal from mdpi

10.3390/ani12040433

PMC8868463

The objective of this study was to describe the incident reporting of harness racing in New Zealand. Retrospective stipendiary stewards’ reports of race day events during the 2015/16 to 2016/17 racing season were examined to describe the reasons and outcomes for race day veterinary examinations of Standardbred horses in New Zealand. The primary reason for examination of horses after a race was due to poor performance. Poor performance was considered if a horse’s performance in the race was lower than its previous race, or lower than expected as reflected by the odds at the tote (reflecting the amount of money placed/gambled on the horse via the official betting agency). The lack of fatalities and injuries reported indicates a low risk profile in harness racing and highlights the stewards’ role in maintaining racing integrity and animal welfare.

The objective of this study was to describe the incident and non-incident reporting of harness racing in New Zealand, the primary injury and reporting outcomes, and to examine horse- and race-level variables associated with the odds of these outcomes. Retrospective stipendiary stewards’ reports of race day events during the 2015/16 to 2016/17 racing seasons were examined. The number of incident and non-incident events and binomial exact 95% confidence intervals (CI) were calculated per 1000 horse starts. Most reports were for non-incidents and an examination was requested for poor performance (11.06 per 1000 starts (95% CI = 10.23–11.89). Races with more than eight participants were 1.9 (95% CI = 1.13–3.4) times more likely to have an incident than races with eight or less participants. The low incidence of significant injuries such as fractures (0.13 per 1000 starts (95% CI = 0.03–0.23) reflects the lower risk of injury in harness racing compared to Thoroughbred racing. The high incidence of poor performance reports highlights the steward’s role in maintaining animal welfare to a high standard.

1. IntroductionWhile smaller in participation numbers and betting turnover than Thoroughbred racing, harness racing represents a significant racing industry and has received considerably less attention in the scientific literature than Thoroughbred racing. Harness racing is conducted using Standardbred horses that race at either a trotting or pacing speed whilst drawing a two-wheeled cart called a sulky [1]. Within New Zealand, the structure of the harness racing industry and Standardbred breeding industry have been described [2,3]. There are limited data on the training of standardbreds internationally, [4] or from within New Zealand [5], and little information on injuries and injuries associated with racing events [6]. This trend is also observed in the international literature with limited data on racing related injuries in standardbred harness racing [4,7].Within the regulatory process of racing, most jurisdictions utilize a process of stipendiary steward reporting and veterinary reports. These reports are routinely collected during race meetings and are published as part of the transparency of racing integrity [7]. Reports are identified as either a non-incident or an incident report. Non-incident reports occur when there has been no identifiable “event” during a race, and routine screening of horses is required as part of the ongoing regulatory and integrity process. The screening requests often focus on a horse or horses that may not have performed up to expectations, or if a horse’s health is questioned (e.g., suspected epistaxis). An incident report is the result of an “event” before or during a race (such as a horse collision, stumble, fall or the horse “breaks its stride”) that requires a horse to be examined. Both reports involve an assessment from the designated veterinarian(s) on duty for the race meeting.Stipendiary stewards’ reports provide an indication of the robustness of the regulatory process and the screening of horses during a race meeting [7]. When combined with race data, this information provides the opportunity to describe the incidence of injuries (from mild lacerations to catastrophic injury) and the odds of these events, or outcomes, with different horse and environment level variables. The collection of such data is important for the monitoring of industry practice and to optimize horse welfare. Such data permit evidence-based changes to be made to management and the structure of racing to meet the industry’s duty of care to the horse racing in it. External to these is the obligation of the industry to minimize injury and loss to meet its social license to operate [8].To date, most of the publications describing race day events from stipendiary stewards or race day veterinary events have examined the Thoroughbred racing industry. International and domestic data have suggested that potential risk factors for race day injury in Thoroughbred racing include track surface, track condition, race distance, race class, age of horse, training intensity, and number of starters [9,10,11]. Within harness racing, Physick-Sheard, Avison, and Sears [4] identified that sex, age, track class, performance history, and workload affect the likelihood of horse mortality on race day. However, they did not examine non-fatal injuries. Australian studies of harness racing steward reports have been carried out to look at the incidence of injuries, mortality, and reasons for poor performance providing a benchmark to assess ongoing improvements in welfare [7]. Lameness was the most common finding in post-race examinations with 2.1 cases per 1000 starts, followed by poor performance/heat stress with 2.04 cases per 1000 starts. To monitor changes in New Zealand harness racing welfare over time, the prevalence of the current incident and non-incident reporting needs to be measured. Therefore, the objective of this study was to describe the incident and non-incident reporting during the 2015/2016 and 2016/2017 harness racing seasons in New Zealand, the primary injury and reporting outcomes, and to examine horse- and race-level variables associated with the rate of the reporting of these outcomes.2. Materials and MethodsData were obtained of all race starts during the 2015/16–2016/17 racing season as an excel spreadsheet from Harness Racing New Zealand, the official registration body for Harness racing in New Zealand. An Excel spreadsheet of all stipendiary steward reports for the same racing seasons were obtained from the Racing Integrity Unity, the official racing compliance organization for all three racing codes within New Zealand. Each stipendiary steward report recorded the date, racecourse, race number, horse name, and other information relevant to the incident reported such as the reason for the requested report and the findings of the veterinary examination. To obtain information about racing surface (grass and all-weather) and condition (fast, good, dead, slow, heavy, easy, and slushy), race type and pace, horse age, and other relevant track and horse information, stipendiary steward reports were cross-referenced with the official race start records and results from the same period.Statistical AnalysisData were exported to RStudio (version 3.5. 1, 2018; R Foundation for Statistical Computing, Vienna, Austria) for manipulation and analysis. After merging of the two datasets, data were cross validated using the horse’s official race name as the unique identifier. Apparent errors or inconsistencies in data (i.e., misspelt horses names) were then checked manually against the official formal transcript of the relevant stipendiary stewards’ report hosted and archived on the Harness Racing New Zealand website. Horses that did not race, or steward reports that were miscategorized (thoroughbred reports), were removed from the dataset. Six records were excluded due to mis-entered values with data that prevented linkage between the two databases. There were 87 reports that recorded an event occurring prior to the beginning of the race meeting that resulted in horses being withdrawn from the race meeting. Using the official descriptors provided within the datasheets, reports were coded for analysis as incident and non-incident reports. Non-incident reports occur when there has been no identifiable “event” during a race, and routine screening of horses is required as part of the ongoing regulatory and integrity process. An incident report is the result of an “event” before or during a race (such as a horse collision, trip or fall) that requires a horse to be examined. Both reports involve an assessment from the designated veterinarian(s) on duty for the race meeting. Multiple phrases and spellings (abbreviations) were used to describe reasons for a horse examination and the outcome findings. These were manually coded and collapsed down into 14 categories, these being Arrhythmia, Cardiac failure, Laceration/abrasion, Lame, Musculoskeletal injury (Fracture), Other musculoskeletal (MS) issues, Unknown musculoskeletal pain, no observable abnormalities detected (NOAD), Poor recovery, Respiratory issues, Previous injury, Bleeders (epistaxis), Unknown and Miscellaneous.Track conditions on grass surfaces were defined as fast, good, dead, slow and heavy. Track conditions on all weather surfaces were defined as fast, good, easy or slushy.Statistical analysis was conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Descriptive statistics (mean, median and proportion) were used to describe the data at a season, track, and race level. Incidence rates are described as the number of events per 1000 starts with their 95% confidence intervals (Parkin [9])Data were tested for normality and if non-parametric, differences between groups were tested using the Kruskall–Wallis test and reported as median and interquartile range (IQR). The distribution of stewards’ reports, and the underlining demographics of the industry and race conditions were initially examined using the Chi-Squared test. Univariable logistic regression was used to screen variables for association with both incident and non-incident reporting. 3. Results3.1. Harness Racing DataDuring the 2015/16–2016/17 seasons, 4211 horses had a race start, over half of the racing population were geldings (53%), followed by mare and fillies (43%), then stallions and colts (4%). Of these 4211 horses, 3034 (72%) had at least one start in the 2015/16 season and 2923 (69%) had at least one start in the 2016/17 season. These horses had 54,702 racing starts, of which 27,819 (50.9%) were during the 2015–2016 season and 26,883 (49.1%) during the 2016–2017 season. Most of starts were for horses in pacing races (72% of starters and 73% of races). The racing took place at 40 racecourses, 22 of which were grass tracks with the remainder being all weather tracks. Most starts (n = 44,280 (80.9%)) were on all-weather surfaces and 10,422 (19.1%) on grass. Of the 44,280 starts on all-weather surfaces, the majority of the racing surface conditions were described as “fast” (n = 32,678 (73.8%)), the remaining being described as on a “good” (n = 6574 (14.8%)), “easy” (n = 2375 (5.4%)) or “slushy” (n = 2653 (6%)). Of the 10,422 starts on grass, surface conditions were described as “fast” (1255 starts or 12%), “good” (6956 starts or 67%), “dead” (1367 starts 13.1%), “slow” (712 starts or 6.8%) and “heavy” (132 starts or 1.3%). There was a median race distance of 2200 m (IQR 2000–2600) with 97% (53088/54702) of starts being over 1609 m (1 m). The median mile rate was 2.04 min (IQR 2.0–2.09) for pacing races and 2.14 min (IQR 2.09–2.20) for trotting races. Most races were at racetracks located in the southern region (South Island) (65.6% 3370/5139) followed by the northern region (26.2%, 1346/5139) (Top half of the North Island) with few races in the central region (Bottom half of the North Island) (8.2%, 423/5139). There was a median of 11 starters per race (IQR 9–13). The horses participated in a median of 7 (IQR 3–10) races per season with no difference between the 2015/16 and 2016/17 season for number of starts. Younger horses participated in fewer starts per season, 2-year-olds in a median of 3 (IQR 2–5) starts per year, 3-year-old horses in 6 (IQR 3–10) starts and 4+yo horses in 9 (IQR 4–15) starts per season (p < 0.001). A detailed description of the distribution of age and sex is presented in Table 1. 3.2. Stewards ReportsThe 1001 records that were included in this study were coded into two categories: (a) screening due to major events occurring before (from the start of the race meeting), during and after the race as “incident reports” (n = 131) and (b) screening due to poor performance or other non-incident concerns, often as part of routine racing integrity screening, coded as “non-incident reports” (n = 870). Most of the stipendiary stewards’ reports described events relating to races (96.3%, 964/1001) rather than those occurring either before the race (2.7%,27/1001), post-race (0.80% 8/1001) or unknown/not described in the report (0.20%, 2/1001).There were only nine (0.16 times per 1000 races, 95% CI = −0.08–0.41) musculoskeletal fractures were reported in the stipendiary stewards’ reports for both seasons. 3.2.1. Incident ReportsAn incident report occurred every 2.4 per 1000 starts (95% CI = 2.0–2.8). The distribution of incident reports across race types reflected the underlying racing population with 77.4% (102/131) associated with pacing races and the remaining 22.6% (29/131) with trotting races (p < 0.05). There was no association of track surface with the frequency of incident reports (p = 0.674). Track conditions were reported as fast for 60% of incidents followed by good (32.1%), easy (2.3%), slushy (3.8%) and slow (0.8%) and dead (0.8%). There were no incidents on heavy tracks. There was no association of track condition with the rate of incident reporting (p = 0.353). There was a positive association of field size with incident reports such that an incident was 1.9 [95% CI = 1.13–3.4, (p < 0.001)] times more likely to occur in a race with 9 or more participants than in a race with 8 or less participants. Incident reports reflected the average number of participants per race on each surface. There were a variety of reasons recorded for the requesting of an incident report with the largest category being due to a horse falling (0.79/1000 starts [95% CI = 0.66–0.91]; 33.0%), primarily during the race (64.3% of falling horses). The major clinical finding for falling horses was integument lacerations and/or abrasions (52.4%) with the remaining 47.6% having no clinical finding. Only 8.8% of horses reported in incident reports (0.13 per 1000 starts, 95% CI = 0.03–0.23) were coded as a musculoskeletal (MSI) injury or sudden collapse. Other reasons for the requesting of an incident report were “galloped on (contact with another horse)” (6.2%), pulled up (16.92%) and those coded as miscellaneous (35.4%). Less than half of the incident reports (38.46%; 0.91/1000 starts, 95% CI = 0.84–0.99) had no observable abnormalities detected in the clinical findings (NOAD). The reporting rates of clinical observations across all incident, and non-incident stipendiary reports are presented in Table 2. Overall, integument lacerations and abrasions accounted for the largest number of incident cases (32.65%), with the remaining categories representing less than 5% of cases each and with no difference in the rates per 1000 starts. 3.2.2. Non-Incident ReportsThere were 871 stipendiary reports that were classified as non-incident and were not associated with a race incident. The major reason for the request of a non-incident report (70.3%, 611/871) was for the routine post-race screening of horses or screening of horses performing below expectations (11.2/1000 starts, 95% CI = 10.32–12.01). Stewards were responsible for the requesting most of the poor performance examinations (95% 582/612). The distribution of age across all poor performers is presented in Figure 1 and reflected the underlying population distribution (p = 0.012). Over the two seasons, 87.3% (533/611) of poor performers were in the ages between 3 and 6 years old. There was a maximum of two reports for poor performance per horse in a season, with 24 horses having two reports each season. There was an association of field size with non-incident reports such that a non-incident report was 2.1 (95% CI = 1.73–2.59) times more likely to be requested in a race with 9 or more participants than in a race with 8 or less participants (p < 0.001).The largest category for clinical finding of a poor performance exam was NOAD during veterinary examination (67.1%, 410/611). A small proportion of clinical findings was classified as poor recovery (6.8%), Arrhythmia (4.6%), respiratory issues (2.87%), and Lacerations/abrasions (2.64%). Musculoskeletal injuries among poor performers (1.61%) were coded as either tendon or ligament strain and myositis. The largest category of findings with the non-incident reports was NOAD (53% of all non-incident reports). The main clinical findings were integument lacerations and abrasions (10.2% of non-incident reports), followed by poor recovery (9.30%), arrhythmia (5.05%) and respiratory issues (4.48%). 4. DiscussionThe distribution of race type reported was similar to previous reports and reflects the large focus of the New Zealand harness racing industry on pacing rather than trotting races. In contrast to racing in the USA and other jurisdictions with an emphasis on racing over a mile (1609 m), many of the races in New Zealand were at a distance greater than 2000 m. Racing over these longer-distance races was reflected in the lower mile rate for both trotters and pacers [6]. The participation structure of harness racing in New Zealand has a large proportion of breeder-owner-trainers and this was reflected in the older age profile of the horses racing in New Zealand [5], which was similar to a report based on the New South Wales (Australia) harness racing population [7]. There was a reduction in the median number of horses racing and the number of starters compared to the previously published literature reflecting the reported contraction in the New Zealand racing industry [6,12].Racing was concentrated on all-weather surfaces, around a few major venues or regions, predominantly in the South Island. This distribution reflects the underlying breeding and participation base for harness racing in New Zealand [13]. The distribution of the stewards reporting reflected this underlying distribution of harness racing in New Zealand. This in association with the relatively high level of screening (non-incident reports) indicates that the reporting should provide a good reflection of the underlying incidents and injuries observed with harness racing. The data used were collected retrospectively and cross validated with the published steward’s reports. At the time of data collection, reports were completed by officials using a pro forma sheet. The use of a pro forma reporting sheet provided some consistency in the type and level of reporting. However, despite this, there were still some limitations in ability to precisely identify the anatomical site affected and occasional inconsistency in descriptors used. During the 2018/19 racing season, an online reporting system was implemented based on the pro forma sheet. This online system will provide data that reflect the data previously captured using the paper-based system and the internal controls within the online system should ensure consistency in terms used with recording.The incidence of musculoskeletal fracture during both observed seasons was lower than the incidence reported in New Zealand Thoroughbred racing (0.48 per 1000 starts) using a similar reporting framework [11]. Fractures were recorded in both incident and non-incident reports, and this reflects how an incident and non-incident report was defined. A steward may class a report with a musculoskeletal fracture as being a non-incident if there was no “event” such as a horse falling or colliding with another horse. It is also possible that some non-displaced fractures are only identifiable, because a horse performed below expectations or presented with a lameness once the horse had cooled down post racing. Within the stipendiary stewards’ reports, the description of the fractures were based on broad anatomical locations with seven of the nine fractures involving the distal limb, and predominately the first phalanx and the metacarpophalangeal joint. Fractures at these sites are typically attributed to accumulated cyclic load [14,15]. The difference in the reported fracture incidence between Thoroughbred racing and harness racing highlights the differences in the pattern and magnitude of load accumulated and the intensity of exercise between codes during both training and racing [1]. The harness racing horses had a median of 7 race starts per season over 2200 m, which was greater than the number of starts and race distance (5 starts/year, 1400 m) reported in the New Zealand Thoroughbred racing industry [13,16]. While harness racing horses in New Zealand typically acquire a greater number of load cycles during training and racing [5,17], the nature and magnitude of the load on the distal limb, and the first phalanx and metacarpophalangeal joint specifically, is less for the trotting/pacing compared to galloping [18]. The lower strain per load cycle with harness racing (trotting and pacing gait) may account for the reduced incidence as fatigue life decreases exponentially with increasing strain [19].The rate of reporting for lacerations was moderately lower than thoroughbred racing, but similar to data reported for harness racing in Australia, which operates under similar rules and conditions to those in New Zealand [7]. The relatively higher frequency of reporting of lacerations/abrasions compared to other categories in part reflects the gait of harness racing (pace and trot) and the relatively higher risk of interference (lower limb being struck by the contralateral limb) [20]. The reporting of lameness was lower than values from Australia using a similar reporting process and the rate of reporting lameness from non-incidents was four-fold greater than with an incident report [7]. These data possibly reflect the consistent reporting that much poor performance can be associated with MSI, and that many low-grade injuries or lameness only become obvious once the horse has cooled down post-race. There was also a low rate of reporting of epistaxis (bleeders) in the dataset compared to an Australian study [7]. This could be explained by the relatively long stand down periods associated with horses that present with epistaxis on race day in New Zealand and hence a possible hesitancy of trainers to present a horse for racing if they suspect it may have an episode of epistaxis [10]. There was no association of track surface with the rate of incident reporting, despite the anecdotal observation that the lower-grade racing was conducted on grass surfaces rather than on all-weather. As most of the racing was conducted on all-weather tracks there may have been significant heterogeneity in the class of racing on all-weather surfaces that prevented the ability to differentiate any surface or race quality effect. In New Zealand, grass track meetings are held in the summer when weather conditions are optimal and the number of races on grass is limited to preserve the ground conditions. The type of meeting that occurs on a grass track differs to what occurs on all-weather surfaces and are part of a summer circuit where trainers take their horses on a trip to several meetings within a regional circuit. Within the current data set, it was also difficult to clearly differentiate race grade and track surface to compare the race grade effect. Harness racing data from Canada indicates that racing at higher ranking tracks was associated with a higher incidence of sudden death and accidents for horses failing to finish a race than in races at lower-ranking race tracks [4]. A similar pattern of greater MSI with higher grade racing was reported in Australian Thoroughbred racing with a higher incidence of musculoskeletal injuries reported on metropolitan racetracks compared to country track [9]. Within New Zealand, the closest analogue to metropolitan tracks for harness racing would be the all- weather surfaces, however, even on all-weather surfaces there is a large variation in race grade (quality), which may explain the lack of apparent clustering observed. There was no association of type of going with the rate of incident reporting. This may be explained in part by most races in New Zealand being on fast or good all-weather tracks. In New Zealand, most all-weather surfaces for harness racing consist of a predominately hardpacked lime substrate and thus there is limited variation in the material properties of fast and good all-weather surfaces and the relative speed differential between races on these going ratings over these longer distances is relatively moderate (no significant differences between categories in mile rate). There is an active program managed by Harness Racing New Zealand to ensure consistency in track surface between tracks. All harness tracks are inspected annually and concerns about track conditions are corrected prior to race day if possible. If a track is not suitable either prior to the start of the race day or during the race day, the track condition deteriorates (e.g., excessive rainfall), a meeting will be cancelled or postponed under the discretion of the stipendiary steward. By having these measures in place, injuries as a result of poor track conditions are minimized.The number of participants in a race increased the likelihood of either an incident or non-incident report occurring. Increases in participants increases the likelihood of accidents such as horses colliding with each other or being galloped on. Similar results are reported in Thoroughbred racing where race-level factors for injuries included the type of race track, the track conditions, race distance, and field size [21]. However, the increase in the number of non-incident reports with the number of participants may reflect stewards inadvertently performing routine checks relative to the starters in the race. Most of the requests for a steward report were due to poor performance and the pattern of non-incident reporting appeared to be routine screening of race participants to maintain racing integrity and horse welfare. The selection of the horses for routine screening is often based on horses that had a performance in the race that was lower than the horse’s previous race, or lower than expectations as reflected by the odds at the tote (reflecting the amount of money placed/gambled on the horse via the official betting agency). Requests for examination based on the category poor performance reflected the distribution of the underlying starter population with respect to age and sex of the horse. The finding of the major category for poor performance being NOAD reflects this observation that much of these examinations were part of the routine screening process and a low frequency of undetected musculoskeletal injuries not associated with an incident during the racing event.The primary issues identified with the social license to operate with horse racing appear to focus on the concept of injury and risk of injury to the equine participants [8]. Routine screening data, such as stewards’ reports provides metrics for industry performance. The level of stewards reporting during harness racing in New Zealand indicates that these data are representative of the industry and provides robust metrics of the industry’s performance. The low incidence of significant clinical findings from this high level of reporting and screening indicates that harness racing in New Zealand is meeting its duty of care to the horses racing in it and the primary issues associated with the social license to operate with horse racing. 5. ConclusionsThere was a robust level of reporting within the harness racing industry during the two seasons examined. The occurrence of non-incident and incident reports was limited emphasizing the smaller risk in New Zealand harness racing compared to Thoroughbred racing. The low fracture rate reported reflects anecdotal reporting of lower rates in harness racing compared to flat racing thoroughbreds and can be described by the different racing and training associated bone strain and fracture risk between the codes. The high level of reporting by the stewards reflects the role stewards have in maintaining of racing integrity.

animals : an open access journal from mdpi

10.3390/ani13081279

PMC10134980

The discovery of antibiotics was a breakthrough in medicine. However, bacterial defense mechanisms driven by genetic variation resulted in resistance to these compounds relatively quickly. Moreover, new classes of antibiotics have not been developed for 30 years. Within the European Union, the EU Parliament and Council Regulation No. 2019/6, which concerns veterinary medicinal products, is currently in force. The current goal is to reduce the use of antibiotics and to stop the rise of drug resistance in bacteria because such antimicrobial resistant organisms can be transmitted to humans through the consumption of animal products or direct contact with animals (dogs, cats, etc.). For this reason, there is a growing interest in essential oils (EOs). As natural mixtures (usually of terpenes and their derivatives), they may consist of about 20–60 components with 1–3 dominant component(s). An important feature of EOs is their hydrophobicity, which allows them to react with lipids present in bacterial cell membranes and mitochondria, disrupting the functioning of cell structures and consequently making them more permeable to other components or antibiotics. In the present manuscript, the activity of two EOs (patchouli and tea tree) was assessed, and their interaction with gentamicin and enrofloxacin was studied.

In this paper, we show the effect of some essential oils (EOs) on staphylococci, including multidrug-resistant strains isolated from pyoderma in dogs. A total of 13 Staphylococcus pseudintermedius and 8 Staphylococcus aureus strains were studied. To assess the sensitivity of each strain to the antimicrobial agents, two commercial EOs from patchouli (Pogostemon cablin; PcEO) and tea tree (Melaleuca alternifolia; MaEO) as well as two antibiotics (gentamicin and enrofloxacin) were used. The minimum inhibitory concentration (MIC) followed by checkerboards in the combination of EO-antibiotic were performed. Finally, fractional inhibitory concentrations were calculated to determine possible interactions between these antimicrobial agents. PcEO MIC ranged from 0.125 to 0.5 % v/v (1.2–4.8 mg/mL), whereas MaEO MIC was tenfold higher (0.625–5% v/v or 5.6–44.8 mg/mL). Gentamicin appeared to be highly prone to interacting with EOs. Dual synergy (38.1% of cases) and PcEO additive/MaEO synergism (53.4%) were predominantly observed. On the contrary, usually, no interactions between enrofloxacin and EOs were observed (57.1%). Both commercial EOs were characterized by natural composition without artificial adulteration. Patchouli and tea tree oils can be good alternatives for treating severe cases of pyoderma in dogs, especially when dealing with multidrug-resistant strains.

1. IntroductionAntibacterial therapies are mainly based on antibiotics. However, they are not always effective and can sometimes be invasive or cause side effects (hair or hearing loss, diarrhea, irritability, lack of appetite, etc.). Moreover, they increasingly encounter antibiotic resistance, which is a therapeutic and economic problem.Purulent dermatitis (pyoderma) is the most common bacterial skin disease of dogs accompanying other dermatological problems, manifesting as a complication of the underlying disease, such as allergies (food allergy, atopic dermatitis, allergy to flea bites), internal diseases (hypothyroidism, adrenal hyperfunction), seborrhea and inflammation of the sebaceous glands, parasites (Demodex canis, scabies, etc.), hormonal fluctuations, anatomical predispositions (e.g., skin folds), or abnormal functioning of the immune system [1]. Puppies that have not yet developed a level of immunity and older dogs or steroid-treated individuals are the most vulnerable [2]. Pyoderma is much more common in dogs with short coats than in those with longer hairs, where the dense hair and undercoat provide a better barrier against bacterial penetration [3].The symptoms that occur in pyoderma are varied and depend on the type of inflammation, the area of the skin and the intensity of the disease. The most common include erythema, blisters, itching, hair loss (alopecia), ulceration, coat and skin discoloration, scabs, pustules, and purulent lesions. It is also possible that skin lesions have an endocrine basis with other symptoms, such as lethargy, weight gain, or excessive thirst [4].In the case of pyoderma in dogs, the most commonly isolated pathogen is a Gram-positive coccus classified as a Staphylococcus pseudintermedius. S. pseudintermedius is believed to colonize the skin and mucous membranes in small numbers in 80% of healthy dogs. Bacteria less commonly found in purulent lesions include other coagulase-positive staphylococci (S. aureus or S. schleiferi subsp. coagulans), Gram-negative bacilli—such as Pseudomonas aeruginosa, Proteus spp., and Escherichia coli—or yeast-like fungi (Malassezia sp., Candida sp.) [5]. These microorganisms are the natural commensal microflora of the skin in dogs; nevertheless, when they are abundant and the animal’s immune system declines, they can be the cause of skin lesions.Antibiotics (and other antimicrobial agents not classified as antibiotics) used in the treatment of pyoderma in dogs should be characterized by a broad spectrum of action and high efficacy against the abovementioned microorganisms, i.e., mainly S. pseudintermedius. The drug must reach high concentrations in the skin and have as few side effects as possible. The most important attribute is strong bactericidal activity. Cephalosporins (e.g., first-generation cephalexin or third-generation cefovecin (Convenia)), fluoroquinolones (enrofloxacin, marbofloxacin, ciprofloxacin), aminoglycosides (amikacin, gentamicin), lincosamides (clindamycin), and amoxicillin/clavulanic acid are among the most commonly used antibiotics in the control of purulent dermatitis in dogs [6].In recent years, there has been growing antibiotics resistance in staphylococci isolated from dogs. In addition to S. aureus (methicillin-resistant Staphylococcus aureus—MRSA), methicillin-resistant Staphylococcus pseudintermedius (MRSP) strains have appeared. This means resistance of these bacteria to antibiotics included in the β-lactam group. Additionally, these pathogens tend to be multidrug-resistant, which poses a problem in selecting the correct antibiotic during treatment [7]. They occur in the pharynx, nasal cavity, rectum, and periosteal area as asymptomatic carriage. Moreover, these bacteria are often isolated from dog bite wounds [8].Because of this, there has been growing interest in essential oils (EOs) and their use in medicine, cosmetology, and the food industry. EOs are obtained from various plant materials (leaves, buds, fruits, flowers, herbs, branches, bark, wood, roots, and seeds) via steam distillation through their maceration with fats or pressing [9]. Essential oils are volatile, liquid, transparent or rarely colored, and soluble in fat and organic solvents. As natural mixtures of an extraordinarily complex nature, they can consist of up to 100–200 chemical compounds in a wide variety of concentrations: several are present in high concentrations (a total of 20–70%) compared to other components (trace amounts). The amount varies depending on the part and species of the plant. They are chemical derivatives of terpenes and terpenoids [10].Despite a number of studies on the composition of individual oils, detailed knowledge of their mechanism of action is still limited. Of particular importance is determining the effects of EOs on various microorganisms, especially how they act in combination with other antimicrobial compounds [11].Essential oils are believed to have important antiseptic, antibacterial, antiviral, antioxidant, antiparasitic, antifungal, and insecticidal activities [12]. An important characteristic of EOs is hydrophobicity, which allows them to dissociate from the lipids present in the bacterial cell membrane and mitochondria, making them more permeable by disrupting cell structures. This ultimately results in bacterial cell death due to the leakage of critical molecules and ions from the bacterial cell at a high rate [13]. EOs can thus serve as a powerful tool for inhibiting the growing phenomenon of bacterial resistance [14]. The overall concept of some antimicrobial synergy is based on the principle that combination of two or more antimicrobial agents may enhance efficacy, reduce/decrease toxicity or side effects of one of agent used, increase bioavailability, lower the dose of, e.g., antibiotics, and reduce the advance of antimicrobial resistance [15]. New and highly effective antimicrobial combinations of drugs that contain natural product(s) have recently become a research priority.The aim of the study was to evaluate the antimicrobial activity of patchouli and tea tree essential oils applied alone and in combination with gentamicin and enrofloxacin as an alternative in the treatment of purulent skin inflammation in dogs against S. pseudintermedius and S. aureus, especially in regard to multidrug-resistant isolates.2. Materials and Methods2.1. Bacterial Strain Origin and IdentificationAll strains were isolated and collected in the veterinary laboratory (West Pomerania, Szczecin, Poland) during routine tests of swabs/skin scrapings from acute pyoderma cases in the years 2019–2021. Strains were then systematically banked (VIABANK™, MWE Medical Wire, Corsham, UK) in order to produce autovaccines and kept in a frozen state (≤−30 °C) until research (no animals were directly involved in this experiment). A total of 12 S. pseudintermedius and 7 S. aureus strains were archived. Additionally, an S. aureus reference strain (ATCC 25923, KWIK-STIK™ Microbiologics, Argenta, Poznan, Poland) and an S. pseudintermedius ED99 type strain (lab collection) were used as an internal control of the entire study. For the purposes of the research presented in the manuscript, the strains were revived onto blood agar, mannitol salt agar (Oxoid, Argenta, Poznan, Poland), and STAPH chromagar (GRASO, Starogard Gdanski, Poland) and incubated overnight at 37 °C.After bacterial growth, isolates were identified on the basis of their ability to ferment mannitol (GRASO, Starogard Gdanski, Poland), Polymyxin B resistance, and ability to form fibrin clots in rabbit plasma (Biomed, Cracow, Poland). The Staphaurex™ Plus Latex Agglutination Test (Remel, ThermoFisher Scientific, Waltham, MA, USA) was also performed. After initial selection, a multiplex polymerase chain reaction (M-PCR) method described by Sasaki et al. [16] was performed to differentiate of coagulase-positive staphylococci (CoPS). By using this method, seven species of CoPS were preliminary differentiated based on the size of the PCR product after amplification of the conserved regions of the thermonuclease (nuc) gene. In addition, the presence of the mecA and blaZ genes were also tested according to Ruzauskas et al. [17]. In this case, the major genetic determinants of resistance to ß-lactam antibiotics were tested.2.2. Antimicrobial Susceptibility TestingThe disk diffusion method was used to determine which antimicrobial agent will inhibit the growth of the selected staphylococci according to the CLSI M100 31st ed. [18] and VET01S 5th ed. [19] recommendations. The following commercial disks (OXOID, Argenta, Poznan, Poland) were used: penicillin G (5 μg), amoxicillin (10 μg), amoxicillin with clavulanic acid (20 + 10 μg), cefalexin (30 μg), doxycycline (30 μg), oxytetracycline (30 μg), trimethoprim/sulfamethoxazole (1:19; 25 μg), neomycin (30 μg), gentamicin (10 μg), amikacin (30 μg), enrofloxacin (5 μg), marbofloxacin (5 μg), ciprofloxacin (10 μg), and polymyxin B (300 U). In order to estimate potential methicillin resistance, an oxacillin disk (1 μg; resistance with the zone of inhibition ≤17 mm recommended for S. pseudintermedius) and a cefoxitin disk (30 μg; a surrogate for oxacillin recommended for S. aureus with the zone of inhibition ≤21 mm in regard to resistance) were used.2.3. Antibiotics and Essential Oils AnalysisA freeze-dried gentamicin (OXOID, ThermoFisher Scientific, Waltham, MA, USA) was dissolved in deionized water to a final concentration of 256 mg/mL, becoming the basis for the appropriate two-fold dilutions in Mueller–Hinton broth (MHB) (GRASO, Gdansk, Poland) to achieve a final concentration ranging from 2.56 mg/mL to 0.01 μg/mL.Enrofloxacin, as a ready-to-use solution for injection (Baytril™ One, 100 mg/mL suspended in 30 mg n-Butanol; Bayer Animal Health, UK), was purchased from Medivet (Szczecin, Poland). Similarly to above, the solution was initially diluted to concentration of 2.56 mg/mL, and then, two-fold dilutions were prepared.Commercial essential oils (Organique/Avicenna, Wroclaw, Poland) from patchouli (Pogostemon cablin; PcEO) and tea tree (Melaleuca alternifolia; MaEO) were used in the study. Only to control the content of patchouli alcohol, a sample of Tisserand Aromatherapy brand of patchouli oil (First Natural Brands Ltd., Sayers Common, West Sussex, United Kingdom) was used in the HPLC-MS study as an internal control. Vials were stored at 4 °C in dark glass bottles. Dimethyl sulfoxide (DMSO) (Avantor, Gliwice, Poland) was used as an organic solvent for essential oils (twofold dilutions expressed as % v/v and mg/mL). A stock solution of the tested oils was prepared in a final concentration ranging from 10% to 0.001% v/v. The concentration was expressed in mg/mL depending on the individual density of the EO batch. In order to exclude an inhibitory effect of DMSO on growth of staphylococci, a concentration gradient of DMSO alone ranging from 0% to 50% (increase by 5%) was performed, and the survival of each staphylococci strain in this gradient was evaluated.2.4. Activity of Antibiotics and Essential Oils against Staphylococci2.4.1. Individual MICTo assess the sensitivity of each staphylococci strain to the antimicrobial agents under study, the minimum inhibitory concentration (MIC) of gentamicin, enrofloxacin, and both essential oils (PcEO and MaEO) against all S. pseudintermedius and S. aureus strains was individually determined via the serial dilution method using sterile 96-well plates (Wuxi Nest Biotechnology, Wuxi, China). Briefly, a decreasing concentration of antibiotics was successively added in the amount of 10 µL to each well containing 85 µL MHB in the rows of the 96-well microplate (causing an additional dilution of 1:10). In the case of EOs, 10 µL of decreasing-concentration, previously prepared stock solution of EOs was used in a similar arrangement as above (1:10; max DMSO content ≤ 10%). Next, the bacterial suspension (5 µL) at a concentration of 2.0 × 107 CFU/mL (DEN-1 densitometer, BioSan, Józefów, Poland) was added to each well (final concentration approx. 1.0 × 106 CFU/mL per well). The MIC was estimated after 24 h of incubation at 37 ± 1 °C. To avoid a false reading (especially at EO with artificial turbidity at the highest concentrations), a 10 µL of 0.01% resazurin (POL-AURA, Olsztyn, Poland) was added to each well. The color changed from blue to pink after an additional 3 h of incubation with resazurin at 37 ± 1 °C, indicating the presence of live bacteria in the well (which means that the antimicrobial agent was ineffective at the concentration tested). MIC was determined on the basis of the dark blue color appearance in the first well after any pink wells (corresponding to the smallest concentration of an antimicrobial agent capable of eliminating staphylococci). All experiments were performed in triplicate.2.4.2. CheckerboardsKnowing the individual effective concentrations of gentamicin, enrofloxacin and EOs, we extended our investigation to study the potential synergistic or antagonistic effect between those antimicrobial agents in the following four combinations: PcEO × gentamicin, PcEO × enrofloxacin, MaEO × gentamicin, and MaEO × enrofloxacin in the 96-well checkerboard.Briefly, a mix of seven serial twofold dilutions of EO in rows (10 µL/well; horizontal orientation) and ten serial twofold dilutions of antibiotics in columns (10 µL/well; vertical orientation) was added to 75 µL/well of MHB for different oil × antibiotic combinations/well. The last row and penultimate column always contained only a single antimicrobial agent supplemented with the pure lacking opposite diluent (DMSO or ddH2O). The last column was reserved for positive and negative controls (wells contain only MHB and both diluents). Then, 5 µL of the particular bacterial suspension at a final concentration of 2.0 × 107 CFU/mL was added to each well (final concentration approx. 1.0 × 106 CFU/mL per well with the exclusion of negative controls which control the purity of MHB and diluents). Positive control confirms the vitality of the strain under conditions of maximum DMSO concentration in MHB. Incubation and readings were similar to the individual MIC. If there was an interaction, the best well was selected. Each checkerboard was performed in triplication.2.4.3. Fractional Inhibitory ConcentrationsTo determine possible interactions between antimicrobial agents, fractional inhibitory concentrations (FICs) were calculated according to van Vuuren and Viljoen [15] as follows:FIC(OxA) = MIC(OxA)/MIC(O) FIC(AxO) = MIC(AxO)/MIC(A) where:OxA—oil in combination with antibiotics O—oil alone AxO—antibiotics in combination with oil A—antibiotics aloneThe ΣFIC was then calculated for each test sample independently as the sum of the FIC:ΣFIC = FIC(OxA) + FIC(AxO)The interpretation of possible interactions in vitro between antimicrobial agents was described as synergistic (ΣFIC ≤ 0.5), additive (0.5 < ΣFIC ≤ 1.0), noninteractive (1.0 < ΣFIC ≤ 4.0), or antagonistic (ΣFIC > 4.0).2.5. Qualitative Analysis of the Composition of Essential OilsComposition of both commercial essential oils were analyzed via high-performance liquid chromatography–mass spectrometry (HPLC-MS) technique. A reversed-phase Zorbax 2.1 × 50 mm Eclipse Plus C18 column (Agilent, Santa Clara, CA, USA) equipped with a guard column was used for the chromatographic separation. An Ultivo G6465B mass spectrometer (Agilent, USA) coupled to a chromatograph (1260 Infinity II Series Liquid Chromatograph, Agilent, USA) was used to detect and identify the constitutes according to mass-to-charge ratio (m/z) working in scanning mode (SCAN) followed by multiple reaction monitoring (MRM) mode.The patchouli oil and tea tree oil were diluted using HPLC hypergrade acetonitrile—ACN (Supelco, Sigma Aldrich, Burlington MA, USA)—to prepare a concentration of 100 mg/mL. Then, the concentration of the oil was further rediluted to obtain lower dilutions for injection into the mass spectrometer (injection volume 1 μL). Mobile phase A was ddH2O containing 0.1% HCOOH (Formic Acid 98–100%, Suprapur, Merck, Germany), whereas mobile phase B was 100% ACN, also containing 0.1% HCOOH.The source of electrospray ionization (ESI) operated in positive (M+H+ and other) and negative (M-H+ and other) modes. The triple quadrupole (QQQ) instrument operated under the following conditions: column temperature 25 °C, flow rate 0.3 mL/min, scan time (0.100 s to 0.500 s), fragmentator 5–120 V, collision energy interval (5.00–50.00 eV), and scanning range (100–750 m/z).Whole system control and data acquisition were performed using MassHunter Acquisition Software ver. C.01.00 (Agilent, Santa Clara, CA, USA). The data obtained were analyzed using Qualitative Analysis Software ver. B.08.00 (Agilent, Santa Clara, CA, USA).3. ResultsThe results of the drug resistance, tests, and PCR analysis of 21 strains under study are summarized in Table 1.All coagulase-positive staphylococcal strains met the required criteria. Staphylococcus aureus ATCC 25923 and Sa 1–Sa 7 isolates have characteristics common to S. aureus, e.g., an immediate reaction in a high-specific latex test capability of fermenting mannitol. A typical PCR band of 359 bp fragment of the nuc gene was also obtained. On the contrary, other strains (Sps 1–Sps 12 and ED99) have always yielded a PCR band of 926 bp (according to Sasaki et al. [16], this fragment of the nuc gene is specific only to S. pseudintermedius) and negative results for the abovementioned tests. They also had quite a characteristic double hemolysis. Additionally, all the isolates were further identified via matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) (AniCon Labor GmbH, Germany) due to the collection of strains for autovaccines. Third-party test reports confirmed the species affiliation of all staphylococci under study.Screening for drug resistance also revealed some patterns. Both reference strains, as well as Sa 4–Sa 7 and Sps 11, were susceptible to all antibiotics tested (except Polymyxin B, because S. aureus is naturally resistant to this antibiotic) and were mecA- and blaZ negative. Sps 9, Sps 10, and Sps 12 strains showed resistance only to single antibiotics (also mecA-negative; Sps 9 was blaZ-positive with resistance only to penicillins). In contrast, the remaining isolates (Sa1–Sa 3 and Sps 1–Sps 8) were resistant to the majority of antimicrobial agents examined, including cefoxitin/oxacillin. Moreover, only these isolates were simultaneously positive for the presence of the mecA gene (527 bp) and the blaZ gene (772 bp). Compiling this information, it can be concluded that isolates Sa1–Sa3 can be considered to represent a methicillin-resistant S. aureus (MRSA), whereas Sps 1–Sps 8 are considered methicillin-resistant S. pseudintermedius (MRSP). However, further in-depth studies are needed to confirm this hypothesis.After preliminary analyses, the first dilution of essential oils was determined at levels of 1% (patchouli; density 0.966 g/mL) and 10% (tea tree; density 0.895 g/mL). A detailed susceptibility analysis to selected antibiotics (gentamicin, enrofloxacin) and essential oils (patchouli, tea tree) using the MIC method is summarized in Table 2.In general, the MIC results for antibiotics were in agreement with the disk diffusion method. A group of staphylococci is especially notable for their high MIC values (high resistance) for both antibiotics (Sps 1 to Sps 8). The Sa3 staphylococcus strain was also characterized by significant dual resistance. One-way moderate or high resistance was also noted in the case of Sps 9 and Sps 12 (gentamicin) as well as Sa 1, Sa 2 and Sps 10 (enrofloxacin). The MIC values are also in agreement with the standards [18,19]. In the case of EOs, definitely more balanced results were obtained. For PcEO, MIC values ranging from 0.25 to 0.5% v/v (2.4–4.8 mg/mL) were recorded most often (81%; all S. pseudintermedius and antibiotic-resistant strains of S. aureus) followed by 0.125–0.25% v/v, which corresponds to a PcEO concentration of 1.2–2.4 mg/mL (19%; only for antibiotic-sensitive strains of S. aureus). The activity of MaEO was more varied regardless of drug resistance. Most often, the MIC ranged from 1.25–2.5% v/v (38% of cases; concentration 11.2–22.4 mg/mL) followed by 0.625–1.25% v/v (19%; 5.6–11.2 mg/mL), 1.25–5% v/v (19%; 11.2 ÷ 44.8 mg/mL) and 2.5–5 (14%; 22.4 ÷ 44.8 mg/mL). In summary, a tenfold stronger effect of PcEO than MaEO was noted. The DMSO content in the wells did not exceed 10%. However, our strains were still able to survive in 15–20% DMSO. This is consistent with the general knowledge of DMSO activity that the content of DMSO should not be more than 10–15%, while DMSO in amounts of 5–7.5% had no effect on MICs [20].The essential oils were tested in combination with antimicrobial drugs against resistant (11 strains) and susceptible bacteria (10 strains), in order to check for their possible synergistic or antagonistic interactions using checkerboard method. Results can be seen in Table 3. For comparison, Table S1 (in the supplementary section) presents the corresponding results when considering the concentration of EOs expressed in mg/mL instead of % v/v. Regardless of the unit chosen, identical interactions were obtained.Gentamicin appears to be highly prone to interacting with EOs. In 53.4% of cases, the PcEO had additive effects on S. aureus and S. pseudintermedius while synergism was observed for MaEO. There was also a significant percentage of dual synergy results (38.1%). Even the worst synergy (∑FIC = 0.5) indicates that four-fold reduction in MICs of antibiotic and EO were observed. In the case of stronger synergies, a remarkable decrease in the MIC values of EOs was observed (e.g., Sps 3/tea tree eight-fold of tea tree and up to sixteen-fold of gentamicin). A double additive effect was noted for only two staphylococci (Sa 7 and Sps 10, however near synergy), probably by the fact that these two staphylococci were the most sensitive to gentamicin (MIC 0.0625–0.125 μg/mL), which may have limited the margin for possible interaction. There were no cases with neutral or negative interaction (even within triplicates).Enrofloxacin in combination with patchouli or tea tree oil mostly acted independently and neutrally—no interactions were observed in 57.1% of cases. Other staphylococci reacted with little reproducibility. However, the PcEO had always an additive affect whereas MaEO acted quite randomly with enrofloxacin (synergy 9.5%, additive 14.3%, noninteractive 19.1%, respectively). Among them, two staphylococci (Sa 1 and Sps 1) are worthy of extra comment because of their origin: both are isolated from the most severe cases of canine pyoderma, and, surprisingly, the best combination of interaction was obtained for them: additive (PcEO) and synergy (MaEO).HPLC-MS AnalysisCommercial EOs are available in diluted, highly concentrated, and—rarely—in undiluted forms. Often the price of such specifics reveals its level of purity or adulteration. An attempt was therefore made to make a preliminary assessment of the composition of the oils used in the study. The volume equivalent to a concentration of 0.1% v/v of each essential oil in the MIC was examined.According to available gas chromatography–mass spectroscopy (GS-MS) analysis of patchouli oil, the presence of up to 30 volatile substances was identified but the main components are: patchouli alcohol (average 20–45% but sometimes up to 72%; molar mass 222.4 g/mol), followed by pogostol (222.4 g/mol; 0.2–6%), pogostone (224.3 g/mol; specific to P. cablin 0.1–27%), norpatchoulenol (206.3 g/mol; 0.1–4%), patchoulene (~8%), seychellene (~6%), α- and δ-guaiene (~18% of each), caryophellene (~8%) and bulnesene (3–23%) (204.4 g/mol of each) [21,22,23]. The nonvolatile chemical profile of PcEO was revealed for the first time using HPLC-Q-TOF-MS by Xie et al. [24], by whom an additional 73 nonvolatile constituents (i.e., 33 flavonoids, 21 organic acids, 9 phenylpropanoids, 4 sesquiterpenes, 3 alkaloids, and 3 other types of compounds) were identified and characterized (pachypodol was most abundant at 344.3 g/mol; other compounds have a molar masses usually greater than 300). In this manuscript, the scan range of patchouli oil was set from m/z 200 to 230 (covering the aforementioned volatile components), and the result is presented in Figure 1.To quantify patchoulol using GC-MS, m/z = 41, 55, 83, 98, 125, 138, 161, 179, 189, 207, and 222 were usually selected as the diagnostic ions (major ions are underlined; [23,25,26]. However, there are no significant data on how patchouli alcohol behaves in the mobile phase using liquid chromatography–mass spectroscopy (LC-MS) studies. Of some surprise in the presented study was the fact that there were no clear peaks, with masses ranging from 221–223 m/z, which should correspond to patchouli alcohol [222 ± H] (marked with an asterisk in Figure 1). For this purpose, the Avicenna EO was compared to a sample of some other essential oil from a highly acclaimed brand—Tisserand EO, an expert in sourcing and blending 100% natural pure essential oils since 1974. As can be seen in Figure 1, both chromatograms and mass spectra are almost identical. In both cases, the most abundant is the peak at m/z 219.2 (100% of both abundance, 1.35 × 106 and 1.05 × 106, respectively), which may be a sought-after oxygenated sesquiterpene: patchouli alcohol. Some of the PcEO sesquiterpene hydrocarbons (SQHCs), e.g., α-patchoulene and β-patchoulene are suspected to be artifacts formed through the dehydration of patchoulol and subsequent Wagner–Meerwein rearrangements during steam distillation [27]. When patchoulol is dehydrated ([M+H-H2O]+ resulted in mass m/z 205), depending on the conditions, various mixtures of patchoulenes and other rearranged hydrocarbons may be obtained (e.g., resulted in m/z 219.2). The neutral loss of -CO, H2O, -OCH3, or -CH3 was commonly observed in MS spectra. The nature of the patchoulol changes that are occurring during LC-MS remains to be explained. The second-most frequently recorded peak was m/z 205.2 (chromatogram abundance 1.05 × 106 with Abund % at 81.1 and 0.95 × 106 with Abund % at 91.25, respectively), which corresponds to an extensive and diverse group of sesquiterpenes of equal mass 204.36 g/mol ([M+H]+), including patchoulenes, guaienes, seychellenes, and bulnesenes, although a certain percentage here may be dehydrated patchoulol. Both essential oils noticeably vary in the third peak: m/z 225.2—it most likely refers to pogostone ([M+H]+). The Tisserand EO seems to be richer in this component compared to the Avicenna EO, but this observation can only be confirmed by a quantitative study. Pachypodol (m/z 343.3, [M-H]+) and several other components were also detected at a low level, which may indicate the natural origin of the essential oils.In the case of tea tree essential oil, the chemical composition of MaEO may be extremely variable, e.g., depending on chemotype (this means that several groups exist within a population of one plant species with the same morphological features differing in compositions of their products), and over 220 chemicals have been identified [28]. Essential oil of Melaleuca terpinen-4-ol type is predominant, whereas in the composition, it should have terpinen-4-ol (35–48%; 154.25 g/mol), γ-terpinene (14–28%; 136.23 g/mol), α-terpinene (6–12%; 136.23 g/mol), 1,8-Cineole/eucalyptol (0.01–10%; 154.25 g/mol), α-pinene (1–4%; 136.23 g/mol), p-cymene (0.5–8%; 134.22 g/mol), terpinolene (1.5–5%; 136.23 g/mol), α-terpineol (2–5%; 154.25 g/mol), sabinene (0.01–3.5%; 136.23 g/mol) (according to ISO 4730:2017-02 [29]). Mass spectra (range from 133 to 160, extremely specific) for the tea tree essential oil used for research (MIC and checkerboards) are summarized in Figure 2.A high abundance of compounds with a mass-to-charge of 137 m/z was observed, which corresponds to a group of several monoterpenes with a mass of 136.23 g/mol ([M+H]+; abundance greater than 1.9 × 105). A two-fold lower abundance (approx. 1.0 × 105) was observed for compounds with a mass of 153 m/z, which probably corresponds to the neutral loss of hydrogen in terpinen-4-ol, eucalyptol, or—less possibly—α-terpineol (154.25 g/mol; [M-H]+). In contrast, the third peak (135 m/z; abundance approx. 1.0 × 105) can be either p-cymene ([M+H]+ from 134.22 g/mol) in lesser amounts or cases of hydrogen loss in a monoterpenes group. The presence of other compounds that are not natively present in MaEO (e.g., sesquiterpenes) and adulteration with fragrance compositions of synthetic origin (linalool, citronellol, etc.) were not found. In conclusion, the presence of the main MaEO-specific compounds (terpinen-4-ol and monoterpenes) was confirmed.4. DiscussionLong-term antibiotic treatments may increase the risk of selecting for multidrug-resistant bacteria, one of the most relevant current threats to public health. Antimicrobial-resistant organisms can be transmitted to humans and other animals in the European Union and other countries through the consumption of products of animal origin, by direct contact with animals or humans, or by other means (Regulation EU No 2019/6) [30]. Alternative therapies, including essential oils (EOs), have become very popular as natural remedies in veterinary medicine. The objective of this study was the establishment of novel approaches to conventional therapies using selected EOs for the treatment of canine skin disorders. The efficacy of EOs in inhibiting a variety of classical and opportunistic pathogens depends on the plant part (e.g., leaf, flower, or bark), origin (e.g., country), seasonal variations, the method of extraction of the essential oil, the procedure used in the antimicrobial assays (e.g., different broth), and the target microbial isolate [31]. Different staphylococcal species (incl. S. pseudintermedius) isolated from canine dermatitis were examined in study by Ebani et al. [32]. Among them, oregano (Origanum vulgare L.) and thyme (Thymus vulgaris L.) EOs resulted highly active against all staphylococcal strains tested. The research conducted by Nocera et al. [7] aimed to test in vitro the antimicrobial activity of 11 EOs (e.g., cinnamon or eucalyptus) against four methicillin-resistant Staphylococcus pseudintermedius (MRSP) and four methicillin-susceptible S. pseudintermedius (MSSP) pyoderma-associated clinical isolates. The obtained findings demonstrated a clear in vitro efficacy of some tested EOs against both MRSP and MSSP strains isolated from dogs. Unfortunately, neither study included both PcEO and MaEO.Patchouli essential oil (PcEO) is obtained by steam distillation or hydrodistillation of the dried leaves of Pogostemon cablin (Blanco) Benth. (Lamiaceae). It has a unique woody odor—utilized in high-end fragrances and cosmetics [27]. This plant originated in Southeast Asia, Madagascar, India, Brazil, Japan, and China, but 90% of patchouli oil around the world is supplied from Indonesia [21]. Given its multicomponent nature, PcEO is also a part of a traditional Chinese medicine that has been used for the treatment of many ailments for centuries, e.g., to treat colds, nausea, fever, headache, and diarrhea [24,33]. Biofilms formed by bacteria are associated with highly enhanced resistance against antimicrobial agents, resulting in therapy failure. However, the PcEO may significantly inhibited the initial adherence phase of S. aureus biofilm development [34].Using the disk diffusion method, Karimi [35] revealed that freshly hydrodistilated Philippine patchouli oil was found to be active only against the Gram-positive bacteria (Staphylococcus aureus ATCC 25923 and other Staphylococcus sp., Bacillus sp., and Streptococcus species). Moreover, both hospital and community clinical human isolates of methicillin-sensitive (MSSA) and methicillin-resistant (MRSA) S. aureus were sensitive to an MIC range of 0.03–0.06% v/v. High antistaphylococcal potential of PcEO has been confirmed for the group of 31 strains isolated from cases of bovine mastitis (MIC ranging from 0.01% v/v to 0.312% v/v) and the reference strain S. aureus PCM 2051 (0.625% v/v) using commercial oil (Pollena Aroma, Poland) [36]. The results of MICs performed by Yang et al. [37] showed that patchouli oil and its main components (patchouli alcohol and pogostone) have good antibacterial activities against Staphylococcus aureus ATCC2925 (MIC at the level of 4.5 mg/mL, 2 mg/mL and 1 mg/mL, respectively). Other studies showed that pure PcEO at > 40 μL/mL concentration reduced the growth of Staphylococcus aureus ATCC 6538 reference strain [38]. The high efficacy of PcEO against staphylococci (even multidrug-resistant ones) is also confirmed by the results in the presented manuscript, where the MIC ranged from 0.125 to 0.5% v/v that correspond to average PcEO concentrations of 1.2–4.8 mg/mL. To our knowledge, we also present the first study of the in vitro activity of PcEO against S. pseudintermedius.Tea tree essential oil (MaEO) is the volatile oil obtained by distillation from the leaves and terminal branchlets of Melaleuca alternifolia (Maiden et Betche) Cheel [29]. As mentioned previously, the chemical composition of MaEO may be extremely variable depending on multiple parameters, such as biomass used (from wild or cultivated trees; only leaves or leaves plus terminal branchlets); chemotype (according to ISO 4730:2017-02); and mode of production (steam distillation versus hydrodistillation) [28].The activity of tea tree oil against S. aureus is definitely better documented in the scientific literature. May et al. [39] reported MaEO MICs and minimum bactericidal concentrations (MBCs) of 0.12–0.5% for S. aureus (including MRSA) as well as time-kill studies in which essential oil with increased concentration of terpinen-4-ol displayed enhanced antimicrobial activity (4 h instead of 6 h for standard tea tree oil). The mechanisms of action of MaEO and three of its components—1,8-cineole, terpinen-4-ol, and α-terpineol—against Staphylococcus aureus ATCC 9144 were also investigated by Carson et al. [40]. At inocula of 5.0 × 105 and 5.0 × 107 CFU/mL, the MICs and MBCs were both 0.25% and 0.5% v/v, respectively. At an inoculum of 5.0 × 109 CFU/mL, the MIC was 0.5% v/v and the MBC was 1% v/v. An identical result was obtained by Nelson [41], while similar results (0.12–0.5% v/v) against various methicillin resistant strains of seven species of Staphylococcus, including S. aureus were reported by Harkenthal et al. [42]. A slightly higher value for MSSA and MRSA was reported by Oliva et al. [43]—0.5–2% v/v. In our study, at inocula of approx. 1.0 × 106 CFU/mL per well, the MIC ranged from 0.625 to 1.25% or 5.6–11.2 mg/mL (susceptible strains of S. aureus) and 2.5 to 10% v/v (multidrug-resistant strains), which refers to the MaEO concentration range of 22.4–89.6 mg/mL. A similarly high MIC (5 ÷ 10%) was reported by De Martini et al. [44] for 17 coagulase-positive Staphylococci (CoPS) isolated from canine otitis externa cases. In contrast to all the above results, extremely low MIC values were also reported. Mann and Markham [45] and Kumari et al. [46] reported an MIC value of 0.02–0.04% v/v of tea tree oil against S. aureus.In a study performed by Meroni et al. [47], a total of 23 S. pseudintermedius strains were collected from clinical samples (pyoderma) from different dogs. The majority of them (61%) were resistant to more than three pharmacological categories and were classified as multidrug-resistant. These authors reported slightly higher or similar MICs (7.6 ± 3.2% v/v) to those presented in this manuscript (0.625 ÷ 5% v/v). In the study of Valentine et al. [48], a total of 25 MRSP and 25 MSSP isolates from dogs with skin and soft tissue infections were included. Tea tree oil has been shown to inhibit the growth of both types of S. pseudintermedius strains, with MICs ranging from 0.12 to 0.96% v/v and from <0.03 to 0.96% v/v, respectively. In an experiment by Han et al. [49], the antimicrobial effects of a topical skin cream (Korean Dara cream®) consisting of four natural oils (emu oil, jojoba oil, avocado oil, and tea tree oil) were evaluated through measurements of MIC (0.23% v/v) against three S. pseudintermedius isolates obtained from the nostrils of healthy dogs.The antibacterial activity of pure terpinen-4-ol on S. aureus reference strains (ATCC 25923, ATCC 13150, NCTC 6571 and NCTC 29213) and clinical isolates was assessed by determining the MIC (0.25% v/v in most cases) and MBC (mostly 0.5 % v/v) in few studies in the scientific literature [50,51,52]. In the presented manuscript, the MIC values were 5–10-fold higher when a commercial MaEO was used instead of terpinen-4-ol. Probably the low content of terpinen-4-ol in Avicenna essential oil caused this result. According to Avicenna Oil’s official certificate of laboratory analysis (batch no. 27947), the terpinen-4-ol content was at the level of 40.4% (2.5 times lower than the pure reagent). The strong antibiofilm activity of terpinen-4-ol against S. aureus was found in the study by Cordeiro et al. [52] in a concentration-dependent manner even at sub-MIC concentrations. Moreover, in silico molecular docking analysis showed a possible interaction between terpinen-4-ol and penicillin-binding protein 2a (PBP2a), which is one of the main molecules involved in staphylococci resistance to beta-lactam drugs. Apart from beta-lactamases being used to inactivate the antibiotic, MRSA and MRSP strain resistance is mediated through the acquisition of a gene cassette containing the mecA gene, which encodes the low-affinity altered transpeptidase PBP2a [53]. Thus, the effective binding of terpinen-4-ol to the PBP2a protein and the consequent inhibition of its activity can be an effective adjuvant tool in the treatment of resistant strains [52]. These observations are in agreement with our study, as the presence of this gene was confirmed in many of the multidrug-resistant strains studied in this manuscript (Sa1–Sa3 and Sps1–Sps8). In addition, resistance to penicillins in staphylococci is also mediated by β-lactamases encoded by the blaZ gene [17]. Unfortunately, only interactions with the main representatives of aminoglycosides and fluoroquinolones (without identifying their individual resistance genes) but not beta-lactam drugs have been studied, so this may be a goal for future studies.Essential oils, due to the small scale on which they are obtained, sometimes have a high price, which encourages dishonest manufacturers and distributors to adulterate them. The main methods of adulterating EOs are to dilute them with vegetable fat, mix them with cheaper EOs, or to add synthetic components to mimic the olfactory properties or the composition of the chemotype [54,55]. While in the first case, the natural fragrance bouquet does not change (it is only less intense), when other compounds are introduced (e.g., various terpenes), the fragrance impression is significantly modified. Such modifications can also affect the antimicrobial activity of EOs [56]. In our study, both commercial Avicenna-brand EOs tested via HPLC-MS were characterized by an appropriate composition that does not vary from the literature data. There was also no adulteration with additional fragrance compounds, such as linalool, citronellol, or limonene, which could alter their properties. Furthermore, the Avicenna patchouli oil had a similar composition to an essential oil from an established brand that was more expensive.5. ConclusionsBoth commercial EOs were characterized by natural composition without artificial adulteration. Patchouli and tea tree oils can be good alternatives for treating severe cases of pyoderma in dogs, especially when dealing with multidrug-resistant strains. Gentamicin, in comparison to enrofloxacin, appears to be highly prone to interacting with EOs. It is noticeable that patchouli oil had several times stronger of an effect on staphylococci compared to tea tree oil. However, tea tree oil is characterized by a stronger synergistic effect, having great potential as long as the natural products contain predominantly terpinen-4-ol. In the future, it is advisable to conduct tests with several oils from different manufacturers (and, e.g., variable chemotypes) as well as their interaction with beta-lactams to confirm the observations obtained.

animals : an open access journal from mdpi

10.3390/ani11061675

PMC8229322

Wilms’ tumor 1-associating protein (WTAP) is a key subunit of the N6-methyl-adenosine (m6A) methyltransferase complex during porcine early embryo development. However, the role of WTAP in embryonic development is still unclear. In this study, we demonstrate that WTAP plays an indispensable role in embryonic development, and the loss of WTAP will promote the apoptosis of embryonic cells, and reduce the rate and quality of embryonic development.

m6A is one of the most common and abundant modifications of RNA molecules present in eukaryotes. The methyltransferase complex, consisting of methyltransferase-like 3 (METTL3), METTL14, and WTAP, is responsible for the m6A modification of RNA. WTAP was identified as an mRNA splicing regulator. Its role as a regulatory subunit of the m6A methyltransferase complex in embryonic development remains largely unknown. To investigate the role of WTAP in porcine early embryonic development, si-WTAP was microinjected into porcine parthenogenetic zygotes. WTAP knockdown significantly reduced the blastocyst rate and global m6A levels, but did not affect the cleavage rate. Betaine was supplemented into the in vitro culture (IVC) to increase the m6A levels. Betaine significantly increased the global m6A levels but did not affect the blastocyst rate. Furthermore, the pluripotency genes, including OCT4, SOX2, and NANOG, were downregulated following WTAP knockdown. The apoptotic genes BAX and CASPASE 3 were upregulated, while the anti-apoptotic gene BCL2 was downregulated in WTAP knockdown blastocysts. TUNEL staining revealed that the number of apoptotic cells was significantly increased following WTAP knockdown. Our study indicated that WTAP has an indispensable role in porcine early embryonic development.

1. IntroductionMethylation of the adenosine base at the nitrogen-6 position (m6A) is one of the most common and abundant post-transcriptional epigenetic modifications of RNA in eukaryotes [1,2]. Previous studies have shown that m6A RNA modification is regulated by adenosine methyltransferases and demethylases [3]. The m6A methyltransferases (or the “writers”), including METTL3 and METTL14, methylate the N6 position of adenosine [4]. The m6A demethylases (or the “erasers”), including FTO and ALKBH5, reverse the RNA methylation process [5,6,7]. Furthermore, the m6A binding proteins (or the “readers”), such as YTHDF2 and YTHDC1, recognize the m6A sites of target mRNAs and regulate the fate of the mRNA [8,9].WTAP was originally identified as a splicing regulator that binds to human Wilms’ tumor 1 protein [10]. It plays an important role in cell cycle progression and mammalian embryo development [11]. The involvement of WTAP in RNA methylation was first observed in studies in Arabidopsis thaliana and yeast [12,13]. In a recent study, WTAP was reported to be the third regulatory subunit of the m6A methyltransferase complex [14]. Although WTAP has no inherent methylation activity, it interacts with the METTL3–METTL14 heterodimer and synergistically forms the m6A methyltransferase complex to promote m6A methylation [15].A growing body of evidence indicates that global mRNA m6A levels are associated with embryonic development [16]. Previous studies have shown that a deficiency in methyltransferases led to reduced global mRNA m6A levels and negatively affected embryo development in mice [17]. Knockdown of WTAP in zebrafish embryos caused defects in tissue differentiation and increased apoptosis [14]. However, the biological role of WTAP in porcine early embryo development is unknown. In the present study, we investigated the effect of WTAP on global mRNA m6A levels and subsequent embryonic development competence by knocking down WTAP in porcine parthenogenetic embryos. Our study demonstrates that WTAP plays an indispensable role in porcine parthenogenetic early embryo development.2. Materials and MethodsAll chemicals and reagents in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA), unless noted otherwise.2.1. Oocyte Collection and In Vitro MaturationPorcine ovaries from slaughtered pre-pubertal gilts were obtained from a local slaughterhouse and transported to the laboratory in 0.9% saline at 35 °C within 2 h. The cumulus–oocyte complexes (COCs) were isolated from 3–8 mm antral follicles aspirated using an 18-gauge needle. COCs, with multiple layers of compact cumulus cells, were selected, and washed three times in hydroxyethyl piperazine ethane sulfonic acid (HEPES) medium with 0.1% polyvinyl alcohol (PVA, w/v) and 0.05 g/L gentamycin. The COCs were cultured in 200 mL of in vitro maturation (IVM) medium, covered with mineral oil and incubated for 42 h at 38.5 °C in an atmosphere containing 5% CO2 at 100% humidity.2.2. Parthenogenetic Activation (PA) of Oocyte and In Vitro CultureTo obtain the porcine haploid embryos, parthenogenetic activation (PA) was used. PA was performed as described in previous reports [18]. Briefly, the metaphase-II (MII) oocytes were activated by two direct-current (DC) pulses of 120 V/mm for 60 µs in the activation medium. The activated oocytes were transferred to PZM-5 medium and cultured in an atmosphere containing 5% CO2 at 100% humidity. The development of the oocytes into blastocysts was examined after 6 days.2.3. Microinjection of siRNA into OocytesBefore microinjection, the oocytes were cultured to MII and subjected to parthenogenetic activation. The siRNA specific for porcine WTAP (si-WTAP) was microinjected into the cytoplasm of the zygote using a Nikon TE2000-U inverted microscope (Nikon, Tokyo, Japan) and an Eppendorf Cell Tram Vario system (Eppendorf, Hamburg, Germany). The siRNA and the negative controls were microinjected into the zygotes in the same way to serve as the negative control, while the non-injected zygotes served as the normal controls. Approximately 10 pL of siRNAs were microinjected into the zygotes at a 20 µM concentration, and the number of zygotes used was indicated in the figure. Following injection, the zygotes were transferred to PZM-5 medium until they developed into blastocysts. The siRNA specifically targeting WTAP or its non-target negative control siRNA was synthesized by Genepharma (Shanghai, China). siRNA sequence: 5′-GCAAGAGUGUACUACUCAATT-3′; negative control siRNA sequence: 5′-UUGUACUACACAAAAGVUACUG-3′.2.4. Betaine TreatmentAfter microinjection, porcine zygotes were cultured in vitro in IVC medium supplemented with betaine (B2629, Sigma, St. Louis, MO, USA) (5 mM, 10 mM, 20 mM). The concentrations of chemical reagents were chosen first based on a previously published report [19], and then, preliminary experiments were performed to determine the optimal concentrations, which were then used in subsequent experiments.2.5. Gene Expression AnalysisTotal RNA was extracted from each group of blastocysts (n = 20) using the AllPrep DNA/RNA Micro Kit (QIAGEN, Dusseldorf, Germany) following the manufacturer’s instructions. cDNA was synthesized using the First-Strand cDNA Synthesis kit (Promega, Fitchburg, WI, USA). Quantitative real-time PCR (qPCR) was performed using the BioEasy SYBR Green I Real-Time PCR Kit (Bioer Technology, Hangzhou, China) on a BIO-RAD iQ5 Multicolor Real-Time PCR Detection System (170-9780, BIO-RAD Laboratories, Hercules, CA, USA). PCR was performed using the following parameters: initial denaturation at 95 °C for 3 min, followed by 40 cycles of denaturation at 95 °C for 10 s, annealing at 60 °C for 15 s, and extension at 72 °C for 30 s. The 2−ΔΔCT method was used to determine the relative gene expression, which was then normalized to the amount of the endogenous control, GAPDH. To test the stability of GAPDH, Bio-Rad iQ5 Software was used for the analysis of amplification curves, melting curves and cycle threshold values (CT values). The experiments were performed at least in triplicate. The primer sequences used in this study are listed in Table 1.2.6. Immunofluorescence StainingBriefly, the blastocysts were washed three times in PBS–PVA. Then, the thinning of zona pellucida was performed using Tyrode’s Solution (Jisskang, Qingdao, China). The embryos were fixed with 4% paraformaldehyde for 30 min at 25 °C. Following fixation, the blastocysts were washed with PBS–PVA and permeabilized in PBS containing 0.2% Triton X-100 for 30 min. The blastocysts were then incubated in PBS containing 1% bovine serum albumin (BSA) for 1 h. Next, the blastocysts were probed with m6A antibodies (1:500, Abcam, Cambridge, UK) and incubated at 4 °C overnight. The blastocysts were washed with PBS three times for 10 min each followed by incubation with Alexa Fluor 488-conjugated secondary antibodies (1:1000, anti-rabbit) for 1 h at RT. The DNA was stained with 10 ng/mL Hoechst 33342 (Thermo Scientific, Waltham, MA, USA) for 15–20 min. The blastocysts were washed thrice with PBS–PVA for 10 min each, air dried, and mounted on a coverslip and a glass slide using an antifade mounting medium (BOSTER, Wuhan, China). To evaluate the average fluorescence intensity in the embryos, image analysis software (ImageJ) was used.2.7. TUNEL AssayThe TUNEL assay was performed as described in previous reports [20]. Briefly, the blastocysts were fixed with 4% paraformaldehyde for 1 h at 25 °C. After fixation, the blastocysts were permeabilized by treatment with 0.1% Triton X-100 for 1 h at 37 °C. The blastocysts were washed three times in PBS–PVA and incubated in the dark for 1 h at 37 °C with TdT and fluorescein-conjugated dUTPs (In Situ Cell Death Detection kit; Roche, Mannheim, Germany). The blastocysts were then stained with 10 μg/mL Hoechst 33342 for 15 min. The blastocysts were washed thrice with PBS–PVA for 10 min each, air dried, and mounted on a coverslip and a glass slide using an antifade mounting medium (BOSTER, Wuhan, China). The number of cells in the blastocysts was analyzed by using ImageJ.2.8. Statistical AnalysisThe data were analyzed using Student’s t-tests with the SPSS 16.0 software (SPSS Inc., Chicago, IL, USA). A p-value of <0.05 was considered statistically significant. The number of embryos used for the statistics in each group of the experiment is equal to n in the figure.3. Results3.1. WTAP Knockdown Impairs Embryo DevelopmentTo investigate the role of WTAP in embryo development, si-WTAP and negative control siRNA were microinjected into zygotes. The expression of WTAP was analyzed by qPCR. The expression of WTAP was significantly (p < 0.005) decreased in si-WTAP-injected embryos compared to that in the negative control siRNA-injected (NC) or non-injected embryos (Con) (Figure 1a). The change in the WTAP level did not affect the cleavage rate (Con, 94.00 ± 2.89%, NC, 91.30 ± 3.47%, si-WTAP, 92.06 ± 3.98%) (Figure 1b), although it significantly (p < 0.005) reduced the blastocyst rate (Con, 49.28 ± 2.38%, NC, 48.28 ± 2.01%, si-WTAP, 32.38 ± 2.76%) (Figure 1c). The m6A expression level was analyzed using immunofluorescence (IF) staining. The results showed that m6A expression was significantly decreased (p < 0.005) in the si-WTAP group compared to the NC group and Con group (Figure 1d,e). These results indicate that WTAP knockdown reduced the global mRNA m6A levels and negatively affected embryo development.3.2. No Effect of Betaine on WTAP-Knockdown Embryo DevelopmentWTAP knockdown reduced the global mRNA m6A levels and the blastocyst rate in porcine parthenogenetic embryos. The IVC medium was supplemented with betaine to investigate its effects on WTAP-knockdown embryo development. However, there was no change in the blastocyst rate following treatment with 5 mM, 10 mM, or 20 mM of betaine (Figure 2a). We used 20 mM of betaine for subsequent studies. The qPCR results showed that the expression of WTAP was not altered in embryos treated with betaine (Figure 2b). However, the results of the IF staining showed that m6A expression was significantly (p < 0.005) increased following treatment with betaine (Figure 2c,d). These results indicate that betaine exposure elevated the global mRNA m6A levels, but did not affect WTAP-knockdown embryo development.3.3. WTAP Knockdown Promoted Embryonic ApoptosisThe mRNA expression of pluripotency- and apoptosis-related genes was analyzed in the blastocysts. The qPCR results showed that compared to the NC and Con groups, the expression of the pluripotent genes SOX2, OCT4, and NANOG were significantly (p < 0.005) downregulated in the si-WTAP group (Figure 3a). In addition, our study showed that the expression of the apoptotic genes CASPASE 3 and BAX were significantly (p < 0.005) upregulated, in contrast to the expression of the anti-apoptotic gene BCL2 (Figure 3b). To investigate the effect of WTAP knockdown on embryonic cell apoptosis, the blastocysts were analyzed by TUNEL staining. TUNEL staining showed that the number of apoptotic cells was increased following WTAP knockdown (Figure 3c). Moreover, WTAP knockdown significantly (p < 0.005) decreased the total number of cells in the blastocysts compared to the NC and Con groups (Figure 3d). Our results showed that the loss of WTAP promoted apoptosis in the embryos.4. DiscussionThe methylation of m6A has been shown to be a reversible process, attributable to modifications by two types of enzymes: methyltransferases and demethylases [21]. However, the identity of the enzymes responsible for each modification and the biological consequences of these modified RNAs are largely unknown [22]. A recent study showed that knockout of METTL3 reduces m6A in mRNAs in mice and the embryos remain in a naive state, leading to early embryonic lethality [17]. Previous studies showed that knockdown of WTAP in zebrafish embryos led to multiple developmental defects, including a smaller head and eyes, a smaller brain ventricle, and a curved notochord. Moreover, knockdown of WTAP led to a striking increase in apoptosis in zebrafish embryos [14]. In the present study, we knocked down the expression of WTAP in pig parthenogenetic embryos by microinjection of si-WTAP, which led to a reduction in the blastocyst rate. TUNEL apoptotic staining showed a significantly increased number of apoptotic cells following WTAP knockdown, which is in agreement with the study carried out on zebrafish.WTAP is the third regulatory subunit in the m6A methyltransferase complex. Previous studies showed that METTL3 and METTL14 can interact to form heterodimers to affect m6A methylation [14]. WTAP interacts with the METTL3–METTL14 heterodimer and synergistically forms the m6A methyltransferase complex to promote m6A methylation. A recent study showed that reduced nucleic acid methylation could impair the maturation and development of pig oocytes [19]. Our results showed that the global mRNA m6A levels and blastocyst rate were reduced when we inhibited the expression of WTAP. This may suggest that RNA methylation plays an important role in both oocyte and embryonic development. Moreover, WTAP regulates transcription and alternative splicing. For example, female-lethal (2)d, a homologue of WTAP in drosophila, regulates the sex determination factor Sel by influencing the alternative splicing of pre-mRNA, and female embryos are lethal when fl(2)d is lost [23,24,25]. Previous studies have shown that betaine is usually used as a methyl donor to increase the global m6A level [19]. Treatment of the porcine parthenogenetic embryos with a methyl donor during IVC significantly boosted the m6A level within the embryos; however, the blastocyst rate and embryonic apoptosis remained unchanged. This may be because WTAP is merely a regulatory subunit without any methylation activity, and the methyl donor did not reverse the embryo damage caused by the deficiency of WTAP (Figure 4).Previous studies showed that WTAP plays an important role in early embryo development and cell cycle regulation [11,26]. Moreover, WTAP may be associated with apoptosis. A previous study showed that WTAP activated apoptosis in smooth muscle cells by regulating the splicing of the apoptosis regulator [27]. Studies have revealed that WTAP-deficient mouse embryos failed to differentiate into the endoderm and mesoderm, and exhibited early lethality [28]. In our study, we found that the pluripotent genes SOX2, OCT4, and NANOG were downregulated in WTAP-inhibited blastocysts. The apoptosis genes CASPASE 3 and BAX were upregulated in WTAP-inhibited blastocysts, while the anti-apoptotic gene BCL2 showed the opposite expression pattern. We speculate that WTAP may affect the embryo development and quality of blastocysts by regulating the expression of pluripotency- and apoptosis-related genes.5. ConclusionsOur study demonstrated that WTAP plays an indispensable role in regulating RNA methylation during porcine parthenogenetic embryo development. Knockdown of WTAP promoted embryonic apoptosis and negatively affected embryo development. Treatment with betaine during IVC significantly increased m6A levels in blastocysts, but it could not improve embryo development when WTAP was lost.

animals : an open access journal from mdpi

10.3390/ani11061528

PMC8225056

Sow mastitis is a serious breast disease that can cause severe inflammation, agalaxia and even lead to death of piglets. Porcine mammary epithelial cells (pMECs) are the main cell types that affect sow milk secretion, therefore, when swine mastitis occurs, the inflammatory response of pMECs directly affects the mammary gland health and sow’s lactation ability. Promoting the health of mammary gland epithelial cells is an important method for treating mastitis. Thus, in the current study, we investigated the effects of artemisinin on the inflammatory response of pMECs induced by lipopolysaccharide (LPS), and proposed a potential anti-inflammatory mechanism. We confirmed that artemisinin can reduce the inflammatory damage of pMECs induced by LPS by inhibiting MAPK and NF-κB signaling pathways. Pretreatment of pMECs with artemisinin showed enhanced anti-inflammatory activity against LPS-induced inflammation. Artemisinin could be a useful, safe and natural anti-inflammatory feed additive to prevent sow mastitis.

Artemisinin performs a variety of biological functions, such as anti-cancer, anti-inflammatory, anti-viral, and anti-oxidant effects. However, the effects of artemisinin on sow mastitis have not been studied. The results of the current study showed that mRNA expression abundance and content of the inflammatory factors interleukin-1β (IL-1β), tumor necrosis factor α (TNF-α), and interleukin-6 (IL-6) were significantly increased when using 50 μg/mL LPS to stimulate pMECs for 24 h (p < 0.05). Pretreatment with 20 μM artemisinin weakened LPS-induced inflammatory damage in pMECs and decreased mRNA expression abundance and the content of inflammatory factors (IL-1β, IL-6, and TNF-α) in pMECs (p < 0.05). Mechanistically, artemisinin inhibited LPS-induced activation of the mitogen-activated protein kinase (MAPK) and nuclear factor-κB (NF-κB) signaling pathways. In summary, the pretreatment of pMECs with artemisinin showed enhanced anti-inflammatory activity against LPS-induced inflammation.

1. IntroductionMammalian mastitis is one of the common diseases worldwide, which leads to increased veterinary expenses, and significant economic losses to the breeding industry every year [1,2]. The clinical signs of mammalian mastitis are agalaxia (decreases in milk production and quality) and dolor, which are accompanied with increased inflammatory mediators in milk. During mammalian mastitis, piglets can develop hypoglycemia and hypothermia, which might finally lead to death [3,4].Breast tissue contains large numbers of mammary epithelial cells for milk synthesis [5,6], which can also react with pathogens. This response is important for determining the outcome of mammary gland infection [7]. A large number of studies have shown that the inflammatory response of mammary epithelial cells induced by lipopolysaccharide (LPS) can not only reduce milk production and the levels of fat and protein in milk, but also destroy the milk–blood barrier [8,9,10]. Therefore, it is important to control and reduce the inflammation of mammary epithelial cells during mastitis. Although the use of antibiotics is still an effective way to treat animal mastitis, the use of antibiotics is severely restricted due to the emergence of more and more serious bacterial resistance and food safety [11]. It is important to seek novel alternatives to antibiotics for the effective and safe treatments of mastitis in veterinary research. Chinese herbal medicine with anti-inflammatory activity is a potentially effective option for the treatment of mastitis [12].Artemisinin (chemical formul, C15H22O5; molecular mass, 282.34 g/moL) is a sesquiterpene lactone compound with a peroxy bridge structure [13]. This specific structure is an important reason for artemisinin and its derivatives being believed to have antimalarial and antibacterial activities [14]. This chemical is isolated from the Chinese plant Artemisia annua L. and is effective for the treatment of severe and multidrug-resistant malaria [15,16,17]. Previous studies have shown that artemisinin has a variety of biological functions, such as anti-bacterial [18], anti-inflammatory [19], anti-viral [20], antioxidant [21,22,23], and immunomodulatory activities [24,25]. Although many studies have described the anti-inflammatory functions of artemisinin, the effects of artemisinin on LPS-induced inflammatory damage in porcine mammary epithelial cells (pMECs) remain to be clarified. We hypothesized that artemisinin might have a protective effect on the LPS-induced inflammatory response of pMECs.2. Materials and Methods2.1. PMECs Isolation, Cell Culture, and TreatmentspMECs were isolated from the mammary gland of a lactating sow as previously described [26] and keep refrigerated in liquid nitrogen. pMECs were cultured in Dulbecco’s modified Eagle’s medium/F12 nutrient mixture (DMEM/F12) supplemented with 5 μg/mL hydrocortisone, 10% fetal bovine serum (FBS), 10 ng/mL insulin-like growth factors-1 (IGF-1), 5 μg/mL insulin–transferrin–selenium, 10 ng/mL epidermal growth factor, and 100 U/mL antibiotics (streptomycin and penicillin) at 37 °C in a humidified atmosphere with 5% CO2. Medium was changed every 48 h, and cells were subcultured by trypsin digestions at a subcultivation ratio of 1:2.LPS (lipopolysaccharide, Escherichia coli serotype O55:B5, Sigma, Waltham, MA, USA) was diluted in DMEM/F12. Artemisinin (purity > 99%, The National Institutes for Food and Drug Control, Beijing, China) was dissolved in DMSO (Sigma, Waltham, MA, USA) and diluted in DMEM/F12.2.2. Cell Viability Assay and Flow Cytometric AnalysisTo test the cell viability, equal amounts of pMECs were seeded in 96-well plates at a density of 2 × 104 cells/mL and cultured in DMEM/F12 containing 10% FBS. After cultured for 48 h, cells are washed three times with PBS, and then treated with different concentrations of artemisinin or LPS for different times. The cell counting kit - 8 (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) was used to analyze cell viability according to the instruction. An automatic microplate reader (Bio-Rad Laboratories Inc., Foster City, CA) was used to measure the absorbance of all wells at a wavelength of 450 nm. Cell viability is the quantification of No. live cells and is expressed as a percentage of the control.To test cell apoptosis, pMECs were pretreated with 0 or 20 μM of artemisinin for 12 h before being stimulated with 50 μg/mL LPS for 24 h. After treatment, all cells were trypsinized, washed 2~3 times with PBS, stained with Annexin V-FITC/PI (Invitrogen Inc., Carlsbad, CA, USA), and analyzed by flow cytometry, according to the instructions.2.3. Real-Time PCRAccording to the experimental requirements, pMECs were treated with LPS or artemisinin, and then the mRNA expression of inflammatory cytokines and apoptosis-related genes were measured by real-time PCR as described previously [27]. Real-time PCR was performed using the ABI StepOnePlus™ real-time PCR system (Applied Biosystems, Grand Island, NE, USA). The amplification of real-time PCR was performed with a reaction volume of 20.0 μL. Samples were analyzed in duplicate.Real-time PCR reaction protocol: heating at 94 °C for 5 minutes for initial enzyme activation, then performing 40 cycles at 94 °C for 30 s, and then annealing at 60 °C for 30 s, last extension at 72 °C for 20 s. In this study, β-actin was stability expressed among different treatments, which was selected as a reference gene by geNorm 3.5 (http://medgen.ugent.be/~jvdesomp/genorm. accessed on 10 December 2009) in pMEC [28]. The expression level of each gene was calculated according to the 2−ΔΔCt method [29]. Table 1 shows the primer sequences of target genes.2.4. Measurement of Inflammatory Factor LevelsCell culture and treatments were performed as described previously [30]. The culture medium in the 6-well plate was aspirated and discarded, the cells were washed with precooled PBS 3 times, and 200 μL of RIPA lysis buffer was added to every well, which was pipetted several times to fully lyse the cells. The lysate was transferred to a 2.0 mL centrifuge tube and centrifuged at 10,000 rpm for 10 min, and the cell supernatant was used to analyze the content of inflammatory factors. Cell supernatants were subsequently used for inflammatory factor analysis with a pig ELISA kit (TNF-α, #CSB-E16980p; IL-1β, #CSB-E06782p; IL-6, #CSB-E06786p), Cusabio Biotech Company, Wuhan, China) according to the manufacturer’s instructions.2.5. Western Blot AnalysisWestern blotting was performed as previously described [31]. Proteins were extracted from pMECs using the radio immunoprecipitation assay (RIPA) lysis buffer (#P0013B, Beyotime, Shanghai, China) and quantified using a BCA protein assay kit (#P0009, Beyotime, Shanghai, China). Proteins (10–20 μg/sample) were separated by SDS-PAGE (Invitrogen Inc.), transferred to nitrocellulose membranes (Millipore, Bedford, MA, USA.), and then hybridized with specific antibodies.Primary antibodies for TLR-4 (1:500, ab13556), p65 (1:1000, ab16502), phospho-p65 (1:1000, ab183559)), β-actin (1:1000, ab5694), p38 (1:1000, ab182453), phospho-p38 (1:1000, ab207483), ERK (1:1000, ab32537), phospho-ERK (1:1000, ab207470), JNK (1:1000, ab126424), and phospho-JNK (1:1000, ab279842) were from the Abcam Company Ltd. (Cambridge, MA, USA). Primary antibodies (dilution, cat. no. follow in parentheses) for IκBα (1:1000, #9242), and phospho-IκBα (1:1000, #2859) were from the Cell Signaling Technology (Woburn, MA, USA).2.6. Statistical AnalysisAll data are derived from at least 3 independent experiments performed in triplicate, and the results are expressed as mean ± SEM. One-way ANOVA was used for statistical analysis, and then Duncan’s test was used for multiple comparisons. p < 0.05 was used as the criterion for judging the significance of the difference, and p < 0.01 was the criterion for judging the extremely significant difference.3. Results3.1. Inflammatory Injury in LPS-Induced PMECsIn order to determine the effect of inflammatory damage on LPS-induced pMEC, cell viability was measured by the CCK-8 assay after the pMECs were treated with various concentrations (5, 10, 25, 50, 100, and 200 μg/mL) of LPS for 12, 24, and 48 h. As shown in Figure 1A, cell viability was significantly inhibited (p < 0.05) after stimulation with 200 μg/mL LPS for 12 h. Cell viability was also significantly inhibited (p < 0.05) after stimulation with 50, 100 or 200 μg/mL LPS for 24 h. In addition, the cell viability was significantly decreased (p < 0.05) after stimulation with all concentrations of LPS for 48 h. As shown in Figure 1B, after pMECs were treated with 50 μg/mL LPS for 24 h, cell apoptosis was significantly increased (p < 0.05), and mRNA expression abundance and the content of inflammatory factors (IL-1β, TNF-α, and IL-6) were also significantly increased (p < 0.05; Figure 1C,D). Thus, after comprehensive consideration of cell viability, apoptosis and inflammatory factors, 50 μg/mL LPS for 24 h was selected to establish inflammatory injury in pMECs for subsequent experiments.3.2. Effects of Artemisinin on the Viability of PMECsIn order to investigate the effect of artemisinin on pMECs viability, cells were treated with various concentrations (0.1, 1, 10, 20, 40, and 80 μM) of artemisinin for 6, 12, 24, and 48 h and analyzed by CCK-8 assay. As shown in Figure 2, low concentrations of artemisinin promoted cell growth and induced cell death in a concentration-dependent manner at higher concentrations. Maximum cell viability significantly increased (p < 0.05) when pMECs were treated with 20 μM artemisinin for 12 h. According to these results, 20 μM artemisinin (treated with 12 h) was chosen to examine the protective effects of artemisinin against inflammatory injury in LPS-induced pMECs.3.3. Effects of Artemisinin on Viability and Apoptosis in LPS-Induced PMECsIn order to determine the protective effect of artemisinin on inflammatory injury in LPS-induced pMECs, cell viability and apoptosis were measured after treatment with artemisinin or LPS. As shown in Figure 3, the LPS-induced decrease in cell viability was significantly inhibited (p < 0.05) by 20 μM artemisinin (Figure 3A), and LPS-induced pMECs apoptosis was significantly reduced (p < 0.05) by artemisinin (Figure 3B,C).3.4. Effects of Artemisinin on LPS-Induced Inflammatory FactorsTo analyze the anti-inflammatory effects of artemisinin, pMECs were treated with 20 μM artemisinin for 12 h before being exposed to 50 μg/mL LPS for 24 h. As shown in Figure 4A,B, the mRNA expression abundance and content of inflammatory factors (IL-1β, TNF-α, and IL-6) induced by LPS were significantly downregulated (p < 0.05) by artemisinin; however, artemisinin supplementation did not affect (p > 0.10) LPS-induced IL-8 mRNA expression abundance. Thus, the anti-inflammatory effect of artemisinin was confirmed in LPS-induced pMECs.3.5. Effects of Artemisinin on the NF-κB and MAPK Signaling Pathways in LPS-Induced PMECsThe nuclear factor κB (NF-κB) and mitogen-activated protein kinase (MAPK) signaling pathways are the most classic inflammatory signaling pathways. To determine the effect of artemisinin on the NF-κB and MAPK signaling pathways, we investigated the critical proteins of these signaling pathway by Western blot analysis. LPS activated cellular NF-κB and MAPK signaling pathways, causing inflammation. The phosphorylation levels of the NF-κB signaling pathway proteins p65 and IκBα were significantly increased (p < 0.05) (Figure 5), and the phosphorylation levels of the MAPK signaling pathway proteins p38, ERK and JNK were significantly increased (p < 0.05) (Figure 6) after the cells were treated with 50 μg/mL LPS for 24 h. When the cells were pretreated with 20 μM artemisinin for 12 h before being stimulated with 50 μg/mL LPS for 24 h, the phosphorylation levels of critical NF-κB and MAPK signaling pathway proteins were significantly decreased (p < 0.05; Figure 5 and Figure 6). These data showed that the activities of NF-κB and MAPK in LPS-induced pMECs were significantly inhibited by artemisinin.4. DiscussionEscherichia coli (E. coli) is one of the main pathogens causing clinical mastitis in animals [32]. LPS is the main component of the outer membrane of E. coli, which is an effective initiator of inflammation and endotoxic shock. In this study, LPS was used to stimulate inflammatory response in porcine mammary epithelial cells in vitro to study the effects of artemisinin on sow mastitis. The results showed that, after pMECs were stimulated with 50 μg/mL LPS for 24 h, cell viability was significantly decreased, inflammatory factor mRNA and protein expression and apoptosis were significantly increased (Figure 1), suggesting that the cells exhibited an inflammatory response. This finding indicates that the pMECs inflammation model was successfully established.Artemisinin has the potential anti-inflammatory effects [19], but the underlying mechanism of artemisinin in regulating inflammation has not been fully explained. Therefore, we studied the anti-inflammatory effect of artemisinin against LPS-induced inflammatory injury in pMECs and examined its mechanism. Previously, it has been confirmed that LPS stimulates mammary epithelial cells to produce a rapid inflammatory response, which is characterized by the release of a large number of pro-inflammatory cytokines (IL-1β, TNF-α, and IL-6) [33]. This response is good for attracting circulating immune effector cells (such as neutrophils) to fight infection, but excessive inflammation can damage cells and tissues [34]. Therefore, the expression of pro-inflammatory cytokines needs to be strictly regulated in the inflammatory response [35]. Studies have shown that artemisinin has a variety of biological functions, such as anti-inflammatory effects. Wang et al. [36] reported that artemisinin pretreatment potently inhibited IL-6 and TNF-αrelease induced by LPS in RAW264.7 cells, and artemisinin synergized with antibiotics to protect against LPS challenge by decreasing pro-inflammatory cytokine release. Yuan et al. [37] reported that proinflammatory factor levels (IL-1β, IL-6, TNF-α) and TLR-2 were significantly ameliorated by artemisinin treatment in LL37-induced rosacea-like mice, and artemisinin significantly decreased the LL37-induced expression of inflammatory biomarkers via the NF-κB signaling pathway in HaCaT cells. Li et al. [38] also reported that dihydroarteannuin can strongly reduce TNF-α levels in the culture supernatant of peritoneal macrophages and the serum of BXSB mice in vitro and in vivo, and the inhibitory effects on TNF-α production resulted from blockade of the NF-κB signaling pathway upstream of IκB degradation. Similarly, in this research, we examined the protective effect of artemisinin against LPS-induced inflammatory injury in pMECs and observed that the LPS-induced expression of IL-1β, TNF-α, and IL-6 was significantly down-regulated by artemisinin (Figure 4). It is worth noting that, in this experiment, we did not analyze the effects of artemisinin on inflammatory response without LPS challenge in pMECs, which is required to be addressed in the future.In the present study, LPS-induced inflammatory damage to pMECs were weakened by artemisinin without cytotoxicity (Figure 3 and Figure 4). In addition, we identified artemisinin as an inhibitor of the NF-κB and MAPK signaling pathways. LPS leads to the rapid and coordinated activation of various intracellular signaling pathways, including NF-κB and MAPK signaling pathways [39]. NF-κB regulates the expression of chemokines, cytokines, anti-apoptotic factors and cell growth factors, and can also trans-activate various physiological and pathological processes, such as innate and adaptive immune responses, cell proliferation, cell death, inflammation and, in some cases, tumorigenesis [40,41]. MAPK contains at least three MAPK families (ERK, p38 and JNK), which could be activated and phosphorylated by LPS, thereby activating transcription factors during inflammation [42,43]. Thus, inhibiting NF-κB and MAPK activation can be used as a potential therapeutic strategy for treating inflammatory events [44,45]. Previous studies have shown that pretreatment of Hep3B cells with artemisinin significantly inhibited TNF-α-induced expression of the NF-κB reporter gene. Artemisinin can also inhibit TNF-α-induced degradation of IκBα and p65 nuclear translocation [46]. In addition, artemisinin could decrease the expression of the inflammatory proteins COX-2 and iNOS and effectively inhibit vascular smooth muscle cells inflammation induced by TNF-α through the NF-κB signaling pathway [47]. Zhu et al. [48] reported that artemisinin attenuates LPS-induced proinflammatory responses by inhibiting the NF-κB pathway in microglial cells. Aldieri et al. [43] also showed that artemisinin inhibited NF-κB activation in cytokine-stimulated human astrocytoma T67 cells. Mechanistically, artemisinin suppressed LPS-induced NF-κB activation by blocking IκBα degradation and p65 phosphorylation and inhibited MAPK activation by blocking ERK, JNK, and P38 phosphorylation (Figure 7). It would be interesting to investigate whether artemisinin could be used as a drug to treat LPS-induced inflammatory. In addition, future studies are needed to study the anti-inflammatory activity of artemisinin in an in vivo model of sow mastitis.5. ConclusionsIn summary, the pretreatment of pMECs with artemisinin showed enhanced anti-inflammatory activity against LPS-induced inflammation. Artemisinin reduces the inflammatory damage of pMECs partially through the inhibition of NF-κB and MAPK signaling pathways.

animals : an open access journal from mdpi

10.3390/ani12070909

PMC8996973

Immune stress is an important stressor in domestic animals that leads to decreased feed intake, slow growth, and reduced disease resistance of pigs and poultry. Especially in high-density animal feeding conditions, the risk factor of immune stress is extremely high, as they are easily harmed by pathogens, and frequent vaccinations are required to enhance the immunity function of the animals. This review mainly describes the causes, mechanisms of immune stress and its prevention and treatment measures. This provides a theoretical basis for further research and development of safe and efficient prevention and control measures for immune stress in animals.

Immune stress markedly affects the immune function and growth performance of livestock, including poultry, resulting in financial loss to farmers. It can lead to decreased feed intake, reduced growth, and intestinal disorders. Studies have shown that pathogen-induced immune stress is mostly related to TLR4-related inflammatory signal pathway activation, excessive inflammatory cytokine release, oxidative stress, hormonal disorders, cell apoptosis, and intestinal microbial disorders. This paper reviews the occurrence of immune stress in livestock, its impact on immune function and growth performance, and strategies for immune stress prevention.

1. IntroductionStress refers to a series of nonspecific responses when the body is stimulated by external and internal abnormal stimuli [1]. In animal husbandry, common stresses include housing density, weaning, immunization, capture, transportation, and exposure to ammonia, temperature, humidity, noise, and high doses of pathogens [2]. Different stresses cause specific biological changes and endanger the animal’s health. Immune stress (IS) is also called immune stimulation. Specifically, in immunology, it refers stimulation of the body by an antigen eliciting an immune response. In a narrow sense, it refers to the systemic nonspecific adaptive immune response following inoculation with an antigen. In a broad sense, it refers to the frequent infection of various pathogenic microorganisms in the body under poor sanitary conditions, resulting in continuous activation of the immune system by invading pathogenic microorganisms and eliciting a nonspecific adaptive response [3]. In recent years, due to the large-scale and high-density farming environment and livestock genetic improvement, animals are vulnerable to a range of pathogens. Therefore, animals require to be vaccinated more frequently. Immunization can protect livestock from pathogens, but it can also stimulate immunity as a special stress factor, which can lead to a series of adverse phenomena such as fever, depression, anorexia, intestinal microflora disorders, nutritional and metabolic changes, and reduced growth, culminating even in increased mortality, resulting in a huge economic loss to farmers and the industry in general. This paper reviews the effects and mechanisms of IS on immune function and growth performance of livestock and poultry and provides a basis for revealing the molecular mechanism of IS. 2. Mechanism of IS on the Neuroendocrine–Immune SystemImmune system activation leads to the release of cytokines (interferon-γ (IFN-γ), interleukin-6 (IL-6), interleukin-1β (IL-1β), nitric oxide (NO), prostaglandin E2 (PGE2), tumor necrosis factor (TNF-α), etc.) by monocytes, macrophages, and lymphocytes [4,5,6,7]. The classical lipopolysaccharide (LPS)-induced IS model has shown that the release of inflammatory cytokines is closely related to toll-like receptors (TLRs) found on the immune cells’ surface [8,9,10]. When stimulated by an immunogen, the TIR region of TLR4 binds to the carboxyl end of myeloid differentiation factor 88 (MyD88), and its amino end binds to the IL-1 receptor-associated kinase (IRAK). After phosphorylation, IRAK is separated from MyD88 and released into the cytoplasm. Activated IRAK then binds to TNF receptor-related factor 6 (TRAF-6) and connects with nuclear factor κB (NF-κB)-inducible kinase (NIK), further activating the IκB kinase complexes (IKKs). Activated IKKs act on IκB t, ubiquitinating and degrading, thus activating the NF-κB. Meanwhile, the mitogen-activated protein kinase (MAPK) signaling pathway is stimulated, resulting in activation of cytoplasmic protein 1 (AP-1) which then enters the nucleus to induce the expression of a variety of inflammatory cytokines and chemokines, macrophage migration, and phagocytosis [11,12,13,14,15,16]. In addition to TLR4-NF/κB and MAPK/AP-1 signaling pathways, the JAK/STAS signaling pathway is also involved in the release of proinflammatory factors during IS [17].Cytokines released by peripheral immune cells subsequently enter the central nervous system (CNS) to activate local immunity through several mechanisms, including (1) the active uptake mechanism via the blood–brain barrier [18], (2) stimulation of the leakage region in the blood–brain barrier of circumventricular organs [19], (3) influencing the afferent neurons of the peripheral vagus nerve to transmit cytokine signals to nerve cells in relevant brain regions, including nucleus tractus solitarius and hypothalamus (also known as the “neural pathway”) [20,21,22], (4) activation of the endothelial cells and microglia in the cerebrovascular system to produce local inflammatory mediators such as inflammatory cytokines, chemokines, cyclooxygenase-2(COX-2), prostaglandin E2 (PGE2), and NO [23,24], (5) activation of immune cells such as monocytes/macrophages and T cells recruited from the periphery to the brain parenchyma via monocyte chemoattractant protein-1 (MCP-1) to produce more cytokines and inflammatory mediators in the brain [25,26]. After entering the CNS, these peripheral cytokine signals are amplified by the local inflammatory network (including the inflammatory signal transduction pathway, cytokines, COX-2 expression, and PGE2 release), so that the CNS produces a large number of inflammatory cytokines (such as TNF-α, IL-1 β, IL-6, IFN-γ, etc.) [23]. Furthermore, a large number of inflammatory factors induce excessive production of reactive oxygen species (ROS) which cause oxidative stress [27] and enlarge central nervous inflammation [28].The hippocampus in the HPA axis is sensitive to stress injury. The sharp increase in inflammatory cytokines coupled with oxidative stress lead to dendritic atrophy and neuronal death in the hippocampus [29], followed by activation of the HPA resulting in the synthesis of corticotropin-releasing hormone (CRH) and arginine vasopressin (AVP) by the intermediate small cell neurons in the hypothalamic paraventricular nucleus. These secretagogues are released to the anterior pituitary through the portal blood. CRH acts on the pituitary and binds to CRHR1 to promote the synthesis and secretion of pro-opioid cortisol (POMC) in the anterior pituitary. POMC is catalyzed by the prohormone-converting enzyme, resulting in the release of adrenocorticoid hormone (ACTH) from the anterior pituitary. ACTH in the bloodstream activates the adrenal melanocortin receptor 2 (MC2R) to stimulate adrenal cortical cells to synthesize and release cortisol (COR) and other glucocorticoids (GC) [6,30].There are abundant glucocorticoid receptors in the hippocampus which can inhibit the stress response of the HPA axis when bound to glucocorticoid [31,32,33]. In contrast, COR can suppress the immune system by targeting genes related to cytokines, chemokines, inflammatory proteins and their receptors, and also affect cell adhesion. At the transcriptional level, COR binds to its receptors, inhibits the transcriptional activity of AP-1 and NF-κB, thereby inhibiting the expression of various cytokines [34]. It can also directly act on immune cells such as macrophages, mast cells, and basophils, thereby inhibiting their immune function, inducing apoptosis of B and T lymphocytes and dendritic cells, inhibiting their proliferation, differentiation, and migration, reducing humoral and cellular immune function, and causing immunosuppression [35,36]. The mechanism of IS effects on the neuro–endocrine–immune system is shown in Figure 1. 2.1. IS Leads to Immune Dysfunction of Livestock The effect of IS on animal immune function is bidirectional. It may enhance or suppress immune function, which depends largely on stress intensity and duration of stress [37]. It is generally believed that short-term (acute) stress can enhance the immune function of livestock, while long-term, low-intensity (chronic) stress can cause immunosuppression [37,38].2.1.1. Acute IS Causes Enhanced Immune Function of Livestock IS significantly increases the plasma concentrations of COR, PEG2, IL-6, TNF-α, and IL-1β [39]. IS induced by LPS increases IL-1, IL-2, IL-6, immunoglobulin G (IgG), and immunoglobulin A (IgA) levels in broiler serum [40]. LPS-induced IS can strongly inhibit C4BPA (a cycle inhibitor of the classical and MBL pathways of the complement system activation pathways) in peripheral blood mononuclear cells (PBMC) in pigs, thereby enhancing the activation of the complement system [41]. Chickens vaccinated with Eimeria tenella express high IgG and sIgA levels [42]. LPS and human serum albumin (HuSA) used to construct a broiler IS model found that IS significantly increases total immunoglobulin, immunoglobulin M (IgM), IgA, and IgG levels in the serum, indicating that IS enhances the broiler humoral immune response [43]. Similarly, IS significantly increases serum IgA levels in Linwu ducks [44], the spleen and thymus index of broilers [45], while LPS-induced IS increases the ratio of CD4+/CD8+ in duck spleen and thymus [46]. IS can increase the concentration of IL-1β, PGE2, and COR in piglet plasma and promote the proliferation of peripheral blood lymphocytes, an indication of enhanced cellular immune function in piglets [47,48,49], similarly to another study which also showed proliferation of lymphocytes and increased ratio of CD4+/CD8+ in peripheral blood lymphocytes of chickens, and a significant increase in serum TNF-a, IL-6, IL-1, and IgG levels [50,51]. Thus, all of the above studies confirm that IS can enhance the immune function of livestock fairly quickly so that the body can better resist a pathogen invasion.2.1.2. Chronic IS Causes Immunosuppression in Livestock In contrast to acute IS, chronic IS causes suppression of cellular immunity in livestock. Chronic IS increases serum ATCH, COR, and expression of Caspase-3 and Caspase-9 in immune cells, decreases the ratio of Bcl-2 to Bax, decreases the index of immune organs (bursa, thymus, and spleen) and the number of neutrophils, CD4+ cells, and CD8+ cells, and induces apoptosis of immune cells [52,53,54]. Chronic IS also affects humoral immunity, a decrease in serum IgG levels and lysozyme activity in livestock [55,56]. Chronic IS also affects intestinal immunity. Chronic IS markedly damages the intestinal structure of livestock, enhances lymphocyte apoptosis in intestinal Peyer’s nodules, reduces intestinal IgA-secreting cells (ASC) and secretory immunoglobulin A (sIgA), causing destruction of the intestinal immune barrier [54] and also the intestinal mechanical barrier by downregulating tight junction proteins and inducing intestinal epithelial cell apoptosis, and the intestinal microbial barrier by interfering with the composition of intestinal microorganisms, with all these effects culminating in serious damage to intestinal immune function [57,58,59]. Thus, chronic IS inhibits the body’s intestinal, cellular, and humoral immunities.3. Mechanism of IS on Growth Performance of Livestock and PoultryAlthough the impact of acute and chronic IS based on the intensity and duration of stress is different, the effects on the livestock growth performance such as the decrease in average daily feed intake (ADFI) and average daily gain (ADG) and increased feed meat ratio (FCR) are somewhat similar.3.1. Effects and Mechanism of IS on Livestock Feed Intake IS significantly decreases the ADFI of livestock [45,47,53,60]. It is believed that this could be due to a reduced feed intake caused by secretion of specific hormones and inflammatory factors.The brain–gut axis plays a vital role in the loss of appetite caused by IS. The hypothalamus, especially its arcuate nucleus (ARC) that is involved in energy perception and integration. In ARC, neuropeptide Y (NPY) and, agouti-related peptide (AgRP), a neuropeptide produced by the AgRP/NPY neuron, and also gamma-aminobutyric acid (GABA) can increase food intake, while POMC, cocaine- and amphetamine-regulated transcripts (CART) reduce food intake [61]. While intestinal hormones (leptin, cholecystokinin (CCK), glucagon-like peptide 1 (GLP-1), resistin, and growth hormone-releasing peptide) can initiate most of the signal transduction and communication in the gut–brain axis, they can also regulate appetite by activating or inhibiting hypothalamic neurons [62]. In IS, a large number of inflammatory factors released by the peripheral immune system enter the circulation. These can induce adipocytes to synthesize and secrete leptin [63,64] and transport it to the brain. With assistance from IL-1β, leptin binds to the leptin receptor in hypothalamic neurons [65], stimulates POMC neurons in ARC neurons to produce anorexic peptide POMC, and inhibits AgRP/NPY neurons from producing appetite peptides NPY and AgRP [20,66,67]. LPS-induced IS can significantly increase resistin and GLP-1 and decrease ghrelin in animals [68,69,70]. Excess GLP-1 can overactivate GLP-1 receptors in the CNS, gastrointestinal tract (GIT), and pancreas, activate GLP-1 receptors in the CNS, all of which can lead to a fullness feeling in the animal. GLP1 can also reduce the speed of gastric emptying and hinder gastric acid secretion, thereby increasing gastric dilatation, limiting excessive food consumption, and enhancing satiety [68]. Ghrelin, the only known peripherally derived appetite hormone, activates the hypothalamic growth hormone secretagogue receptor (GHSR-1a) [61], mediates the activation of AgRP/NPY neurons in ARC [71], and inhibits the satiety POMC neurons [72], promoting appetite. Therefore, a decline in the level of ghrelin leads to a decrease in appetite. The mechanism of appetite suppression by IS is shown in Figure 2.3.2. Effect of IS on the Digestion and Absorption of Livestock and PoultryMucosal integrity, digestive enzymes, transport proteins, and microbes of the intestines play a major role in the digestion and absorption of nutrients. Nutrient absorption mostly occurs in the small intestine (SI). Iintestinal villi height (VH) growth will increase the contact area between the nutrients and SI, resulting in improved absorption rate of nutrients. This is supported by shallower crypt depth (CD) which facilitates the proliferation of intestinal epithelial cells to further improve absorption from the SI. Thus, both VH and CD, and the ratio of VH/CD (known as V/C) are often used to evaluate intestinal absorptive capacity [73]. Furthermore, intestinal microbes can ferment indigestible substrates such as dietary fiber and endogenous intestinal mucus to produce short-chain fatty acids (SCFA) which are easily absorbed [74]. Thus, intestinal microbial composition is important for optimal intestinal function.IS can seriously affect intestinal health. LPS-induced IS can reduce VH and enhance CD and reduce alkaline phosphatase (AKP) activity in the duodenum and jejunum, and also decrease the number of duodenal sodium-dependent glucose cotransporter 1 (SGLT1) and the plasma D-xylose level [75]. IS decreases the level of lactase, maltase, and sucrase markedly in the jejunum and ileum of piglets [76,77]. Furthermore, IS increases SI interleukin-8 (IL-8) secretion, ROS production, and TNF-α mRNA abundance, but decreases the level of SGLT1, excitatory amino acid transporter 1 (EAAC1), H+/peptide cotransport 1 (PEPT-1), and intestinal fatty acid binding protein 2 (I-FABP2) in the intestine of weaned piglets [78]. It was also found that the abundance and distribution of tight junction proteins (zonula occluden-1 (ZO-1), Occludin, and claudin-1) mRNA in the SI declines when exposed to IS [79]. IS makes duck SI walls thinner, lowers the V/C value, reduces the expression of tight junction proteins (ZO-1, Occludin, and Claudin-1), and enhances intestinal permeability [44]. IS not only downregulates the expression of digestive enzymes (maltase, AKP, and Na-K-ATPase) and tight junction proteins (CLDN-1, OCLD, ZO-1, ZO-3, EpCAM, and JAM2), but also reduces the phosphorylation of mTOR protein in the jejunum and inhibits the proliferation of jejunum cells [80]. IS causes rumen and intestinal flora disorders in livestock, resulting in SI damage, reduced intestinal flora diversity, and an increased harmful-to-beneficial bacteria ratio [56,81,82,83,84,85,86].Intestinal injury cause inflammation, oxidation, apoptosis, and autophagy. IS induces the expression of key genes related to TLR4 and the nucleotide-binding oligodomain proteins (NODs) signaling pathway and increases the expression of inflammatory factors (IL-1β, IL-2, IL-6, IL-8, INF-γ, and TNF-α), iNOS, and COX-2 in the SI [44,76,77,78,87]. IS also reduces SI superoxide dismutase (SOD) and glutathione peroxidase (GSH-Px) activities and increases malondialdehyde (MDA) and reactive oxygen species’ (ROS) levels [44,86]. In addition, IS induces the expression of apoptosis-related genes (Caspase-3, Caspase-8, Bax, A20, and MT) and autophagy-related pathway key proteins (PTEN-induced putative kinase 1 and parkin), increases the ratio of light chain 3-II (LC3-II) to LC3-I in the intestine, all of which leads to apoptosis and autophagy in the SI [44,80].As mentioned above, IS can lead to a release of a large number of inflammatory cytokines and chemokines by peripheral immune cells. Inflammatory cytokines in the SI and over-recruited immune cells mediate SI damage and decrease ADFI, mediating the destruction of SI integrity and apoptosis of epithelial cells [88]. However, there is little research on the mechanism of the changes of intestinal digestive enzymes and related transporters caused by IS. Some believe that the changes of digestive enzymes and transporters are closely related to the release of glucocorticoids [89]. Thus, in general, IS causes microbial flora disorders, reduces intestinal digestive enzymes and transporters, and damages SI structural integrity, culminating in digestive and absorption disorders. 3.3. Effect of IS on Nutrient Metabolism in Livestock Although the reduction of feed intake, digestion, and absorption are the most important factors in IS-induced growth suppression of livestock, the redistribution of nutrients also plays a role [90]. After the regulation of multiple hormones (growth hormone (GH), thyroid hormone, IGF-1, and GC) and cytokines, the body will redirect nutrients originally used for muscle synthesis and growth to the immune system to maintain a highly activated immune system to prevent diseases, resulting in an increase in FCR [90,91,92,93,94,95].The activation of the HPA axis caused by IS leads to an increase in the secretion of catabolic hormones such as ACTH and CORT and a decrease in the secretion of anabolic hormones such as GH and IGF-1. In addition, a large number of inflammatory cytokines are secreted, which inhibits the synthesis of energy and promotes the decomposition of energy [96]. Mechanistically, the proinflammatory cytokines released by the body during IS directly act on the thermoregulatory center via the blood–brain barrier, resulting in an increase in body heat production and a reduction in body heat dissipation, which results in an increase in body temperature [7,97]. IS increases energy consumption because in thermostatic animals, the metabolic rate needs to be raised by 10–12.5% to increase the body temperature by 1 °C [98].In carbohydrate metabolism, IS promotes gluconeogenesis and glycogen hydrolysis in the liver, resulting in an increase in glucose production, a reduction in the uptake of glucose in peripheral tissues such as skeletal muscle and myocardium, resulting in an increase in blood glucose concentration. Because a lot of aerobic energy and hence oxygen is required for the immune response, an anaerobic situation is created that results in the conversion of glucose into lactic acid [3,52].In protein metabolism, IS increases skeletal muscle protein degradation rate, reduces skeletal muscle protein deposition, speeds up peripheral protein decomposition, enhances liver degradation of valine, leucine and isoleucine, and produces a large amount of liver acute phase protein (APP). In addition, the induction of inflammatory cytokines upregulates the expression of liver proteins involved in immune defense function, amino acid catabolism, ion transport, wound healing, and hormone secretion [92,99].In lipid metabolism, IS induces the degradation of a large number of lipoproteins in the body and produces a series of inflammatory factors such as TNF-α, which can inhibit the synthesis of fatty acids (FA) in adipose tissue and promote fat degradation [100]. TNF-α also stimulates the liver to synthesize FA de novo and lipolysis in adipose tissue. In addition, the AMP-dependent protein kinase (AMPK) lipid metabolism pathway is activated, inhibiting the activity of acetyl-CoA carboxylase (ACC), thereby reducing the conversion of malonate-CoA. This activates carnitine palmitoyltransferase-1 (CPT-1), which in turn enhances the activity of CPT-1 and the expression of peroxisome proliferator-activated receptor-α (PPAR-α) mRNA, thereby stimulating liver lipid metabolism, accelerating FA oxidation and reducing fat deposition [101].Thus, IS reduces the appetite of livestock, redistributing nutrients in the body to the immune system, resulting in a decline of ADFI, enhancing catabolism, reducing intestinal digestion and absorption function, all of which finally leads to a decline in growth performance of livestock and poultry, resulting in economic loss to the livestock industry.3.4. Other Effects of IS on Livestock IS can cause the activation of the OPG/RANKL pathway in tibia and increase the expression of inflammatory cytokines. Furthermore, IS can enhance the generation of osteoclasts, leading to impairment of bone development [102]. IS can reduce chicken muscle pH, which can result in pale muscle [95]. Mechanistically, IS can lead to accumulation of lactic acid in muscle, resulting in a lowering of muscle pH, which can lead to an increase in protein denaturation and muscle fiber damage, damaging the hydration and texture of meat [103]. In addition, protein denaturation may lead to a decrease in light transmittance to the meat’s surface, resulting in pallor meat [104]. IS can also lead to a sharp decline in lactation of cows [105] due to inhibition of the release of prolactin, GH, and IGF-I, resulting in reduced milk quality and yield [106]. LPS-induced IS decreases the feed intake and egg laying rate in hens, leading to a significant increase in eggshell thickness, strength, and a reduction in albumin quality and content [107].4. Prevention and Control Technologies for ISBecause of intensive farming, the frequency of IS has increased, resulting in a negative impact on the animal industry. It is particularly important to develop safe and effective techniques to prevent and alleviate the effects of IS. Controlling stress, improving management, and resisting the stress response by nutritional intervention are the most popular methods used currently.4.1. Vaccination ProgramIn order to determine which kind of vaccines are necessary for an individual farm, it is important evaluate the endemic infectious agents in the region, the age of the animal, genetic and health status of the breeding animals, the distance to other farms, and the level of biosecurity to be implemented in the farm [108]. In poultry production, vaccines for Newcastle disease (ND), infectious bronchitis, and infectious bursal disease are used in most countries [109]. Furthermore, vaccine programs should be designed based on other impact factors such as interaction between different vaccines, interference with maternal antibodies, vaccine type, vaccination method, and vaccination frequency. Unnecessary vaccination should be avoided to prevent excessive stimulation of the animal immune system [109,110].Combined immunization is a scheme worthy of consideration. Combined immunization refers to simultaneous injection with two or more antigens in combination to reduce animal stress [111]. Some have used an inactivated vaccine against serotype 4 and 8 bluetongue disease to immunize sheep and found that simultaneous double-injected booster vaccination yields the highest median serotype-specific neutralization antibody 26 weeks after the first vaccination and a positive serum antibody level at a maximum even after one year [112]. Trivalent vaccine (Mycoplasma hyopneumoniae, porcine circovirus type 2 (PCV2) and porcine reproductive and respiratory syndrome virus (PRRSV)) used in pigs has a protective effect on the infection of three pathogens, and the effect is similar to that of a monovalent vaccine, and their growth performance is superior to the uninfected pigs [113]. Similarly, a combined vaccine (Mycoplasma hyopneumoniae and PRRSV) is effective in pigs [114]. A recombinant herpesvirus of turkey laryngotracheitis vaccine (rHVT) combined with chicken embryo origin laryngotracheitis vaccine (CEO) provides stronger protection than rHVT alone does. rHVT can reduce the virulence return and replication of CEO, which is safer than CEO alone. rHVT–CEO vaccination strategy is another way to achieve better disease control [115]. In general, combined immunization can exert a positive effect without interfering with the effects of each vaccine and reduce the number of immunizations, reducing cross-infection, stress, and improving health and production efficiency.4.2. Improved Feeding RegimeImproving the hygiene of the feeding environment and strengthening farm management can minimize the occurrence of IS. Good hygiene conditions can reduce the contact chance between the animal and pathogenic microorganisms in the environment, thereby reducing the excessive stimulation of the host immune system [116]. A decrease in feeding density has also been used to prevent IS. High-density feeding has been shown to increase IS and thereby affect the growth performance of geese. A reasonable stocking density is not only good for animal welfare but also for better animal growth [117]. Currently, other measures such as “all-in, all-out”, group feeding, and improvement in uniformity of animals are widely implemented in poultry farms, which have reduced the chance of poultry contact with bacteria and viruses and thereby reduced the occurrence of IS. 4.3. Nutritional RegulationAlleviating animal IS through nutritional regulation has been highlighted in recent years because animals consume more when under IS. This is because of the amount of nutrients required to synthesize immune effector molecules [118]. The addition of amino acids, fats, vitamins, and plant extracts in feed have been shown to alleviate IS and enhance host resistance to IS.4.3.1. Amino Acid AdditivesIS not only increases amino acid requirements of the host but also affects amino acid balance. Adding L-arginine to the diet can effectively alleviate the damage to the intestinal mucosal barrier caused by IS in order to maintain intestinal integrity [54,119]. Mechanistically, L-arginine inhibits the TLR4 signaling pathway, decreases the percentage of CD14+ cells with resultant overexpression of proinflammatory factors in animals in IS [9]. In addition, other amino acids and their derivatives, such as glutamic acid [76], glutamate precursor α-ketoglutarate [120,121], L-theanine [122], cysteine [123], N-acetylcysteine [77,124,125], glycine [126], glycyl-glutamine [88], glutamine [127], and asparagine [128,129] have been used as anti-IS additives in feed.4.3.2. Fatty Acid AdditivesPolyunsaturated FA can relieve IS by reducing the release of inflammatory factors. Adding conjugated oleic acid to the diet can inhibit the production of IL-1β, enhance proliferation of lymphocytes by inhibiting the expression of PGE2, and thereby improve the growth performance of pigs [130]. n3 FA can also reduce the expression of PGE2 and the level of inflammatory cytokines in the serum induced by LPS [131]. Fish oil can reduce the release of inflammatory cytokines and improve the growth performance in IS pigs [47,132]. While adding fish oil to the diet can increase cellular immunity and reduce the inflammation index in animals exposed to IS [133], increasing the ratio of n3 to n6 fatty acids in the diet can inhibit the inflammatory response, thereby improving the animals’ performance [134].4.3.3. Vitamin AdditivesVitamins are necessary for animal metabolism and growth, acting in a variety of ways. Vitamins act to minimize stress, inflammation, and, in general, immune function regulation. Vitamin C can regulate the inflammatory response and oxidative stress caused by LPS and reduce hippocampal cell apoptosis [135]. Vitamin C can also alleviate the damage caused by IS in piglets [136]. Under stress conditions, the requirements for vitamin C are increased, and therefore it is widely used in animal feed to alleviate the health impacts of IS. Adding high doses of vitamin E can increase the feed intake of laying hens during IS and also increase the level of antibodies after vaccination for NDV and avian influenza virus [137]. Vitamin E can inhibit the increase in proinflammatory cytokines, PGE2, and cortisol caused by IS [131,138]. In vitro studies have shown that vitamin A can improve intestinal barrier integrity and reverse LPS-induced intestinal barrier damage by enhancing the expression of tight junction proteins [139].4.3.4. Trace Element AdditivesTrace elements affect immune regulation and can function to reduce stress. For example, MgSO4 significantly attenuates LPS-induced acute lung injury, apoptosis, oxidative stress (reducing MDA levels), and lung inflammation [140] and inhibits the NF-κB signaling pathway in mice [141]. Chromium yeast can effectively reduce IL-1β, TNF-α, and COR in piglet serum, thereby alleviating IS [142]. Clinoptilolite (aluminosilicate containing sodium, potassium, and calcium) can inhibit the infiltration and hyperactivation of neutrophils in broilers and reduce plasma and SI mucosal inflammatory cytokines, thereby reducing the inflammatory response caused by LPS [10]. Copper also has an anti-IS additive effect [143,144].4.3.5. Probiotics AdditivesProbiotics, also known as microecological preparations, are a group of active microorganisms or metabolites that benefit the host. The role of probiotics is to improve intestinal health and immune function by regulating the colonization and composition of the intestinal flora, intestinal pH, and also improving communication through the intestinal–brain axis pathway [145]. Feeding some probiotics (Lactobacillus acidophilus, Lactobacillus casei, Bifidobacterium thermophilum, and Enterococcus faecium) in the diet can alleviate intestinal inflammation, oxidative stress, and morphological damage caused by enterotoxigenic E. coli and enhance the growth performance of piglets [146]. Adding Bacillus subtilis to broiler diets can normalize the expression of gut barrier-related genes (JAM2, occludin, ZO1, and MUC2) under IS [147]. Fermentation products of Bacillus subtilis not only reduce the expression of LPS-induced SI inflammation genes but also enhance the expression of intestinal barrier genes, thereby improving the growth performance of broilers under IS [58]. In addition, probiotics such as Bacillus amyloliquefaciens [55], Lactobacillus casei [148], Lactobacillus acidophilus [149], Lactobacillus delbrueckii [150], Enterococcus faecalis [151], Clostridium butyricum [151], and also their fermentation products, can alleviate IS. 4.3.6. Plant Extract AdditivesSeveral low-molecular-weight secondary metabolites of plants can also play a role in modulating immune stress along with their antibacterial, antiviral, anti-inflammatory, and anti-tumor functions. Since many plant extracts exhibit low toxicity, high safety, and health improvement functions, it has become a research hotspot in the field of feed additives.Table 1 shows the more important IS resistance plant extracts reported in recent years. These plant extracts include esters, phenols, polysaccharides, glycosides, flavonoids, alkaloids, and crude plant extracts [16,40,152,153,154,155,156,157,158]. Anti-IS mechanisms of plant extract can be categorized into three types, namely (1) relieving inflammation and oxidative stress by inhibiting inflammatory response and oxidative pathways, (2) maintaining intestinal balance by regulating intestinal barrier-related proteins and promoting intestinal cell proliferation, and (3) maintaining hormone balance by regulating the endocrine system. 5. ConclusionsOverall, IS significantly impacts the host neuro–endocrine–immune axis and the brain–gut axis, which can result in hormone secretion disorders, cell apoptosis, intestinal flora disorders, intestinal barrier destruction, oxidative stress, and metabolic disorders, all of which can compromise the immune function and growth performance of livestock. Therefore, efficient and feasible control measures, such as providing good quality nutritious feed, reducing excessive stimulation of the immune system by organizing vaccine procedures at the appropriate time, and using selected plant extracts and potential new drugs for control, are required, as we are experiencing an upsurge of new infectious diseases in livestock.

animals : an open access journal from mdpi

10.3390/ani11061695

PMC8227609

One reason for lameness in cats is the rupture of the cranial cruciate ligament. This ligament is located in the stifle joint and contributes to its stabilization during excessive forward movement and internal rotation of the tibia. One method for the surgical treatment of cranial ligament rupture is the placement of an extracapsular suture. Different materials and methods of suture fixation have been used in dogs and cats. This study investigated the use of a novel polylactide absorbable bone anchor that was implanted with ultrasound technology for suture fixation and compared this with suture fixation alone and fixation with a nonabsorbable bone anchor using an ex vivo modified limb-press model. For evaluation, distance measurements on radiographs were performed and the angles between defined bony structures were calculated. The acquired measurements accounted for both craniocaudal and mediolateral movements, and the results showed that the absorbable anchor could neutralize excessive movement within the stifle joint in two of three measurements and seems to be a good alternative to well-known surgical methods.

Background: This study evaluated joint stability after surgical repair of cranial cruciate ligament (CrCL)-deficient stifle joints in cats using a novel absorbable polylactide bone anchor in an ex vivo model. Methods: Thirty-six hindlimbs from cats with intact (Gi group) and transected CrCLs were treated with fabellotibial suture alone (GFW group), suture combined with an absorbable polylactide bone anchor (GWD group), or suture combined with a nonabsorbable bone anchor (GFT group), positioned in a limb press with predefined joint angles (stifle joint: 120 ± 5°; hock joint: 120 ± 5°) and loaded with 10%, 20%, and 30% of body mass (BM). Predefined points were measured on lateral radiographs and with a coordinate measurement machine. Distances on radiographs (mm) were measured and angles (°) were calculated to represent the craniocaudal movement and the internal rotation of the tibia. Results: There were no differences for craniocaudal movement between Gi and GFW or GFT, but for GWD regarding angle measurement at 30% BM. For internal rotation, there was no significant difference between Gi and GFW or GWD, but for GFT. Conclusion: The used absorbable polylactide bone-anchor was able to stabilize the stifle joint regarding internal rotation and craniocaudal movement as calculated from distance measurements.

1. IntroductionThe cranial cruciate ligament (CrCL) is an important anatomical structure in the stifle joint. Its functions include prevention of cranial tibial drawer, excessive internal rotation of the tibia, and hyperextension of the stifle joint [1,2,3]. CrCL disease occurs less commonly in the feline species as opposed to canines and humans [1,3]. Conservative management may seem to resolve the lameness and its associated pain but most of these cats will continue to have some degree of cranial tibial translation [3,4,5,6]. Surgery is thus recommended to prevent continued stifle instability and provide a quicker and more reliable return to function [3,5]. Tibial-plateau-leveling osteotomy (TPLO) and tibial tuberosity advancement (TTA) are commonly used to treat CrCL disease in dogs [7,8,9,10,11,12,13]. TPLO and TTA however has only been investigated in a few studies in cats, whereas extracapsular stabilization with fabellotibial suture has been reported more commonly [5,12,14,15].Limb-press models have been widely used in research studies to investigate different stabilization methods in dogs and cats [8,10,13,14,16,17,18]. Stability in the stifle in these limb-press models has been evaluated using either radiographic measurements [16,18] or a three-dimensional coordinate-measuring system when comparing different surgical techniques [18].The use of ultrasonically implanted absorbable bone anchors has been investigated in sheep for treatment of mandibular fractures, spinal fractures, skull reconstruction and for implantation in the femur and tibia [19,20,21,22]. In humans, these implants have been used for tendon and ligament repair and for joint stabilization [23,24]. In cats and dogs, there are no published reports on their clinical use. In contrast the use of conventional bone anchors for suture placement in the femur during extracapsular stabilization has been well described in dogs, and it has been shown to be an effective option for the surgical treatment of CrCL rupture [25,26,27]. However, there are no such studies evaluating the use of bone anchors for fabellotibial suture fixation in cats. In this study we investigated the surgical stabilization of the CrCL-deficient stifle joint in cats, using either an ultrasonically implanted absorbable polylactide bone anchor or a conventional anchor in an ex-vivo model.The absorbable polylactide anchor is implanted with BoneWelding technology (VetWelding AG, Stansstad, Switzerland), which uses ultrasound for implantation [21,28]. The ultrasound is used to liquefy the anchor polymer at the bone contact interface. The liquid then flows into cancellous bone cavities and immediately solidifies, creating a stable bond with the bone [21,28]. At 37 °C, there is no alteration in the anchor’s pull-out strength for a period of 12 months [29,30]. This period is estimated to be about 6 months for live animals, given that their body temperature is usually higher [29,30]. This time period is also sufficient for the formation of periarticular fibrosis which is ultimately responsible for long-term joint stabilization [29,30]. The benefits of this implantation method include the ability to withstand higher pull-out forces, faster implantation, with no evidence of inflammation at the implantation site compared with the use of conventional bone screws [19,20,21,28,31].We hypothesized that a fabellotibial suture technique using this ultrasonically implanted absorbable polylactide bone anchor would neutralize excessive craniocaudal drawer and prevent internal rotation of the tibia, after CrCL rupture in cats. We also hypothesized that this technique would be comparable to using a fabellotibial suture either on its own or in combination with a nonabsorbable bone anchor.2. Materials and Methods2.1. SpecimensHindlimbs (n = 36) from cats euthanized for reasons unrelated to the study were harvested from the Hospital for Small Animals of the University of Veterinary Medicine Vienna, Austria. Prior signed consent from the owners was obtained for using the cadavers for teaching or research. The limbs were free from stifle pathology as evaluated by radiography. The recorded body mass (BM) was 3.6 ± 1.1 kg (range 1.7–6.2 kg), and the age was 12.4 ± 5.3 years (range 2.0–19.4 years).The limbs were prepared as previously described [16,18]. Briefly, they were disembodied, and the soft tissues from the femoral head to the proximal metatarsus were removed. Stifle and talocrural joint capsules, patellar tendon, fabellae, and collateral ligaments were preserved. The femoral head and neck were removed using a saw. The hindlimbs were wrapped in saline-soaked towels and stored at −20 °C until testing.Prior to testing, the hindlimbs were thawed at room temperature. The femoral shaft was fixed in a copper tube containing polymethyl methacrylate. To simulate the quadriceps mechanism, a wire was guided through a predrilled hole in the center of the patella and connected to a force gauge mounted on the top plate of the limb press. Two 2.0 mm cortical bone screws (Braun, Austria) of 6.0 mm length were horizontally placed in a medio-proximal orientation to both fabellae. A wire was placed around the screws and connected to a turnbuckle which, in turn, was connected to a wire guided through a predrilled hole in the proximal aspect of the calcaneus to simulate the Achilles tendon mechanism.2.2. Limb PressFor mechanical testing, a limb press was used that has been adapted for feline hindlimbs by Kneifel et al. [16,18]. It comprised a rectangular base plate connected by a column (90° angle) in each corner to an equal-sized top plate. Connection was achieved over drilled holes allowing the top plate to slide up and down and to be secured with screws. To measure the created patellar tendon load (PTL), a device mounted on the top plate held a force gauge (Sauter FA-100; Sauter GmbH, Vienna, Austria) connected to the wire, simulating the quadriceps mechanism. The paw was placed on a pedestal with a rough surface and additionally fixed with a K-wire. The pedestal was placed in the center of a scale (WPT 30F1/K, 0.01 kg digit; Radwag Wagi Elektroniczne, Krakow, Poland) to measure the body mass applied to the hindlimb (Figure 1). In addition, at 0.5 cm distal to the stifle joint surface, the caudal aspect of the tibia was connected to a spring mechanism to keep the tibial plateau from translating cranially relative to the femur.On the back, a radiographic plate was attached parallel and sagittally to the limb.2.3. Surgical TreatmentFor different treatment methods, the hindlimbs were randomly divided into the three groups. For each group, appropriate placement of the created bone tunnels was obtained using radiographs.2.3.1. Fabellotibial Suture AloneThe first group (GFW) underwent a fabellotibial suture technique with #2 braided nonabsorbable multistrand polyethylene suture (FiberWire; Arthrex Vet Systems, Frechen, Germany) without the use of a bone anchor.The suture was placed through the lateral femorofabellar ligament, guided medially through a bone tunnel cranial to the proximal aspect of the extensor groove of the tibia [32], and guided back laterally under the distal aspect of the patellar tendon and knotted with the opposite end of the suture. Tightening was performed with the mentioned force gauge, at 20 N, as previously described, while the stifle joint was held at an angle of 100° [16,18,33,34] and fixation was performed with six square knots.2.3.2. Fabellotibial Suture with a Nonabsorbable Suture AnchorThe second group (GFT) underwent fabellotibial suture treatment with a 2.8 mm × 11.7 mm threaded nonabsorbable suture anchor (FASTak, Arthrex Vet Systems) preloaded with #2 braided nonabsorbable multistrand polyethylene suture (FiberWire; Arthrex Vet Systems). The hole for anchor placement in the caudal aspect of the lateral femoral condyle was drilled using a 2.0 mm drill bit (Arthrex Vet Systems). Further suture placement and tightening were performed as described for the GFW group.2.3.3. Fabellotibial Suture with an Absorbable Suture AnchorThe third group (GWD) received a 2.3 × 7.2 mm polylactide absorbable bone anchor (Weldix; VetWelding AG) which can be preloaded with different types of suture materials from sizes USP#4-0 to #2. For anchor placement, an ultrasound device (BoneWelder Vet; VetWelding AG) with a frequency of 30 kHz was used.During implantation, ultrasonic vibrations established shear forces at the anchor–bone interface, which cause the polymer to liquefy and flow into the surrounding cancellous bone structure. The polymer immediately solidifies, creating a stable bond between implant and bone. This is also the case for the used suture material, which is locked in place during this process.In our study, the anchor was preloaded with #2 braided nonabsorbable multistrand polyethylene suture (FiberWire; Arthrex Vet Systems) (Figure 2). A 1.8 mm twist drill with a drill stop (VetWelding AG) was used to create the insertion site in the caudal aspect of the lateral femoral condyle.Further suture placement and tightening were performed as described for the GFW and GFT groups.2.4. Mechanical TestingAll hindlimbs (n = 36) were tested in the same setup with intact (Gi) and transected (GT) CrCL, including after surgical repair (GFW, GFT, and GWD groups).The hindlimb was placed in the limb press, and the paw was secured with a small K-wire. Throughout the experiment, the hindlimb was sprayed with saline to keep it moist. Next, 10%, 20%, and 30% of BM was applied to the hindlimb by sliding the top plate up and down and tightening or loosening the turnbuckle in the Achilles tendon mechanism. The angles of the stifle and talocrural joints did not exceed 120 ± 5° [14,16,18]. Radiographs were obtained for verification (Gierth HF 80/15, X-ray tube; Riesa, Germany) and immediately developed (Kodak Point-of-Care CR-360 System; Carestream Health Inc., Rochester, NY, USA). Distances were measured using the DICOM viewer (dicomPACS view, version 5.2.11; Oehm and Rehbein GmbH, Rostock, Germany), and joint angles were measured as previously described (Figure 3) [14,16,18]. BM and percentage of BM for each cat can be found in the Supplementary Materials (Tables S1–S26).2.4.1. Patellar Tendon LoadThe PTL reflects the tension created within the quadriceps mechanism. It was evaluated thrice at each BM load with an intact CrCL with the value read directly from the force gauge, and the mean was calculated and rounded to the nearest whole number. This value was used for all further testing in both hindlimbs of the same cadaver, guaranteeing the same tension within the quadriceps mechanism, even when altering the top plate or the Achilles tendon mechanism. In cases where the PTL changed when altering the entire mechanism to apply the calculated weights, it was corrected again.PTL for each cat and each BM load are listed in the Supplementary Materials (Tables S1–S26).2.4.2. Measurement ProcedureThe hindlimb was placed in the limb press, and the described body mass and evaluated PTL were applied. If the joint angles evaluated on the radiographs were within the predefined range, we measured the distance between the most cranial aspect of the medial tibial condyle contributing to the articular surface [35] and the center of the circle (r = 6 mm) superimposed over the caudal aspect of the femoral condyles. Subsequently, 3D measurements for later angle calculation were performed using a coordinate-measuring machine (CMM) (Microscribe M; Ravware Inc., Raleigh, NC, USA). The predefined femoral points measured were the proximal aspect of the trochlear groove (F1) and the lateral aspect of the lateral fabella (F2). A point 1 cm distal to the tibial crest (T1) and the fibular head (T2) represented the points on the lower leg. Each point was measured thrice, and the mean was calculated. After first collecting measurements from a hindlimb with an intact cranial CrCL, the cranial CrCL was then transected with a #11 blade, the arthrotomy was closed with interrupted sutures using a 2/0 monofile absorbable suture material (Biosyn; Covidien, Dublin, Ireland), and the hindlimb was, again, placed in the limb press. Previous loads were applied, radiographs were taken, and the coordinates of the predefined points were measured again. After surgical repair, the hindlimb was again mounted on the limb press, radiographs were taken, and the coordinates were measured.No parts of the limb press were altered when removing a hindlimb, during the time the hindlimb was not in the apparatus, and when replacing the hindlimb.2.5. Calculations2.5.1. DistancesTwo points each on the femur and tibia were defined, and the distance (D) between them was measured on previously obtained radiographs. X1 was the most cranial point of the tibia and X2 the center of a circle (r = 6 mm) superimposed over the caudal aspect of the femoral condyles (Figure 3). If a double condylar sign with a distance exceeding 1 mm was present, the hindlimb was replaced and the measurements performed again.Differences between the distances measured in the hindlimbs under all conditions, i.e., intact (Di) and transected (DT) CrCLs and after surgical repair (DFW = fiber wire; DFT = FASTak; DWD = Weldix), were calculated to determine the cranial motion of the tibia under each condition, including DT − Di for assessing the instability gained and Di − DFW/FT/WD for evaluating the remnant instability after each treatment. A difference of 0 indicated the restoration of stability.Calculated distances for each hindlimb can be found in the Supplementary Materials (Tables S1–S26).2.5.2. AnglesTo evaluate the movement between the femur and tibia in more than one plane, the angles between the two measured coordinates were calculated using a 2-argument arctangent. Points F2 and T1 were used to evaluate the craniocaudal movement of the tibia relative to the femur in a sagittal plane, creating angle α. Points F1 and T2 were used to calculate angle β, representing the inward rotation of the tibia in the transversal plane. As the tibia moves forward and rotates internally in the CrCL-deficient stifle joint, angle α decreases and angle β increases (Figure 4 and Figure 5). The angles were calculated for intact (αi, βi) and transected (αT, βT) CrCLs and after treatment (αFW/βFW = fiber wire; αFT/βFT = FASTak; αWD/βWD = Weldix). αT − αi and βT − βi were calculated to evaluate the instability gained after transection, and αi − αFW/FT/WD and βi − βFW/FT/WD were calculated to evaluate the remnant instability after each treatment. A difference of 0 indicated the restoration of stability.Calculated angles for each hindlimb can be found in the Supplementary Materials.2.6. Statistical AnalysisPrior to the commencement of the study, the sample size was estimated using G*Power v3.1 (Heinrich-Heine-Universität Düsseldorf; Düsseldorf, Germany) with a power of 80% to select the adequate number of specimens to be tested. Statistical analyses were performed using IBM SPSS version 19 (IBM; Armonk, NY, USA). To evaluate the correlation between the technique, calculated differences, and load, a general linear model with repeated measurements was used, followed by post hoc tests using Bonferroni’s alpha correction for multiple comparisons.To evaluate the impact of the technique on the differences for each load, one-sample t-tests were used. Pearson’s correlation coefficient was used to detect the correlation between the BM and calculated differences, and the Kolmogorov–Smirnov test was used to determine whether data were normally distributed. p < 0.05 was considered to indicate statistically significant differences.3. Results3.1. DistancesThe distances measured after CrCL transection were all significantly higher than those in stifle joints with an intact CrCL and increased with higher loads. Following stabilization with braided nonabsorbable multistrand polyethylene suture (FiberWire; Arthrex Vet Systems) (GFW group) and a polylactide absorbable bone anchor (Weldix; VetWelding AG) (GWD group), the measured distances at all loads were not significantly different compared with those of stifle joints with an intact CrCL according to the one-sample t-test (Table 1).The nonabsorbable bone anchor (FASTak; Arthrex Vet Systems) preloaded with #2 braided nonabsorbable multistrand polyethylene suture (FiberWire; Arthrex Vet Systems) (GFT group) showed a statistically significant difference regarding craniocaudal movement compared with the distances measured for an intact CrCL (Table 1). There was a significant effect of load (F = 163, p < 0.01) and condition (F = 545, p < 0.01) on the distances, but we could not show an interaction between applied loads and the technique in this setting (F < 1, p = 0.71). Pearson’s correlation analysis showed no significance between the BM and distances under all conditions and at all loads.3.2. AnglesAngle α was significantly lower, while angle β was significantly higher in stifle joints with a transected CrCL than in those with an intact CrCL. At a load of 30% BM, there was a significant difference between GWD group and the Gi group regarding angle α (p = 0.04) (Table 2). The calculated angle β values showed no difference compared to the Gi group for all treatments (p > 0.05) (Table 3).The load (Falpha = 10, Fbeta = 14, p < 0.01) and condition (Falpha = 162, Fbeta = 221, p < 0.01) significantly affected angles in the generalized linear model. In addition, the load–condition interaction affected the measured angles (Falpha = 9, Fbeta = 12, p < 0.01). Pearson’s correlation showed a correlation between the body mass and the measured angle β in the transected and surgically repaired CrCL at a load of 30% BM.4. DiscussionThis study evaluated the use of a novel ultrasonically implanted absorbable polylactide bone anchor for fixation of fabellotibial suture in cats with CrCL rupture. The results partially confirmed our hypothesis, that using the fabellotibial suture technique with an ultrasonically implanted absorbable polylactide bone anchor can neutralize excessive craniocaudal movement and inward tibial rotation.With our model, we showed that the use of the novel absorbable bone anchor neutralized the craniocaudal movement of the tibia based on conventional 2D measurement on radiographs but not at a load of 30% BM when calculating the corresponding angle. By contrast, the use of a nonabsorbable bone anchor for suture fixation failed to neutralize excessive cranial movement of the tibia according to 2D measurements, but this was not the case for 3D angle calculations. Internal tibial rotation was neutralized using all the techniques at all loads according to 2D and 3D measurements.These conflicting results between measurements on radiographs representing craniocaudal movement of the tibia and angle α, which represents the same movement, lead to the suggestion that angle α may not properly reflect craniocaudal movement and may also yield false-positive results. This suspicion is supported by a recent study that suggested that a comparison between distance measurements and angle measurements using a CMM (Microscribe M; Revware Inc., Raleigh, NC, USA) is not possible, even if calculated distances are evaluated based on 3D measurements and not calculated angles [18].To determine the possibility of comparing both methods, replication studies are recommended, keeping in mind that radiographs are the source used under clinical circumstances.With regard to distance measurements, the differences between bone anchors may be due to different materials and implantation processes. The fact that the polylactide bone anchor liquefies during implantation, infiltrates bone cavities, and creates a rigid bond between bone and implant after immediate solidification might explain its superiority over the conventional screw-in anchor [20,21,28,31].Another influencing factor might be the kind of suture attachment in different anchors. In the nonabsorbable anchor, the suture is not firmly attached and can become displaced after implantation, unlike the absorbable suture. This is because during ultrasonic implantation of an absorbable polylactide anchor, the material melts over the suture and creates an inextricable connection between anchor and suture, with the anchor itself rigidly bonding to the bone.Regarding implantation, ultrasound-guided implantation of a polylactide anchor, in contrast to fabellotibial suture and nonabsorbable anchor placement, requires practice. Nevertheless, once the process of implantation is understood, the device is easy to use and does not require more time than that for conventional screw anchor placement. What needs to be considered is that the load needed for proper placement varies with the type of preloaded suture and must be higher when the suture material is braided and of a larger diameter. We experienced this during pretesting, when implantation was practiced using different suture materials. If excessive load is applied on the sonotrode during implantation, the anchor head breaks, as previously reported [19,20,21,31,36]. The same outcome can occur if the sonotrode is not placed straight on the anchor head during insertion.Another phenomenon that needs to be considered when using absorbable polylactide anchors is the so-called backflow—i.e., when liquified material bulges out of the predrilled hole and creates a small bump on the bone surface. Backflow most likely occurs when the bone is not at the optimal temperature. This occurred in five of our tested hindlimbs and was resolved by immersing them in warm water for a few minutes prior to implantation to ensure body temperature inside the bone. Despite the backflow, all implants remained macroscopically stable and regained stability with regard to internal rotation and distance measurement. Although this phenomenon is unlikely to occur in live animals and may not lead to any soft-tissue irritation, it should be considered. Furthermore, the potential for irritation or even damage to the suture from protruding anchor material is not known and should be looked at in further studies. Regarding the increase in temperature during ultrasound-assisted implantation, no cellular reaction in the area around the pin has been found [19,20,21,31,36]. With regard to healing, different studies on the femur, tibia, jaw, and spine of sheep have shown high biocompatibility, with low scores or even the absence of inflammation-related cellular reactions [19,20,21,36]. To evaluate these findings in a clinical context, studies with long-term patient follow-up for tolerability of such implants are needed.In our study, all the sutures were tied with the stifle joint held at 100°, as described in a study that reported that the tightening of lateral suture fixations results in the least loss of tension within that angle [34]. This study was performed on dogs with CrCL-deficient stifle joints, and the same angle was selected in previous studies on cats [16,18]. With regard to distance measurement in the GFT group, craniocaudal instability neutralization was not possible within our setup where measurements were performed at a joint angle of 120 ± 5°. Considering this, further studies are necessary to prove whether suture tightening at a 100° angle is optimal in cats and with different surgical techniques. A force of 20 N was selected based on previous studies [16,18,33], but care needs to be taken because, to the best of our knowledge, no study has evaluated whether this force is most accurate for suture tightening in cats, and further evaluations are needed.When interpreting our results, we were aware of the inherent limitations of an ex vivo biomechanical model. Such a model can never completely reflect the physiologic situation during hindlimb movement in live animals, especially as the functions of muscles and surrounding soft tissues in contributing to additional stability cannot be evaluated. The Achilles tendon, with its connection to the gastrocnemius, plays a major role in creating the cranial tibial thrust [37,38]. This leads to compression of the tibia against the femur and is neutralized by an intact CrCL, amongst other anatomical structures [38]. In the CrCL-deficient stifle joint, the Achilles tendon is responsible for excessive cranial translation of the tibia relative to the femur [37,38]. Even if we mimicked this anatomical structure in our experimental setting and a cranial and medial translation of the tibia was achieved within the CrCL-deficient stifle joints, the extent to which the ex vivo situation is able to reflect reality remains unclear. The contribution of creep due to repeated testing of the stifles and its possibility of altering the results should also be kept in mind.Another important limitation was that testing was performed only at specific joint angles, making it impossible to explain the behavior of each kind of treatment during the various ranges of motion the joint experiences when a cat is walking. However, angles of stifle and hock joints were chosen as they occur in the early stance phase of a walking cat. This is the point where the stifle joint angle is the widest, which results in maximum instability for the CrCL [39,40]. A further limitation was that the maximum applied load was 30% BM. This load was supposed to mimic the load applied during normal walking of a cat but would be much lower than that required to accurately mimic a jumping cat. To avoid cable breakage and fixation loosening [16], it was not possible to apply higher loads in our setup, and this needs to be considered when interpreting the results. It can be assumed that any instability already visible at a load of 30% BM will be even more obvious with higher loads, as is the case, e.g., when a cat jumps.For further clinical evaluation of sutures, it would be beneficial to perform cyclic testing of potential implant migration. At least, there was no possibility of evaluating postoperative clinical performance, including resolution of lameness, time of healing, and potential complications, including possible screw loosening or implant migration and, within that, a possible loss of stability.The bias potentially created by our knowledge of which method was used when the radiographs were analyzed and coordinate measurements were performed is also a possible limiting factor.5. ConclusionsThe absorbable polylactide bone anchor tested in combination with a braided nonabsorbable multistrand polyethylene suture has the ability to neutralize excessive craniocaudal movement and internal tibial rotation following CrCL rupture in cats as demonstrated by distance measurements in radiographs for our ex vivo model and seems to be a suitable alternative to conventional suture anchors. Nevertheless, the conflicting results between distance measurements and angle calculations as well as the limitations of the experimental setup mean that further studies on this subject are necessary along with the consideration of possible limitations when interpreting the results.

animals : an open access journal from mdpi

10.3390/ani13050905

PMC10000078

One of the limiting factors in sheep husbandry is reproductive seasonality, which is regulated by nocturnal melatonin secretion. Subcutaneous implants of this hormone have been used to modulate this seasonality. Nowadays, consumers are increasingly concerned about organic, hormone-free production. In order to adapt sheep production to these new demands, it would be of interest to replace synthetic melatonin with phytomelatonin, which is present in plants and can be included in sheep diet. In addition, if phytomelatonin comes from by-products from the food industry, a further step would be taken towards the objectives of the circular economy. Thus, the main objective of this work was to evaluate the effect of phytomelatonin-rich diets on ram sperm quality and seminal plasma composition. With this work, we found that a phytomelatonin-rich diet, including a mix of grape pulp, and pomegranate and tomato pomaces, can increase melatonin levels in seminal plasma, improve sperm viability and morphology, and protect sperm cells against oxidative damage.

The aim of this study was to evaluate the effect of a phytomelatonin-rich diet, including by-products from the food industry, on ram sperm quality and seminal plasma composition. Melatonin content in several by-products before and after in vitro ruminal and abomasal digestion was determined by HPLC-ESI-MS/MS. Finally, 20% of a mix of grape pulp with pomegranate and tomato pomaces was included in the rams’ diet, constituting the phytomelatonin-rich diet. Feeding the rams with this diet resulted in an increase in seminal plasma melatonin levels compared with the control group (commercial diet) in the third month of the study. In addition, percentages higher than those in the control group of morphologically normal viable spermatozoa with a low content of reactive oxygen species were observed from the second month onwards. However, the antioxidant effect does not seem to be exerted through the modulation of the antioxidant enzymes since the analysis of the activities of catalase, glutathione reductase and glutathione peroxidase in seminal plasma revealed no significant differences between the two experimental groups. In conclusion, this study reveals, for the first time, that a phytomelatonin-rich diet can improve seminal characteristics in rams.

1. IntroductionLivestock farming systems in the Mediterranean regions of the southern European Union countries, such as sheep husbandry, are important [1] given that they are linked with the use of semi-natural and natural areas, and involve well-adapted autochthonous breeds [2], which is the case for the Rasa Aragonesa breed [3]. However, these systems are currently threatened by economic, institutional, environmental and social factors [4]. Furthermore, one of the limiting factors is reproductive seasonality, which is regulated by nocturnal melatonin secretion [5] in the pineal gland. Melatonin is also synthesized in the male reproductive tract [6] and is present in seminal plasma [7], having direct effects on ram sperm functionality. Experiments conducted with melatonin added in vitro to ovine sperm samples have demonstrated that this hormone. modulates sperm capacitation, decreases oxidative stress and apoptosis markers [8,9,10].Regarding the in vivo effects of melatonin on small ruminants, studies in the 1980s were performed by orally administering melatonin [11,12,13]. This hormone, absorbed onto food pellets or added in saline solution, resulted in sustained elevated blood levels of melatonin for at least 7 h in ewes [11]. Moreover, Arendt et al. reported that daily oral administration allowed advancing the reproductive season in ewes [12] although the efficacy of this treatment decreased if the administration was reduced to three times a week [13]. Most of these experiments were performed on ewes, but few studies examined the effects of the oral administration of melatonin on rams. Among these, the work performed on Suffolk rams by Kusakari and Ohara revealed that melatonin feeding could increase the reproductive activity of these rams during periods of seasonal regression [14].Since the 1990s, with the development of melatonin subcutaneous implants, the use of these devices has displaced studies with oral melatonin, as they are more practical [15,16,17,18,19,20,21,22,23]. In this case, our group and others demonstrated that implants had beneficial effects on reproductive traits, sperm motility and fertilization parameters, the seminal plasma hormonal profile and the activity of some antioxidant enzymes in rams [15,16,17,18,19]. There is abundant research into the effects of melatonin implants on the semen of other ruminants [20,21,22,23].However, in order to adapt sheep production to the new demands of consumers, who are increasingly concerned about organic, hormone-free production, it would be of interest to replace the synthetic melatonin present in implants with other natural sources of melatonin such as phytomelatonin, which is the melatonin present in plants and which can be administered with the diet.Plants contain phytomelatonin in highly variable concentrations [24]. The highest levels have been found in seeds such as mustard (Brassica nigra and Brassica hirta, 129 and 189 ng/g of dry matter, respectively), goji (Lycium barbarum, 103 ng/g), fenugreek (Trigonella foenum-graecum, 43 ng/g), almond (Prunus amygdalus, 39 ng/g), sunflower (Helianthus annuuss, 29 ng/g), fennel (Foeniculum vulgare, 28 mg/g) and alfalfa (Medicago sativum, 16 ng/g) [25], and also in fruits, such as tomato (Solanum lycopersicum, 2–114 ng/g, depending on the variety, the fraction and the method of analysis), cherry (Prunus avium, 8–120 ng/g) and grapes (Vitis vinifera, 5–96 ng/g) [24].Modulating blood melatonin levels in mammals through the intake of these products has become a strategy of great interest [26]. In fact, several studies have shown that the consumption of certain melatonin-rich vegetables, seeds and plant products increases the levels of this hormone in blood [27,28]. Given that melatonin readily crosses physiological barriers such as the blood–testis barrier [29], we hypothesized that a phytomelatonin-rich diet could have a beneficial effect on ram semen. In the present study, by-products from food industry derivatives were included in the diet in order to achieve circular economy objectives. Specifically, our aims were to evaluate the effect of the phytomelatonin-rich diet on melatonin concentration and the activity of the antioxidant enzymes in seminal plasma, and on sperm quality of rams.2. Materials and MethodsUnless otherwise stated, all reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA).2.1. Determination of Melatonin Content in Vegetables and Residues after In Vitro DigestionThe melatonin content in several agri-food by-products was analyzed to choose which could potentially be used as a supplement in the animals’ diet. The analyzed products were by-products of the juice (pomegranate pomace and peels), the canning (tomato pomace), the brewing (brewer’s spent grain, malt sprouts and spent yeast) and the wine (grape pulp and grape seeds) industries. Another product with reported high melatonin content, sunflower meal [25,30], was included as a control.Melatonin was determined following the protocol published by Rebollo-Hernanz et al., 2019 [31]. Briefly, by-products were homogenized until turning them into flour. Melatonin was extracted by treating the flours with MeOH under continuous stirring at 4 °C (16 h in darkness). After centrifugation, the supernatants were filtered under vacuum and dried using N2. The residues were resuspended in Milli-Q water and melatonin was isolated using solid-phase extraction (SPE, cartridge C-18, Waters) and measured by HPLC-ESI-MS/MS triple quadrupole. Melatonin content was expressed as ng g−1 sample. From each sample, triplicate extractions were made and each one was injected twice into the column.Melatonin content in the feed residues remaining after in vitro studies or ruminal and abomasal digestion was estimated by the same protocol.2.2. Analysis of Chemical Composition of the By-ProductsThe by-products selected on the basis of their melatonin content were: grape pulp, tomato pomace, pomegranate peels and pomegranate pomace, to which sunflower meal was added as control.The AOAC (Association of Official Analytical Chemists) methods [32] were used for the analysis of dry matter (DM, method 934.01), organic matter (OM, method 942.05), crude protein (CP, method 976.05) and ether extract (EE, method 2003.05) content. Concentration of neutral detergent fibre (NDFom) was analyzed as described by Mertens [33] in an Ankom 200 Fibre Analyser (Ankom Technology, New York, NY, USA), using α–amylase and sodium sulphite, with the results being expressed exclusive of residual ashes. The acid detergent fibre (ADF, method 973.18) and acid detergent lignin (ADL) were determined as described by AOAC [32] and Robertson and Van Soest [34], respectively.2.3. Analysis of In Vitro Digestibility of the By-ProductsThe four selected by-products and the sunflower meal were ground (1 mm particle size), analyzed for chemical composition and incubated in vitro in a closed batch system. Rumen fluid was obtained from four adult, rumen-cannulated ewes. Extraction procedures were approved by the Ethics Committee for Animal Experimentation of the University of Zaragoza (protocol PI48/20). The care and management of animals followed the Spanish Policy for Animal Protection RD 53/2013, which complies with EU Directive 2010/63 on the protection of animals used for experimental and other scientific purposes.On each incubation run, the rumen contents (approximately 300 mL of each animal) were sampled before feeding, filtered through a cheesecloth, mixed, and immediately transferred to the lab for incubation.Four in vitro closed batch incubation series were carried out. Incubations were run at 39 °C in a water bath for 24 h under anaerobic conditions following Theodorou et al. [35] with modifications [36]. Five gas bottles were filled with 800 mg of sample sealed in nylon bags and then 80 mL of incubation medium including rumen inoculum (0.2 of total incubation volume) were added. Three bottles without substrate were also included as blanks.Pressure produced on the bottles was measured with a HD8804 manometer fitted to a TP804 pressure gauge (DELTA OHM, Caselle di Selvazzano, Italy). Readings were corrected for the atmospheric pressure and converted to volume (mL) using a pre-established linear regression (n = 103, R2 = 0.996) and expressed per unit of incubated organic matter (OM). At the end of the 24 h incubation, 40 mL of the liquid phase and the solid residue were collected from one bottle per treatment and immediately frozen until analysis of the melatonin content. The solid residues of the remaining three bottles were washed with tap water and dried (60 °C, 48 h) to determine ruminal digestibility. Thereafter, dry residues pooled by treatment for each incubation run were digested with HCl-pepsin [37] to estimate abomasal digestibility. The melatonin concentration in the residues after ruminal and abomasal digestion was analyzed as described in Section 2.1.2.4. Animals and DietsSixteen 2-year-old Rasa Aragonesa rams were randomly assigned into two groups: eight rams were fed with 500 g of a commercial diet (Nutrifeed ovejas 800 MS®, Agroveco, Spain) and the other eight were fed with a 500 g phytomelatonin-rich diet per day for five months (from February to July, non-breeding season). The phytomelatonin-rich diet consisted of a mixture of various sources of phytomelatonin (20%), such as pomegranate and tomato pomaces and grape pulp (all of them derivatives of the agri-food industry), which was added to the commercial diet (80%). Diets were formulated to contain the same amount of protein, fat and fibre (Section 2.2). All animals were fed straw ad libitum.All rams were housed at the Experimental Farm of the University of Zaragoza (Zaragoza, Spain) and all experimental procedures were accomplished as described for the Project License PI39/17 approved by the Ethics Committee for Animal Experiments, University of Zaragoza (Spain), in accordance with the Directive 2010763/UE of the European Parliament on the of animals used for scientific purposes.2.5. Semen Collection and Seminal Plasma ExtractionEjaculates from each ram were obtained by artificial vagina prior to the beginning of the experiment (month 0, February) and every fifteen days for 5 months (month 5, July), so two complete semen analyses from all animals of each group were performed every month. After semen collection, sperm motility, morphology, membrane integrity, intracellular levels of reactive oxygen species (ROS) and phosphatidylserine (PS) inversion were assessed.Seminal plasma was obtained by centrifugation at 14,000× g for 10 min at 4 °C. The supernatant was collected and centrifuged again under the same conditions, and the recovered seminal plasma was stored at −20 °C until the analysis of the melatonin concentration and the activity of antioxidant enzymes (glutathione reductase, glutathione peroxidase and catalase).2.6. Sperm Motility AnalysisA computer-assisted sperm analysis system (CASA) was used for analyzing sperm motility (ISAS v. 1.04, Proiser S.L., Valencia, Spain). For this assessment, a dilution of each semen sample was made in a medium with the following composition: 0.25 M sucrose, 100 mM EGTA, 0.5 mM sodium phosphate, 50 mM glucose, 100 mM HEPES and 20 mM KOH. A drop of 8 μL of each diluted sample (3 × 107 cells/mL), was placed between a pre-warmed slide and a coverslip and maintained at 37 °C in a heated slide holder during analysis. Spermatozoa were recorded using a video camera (Basler A312f, Basler Vision Components, Exton, PA, USA) mounted on a microscope (Nikon Eclipse 50i, Nikon Instruments Int, Tokyo, Japan) equipped with a 10× negative-phase contrast lens. The recording was performed at 25 frames/s and 25 consecutive digitalized images were taken for a single field. Five fields of each drop were recorded, and percentages of total motile and progressive motile spermatozoa in all samples were evaluated.2.7. Sperm Morphological StudyThe sperm morphology was evaluated by eosin-nigrosine staining [38]. A volume of 10 µL of each dye was added to 20 μL of each sample (4 × 107 cells/mL). After mixing, a drop of the stained sample (20 μL) was placed on a slide and spread with the aid of another. The smears were air-dried and observed by bright field microscopy using a Nikon Eclipse E-400 microscope (Kanagawa, Yokohama, Japan). At least 200 spermatozoa were analyzed at 1000× magnification, and the percentage of normal morphology cells was evaluated, considering abnormal those cells that showed primary (detached head) and secondary (bent tail, coiled tail or proximal or distal droplet) abnormalities [38].2.8. Flow Cytometry AnalysesAll flow cytometry measurements were performed using a Beckman Coulter FC 500 flow cytometer (Beckman Coulter Inc., Fullerton, CA, USA) equipped with CXP software, two lasers of excitation (argon-ion laser, 488 nm; and solid-state laser, 633 nm) and five filters of absorbance (FL1-525, FL2-575, FL3-610, FL4-675 and FL5-755; ±5 nm each bandpass filter). A flow rate stabilized at 200–300 cells/s was used, and a minimum of 20,000 events were recorded in all experiments. The sperm population was gated for further analysis on the basis of its specific forward (FS) and side scatter (SS) properties and other non-sperm events were excluded.2.8.1. Evaluation of Sperm Membrane IntegrityTo determine cell membrane integrity (viability), a modification of the procedure described by Harrison and Vickers [39] was followed. Samples (500 μL; 5 × 106 cells/mL) were fixed with 3 μL formaldehyde (0.5% (v/v) in water) and stained with 3 μL of 10 μM carboxyfluorescein diacetate (CFDA) and 3 μL of 7.3 μM propidium iodide (PI). After incubation (15 min, 37 °C in darkness), samples were analyzed by flow cytometry. The monitored parameters were FS log, SS log, FL1 log (CFDA) and FL4 log (PI).2.8.2. Intracellular Content of ROSROS levels were assessed by using 2′,7′-dichlorohydrofluorecein-diacetate (H2DCFDA), which is freely permeable across cell membranes [40] and converted into non-permeable and non-fluorescent 2′,7′-dichlorodihydrofluorescein (H2DCF) by intracellular esterases. The H2DCF is oxidized by H2O2 to dichlorofluorescein (DCF), which emits fluorescence at 530 nm in response to 488 nm excitation [41]. This probe was combined with PI to exclude the nonviable population from the analysis [42]. For the assessment, sperm samples (final concentration 5 × 106 cells/mL) were stained with 5 μL of 10 μM H2DCFDA, and 3 μL of 1.5 mM PI. After 15 min of incubation (37 °C in darkness), samples were fixed with 5 μL formaldehyde (0.5% (v/v) in water) and analyzed. The monitored parameters were FS log, SS log, FL1 log (H2DCFDA) and FL4 log (PI).2.8.3. Detection of Membrane Phosphatidylserine TranslocationAnnexin V is a protein with a high affinity for phosphatidylserine (PS); hence, it can be used as a sensitive probe to detect PS exposure upon the cell membrane [43]. For the analysis, aliquots of 50 μL of sperm samples were diluted (final concentration 4 × 106 cells/mL) with 250 μL of 1X binding buffer (provided in the commercial kit; Binding Buffer Apoptosis Detection Kit, Life Technologies, Carlsbad, CA, USA) and stained with 3 μL of 1.5 mM PI and 2 μL FITC-Annexin V (conjugate also provided in the kit). After 15 min of incubation (37 °C in darkness), samples were assessed by flow cytometry. The monitored parameters were FS log, SS log, FL1 log (Annexin-V) and FL4 log (PI).2.9. Melatonin Evaluation in Seminal PlasmaMelatonin concentration in ram seminal plasma was measured using a commercial competitive immunoassay (Direct saliva melatonin ELISA kit, Bühlmann Laboratories AG, Schönenbuch, Switzerland) as previously described [44]. Absorbance was measured at 450 nm on a microtiter plate reader (SPECTROstar Nano, BMG Labtech, Ortenberg, Germany).2.10. Antioxidant Enzyme Activity Assays in Seminal Plasma2.10.1. Glutathione Reductase (GRD, EC.1.6.4.2)All measurements were performed as previously described [16] and all samples were loaded in duplicate. The reaction mixture contained 501.38 mM sodium phosphate buffer at pH 7.2; 0.5 mM EDTA; 85 μM NADPH+ + H+ and 0.8 mM GSSG. 5 μL of seminal plasma were added to complete a final volume of 200 μL. The enzymatic activity was evaluated for 3 min at 340 nm with a microtiter plate reader (SPECTROstar Nano, BMG Labtech, Ortenberg, Germany).2.10.2. Glutathione Peroxidase (GPx, EC.1.11.1.9)Measurements were performed as previously described [16,45] based on a modification of the procedure described by Plagia and Valentine [46]. The reaction mixture contained 501.38 mM sodium phosphate buffer at pH 7.2; 0.5 mM EDTA; 85 μM NADPH+ + H+, 54 mUI GRD, 2 mM GSH and 1.2 mM t-BuO2H. A total of 6 μL of seminal plasma was added to complete a final volume of 200 μL. The absorbance change at 340 nm was monitored for 3 min with the microtiter plate reader (SPECTROstar Nano, BMG Labtech, Ortenberg, Germany).2.10.3. Catalase (CAT, EC. 1.11.1.6)For catalase evaluation, the reaction mixture contained 62.5 mM sodium phosphate buffer at a pH 7, 200 mM H2O2 and 4 μL of seminal plasma to complete a final volume of 200 μL. The absorbance change at 240 nm was monitored for 120 s with the microtiter plate reader (SPECTROstar Nano, BMG Labtech, Ortenberg, Germany). For these measurements, a quartz microplate was used.2.11. Statistical AnalysesResults of in vitro digestion were analyzed statistically by ANOVA with the Statistix 10 package (Analytical Software. Statistix 10 for Windows; Analytical Software: Tallahassee, FL, USA, 2010), considering the incubation run as a block. Treatment differences among means with p < 0.05 were accepted as representing statistically significant differences. When significant, differences were contrasted by the Tukey t-test.The chemical compositions of the diets (control vs. phytomelatonin-rich) were compared by the Mann–Whitney test.Data are shown as mean ± S.E.M. (standard error of the mean). The effects of the diet and time on sperm motility, morphology, viability, intracellular levels of ROS, phosphatidylserine translocation, melatonin concentration and the activities of antioxidant enzymes in seminal plasma were analyzed using the mixed-model ANOVA and Fisher’s LSD as a post hoc test. Statistical analyses were performed with SPSS Statistics v.26 (IBM Analytic, Armonk, NY, USA) and GraphPad Prism v.8 (La Jolla, CA, USA).3. Results3.1. Melatonin Content in VegetablesAccording to the HPLC-ESI-MS/MS analysis, the by-products containing the highest levels of phytomelatonin were pomegranate pomace and peels, tomato pomace and grape pulp (Table 1). We therefore selected these by-products for subsequent analysis, together with the sunflower meal.3.2. Melatonin Content after In Vitro Digestion and Diet CompositionThe chemical composition of the selected by-products is summarized in Table S1 (Supplementary material). After incubation of these by-products in ruminal liquid and the treatment of residues with HCl-pepsin, the melatonin content in the obtained fractions was measured, and the results are shown in Table 2.The in vitro ruminal and abomasal substrate digestibility and rumen in vitro fermentation pattern of these four by-products and also the sunflower meal are summarized in Table S2 and Figure S1 (Supplementary Material), respectively.Based on these results, for the in vivo experiment, we selected pomegranate and tomato pomaces and grape pulp mixed in equal proportions. We opted for the pomegranate pomace instead of the pomegranate peels since the rumen fermentation was better (Figure S1), as well as the acceptability by the animals. The mixture accounted for 20% of the daily ration ingested by the treated animals. The chemical analysis of the phytomelatonin-rich and commercial (control) diets (Table S3, Supplementary Material) revealed that the differences in CP, EE and fibre were not significant (p < 0.05).3.3. Effects of Phytomelatonin-Rich Diets on Seminal Plasma3.3.1. Effect on Melatonin Levels in Seminal PlasmaMelatonin levels in seminal plasma from rams fed with the phytomelatonin-rich diet showed an increase from month 2, which was statistically significant (p < 0.05) in month 3 when compared with rams from the control group. Afterwards, melatonin levels tended to decrease (month 4) and then rise (month 5) both in rams fed with phytomelatonin and rams fed with the commercial diet (Figure 1).3.3.2. Effect on Antioxidant Enzymes Activity in Seminal PlasmaThere were no statistical differences in the activity of catalase, glutathione reductase and glutathione peroxidase between the two experimental groups throughout the duration of the study (Figure 2a, Figure 2b and Figure 2c, respectively).3.4. Effects of Phytomelatonin-Rich Diets on Sperm Quality3.4.1. Effect on Sperm MotilityFor total motility, there was a significant effect of the diet (p < 0.01). Percentages of total motile sperm cells in samples from animals fed with the phytomelatonin diet tend to be higher than in the control group samples (Figure 3), and the post hoc test showed that this increment was statistically significant (p < 0.01) just in the first month after diet administration. However, there was no significant effect of diet on progressive motility.3.4.2. Effect on Sperm MorphologySignificant effects of the administered diet were observed on the percentages of morphologically normal spermatozoa (p < 0.0001). After month 2, rams fed with the phytomelatonin-rich diet showed higher percentages of morphologically normal cells than the control group until the end of the experiment (Figure 4), with the maximum differences reached at months 2 and 3 (p < 0.001).3.4.3. Effect on Sperm Viability, ROS Levels and PS TranslocationRegarding the percentages of viable, viable with low levels of ROS and viable with no PS translocation spermatozoa, the phytomelatonin-rich diet had a significant effect on sperm viability (p < 0.001), and on the content of ROS (p = 0.01).From the second month of the study, sperm viability was significantly higher in rams fed with the phytomelatonin-rich diet than in the control group (p < 0.01). These differences were also found until the end of the study, especially in the third and the fifth months (p < 0.001, Figure 5a). A similar pattern was observed when analyzing the production of ROS, since rams fed with the phytomelatonin-rich diet had significantly higher percentages of viable sperm with low ROS levels than the control group from month 2, this being more marked at month 3 (p < 0.001, Figure 5b). Nevertheless, this effect was not observed when studying the phosphatidylserine inversion, as there were no statistical differences in the percentages of viable spermatozoa without PS translocation between the two groups during the experiment (Figure 5c).4. DiscussionResults of the present experiment have revealed that rams fed with a phytomelatonin-rich diet experience an increase in the content of melatonin in their seminal plasma, improving sperm viability and morphology, and protecting sperm cells against oxidative damage. In previous studies, the beneficial effects of melatonin subcutaneous implants on testis size, ejaculate volume, sperm concentration, motility and morphology, among others, have been reported [15,17,18]. Likewise, melatonin implants increased melatonin levels in seminal plasma during the non-reproductive season [16,19]. However, as consumers are increasingly aware of animal welfare and organic production, it is of interest to adapt animal production to new consumer demands. Replacing the synthetic melatonin used in subcutaneous implants with other natural sources of melatonin, such as phytomelatonin (melatonin from plants), could be a good option, provided that it does not adversely affect the nutritive feed value. In addition, obtaining these sources from by-products of the agri-food industry is consistent with the objectives of the circular economy.In accordance with the existing bibliography [24,25,26,47] and having regard to the availability of by-products from the food industries in our region, we selected by-products from the wine, beer, juice and canning industries. Analysis of these products by HPLC-ESI-MS/MS revealed that pomegranate pomace and peels, grape pulp and tomato pomace had the highest melatonin content. After evaluating the chemical composition, the in vitro digestibility and the melatonin remaining after in vitro ruminal and abomasal digestion, we selected pomegranate pomace, grape pulp and tomato pomace as the ingredients for a phytomelatonin-rich diet. Pomegranate peels showed higher melatonin content after in vitro digestion than pomegranate pomace but a worse rumen fermentation pattern, digestibility, and acceptability by the animals. Sunflower meal could also have been a good alternative, but in this study it was included only as a control since the work focused only on by-products.Feeding the rams with a phytomelatonin-rich diet containing a 20% proportion of a mix of by-products (tomato pomace, pomegranate pomace and grape pulp) resulted in an increase in seminal plasma melatonin levels compared with the control group in the third month of the study. However, the use of melatonin subcutaneous implants led to an earlier, higher increase in seminal plasma melatonin levels than that resulting from administering the phytomelatonin-rich diet [16,19]. This could be explained because the quantity of melatonin that can be administered through subcutaneous implants is much higher than the quantity found in the agri-food by-products used in the diet (18 mg in melatonin subcutaneous implants vs. 23.76 ± 1.37, 45.94 ± 4.19 and 35.81 ± 0.40 ng/g for tomato pomace, grape pulp and pomegranate pomace, respectively).Regarding the possible effects of the diet on sperm quality, parameters including sperm motility, morphology, membrane integrity, intracellular ROS levels and PS translocation were evaluated. We observed that total motility tends to be higher in semen collected from animals fed with the phytomelatonin-rich diet than in the control group samples, although this increase was only significant in the first month after administering the diet. This finding is in agreement with results previously described for Black Racka rams treated with subcutaneous melatonin implants during the non-breeding season, since treated animals from this breed showed better total sperm motility rates than the control group [17]. However, feeding a phytomelatonin-rich diet had no significant effect on progressive motility, contrary to what was described when melatonin implants were used in rams of the same [48] or other breeds [15,17].The phytomelatonin-rich diet led to higher percentages of morphologically normal and viable spermatozoa than in the control group from the second month to the end of the experiment. Beneficial effects on morphology were also reported in some studies using melatonin implants [15] but not others [17]. In the case of viability, no changes were reported after the use of melatonin implants in rams [15], but beneficial effects were described in other species, including bull [23], buck [22] and buffalo [21]. In ovine, improvement in sperm viability was observed when melatonin was added directly to spermatozoa in vitro [49,50]. However, in this work, we describe, for the first time, how melatonin supplemented through diet can enhance the percentage of morphologically normal and membrane-intact sperm cells.A similar pattern occurred with the percentage of live sperm cells with low content of reactive oxygen species (ROS). It has been demonstrated that in vitro incubation of sperm samples with melatonin can reduce oxidative stress damage caused by ROS in human and boar spermatozoa [50,51]. Recently, our research group has documented that melatonin, when added to sperm samples in capacitating conditions, can reduce ROS levels and partially prevent sperm capacitation [9]. However, in the present study, no significant differences in the percentages of viable spermatozoa without inversion of phosphatidylserine were observed between the samples from the two groups. Although in vitro melatonin has an antiapoptotic effect on sperm from different species [51,52], including ovine [49], its oral administration through the assayed diet does not seem to influence this parameter in rams.According to the data presented, we can infer that a diet rich in phytomelatonin can protect sperm cells against oxidative damage by decreasing intracellular ROS levels. This antioxidant effect could be exerted directly or indirectly by influencing the activities of antioxidant enzymes [53]. In the present study, the analysis of the activities of catalase (CAT), glutathione reductase (GRD) and glutathione peroxidase (GPX) in seminal plasma revealed no significant differences between the two experimental groups. Although melatonin implants modulate the activity of GPX and GRD, but not CAT, in rams [16] and bucks [20], the tested phytomelatonin-rich diet seems not to modify the activities of these antioxidant enzymes, probably because the amount of exogenous melatonin provided with the diet is much smaller than that produced by implants [16]. As no changes in the activities of the aforementioned antioxidant enzymes were detected, we hypothesized that phytomelatonin could modulate the levels of reactive oxygen species directly in the spermatozoa, either by crossing the sperm plasma membrane or by binding to its specific membrane receptors. Nevertheless, the molecular mechanisms by which the phytomelatonin-rich diet exerts this antioxidant action on spermatozoa remain unknown. Furthermore, we cannot overlook the fact that other antioxidants present in the by-products used in the diet may contribute synergistically with phytomelatonin to produce this effect. For instance, polyphenols and anthocyanins present in pomegranate have been shown to have powerful antioxidant activity [54], as well as lycopene, ascorbic acid, vitamin E and flavonoids present in tomato [55], or catechins, flavonols, benzoic acid and cinnamic acid present in grape pulp [56].5. ConclusionsThis study reveals, for the first time, that a phytomelatonin-rich diet increases melatonin levels in seminal plasma, improves sperm viability and morphology, and protects sperm cells against oxidative damage by decreasing intracellular ROS levels; all these effects occurring, at the latest, three months after the beginning of the feeding.

animals : an open access journal from mdpi

10.3390/ani13081284

PMC10135235

Fowl typhoid is an infectious bacterial disease of poultry that causes significant economic losses in the poultry sector. Antimicrobial therapy, biosecurity practices and vaccination programs are used to prevent and treat fowl typhoid. Misuse of antibiotics generally results in emergence of multidrug resistance in Salmonella Gallinarum. Although vaccination reduces flock losses, disease is still prevailing in many developing countries including Pakistan. To develop an alternative to these approaches, the present study was designed to explore Lactobacillus spp. of poultry origin which can inhibit Salmonella Gallinarum in vitro. Twenty one lactobacilli isolated from poultry have antibacterial activity against Salmonella Gallinarum. These isolates were further selected for the characterization of in vitro probiotic properties. Out of 21, two Limosilactobacillus fermentum strains isolated in this study have the ability to survive physicochemical barriers of the gut such as tolerance to low pH, bile salts, auto-aggregation and co-aggregation activity with Salmonella Gallinarum, and caused significant reduction in growth of Salmonella Gallinarum. These isolates may be used for the development of anti-Salmonella Gallinarum probiotics after in vivo study in broilers.

Fowl typhoid, a septicaemic disease of poultry, is caused by Salmonella Gallinarum and leads to severe economic losses. The aim of the present study was to isolate, select and characterize indigenous probiotic lactobacilli with anti-Salmonella Gallinarum activity. A total 55 lactobacilli were isolated from the caeca and ileum parts of healthy chickens and identified to species level by 16S rDNA sequencing. All the isolates were initially screened for antimicrobial activity and selected isolates were further subjected to in vitro evaluation of probiotic properties. Lactobacilli isolates (n = 21) showed varying levels of activity (08–18 mm) against Salmonella Gallinarum. These selected isolates also showed tolerance to acidic conditions (pH 3 and 4). Out of these 21 isolates, 13 showed growth (>0.5 OD at 600 nm) 0.3% bile salts. Moreover, these isolates also had the ability of auto-aggregation (20.05 ± 0.62%–50.70 ± 1.40%), and co-aggregation with Salmonella Gallinarum (5.22 ± 0.21%–42.07 ± 0.70%). Results revealed that lactobacilli had a higher level of resistance to vancomycin (100%), streptomycin (100%), ciprofloxacin (95%), gentamicin (90%), doxycycline (90%), oxytetracycline (85%), and bacitracin (80%), and a lower level of resistance to penicillin (33%), erythromycin (28%), chloramphenicol (23%), fusidic acid (23%) and amoxicillin (4%). The Limosilactobacillus fermentum PC-10 and PC-76 were sensitive to most of the antibiotics. The overall results revealed that two Limosilactobacillus fermentum strains (PC-10 and PC-76) fulfill the in vitro selection criteria of probiotics, i.e, tolerance to low pH, resistance to bile salts, auto-aggregation, co-aggregation with Salmonella Gallinarum, and absence of acquired antibiotic resistance. The Limosilactobacillus fermentum PC-10 and PC-76 also inhibited the (>5 log10) growth of Salmonella Gallinarum in co-culture assay. It is concluded that Limosilactobacillus fermentum PC-10 and PC-76 may be further investigated and developed as anti-Salmonella Gallinarum probiotics for poultry.

1. IntroductionThe poultry industry is one of the most dynamic sectors of livestock in Pakistan. It is developing at an impressive growth rate of 7.5 percent during the last few years and provides employment to about 1.5 million people. Poultry meat contributes more than 40% of the total meat production in Pakistan [1]. The poultry sector is still facing the problem of fowl typhoid, which is a severe septicaemic infection of poultry caused by Salmonella enterica serovar Gallinarum. However, this disease has been eradicated from commercial poultry in many developed countries such as Japan, Australia and Canada, and others in Western Europe and North America [2]. In developing countries, including Pakistan, fowl typhoid is endemic and found in both growing and adult chickens [3]. It can cause 10–90% mortality in chickens which eventually leads to massive economic losses in the poultry sector [4].In the poultry sector, various approaches such as antibiotic therapy, biosecurity plans and vaccines are currently being used for the prevention and control of fowl typhoid [5]. Vaccination is one of the most effective and practical strategies to control this disease [6]. The killed vaccines produce negligible cell-mediated immune response and therefore cannot be used in place of the live vaccines against Salmonella Gallinarum. The live vaccine developed from the attenuation of Salmonella Gallinarum strain SG9R is widely used for the control of fowl typhoid. The continuous impediment of using live vaccine is the possibility of reversion of the SG9R to pathogenic phenotype due to a point mutation in the rfaJ gene. Moreover, the intact plasmid of Salmonella Gallinarum SG9R also harbors various virulence genes that can cause fowl typhoid in chicks at a young stage or in immunocompromised brown-egg-laying hens [7].The extensive misuse of antibiotics in chicken has led to the development and transmission of antibiotic resistance in bacteria such as extended spectrum beta lactams (ESBL), which is one of the major problems for both poultry industry and public health [5]. Emergence of antibiotic-resistant strains and restrictions on antibiotic use in poultry have compelled scientists to find alternate approaches such as probiotics in order to increase production and also to improve disease resistance [8].According to the Food and Agriculture Organization (FAO), probiotics are defined as ‘live microorganisms which when given in appropriate amounts provide health advantage to the host [9]. Lactobacillus is one of the genera most commonly used as probiotics to improve intestinal health and production in poultry [10]. Lactobacilli are common inhabitants of the gut of animals and have acquired Generally Recognized as Safe (GRAS) status. Lactobacilli can inhibit the growth of enteric pathogens by producing organic acids, hydrogen peroxide and bacteriocins. Production of organic acids lowers the pH in the microenvironment of the gut and makes it unsuitable for most pathogenic bacteria, such as Salmonella. The bacteriostatic proteins, such as bacteriocins, synthesized by lactobacilli also inhibit the growth of many pathogens. In addition, most of these bacteriocins are pH-stable and are not affected by fluctuation in acidity in different parts of the gut [11]. Since the probiotic properties of lactobacilli are strain-specific and novel environments may have novel probiotic strains, the current study was designed as the first step of a multistep project to isolate and characterize the lactobacilli from caeca and ileum of indigenous poultry. The isolates were screened for their antibacterial effect against Salmonella Gallinarum and other desirable probiotic traits, i.e., tolerance to low pH and bile salts activity, absence of acquired antibiotic resistance, auto-aggregation, co-aggregation, and inhibition of Salmonella Gallinarum in broth culture in an effort to develop indigenous probiotics for poultry.2. Materials and Methods2.1. Sample CollectionSamples including liver, spleen and caeca were collected from backyard (Golden Misri) and commercial poultry (broiler) in different areas of Punjab Province, i.e., Lahore, Arifwala, Pakpattan and Okara where outbreak of fowl typhoid was reported. Postmortem examination of dead and sick birds suspected of fowl typhoid was conducted and infected liver, spleen and caeca organs were collected aseptically in sterile containers. Samples were kept in an icebox to maintain the cold chain during transportation to the Institute of Microbiology, University of Veterinary and Animal Sciences (UVAS), Lahore.2.2. Isolation of Salmonella GallinarumThe samples (1 gm) were added to freshly prepared 5 mL selenite broth tubes for the enrichment of Salmonella followed by incubation under aerobic conditions at 37 °C for 24 h. The liver and spleen tissues were homogenized before enrichment. After the enrichment, contents (1 mL) from the selenite broth tube were cultured on Salmonella Shigella (SS) agar. After incubation of SS agar plates, presumptive black-colored bacterial colonies were selected from SS agar plates and were subcultured for purification [12]. The isolates were identified by microscopic (Gram staining) and biochemical characterization (motility test, indole production test, Voges–Proskauer test, catalase test, H2S production test, methyl red test, oxidase test, citrate utilization test and urease test) following the Bergey’s Manual of Determinative Bacteriology [13,14].2.3. Identification of Salmonella GenusThe Salmonellae genus was confirmed by polymerase chain reaction (PCR) using a 30-bp forward primer (F-5′CGGAACGTTATTTGCGCCATGCTGAGGTAG3′) and a 27-bp reverse primer (R-5′GCATGGATCCCCGCCGGCGAGATTGTG3′) targeting the hilA gene [15]. The Salmonella Gallinarum was confirmed by amplifying the glgC gene using the serovar specific primers (F-5′GATCTGCTGCCAGCTCAA-3′) and (R-5′GCGCCCTTTTCAAAACATA3′) as described previously [16]. The 25 µL PCR reaction mixtures contained nuclease-free water (7.5 µL), 12.5 µL of DreamTaq Green PCR master mix 2× (Thermo Scientific, Waltham, MA, USA) and 1.5 µL of each of the forward and reverse primers (10 µM) and template DNA (2 µL). The amplification was performed by the following program: 1 cycle of 94 °C for 5 min, followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 30 s, and a final extension at 72 °C for 7 min [16].2.4. Isolation and Preliminary Identification of LactobacilliSamples of caeca (n = 30) and ileum (n = 30) of healthy backyard (Golden Misri) poultry from flocks having an outbreak of fowl typhoid were collected in a sterile container from different areas of Punjab. The contents of caeca and ileum were transferred aseptically into normal saline tubes and 10-fold serial dilutions were made. The lactobacilli were cultured on De Man Rogosa Sharpe (MRS) media plates supplemented with fluconazole (100 µg/mL). Plates were incubated under anaerobic conditions at 37 °C for 48 h. Colonies with different morphological characteristics were selected from the culture plates for Gram staining and catalase test. After confirmation, selected colonies were purified by subculturing two or three times on MRS agar and then stored in MRS broth (Hi Media) with 30% glycerol added [17]. 2.5. Identification of Lactobacilli by Molecular MethodThe isolates identified as Gram-positive rods and catalase negative were confirmed by Lactobacillus genus-specific PCR. The DNAs of all the isolates were extracted by a commercial DNA extraction kit (Fair Biotech) following the instructions of the manufacturer (Fair Biotech, Taiwan, China). The isolates were identified to genus level by genus-specific PCR using forward primer XB5 (5′GCCTTGTACACACCGCCCGT′3) and reverse primer LBLMA-I (5′CTCAAAACTAAACAAAGTTC′3), and to species level by amplifying the 16S rRNA gene using universal primers (8FLP and XB4) as described previously [18]. Briefly, the reaction mixture (25 µL) contained 12.5 µL of 2× master mix, 1.5 µL of 10 µM of respective primers, 7.5 µL of nuclease-free water and template DNA (2 µL). The amplification was done using Simpliamp Thermal Cycler (Applied Biosystems, Waltham, MA, USA) using the following program: initial denaturation at 94 °C for 10 min followed by 35 cycles of denaturation at 94 °C for 1 min, annealing at 55 °C for 1 min, elongation at 72 °C for 1 min and a final elongation step at 72 °C for 10 min. The amplicons of genus-specific amplification (250 bp) and 16S rRNA gene (1400 bp) were resolved on 1.5% agarose gel electrophoresis and sequenced by a commercial sequencing service (Macrogen, Seoul, South Korea). The sequences were analyzed by BioEdit and species were identified by comparing the retrieved sequences with the NCBI GenBank database using the Basic Local Alignment Search Tool [19]. The sequences were also submitted to NCBI GenBank and their accession numbers were obtained.2.6. Antagonistic Effect of Lactobacilli Isolates on Salmonella GallinarumAntibacterial activity of lactobacilli was determined by well diffusion assay as described previously. Briefly, freshly grown cultures of lactobacilli in MRS broth were centrifuged at 10,000 rpm for 5 min and supernatant was collected. Cell-free supernatants (CFSs) were prepared by filtration of supernatants through a 0.22 µm syringe filter. The CFSs were divided into two parts and the pH of one part was adjusted as 7.0. The standard inoculum (0.5 McFarland) of Salmonella Gallinarum prepared in normal saline was swabbed on Mueller Hinton Agar plates with a sterile swab. The wells were made and sealed with molten agar. The CFSs (100 µL) were added in wells and plates were incubated at 37 °C for 24 h under aerobic conditions. Antibacterial activity was determined as the diameter of the zone of inhibition (mm) around the wells [20].2.7. In Vitro Determination of Probiotic Properties of Lactobacilli2.7.1. Tolerance to Low pHThe tolerance of lactobacilli to low pH was determined as described previously [21]. Briefly, exponentially grown isolates were suspended in phosphate buffer saline to prepare bacterial inoculum (containing approximately 3 × 108 CFU/mL). The 100 µL of these suspensions were added to 10mL MRS broth tubes at different pH (pH 3, 4, 7) and incubated for 90 min. After incubation, 10-fold serial dilutions were plated on MRS media plates for enumeration of lactobacilli. Tolerance to low pH was determined by comparing the number of viable cells (CFU/mL) recovered under normal conditions (pH 7) and after exposure to acidic conditions (pH 3 and 4). A mean log10 reduction of more than 0.4 (60%) at pH 3 was considered as a cut-off for tolerance to low pH.2.7.2. Resistance to Bile SaltsThe tolerance of lactobacilli to bile salts was determined as described previously by Bao et al. [20]. Briefly, 200 µL of MRS broth supplemented with different concentrations of bile salts (0.3%, 1% and 1.8%) was added on a 96-well microtiter plate. The inoculum of lactobacilli (equivalent to 1.0 MF) was prepared in MRS broth and 20 µL of the inoculum was added into respective wells of the microtiter plate. Plates were incubated at 37 °C for 24 h and growth of lactobacilli was monitored by measuring the Optical Density (O.D) at 600 nm by a spectrophotometer (Multiskan Sky, Spectrophotometer, Thermoscientific, Waltham, MA, USA) at 0 h and 24 h. An increase in 0.5 OD of the isolates after 24 h growth was considered as good tolerance and survival ability.2.7.3. Antibiotic Resistance ProfileAntibiotic susceptibility pattern of lactobacilli to different antibiotics was determined by broth microdilution method on a 96-well microtiter plate using susceptibility test media as described previously [22]. Briefly, 100 µL of LSM broth containing doubly diluted antibiotics was added to the 96-well plates. Plates were inoculated with 100 µL (105 CFU/mL) of inoculum of respective lactobacilli. Inoculum of lactobacilli was prepared by suspending the exponentially grown isolates from MRS plates into normal saline and adjusting the OD equivalent 1 McFarland and then further diluting (1:1000) it into LSM broth. Plates were incubated at 37 °C for 24 h. Minimum inhibitory concentrations (MICs) of the antibiotics were read as the lowest concentration of antibiotic which resulted in no visible growth. On the basis of MICs, the isolates were designated as resistant, intermediate or sensitive following the breakpoints and guidelines provided by the European Food Safety Authority or as described previously [18,23].2.7.4. Auto-Aggregation and Co-Aggregation AssayAuto-aggregation represents the ability of the strain to clump together and indirectly reveal its adhesion capacity to gut epithelium while co-aggregation indicates the ability to clump with other bacterial strains and exert its inhibitory effect. Auto-aggregation and co-aggregation activity of lactobacilli were analyzed by following the method as described previously [20]. Briefly, for the determination of auto-aggregation activity, freshly grown lactobacilli isolates in MRS broth were centrifuged at 6000 rpm for 15 min. The supernatants were discarded and pellets were washed and resuspended in phosphate buffer saline (OD equal to 1 McFarland) followed by incubation at 37 °C up to two hours. After incubation, optical densities were measured at 600 nm and percentage auto-aggregation capacities of the isolates were calculated by following the formula:%auto-aggregation = A0 − At/A0 × 100 where A0 represents the OD value of an isolate at 0 h and At is the OD value after incubation for a specific time.To determine the co-aggregation of lactobacilli with Salmonella Gallinarum, equal volumes (1 mL) of lactobacilli and Salmonella Gallinarum suspensions in PBS (equal to I McFarland) were mixed and incubated at 37 °C. Similarly, lactobacilli and Salmonella Gallinarum were also resuspended in PBS separately and incubated in the same conditions. Optical densities were measured at 600 nm at different time intervals (1 h and 2 h) and co-aggregation was determined by using the following formula:%Co-aggregation = (OD1 + OD2) − 2(OD3)/(OD1 + OD2) × 100 where OD1 represents the value of strain 1 (lactobacilli), OD2 is the value of strain 2 (Salmonella Gallinarum) and OD3 is the value of mixture of strain 1 and 2. The auto-aggregation and co-aggregation of the isolates were analyzed using Lactobacillus rhamnosus GG (LGG), procured from the Institute of Microbiology, UVAS, Lahore, as control2.7.5. Inhibition of Salmonella Gallinarum in Broth CultureTo study the growth kinetics of Salmonella Gallinarum and selected lactobacilli in co-culture conditions, 1 mL of each of freshly grown culture of Salmonella Gallinarum and selected Lactobacillus isolate were added to 10 mL of nutrient broth and incubated at 37 °C for 24 h. Both of the Lactobacillus isolates and Salmonella Gallinarum were enumerated at different time intervals (0, 6, 12 and 24 h) on MRS agar and salmonella shigella agar, respectively [24]. The counts were converted into mean log10 CFU/mL and log10 reduction of Salmonella Gallinarum was determined.2.8. Statistical AnalysisThe data of optical density and enumeration were presented as Mean ± Standard Deviation (SD) and statistically significant differences among isolates were determined by one-way ANOVA followed by post hoc Tukey’s Multiple Comparison Test using graph pad prism 8.0 software.3. Results3.1. Isolation and Identification of Salmonella GallinarumIn the present study, Salmonella Gallinarum (n = 5) were isolated from field outbreaks in different districts of Punjab. The black centered colonies, typical of Salmonella, were observed on SS agar after culturing the samples. The selected Salmonellae (n = 50) were small Gram-negative rods, in single or in pair form. Biochemical tests revealed that all the salmonellae were positive for catalase, oxidase, methyl red, citrate utilization and triple sugar iron test while negative to indole, urease and VP test. All isolates were confirmed by genus-specific PCR of Salmonella. Out of 50 salmonellae, five isolates were identified as non-motile and assumed to either belong to Salmonella Pullorum or Salmonella Gallinarum. All these five non-motile Salmonella were identified as Salmonella Gallinarum by serovar-specific PCR which amplified a 194 bp region of glgC gene.3.2. Isolation and Identification of Lactobacilli IsolatesPoultry caeca and ileum samples were cultured on MRS agar plates and a total of 95 isolates (catalase-negative, Gram-positive rods) were selected as presumptive lactobacilli. Out of these 95, 55 isolates were confirmed as lactobacilli by genus-specific PCR. Out of 55, 21 selected isolates were further identified to species level by sequencing their partial 16S rRNA gene or 16S rDNA-23SrDNA interspacer region. The sequencing results revealed that the selected lactobacilli were Limosilactobacillus reuteri (09), Ligilactobacillus salivarius (02), Limosilactobacillus fermentum (02), Lactobacillus crispatus (05) and Lactobacillus johnsonii (03). The GenBank accession numbers of these lactobacilli are ON819853-ON819863, ON819865, ON819867, ON819869, ON819870, ON819876, OP703611, ON920521, ON920522, ON920524 and ON920525. An evolutionary tree constructed by the neighbor joining method on the basis of 16SrRNA gene sequences of the selected isolates is given in Figure 1.3.3. Antibacterial Activity of Cell Free Supernatants of LactobacilliActivity of the CFSs of all lactobacilli isolates directly and after adjusting the pH (7.0) was determined against Salmonella Gallinarum by well diffusion assay. Out of 55, CFSs of 21 lactobacilli isolates when used directly showed activity against Salmonella Gallinarum. The CFSs of PC-76 and PC-10 showed higher antimicrobial activity (18 ± 0.5 mm and 15 ± 0.5 mm, respectively) as compared to CFSs of other isolates. Activity of CFSs of all isolates against Salmonella Gallinarum is shown in Figure 2.3.4. Tolerance to Acidic pHThe effect of acidic conditions (pH 4, 3) on the viability of selected 21 lactobacilli isolates is shown in Table 1. All isolates showed more microbial growth at pH 7 and exhibited varying levels of tolerance to acidic pH (3 and 4). The isolates PC-07 and PC-17 showed significantly higher growth rate (p < 0.05) at pH 4 (mean log10 CFU/mL 7.83 ± 0.1 and 7.78 ± 0.1, respectively) compared to other isolates, while PI-47 exhibited less growth (mean log10 CFU/mL 5.49 ± 0.1). Moreover, at pH 3, PC-07 and PC-13 showed higher growth rate (mean log10 CFU/mL 7.84 ± 0.1, 7.83 ± 0.1) while PC-65 and PI-47 exhibited lower tolerance to acidic pH, as is evidenced by a decrease in mean log10 CFU/mL 4.69 ± 0.1 and 4.84 ± 0.1, respectively. Out of 21, 14 isolates (PC-01, PC-04, PC-11, PC-15, PC-17, PC-28, PC-47, PC-55, PC-60, PC-65, PC-68, PI-80 and PI-83, PI-93) showed poor tolerance to acidity (pH 3) as marked by more than 0.4 log reduction in their counts as compared to pH 7. The isolates PC-07, PC-10, PC-12, PC-13, PC-24, PC-25 and PC-76 showed good tolerance to pH 3 (<0.4 log reduction). Remarkably, few isolates (PC-12, PC-13, PC-15, PC-76 and PI-83) showed better tolerance to pH 3 as compared to pH 4.3.5. Bile Salt Tolerance of LactobacilliThe lactobacilli isolates showed variable levels of tolerance and growth in MRS broth supplemented with different concentrations of bile salts, as shown in Figure 3. All of the isolates showed higher growth in MRS broth, followed by MRS broth supplemented with 0.3%, 1.0% and 1.8% bile salts. The isolate PC-04 showed higher growth (OD 1.32 ± 0.03) at 0.3% bile salt concentration followed by PC-10, PC-55, PC-60 (OD 1.18 ± 0.05, 1.19 ± 0.03, 1.18 ± 0.05), respectively. The PC-12 and PC-17 showed highest tolerance (OD 0.92 ± 0.03, 0.90 ± 0.03) when exposed to 1% bile salt concentration. At a higher concentration of bile salts (1.8%), lactobacilli isolate PC-04 exhibited the higher tolerance (OD 0.67 ± 0.01) and was meaningfully different (p < 0.05) among all other lactobacilli isolates. Out of 21 isolates, only 13 (PC-01, PC-04, PC-07, PC-10, PC-12, PC-17, PC-55, PC-60, PC-65, PC-76, PI-80, PI-83 and PI-93) isolates showed tolerance and growth (>0.5 OD) at 0.3% bile salts, while PC-01, PC-04, PC-10, PC-12, PC-55, PC-65, PI-83 and PI-93 also showed good growth in MRS broth supplemented with 1.8%.3.6. Auto-Aggregation and Co-Aggregation Activity of LactobacilliAuto-aggregation and co-aggregation activity of selected lactobacilli isolates is given in Table 2. The auto-aggregation activity of isolates after 2 h was in the range of 20.05 ± 0.62% to 50.70 ± 1.40%. The highest auto-aggregation (p < 0.05) was observed for PC-10 (50.70 ± 1.40%) followed by PC-11 (46.75 ± 3.12%), while the lowest auto-aggregation activity was noted for PC-01 (20.05 ± 0.62%) in comparison to control LGG isolate (39.00 ± 0.20%). All the isolates exhibited variable co-aggregation activity with Salmonella Gallinarum after 2 h (5.22 ± 0.21–42.07 ± 0.70%). The PC-76 showed significantly higher (p < 0.05) co-aggregation activity after 1 h (29.15 ± 0.72%) and 2 h (38.30 ± 0.45%) compared to other isolates and significantly lower with LGG isolate (32.30 ± 0.55, 42.07 ± 0.70%, respectively).3.7. Antibiotic Susceptibility ProfileThe MICs of different antibiotics were determined by the broth microdilution method and results were interpreted using the breakpoints recommended by the European Food Safety Authority. The results of the study revealed that lactobacilli had higher levels of resistance to vancomycin (21/21, 100%), streptomycin (21/21, 100%), ciprofloxacin (20/21, 95%), gentamicin (19/21, 90%), doxycycline (19/21, 90%), oxytetracycline (18/21, 85%) and bacitracin (17/21, 80%), and lower levels of resistance to penicillin (7/21, 33%), erythromycin (6/21, 28%), (16/21, 76%), chloramphenicol (05/21, 23%), fusidic acid (05/21, 23%) and amoxicillin (01/21, 04%). Resistance phenotypes of each isolate are given in Table 3. Limosilactobacillus fermentum PC-10 and PC-76 were sensitive to most of the antibiotics. The MICs value of different antibiotics against lactobacilli are given in Supplementary data file Table S1.3.8. Limosilactobacillus fermentum PC-10 and PC-76 Inhibit the Growth of Salmonella Gallinarum in Co-Culture AssaysLimosilactobacillus fermentum PC-10 and PC-76 were co-cultured with Salmonella Gallinarum in nutrient broth and their counts were determined at 0, 6, 12 and 24 h using respective selective media. The kinetics in terms of mean log10 CFU/mL of the PC-10, PC-76 and Salmonella Gallinarum are presented in Figure 4. Results revealed that PC-10 caused a non-significant (p > 0.05) decrease in Salmonella Gallinarum count after 6 h (mean log10 CFU/mL 8.14 ± 0.10) and a remarkable reduction (p < 0.05) after 12 and 24 h (mean log10 CFU/mL 2.39 ± 0.05 and 1.84 ± 0.01, respectively) as compared to initial counts (mean log10 CFU/mL 8.17 ± 0.15). Similarly, PC-76 also reduced the viable counts of Salmonella Gallinarum at 6, 12 and 24 h (mean log10 CFU/mL 7.90 ± 0.25, 2.20 ± 0.18, and 1.72 ± 0.10, respectively) as compared to its initial counts (mean log10 CFU/mL 8.11 ± 0.31). Following longer incubation, the Limosilactobacillus fermentum PC-10 and PC-76 reduced the pH of media to 3.7 and 3.9, respectively, causing a remarkable reduction (>6 log) in Salmonella Gallinarum counts in 24 h.4. DiscussionAntibiotics are commonly used for the control and treatment of fowl typhoid and other bacterial infections of poultry. The overuse and misuse of antibiotics has resulted in the emergence of antibiotic resistance in bacteria. Transmission of antibiotic-resistant bacteria from poultry to the human food chain is one of the major risks to public safety [25]. Therefore, it is urgent to explore alternative strategies such as probiotics for the control of bacterial infections of poultry. Bacterial species belonging to the genus Lactobacillus are the most commonly used poultry probiotics. Lactobacilli do not pose any harmful effect on the host and are Generally Recognized as Safe (GRAS). Lactobacillus spp. are natural inhabitants of green plants, fermented foods and the gastrointestinal tract of humans and animals, and can be employed in the food industry for medical and therapeutic purposes [26,27]. The naturally resident microbiota of chicken gut is comprised of a large diversity of microbes, and is an excellent source for the selection of effective probiotic strains [28].In this study, a total of 55 lactobacilli were isolated from the caeca and ileum of healthy chicken in flocks having an outbreak of fowl typhoid. The rationale behind selecting such flocks was that if a chicken remains healthy during the outbreak of fowl typhoid, it might have superior microbiota which might be explored for the selection of probiotics. Out of 55, 21 lactobacilli were selected for further analysis on the basis of their ability to inhibit Salmonella Gallinarum. Various research studies have isolated lactic acid bacteria from fermented foods, intestinal contents and droppings of poultry and analyzed their antimicrobial activity against Salmonella enteritidis [29,30,31]. The species-level identification of 21 lactobacilli selected in this study revealed that these were Limosilactobacillus reuteri, Ligilactobacillus salivarius, Limosilactobacillus fermentum, Lactobacillus crispatus and Lactobacillus johnsonii. Our findings are similar to previous studies which have also reported that lactobacilli including L. reuteri, L. johnsonii, L. crispatus, L. acidophilus, L. salivarius and L. aviaries are most common inhabitants of the poultry gut [32,33]. Inhibitory effect of probiotics against gut pathogens is considered another important characteristic in the selection of effective probiotic strains [34]. Since the neutralized CFSs of the lactobacilli reported in this study showed no inhibitory activity against the Salmonella Gallinarum, it is assumed that the activity of these isolates was because of the production of organic acids. It has also been reported previously that acid neutralization of CFSs of lactobacilli can result in loss of their antibacterial activity [29]. Other than lactobacilli, Bacillus subtilis strains have also been reported to have activity against Salmonella Gallinarum [35].Although the in vitro characterization of potentially beneficial microbes cannot entirely mimic gastrointestinal conditions, it still is considered a cheap, reliable and fast method for screening of large microbial populations for desired beneficial properties [34]. One of the essential probiotic properties of lactic acid bacteria is their viability at the acidic pH of the stomach and ability to withstand the high concentration of bile salts. These characteristics are regarded as positive indicators for the survival of bacteria in the gut [36]. This study revealed a significant loss in viability of most of the strains (14/21) after 90 min exposure to an acidic environment (pH 3 and pH 4). Only seven (33.33%) strains which showed <0.4 log or <60% reduction in viability were considered acid-tolerant strains. Similar to the findings of this study, a previous report also recovered more than 50% of lactobacilli at pH 3 [37]. It has also been reported previously that lactobacilli are less tolerant to pH 3 and pH 2 as compared to pH 4, which is in accordance with the results of the current study [30]. The concentration of bile salt in the poultry gut is in the range of 0.1–1.0% and following the standard criteria, probiotic strains should be able to survive at 0.10 to 0.30% bile salt concentrations [38,39]. Results similar to the findings of this study have also been reported previously [17,30,40]. Since the probiotics should tolerate and grow in good numbers in the gut to exert their beneficial effect, an increase of at least 0.5 in optical density at 0.3% bile salt concentration was considered a cut-off as selection criteria. Most of the isolates (13/21) were able to grow well in 0.3% bile salts, while 11 were resistant to 1.0% bile salts and eight were resistant to 1.8% bile salts.Auto-aggregation is another property of lactobacilli which depicts its ability to form a biofilm that protects the host from invading pathogens. This activity also indirectly reveals the capacity of lactobacilli to attach to the surface of intestinal epithelial cells [41]. Co-aggregation of lactobacilli with pathogens indicates their ability to attach with pathogens in vivo and create a microenvironment where their antimicrobial metabolites can inhibit the pathogens [42]. In this study, auto-aggregation activity and co-aggregation of lactobacilli with Salmonella Gallinarum were in the range of 20.05 ± 0.62%–50.70 ± 1.40% and 5.22 ± 0.21%–42.07 ± 0.70%, respectively. Another study recorded higher co-aggregation of lactic acid bacteria with pathogenic Staphylococcus aureus (10.15–38.5%) while their results showed the low co-aggregation (5.56–8.62%) activity of L. fermentum with E. coli 0157H7 [43].Various studies indicated that antibiotic resistance genes of lactobacilli can be passed to resident organisms in host GIT. So, considering the safety issue of probiotics, antibiotic resistance patterns of isolates were investigated. The Lactobacillus species are normally resistant to antibiotics that inhibit DNA synthesis (quinolones). They are sensitive to all cell wall and protein synthesis inhibitors except vancomycin and aminoglycoside [44]. A higher level of resistance to aminoglycoside and vancomycin, ciprofloxacin and tetracycline is reported in lactobacilli isolates in this study which is in accordance with the previous studies. Lactobacilli resistant to erythromycin, penicillin or tetracycline are considered a threat to public safety as the resistance to these antibiotics is generally acquired in lactobacilli. The isolates showing the acquired resistance were rejected for probiotic potential. The Limosilactobacillus fermentum PC-10 and PC-76 were sensitive to most of the antibiotics and did not contain the acquired resistance; therefore, these are safe for addition to the food chain and pose no risk of antibiotic resistance transfer. Many previous studies have also reported similar results of antibiotic resistance in lactobacilli from poultry or food chain [8,29].The co-culture assay is used to assess the antagonistic effect of one organism on the growth of another organism, when both are cultured together in broth [45]. In this study, Limosilactobacillus fermentum PC-10 and PC-76 were selected for co-culture assay on the basis of their antimicrobial activity against Salmonella Gallinarum in well diffusion assay, their probiotic potential, and absence of resistance to most of the antibiotic studies. The results revealed that both of the lactobacilli strains can significantly inhibit (>5 log reduction) the growth of Salmonella Gallinarum. Similar results have also been reported previously where strains of L. fermentum and L. gasseri inhibited the growth of Salmonella Enteritidis [29] Therefore, the isolated strains of lactobacilli in the present study, Limosilactobacillus fermentum PC-10 and PC-76, can be used to develop a probiotic product after in vivo studies in chicken. To the best of our knowledge, this is the first study in Pakistan which analyses the in vitro probiotic properties of Limosilactobacillus fermentum against Salmonella Gallinarum. Two novel strains (PC-10 and PC-76) isolated from the natural habitat of chicken have the ability to inhibit the growth of Salmonella Gallinarum and have suitable in vitro probiotic properties. These isolates can be employed in poultry after studying their beneficial effect in broilers.5. ConclusionsIt is concluded that chicken-derived Limosilactobacillus fermentum (PC-10 and PC-76) have in vitro probiotic potential and antagonistic activity against Salmonella Gallinarum. These strains may be used to develop indigenous probiotics against Salmonella Gallinarum infection after in vivo evaluation in chicken.

animals : an open access journal from mdpi

10.3390/ani12050624

PMC8909879

General anesthesia in rabbits is associated with significantly higher mortality than in dogs and cats. In addition, as prey animals, rabbits tend to mask signs of pain, making early detection particularly difficult. Loco-regional anesthesia represents a fundamental component of a multimodal approach to pain management and is an effective strategy to reduce the need for systemic anesthetic and analgesic drugs, thereby limiting their associated side effects. The saphenous nerve is the largest sensory branch of the femoral nerve and provides sensory supply to areas of the hind-limb. Saphenous nerve blockade allows for analgesia of the front and inner parts of the pelvic limb without affecting femoral nerve motor function, thus allowing optimal pain relief but permitting continued mobilization, which may be of particular importance in rabbits as it allows preservation of normal physiological prey behavior (e.g., hiding, kicking). This cadaveric study describes an ultrasound-guided saphenous nerve block technique in rabbits and compares the length of the nerves stained following injection of two different dye volumes. The results show that both volumes consistently stained the saphenous but not the femoral nerve. This technique has the potential to provide hind-limb analgesia while preserving femoral motor function in rabbits.

Ultrasound-guided (US-guided) loco-regional anesthesia techniques allow direct visualization and blockade of sensory nerves. The saphenous nerve (SN), a terminal branch of the femoral nerve (FN), is strictly a sensory nerve for which electrical locator devices are ineffective for localization as no effector muscle contractions can be evoked. US-guided SN block in species other than rabbits produces hind-limb analgesia without affecting FN motor function. The aims of this study were to develop a US-guided SN block technique in rabbits and to compare the spread obtained using two different dye volumes. Twelve hind-limbs from six cadavers (1.6 ± 0.1 kg) were included; after randomization, the SN block was performed on the right or left hind-limb, injecting 0.05 mL kg−1 or 0.1 mL kg−1 of tissue dye in lidocaine (1:50 v:v). Subsequent dissections allowed nerve staining measurements. All SNs were identified, and 17.8 ± 4.6% and 31.0 ± 8.9% of the SN length were stained using low-volume and high-volume of the dye, respectively. Regardless of the volume used, the SN was consistently stained while the motor branch of the FN was not. This US-guided technique may provide hind-limb analgesia without affecting FN motor function in rabbits undergoing mid-distal hind-limb surgeries.

1. IntroductionUnlike dogs and cats, rabbits have been domesticated for only a relatively short time; however, as their popularity as companion animals increases, there is greater demand for the provision of anesthesia in this species [1]. Unfortunately, possibly related to their peculiar anatomical, physiological, and behavioral features, there is a higher risk of peri-anesthetic death in rabbits (between 1.39% [2] to 4.8% [1]), compared to cats (0.24%) and dogs (0.17%) [2]. In most cases, the primary etiology of peri-anesthetic death is unknown but is often ascribed to either cardiovascular or respiratory causes [1,2]. Additionally, a high incidence of non-fatal gastrointestinal complications is also common in rabbits (38%) [1]. No specific risk factor has been associated with the higher peri-operative mortality rate recorded in rabbits, and this is most likely multifactorial (e.g., dietary changes, pain, disease, and medication side effects) [3].Prey animals such as rabbits are reluctant to show signs of discomfort, making them particularly challenging for early detection of pain as well as the evaluation of the efficacy of an analgesic treatment if this is based on behavioral changes [1,4,5,6,7]. Therefore, adequate peri-operative analgesia plays a vital role in the outcome [4,5,6]. While no single technique or drug regimen has been shown to eliminate peri-anesthetic morbidity and mortality in any species, a pre-emptive analgesia strategy and a multimodal approach to pain management should ensure optimal peri-operative analgesia in rabbits, reducing the risk of under-detection of pain and the occurrence of possible related complications in the peri-anesthetic period, such as ileus [1,4,5,8].Loco-regional anesthesia techniques, such as peripheral nerve blocks (PNB) and interfascial blocks, are fundamental components of a multimodal approach to pain management and are effective in reducing the need for systemically-administered analgesic and anesthetic drugs, potentially limiting the latters’ side effects [9,10]. However, in addition to preventing the transmission of nociceptive signals to the cortex, many loco-regional techniques will also result in a degree of motor block that may affect the rabbit’s normal behavior pattern. Temporary absence of physiological prey behaviors may increase hospitalization-associated stress, reduce food intake, and potentially increase the risk of stress-related complications, such as ileus [1]. Furthermore, prolonged immobility during the post-operative period secondary to impaired motor function may limit efficient ventilation, as reported in humans [11]; this may be particularly significant in rabbits, where there is a high incidence of pre-existing pulmonary disease, even in apparently healthy individuals [1].Hind-limb innervation in dogs derives from the sciatic nerve and femoral nerve (FN); the latter runs inside the iliopsoas muscle and receives contributions from the fourth, fifth, and sixth lumbar spinal cord segments (L4–L6) [12]. Before leaving the iliopsoas muscle, the FN gives origin to the saphenous nerve (SN) and is located cranially to the femoral artery (FA) in a different interfascial plane [9]. The SN runs distally, cranial to FA, and within the same fascia (medial femoral fascia), forming a so-called ‘neurovascular bundle’ [9,13,14]. The SN supplies branches to the stifle joint and ends in the skin over the first digit (when present) [14,15]. Overall, in dogs, the SN is responsible for the sensory innervation of the medial aspect of the distal thigh, stifle joint, tibia, tarsus, and metatarsus, and the cranial aspect of the stifle [13].Similarly, hind-limb sensory and motor innervation in rabbits is provided by the FN and sciatic nerve, which originate from the lumbosacral plexus (L4–L7), and the first and second sacral cord segments (S1–S2) [16,17,18].In dogs, the peripheral block of the SN has been shown to provide desensitization of the medial and cranial aspects of the distal limb (including the stifle) without affecting the motor function of the quadriceps femoris muscle [9,19]. Due to the neuro-anatomical similarities between dogs and rabbits [16,17,19,20], it is likely that this also applies to rabbits. Two previous studies reported successful blockade of FN (and sciatic nerve) using electrical stimulation in rabbits undergoing pelvic limb orthopedic procedures [17,18]. However, by blocking the FN (instead of SN), the motor function of the quadriceps, gastrocnemius, and cranial tibial muscles [16] will potentially be impaired during the post-operative period, affecting voluntary movements and possibly impacting negatively on the animal’s welfare.Electrical locator devices are ineffective in localizing sensory nerves since motor fibers are lacking in these types of nerves and muscle contractions cannot be evoked [9,19]. To perform a SN block, as well as other sensory nerve blocks, ultrasound-guided (US-guided) techniques should be applied, as direct visualization of the targeted nerve and landmarks is possible. Besides these advantages, the US-guided techniques also allow real-time visualization of the local anesthetic spread around the nerves while reducing the risk of damage to other vital structures (e.g., arteries or veins) [9,19].In dogs, the US-guided SN block technique has been designed as an interfascial block injection, as the local anesthetic is injected into the medial femoral fascia that contains the SN and both the FA and FN [9,19].Recently, a case report describing the use of a combined US-guided and nerve stimulation-guided saphenous and sciatic nerves blocks in a pet rabbit undergoing calcaneal fracture repair demonstrated the effectiveness in peri-operative pain management with no impairment of motor function in the affected limb [20].To the authors’ knowledge, there are no previous studies that have investigated the sono-anatomy of the SN and its surrounded structures in rabbits and designed a US-guided SN block technique specifically for this species. Additionally, no previous research in companion animals has investigated the resulting length of SN stained once the US-guided SN block technique is performed and if this would affect the femoral motor branch (FMB). The aims of the present study were (1) to describe the gross anatomy of the SN and surrounded structures in rabbits; (2) to describe the sono-anatomy of the SN in relation to the surrounding structures and to design a US-guided SN block technique specifically for rabbits; and (3) to compare the length of the nerves (saphenous and femoral motor branch) stained after injecting two different dye volumes (low volume, 0.05 mL/kg; high volume, 0.1 mL/kg). Our hypotheses were that the SN could be easily identified using the US-guided technique and that the spread obtained with both injected dye solution volumes would effectively stain the SN while not diffusing to the FMB, which may potentially preserve the motor function in vivo.2. Materials and MethodsThis study included a total of six cadavers of adult Californian rabbits (Oryctolagus cuniculus). No animals were euthanized for the purpose of this study, and no ethical committee approval was required as this was purely a cadaveric study. The rabbits were obtained frozen from a certified local butchery and were slowly thawed at room temperature over a period of 24 h. All cadavers were presented skinned, partially eviscerated, and exsanguinated. Cadavers that could still be frozen (based on palpation) or with damaged vertebrae, pelvic bones, nerves, muscles, or other inguinal structures were excluded.2.1. Gross Anatomical InvestigationThe first cadaver (2 hind-limbs) was used only for gross anatomical dissections in order to evaluate the SN anatomy and its relationship with the surrounding muscles and vessels in the rabbit. Both hind-limbs were dissected with the rabbit cadaver placed in dorsal recumbency and the hind-limb extended in a natural position but externally rotated around the longitudinal plane. For each hind-limb, following the identification of the superficial muscles (Figure 1), dissection was performed by removing the vastus medialis, sartorius, gracilis, and adductor muscles before visualizing the medial femoral fascia containing the neurovascular bundle. After identification of the iliopsoas muscle, the origin of the FN was identified and its path was followed along the medial aspect of the hind-limb. The FN and its branches, the SN, and motor branch (FMB), were also identified.2.2. Sono-Anatomy Study and US-Guided Saphenous Nerve Block DesignThe other five cadavers (10 hind-limbs) were used for sono-anatomy investigation of the SN and surrounding structures in order to design the US-guided SN block. For all 10 legs, the medial aspect of the hind-limb was first ultrasound screened using a veterinary dedicated ultrasound system (Fujifilm Sonosite lnc., S II Veterinary Ultrasound System, Bothell, WA, USA). Ultrasound gel (Aquasonic 100, Parker, Fairield, NJ, USA) was used to establish contact between the ultrasound probe (L25x, 13-6 MHz Linear Transducer, Bothell, WA, USA) and the muscle layers. The ultrasound probe was placed over the proximal inguinal area, transversally to the long axis of the pelvic limb, where the FN was visualized before branching into the SN and FMB. The US probe was then slid distally to the level of the middle thigh, where the acoustic window displayed the femur, the vastus medialis, adductor, pectineus, and sartorius muscles, and the neurovascular bundle (Figure 2). At this level, the terminal portion of the pectineus muscle was visualized. Based on our gross anatomical study, all these aforementioned structures were considered as landmarks to design the US-guided SN block technique. In order to consistently perform the US-guided SN block in all cadavers, the terminal portion of the pectineus muscle was then chosen as the main landmark. To better visualize the targeted acoustic window, adjustments were performed with minor ultrasound probe movements (minor rotation and minor sliding movements) until the neurovascular bundle was visualized at the center of the US window, caudally to the vastus medialis muscle and femur, cranially to the adductor and pectineus muscle, and below the sartorius muscle.After each sono-anatomy study, the hind-limb was assigned using an online randomizer generator (www.random.org, accessed on 5 February 2022) to receive either a low (0.05 mL/kg) or high (0.1 mL/kg) volume of the injectate, which comprised a mixture of a permanent tissue dye solution (Tissue marking dye yellow, Mopec, Madison Heights, MI, USA) with lidocaine 2% (Lidocaine, Hameln Pharmaceuticals, Gloucester, UK) mixed in a 1:50 v:v ratio [21].For the US-guided SN block technique, a 21-gauge, 50 mm echogenic insulated needle (Echoplex, Vygon, Swindon, UK) with an attached extension line primed with the injectate solution was inserted through the tensor fascia latae muscle in a cranio-caudal orientation towards the neurovascular bundle. The along visual axis technique (the long axis of ultrasound probe was along the operator’s visual axis and ultrasonic beam and vertical to the surface) and in-plane needling approach were used to perform the US-guided SN block with lidocaine-dye mixture injections in all cadavers [22]. The position of the tip of the needle was deemed satisfactory when it pierced the medial femoral fascia containing the SN and the FA and Femoral vein (FV). This technique was consistently applied to all cadavers for all hind-limbs by the same operator (R.F.).2.3. Anatomical Dissection following US-Guided Saphenous Nerve BlockThe rabbit cadavers were kept in dorsal recumbency and a sagittal incision was made from the inguinal crease to the medial aspect of the stifle along the center of the thigh. Careful blunt dissection was performed to remove the sartorius, gracilis, vastus medialis, and adductor magnus muscles. This allowed the exposure of the pectineus and iliopsoas muscles, the FA and FV, and the FN and its branches (motor and SN branches). Following this, blunt dissection of the FA, FV, and of the SN allowed their separation from the surrounding structures, from the inguinal crease to the proximal level of the stifle. The medial femoral fascia containing these structures was preserved to allow the evaluation of the spread limits of the dye solution and whether the SN and the femoral motor branch (FMB) were stained in the hind-limbs injected with either the low dye solution volume (0.05 mL/kg) or with the high volume (0.1 mL/kg).The SNs were considered to have been successfully stained when the dye solution was detected within the medial femoral fascia and around the entire circumference of the SN for a length of ≥1 cm [23], but not outside the fascia. The total length of the SN was considered from the bifurcation of the FN to the proximal level of the stifle joint. The nerves’ staining length was measured in all cadavers with a standard ruler.2.4. Statistical AnalysisStatistical analysis was performed using Microsoft Excel (v16.0, Microsoft Corporation, Santa Rosa, CA, USA). The parametric data were tested for normal distribution by the Shapiro-Wilk test and are presented as mean ± standard deviation (SD). The Fisher’s exact test was performed to determine statistical significance using two categories, category-1 SN stained for ≥1 but ≤2 cm, category-2 SN stained for >2 cm [12,23]. A p-value less than 0.05 was considered statistically significant.3. ResultsAll cadavers were in satisfactory condition after thawing, and no significant damage was observed; therefore, all cadavers were included in the study. Due to the cadaveric preparation, it was not possible to determine the sex or age of the rabbits; however, the mean carcase weight was 1.6 ± 0.1 kg.3.1. Gross Anatomical InvestigationIn both hind-limbs, it was possible to observe the FN giving origin to the SN and to the FMB after leaving the psoas compartment at the proximal inguinal area. At this level, the FMB ran to the anterior aspect of the thigh, where it gave origin to other branches. The SN was identified within the medial femoral fascia as part of the neurovascular bundle, cranially to the FA and FV. The SN was localized below the sartorius muscle caudally to the rectus femoris and vastus medialis muscles and cranial to the pectineus and adductor muscles. In the middle compartment of the thigh, the pectineus muscle was identified as a triangle-shaped muscle located below the neuromuscular bundle. The distal insertion of the pectineus muscle was located at the end of the first proximal third of the femur on its posteromedial surface. The femur was located cranially to the neurovascular bundle, adductor, semimembranosus, and pectineus muscle, caudally to the FMB, rectus femoris, tensor fascia latae muscles, and laterally to the vastus medialis muscle. Distally, at the proximal aspect of the stifle joint, the SN ran deeper to innervate the articular capsule (Figure 3).3.2. Sono-Anatomy Study and US-Guided Saphenous Nerve BlockDespite all cadavers being considered correctly thawed for the anatomical study (based on palpation), in one of the hind-limbs, a mild degree of ice crystallization was identified in the middle thigh muscles during the sonography. However, as this finding did not interfere with the identification process of the acoustic target window and the neurovascular bundle, this hind-limb was also included.In all hind-limbs, it was possible to identify the targeted acoustic windows at the level of the middle thigh using the landmarks as described in the anatomical study. The landmark structures (muscle, bone, and neurovascular bundle) were all visualized after mild probe movements within 1.5 cm depth for all cadavers.According to the anatomical findings, the US-guided SN block technique was designed, placing the ultrasound probe approximately at the central portion of the middle thigh transversally to the long axis of the hind-limb. At this level, the visualization of the neurovascular bundle was at the center of the acoustic window (Figure 4).All the muscles (sartorius, adductor, semimembranosus, pectineus, and vastus medialis muscles) were displayed as structures with heterogeneous echogenicity. The femur was displayed as a hyperechoic structure with acoustic shadow. At the center of the targeted acoustic window, the terminal portion of the pectineus muscle was visualized as a triangular-shaped structure with heterogeneous echogenicity caudally to the femur. In addition, the neurovascular bundle was displayed as a piriform-shaped structure with a hyperechoic outer layer (medial femoral fascia) containing three hypoechoic round-shaped structures (SN, FA, and FV). It was not possible to confirm the position of the SN in relation to both the FA and FV due to the absence of blood flow (Figure 4). The neurovascular bundle was always located below the sartorius muscle and medially to the pectineus muscle within 0.5 cm depth for all cadavers. In addition, the vastus medialis muscle was located cranially, the adductor and the semimembranosus muscles were located caudally to the neurovascular bundle (Figure 4).In all of the 10 hind-limbs, the US-guided SN block was feasible at the first attempt. The timing to successfully perform the block was around 5 min. During needling, the shaft of the needle was always visible, as was the tip when piercing the medial femoral fascia. The dye solution was injected in five hind-limbs with a low volume (0.05 mL/kg; resultant mean volume of 0.08 ± 0.006 mL) and in the other five with a high dose (0.1 mL/kg; resultant mean volume of 0.16 ± 0.009 mL). During injection, an anechoic area was formed within the medial femoral fascia, which improved the visualization of the SN for all cadavers (Figure 5a,b).3.3. Anatomical Dissection following US-Guided Saphenous Nerve BlockIn all cadavers, the dye solution was injected within the target area (medial femoral fascia) without involving the surrounding structures (Figure 6a and Figure 7a). It was possible to separate all the surrounding anatomical structures maintaining the neurovascular bundle integrity. The dye solution was visible within the medial femoral fascia containing the SN and both FA and FV in all of the 10 pelvic limbs (Figure 6b and Figure 7b). All SNs, regardless of the dye solution volume injected, were stained around their entire diameter and for more than 1 cm of length (Figure 6c,d and Figure 7c,d). In addition, none of the FNs nor FMB were stained with either high or low volumes of dye solution.Individual total SN length, SN length of stain, and percentage of the total length of the nerve stained, reported as mean and standard deviation, are described in Table 1. The difference in the tissue dye solution spread along the SN was statistically significant between the two groups (p = 0.0476; Fisher’s Exact test).4. DiscussionThe present study is the first investigating the sono-anatomy of the middle thigh area in rabbits and the first to successfully design a US-guided SN block technique specifically for this species. In addition, it has also demonstrated that the length of the saphenous nerve stained was proportional to the dye solution volumes injected, but even the lower volume successfully stained the SN in a manner that should provide adequate blockade. The technique here described allowed the staining of the SN without involving the motor component of the FN (femoral motor branch). This US-guided SN block has the potential, therefore, to produce sensory blockade without affecting the motor function of the quadriceps femoris muscle in rabbits.Based on this study, the anatomy and sono-anatomy of the medial aspect of the hind-limb and the relationship between the SN, femur, FA, FV, and muscles in rabbits are similar to that described in dogs [9,19]. However, in the rabbit, the pectineus muscle terminates on the proximal third of the femur, while in dogs, this occurs on the distal aspect [24]. In dogs, the main landmark for the US-guided SN block is the femoral artery [9,19]; however, due to the nature of this study in cadavers, other landmarks (muscles and femur) had to be used. The pectineus muscle was considered the main landmark in this study, as it allowed a consistent location to perform the US-guided SN block in all hind-limbs.This technique could be considered as an interfascial plane block because the target point was the medial femoral fascia that contains the SN, and not directly the SN itself. Due to the small size of the SN in rabbits and to the proximity of vital structures (FA and FV), this block was designed as an interfascial block to reduce the risk of intravascular injection when performed in live animals, and also, therefore, a suitable option if the nerve (SN) is not clearly visualized or distinguished from the vascular structures. Furthermore, this may facilitate the contact of the nerve surface with the local anesthetic, as previously reported in cats [23].The main landmark (pectineus muscle) and the target structures (neurovascular bundle inside the medial femoral fascia) were all within a maximal depth of 1 cm, which may have favored the visualization. Nonetheless, the small size of the SN in rabbits may make it challenging to perform the needling due to the proximity of the target structures to the ultrasound probe [25].The volumes of the dye solution used in this study were decided according to previous similar studies performed in dogs targeting peripheral nerves [12,26], but lower than the volumes used to perform other interfascial plane nerve blocks in small animals (up to 0.6 mL/kg) [27,28,29]. In fact, in interfascial plane nerve blocks, the use of larger volumes of local anesthetic allows a more consistent spread around the targeted nerves, increasing the chances of successful nerve blockade [29]. In our study, although an adequate SN staining length was obtained in both groups, it is possible that ultrasound visualization and dye solution distribution within the medial femoral fascia may improve with the use of higher volumes. However, volumes of dye solution larger than those studied may spread to the FN and FMB; further research is required to investigate this. Nevertheless, the choice of the volume of injectate used should always account for the maximum dose of the local anesthetics allowed in the species considered [30].In dogs, the use of 0.1 mL/kg of local anesthetic allowed complete sensory block of the saphenous nerve [19], while in our study, all SNs were adequately stained (complete staining around its diameter for more than 1 cm in length) [23], suggesting that this technique may result in an effective sensory block of the medial aspect of the mid-distal hind-limb even with a volume of 0.05 mL/kg. Future in vivo studies are needed to confirm these assumptions.The present study has some limitations that need to be addressed. The echogenicity of fascial planes, muscles, and vascular structures may be altered during freezing and thawing and may not accurately reflect the sono-anatomy of the live rabbit. All cadavers were skinned, which may have improved the US visualization of the target structure and landmarks but may have made needling more challenging due to the proximity of the target structures to the probe [25].The pressure of the injection was not measured during the US-guided SN block, and it is recognized that the spread of dye solutions within an interfascial plane may be unpredictable in cadavers due to differences in tissue integrity [31]. This may also influence the resistance to injection, which, if increased, could promote a wider spread of the dye solution when compared to live animals.Further studies are needed to investigate whether the injection of a larger volume of dye solution using the described US-guided SN block technique would also reach the FN and FMB, producing motor block, or if performing the US-guided SN block more distally would lead to similar results to our study. In addition, clinical studies are required to evaluate the analgesic efficacy of this technique in rabbits.5. ConclusionsIn rabbits, the US-guided SN block herein described allowed a consistent adequate stain of the saphenous nerve in all cases without affecting the motor branch of the femoral nerve, even with the lower volume of dye solution (0.05 mL/kg). Furthermore, even with the higher volume (0.1 mL/kg) of dye solution, the femoral nerve and its motor branch were not stained in any case. This US-guided SN block technique (regardless of the volumes used) has the potential to produce a sensory blockade of the mid-distal hind-limb, including the medial and cranial aspect of the stifle, with no quadriceps muscle motor function impairment. This technique has the potential to preserve the rabbits’ normal behavior pattern in the post-operative period while reducing the risk of undetected pain; however, these conclusions need to be evaluated with in vivo studies.

animals : an open access journal from mdpi

10.3390/ani11092749

PMC8472355

Macaques in captivity are prone to becoming overweight and obese, which may cause several health and welfare problems. Diet likely plays an important role herein. In an attempt to reduce overweight incidence and related health problems, a minor dietary change was implemented in our long-tailed macaque breeding colony. The provisioning of bread was replaced by grains and vegetables, while the basic diet of monkey chow remained the same. Overweight status did not differ after dietary change, but some biochemical parameters related to glycemic response and lipid metabolism improved. This study emphasizes the importance of evaluating husbandry changes and shows that relatively minor dietary adjustments may improve animal health and welfare.

Macaques in captivity are prone to becoming overweight and obese, which may cause several health problems. A diet that mimics the natural diet of macaques may prevent these problems and improve animal welfare. Adjusting captive diets towards a more natural composition may include increasing fiber content and lowering the glycemic index, i.e., reducing the impact on blood glucose levels. Such a dietary change was implemented in our long-tailed macaque (Macaca fascicularis) breeding colony. The basic diet of monkey chow pellets remained the same, while the supplementary provisioning of bread was replaced by grains and vegetables. This study is a retrospective evaluation, based on electronic health records, that investigated whether this minor dietary change had a beneficial effect on relative adiposity and overweight-related health parameters in 44 non-diabetic, group-housed, female long-tailed macaques. Relative adiposity was measured with a weight-for-height index and blood samples were collected during yearly health checks. Glycemic response and lipid metabolism were evaluated using several biochemical parameters. Relative adiposity and overweight status did not differ after dietary change. Yet, relatively heavy individuals generally lost body weight, while relatively lean individuals gained body weight, leading to a more balanced body weight dynamic. Dietary change did not affect HbA1c and triglyceride levels, while fructosamine and cholesterol levels were significantly reduced. Thus, the minor dietary change had no significant effect on overweight status, but some biochemical parameters related to the risk of diabetes and cardiovascular disease were positively affected. This study emphasizes the importance of evaluating husbandry changes and that critically reviewing husbandry practices can provide valuable insights to improve animal health and welfare.

1. IntroductionMacaques in captivity are susceptible to becoming overweight and obese. Similar to humans, this can cause several health problems, such as type 2 diabetes mellitus (T2DM) and cardiovascular disease [1,2]. Diet likely plays an important role in becoming overweight and the related health problems [3]. A diet that mimics the natural diet of macaques may prevent these problems. Wild macaques mainly eat wild fruits, supplemented with seeds, flowers, leaves, buds, bark and small animals, e.g., insects [4,5,6,7]. This natural diet is high in fiber and low in fat [8,9], resulting in little to no overweight in wild macaques [10,11]. In contrast, diets in captivity tend to be low in fiber and high in easily digestible carbohydrates, such as sugar [12]. Accordingly, 10–15% of captive macaques develop obesity during their life [13].Adjusting the diet towards a more natural composition may decrease overweight-related health problems and improve animal welfare in captivity. For example, increasing fiber and decreasing sugar content in the diet led to a reversal of prediabetes and more natural behaviour in great apes [12]. Fruits, vegetables, and grains generally contain a high amount of fiber and have a low glycemic index (GI ≤ 55) [14,15,16]. Other food items, such as bread, have a high GI (GI ≥ 70) as they contain carbohydrates that are quickly digested and metabolized [14,17]. This leads to postprandial hyperglycemia, i.e., a high increase in blood glucose after consumption, which has been proposed to increase the risk of T2DM and cardiovascular disease in humans [16].Various biochemical parameters can be used to assess the risk of developing overweight-related health problems such as TD2M and cardiovascular disease. As animals are progressing towards T2DM, the glycemic response becomes impaired and blood glucose levels increase [18]. As a result, glycated proteins are formed, e.g., fructosamine and glycated hemoglobin (HbA1c), both accurate biomarkers to measure the intermediate and long-term glycemic response, respectively [19,20]. In addition, obese macaques experience changes in markers for lipid metabolism, e.g., increased total cholesterol and triglyceride concentrations, which are risk factors for the development of both T2DM and cardiovascular disease [18,21]. These four biochemical parameters provide information regarding health risks and can thus be useful in the diagnosis and management of overweight-related health problems in macaques [20].In an attempt to reduce overweight incidence and overweight-related health problems, a dietary change was implemented in our long-tailed macaque (Macaca fascicularis) breeding colony. The supplementary provisioning of bread was replaced by grains and vegetables, while the basic diet of monkey chow pellets remained the same. Although wild long-tailed macaques mainly eat fruits, it would not be appropriate to feed similar amounts of fruit in captivity. Cultivated fruits have a different nutritional composition, i.e., less protein and fiber and more sugar, compared to wild fruits [12,22]. Since the nutritional composition of cultivated vegetables is more like wild fruits, more vegetables than fruits were provided. The dietary change led to an increased fiber content and a lower GI. The implementation of this dietary change had little impact on daily husbandry practices, i.e., feeding times and routines remained unchanged.This study is a retrospective evaluation that investigated whether this minor dietary change had a beneficial effect on relative adiposity and overweight-related health parameters in non-diabetic, group-housed, female long-tailed macaques. The evaluation was based on data retrieved from electronic health records. Relative adiposity and biochemical parameters were measured during annual health checks before and after dietary change. Overweight status was determined with a species-specific weight-for-height index, which represents relative adiposity levels of long-tailed macaques [11]. Biochemical parameters related to glycemic response, i.e., fructosamine and HbA1c, and lipid metabolism, i.e., cholesterol and triglyceride, were compared to evaluate the effect of dietary change on the risk of T2DM and cardiovascular disease.2. Materials and Methods2.1. Subjects and HousingSubjects of this study were 44 full-grown adult female long-tailed macaques from the breeding colony of the Biomedical Primate Research Centre (BPRC), an AAALAC accredited facility, in Rijswijk, the Netherlands. The animals were aged between 6 and 22 (10.7 ± 0.61) years old and weighed between 3.4 and 9.15 (5.5 ± 0.20) kg at the time of initial data collection. Pregnant and (pre)diabetic individuals were excluded to prevent the possibility of pregnancy or disease progression interfering with our outcome parameters. All females lived with their offspring and typically with one adult breeding male in multi-generational groups (N = 9 groups). The groups were formed by adhering to natural group dynamics, i.e., females are philopatric, while males are removed from their natal group at puberty. The amount of data on adult males was therefore insufficient to include them in the data analyses.Individuals had access to enriched indoor (±72 m2 and 2.85 m high) and outdoor (±250 m2 and 3.1 m high) compartments. The inside enclosure contained sawdust bedding, while the outside enclosure had a sand bedding where natural plant growth was possible. Environmental enrichment consisted of several climbing structures, beams, fire hoses, car tires, sitting platforms, and a swimming pool to stimulate natural behaviour [23]. Drinking water was freely available throughout the day via automatic water dispensers.2.2. Diet and Dietary ChangeThe basic diet of the macaques consisted of monkey chow pellets (Ssniff, Soest, Germany) that were daily fed in the morning. The amount of monkey chow per individual was calculated based on the basal metabolic rate and depended on their age, sex, and body weight [24]. Adult females were calculated to require on average 90 g of monkey chow per day. In addition, one slice of wheat bread (~30 g, three times a week), 120 g of fruit/vegetables (three times a week), or 15 g of a grain mixture (once a week) were provided per individual in the afternoon (Table 1). Since the sum of all individuals’ needs was provided to the group and the distribution of food among group members could not be controlled, actual food intake likely varied per individual. Food enrichment was provided occasionally but its contribution to daily nutritional intake was carefully controlled and considered negligible.A dietary change took place in June 2019 to reduce the diet’s glycemic index and enhance fiber content. The supplementary provisioning of wheat bread in the afternoon (three times a week) was replaced by maize silage, grain mixture, and vegetables. As a result, the ratio of fruit to vegetables changed from approximately 1:3 before dietary change to 1:5 after dietary change. The 10 most commonly fed fruit/vegetables were banana, bell pepper, cabbage, chicory, Chinese cabbage, cucumber, endive, leek, lettuce, and tomato (Table A1). The dietary change led to a 15.4% increase in fiber content in the average daily diet, while the amount of energy, protein, carbohydrates, and fat remained approximately the same (Table 2). Besides an increase in fiber content, the removal of bread led to a lower overall GI after dietary change.2.3. Data CollectionThis retrospective evaluation was based on data retrieved from BPRC’s electronic health records, which included data from annual health checks. These health checks are a routine veterinary procedure related to the regular health management of the colony [25]. No additional procedures were performed, and all procedures complied with regulations in the European Directive 2010/63 and the Dutch law. The health checks prior to dietary change took place in spring 2018. Since the dietary change took place in June 2019, data from the health checks in autumn 2020 were used for testing the effect of the dietary change. Subjects served as their own control to exclude possible confounding factors, e.g., dominance rank, genetics, etc.Prior to the health checks, individuals were fasted overnight, while water was freely available throughout the night. At the assessment in spring 2018 (before dietary change), individuals were sedated with an intramuscular injection of ketamine (10 mg/kg, Ketamine 10%; Alfasan, Woerden, The Netherlands). There was a subsequent change in the routine anesthesia protocol for the benefit of the animals. In autumn 2020 (after dietary change), monkeys were thus sedated with a combination of ketamine (10 mg/kg, Ketamine 10%; Alfasan, Woerden, The Netherlands) and medetomidine (0.05 mg/kg, Sedastart; AST Farma, Oudewater, The Netherlands) IM, which was reversed after the procedures with atipamezole (0.25 mg/kg, Sedastop; AST Farma, Oudewater, The Netherlands) administration IM. Medetomidine induces muscle relaxation and results in mild hyperglycemia [26,27,28].As part of the health check, body weight and height were determined, as described earlier [11]. Briefly, a standard scale was used to measure body weight to the nearest 0.1 kg. Height was measured as crown–rump length by placing the monkeys on their back on a measuring mat (SECA, Hamburg, Germany). Height was measured to the nearest 0.1 cm. Body weight and height were used to calculate a species-specific weight-for-height index (hereafter referred to as WHI). WHI was calculated as weight (in kilograms) divided by height (in meters) to the power of 2.7 (WHI2.7 in [11]). This measure of relative adiposity was preferred over solely using body weight as the latter does not take into account individual variation in height. Although all females were full-grown and skeletally matured, height was highly variable (range: 40.1–47.7 cm). We determined overweight status and individuals were considered overweight when their WHI exceeded the upper boundary of 62 kg/m2.7 [11].Furthermore, blood samples were collected for complete blood count and blood chemistry. The samples were analyzed for fructosamine (umol/L), HbA1c (%), total cholesterol (mmol/L) and triglyceride (mmol/L) levels using a Cobas Integra 400 plus (Roche Diagnostics, Rotkreuz, Switzerland). Blood samples were collected from the vena femoralis into EDTA and serum tubes (Vacuette, Greiner Bio-One international GmbH, Alphen aan den Rijn, The Netherlands), left for 30 min and centrifuged at 3000 rpm for 10 min. Afterwards, the remaining serum was transferred to polypropylene tubes and stored below −20 °C.All biochemical parameters, except for triglyceride after dietary change, were analyzed on the same day of the sample collection. Triglyceride levels after dietary change were analyzed roughly five months after sample collection. A correction was applied to the data as triglyceride levels in serum are only stable up until three months when stored at −20 °C [29]. The correction was based on the regression equation between triglyceride levels after five months and deviation from the actual value from thirty samples for which the original triglyceride values were available (Appendix B).2.4. Data AnalysesStatistical testing was performed in IBM SPSS Statistics version 26. The effect of dietary change on body weight, WHI, and biochemical parameters was tested with a paired samples t-test or Wilcoxon signed ranks test. Normal distribution of the data was checked with the Shapiro–Wilk test. Pearson and Spearman correlations were used to test the association between age and WHI and associations between the different biochemical parameters. Linear regression analyses were used to evaluate whether age and WHI affected delta WHI, fructosamine, HbA1c, cholesterol, and triglyceride levels. Normality and homoscedasticity of the residuals were visually checked. Delta WHI was calculated as WHI after dietary change (in 2020) minus WHI before dietary change (in 2018). Finally, Fisher’s exact test was used to compare the proportion of overweight individuals before and after dietary change. Descriptive statistics in the results are reported as mean ± SE. The level of significance was α = 0.05 and all tests were two-tailed.3. Results3.1. Relative Adiposity and Overweight StatusMean body weight was 5.5 ± 0.20 kg before dietary change and 5.4 ± 0.19 kg after dietary change, which is not a statistically significant difference (paired samples t-test, t = 0.959, n = 44, p = 0.343). Similarly, WHI did not differ after dietary change (paired samples t-test, t = 0.991, n = 44, p = 0.327; Figure 1). WHI was independent of age in our study population both before (Spearman correlation, r = 0.180, n = 44, p = 0.242) and after dietary change (Spearman correlation, r = 0.053, n = 44, p = 0.731). Delta WHI was independent of age (F (1, 41) = 0.689, p = 0.411), but was significantly associated with baseline WHI (F (1, 41) = 11.731, p = 0.001). Delta WHI was significantly higher in individuals with a low baseline WHI, implying that WHI increased in relatively lean individuals, while WHI decreased in relatively heavy individuals after dietary change (Figure 2).The year before dietary change, four individuals (9.1%) had WHIs above the upper boundary for overweight, while only two individuals (4.5%) were overweight after dietary change. Overweight status did not significantly differ before or after dietary change though (Fisher’s exact test, p = 0.676).3.2. Biochemical ParametersTable 3 shows descriptive statistics on fructosamine, HbA1c, cholesterol, and triglyceride levels before and after dietary change. Correlations between the different biochemical parameters were weak or absent (Table A2).3.2.1. Glycemic ResponseFructosamine levels were independent of age and WHI both before (F (1, 41) = 0.053, p = 0.820; F (1, 41) = 1.520, p = 0.225) and after dietary change (F (1, 41) = 1.698, p = 0.200; F (1, 41) = 0.026, p = 0.873). After the dietary change, fructosamine levels were significantly reduced (paired samples t-test, t = 7.060, n = 44, p < 0.0005; Figure 3a).WHI had no significant influence on HbA1c levels before (F (1, 41) = 0.057, p = 0.812) or after dietary change (F (1, 41) = 1.214, p = 0.277). Age did not affect HbA1c levels before dietary change (F (1, 41) = 0.680, p = 0.414), yet age was positively associated with HbA1c levels after dietary change (F (1, 41) = 7.261, p = 0.010), i.e., older individuals had higher HbA1c levels. The regression equation indicated that HbA1c values increased 0.028% per year of age (R2 = 0.174). Dietary change had no significant effect on HbA1c levels (paired samples t-test, t = −0.759, n = 44, p = 0.452; Figure 3b).3.2.2. Lipid MetabolismCholesterol levels were independent of age and WHI before (F (1, 41) = 0.164, p = 0.687; F (1, 41) = 0.257, p = 0.615) and after dietary change (F (1, 41) = 0.082, p = 0.775; F (1, 41) = 2.567, p = 0.117). Cholesterol levels were significantly reduced after dietary change (paired samples t-test, t = 3.971, n = 44, p < 0.0005; Figure 4a).Triglyceride levels were independent of WHI before (F (1, 41) = 0.793, p = 0.378) and after dietary change (F (1, 41) = 3.353, p = 0.074), while triglyceride levels significantly increased with age before (F (1, 41) = 7.146, p = 0.011) and after dietary change (F (1, 41) = 5.491, p = 0.024). The regression equations showed that triglyceride levels increased with every additional year of age with 0.072 mmol/L before dietary change (R2 = 0.177) and 0.051 mmol/L after dietary change (R2 = 0.181). Triglyceride levels were not significantly different after dietary change (Wilcoxon signed ranks test, Z = −0.604, n = 44, p = 0.546; Figure 4b).4. DiscussionThis study evaluated, based on electronic health records, the effect of a minor dietary change on relative adiposity and overweight-related health parameters in non-diabetic, group-housed, female long-tailed macaques. Relative adiposity and biochemical parameters related to glycemic response and lipid metabolism were compared before and after the supplementary provisioning of bread was replaced by grains and vegetables. Relative adiposity and overweight status did not differ after dietary change. Yet, relatively heavy individuals generally lost body weight, while relatively lean individuals gained body weight, leading to a more balanced body weight dynamic. Dietary change had no effect on HbA1c and triglyceride levels, while fructosamine and cholesterol levels were significantly reduced. Thus, the minor dietary change had no significant effect on overweight status but had a positive effect on some biochemical parameters related to the risk of T2DM and cardiovascular disease.4.1. Relative Adiposity and Overweight StatusBody weight, WHI, and overweight status did not differ after dietary change. Based on the increased fiber content, a reduction in relative adiposity was expected. Fiber intake increases satiety and decreases the feeling of hunger after a meal, which results in reduced energy intake, even when food is available ad libitum [30]. The importance of fiber in the diet of captive primates is increasingly being recognized and the provisioning of browse (e.g., willow twigs), which is high in fiber, is therefore often recommended and becoming more popular in zoos and other institutions [31,32]. A reduction in body weight after transitioning to a high-fiber diet was found in a vervet monkey (Chlorocebus aethiops sabaeus) breeding colony [33], but not in our study. However, the relative increase in fiber content was almost tenfold higher in the vervet monkey study (140%) compared to our study (15.4%). Higher fiber contents may be needed for relative adiposity to decrease overall.Nevertheless, dietary change had a differential effect on relative adiposity of individual animals, depending on their baseline value. Relatively lean individuals gained body weight, while relatively heavy animals generally lost body weight after dietary change. This finding may be explained by an unexpected secondary effect of the dietary change. Although food intake was not measured in this study, this finding suggests that dietary change resulted in a different distribution of food among group members. Wheat bread was easy to monopolize, resulting in some individuals obtaining several slices, while others obtained none (personal observation, cf. [34,35]). In contrast, grains, maize silage, and leafy vegetables were likely divided more equally, as these items were spread through and/or in front of the cages. The more equal distribution of these food items may have led to relatively lean individuals obtaining more food than before, thus gaining body weight, while relatively heavy individuals obtained less food, thereby losing body weight. Even though relative adiposity did not decrease overall, dietary change had a differential effect on individual animals resulting in a more balanced body weight dynamic.Relative adiposity was not related to any of the biochemical parameters in this study, while other studies found several associations between being overweight and indicators of glycemic response and lipid metabolism in macaques. Cholesterol and triglyceride levels are generally higher in obese male and female rhesus macaques (Macaca mulatta) compared to their non-obese counterparts [1,21,36]. Body weight is also positively correlated with triglyceride and glucose levels in adult female long-tailed macaques [37]. These studies often included highly obese subjects with body fat accounting for up to 61% of their body weight [1]. This body fat percentage would equal a body weight of roughly 12.65 kg [38], while the heaviest monkey in our study initially weighed 9.15 kg. Accordingly, no effect of body weight on cholesterol, triglyceride, or glucose levels is found in long-tailed macaques with relatively low body weights [39]. Thus, the relatively low overweight prevalence and little variation in relative adiposity between individuals in our study may explain the lack of significant associations between WHI and biochemical parameters.The absence of these relationships implies that this long-tailed macaque population is generally healthy regarding overweight-related health parameters. All biochemical parameters, i.e., fructosamine, HbA1c, cholesterol, and triglyceride levels, also fit well within previously reported ranges for this species [18,19,39,40,41,42,43,44]. However, this does not mean that overweight-related health problems do not occur in our breeding colony.4.2. Glycemic ResponseSince a high-GI food (bread) was replaced with low/medium-GI foods (grains and vegetables), the glycemic response was expected to improve, thereby reducing the risk of T2DM. Glycemic response was measured using fructosamine and HbA1c levels, which produced different results regarding the effect of dietary change. Fructosamine levels decreased, while HbA1c levels showed no significant difference after dietary change. These results may be explained by the difference in sensitivity of albumin and hemoglobin to bind to glucose. Fructosamine is formed when plasma glucose binds to albumin, while HbA1c results from glycation of hemoglobin [20]. Albumin has been suggested as being more sensitive to postprandial glycemic variation compared to hemoglobin and therefore larger alterations in blood glucose would be needed to affect HbA1c levels similar to fructosamine levels [45,46]. As a result, varying fiber and glucose intake in humans does not affect HbA1c, but significantly influenced fructosamine levels [46]. Similarly, fructosamine levels differ between long-tailed macaques fed a standard or high-fat diet, while no difference is found in HbA1c [19]. These findings are consistent with the outcome of our study. Although no significant effect on HbA1c levels was found, the decrease in fructosamine suggests that the dietary change had a positive impact on glycemic response. As the decrease in fructosamine was observed across the study population, this was likely a primary effect of dietary change and independent of the potentially new food distribution.A third possible biochemical parameter to quantify glycemic response is plasma glucose concentration. Although plasma glucose levels were measured, a fair comparison was not possible as medetomidine is known to affect glucose levels and this was added to the anesthesia protocol [27,28]. Moreover, plasma glucose levels provide information about instant glucose levels, while fructosamine levels reflect blood glucose levels from the past two to three weeks and HbA1c represents the previous two to three months [20]. Fructosamine and HbA1c are thus more suitable parameters to detect long-term changes in glycemic response as they reflect glucose levels over a longer period.Nevertheless, there was no significant association between fructosamine and HbA1c levels. Cefalu et al. (1993) found that fructosamine and HbA1c are significantly correlated (r = 0.61) in a long-tailed macaque population, which included both diabetic and non-diabetic monkeys [19]. Fructosamine and HbA1c also correlate well in diabetic humans (0.55 < r < 0.88; [47,48,49,50]), but no correlation has been found in non-diabetic humans (r = 0.01; [50]). Our study included only non-diabetic individuals, which might explain the lack of correlation between fructosamine and HbA1c.Furthermore, HbA1c levels increased with age after dietary change, but this was not found before dietary change. HbA1c levels are also positively associated with age in non-diabetic humans [51,52,53], but not in other studies with macaques and squirrel monkeys (Saimiri species; [44,54]). Since higher age is a risk factor for the development of T2DM in both humans and primates [18,55], the link between age and HbA1c and their relation to T2DM in primates may need further investigation.4.3. Lipid MetabolismMultiple studies show that an increased fiber intake has a positive effect on lipid metabolism, i.e., leads to reduced total cholesterol, LDL and triglyceride levels, in humans and rats [56,57]. Especially water-soluble fibers seem to have this cholesterol-lowering effect in humans [58]. Therefore, it was expected that the serum cholesterol and triglyceride levels would decrease after the dietary change. In line with this expectation, cholesterol levels decreased after dietary change. In contrast, triglyceride levels showed no significant difference after dietary change.In the present study, higher triglyceride levels were found in older individuals compared to younger monkeys both before and after the dietary change. Similar age-effects have been reported in other studies with both long-tailed macaques and rhesus macaques [42,43,59,60]. Possibly, no significant effect of dietary change on triglyceride levels was found because the effect of dietary change was counteracted by an age-effect. Furthermore, triglyceride levels after dietary change had to be corrected due to the period between blood sample collection and analysis. This correction may have introduced some bias, thereby reducing reliability of the triglyceride data.Altogether, the minor dietary change had a beneficial effect on at least one of the two biochemical parameters related to cardiovascular disease.5. ConclusionsThis study evaluated in retrospect the effect of a minor dietary change on relative adiposity and overweight-related health parameters in non-diabetic, group-housed, female long-tailed macaques. The basic diet of monkey chow pellets remained the same, while the supplementary provisioning of bread was replaced by grains and vegetables. Although this minor dietary change had no significant effect on overweight status, dietary change had a differential effect on individual animals resulting in a more balanced body weight dynamic. Also, some biochemical parameters related to the risk of diabetes and cardiovascular disease were positively affected. These results emphasize the importance of evaluating husbandry changes and shows that critically reviewing husbandry practices can provide valuable insights to improve animal health and welfare.

animals : an open access journal from mdpi

10.3390/ani12020179

PMC8772703

The reduction of dependence on fishmeal as a main protein source for aquafeeds remains a big problem in reaching sustainable aquaculture. Several alternatives to this ingredient are being tested and developed, insects being one of the most promising. The present study included two different insect species (black soldier fly, Hermetia illucens, and yellow mealworm, Tenebrio molitor) in the formulation of diets for rainbow trout (Oncorhynchus mykiss) against one typical fishmeal-based diet. Different parameters related to both the efficiency of these diets and their physiological repercussions were analysed. Yellow mealworm proved to be the best alternative for the growth and nutrition of rainbow trout, possibly due to some changes described in protein utilization and intestine histology, while other parameters revealed the possible usage of insect meals as functional ingredients due to their repercussions on preventing tissue damage.

The demand of optimal protein for human consumption is growing. The Food and Agriculture Organization (FAO) has highlighted aquaculture as one of the most promising alternatives for this protein supply gap due to the high efficiency of fish growth. However, aquaculture has been facing its own sustainability problem, because its high demand for protein has been traditionally satisfied with the use of fishmeal (FM) as the main source. Some of the most promising and sustainable protein substitutes for FM come from insects. The present manuscript provides insight into an experiment carried out on rainbow trout (Oncorhynchus mykiss) with a 50% replacement of FM with different larvae insect meals: Hermetia illucens (HI), and Tenebrio molitor (TM). TM showed better results for growth, protein utilization and more active digestive function, supported by intestinal histological changes. Liver histology and intermediary metabolism did not show relevant changes between insect meals, while other parameters such as antioxidant enzyme activities and tissue damage indicators showed the potential of insect meals as functional ingredients.

1. IntroductionAlbeit at a slower speed than some decades ago, the global population is expected to keep increasing and reach 8.5 billion in 2030, 9.7 billion in 2050, and 10.9 billion in 2100 [1]. As a consequence of this increment, the demand of adequate protein for human consumption is also increasing. Aquaculture is one of the most promising alternatives to satisfy this demand due to the high efficiency of fish growth [2], the rapid development of the aquaculture industry itself, and the adequate calories—protein ratio of fish [3]. However, because many of the fish cultivated for human consumption require high protein levels to grow appropriately, aquaculture has been facing its own sustainability problems in the last few decades. These protein requirements have been traditionally satisfied with the use of fishmeal (FM) from wild-caught fish and as a by-product of extractive fishing practices [4]. Due to the fast growth of aquaculture, these ingredients are considered as non-sustainable in the long term.Many efforts have been carried out from both research and aquaculture industries to partially replace FM with sustainable ingredients in fish feeds, without impairing fish growth and while giving insight into these sustainable ingredients. Alternatives such as vegetable ingredients [5,6], yeast [7,8], or microalgae [9,10] are some of the ingredients that are being studied currently. Following this line, the present study is focused on insects as one of the most interesting protein substitutes for FM [11,12,13,14]. Setting aside the interspecific differences, as well as the harvesting time of larvae [15], the amino acidic proportions of the most typically studied insects tend to match that of FM [16,17]. Insects also reproduce and grow easily, have very efficient growth ratios, and require low amounts of space and energy to be produced [18]. Hence, their potential as a good source of sustainable animal protein is promising. Because the Food and Agriculture Organization (FAO) has mentioned zero hunger, sustainable communities, and life below water as three of its 17 Sustainable Development Goals of the 2030 Agenda [19], it is easy to assume that both aquaculture and insect production might consequentially play important roles in the upcoming years or decades.Several manuscripts have proven the efficiency of insect meals (IMs) in different fish species [20,21,22,23,24], revealing the importance of both the insect and the fish species involved. For salmonids, the inclusion of IMs in feeds is, in general, well accepted. As an example of unaffected growth, Terova [25] replaced up to 30% FM with Hermetia illucens (HI) in rainbow trout feed (Oncorhynchus mykiss). In the case of Atlantic salmon (Salmo salar), it was already proven that FM could be replaced entirely, using HI as one of the chosen protein sources [26]. Another experience [27] tested two hydrolysed IMs (yellow mealworm, Tenebrio molitor (TM) and superworm, Zophobas morio) on fingerlings of sea trout (Salmo trutta m. trutta) at 40% FM replacement and noticed almost no changes in growth or protein use. Moreover, another experiment [28] did not note changes in growth for rainbow trout, substituting FM completely with TM. These and other published data support the idea that, with the due differences among species, a partial replacement of FM with IM has no adverse effects on the growth of most fish. Moreover, functional properties such as a possible enhancement of both the immunological and the antioxidant systems have been attributed to IMs, possibly due to compounds such as chitin or its derivatives [14,21,29,30,31,32,33].The European Commission approved the use of seven insects as ingredients in aquafeeds [34]. Due to their relative availability, HI and TM are two of the most broadly studied insects for animal nutrition. Thus, the IM industry has a big potential and requires research studies to validate the use of IMs as an alternative ingredient in feed for aquaculture.Following the results of a previous study with 15–30% FM replacement (5–10% IM inclusion level) [32], but increasing the FM replacement to 50% (18% IM inclusion level) in feed for rainbow trout, the present manuscript provides insights on the effects of two different IMs for several aspects, from growth to final composition of the fillets, while evaluating the physiological status of the fish and their possible consequences on health and welfare status.2. Materials and Methods2.1. Experimental DietsWhole dried insects from two different species in larval stage, Hermetia illucens (HI; Entomotech S.L., Almería, Spain) and Tenebrio molitor (TM; Mealfood Europe S.L., Salamanca, Spain) were used for this study, processed as insect meals (IMs). IMs were analysed before the formulation of the diets (Table 1). A total of three isoproteic (43.3%) and isolipidic (17.4%) diets were formulated (Table 2): a control diet with no IM (C), and two diets with 18% diet inclusion (50% fishmeal replacement) of the cited IMs: H18 (HI), and T18 (TM). Ingredients were provided by ‘Lorca Nutrición Animal S.A.’ (Murcia, Spain). Methionine and lysine were added to diets to meet the nutritional requirements of rainbow trout [35,36], manufactured by LifeBIOENCAPSULATION S.L. (Almería, Spain), and extruded as pellets of 3 mm. The dough was passed through a single screw laboratory extruder (Miltenz 51SP, JSConwell Ltd., Palmerston North, New Zealand). The extruder barrel had four sections, with a temperature per section of (from inlet to outlet) 100 °C, 95 °C, 90 °C and 85 °C, respectively. Pellets were kept in a drying chamber at 30 °C for 24 h (Airfrio, Almería, Spain) and stored in sealed plastic bags at −20 °C until they were used.2.2. Experimental Animals and Rearing ConditionsA total of 360 female rainbow trout with an initial weight of 14.6 ± 0.2 g from a commercial farm (Piscifactoría Fuente del Campillo, Guadalajara, Spain) were transported to the experimental facilities of the Aquaculture Research Centre of “Instituto Tecnológico Agrario de Castilla y León” (ITACyL). Fish stayed in acclimation for 15 days before the beginning of the growth trial, and then they were randomly allocated into 12 cylindrical fiberglass tanks (four replicas per treatment; 500 L) of a recirculating system, in groups of 30 animals. Once a day (9 a.m.), fish were fed by hand until apparent satiation was reached (maximum of 3% daily feed intake). During the growth trial (77 days), water temperature (12.5 ± 1 °C), water dissolved oxygen (9.2 ± 1 mg/L), and room photoperiod (12 h light: 12 h dark) were monitored. Water ammonia and nitrite levels were analysed daily, and kept at optimal levels (ammonia < 0.1 mg/L and nitrite < 0.1 mg/L). The care and handling of rainbow trout were conducted according to specific regulations: The Directive of the European Union Council (2010/63/EU) [37] and the Spanish Government (Real Decreto 53/2013) [38]. The experiment was approved previously by the Bioethical Committee of “ITACyL” (Authorization number: 2017/19/CEEA).2.3. Growth Trial and Samples CollectionMortality and feed intake were monitored on a daily basis. Fish were measured and weighed every 21 days through a simple biometry procedure with a graduated ictiometer (±0.1 mm) and scale (±0.1 g), being previously fasted for one day and anesthetized with tricaine methanesulfonate (MS-222; 180 mg/mL). In order to take samples of the different tissues, the fish were sacrificed by an overdose of MS-222 (300 mg/mL).Before the feeding trial, eight fish were randomly sacrificed to analyse the initial value of the protein in the fillet.During the final two weeks of the experiment, faeces were gathered every 24 h in a settling column using a modified Guelph method [39], and frozen at −80 °C until they were analysed. At the end of the experiment, eight fish per diet (2 fish per tank) were randomly sampled and sacrificed. According to time sequence, the following were collected to be analysed individually: skin mucus, blood, liver, stomach, intestine with pyloric caeca, and fillet samples. Skin mucus samples were collected by scraping the dorso-lateral surface of the fish skin from cranial to caudal according to de Mercado et al. [40] and frozen at −80 °C until processing. Blood samples were collected with heparinized syringes and their plasma was separated by centrifugation at 3500× g and 4 °C, for 15 min. Individual plasma samples were frozen at −80 °C until their analysis.For enzyme determinations, samples were frozen in liquid nitrogen and kept at −80 °C until they were analysed. For tumour necrosis factor-alpha determination (TNF-α), distal intestine samples were kept in Allprotect Tissue Reagent (QiaGEn) and stored at −20 °C until protein extraction. The samples for histomorphology analyses were fixed in 4% buffered formalin for 48 h before dehydration and processing. For chemical analyses, the samples were directly frozen at −80 °C.2.4. Histomorphology2.4.1. Samples ProcessingThe fixed samples were dehydrated in increasing ethanol solutions (25%, 50%, 75%, and 100%) and embedded in synthetic paraffin. Histological sections (3–4 µm) were obtained by a rotary microtome (FINESSE ME+ Thermo Scientific©, Waltham, MA, USA), stained by hematoxylin and eosin technique for histomorphology studies and observed with light microscopy. All of the evaluations were performed by graded objective lens in five random regions for each stained tissue section with an Olympus EP50 microscope camera and an Olympus CX31 microscope.2.4.2. Distal Intestine and Pyloric Caeca Histomorphology AnalysesQuantitative studies included the measurement of heights of villi and enterocytes, as well as widths of villi, stratum compactum, muscular layers (longitudinal and circular), and lamina propria as mean of three measures (apical, intermediate, and basal). The level of inflammatory infiltration in lamina propria, the level of loss of supranuclear vacuolization of enterocytes, and the relative position of enterocyte nuclei were measured through a subjective analysis.2.4.3. Liver Histomorphology AnalysisHepatocyte cytoplasm and hepatocyte nucleus measures were taken as quantitative variables. A qualitative analysis concerning the search of inflammatory patterns (necrosis and inflammation focuses) and hepatocyte intranuclear vacuolization was also carried out.2.5. Analytical DeterminationsFor intermediary metabolism and antioxidant status, liver samples were individually homogenized in nine volumes of ice-cold 100 mM Tris-HCl buffer, containing 0.1 mM EDTA and 1 g/kg (v/v) Triton X-100, pH 7.8. This was followed by centrifugation at 30,000× g for 30 min, at 4 °C. For further enzyme assays, the supernatants were stored at −80 °C as aliquots.The concentration of soluble protein in samples was determined by Bradford method [41], employing bovine serum albumin as a standard.2.5.1. Chemical AnalysesAOAC methods [42] were used to analyse fat content and moisture of IMs, diets, and fish fillets. Protein content was determined with the Dumas method [43], using a nitrogen analyser (FP 528, LECO, St. Joseph, MO, USA), and with a conversion factor of 4.67 for HI, 4.75 for TM [44], and 6.25 for feeds and faeces. Acid-insoluble ash was used as marker in feeds and faeces to determine the apparent digestibility of the protein [45]. Phosphorus (P) was determined by molecular absorption spectrophotometry according to ISO standard [46], with a spectrophotometer (UV/Vis UV2, UNICAM, Cambridge, UK). Calcium was determined as described by Pessoa [47], with X-ray fluorescence method of Dispersive Energy. The method described by Gamage and Shahidi [48] was used to isolate chitin from IM, which was washed with acetone, dried, and weighed afterwards. For amino acids, samples of HI and TM were hydrolysed with 6 N HCl for 22 h at 110 °C [16]. The determination was performed by ion-exchange liquid chromatography and postcolumn continuous reaction with ninhydrin (Biochrom 30; Cambridge, UK). Tryptophan was not determined.2.5.2. Digestive Enzymes DeterminationIntestine with pyloric caeca and stomach were processed separately to determine digestive enzymes. Samples were first individually homogenized at 4 °C with distilled water (250 mg/mL). Acid protease activity was determined from stomach extracts, while amylase and alkaline protease activities were determined from the intestine and pyloric caeca extracts. The activity of amylase was determined through the Somogy–Nelson method [49], with soluble starch 20 g/kg as substrate, defining one unit of activity as the quantity of enzyme able to produce 1 μg of maltose per minute and mg of protein. Walter method [50] was used to measure the activity of alkaline protease, employing casein 10 g/kg as substrate. Anson method [51] was used to measure the activity of acid protease activity, with hemoglobin 5 g/kg as substrate. For both proteases, one unit of activity was defined as 1 μg of tyrosine produced per minute and mg of protein. The standard temperature for all digestive enzyme analyses was 37 °C.2.5.3. Liver Intermediary MetabolismThe method described by Furné [52] was used to determine the enzymatic activity of fructose 1,6-bisphosphatase (FBPase; EC 3.1.3.11), pyruvate kinase (PK, EC 2.7.1.40), glutamate pyruvate transaminase (GPT; EC 2.6.1.2), glutamate oxaloacetate transaminase (GOT; EC 2.6.1.1), and glutamate dehydrogenase (GDH; EC 1.4.1.2). Enzymes were analysed at 25 °C, and changes in absorbance were monitored with a PowerWaveX microplate scanning spectrophotometer (BioTek Instruments, Winooski, VT, USA) to determine the enzyme activity.2.5.4. Non-Specific Immune StatusPlasma Immunological DeterminationsLysozyme activity was performed using a turbidometric method [53] with Micrococcus lysodeikticus (Sigma, St. Louis, MO, USA). After reaction for 20 min at 35 °C, the absorbance was measured at 450 nm. A standard curve with hen egg-white lysozyme was used.Total esterase activity was assayed according to Mashiter and Morgan [54] at 25 °C. P-nitrophenyl acetate (0.8 mM) was used as a substrate, and 1.6 mM acetazolamide as an inhibitor of carbonic anhydrase activity. The absorbance increase was measured at 405 nm for 5 min, after incubation for 10 min.Anti-protease was measured according to Thompson [55]. The production of 4-nitroaniline was determined by the variation of the OD (optical density) at 410 nm for 30 min. Trypsin activity in absence of plasma was used as control (CAS 90002–07–7, Acofarma, Spain).Phosphatase activity was determined according to Huang [56]. P-nitrophenyl phosphate (Sigma) was used as a substrate, a buffer at pH 10 (NaHCO3/NaOH 0.05 M, MgCl2 1 mM) was used for alkaline phosphatase activity, and a buffer at pH 5 (CH3COOH/CH3COOHNa 0.1 M, MgCl2 1 mM) to measure acid phosphatase activity. The measurement was performed at 405 nm for 30 min, at 37 °C.Peroxidase activity was determined according to Mohanty and Sahoo [57]. A solution of 20 mM TMB (3, 3′, 5, 5′-Tetramethylbenzidine) was used as substrate. The samples were read at 450 nm after blocking reaction for 2 min. Plasma-free standard samples were measured as controls. The activity was expressed in OD (optical density).Total immunoglobulin was determined according to Panigrahi [58]. After precipitation with polyethylene glycol, the immunoglobulins were separated from the total proteins. Total immunoglobulin content was calculated by subtracting the protein content resulting from the total protein content in the untreated plasma.TNF-α Detection in Distal Intestine and Skin MucusFor protein extraction, samples of the distal intestine and skin mucus were homogenized using beads and an ice-cold lysis buffer (Tris 20 mM, NaCL 100 mM, Triton X-100 0.05%, EDTA 5 mM, protease inhibitor cocktail 1X), in a bead mill homogenizer (Qiagen RETSCH tissuelyser). The homogenate was centrifuged for 25 min, at 12,000× g and 4 °C. The soluble proteins contained in the supernatant were stored at −20 °C until use. The cytokine TNF-α was determined following the indirect ELISA method described by Morales-Lange [59], with slight modifications according to Weththasinghe [33]. Briefly, 100 µL of sample diluted to 45 ng/µL in a carbonate buffer (60 mM NaHCO3, pH 9.6) were seeded into 96-well plates (NUNC MAXISORPTM, Invitrogen), and incubated overnight at 4 °C. After blocking (5% Blotting-Grade Block, BioRad, Hercules, CA, USA; 2 h at 37 °C), plates were incubated for 90 min at 37 °C with 50 μL of the primary antibody (rabbit anti-TNFα, diluted 1: 200). Next, 50 μL of the secondary antibody (mouse anti-rabbit IgG-HRP, diluted 1: 7000) were added and incubated for 60 min at 37 °C. Finally, 100 μL of chromagen substrate 3,3′,5,5′- tetramethylbenzidine single solution (TMB, Thermofisher, Waltham, MA, USA) was added and incubated for 30 min at room temperature. The reaction was stopped with 50 mL of 1 N sulfuric acid and read at 450 nm on a Spectra Max microplate reader (Spectra Max M2; Molecular Devices, San José, CA, USA). The calibration curve was performed using serial dilutions of the corresponding epitope peptide ranging from 0 µg/mL to 1.2 µg/mL.2.5.5. Liver Antioxidant Status and Fish Welfare IndicatorsThe procedure described by Pérez-Jiménez [60] was followed to determine superoxide dismutase (SOD, EC 1.15.1.1), catalase (CAT, EC 1.11.1.6), glutathione peroxidase (GPX, EC 1.11.1.9), glutathione reductase (GR, EC 1.6.4.2), and glucose-6-phosphate dehydrogenase (G6PDH; EC 1.1.1.49). The enzyme analyses were carried out at 25 °C, and the enzyme activity was determined using a PowerWaveX microplate scanning spectrophotometer (BioTek Instruments, Winooski, VT, USA), through the monitorization of absorbance changes. Preliminary assays allowed the establishment of the optimal substrate and protein concentrations to measure the maximal activity of each enzyme. The millimolar extinction coefficients used for NADH/NADPH, DTNB, and H2O2 were 6.22, 13.6, and 0.039/mM·cm, respectively. The enzyme needed to inhibit half of the ferricytochrome C reduction rate was defined as one unit of SOD activity. For other enzymes, the amount of enzyme required to transform 1 μmol of substrate per minute was defined as one unit of enzyme activity. Malondialdehyde (MDA) level was used to quantify lipid peroxidation. MDA reacts in the presence of thiobarbituric acid to produce coloured thiobarbituric acid reacting substances (TBARS).Plasma glucose and lactate levels were analysed with commercial colorimetric kits, following the instructions of the manufacturer (Glucose-TR, ref. 41011, Spinreact; Lactate, ref. 1001330, Spinreact). The absorbance was measured using a microplate reader (ELx800TM; BioTek Instruments, Inc., Winooski, VT, USA), in 96-well microplates.2.6. Statistical AnalysisThe software used for the statistical analyses was SAS system version 9.0 (SAS Institute Inc., Cary, NC, USA). A general linear model (PROC GLM) analysis of variance (ANOVA) was used to process the data, and the comparison of the means was performed by a Tukey test. Differences were considered significant when the p-value was <0.05. Values are showed as mean ± standard error of the mean.3. Results and Discussion3.1. Growth PerformanceIn general, the performance of all diets was within normal values, with an efficient FCR (Table 3). Fish fed with T18 showed the best overall growth performance with very similar values to C while H18 showed lower numbers for growth, being statistically different from T18, or even to C when talking about SGR, FCR, and the apparent digestibility coefficient of the protein (ADCprot). Even though there are small discrepancies in the literature about the performance of HI as an ingredient for fish [61,62,63,64], the present results seem to follow the general conclusions of other trials in rainbow trout. Rainbow trout seem to have a higher tolerance to the inclusion of TM in diets [65,66] than that of HI [61,67], which could be due to the different levels of chitin in the composition of the insects, or their amino acid profiles. Chitin might have a positive influence over fish physiology as a functional ingredient [68], but the presence of this molecule tends to lower the digestibility of crude protein [69,70]. Because HI has higher levels of chitin in its body composition than TM (Table 1), this, together with results in other experiences [61,65,67,71] suggest that a 15–18% inclusion of HI in rainbow trout feed is a possible maximum level of inclusion for this species, while an 18% of TM or even more, is still compatible with optimal growth performance. In addition, the digestibility of the protein was higher in TM than HI; although the amino acid profile between IMs differed, the diets were supplemented with methionine and lysine to cover the nutritional requirements. This should had led to a similar growth between insect-based treatments, but the higher growth of T18 over H18 means that its higher digestibility played an important role, and consequently led to a higher growth than HI.The mortality during the trial was around 3% without remarkable differences between diets (data not shown).3.2. Histomorphology3.2.1. Distal Intestine and Pyloric CaecaOn distal intestine (Figure 1), no significant differences were described for villi, stratum compactum, longitudinal muscular layer, or lamina propria widths. Villi height was higher for T18 fish than H18 (p < 0.05), with an intermediate value for C fish. Enterocyte height was higher in T18 than in C (p < 0.05), with no significant differences between H18 and T18. The circular muscular layer was wider on C than on H18, while the total muscular layer was wider on C than on both IM treatments (p < 0.05). For qualitative analyses (Figure 2), slightly higher levels of both inflammatory infiltration and loss of intracellular vacuolization were highlighted for C diet. The different degrees of supranuclear vacuolization did not affect the enterocyte structure because most nuclei were positioned on the basal part of the enterocytes, showing no differences between treatments.Few significant differences were found in pyloric caeca compared to the distal intestine. T18 showed the highest values for villi height (compared to H18; p < 0.05), while the rest of the results remained stable (Figure 1). For the qualitative analyses, no differences were observed for inflammatory infiltration or intracellular vacuolization, and most nuclei were described on the intermediate part with no differences between treatments.The histology of insect-fed fish intestine has been extensively studied in the last few years [27,33,63,72,73,74]. However, due to the large number of variables involved in the studies, such as the fish species, the insect used to elaborate the feeds, the chosen intestine sections, or the analysed parameters, there is still work to be conducted. Villi height is one of the most frequently analysed parameters, being an indicator usually associated with gut health and growth performance. In this way, the results of this study match partially those of the current literature, because it has been described that similar inclusions of HI in fish feed tend to decrease villi height [67,71,75,76,77]. There are other cases in which no changes or even an increase in villi height was described with the addition of HI meal [63,78,79]. The case of TM is less studied and it seems that, with one exception [63], most studies describe no changes in villi height when dealing with this ingredient [27,73,80].On pyloric caeca, the results were similar. However, this is a less studied variable for insect-fed fish, and to our knowledge, only two studies evaluated this parameter after a growth trial with HI treatments; one that matched the data of this study [77], and one that offered opposite results [81]. Considering that digestive efficiency and growth are directly related to gut anatomy alterations such as a decrease in the absorption surface [82,83,84], it is no surprise that the results of the present manuscript agree with the lower growth performance and protein digestibility described for HI, as well as the higher results on TM.The circular muscular layer was also affected in the present study. The main finding was a decrease in the width for H18 with significant difference with respect to C diet (p < 0.05). Because contraction and relaxation of circular and longitudinal muscular layers lead to peristaltic movements, a different width of the circular muscular layer could alter the movement of the feed along the gut, affecting the intestinal bacterial growth [85] and ultimately the digestibility of the nutrients, consistent with the lower digestibility of the protein observed in fish fed with H18. Similar results were showed by Lu [76], while other authors have not described changes in muscular layer width [27,79,81]; there is even a case in which different insect species gave different results [63] with the inclusion of IMs. This divergence of data is probably caused by the different species of fish used.It is interesting to notice that the differences in the degree of enterocyte supranuclear vacuoles loss (fewer vacuoles in C) are in consonance with the results on enterocyte height (lower in C, with a significant difference between C and T18). The presence or absence of lipidic vacuoles in enterocytes has been related to their height for other ectothermic species [86,87]. In this way, the different nature of the fat between diets (insect fat in H18 and T18), and their absorption process may have played an important role in the degree of supranuclear lipidic vacuoles in enterocytes. Furthermore, the work of Kumar [88] described how IMs could cause a protective effect against the problems of soybean meal [89] in salmonids. Considering that C diet also showed a slightly higher submucosa inflammatory infiltration, the lower degree of vacuolization and the consequential lower enterocyte height with respect to IM diets could be due to the lack of this protective effect. The three diets had a relatively high amount of vegetable ingredients, but even though H18 and T18 had the highest amounts of these ingredients, they showed the lowest levels of inflammatory signs.3.2.2. LiverNo significant differences were found in any of the measured variables: hepatocyte nucleus and cytoplasm diameters, inflammatory patterns, or level of hepatocyte vacuolization (Figure 3). Liver histology has also been extensively studied for insect-fed fish [71,72,74,75,81]. One of the most frequent findings in several fish species, including rainbow trout, is that an increasing proportion of IMs in the feed tends to increase the number of lipidic vacuoles in hepatocytes, while other related variables are mostly unaffected [71,78,81]. This discrepancy with our results could be due to the different size and feeding period of the fish involved, because the cited study in rainbow trout was performed with bigger fish. The work of Kumar [88], however, had similar conclusions to those of the present study.3.3. Digestive EnzymesAcidic proteases showed no significant differences between treatments (Table 4; p > 0.05), while alkaline proteases and amylase showed significant increases in T18 treatment. The higher values of alkaline proteases for T18 are in consonance with the bigger digestibility of the protein demonstrated by this diet over H18 (Table 1). Because IM diets showed higher values of alkaline proteases than C, as well as lower acidic–alkaline ratios (non-statistically significant for H18), this supports the idea that insect-based diets suffered a more active digestion in the intestine. Contrarily, the work of Coutinho [23] showed a lower activity of total alkaline protease, trypsin, and lipase even on low inclusions of TM (defatted) for meagre (Argyrosomus regius). However, this work and the work of Guerreiro [22] showed that meagre might not be the best candidate to use IMs as a source of protein, which supports the point that differences between species must always be considered. In the present study, the increase in alkaline proteases could have been caused by the added effort of having to break the β(1–4) glycosidic bonds of chitin polymers, and especially its metabolites, because it was proven that chitinase activity is low in rainbow trout [90]. However, because chitinase (mostly located in the stomach) and chitobiase (mostly located in the intestine) activities were not measured, this would remain as a theory to encourage more future research.The results for amylase activity follows closely the tendency of a previous study [32]. It has been described that amylase activity increases with the amount of carbohydrates in the diet [91]. Moreover, Rapatsa and Moyo already proved [92,93] that higher levels of Imbrasia belina meal increased the levels of amylase, which was conferred to the remaining vegetable contents of Imbrasia belina, more than to the intrinsic components of that insect. Vegetables are the most common substrates for the feeding of insects. As part of the private knowledge of the insect provider, these data are not available in the present study, but the results in the amylase activity could be due, on the one hand, to different feeding habits of the insects, and on the other hand, to the different levels of wheat meal between H18 and T18 diets.3.4. Liver Intermediary MetabolismNo differences were found between treatments for any of the measured enzymes (FBPase, PK, GPT, GOT and GDH), which means that intermediary metabolism in the liver was mostly unaffected (Table 5). As stated previously, protein use (Table 3) was in general more efficient on T18 than on H18, which gives the idea that protein availability was better for fish fed with T18. The literature concerning the analysis of intermediary liver metabolism after a feeding trial with IMs is scarce, but supports the point that these ingredients do not disrupt the function of these enzymes [23,28,94,95]. In a previous experience [32] with similar diets and rearing conditions, but with 10% inclusion level of IMs, an increase in GOT activity was observed; in the present study, even though T18 showed a very similar trend (p-value = 0.059), the ANOVA did not reveal a significant difference. Considering that liver histology did not show differences either, this also supports the present data.3.5. Non-Specific Immune StatusThe immune status was evaluated in different tissues. Tumour necrosis factor-alpha (TNF-α) was determined as a pro-inflammatory indicator [96] in distal intestine, as a first immune barrier from which an immune response can be initiated [97], and in skin mucus, for its role in innate immunity and fish health [98]. In addition, different parameters in plasma related to non-specific immune responses were determined.No significant differences were found between treatments for TNF-α in distal intestine and skin mucus, or for lysozyme, esterase, anti-protease, alkaline phosphatase, peroxidase, or total immunoglobulins (IG) in plasma (Table 6).The only statistically significant difference was on acid phosphatase in plasma, where T18 showed the lowest level, being different to C (p < 0.05; Table 6). Acid phosphatase is related to tissue damage in other species [99,100]. IMs are known for causing varied effects on the immunological system of fish, while the precise mechanisms that produce them are still unknown [14,30,33,88]. In general, the immunostimulant effect of Ims is well accepted [30,101,102], and a frequent justification for this is the influence of chitin and its derivatives, together with the presence of antimicrobial peptides in insects [13,103]. However, a publication by Xu [104] described how the replacement of soybean oil with an ω-3 enriched insect oil modified the genetic expression of IL-1β, IL-10, and TNF-α on the liver and kidney of juvenile mirror carp, as well as the amount of serum lysozyme. Moreover, Kumar [88] carried out two parallel experiments, one based on the addition of HI meal, and one on the replacement of fish oil with insect oil, and described different results for the same immunological parameters and the same organs of rainbow trout. Because the immunological system is complex and multifactorial, the different components of IMs such as chitin, insect fat, or other that might not be considered, could lead to very different interactions with it. Even though these immunological benefits are usually attributed to IMs, its chitin, or both [30,101], concluding results will not be reached until the precise mechanisms involved are described.3.6. Liver Antioxidant Status and Fish Welfare IndicatorsWith the exception of GPx activity, which was higher for C than for IM-based diets (p < 0.05), T18 showed an overall more active antioxidant status. There were no differences for G6PDH or GR. SOD activity was higher in T18 than in C and H18 (p < 0.05). Higher levels of CAT activity were highlighted in T18 than in H18, and lower levels of MDA are described in T18 than in C diet (p < 0.05). No differences were found between treatments for glucose or lactate plasmatic levels (Table 7).The antioxidant system is a complex biochemical structure composed of several molecules and enzymes that fight against the derived toxicity of reactive oxygen species (ROS) resulting from cellular metabolic processes. Roughly, SOD, CAT, and GPx are involved as direct defensive mechanisms against ROS, while GR regenerates the substrate of GPx (glutathione) and G6PDH works on the maintenance of this system by providing NADPH, which is used as a fuel to allow the activity of GR. An imbalance between ROS production and antioxidant mechanisms derivate in oxidative stress, resulting in cellular membrane lipids damage. MDA is a product formed from the breakdown of polyunsaturated fatty acids due to lipid peroxidation, and it may be used as a marker of cellular damage [105]. Because all these molecules work together to prevent oxidative injury, it is not strange to find coincidences in their activities, such as the ones shown between SOD and CAT, higher on T18, as well as its consequential MDA decrease on T18. This would also be in consonance with the previously mentioned results of plasmatic acid phosphatase, because a more efficient antioxidant system should be reflected on a lesser amount of cellular damage. In general, the current bibliography strongly supports the idea that IMs help to prevent the derived toxicity of oxygen in two different ways: indirectly, by enhancing the antioxidant system, or directly, by preventing the oxidative damage itself. The first case can be easily recognized in those experiences where higher activities of antioxidant elements are highlighted, ideally, but not always, followed by a consequential decrease in the concentration of oxidative damage indicators (typically, MDA) when a significant amount of IM is added to the diet [21,30,31,32]. The second case can be more complicated, because the mechanisms that regulate the reduction of oxidative damage through the addition of IMs are not yet well known. As previously stated for the immunological system, a frequent hypothesis used to justify this involves the activity of chitin and its derivates in fish physiology, because it was described that this molecule could produce a direct scavenging effect on radical species, as well as an increase in intracellular glutathione [29]. The work of Moutinho [20] described a decrease in both SOD and CAT activities and a lower concentration of MDA on European seabass liver after an increasing amount of HI was added to the diet, suggesting this preventive effect. Moreover, the work of Sánchez-Muros on tilapia [106] described an increase in liver SOD activity, muscle ROS, and intestine ferric-reducing antioxidant power for a diet based on FM and soy meal, while two experimental diets with TM meals produced the opposite effects, even though one of them had the same amount of soy meal as the first, giving again the idea of a preventive effect. Interestingly, the work of Xu [104] highlighted an increase in SOD activity in the liver of juvenile mirror carp after the administration of HI oil, which suggests that chitin might not be the only element involved in the enhancement of the antioxidant system of fish after using whole IMs.Glucose and lactate are known to be indicators of animal welfare, because their levels in plasma are increased after stressful situations [107,108,109]. Other trials based on the evaluation of IMs as ingredients for fish compared the levels of plasmatic glucose, but in general, not many changes have been described [110,111]. The present study matched this case; on the one hand, this means that the fish homeostasis was correct and compensated even in those cases where other indicators (acid phosphatase and MDA) suggested the presence of tissue damage; on the other hand, IMs did not affect either of these animal welfare indicators in a positive way.3.7. Proximate Composition of the FilletThe proximate composition of rainbow trout fillets was appointed in Table 8. Moisture levels were lower for T18 than for C. Protein and ash levels were higher for insect-based treatments than for C diet, while fat showed no significant differences. Despite the lower values of phosphorus in IMs, the phosphorus content in the fillets was similar among treatments, so IMs were able to satisfy the nutritional requirements.In general, dry components were higher on fillets of fish fed with insect-based diets. Several studies have evaluated the composition of rainbow trout fillets after a growth trial with HI and TM IMs. Some of them described no differences in fish fillet compositions [62,67,112], while others described small changes in raw protein or lipids, the decrease on these parameters being more frequent than the opposite trend [61,65,113]. However, all the cited manuscripts describe experiences with bigger sizes of rainbow trout than those used for the present study. Our previous study [32] did not show any differences on fillet protein, fat, or ash, which means that the increased inclusion of HI and TM in the diets (18% vs. 5/10%) may have played an important role on these changes. Considering this, it is possible that smaller fish could have dealt differently with higher amounts of IMs, contrary to what was described for bigger fish.4. ConclusionsThis study shows the importance of adequately selecting a type of IM before its inclusion as an ingredient in aquafeeds. Although in terms of absolute values for growth performance, the use of HI or TM in feeds for rainbow trout was efficient, fish fed with TM grew better than fish fed with HI. These differences have been marked by the higher use of the protein and more active digestive function, supported with intestinal histological changes observed, particularly the increase in villi height for T18. It is also remarkable that a small increase in enterocyte height was described for insect-based diets, which could be related to the different absorption of insect fat.No changes were noticed for liver histology or intermediary metabolism. The antioxidant and immunological systems suffered a slight activity improvement for insect-based diets reflected on the decrease in tissue damage indicators (MDA and acid phosphatase), but this did not modify the overall health and welfare status of fish. Although more research is encouraged to isolate and identify the specific physiological mechanisms that make IMs improve the performance of both the antioxidant and the immunological systems of fish, this study supports the idea that IMs act as potential functional ingredients.Minor changes in the composition of the fillets were observed, with a higher amount of protein in fish fed with insects. More research is encouraged to elucidate the long-term feeding effects.

animals : an open access journal from mdpi

10.3390/ani13111772

PMC10252083

Climate change has strengthened the incidence of heat waves, causing significant losses in livestock units productivity and farm profit. Among different food producing animals, rabbits perform a crucial role in meat production in many countries worldwide. However, these animals are more sensitive to thermal stress compared with other livestock species, mainly due to the lack of functional sweet glands. Thus, maintaining rabbit farming systems under recent climate changes becomes a challenge. Phytochemicals may present an effective intervention to mitigate heat stress (HS) effects on rabbits via different modes of action. Recently, more attention has been paid for such substances, particularly, with the emergence of nanotechnology approaches. This technology has recently employed to improve the availability, solubility, and efficacy of phytochemicals, overcoming natural biological barriers and some industrial obstacles. In this sense, we explored the protective role of Origanum majorana (marjoram essential oil nanoemulsion, MEONE) against heat stress in growing rabbits, considering growth performance, immunity, and inflammatory and oxidative stress pathways. In addition, the economic efficiency of this nanotechnology-based intervention was estimated. The results indicate that the addition of 400 mg/kg diet of MEONE inhibited inflammatory biomarkers, DNA damages, and oxidative stress caused by heat stress, enhanced the growth performance, and relative economic efficiency of newly weaned rabbits.

With the recent trend of global warming, HS-instigated diminishing could extremely jeopardize animal health, productivity, and farm profit. Marjoram essential oil (MEOE) is a worthy source of wide range phytogenic compounds that may improve heat tolerance, redox and inflammatory homeostasis, and immunity of newly weaned rabbits, specifically if included in the diets in a nano form. One hundred newly weaned rabbits were randomly distributed into four homogeneous groups. The first group (control group) included rabbits that received basal diet without supplementation. In contrast, the other three groups included rabbits that received basal diets supplemented with 200 (MEONE200), 400 (MEONE400), and 800 (MEONE800) mg MEONE/kg diet, respectively. Among MEONE-treated groups and control groups, MEONE400 group showed the highest (p < 0.001) growth performance traits, including final body weight, average daily gain, feed efficiency, and the performance index. Compared to the control, all MEONE-supplemented groups possessed lower rectal temperatures and respiration rates, recording the lowest values in the MEONE400 group. The oxidative stress biomarkers and immunoglobulins G and M were significantly improved in the MEONE400 and MEONE800 compared with the control and MEONE200 groups. The addition of MEONE (400 or 800 mg/kg) decreased the concentrations of serum interleukin-4 (p = 0.0003), interferon gamma (p = 0.0004), and tumor necrosis factor-α (p < 0.0001) but significantly elevated (p < 0.001) the activity of nitric oxide, amyloid A and lysozyme. Liver functions (lower concentrations of liver enzymes) were significantly improved in all MEONE-treated groups compared to the control group. There was a considerable significant effect of dietary supplementation of MEONE400 on economic efficiency. In conclusion, the addition of 400 mg/kg to the diets of newly weaned rabbits can be recommended as an affective intervention to mitigate the negative impacts of HS.

1. IntroductionThe whole temperature in the earth is predictable to be increased by 4 °C in the subsequent 100 years [1]. Increasing global temperatures will have deleterious consequences on the livestock sector by reducing the quality and quantity of feedstuff, augmented competition for natural resources, deprivation of biodiversity, the spread of livestock syndromes, and augmented heat stress (HS) [2].Among these obstacles, HS seems the largest threats fronting animal populations in current times, and the deleterious effects of HS have begun to alter animal health, physiology, and behavior, ultimately with negative consequences for animal fecundity and survival [1]. Rabbits are categorized as the best meat producers due to their high meat quality, superior growth rates, and feed efficacy [3]. In addition, rabbit breeding is a veritable approach of decreasing poverty due to its low cost of breeding in developing countries. Moreover, the consumption of rabbit meat does not contravene any religious or social taboos. Global rabbit meat production is presently assessed at 1,482,441 tons equivalent carcasses [4].In this scenario, there is an emergent concern in escalating the rabbit industry as an economical and environmentally sustainable livestock farming system [5]. With this regard, HS has substantial impacts on rabbit production and its physiological aspects, especially in tropical and subtropical areas. According to many reports, the thermoneutral zone of rabbits is 15–25 °C; while higher temperatures cause discomfort and/or stress conditions [6,7]. Ardiaca et al. [8] indicated that the body temperature of rabbits can increase up to 40 °C in stressful situations without being pathological. However, above 40 °C, cell membranes start to be destroyed by protein denaturation [9]. From 24 °C onwards, weaned rabbits, during the fattening period, start to have respiratory problems, with fatigue, increased heart rate, lack of appetite, and decreased basal metabolism. Thus, newly weaned rabbits are expected to be more sensitive to HS, as they usually suffer weaning psychological shock and severe physiological/nutritional changes post-weaning, making them more susceptible to negative impacts of HS [7]. Many reports have shown that HS can negatively affect economic traits, including growth, feed efficiency, and meat quality as well [10,11]. In addition, HS may interfere with growing rabbits welfare and health by impairing the immune system and redox status and evoking inflammatory reactions [12,13]. For this concept, it is important to bring our attentions toward establishing good management for growing stage in rabbits. Several approaches have been employed to diminish the detrimental effects of HS in growing rabbits’ improve health, welfare and meat quality [6,14,15], and maximize the rabbit industry economy.Origanum majorana (marjoram) is one of the most prevalent condiments applied for many food preparations, agricultural, pharmaceutical, and biomedical uses, aiming its essential oils [16]. The marjoram essential oil nanoemulsion (MEOE) has strong antioxidant, anti-inflammatory, and antimicrobial abilities [17]. It has considerable amounts of terpinenes (e.g., terpinen-4-ol, trans-sabinene-hydrate, γ-terpinene, α-terpinene, carvacrol, and thymol) [18,19,20].Marjoram essential oil alleviated the renal damages induced by ivermectin [21] via levitation antioxidant capabilities, lessening inflammation in rabbits. Moreover, Origanum majorana has beneficial effects on feed efficiency, antioxidant abilities, growth indices, and immunological reactions in lamb and fish [22,23,24,25].Nevertheless, it is well known that the use of essential oils as a dietary supplementation is restricted due to their limited permeability, solubility, bioavailability, and storage instability. Recent studies have shown that transforming phytochemicals into nano form can confer them several biological and industrial advantages, making them more kinetics in the biological systems and more suitable for handling under industrial conditions [26,27].Based on previous findings, we hypothesized that the dietary inclusion of MEONE may have protective effects against HS by boosting the growth and improving the health of newly weaned growing rabbits. Moreover, we considered the economic efficiency of the supplement to prove the cost efficiency of the supplementation when a technology, such as nanoemulsion, was used.2. Materials and Methods2.1. Preparation of Marjoram Essential Oil NanoemulsionThe MEOE was purchased from the pure life company, Giza, Egypt. A single layer of MEONE, oil-in-water, was prepared [28]. In brief, nanoemulsions were prepared from MEOE 2.5 mL and surfactant (tween 80) mixture by slowly dropping water (up to 10 mL) appointed by a magnetic stirrer at 25 °C. The dropping rate of water was preserved persistent at around 1.0 mL/min. Then, the emulsion was scattered for 30 min through an ultrasonic bath (Sonix, Springfield, VA, USA, SS101H230). It was additionally homogenized by employing an ultrasonic probe (Serial No. 2013020605, Model CV 334) involved to a homogenizer (Sonics Vibra-cell™, Model VC 505, Inc., USA) under the subsequent conditions: amplitude: 60%, timer: 5 min, and pulser: 1 s ON/1 s OFF to yield nano emulsions.2.2. Physichochemical Properties of MEONEThe internal morphology of the freshly prepared MEONE was visualized using transmission electronic microscope (TEM, JEOL JEM-2100, JEOL Ltd., Tokyo, Japan) at 160 kV. Digital Micrograph and Soft Imaging Viewer software (Gatan Microscopy Suite Software, version 2.11.1404.0) were utilized to complete the image capture analysis process. Nano emulsion average vesicular sizes (Z-average), polydispersity index (PDI) and surface charge of nanoemulsion particles (Z-potential) were measured using Zetasizer Nano ZS analyzer (Malvern Instruments, Malvern, UK).2.3. Animals, Housing and Experimental DesignAll the procedures that used the rabbits for this trial were approved by the scientific committee of the Institutional Animal Care and Use Committee (IACUC) of Zagazig University (Approval number: ZU-IACUC/2/F/367/2022).One hundred weaned New Zealand White (NZW) male rabbits (707 ± 14.35 g, six weeks of age) were involved in this study. The weaned rabbits were acquired from the Rabbit Research Unit, Faculty of Agriculture, Zagazig University, Egypt. All rabbits were kept in the same environmental and management conditions. The animals were housed in galvanized wire battery cages (50 × 45 × 40 cm3) provided with conventional feeders and an automatic system of nipple drinkers in a well-ventilated rabbitry. The growing rabbits were randomly distributed into four groups, the control group, where rabbits received basal diet without supplementation, and three treated groups that included rabbits receiving basal diets supplemented with 200 (MEONE200), 400 (MEONE400), and 800 (MEONE800) mg MEONE/kg diet, respectively. The experimental period lasted eight weeks (July and August, presenting summer season). Growing rabbits were fed with basal diets formulated to congregate the nutrient requirements as endorsed by NRC [29]. The ingredients and chemical composition of the basal diet is presented in Table 1. The diet ingredients are free of antibiotics. During the experimental period, ambient temperature (AT) and relative humidity (RH) were assessed using a hygrothermograph (ST-50A, SEKONIC, Tokyo, Japan) placed in the animal facility. The hygrothermograph was placed above (30–50 cm) the cages for measuring the RH and AT at the constant time every day. These parameters (RH and AT) were recorded daily at 14.00 PM. Both variables were used to calculate the temperature humidity index (THI) values and thereby detect the severity of HS during the experimental period. The following question was used for detecting the THI values THI = db − [(0.31 − 0.31(RH)] × [(db °C − 14.4)], where db is the dry bulb temperature (AT) in Celsius and RH is the percentage relative humidity. The THI values were subsequently classified as follows: <27.8 = absence of heat stress; 27.8 to 28.9 = moderate heat stress; 28.9 to 30.0 = severe heat stress, and >30.0 = extremely severe heat stress [7]. The previous method was used to detect the severity of HS in rabbits.2.4. Growth Performance and Physiological ResponsesLive body weight (LBW) of individual rabbits was determined weekly, and average daily weight gain (DWG) was calculated according to the following equation: DWG = [(Initial body weight − Final body weight)/duration]. Feed intake (grams per rabbit) was weekly determined during the whole experimental period. Feed intake (FI) was determined by weighing the residuals of daily offered feed. The feed conversion ratio (FCR; g feed/g gain) and performance index (PI, % = [(FBW, Kg × 100)/FCR] were estimated accordingly.Rectal temperature (RT; °C) and respiratory rate (RR; breath/min) were weekly measured. RT was assessed by inserting a digital thermometer (Model No.: YF-160A, TYPE-K) to a depth of approximately 4 cm into the rectum for 1 min. At the same time, RR was evaluated by visually calculating the number of movements of the flanks of the rabbits in a resting position for 1 min with a stopwatch. The thermometer was disinfected between each animal via apply rubbing alcohol (60%) with a cotton swab.2.5. Slaughtering and Carcass CharacteristicsAt the end of the treatment, ten rabbits from each experimental group were randomly picked out and fasted for 12 h and immediately slaughtered after weighing them individually. We followed the Islamic method for slaughtering the rabbits (we used sharpness knife qualified person for slaughtering, then used a special words during the slaughtering), as indicated in a recent study by Bouzraa et al. [30]. The viscera, tail, and pelt were removed after complete bleeding, and then the carcass and its components were weighed as total edible parts. The edible giblets (heart, liver, and kidneys) were weighed as a percentage of live body weight. Similarly, the dressing percentage was calculated by dividing the hot-dressed carcass weight by pre-slaughter weight and expressed as a percentage.2.6. Blood Sampling and AnalysesThe blood samples were collected form six rabbits/group the marginal ear vein in sterilized blank tubes. The blood samples were left for 2 h, allowing clot formation, then the serum samples were obtained by centrifugation at 3000× g for 15 min (T32c; Janetzki, Wallhausen, Germany). Serum samples were detached and preserved at −20 °C for further analyses.For assessing the redox status of rabbits, the activities of Total antioxidant capacity (TAC), superoxide dismutase (SOD), and glutathione (GSH) were determined using specialized quantitative sandwich ELISA kits (TAC: MBS8807700, SOD: MBS8807589, and GSH:MBS9718983, My BioSource, San Diego, CA, USA) following the guideline instructions. Moreover, the concentrations of malondialdehyde (MDA) and protein carbonyl (PC) were determined as lipid and protein oxidation biomarkers, respectively, using a specialized ELISA kit (MDA: MBS8806802, PC: MBS2601439, My BioSource, San Diego, CA, USA). At molecular level, the concentration of n 8-hydroxy-2′-deoxyguanosine (8-OHdG) as an indicator for oxidative DNA impairment was assessed using a commercial competitive sandwich ELISA kit (Trevigen, Gaithersburg, MD, USA).For immune function evaluation, the concentrations of immunoglobulin M (IgM) and G (IgG) were assessed by ELISA kits [31], and the concentration of myeloperoxidase (MYO: MBS724170, My BioSource, San Diego, CA, USA) was also determined.For inflammatory assessment and liver function, the levels of interferon gamma (IFNγ), tumor necrosis factor (TNF-α), and interleukin 4 (IL-4) in rabbit serum were determined using commercially available sandwich ELISA kits (IFNγ: MBS2601171, TNF-α: MBS7612133, and IL-4: MBS733925, My BioSource, San Diego, USA,). The nitic oxide (NO) and lysosome activity were detected as described by the methodology of Rajaraman et al. [32] and Sun et al. [33], respectively. The concentration of amyloid A was assessed with commercially available sandwich ELISA kits (Biosource, Camarillo, CA, USA). Glutamyl transferases (GGT) and lactate dehydrogenase (LDH) were colormetrically analyzed using kits obtained from Bio-diagnostic Company (Giza, Egypt). All laboratory analyses and biochemical assessments followed the ISO/IEC 17025 protocols (the last version in 2005). More details on the standards and yields (detection range, sensitivity, and inter- and intra-assays precision) of ELISA kits used in the experiment are presented in Table S1.2.7. Economic EfficiencyEconomic efficiency (EE, %) comprising both costs and net revenue was estimated. The dominant prices by USD of the experimental diets and rabbit’s meat in Egypt during the experimental period were as follows: price of 1 kg/live body weight on selling was 2.5 USD, 1 kg a feed cost was 0.33 USD, and 1 mL MEONE was 0.031 USD. All collected prices were calculated form the local market, then adjusted as the exchange rate from Egyptians Pounds to Dollars.These data were used to estimate the following economic items as follows: Total feed costs = total FI per rabbit × price/kg. Net revenue = price of rabbit-total feed cost, and EE = net revenue/total feed cost.2.8. Statistical Model and Analysis ProcedureThe Levene and Shapiro–Wilk tests were conducted in order to check for normality and homogeneity of variance [34]. One-way anova of statistical analysis system (Proc Anova; SAS, 2012 version 8, Cary, NC, USA) was used for assessing growth performance, feed utilization, blood biochemical, oxidative DNA marker, and economic efficiency, Multiple comparisons among means were carried out by the Duncan’s Multiple Range Test. Results were expressed as means ± SE. statistical significance was accepted at p-value < 0.05. Figures were fitted by the Graph-Pad Prism software 9.0 (Graph Pad, USA).3. Results3.1. Meteorological ParametersThe overall means of ambient temperature (AT), relative humidity (RH), and temperature–humidity index (THI) during the whole experimental period were 30.91 ± 0.14 °C, 72.01 ± 0.64%, and 29.47 ± 0.11, respectively (Table 2). The THI obtained in the present study indicated that the growing rabbits suffer from severe heat stress.3.2. Characterization of Marjoram Essential Oil Nanoemulsion FormationThe TEM image for MEONE is presented in Figure 1A. The image shows spherical particles morphology of the tested nano-emulsion oils with little or no aggregation identified. The mean of particles size was 215 nm (Figure 1B). The values of PDI and Z-potential were 0.177 and −15.6 mV, respectively (Figure 1C).3.3. Growth Performance, Feed Utilization, and Physiological ResponsesResults shown in Table 3 revealed that the FBW was significantly increased in the MEONE200 and MEONE400 groups compared to the control group, whereas MEONE800 resulted in an intermediate value. Treatment with MEONE400 resulted in the highest significant PI and ADG. This improvement was observed at week 10 of age and afterwards, compared to other groups and control group (Table 3). However, FI was not affected by the treatment; FCR (lower FCR) was significantly improved in all treated groups compared with control group (Table 3). Treatment with MEONE400 significantly reduced RT and RR compared to the control group (Figure 2A,B).3.4. Carcass TraitsResults in Table 4 show the carcass traits of heat-stressed rabbits fed different levels of MEONE compared with control rabbits. Only the dressing percentage and liverrelative weight were significantly affected by the dietary treatment (p = 0.0350 and 0.0296, respectively). The dressing percentage was significantly higher in the MEONE200 and MEONE400 groups than the control group (p < 0.05). Meanwhile, the relative weight of liver and total edible giblets were significantly higher in MEONE400 groups than their counterparts in the control group (p < 0.05).3.5. Redox StatusThe results in Table 5 show significant effects of dietary inclusion of MEONE on both of oxidative and anti-oxidative stress variables of heat-stressed growing rabbits, the lowest values of MDA and MYO were recorded in the MEONE800 group; however, the lowest values of PC were observed in the MEONE400 group compared to the HS-group (p < 0.05). With respected to the anti-oxidative stress markers, the elevated values of SOD and GSH were observed in the MEONE400 and MEONE800 groups, respectively. Meanwhile, the values of TAC maximized in the MEONE400 group. Despite the MEONE level, treatment with MEONEDNA significantly decreased oxidative marker (Ohdg) compared to the control group (Table 5).3.6. Immunity and Inflammatory Responses and Liver FunctionDietary supplementation with different levels of MEONE significantly enhanced the IgG and IgM levels (p < 0.001), recording the greatest values in the MEONE800 group (p < 0.05; Table 6). With regard to inflammatory cytokines, the addition of MEONE declined the concentrations of IL-4 (p = 0.0003), IFN Y (p = 0.0004), TNF-α (p < 0.0001), and amyloid A (p < 0.0001) in the serum, but significantly increased the activity of NO (p = 0.0001), and LZM (p < 0.0001; Table 6). Heat stressed rabbits fed diets supplemented with MEONE (200, 400, or 800 mg MEONE/kg) had lower GGT and LDH compared to non-supplemented rabbits (Table 6).3.7. Economic EfficiencyResults of economic efficiency are shown in Table 7. There were pronounce significant effects of dietary supplementation with MEONE on economic efficiency; the highest values of NE, economic efficiency, and relative economic efficiency were achieved by rabbits given a diet enriched with 400 mg MEONE, followed by 200 mg MEONE and 800 mg MEONE, while the control group achieved the lowest corresponding values.4. DiscussionIn tropical/subtropical regions, the harmful effects of HS on livestock are not easy to mitigate, specifically when the high ambient temperature is combined with high humidity. Therefore, breeders may need to apply more than one strategy (controlling housing system, nutritional manipulation, selection of heat tolerance animals) to protect rabbits against negative consequences of HS. Among HS alleviating strategies, nutritional interventions can present practical and effective solutions either applied alone or in combination with other alleviating strategies [7]. In this sense, several studies recommended the enrichment of diets of heat-stressed growing rabbits with phytogenic supplementation, including essential oils. Essential oils, such as marjoram, have many therapeutic activities, including antioxidants, anti-inflammatory, antimicrobial, antiviral, spasmolytic, carminative, and hepatoprotective effects [35], making them an ideal feed supplement for early weaned rabbits. Despite the beneficial biological effects of essential oils on animal health, these photogenic substances have some limitations related to its bioavailability and other industrial limitations [36,37]. Most essential oils are lipophilic molecules, and therefore their absorbance along the gastrointestinal tract can be improved if they are offered as emulsions. In this study, we used oil-in-water single-layer nanoemulsion procedure to fabricate MEOE in nanoform. The results of the physichochemical properties (size ≈ 200 nm and negative zeta potential, Z-potential ≈ −11 mV) confirmed the relevance of the resultant nanoemulsion for better intestinal absorbance and cellular uptake, increasing the bioavailability of the active components of MEOE. In this sense, small size, low-magnitude negative charge, and moderate hydrophilicity help nanoparticles pass through the small intestinal mucus layer more easily [38]. It has been found that the anionic nanoparticles induce tight junction relaxation, increasing intestinal permeability. This permeation-enhancing effect is a function of nanoparticle size and charge, with smaller (≤200 nm) and more negative particles conferring enhanced permeability [39].In the present study, the THI value indicated that the growing rabbits are exposed to severe heat stress (28.9 to <30.0; [7]. Nevertheless, MEONE-supplemented rabbits possessed better heat tolerance capacity, growth performance, and health status (lower oxidative and inflammatory stress indicators). The administration of MEONE decreased the RT of growing rabbits exposed to serve heat stress, what in general agreement with several previous studies observed significant effects of natural antioxidants in decreasing the rectal temperature in rabbits due to its contents from bioactive components, such as flavonoids and flavones [39,40]. Moreover, in hot environments, rabbits begin to increase their respiratory activity in order to maintain their heat balance and, thus, losing more body heat throughout t evaporation from the respiratory tract, which explains the higher respiratory rate observed in control group compared to MEONE treated group, which also suggest the action of MEONE in heat regulation. These findings can also explain the improvement in growth performance of MEONE-supplemented rabbits, as these rabbits did not expenditure energy in heat regulation process, respiratory evaporation; instead, most of available energy is directed to the important biological functions, such as growth. This can be confirmed by the findings of our study as the MEONE-supplemented rabbits had better FBW, ADG, PI, and FCR compared to the control rabbits, recording the highest significant values with 400 mg MEONE/kg diet. The current results coincide with the results of Abdelhadi et al. [41], who showed significant effects of adding essential oils and nanoemulsified essential oils (garlic, pomegranate, and tea tree essential oils) to rabbit diets on growth performance and feed utilization. The improved growth performance in this study could be linked to the enhanced feed utilization and the nutrients digestibility. Essential oils perform a vital role in boosting the digestion through stimulating the secretion of digestive enzyme and fluids and improving guts eubiosis [42]. In this context, the relative weight of liver was significantly increased by MEONE400 supplementation. These results may signalize that the addition of MEONE in rabbit diets leads to improve the health status of liver, which performs a decisive role in synthesis of blood protein and other enzymes linked to heat tolerance [6,43,44].It is well known that the high AT commonly upregulating the synthesis of free radicals and cytokines, causing oxidative and/or inflammatory stress [45]. Therefore, in this study, we have focused on studying the changes in these two pathways in heat-stressed and MEONE-supplemented rabbits. Giving in mind that essential oils can alleviate the negative effects of HS through controlling these two major cells damaging pathways, oxidative and inflammatory pathways [6,14,41,46,47,48].The current study indicated that the fortification of heat-stressed rabbit diets with MEONE at high doses (800 mg/kg diet) decreased oxidative stress by lowering MDA and MYO levels by 30.59 and 87.31%, respectively. However, the concentration of PC decreased by 46.41% in the MEONE400 treated group compared to the control group. In the same manner, the evidence of anti-oxidative stress improved by the addition of MEONE to rabbit diets; the levels of TAC and GSH maximized in the high-dose treated group, while the activity of SOD maximized in MEONE400 treated group. In accordance with the present results, several former studies indicated that the essential oils with its bioactive component, including flavonoids and flavones, could enhance the oxidative stability and effectually deferring the lipid oxidation or proteins (protein carbonyl) and other nutrients by inhibiting the diffusion of oxidation reactions [49,50]. Interestingly, it was observed that the protective effects of MEONE extended also at molecular level, as MEONE administration reduced the concentrations of serum 8-OHdG, a new biomarker for the oxidative damage of DNA. Thus, it can be concluded that MEONE confers animals an integrated oxidative defense system that helps them to easily tolerate with oxidative stress evoked by severe heat stress.Under heat stress conditions, a significant impairment in immune function can be observed. This is mainly mediated by the hypothalamic-pituitary-adrenal axis [51], which evokes the glucocorticoid functions as an anti-immune response element [52]. The increase in glucocorticoid concentration impairs both cellular and humoral immune systems. Consequently, the heat stress causes major losses in rabbit production due to the deterioration in the immune function that makes rabbits susceptible to pathogens [7]. In this study, the additions of MEONE in the diets of early weaned growing rabbits improve cellular and humoral immune systems. Rabbits supplemented with 400 or 800 mg MEONE/kg diet had significantly improved levels of IgG and IgM (cellular immunity) compared to rabbits in other groups. Moreover, the supplementation of rabbit diets with MEONE resulted in significant improvements in humoral immunity, as indicated by the increased lysosomal activity and concentration of amyloid A. The improved lysozyme activity may contribute to the elimination of pathogens because of its enzymatic degenerative potential [48,53], and amyloid A has a significant immunological activity either by being chemotactic for mast cells and neutrophils or by stimulating the synthesis of cytokines [54]. These positive effects of MEOE on immune system can be attributed to its wide biological activities, including anti-allergic, anti-inflammatory, antiviral, and antimicrobial activities [55].Regarding inflammatory pathways, it is interesting to note that MEONE had the ability to suppress the de novo of inflammatory cytokines and, thus, inflammation reactions. High AT leads to an increase in the apical contents of pro-inflammatory cytokines, such as IFNγ and IL-4, which, in turn, raise the permeability of intestine of the animal to pathogens [56]. According to the findings of this study, MEONE-treated rabbits had a significant decrease in the pro-inflammatory cytokines (IL-4, TNF-α, and IFNγ), indicating the potential anti-inflammatory activity of MONE. On the other hand, MONE treatments resulted in signification increases in NO concentrations. Under physiological concentrations, this small molecule can act as an anti-inflammatory agent, neurotransmitter, immune regulatory factor, and vasodilator [57]. Recently, it has been concluded that NO can show an indispensable role in defending animals during exposure to adverse pathogen attacks or environmental conditions throughout the enhancing of non-specific immunity [14,58]. Moreover, the vasodilation role of NO can partially explain the improved heat tolerance ability (lower RT and RR) of MEONE-supplemented rabbits. The re-partitioning of the blood toward peripheral blood vesicles (e.g., ear blood vesicles in rabbits) is an important physiological process for body heat loss. Interestingly, the present results indicated that the high level of NO in the blood serum of growing rabbits treated by MEONE is associated with increases in the antioxidant markers, such as GSH and TAC. This supports that the NO concentrations were at physiological levels, as NO is classified as a free radical as well.Our results indicated the safety of the MEONE to growing rabbits as indicated by the liver function. High levels of gamma glutamyl transferase (GGT) in the blood may be a sign of liver impairment (disease or damage to the bile ducts). Former studies have been indicated that high ambient temperature elevates the levels of GGT may be due to disturbance of enzyme synthesis, damage in the liver, or changes in the permeability of hepatic cells membrane, which causes the enzymes to leak into the bloodstream [59]. In the current study, MEONE has demonstrated a healthy defensive property against heat stress-induced hepatotoxicity by reducing the mentioned enzymatic activity.It is known that under subtropical conditions, HS has adverse impacts on the growth performance of rabbits, leading to major economic losses due to high rates of animals mortality, low feed efficiency, and costs related to health management [60]. In this study, the inclusion of MEONE in the diets of growing rabbits contributes to raising the economic values of the supplemented diets, making the fattening process more profitable.5. ConclusionsTo summarize, exposing growing rabbit to heat stress negatively affect growth; feed utilization; meat quality; health status, mainly via the elevation of oxidative stress and inflammatory responses; and the impairment of immune system functions. Under the conditions of our study, supplementation with 400 mg MEONE/kg diet improved growth and feed efficiency and the profitability. Moreover, serum antioxidant and anti-inflammatory capacities were improved, which are associated with adequate immune functions. These effects may be due to the improved bioavailability of MEOE after fabricating in a nano form. Further advanced studies are needed to explore the manner by which these nanoparticles are absorbed and circulating to emphasize the actual need for nanomulsion technology in phytogenic feed additives processing. Moreover, active components of MEOE and their roles at cellular/molecular levels as active thermoregulatory agents need more investigations.

animals : an open access journal from mdpi

10.3390/ani13081299

PMC10135208

Although glucocorticoids are frequently used for a variety of purposes, the effects and mechanisms of glucocorticoids on platelets are not fully understood. The present study investigates the effect of prednisolone and its mechanism of action on the regulation of platelet function. The results demonstrated that prednisolone affects platelet function by inhibiting thromboxane A2 (TxA2) generation through the regulation of cPLA2 phosphorylation, providing knowledge of glucocorticoids in coagulation and bleeding disorders.

Glucocorticoids have been commonly used in the treatment of inflammation and immune-mediated diseases in human beings and small animals such as cats and dogs. However, excessive use can lead to Cushing’s syndrome along with several thrombotic and cardiovascular diseases. Although it is well-known that glucocorticoids exert a significant effect on coagulation, the effect of cortisol on platelet function is much less clear. Thus, we aimed to study the effects of prednisolone, one of the commonly used glucocorticoids, on the regulation of platelet function using murine platelets. We first evaluated the concentration-dependent effect of prednisolone on 2-MeSADP-induced platelet function and found that the 2-MeSADP-induced secondary wave of aggregation and dense granule secretion were completely inhibited from 500 nM prednisolone. Since 2-MeSADP-induced secretion and the resultant secondary wave of aggregation are mediated by TxA2 generation, this result suggested a role of prednisolone in platelet TxA2 generation. Consistently, prednisolone did not affect the 2-MeSADP-induced aggregation in aspirinated platelets, where the secondary wave of aggregation and secretion were blocked by eliminating the contribution of TxA2 generation by aspirin. In addition, thrombin-induced platelet aggregation and secretion were inhibited in the presence of prednisolone by inhibiting the positive-feedback effect of TxA2 generation on platelet function. Furthermore, prednisolone completely inhibited 2-MeSADP-induced TxA2 generation, confirming the role of prednisolone in TxA2 generation. Finally, Western blot analysis revealed that prednisolone significantly inhibited 2-MeSADP-induced cytosolic phospholipase A2 (cPLA2) and ERK phosphorylation in non-aspirinated platelets, while only cPLA2 phosphorylation, but not ERK phosphorylation, was significantly inhibited by prednisolone in aspirinated platelets. In conclusion, prednisolone affects platelet function by the inhibition of TxA2 generation through the regulation of cPLA2 phosphorylation, thereby shedding light on its clinical characterization and treatment efficacy in dogs with hypercortisolism in the future.

1. IntroductionPrednisolone is one of the most commonly used glucocorticoids in the treatment of various diseases including inflammatory and immune-mediated diseases. Prednisolone has the ability to reduce the recruitment of inflammatory cells including eosinophils, lymphocytes, mast cells, and dendritic cells by reducing the production of chemotactic mediators and adhesion molecules and the ability of inflammatory cells to survive [1]. Due to the high efficacy of glucocorticoids, the prescription rate has climbed high during the past years, leading to excessive glucocorticoid abuse that raises blood cortisol levels, resulting in iatrogenic Cushing’s syndrome [2,3,4,5].Cushing’s syndrome is a disorder characterized by polyuria, polydipsia, muscle and skin atrophy, weight loss, truncal obesity, and hair loss [6]. Excessive use of glucocorticoids in thrombotic, hemostatic, and cardiovascular diseases is associated with increased morbidity and mortality [7]. Glucocorticoid medication has been shown to increase the incidence of hypertension, hyperglycemia, and hyperglyceridemia leading to coronary and ischemic heart disease, heart failure, and unexpected death [8,9]. Moreover, high levels of glucocorticoids have been shown to elevate levels of von Willebrand factor (vWF), anti-hemophilic factor, fibrinogen, plasminogen activator inhibitor-1, and platelet count [10,11,12], which can result in embolic disorder and further increase the risk of the coagulative disorder [13]. The pathogenesis and clinical symptoms of Cushing’s syndrome are similar in both humans and dogs. However, the incident rate is 1000 times higher in dogs compared to humans [6]. As a result, Cushing’s syndrome-induced thromboembolic complications are around four times more common in dogs which leads to a four times higher mortality rate, making it a disease of interest in veterinary medicine [14].It has been well-established that platelets play a crucial role in thrombosis and hemostasis and have been known to play a vital role in the development of blood-related disorders including diabetes and cardiovascular diseases in humans. Various signaling mechanisms are involved in the activation of platelets. Initially, platelets are activated by the exposure of collagen and vWF from the extracellular matrix upon vascular damage. Activated platelets further release adenosine diphosphate (ADP) and generate thromboxane A2 (TxA2) that enables the recruitment of circulating platelets, activates the recruited platelets, and forms the hemostatic plug in the presence of thrombin by converting fibrinogen to fibrin. Collagen and vWF induce platelet activation by the glycoprotein (GP) VI-mediated signaling pathway through phospholipase C (PLC) γ2 activation [15]. In contrast, G protein-coupled receptor (GPCR) agonists ADP, TxA2, and thrombin cause platelet aggregation by the activation of the PLCβ pathway. ADP stimulates Gq-coupled P2Y1 and Gi-coupled P2Y12 receptor-mediated signaling pathways [16]. Likewise, thrombin activates platelets by coupling to Gq and G12/13 through protease-activated receptors (PARs) [17]. TxA2 activates platelets through the thromboxane prostanoid (TP) receptor by coupling to Gq and G12/13 [18,19]. In addition, ADP-induced TxA2 generation is regulated by calcium binding to cytosolic phospholipase A2 (cPLA2) [20,21]. Thrombin-induced TxA2 generation is mediated by the P2Y12 receptor [22]. The generated TxA2 then causes w positive-feedback effect on platelet functional response [23].Several studies have suggested the occurrence of hyper-coagulation in patients with Cushing’s syndrome whose blood cortisol levels are high [12,24,25]. The study of Manetti et al. has shown that venous thromboembolism occurred in 7.5% of Cushing’s syndrome patients, whereas it did not occur in patients with remission of the disease [26]. Additionally, glucocorticoid administration is regarded as a first-line treatment option in several bleeding disorders, including immune thrombocytopenic purpura and acquired hemophilia A [27]. In contrast, in vitro studies using platelet-rich plasma (PRP) have shown that high glucocorticoid levels exert an inhibitory effect on platelet function [28,29,30]. Until now, the mechanisms involved in the effects of glucocorticoids on the production of thromboxane in platelets have not been elucidated.In this study, we demonstrated the effect of prednisolone on platelet function and its underlying mechanism using washed murine platelets. We have shown that prednisolone inhibits 2-MeSADP-induced platelet secretion and the resultant secondary wave of aggregation. In aspirinated platelets, prednisolone had no effect on the 2-MeSADP-induced platelet aggregation as well as secretion, suggesting that prednisolone exerts its antiplatelet effect by regulating TxA2 generation. We have further shown that 2-MeSADP- and thrombin-induced TxA2 generation and 2-MeSADP-induced cPLA2 phosphorylation are inhibited in the presence of prednisolone. In conclusion, prednisolone affects platelet function by inhibiting TxA2 generation through the regulation of cPLA2 phosphorylation.2. Materials and Methods2.1. Materials2-MeSADP, thrombin, apyrase, prostaglandin E1 (PGE1), sodium citrate, prednisolone, and acetylsalicylic acid (ASA) were purchased from Sigma (St. Louis, MO, USA). Anti-phospho-cPLA2 (Ser505), anti-cPLA2, anti-phospho-ERK (Thr202/Tyr204), and anti-total-ERK antibodies were purchased from Cell Signaling Technology (Beverly, MA, USA). HRP-linked secondary antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Thromboxane B2 (TxB2) ELISA kit was purchased from Enzo Life Sciences (Exeter, UK). All other reagents were of reagent grade.2.2. AnimalsThirty C57BL/6 mice were purchased from Orient Bio, Inc. at the age of eight weeks (Seongnam-si, Gyeonggi-Do, Republic of Korea) and housed in a semi-pathogen-free facility. No more than 5 mice were housed per cage and mice were freely provided with food and water under constant conditions of temperature (23 °C) and humidity (60%). All animal procedures were approved by the Chungbuk National University Animal Ethics Committee (CBNUA-873-15-02).2.3. Platelet PreparationBlood was collected via cardiac puncture from healthy mice in 3.8% sodium citrate as described previously [31]. PRP was obtained by centrifugation at 100× g for 10 min at room temperature (RT) and 1 µM PGE1 was added. For aspirin treatment, the PRP was treated with 1 mM ASA for 30 min at 37 °C. By centrifuging plasma at 400× g for 10 min, platelet pellets were obtained from the plasma and re-suspended in Tyrode’s buffer (pH 7.4) with 0.05 units/mL of apyrase. The platelet count was adjusted to 1 × 108 cells/mL.2.4. Platelet Aggregation and Secretion AssayThe light transmission through a 250 μL sample of washed platelets (1 × 108 cells/mL) at 37 °C in a Lumi-aggregometer (Chrono-Log, Havertown, PA, USA) was measured to assess agonist-induced platelet aggregation while stirring at 900 rpm. Amounts of 250 nM, 500 nM, and 1000 nM prednisolone were pre-incubated for 5 min with washed platelets before stimulation with agonists. ATP release from platelets was assessed using a luciferin luciferase reagent to detect platelet dense granule secretion.2.5. Measurement of TxA2 GenerationWashed platelets (1 × 108 cells/mL) were stimulated for 3.5 min in a platelet aggregometer under 900 rpm stirring conditions at 37 °C and the reaction was halted by snap-freezing the sample in liquid nitrogen. Samples were stored at −80 °C until TxB2 analysis was performed. Briefly, samples were centrifuged at 3000× g for 10 min at 4 °C after being thawed at RT, and the supernatant was diluted (1:20) with buffer. The diluted samples were measured to evaluate the levels of TxB2 in duplicates using a TxB2 ELISA Kit (Enzo Life Sciences, Exeter, UK), according to the manufacturer’s instructions.2.6. Western Blot AnalysisWashed platelets were activated with 2-MeSADP for 2 min at 37 °C under stirring conditions, and the reaction was stopped by adding 6.6 N perchloric acid. Platelet samples were washed with distilled water, re-suspended with sample buffer, and denatured for 10 min. Platelet samples were separated on a 10% SDS polyacrylamide gel and transferred to polyvinylidene difluoride membranes. Nonspecific binding sites were blocked by SuperBlock® blocking buffer (Thermo Fisher Scientific, Waltham, MA, USA) at RT for 1 h, and membranes were incubated overnight with anti-phospho-cPLA2 (Ser505), anti-cPLA2, anti-phospho-ERK (Thr202/Tyr204), or anti-total-ERK antibodies diluted to a concentration of 1:1000 with gentle agitation. After 3 washes for 5 min each with Tris-buffered saline with 0.1% Tween 20, the membranes were incubated with appropriate secondary antibody at RT for 1 h. After washing, membranes were incubated with chemiluminescence substrate (Pierce, Rockford, IL, USA) for 5 min at RT, and immune reactivity was detected by using iBrightTM CL1500 (Thermo Fisher Scientific, Waltham, MA, USA).2.7. Statistical AnalysisAll statistical calculations were performed using Prism software (version 9.1). The data were presented as mean ± standard error (SE). Statistical significance was established using the one-way analysis of variance (ANOVA) followed by a post hoc Dunnett’s multiple comparison test. The normality test was conducted by using a Shapiro–Wilk test and data were found to be normally distributed (p > 0.05).3. Results3.1. Prednisolone Regulates 2-MeSADP-Induced Secondary Wave of Platelet Aggregation and Secretion in PlateletsIn order to determine the role of prednisolone in platelet function, we first stimulated the platelets with 2-MeSADP and measured the platelet aggregation and dense granule secretion. As shown in Figure 1A, platelet aggregation and dense granule secretion induced by 2-MeSADP were inhibited in the presence of prednisolone in a concentration-dependent manner. 2-MeSADP-induced dense granule secretion and the secondary wave of aggregation were not completely inhibited by 250 nM prednisolone, while both dense granule secretion and the secondary wave of aggregation were completely inhibited in the presence of 500 nM prednisolone. Therefore, we used 500 nM prednisolone throughout the experiments. Interestingly, prednisolone treatment did not affect 2-MeSADP-induced primary aggregation. Thus, our findings suggest that prednisolone inhibits 2-MeSADP-induced dense granule secretion and the resultant secondary wave of aggregation.3.2. The Effect of Prednisolone in 2-MeSADP-Induced Platelet Function in Aspirinated PlateletsConsistent with the previous result, prednisolone inhibited 2-MeSADP-induced platelet secretion and the resultant secondary wave of aggregation at both low and high concentrations of agonist (Figure 2A). It is well-established that ADP-induced platelet secretion and the secondary wave of aggregation require TxA2 generation [32,33]. In order to evaluate the effect of prednisolone on TxA2 generation, we stimulated the platelets with 2-MeSADP in the presence of aspirin, which blocks the positive-feedback effect of the generated TxA2. As shown in Figure 2B, prednisolone showed no further inhibition at both low and high concentrations of 2-MeSADP-induced platelet aggregation in aspirinated platelets compared to non-aspirinated platelets. These results suggest that prednisolone may inhibit 2-MeSADP-induced platelet aggregation and dense granule secretion by inhibiting TxA2 generation.3.3. Prednisolone Regulates Only Low Concentrations of Thrombin-Induced Platelet Aggregation and SecretionTo check the role of prednisolone in PAR-mediated platelet function, we stimulated the platelets with various concentrations of thrombin. As shown in Figure 3A, prednisolone inhibited a low concentration of thrombin-induced platelet aggregation and dense granule secretion. However, prednisolone could only partially inhibit platelet secretion when stimulated with a high concentration of thrombin. Additionally, prednisolone did not show any further inhibitory effect on thrombin-induced platelet aggregation and dense granule secretion in aspirinated platelets (Figure 3B). It has been known that platelet aggregation and dense granule secretion only require a positive-feedback effect of generated TxA2 when stimulated with a low concentration of thrombin [34,35]. Taken together, our data suggest that prednisolone affects low concentration of thrombin-induced platelet aggregation and dense granule secretion by regulating the positive-feedback effect of TxA2 generation.3.4. Prednisolone Inhibits 2-MeSADP- and Thrombin-Induced TxA2 Generation in PlateletsTo confirm the effect of prednisolone on TxA2 generation, we stimulated the platelets with 2-MeSADP and thrombin and measured the amount of TxA2 release. As shown in Figure 4, both low and high concentrations of 2-MeSADP- and thrombin-induced TxA2 generation were significantly inhibited in the presence of prednisolone, confirming the inhibitory effect of prednisolone on TxA2 generation.3.5. Prednisolone Inhibits 2-MeSADP-Induced cPLA2 and ERK PhosphorylationERK phosphorylation is one of the most important upstream signaling molecules in TxA2 generation in platelets [36]. To understand the molecular mechanism involved in the regulation of TxA2 generation by prednisolone, we stimulated the platelets with 2-MeSADP and measured the cPLA2 ERK phosphorylation in non-aspirinated and aspirinated platelets. As shown in Figure 5, prednisolone completely inhibited 2-MeSADP-induced cPLA2 phosphorylation in both non-aspirinated and aspirinated platelets. However, 2-MeSADP-induced ERK phosphorylation was significantly inhibited by prednisolone only in non-aspirinated platelets, while prednisolone had no effect on 2-MeSADP-induced ERK phosphorylation in aspirinated platelets where there is no contribution of generated TxA2. These results confirm that prednisolone directly inhibits cPLA2 phosphorylation to regulate TxA2 generation in platelets.4. DiscussionGlucocorticoids are one of the most commonly used steroid medications to treat numerous types of allergies, inflammatory diseases, autoimmune disorders, and cancers [37]. Risks associated with glucocorticoid therapy are associated with dose, duration of therapy, and specific therapy used. Furthermore, chronic glucocorticoid administration can result in toxicities and undesirable effects such as Cushing’s syndrome even at physiological doses [38]. Thus, it is important to comprehend the precise mechanisms of the pathophysiological and therapeutic effects of glucocorticoids. One of the various side effects of glucocorticoid administration is hypercoagulopathy, which is known to be triggered by thrombocytosis, hyperglycemia, hypertension, and dyslipidemia [8,9,39]. However, there are discrepancies on how glucocorticoids affect platelet function, which is known to play a crucial role in coagulopathy. In contrast to the known role of glucocorticoids in hypercoagulability, previous studies have reported the inhibitory effect of glucocorticoids on platelet function using PRP [28,29,30,40]. PRP contains various coagulative factors that influence platelet functional responses. To rule out the contribution of the contents in PRP and thoroughly understand the molecular mechanism involved in the regulation of platelet function by glucocorticoids, we used a washed platelet system and investigated the effect of prednisolone, one of the most commonly used glucocorticoids in platelets.Sex hormones, another type of steroid, have a well-established role in platelet function [41,42,43]. As corticosteroids and sex hormones are generated from cholesterol via the same steroidogenic pathway, they share similar effects on target cells [44,45]. 17β-estradiol and progestin, sex hormones, have been known to inhibit ADP-induced platelet activation [41]. Consistently, we observed that 2-MeSADP-induced platelet dense granule secretion and the resultant secondary wave of aggregation were completely inhibited in the presence of prednisolone. It has been shown that ADP-induced secretion and the resultant secondary wave of aggregation are dependent on the positive-feedback effect of generated TxA2 [46]. When the contribution of TxA2 generation was blocked by aspirin treatment, prednisolone had no further inhibitory effect on 2-MeSADP-induced platelet aggregation and secretion compared to non-aspirinated platelets, suggesting that prednisolone inhibits the 2-MeSADP-induced secondary wave of aggregation and secretion through the inhibition of the TxA2 generation in platelets.Thrombin is the most potent agonist that activates platelets via the PAR-mediated pathway [47]. It has been shown that 17β-estradiol and progestin inhibit thrombin-induced platelet activation as well [41]. In contrast, others have reported that 17β-estradiol and progesterone potentiate thrombin-induced platelet activation [42,43]. We found that thrombin-induced platelet aggregation and dense granule secretion were inhibited in the presence of prednisolone. However, unlike the effect of prednisolone on 2-MeSADP-induced platelet function, we observed that prednisolone only inhibited low-thrombin concentration-induced platelet aggregation and secretion, and only partially inhibited high-thrombin concentration-induced secretion. In contrast to ADP, it has been known that TxA2 generation and the subsequent granule secretion play a role at low-thrombin concentration-induced platelet activation [48]. Therefore, our data suggest that prednisolone affects thrombin-induced platelet function by explicitly inhibiting TxA2 generation. More importantly, we found that 2-MeSADP-and thrombin-induced TxA2 generation was significantly inhibited by prednisolone, confirming the effect of prednisolone on TxA2 generation.Glucocorticoids are known to inhibit cPLA2 in several cells [49,50]. cPLA2 is an upstream molecule of TxA2 generation that separates arachinonic acid, a precursor of TxA2, from phospholipids [51]. Glucocorticoids have also been shown to activate ERK1/2 in PC12 cells (a pheochromocytoma cell) and vascular smooth muscle cells [52,53,54]. However, in mast cells, dexamethasone has suppressed the phosphorylation of proteins that are associated with the activation of the mitogen-activated protein kinases (MAPKs) [55]. So far, the effect of glucocorticoids on cPLA2 and MAPKs has not been determined in platelets. In our study, we found that 2-MeSADP-induced platelet cPLA2 phosphorylation was completely inhibited by prednisolone in both non-aspirinated and aspirinated platelets. However, 2-MeSADP-induced ERK phosphorylation was significantly inhibited only in non-aspirinated platelets, whereas no additional inhibitory effect of prednisolone on ERK phosphorylation was observed in aspirinated platelets. It has been known that ERK plays a major role in TxA2 generation downstream of P2Y1, P2Y12, and PARs in platelets [22,33,56]. Further, it has been known that TxA2 can induce ERK phosphorylation by itself in platelets [57]. Thus, our data indicate that prednisolone affects platelet activation by inhibiting TxA2 generation by directly regulating cPLA2 phosphorylation, and that the effect of prednisolone on ERK phosphorylation is mediated by inhibition of TxA2 generation in platelets. High glucocorticoids have been shown to be effective in combination with first-line medication for some bleeding disorders because of their coagulation function [27]. It has also been demonstrated that in a majority of bleeding disorders, it is difficult to control the use of glucocorticoids as it may also lead to hypercoagulability such as in venous thromboembolism [58]. Furthermore, using glucocorticoids has been linked to the development of a number of metabolic disorders [59]. In contrast, glucocorticoids have also been known to cause bleeding disorders in the uterus and gastrointestinal tract [58,60]. Our study has demonstrated that prednisolone inhibits platelet activity, pointing to its potential role in hypo-coagulation and contradicting the prevalent notion that glucocorticoids promote coagulation.5. ConclusionsCushing’s syndrome is a disease that severely affects both humans and dogs and is potentially life-threatening if it remains untreated [61]. It significantly reduces the quality of life and, thus, requires effective treatment [62]. In this study, we demonstrate that prednisolone, a glucocorticoid, inhibits platelet function by the inhibition of TxA2 generation through the regulation of cPLA2 phosphorylation. Moreover, it indirectly regulates generated TxA2-mediated ERK phosphorylation to control platelet functional responses. Overall, our study provides evidence that could be applied to the therapeutic use of prednisolone in coagulation and bleeding disorders as well as the clinical management of Cushing’s syndrome in dogs in the future.

animals : an open access journal from mdpi