Datasets:

taln-ls2n

/

pubmed

like 2

J_Med_Internet_Res-6-3-1550604

Internet Usage by Low-Literacy Adults Seeking Health Information: An Observational Analysis

Background Adults with low literacy may encounter informational obstacles on the Internet when searching for health information, in part because most health Web sites require at least a high-school reading proficiency for optimal access. Introduction Although a tremendous volume of educational health materials is disseminated in the United States, not all Americans find this information accessible or usable. In particular, adults with poor health and low functional literacy face great risks of poor health outcomes and preventable disease progression [1-4]. While many low-literacy adults could benefit from enhanced health knowledge, most current health education materials are written at a 10th grade or higher reading level [3]. Inability to access or understand health education materials inhibits important preventive or treatment measures, and may decrease the likelihood of identifying a symptom of disease. Low health literacy is also a barrier to enrollment in clinical trials [5,6] and minimizes adherence to instructions given by health professionals [7]. These obstacles are compounded by low income levels pervasive in the undereducated population [1], which can prevent individuals from pursuing regular primary care, paying health insurance premiums, or purchasing medications when prescribed. Collectively, these factors help to explain why low-literacy adults are twice as likely to be hospitalized as individuals with high functional literacy [8]. The expense of poor health and low functional literacy on the health system is estimated at $73 billion each year [9]. High cost estimates have encouraged many health-care providers to search for innovative ways to improve health literacy. The Internet has been embraced as an easy-to-use, convenient, and comprehensive clearinghouse for information on diseases, disorders, treatments, and preventions. Even when receiving physician care, between 40% and 54% of medical patients use the Internet to learn about treatment options and tobetter understand their medical conditions [10]. However, the low-literacy population has largely been excluded from the veritable boom of Internet health resources. The expense of Internet services and personal computers may be too high for this population. In addition, most text-based health information on the Internet is too advanced to be optimally effective for low-literacy populations. On average, Internet health-education materials are written at a 10th grade or higher reading level, and 100% of English-language health Web sites examined in a 2001 study required at least high school-level reading proficiency [11,12]. Another study concluded that of 1000 Web sites reviewed, only 10 had a level of writing and content accessible to low-literacy adults [13]. Kalichman et al suggest that individuals who read English below a 6th grade level are not likely to make effective use of the Internet [14]. Further, Zarcadoolas et al report that complex Web features, such as animated links, may be challenging for low-literacy adults to identify and utilize [15]. The 1992 National Adult Literacy Survey (NALS) revealed that more than 90 million Americans either read at a low-literacy level or are functionally illiterate [1]; the paucity of Internet health resources appropriate for these individuals perpetuates discrepancies in health outcomes between the educated and undereducated. While no studies to date have determined how many low-literacy adults regularly use the Internet to find health information, the dearth of educational materials suitable for these individuals may impair optimal usage and navigation. One study has reported interventions that enabled low-income HIV-positive individuals to use the Internet and to critically evaluate information that they encountered [16]. Health-related Internet use has also been shown to enhance knowledge about HIV and to be correlated with active coping in a study of HIV-positive patients [17]. Although these studies focus on low-income status rather than low-literacy status, the established correlation between these two factors suggests that low-literacy adults may likewise benefit from augmented health education via the Internet. We conducted an observational study of low-literacy adults to assess how they searched for Internet health information in as close to a natural setting as possible. Our investigative questions include the following: if low- to mid-level literacy adults are given access to the Internet, can they find basic health information that they can understand? Will their search strategies be effective in identifying information that they can use and comprehend? How do they rate current health Web sites in relation to their needs and interests? Will they be able to conduct successful self-directed searches? In our investigation, we also categorized navigational strategies used by low-literacy adults and the reading level of materials they accessed. Methods We enrolled 13 adult literacy students (3rd to 8th grade reading levels) from Bidwell Training Center, a vocational school in Pittsburgh, Pennsylvania. The protocol used was approved by the University of Pittsburgh Institutional Review Board. Bidwell students are organized individually and/or in small groups for reading instruction; they meet together once a week for program announcements. The literacy program coordinator introduced the study to 20 students in this large-group setting. Thirteen interested students then self-selected into the study. All subjects participated in a computer skills workshop in May 2003, where they were presented with basic search and navigation strategies and learned how to use the Google search engine. We selected Google because it is a widely used search engine and has a "Did You Mean…" feature that corrects misspelled search terms. We anticipated that this might be a feature that low-literacy subjects would find particularly helpful. Among other topics, subjects were taught how to use the "Back" button and the "Forward" button, how to scroll down a page, how to identify links, and how to conduct basic searches. Each subject also filled out a brief questionnaire to give insight on their educational background, ethnicity, health insurance status, and previous experience with computers and the Internet. The questionnaire was written at a 3rd grade reading level (Flesch-Kincaid Reading Scale). An investigator met individually with each of the participants within 3 weeks of the computer skills course for the observational portion of the study. Participants were 1) administered the REALM test (Rapid Estimate of Adult Literacy in Medicine) [18] to assess their health literacy level; 2) asked several questions to gauge their comfort level on the Internet and prior Internet experiences; and 3) taught how to "think aloud," or continually express their thoughts while using the computer. Investigators engaged each participant in several think-aloud examples in order to actively illustrate this process. The investigator then asked the participant to use the Internet and Google search engine and think aloud while finding information on a subject of his or her choice. This preliminary question allowed participants to practice and review their Internet searching techniques. Participants were permitted to ask the investigator technical and navigation-related questions during this part of the study. These questions included, but were not limited to, whether to put spaces between words in search terms and how to initiate a search once a search term had been specified. Participants were then asked to find answers on the Internet to 3 health-related questions developed by members of the research team. Participants were instructed to use the Google search engine so that their answers could be standardized. A committee consisting of a physician, a faculty member specializing in human-computer interactions, a community health educator, and an information sciences specialist compiled various answers to these questions that would qualify as accurate and complete. Subjects who were able to generate any of these answers during their online searches were considered to have answered the questions correctly; subjects who were not able to generate these answers were determined to have answered the question either incorrectly or incompletely. Examples of responses for each question that would have been considered correct are included in the Results section The investigator read the 3 questions aloud and also provided them to the participant in written form (Arial font, 20 pt): 1. Think of a health question you are interested in for yourself or for someone you know. Find out information about this question on the Internet. 2. Imagine that someone you care about has lung cancer. This person would like to know about treatments for lung cancer. Can you find out the three main types of treatments using the Internet? 3. Imagine that you are at a doctor's office and you are told you have a disease called diabetes (sometimes called sugar). You are given a pill called Metformin to take for it. What does Metformin do? Subjects identified answers to the investigator, who then asked them to articulate the answers in their own words. Participants who seemed frustrated or unreceptive, or who asked to move to a new question were directed to the next task. Participants were allowed to use any Web sites they felt would help them answer the questions. Participants also were not provided with dictionaries-our objective was to examine how they navigated the Internet without assistance from external sources. Subjects were given up to 15 minutes to complete each task, as measured by the investigator. To minimize anxiety, they were not informed of the time limitation. After the 15-minute period, investigators used a series of prompts to gradually guide subjects, if necessary, to the next task. Next, investigators accessed the colon and rectum cancer Web page on the American Cancer Society (ACS) Web site [19]. Participants were asked to navigate through links on this page and find 2 ways to help prevent colon and rectum cancer. Investigators recorded the amount of time spent answering this question and the number of links participants clicked on to find the answers. After this task was completed, investigators asked the participants several subjective questions to qualify their experience on the Internet. Participants were then given $25 compensation, which ended their direct involvement in the study. Investigators wrote notes on each participant's progress, and asked for participant feedback about the Internet both before and after searching the Internet. Investigators did not coach subjects on proper technical or navigational techniques after the initial practice question until subjects had completed their tasks. In 2 cases, investigators directed subjects to Google's "Did You Mean…" search term correction option in order to adjust for spelling mistakes; these subjects had repeatedly demonstrated very poor spelling proficiency before this intervention. Camtasia Studio screen-capture software recorded individual keystrokes and think-aloud recordings. Questionnaires and think-aloud methods were used to ascertain the criteria used by participants in evaluating Internet health Web sites. Investigators also calculated the 1) literacy levels of Web sites accessed by the participants, 2) the amount of time spent on each Web site, 3) the number of questions answered thoroughly and correctly by each participant according to pre-determined standards, 4) the average number of sites used to answer each question, and 5) the number of participants who accessed sponsored sites, or paid advertisements appearing on the Google retrievals page, and how many used that information to answer questions. Results Qualitative and quantitative results were analyzed in this study. Participants In this study, the subject population was reduced from 13 to 8. Two participants were excluded because they did not attend the one-on-one searching session with the investigator. Two other participants were excluded because they were non-native English speakers who did not understand the tasks presented to them. One participant was later excluded because technical problems prohibited the retrieval of her computer searches. The average age of our 8 remaining participants was 41.5 years. Five subjects were male and 3 were female. Seven identified themselves as African Americans and 1 self-identified as of Asian descent. The Asian participant was an English-as-a-second language (ESL) speaker with a university education from his native country. Seven of the 8 participants reported having health insurance. Seven of the 8 also had at least some high school or trade school education; 1 participant did not report educational experience on the intake questionnaire. Of these subjects, 2 reported on the intake questionnaire that they had never previously used a computer or the Internet. Two reported that they had previously used a computer, but had not used the Internet. Subjects generally used computers with greater frequency than they used the Internet. Three participants reported on the questionnaire that they used the Internet 2 or more times a week; they later said verbalized that their main online interests were news, sports, cars, and/or entertainment information. The other 5 participants reported on the questionnaire that they used the Internet either occasionally or not at all. Usage reports from the intake questionnaires are provided in Table 1. Table 1 Self-reported, written questionnaire responses about prior Internet and computer usage by subjects (n=8) Subject Have you ever used a computer? If so, how often? Where do you use computers? Do you use the Internet? If so, how often? Where do you use the Internet? 1 No "No where" [sic] No "I've never used the Internet" 2 Less than once a month "When I was in jail" No (N/A) 3 2 or more times a week "At school, Bidwell Training Center in Ms. Cooper's class." No "At the Carnegie Library in Beechview where I live" 4 No "No" No "No" 5 2 or more times a week "To type" Yes; Less than once a month "In school" 6 No; 2 or more times a week "At home" Once a week "At home" 7 2 or more times a week "Home" 2 or more times a week (at home) "Home" 8 Once a week "Different location" 2 or more times a week "Different location" As seen in Table 1, the self reports of prior Web and computer experiences are unclear in several cases. Subject 3 reported no prior Internet usage in one part of the questionnaire, but reported in a subsequent answer Web usage at a local public library. In addition, as Table 1 indicates, subject 8 reported more frequent usage of the Internet than of computers; subject 6 (ESL student) first indicated no prior computer usage, then later reported on the questionnaire computer usage of twice a week. Because there were seemingly divergent perceptions of what constitutes a computer or Internet experience, perceived computer/Web adeptness cannot be correlated with our participants' experience using this technology. Therefore, while this study will indicate differences in results between the 3 people with frequent Internet experiences (defined in this study as usage of at least once a week) and the 5 individuals without, the study will not attempt to conclude whether the skill level of subjects in the study correlated with the sustainability of their prior computer and Web experiences. Search Engine Usage Participants reviewed their navigational skills during their preliminary question, where they were encouraged to look for information on any subject that interested them. They used Google to search for a variety of topics, ranging from entertainment to health-related information. Participants occasionally searched for information on more than one topic. Participants used the search items listed in Table 2 in order to answer the preliminary question and questions 1 to 3. Semicolons between words or phrases separate multiple search terms used by a subject to answer a question. The subjects are listed in Table 2 in the same order (ie, 1, 2, 3…) as they appeared in Table 1. Table 2 Search terms used by subjects to answer preliminary questions and questions 1 to 3 (n=8) Subject Preliminary Question 1 Question 2 Question 3 1 lena horn† Lung cancer Lung cancer Metformin 2 health care;health care mental Sports and health health care about lung cancer A pill called metformin 3 (no clear search topic) Herpes Cancer Metformin 4 Wwwsoulfood; wwwsoulfoodcom; soulfood AIDS lung caner† Diabetes 5 Will Smith; sipers†; spiders High blood Lung cancer Metformin 6** Bi;;;dwell training center† Health health lung caner† Health diabetes 7 sonny Rollins Tuberculosis Treatments for lung cancer Metformin 8 Babyface recording artist Pain Cancer Pdr* * Physicians' Desk Reference ** English-as-a Second Language subject † misspellings for: "lena horne," "bidwell training center," "lung cancer," and "spiders"; the Google correction option was used in two instances when the subject was prompted by investigators to amend search terms. Questions 1 to 3 were given to our participants in writing, as well as orally; this may have affected their selection of search terms. For question 2, one participant wrote "treatments for lung cancer" in the search term box, a phrase that is written explicitly in that question. Another participant was similarly prompted by the wording of question 3 to write "a pill called metformin" as his search term. Individuals who used the Internet at least once a week are labeled in Table 1 and subsequent tables as subjects 6 to 8. Search terms generated by these frequent Internet users did not differ greatly from search terms generated by individuals who had little Internet experience. The one exception was subject 8, who attempted to answer question 3 by using the online Physicians' Desk Reference, a site about which she had once heard good reviews. In general, this group found generating original search terms to be somewhat challenging. Many did not initially remember whether to put spaces between the words in search terms. Even a subject who reported using the Internet once a week hesitated when writing the search term for question 1, finally stating, "Yeah, you do have to space [between words]… I had to remember if you had to space." With one exception, participants were able to correct their terms by inserting the proper spaces. Spelling of search terms was generally a problem for only 2 participants, one of whom (subject 7) spoke English as a second language. Subjects tended to self-correct for spelling in the search term box before pressing the "Google Search" button or Enter key. Several participants also had difficulty understanding what type of terms to put in. When conducting a preliminary search for information on the television show, Soul Food, one participant typed into the search term box, "wwwsoulfood." When this retrieved no results, the subject looked at the URL for guidance and then typed "wwwsoulfoodcom" into the Google search term box. This again did not yield any results. The participant next entered "soulfood" into the search term box. The investigator finally directed the subject to Google's "Did You Mean…" option so that the subject could answer the question. However, this participant had continued difficulties generating correct search terms; later in the study, he used "lungcaner" as a search term to find information about lung cancer. Nearly all participants retained skills such as scrolling and clicking on links from the computer workshop or previous Internet experiences. They also learned other navigational strategies through repetition and practice. For example, one participant who was conducting a preliminary search for information about Will Smith looked at the Google retrievals and stated, "So it [search engine] must go to other Smiths ... I wonder if I was supposed to put in 'Will Smith the actor'?" Quickly, the subject had learned that increasing the specificity of search terms generally improves the specificity of results. Six of the 8 participants did not venture past page 1 of the Google retrievals. One participant was surprised by the number of search results, saying, "You find a lot of stuff on this thing [the Internet]." Another participant explained why she stayed on page 1: "Oh boy, I've got a lot to choose from. I don't want to go to the other ten [pages of retrievals] because it might give me other information I don't really need ... the first page gives me just enough of what I need to know." This participant had deduced that first-page retrievals typically have the most relevant sites to the particular search term used. Later, this subject stated, "I didn't answer the questions, but I looked up the information, and it [Internet] gave me what it wanted me to have." This statement implies that the subject believed that the Internet was more in control of the searching than the subject, revealing a possible belief that the search engine and search terms selected are not the primary determinants of what type of information is retrieved. Sites Accessed Ability to Answer Questions In question 1, participants were asked to use information on the Internet to find the answer to a health-related query of their choice. Most participants identified only a subject area, and did not clearly articulate a specific question despite verbal prompting by the investigators. Several participants initially stated a topic, but changed it as they retrieved unrelated material that they found more interesting. While recordings from the think-alouds would have been helpful in designating the search topics, we found that despite investigators' prompts and encouragement, subjects were very reluctant to verbally report their real-time experiences navigating through the Web. As one subject stated, "Shucks, I can't think aloud." It is therefore difficult to gauge whether participants were able to find adequate information for which they searched, especially during the unstructured searching period required to answer the first question. Question 2 required participants to locate the 3 main types of lung cancer treatments (acceptable answers: chemotherapy, surgery, radiation). This question models the navigation of a typical Internet health-information seeker who searches for disease-related information. Of all 8 participants, only subject 5 was able to answer this question accurately and completely. Subject 3 verbalized one viable option-chemotherapy-based on information accessed online. The remaining participants either did not answer the question or identified an alternative medicine as one of the principal types of lung cancer treatments available. Question 3 required participants to find out the role of metformin, or Glucophage, in diabetes treatment (one acceptable answer: metformin lowers sugar in the blood). This question models a doctor-patient interaction in which a patient who is prescribed an unfamiliar medication independently searches for information about its effects. Six of 8 participants were unable to find information on the Internet to answer the question. The 2 participants, subjects 3 and 7, who found the information, read directly from text on the site and did not articulate the information in their own words. Surprisingly, subjects who reported sustained prior Internet experience in the questionnaire were no more successful at answering questions than subjects with little Internet experience. This could have been a result of the generalized search terms that they used to look for answers. Prior Internet experience does not seem to lead to satisfactory search/navigation skills for members of this group in searching for health information. Information Accessed Sites used by subjects 3, 5, and 7 to successfully answer questions 2 and 3 were written at a 12th grade reading level (Flesch-Kincaid). It is noteworthy that these subjects were able to identify the answer in the text and read it aloud. In 2 out of 3 cases, they were unable to express these answers in their own words, which suggests a minimal comprehension of the material accessed. Seven of the 8 participants accessed sponsored site information while attempting to answer questions. Businesses pay a service fee to Google to have their site names appear as sponsored sites when triggered by a particular search term or keyword. Sponsored sites are outlined in color and/or appear in boxes on the right side and heading of the Google retrievals page. In general, alternative treatments and commercial therapies and medications appear under this listing; many of these sites may contain information that is uncorroborated by legitimate scientific sources. Five participants used information provided by the sponsored sites to answer questions. Two out of 3 of the subjects who used the Internet at least once a week also used this information to answer questions. Half of the participants searching for lung cancer cures arrived at the same site: an Asian dietary supplement site claiming to cure cancer by removing free radicals from the body [20]. Another popular sponsored site promoted a radio frequency technique to hinder cancer progression [21]. The titles of these sites as they appeared in the sponsored sites submenu were: "New Cancer Treatment" and "Cancer Treatment." The Flesch-Kincaid formula indicated that the information on both sites was written at a 12th grade or higher reading level. Information on sponsored sites, therefore, was not necessarily any easier to read or interpret than information on non-sponsored sites accessed by subjects in this study. General Site Profiles Observational logs and records on the Camtasia software show little correlation between our subjects' ability to identify answers and the amount of text on a page; analysis using the Camtasia software also showed little conclusive difference in the amount of time that the subjects spent on each site despite variances in the amount of text on the pages accessed. Therefore, subjects did not seem to prefer or navigate towards Web pages/sites with less text. Participants, on average, used between 1 and 2 Web sites to answer questions 1 to 3. Table 3 records the number of links from the Google retrievals page that were selected by subjects. The results for subjects 1 to 5-the participants with minimal prior Internet experience-are also presented separately from the results for participants with sustained prior Internet experience (subjects 6 to 8). The Flesch-Kincaid reading scale used in this study scores text at a 1st to 12th grade reading level. Given this scale, sites ranked at the 12th grade level require at least that level of reading ability. That is, material scored at a 12th grade level may actually be written at a college level. In our study, the average site accessed required at least a 10th grade reading level. Table 3 Average number of links used to answer questions Avg. Number of Links Used (Average Total) Avg. Number of Links Used (Subjects 1-5) Avg. Number of Links Used (Subjects 6-8) Preliminary 1.875 2.4 1.0 Question 1 1.14 1.2 1.67 Question 2 1.82 1.8 2 Question 3 1.5 1.6 1.33 AVG. 1.58 1.75 1.5 Table 4 Average (rounded) reading level of sites accessed Avg. Reading Level of Sites Accessed Avg. Reading Level of Sites Accessed (Subjects 1-5) Avg. Reading Level of Sites Accessed (Subjects 6-8) Preliminary 10.50 10.7 10.0 Question 1 10.50 9.4 11.2 Question 2 11.1 11.3 11.0 Question 3 11.8 11.8 11.9 AVG. 11.0 10.8 11.0 Table 5 Average time spent on sites Avg. Total Time Spent Per Site (min) Avg. Total Time Spent Per Site (Subjects 1-5) Avg. Total Time Spent Per Site (Subjects 6-8) Preliminary 7.2 8.7 4.7 Question 1 10.3 10.6 9.8 Question 2 8.7 8.7 8.7 Question 3 6.6 8.3 5.8 AVG. 8.2 9.1 7.25 Participants spent an overall average of 8.2 minutes on individual sites. All participants voluntarily finished answering questions 1 to 3 before the 15-minute time limit was reached. After completion of these first 3 questions, subjects were directed to a specific site; question 4 was posed about information directly linked to that site. We chose to use the ACS colon and rectum cancer Web page site, which contains links to a variety of prevention resources written at 6.3-12.0 grade levels (Flesch-Kincaid Reading Scale). The page to which we directed subjects consists of a listing of links to defined topic areas, one of which was closely related in wording to question 4. On the ACS site, 5 out of 8 people were able to answer question 4 correctly. Three of the 5 reported prior Internet experience; 2 reported none. These subjects used 3.8 links on average to answer the question. The 3 subjects who did not access the material used 6.5 sites on average before they were either stopped by the investigator or quit voluntarily. Two of these subjects had never used the Internet prior to enrollment in the study. Attitudes and Self-reporting While most participants were unable to answer all of the questions asked, 7 out of 8 reported feeling very comfortable or comfortable with their Internet searching experience. The eighth participant felt moderately comfortable. Also, 5 out of 8 found it at least moderately easy to find readable and understandable information on the Internet. Two of the remaining participants found it very difficult to find readable information, and one participant reported that finding understandable information is easy if the Web user has strong reading skills. Despite their dependence on sponsored sites and alternative Web sites to answer questions, 7 out of 8 subjects reported that they found it very easy to locate trustworthy information on the Internet. The eighth subject noted that it is moderately easy to find information that is trustworthy on the Internet. However, one subject said, "I believe that on the Internet, you have your shysters ... just like anything." Subjects felt positive about continuing their online experiences, and all expressed some enthusiasm about improving their skills. One participant stated, "I'm getting a computer ... it can help your typing skills." Another subject said, "The computer is real interesting. I'm a see if I can get one so I can learn [how to use it]." After the study was completed, many participants asked investigators to continue teaching them Internet skills or to continue helping them locate Internet resources on a variety of subjects. Discussion This observational study is the first to examine Internet use by low-literacy adults seeking health information [11]. Irrespective of prior experience using the Internet and/or computers, low-literacy adults participating in our study did not use optimal search terms to answer questions, encountered difficulties finding health information at the appropriate reading level, and were unable to successfully interpret Internet health information as it was presented. While basic navigational skills (eg, using the "Back" button) were easily retained, areas that required reading and comprehension were problematic for most subjects-evidenced by their inability to answer questions and comments made during their think-alouds. Therefore, the literacy level needed to read health information on the Internet does appear to inhibit information-seeking efforts of low-literacy adults. Searching strategies were sub-optimal in several respects. First, the search terms used by subjects were predominately non-specific (Table 2). Although we anticipated that subjects who used the Internet more often would generate more specific search terms than did their peers, we did not observe this in the study. Difficulty Generating Search Terms Without guidance, subjects had difficulty generating original search terms that would yield specific results. A recent study reveals that adolescents used similarly general search terms when searching the Internet for health information [22]; this corroborates results from another study, which found that among subjects with an average of 33 months of Internet experience, self-selected search terms to find health information were unexpectedly general [23]. These observations highlight search terms as a potential barrier to specific, targeted Internet health information for different types of Internet users with varying levels of Web expertise. A categorizing search engine might be particularly effective for use by these groups; it minimizes the need for individuals to both create a specific search term and independently read and assess all retrievals. A sample search to answer question 2 was conducted using the Vivisimo search engine [24]. The search term "lung cancer" yielded a series of folders about lung cancer separated by subject matter; one folder specifically focused on lung cancer treatments. Individuals clicking on that option could access all sites on lung cancer treatments retrieved by the engine, circumventing the need to sift through thousands of retrievals to locate treatment-focused sites. A future study could monitor the ease with which low-literacy individuals could conduct self-directed searches using an automatically sorting search engine. Reluctance to Use Links Search strategies observed in this study were also sub-optimal because most subjects exhibited some unwillingness to click on links to Web sites on the Google retrievals page. On average, subjects clicked on one to two links to answer questions. Even when the subjects did not appropriately answer questions or only partially answered questions, most seemed reluctant to click on additional links on the Google retrievals page, and 7 of 8 did not go to subsequent retrievals pages. These results did not seem to correlate with prior Internet experience. Subjects also rarely re-typed search terms in order to access more relevant retrievals. These results differ from those of a previous observational Internet study, whose participants preferred to choose links from page-one retrievals and then re-type original search terms if they were unable to find appropriate information [23]. As stated earlier, our subjects had such difficulty generating original search terms, figuring out appropriate spelling, and determining whether to place spaces between words in search terms, it is conceivable that this is why they avoided this strategy. Another reason why subjects' generation of search terms and selection of links were so limited may have been because the subjects were not interested in the health materials or the questions. Subjects may have also found the Google retrievals page confusing and intimidating. While the think-alouds are inconclusive about which of these factors contributed most to the weak search strategies observed, the post-session questionnaire reveals that the majority of participants reported that it was easy to search the Internet. Future research may help to illuminate the factors that contribute to the inconsistencies between subjects' perceived unwillingness to explore the Internet's health resources and their positive feedback about navigating through these resources. High Literacy Levels of Health Web Sites The health sites participants accessed to answer questions 1 to 3 had, on average, an 11th grade reading level (Flesch-Kincaid Reading Scale), which was consistent with the findings of previous studies [3,25]. Clearly, all of our subjects experienced difficulties using these sites to answer questions. The literacy level of the materials that the subjects did access may have limited their ability to read and understand materials as presented to them, and may have also impaired their ability to select the appropriate links for finding information. However, a majority of subjects were able to find specific information on the ACS Web site. As one subject reported about the site, "This is a real good one 'cause it breaks it right down for you." This Web page consisted of a series of links: general links on the left and right sides of the page and links to colorectal cancer in the center. Subjects who were unable to answer the questions seemed to find the lists of links on the page confusing, and picked links that took them to unrelated pages on the ACS site rather than to specific pages containing colon and rectum cancer information. While the selection of only 1 link on the colon and rectum cancer Web page was necessary in order to answer the question, these subjects on average picked more than 6 separate links before quitting. Therefore, layout of health Web sites evidently affects the ability of low-literacy adults to find pertinent health information. Despite the navigational difficulties observed on the ACS Web page, the ability of 5 subjects to correctly answer question 4 probably resulted from the fact that the information needed to answer question 4 was written at an 8th grade reading level-significantly lower than the11th grade reading level required on average to read information retrieved in the first 3 searches. This suggests that low-literacy individuals can identify and utilize easier-to-read materials on Web sites. The Internet may indeed be a useful health resource to this population if materials are written at an appropriate reading level. Considering the navigational struggles of our subjects, the actual process of locating low-literacy sites on the Web may prove a more daunting challenge to this population. Difficulty Measuring Participants' Comprehension of Information While most were able to competently navigate through lower literacy materials, subjects' comprehension of Internet health information was difficult to measure in our study. Some participants found correct answers and read them to the investigators directly from the Web text, but none were able to articulate the answer in their own words when prompted. In their analysis of the1992 National Adult Literacy Survey (NALS) results, Kirsch et al reported that low-literacy adults may successfully perform simple comprehension exercises such as locating a single piece of information from text, but often find it more difficult to integrate and synthesize that information [1]. Furthermore, subjects in our study may have been able to use cues from sentence structure to locate an answer, and then relied on their pronunciation skills in order to read the answer as written. However, their ability to identify relevant health information within text is not necessarily a measure of their ability to comprehend that information. In addition, several subjects seemed to compensate for their low literacy skills by using external information resources. One subject who examined a Web site on mental health law (12th grade level) expressed great enthusiasm about a particular topic that he said was presented on the site. A perusal of the site after the session showed that this topic was not addressed on any of the pages he had accessed. This participant may have compensated for his struggles in reading the site by citing facts with which he was personally familiar. Another subject used a similar approach when accessing a lung cancer site. When asked about the type of information he was reading, the subject responded that the page focused on smoking cessation. However, there were no smoking-related topics on the pages examined by the subject. The subject was able to correlate lung cancer with smoking, and may have relied on this information in order to answer the investigator's query. Overall, some subjects may have been able to rely less on actual comprehension skills and more on background knowledge in order to infer answers. Positive Web-site and performance feedback reported by most of the participants could have also been fueled by a desire to compensate for reading and comprehension difficulties. Participants were aware that the majority of the investigators were affiliated with a local hospital system; some may have felt compelled to answer positively about Internet health information because they were reporting to health-care professionals. Additionally, the participants may have been unwilling or ashamed to admit that they had difficulty understanding the information on the Internet. Individuals with low literacy tend to be embarrassed by their reading inadequacies [26]. Participants may have felt compelled to report more positively about their Internet experiences in order to de-emphasize their difficulties navigating the Web. These considerations might begin to explain that while most participants struggled when using the Internet, most 1) felt they did a good job searching for information, and 2) found information on the Internet readable and understandable. Collectively, then, poor comprehension of health information on the Internet coupled with a desire to compensate for self-perceived inadequacies in reading may have negatively affected the ability of our subjects to objectively evaluate Web sites. In this study, these factors may also have diminished the accuracy of their think-alouds and feedback in relation to their actual Internet experiences. Inaccurate Self-assessment An alternative reason why subjects reported positive experiences on the Internet could be that subjects were unaware of the magnitude of their Internet searching difficulties. A study by Moon et al indicates that 70% of subjects told investigators that they read "really well," while in actuality, their mean REALM scores reflected a 7th to 8th grade reading level [27]. This suggests that individuals may actually overestimate their reading ability in relation to standard educational parameters; it may also relate to a similarly heightened perception of Internet competence. Furthermore, because the majority of our subjects had minimal Internet experience, they may not have been able to objectively gauge the limitations of their Internet skills in relation to the skills of more advanced users. While the investigators were able to categorize their searching as sub-optimal, our participants could have considered their searching strategies to be adequate, if not standard. Preference for Sponsored Sites Subjects' reliance on sponsored-site information to answer questions, regardless of the high literacy level required to read those sites, suggests that other factors promote the selective advantage of sponsored sites over non-sponsored sites. In fact, the design of sponsored sites on the Google retrieval page follows many of the guidelines for creating optimal layouts for health information targeted to low-literacy adults [28]. First, the sponsored sites are organized by topic, and are also segmented in colored boxes that stand out from the rest of the Google retrievals. They do not contain the "teaser information" and keywords associated with normal Google links, and minimize the amount of text used. Most are easier to read than the normal Google links, are automatically categorized by subject, and are visually stimulating. In addition, despite misspellings of search terms, sponsored sites are often applicable to the intended subject. For example, a search of "lung caner" instead of "lung cancer" yields sponsored sites on lung cancer, though most of the non-sponsored Google retrievals are irrelevant. When individuals misspell search terms, which the low-literacy subjects in our study did fairly commonly, they might easily gravitate to sponsored-site information to answer their health questions. Of concern is that subjects did not seem to differentiate between the information on the sponsored sites and information on non-sponsored sites. Subjects used these sites interchangeably to answer questions. One study suggests that critical interpretation of Web sites is based on the Internet acumen and interests of the information-seeker; if coupling the motivation to find a topic and the ability to do so successfully, the information-seeker will be well-equipped to evaluate Web sites objectively and perceptively [29]. This approach offers 3 possible explanations for our results. First, our questions may have been of little interest to our subjects; this may have diminished their motivation in answering questions and affected impacted their critical analysis of sites. Second, many of our subjects had little sustained exposure to various Web sites before the study. Those subjects in particular may not have been able to critically compare Web sites as readily as individuals who had previously seen both good and bad Web sites and developed their own rating system. In this context, most health information on the Internet may have seemed trustworthy and interchangeable to some of the subjects. Third, the searching problems observed even among those subjects with previous Internet experience underscore the fact that none of these subjects reported that their prior Web usage included searches for health information. While these subjects had successfully found items of personal interest in previous Web searches, they were unable to navigate to health materials that were any more accurate or easy-to-read than those found by the rest of the subjects. Therefore, health searches may present unique challenges to a low-literacy population that counter the ability to find accurate, trustworthy health information. This may result from the high literacy level required for reading health information and health Web sites in addition to the complexity of health terminology. Limitations of Methodology Standard methodologies used in this study to determine health literacy and to generate continual feedback were sub-optimal. First, REALM test results were inconclusive. Subjects were placed into the literacy program at Bidwell Training Center after taking the national Tests for Adult Basic Education (TABE). However, in our study, these subjects tested significantly higher on the REALM than expected for individuals with the reading levels indicated by their TABE scores as reported by Bidwell Training Center (3rd to 8th grade reading skills). Subjects may have strong phonetic skills that help compensate for poor word recognition and comprehension. This observation is supported in a study by Wilson et al [30], which similarly noted that lower literacy participants who used the REALM tested at several grade levels above their actual reading level. The REALM may not be an optimal tool for accurately determining the health literacy of low-literacy adults. Whereas complete think-alouds could have helped us better understand subjects' navigational priorities and comprehension levels, the protocols we used in this study were ineffective at prompting verbalization. None of the participants consistently articulated their step-by-step navigational process at all points during their searching session. Investigators continually prompted the subjects through the exercise, but were unable to stimulate free-thinking, consistent, and self-motivated think-alouds. One potential explanation originates from the observation that our study population was not uniformly familiar with the Internet. Therefore, some subjects may have felt overly challenged by simultaneously learning how to use the Internet and verbalizing their navigational strategies. According to previous studies [31], these subjects were probably in an "acquisition role." Such studies disclosed that a learner who is new to a certain task focuses primarily on acclimatization, and finds it overwhelming to concurrently think aloud. Since traditional think-aloud protocols may be ineffective for this group, an interactive protocol may be of assistance for future studies. In such a protocol, subjects would directly be asked about specific site features, and asked to rate and make comparisons between health sites. This may highlight precise preferences the subjects might have for Web-site information, content, design, and presentation, and may result in a more cohesive rating system. Overall, however, our subjects were very enthusiastic about learning how to use the Internet, and all indicated an interest in improving their skills for future use. In this study and other studies [13,15], members of the low-literacy population have expressed excitement about using the Internet. In order for the Internet to further empower these individuals to make informed health decisions, the development of easy to read and easy to comprehend health materials is imperative. If Google's sponsored sites are usedas a guide, low-literacy adults prefer information that is aesthetically pleasing, has minimal text, and is organized by subject matter. Search engines that are able to consolidate these features for searches will probably be of greater use to this population. However, low-literacy adults must improve their navigation and searching skills to efficiently locate low-literacy materials on the Internet. With sufficient practice, they are likely to develop the skills to use the Internet to find specific health information, and learn to critically evaluate the information they access. Indications for Future Research One caveat to the present study is that our sample size precluded the analysis of factors besides low literacy that could influence the results we observed. We believe, however, that our findings with this sample group in an observational study were representative of the way low-literacy adults interact with the Internet. It will be important to validate and analyze in a larger study the appeal of sponsored sites (as opposed to other retrieved links) to low-literacy adults. It will also be worthwhile to determine the relative importance of limited literacy in comparison to socioeconomic and cultural factors in effective use of the Internet by this population. Future work will identify the exact components of sites that engage and promote learning by low-literacy adults. Greater understanding of these factors will hasten the day when the Internet becomes an effective vehicle for optimizing the health knowledge and acumen for those at high risk of poor health outcomes.

Arch_Orthop_Trauma_Surg-3-1-2092411

Drilling k-wires, what about the osteocytes? An experimental study in rabbits

Introduction The function of osteocytes regarding osteonecrosis has been underestimated for a long time. Recently it has been suggested that apoptosis of osteocytes results in strong osteoclastic bone resorption. Death of osteocytes due to drilling may therefore increase the risk of K-wire loosening. The purposes of our in vivo study were to assess the minimal drill time needed to notice disappearance of osteocytes and to measure the distance of the empty osteocyte lacunae surrounding the drill tract in relation with the insertion time, directly and 4 weeks after drilling Kirschner (K-) wires into the femur and tibia of rabbits. Introduction Experiments concerning thermal damage to bone tissue due to drilling Kirschner wires (K-wires) show several variables interfering with heat generation [7, 9–14, 32]. These variables can be categorized into three groups: drilling technique, K-wire characteristics and external factors. The drilling technique is a subtle balance between drilling speed, insertion time and pressure [1, 2, 5, 30, 31, 34]. The characteristics of K-wires differ in diameter and K-wire tip [17, 23, 29]. The main external factor is irrigation with a water spray during drilling [3, 20, 24, 36]. It is believed that critical temperature for bone injury is around 56°C because alkaline phosphatase denaturates at this temperature. Osteocytes, however, may even be more sensitive to heat [23, 24]. Eriksson [9, 10, 12] observed that a temperature of 47°C for only 1 min results in bone resorption. Bone is capable of remodeling by bone resorbing osteoclasts, bone forming osteoblasts and osteocytes [15, 19, 22, 26]. The latter differentiate out of osteoblasts which have ceased bone production and have become embedded in the bone matrix, in vacuoles cushioned by fluid and large molecules, forming a network [8, 15, 19, 22, 25]. Despite the fact that the osteocytes are the major constituents of bone, their role in bone resorption regulation has remained controversial for a long time [4, 16, 21, 22, 27]. In the past decade, however, understanding of osteocyte physiology has increased dramatically and it has become clear that osteocytes play an important role in bone remodeling [6, 15, 18, 25, 26, 35]. Now we know that micro damage to bone, e.g. by drilling, is associated with an increase in osteocyte death by apoptosis [25]. This process triggers local bone resorption resulting in K-wire loosening, because death of osteocytes turns off the inhibition of osteoclasts [6, 15, 16, 22, 25, 35]. The status of osteocytes after drilling into bone has been investigated before. Thompson [33] describes the absence of osteocytes as an acute cellular reaction, which increases in severity with increase of drilling speed. Pallan et al. [28] continued Thompson’s [33] work and describes the delayed cellular changes in bone after pin insertion which were left in situ for maximum 10 weeks. He also concluded that higher speeds produce relatively higher temperatures and increased tissue response after a long time period [28]. It is well known that in clinical daily practice K-wires are often drilled without proper cooling because most drilling devices do not have an incorporated cooling system and therefore cooling has to be done manually. In case percutaneous drilling is performed, the cooling effect on bone is minimal because of the surrounding soft tissues. An in vivo study was therefore designed which simulated daily practice concerning K-wire drilling to analyze the absence of osteocytes directly (t = 0) and 4 weeks (t = 4) after insertion. Our first aim was to measure the minimal drill time needed to notice disappearance of osteocytes. The second aim was to measure the distance of the disappeared osteocytes around the periphery of the drill holes in relation to drilling time. Materials and methods Animals and anesthesia A total of 14 healthy, New Zealand white rabbits of female sex weighing a mean of 2.81 kg (2.66–3.09 kg) were used in this investigation. The rabbits were solely housed on a 12 h/12 h (light/dark) cycle and provided with standard diet food and water ad libitum. All animals were housed in the Central Animal Laboratory, Utrecht University, Utrecht, The Netherlands and received care in compliance with the European Convention Guidelines. The animals were pre-anesthetized with a combination of methadone (10 mg/ml at a dose of 2.5–5.0 mg i.m.), ventraquil (10 mg/ml at a dose of 2.5–5.0 mg i.m.) and etomidaat (2 mg/ml at a dose of 2.0–8.0 mg i.v.). After introduction of anesthesia the rabbits were cuffed and mechanically ventilated with O2:N2O, proportion 1:1, and 2% isoflurane. During the surgical procedure methadone (2.0–5.0 mg i.v.) was given. At the end of the operation nalbuphine (10 mg/ml at a dose of 1.0–2.0 mg/kg i.v.) was administered. After surgery the rabbits were housed at the intensive care for the rest of the day and night. At the end of the study the rabbits were euthanized by an i.v. overdose of pentobarbital. Surgical technique An operation device was created by the first author (Fig. 1). The operation device consists of a base plate. On this base plate, a dynamic plate was fixed which could be moved up or down. In front of the entire length of this dynamic plate a sideways moving slide was fixed on the base plate. On this sideways moving slide another back-forward slide including the drill was fixed. Forward movement of the drill was initiated using a 1.5 kg weight. Fig. 1Operation device consisting of a base plate with a slide for sideways movement on top of this. On this slide another back-forward slide with a drill was fixed. The forward movement was initiated by a weight of 1.5 kg. Another dynamic plate was fixed on the base of the plate to move up and down. During surgery, the rabbit was fixed on this dynamic plate During surgery the rabbit was fixed on the dynamic plate. This plate made it possible that the femur or tibia were on the same height as the K-wire. The sideways moving slide was responsible for the exact position of the K-wire in front of the femur or tibia. After the rabbits were preanesthetized, X-rays were made to exclude deformities. Thereafter the animal’s hind limb was carefully shaved and prepared with a povidone-iodine solution. After this procedure, the hind limb was fixed on the testing machine. With the animal surgically draped, a straight-line skin incision was made on the lateral aspect of the femur extending from just below the anterior–inferior spine to the distal femur, followed by a straight-line skin incision on the lateral aspect of the tibia extending from just below the joint line proximally to about the joint line distally. Dissection was carried out down to the periostium. Synthes Trocar tipped K-wires of 70 mm length and 0.6 mm thickness were drilled through the diaphysis. One K-wire was drilled into the femur and one into the tibia. Drilling was performed by a rotary engine fixed at 1,200 rpm. This is the maximum drilling speed used in our daily practice. Cooling was not performed. After insertion, the K-wires were cut short and the K-wire ends were bent to the cortex. After the wounds were closed in layers, X-rays were made to check the position of the K-wires and the condition of the bone. Insertion time could be measured very accurately by analyzing the operations recorded on video camera. All the experiments were performed by the same investigator. Histological technique After termination, the femur and tibia were removed from the hind limb and fixed in 4% formaldehyde solution. They were then decalcified, cut transversely next to the K-wire hole, after the K-wire was removed gently by a pair of tweezers, and embedded in paraffin according to standard procedures. Four micrometer thick serial sections were cut until the drill hole was visible, stained with hematoxylin and eosin and evaluated under a light microscope at 400× magnification for the presence or absence of osteocytes in the osteocyte lacunae surrounding the drill holes by a single investigator. The best section was used for evaluation. The distance over which the osteocytes had disappeared perpendicular to the drill holes was measured with an interactive morphometry device (Q-PRODIT, Leica, Cambridge, UK). In each section, four distances from the drill hole to the first osteocyte bearing lacuna were measured and averaged. Statistics Pearson’s Chi-square test was used to determine the drilling time needed for osteocytes to disappear. Pearson correlation was used to highlight any significant correlation between the drilling time and the distance of the disappeared osteocytes surrounding the drill holes. A P-value <0.05 was considered statistically significant. The data were analyzed using SPSS 12.0.1 for windows. Results After surgery (t = 0) six rabbits were terminated. Both hind limbs were used resulting in 24 assessments. From the remaining eight rabbits only one hind limb was operated. Four weeks later at termination this resulted in 16 assessments (t = 4). Two t = 0 and one t = 4 assessments were lost because no sections could be produced showing the drill hole, leaving 22 t = 0 and 15 t = 4 assessments. Histological response to drilling was seen in most sections next to the drill hole (Fig. 2). At t = 0 and t = 4, all osteocyte lacunae next to the drill holes were empty in 11/22 (50.0%) and 13/15 (86.7%) of bones, respectively. At t = 4, a drilling time longer than 27 s resulted in a significant loss of osteocytes surrounding the drill holes (P = 0.008). At t = 0, a drilling time longer than 37 s resulted in a significant loss of osteocytes surrounding the drill holes (P = 0.011). Fig. 2Status of osteocytes due to K-wire drilling at 1,200 rpm. Disappeared osteocytes due to drilling are marked by a dot. The osteocytes are encircled. The dotted line indicates the border between the present and disappeared osteocytes (hematoxylin and eosin) A statistically significant positive correlation was seen between the distance of the empty osteocyte lacunae surrounding the drill hole in relation with drilling time, directly (P = 0.008, Fig. 3a) and 4 weeks after K-wire insertion (P < 0.000, Fig. 3b). Fig. 3a Distance of empty osteocyte lacunae, in micrometers, surrounding the drill hole in correlation with drilling time, in seconds, at t = 0. b Distance of empty osteocyte lacunae, in micrometers, surrounding the drill hole in correlation with drilling time, in seconds, at t = 4 Discussion The two most important findings of this experiment are that (1) osteocytes disappear especially beyond a drilling time of 27 s and (2) that there is a positive correlation between the distance of the empty osteocyte lacunae in relation with drilling time. As far as we know there are no experiments studying the presence/absence of osteocytes while using drilling time as a variable. Therefore it is interesting to notice that our results, representing daily practice by using a drilling speed of 1,200 rpm, clearly indicate that osteocytes disappear after a drilling time above 27 s in this animal model, when no cooling is used. Pallan [28] and Thompson[33] however, noticed absence of osteocytes already after a drilling time varying from 5 to 12 s. The limitation of their experiments is that the insertion force was not standardized. That could be a possible factor explaining the difference in insertion time with matching histological changes. The effect on osteocytes in our study was more pronounced 4 weeks after insertion compared to the effect on osteocytes directly after surgery. This confirms the results from Thompson [33] where the effect on osteocytes was more pronounced 72 h after the operation than at either 24 or 48 h. Apparently, not all damage to osteocytes is acute and microscopically visible, directly after K-wire insertion. This can be explained by the study of Eriksson et al. [13]. They compared histology and histochemistry for detection of bone viability directly after surgery. The histological sections revealed a 200 μm wide zone of empty osteocyte lacunae around the periphery of the drill hole while histochemistry of the same bone specimens showed a zone of 500 μm lacking diaphorase enzyme activity. This indicates that the empty osteocyte lacunae directly after surgery underestimates the extent of the drilling trauma and becomes clear as time passes. Neovascularisation is probably a reason for the increase of osteocytes disappearance in time. During heating, due to drilling, blood flow will stop in minor vessels preventing the clearance of the osteocyte lacunae [10, 13]. After neovascularisation the osteocyte lacunae can be cleared finally. Our study also showed an interesting positive correlation between the distance of the empty osteocyte lacunae in relation with time. The zones of disappeared osteocytes varied from 106 to 1,285 μm and the drilling times from 41 to 262 s. A correlation between drill time and distance was not shown before. What we do know is that Eriksson et al. showed a 200 μm wide zone of empty osteocyte lacunae around the periphery of the hole after drilling into femurs of New Zealand white rabbits while conventional irrigation was administered[13]. These results are difficult to compare because Eriksson et al. did not measure time, irrigated with saline, did not standardize the insertion force and drilled with a running speed of 20,000 rpm. Furthermore, Pallan [28] showed that the histological changes, including absence of osteocytes, were never apparent more than 250 to 500 μm from the pins [28]. Like Eriksson they did not standardize insertion force and drilling time varied from 5 to 12 s. No osteoclastic activity was noticed in the already existing cortex. This was expected as Pallan [28] hardly noticed osteoclastic activity after 6 weeks, and only slight osteoclastic activity after 8 and 10 weeks and our experiments covered a time span of only 4 weeks. We are convinced that delayed tissue response corresponds with late K-wire loosening seen in daily practice. Although we did not compare drilling, with and without cooling, this study demonstrates the need for a short drill time, less than 27 s to prevent the disappearance of osteocytes and to limit the bone resorption cascade.

Purinergic_Signal-4-1-2246002

The inflammatory effects of UDP-glucose in N9 microglia are not mediated by P2Y14 receptor activation

In this study we evaluated the functionality and inflammatory effects of P2Y14 receptors in murine N9 microglia. The selective P2Y14 receptor agonist UDP-glucose (UDPG) derived from microbial sources dose dependently stimulated expression of cyclooxygenase-2 and inducible nitric oxide synthase, and potentiated the effects of bacterial lipopolysaccharide on nitric oxide production. However, another selective P2Y14 receptor agonist, UDP-galactose, did not affect these endpoints either alone or in combination with lipopolysaccharide. Interestingly, synthetic UDPG also had no detectable pro-inflammatory effects, although P2Y14 receptors are both expressed and functional in N9 microglia. While synthetic UDPG decreased levels of phosphorylated cyclic AMP response element binding protein, an effect that was blocked by pertussis toxin, the pro-inflammatory effects of microbial-derived UDPG were insensitive to pertussis toxin. These data suggest that the pro-inflammatory effects of microbial-derived UDPG are independent of P2Y14 receptors and imply that microbial-derived contaminants in the UDPG preparation may be involved in the observed inflammatory effects.

Mol_Syst_Biol-3-_-2174632

A map of human cancer signaling

We conducted a comprehensive analysis of a manually curated human signaling network containing 1634 nodes and 5089 signaling regulatory relations by integrating cancer-associated genetically and epigenetically altered genes. We find that cancer mutated genes are enriched in positive signaling regulatory loops, whereas the cancer-associated methylated genes are enriched in negative signaling regulatory loops. We further characterized an overall picture of the cancer-signaling architectural and functional organization. From the network, we extracted an oncogene-signaling map, which contains 326 nodes, 892 links and the interconnections of mutated and methylated genes. The map can be decomposed into 12 topological regions or oncogene-signaling blocks, including a few ‘oncogene-signaling-dependent blocks' in which frequently used oncogene-signaling events are enriched. One such block, in which the genes are highly mutated and methylated, appears in most tumors and thus plays a central role in cancer signaling. Functional collaborations between two oncogene-signaling-dependent blocks occur in most tumors, although breast and lung tumors exhibit more complex collaborative patterns between multiple blocks than other cancer types. Benchmarking two data sets derived from systematic screening of mutations in tumors further reinforced our findings that, although the mutations are tremendously diverse and complex at the gene level, clear patterns of oncogene-signaling collaborations emerge recurrently at the network level. Finally, the mutated genes in the network could be used to discover novel cancer-associated genes and biomarkers. Introduction Cells use sophisticated communication between proteins in order to initiate and maintain basic cellular functions such as growth, survival, proliferation and development. Traditionally, cell signaling is described via linear diagrams and signaling pathways. As many more ‘cross-talks' between signaling pathways have been identified (Natarajan et al, 2006), a network view of cell signaling emerged: the signaling proteins rarely operate in isolation through linear pathways, but rather through a large and complex network. As cell signaling is crucial to affect cell responses such as growth and survival, alterations of cellular signaling events, such as those that arise by mutations, can result in tumor development. Indeed, cancer is largely a genetic disease that is caused by acquiring genomic alterations in somatic cells. Alterations to the genes that encode key signaling proteins, such as RAS and PI3K, are commonly observed in many types of cancers. During tumor progression, it is proposed that a malignant tumor arises from a single cell, which undergoes a series of evolutionary processes of genetic or epigenetic changes and selections so that a cell within the population can acquire additional selective advantages for cellular growth or survival, resulting in progressive clonal expansion (Nowell, 1976). Genetic mutations of the signaling proteins might overactivate key cell-signaling properties such as cell proliferation or survival and then give rise to the cell with selective advantages for uncontrolled cellular growth and promoting tumor progression. In addition, mutations may also inhibit the function of tumor-suppressor proteins, resulting in a relief from normal constraints on growth. Furthermore, epigenetic alterations by promoter methylation, resulting in transcriptional repression of genes controlling tumor malignancy, is another important mechanism for the loss of gene function that can provide a selective advantage to tumor cells. Enormous efforts have been made over the past few decades to identify mutated genes that are causally implicated in human cancer. Furthermore, a genome-wide or large-scale sequencing of tumor samples across many kinds of cancers represents a largely unbiased overview of the spectrum of mutations in human cancers (Stephens et al, 2005; Sjoblom et al, 2006; Greenman et al, 2007; Thomas et al, 2007). Most of these efforts have been made by the Cancer Genome Project (CGP, http://www.sanger.ac.uk/genetics/CGP/), which aims to identify cancer-mutated genes using genome-wide mutation-detection approaches. Similarly, genome-wide identification of epigenetic changes in cancer cells has been conducted recently (Ohm et al, 2007; Schlesinger et al, 2007; Widschwendter et al, 2007). These studies showed that a substantial fraction of the cancer-associated mutated and methylated genes is involved in cell signaling, which is in agreement with the previous finding that the most common domain encoded by cancer genes is the protein kinase domain (Futreal et al, 2004). Although there is a wealth of knowledge regarding molecular signaling in cancer, the complexity of human cancer prevents us from gaining an overall picture of the mechanisms by which these genetic and epigenetic events affect cancer cell signaling and tumor progression. Where are the oncogenic stimuli embedded in the network architecture? What are the principles by which genetic and epigenetic alterations trigger oncogene-signaling events? Given that so many genes have genetic and epigenetic aberrations in cancer signaling, what is the architecture of cancer signaling? Do any tumor-driven signaling events represent ‘oncogenic dependence' (the phenomenon by which certain cancer cells become dependent on certain signaling cascades for growth or survival)? Who are the central players in oncogene signaling? Are there any signaling partnerships generally used to generate tumor phenotypes? To answer these questions, we conducted a comprehensive analysis of cancer mutated and methylated genes on a human signaling network, focusing on network structural aspects and quantitative analysis of gene mutations on the network. Results and discussion The architecture and the relationships among the proteins of a signaling network are important for determining the sites at which oncogenic stimuli occur and through which oncogenic stimuli are transduced. Extensive signaling studies during the past decades have yielded an enormous amount of information regarding regulation of signaling proteins for more than 200 signaling pathways, most of which have been assembled and collected in public databases in diagrams. We manually curated the data on signaling proteins and their relations (activation and inhibitory and physical interactions) from the BioCarta database and the Cancer Cell Map database (see Materials and methods). We merged the curated data with another literature-mined signaling network that contains ∼500 proteins (Ma'ayan et al, 2005). As a result, we have built a human signaling network containing 1634 nodes and 5089 links. Integrative network analyses have provided numerous biological insights (Wuchty et al, 2003; Han et al, 2004; Ihmels et al, 2004; Luscombe et al, 2004; Kharchenko et al, 2005; Wang and Purisima, 2005; Cui et al, 2006). Thus, the integration of the data on mutated and methylated cancer-associated genes onto the network will help us to identify critical sites involved in tumorigenesis and increase our understanding of the underlying mechanisms in cancer signaling. To integrate mutated and methylated genes onto the network, we first collected the cancer mutated genes from the Catalogue Of Somatic Mutations In Cancer (COSMIC) database, which collects the cancer mutated genes through literature curation and large-scale sequencing of tumor samples in the CGP. We then combined these data with the cancer mutated genes derived from other genome-wide and high-throughput sequencing of tumor samples (Stephens et al, 2005; Sjoblom et al, 2006; Greenman et al, 2007; Thomas et al, 2007). The merged gene set represents a mixture of the past directed approach and current systematic screening of cancer mutations. The cancer-associated methylated genes were taken from the genome-wide identification of the DNA methylated genes in cancer stem cells (Ohm et al, 2007; Schlesinger et al, 2007; Widschwendter et al, 2007). Finally, 227 cancer mutated genes and 93 DNA methylated genes were mapped onto the network. Among the 227 cancer mutated genes, 218 (96%) and 55 (24%) genes were derived from large-scale gene sequencing of tumors and literature curation, respectively (see Materials and methods, Figure 1A). In general, cancer genes can be divided into two groups: positive regulators (oncogenes) that promote cancer cell proliferation and the negative regulators (tumor suppressors) that restrain it. By comparing the mutated genes with the known tumor suppressors, we found that only 6.6% (15 genes) of the mutated genes are known tumor suppressors and that the majority of the mutated genes are oncogenes (Supplementary Figure 1). On the other hand, methylated genes are mainly found to encode tumor suppressors in cancer cells (Supplementary Figure 1) (Ohm et al, 2007; Widschwendter et al, 2007). Cancer mutated genes are enriched in signaling hubs but not in neutral hubs Genes that, when mutated or silenced, result in tumorigenesis often lead to the aberrant activation of certain downstream signaling nodes resulting in dysregulated growth, survival and/or differentiation. The architecture of a signaling network is important for determining the site at which a genetic defect is involved in cancer. To discover where the critical tumor signaling stimuli occur on the network, we explored the network characteristics of the mutated and methylated genes. The signaling network is presented as a graph, in which nodes represent proteins. Directed links are operationally defined to represent effector actions such as activation or inhibition, whereas undirected links represent protein physical interactions that are not characterized as either activating or inhibitory. For example, scaffold proteins do not directly activate or inhibit other proteins but provide regional organization for activation or inhibition between other proteins through protein interactions. In this case, undirected links are used to represent the interactions between scaffold proteins and others. On the other hand, adaptor proteins are able to activate or inhibit other proteins through direct protein interactions. In this situation, directed links are used to represent these relations. There are two kinds of directed links. An incoming link represents a signal from another node. The sum of the number of incoming links of a node is called the indegree of that node. An outgoing link represents a signal to another node. The sum of the number of outgoing links of a node is called the outdegree of that node. We call incoming and outgoing links as signal links, whereas the physical links are neutral links. We first examined the characteristics of the nodes that represent mutated genes on the network. We compared the average indegree of the mutated genes with that of the nodes in the whole network. We found that the average indegree of the mutated nodes is significantly higher than that of the network nodes (P<1.1 × 10−6, Wilcoxon test, Supplementary Figure 2). A similar result was obtained for the average outdegree of the mutated nodes (P<6.0 × 10−14, Wilcoxon test, Supplementary Figure 2). In contrast, there is no difference of the average neutral degrees between the mutated nodes and other nodes in the network. To refine these results further, we calculated the correlations between the indegree, outdegree and neutral degree of the network nodes. We found a significant correlation between the indegree and the outdegree of the network nodes (R=0.41, P<2.2 × 10−16, Spearman's correlation), but no correlation between the indegree and neutral degree of the nodes (R=−0.02, P=0.54, Spearman's correlation). Taken together, these results suggest that cancer mutations most likely occur in signaling proteins that are acting as signaling hubs (i.e., RAS) actively sending or receiving signals rather than in nodes that are simply involved in passive physical interactions with other proteins. As these hubs are focal nodes that are shared by, and/or are central in, many signaling pathways, alterations of these nodes, or signaling hubs, are predicted to affect more signaling events, resulting in cancer or other diseases. In previous studies, we found that cancer-associated genes are enriched in hubs (Awan et al, 2007). However, these results indicate that cancer-associated genes are enriched in signaling hubs but not neutral hubs. We also investigated the relations between the node degree and the methylated genes in the network. Methylated gene nodes do not appear to differ significantly from the network nodes with regard to their indegree, outdegree and neutral degree, respectively (P=0.32, P=0.16, P=0.09, Supplementary Figure 2). These results suggest that cancer mutated genes and methylation-silenced genes have different regulatory mechanisms in oncogene signaling. Activating and inhibitory signals enhance and alleviate oncogene-signaling flows, respectively Signaling flow branching represents the splitting of one signal at a source node (Figure 1B), whereas signaling flow convergence represents the consolidation of the signals at a target node from two source nodes (Figure 1B). Both types of the signaling flows are the basic elements of the network architectural organization. In the network, when the upstream and downstream nodes of a particular signal transduction event get altered either genetically or epigenetically, we considered the transduction event (link) to be most likely selected and used in cancer signaling and defined it as an oncogenic signal transduction event (Figure 1B). If a particular oncogenic signal transduction event is frequently found in many tumor samples, we infer that the tumor cells are ‘dependent' on this highly used signaling event and call it ‘oncogene-signaling-dependent event' (Figure 1B). To investigate how cancer signaling is distributed on these signal transduction routes, we extracted all the branching and convergent signaling flow units that contain at least one oncogenic signal transduction event and conducted a quantitative analysis by overlaying the gene mutation frequency onto these units. The mutation frequency of a gene was defined as the number of tumor samples that contain that mutated gene divided by the total number of the tumor samples that are used to screen the mutations for that gene. The mutation frequency of each mutated gene was obtained by using the COSMIC database, which contains the data on more than 200 000 tumor samples screened for cancer gene mutations. For the signaling branching units, we divided the signaling flows into two groups: activating and inhibitory group (Figure 1B) and compared the gene mutation frequencies of the upstream nodes with those of the downstream nodes in each group. Interestingly, in the activating group, the upstream nodes often have lower mutation frequencies than those of the downstream nodes. In contrast, in the inhibitory group, the upstream nodes often have higher mutation frequencies than those of the downstream nodes (Table I). Statistical tests confirmed that these observations are statistically significant (Table I). Similar results were obtained for the signaling convergent units as well (Figure 1B, Table I). These results suggest that the oncogene-signaling event triggered by mutations is preferentially associated with activating downstream signaling paths or conduits. Conversely, oncogene-signaling event triggered by mutations are less likely to be associated with downstream inhibitory signaling paths. In general, there are far more activating signaling flows than inhibitory ones in the network. Thus, we hypothesized that the downstream genes of the network, especially the genes of the output layer of the network, would have a higher mutation frequency. To test this possibility, we compared the average gene mutation frequency of the nuclear proteins, which represent the output layer members of the network, with that of the other network genes. Indeed, the nuclear genes have higher mutation frequency than others (P=0.01, Wilcoxon test), which complements with our previous finding that cancer-associated genes are enriched in the nuclear proteins (Awan et al, 2007). In contrast, the distributions of the methylated genes have no such preference, suggesting that DNA methylated genes do not tend to directly affect the output layer of the network. These results strongly suggest that the genes in the output layer of the network, which play direct and important roles in determining phenotypic outputs, are frequent targets for activating mutations. The importance of this output layer is reinforced by our previous observations that the expression of the output layer genes of the signaling network is heavily regulated by microRNAs (Cui et al, 2006). Mutated and methylated genes are enriched in positive and negative regulatory loops, respectively The complex architecture of signaling networks can be regarded as consisting of interacting network motifs, which are statistically overrepresented subgraphs that appear recurrently in networks. A signaling network motif, also known as regulatory loops in biology, is a group of interacting proteins capable of signal processing. They bear specific regulatory properties and mechanisms (Babu et al, 2004; Wang and Purisima, 2005). The structure and the intrinsic properties of the frequently occurring network regulatory motifs give us a functional view of the organization of signaling networks. Thus, the study of the distributions of the mutated and methylated genes in the network motifs will provide insights into cancer-signaling regulatory mechanisms. We first examined the mutated genes on the feed-forward loops, in which the first protein regulates the second protein, and both proteins regulate the third protein. We classified the feed-forward loops into four subgroups (labeled 0–3) based on the number of nodes that are mutated genes. We calculated the ratio (Ra) of positive (activating) links to the total directed (positive and negative) links in each subgroup and compared it with the average Ra in all the feed-forward loops, which is shown as a horizontal line in Supplementary Figure 3. The Ra (∼0.7) in subgroup 0 is less than the average Ra (∼0.74) of all the feed-forward loops (P<1.9 × 10−9, Fisher's test). However, as the number of mutated nodes rises, the Ra for the corresponding group increases to a maximum of ∼0.93 (Supplementary Figure 3, Supplementary Table 1). We obtained similar results, when we extended the same analysis to all the 3-node- and 4-node-size network motifs (Figure 2, Supplementary Table 1). These motifs show a clear positive correlation between positive link ratio and the number of mutated genes in the motifs. These results suggest that cancer gene mutations occur preferentially in positive regulatory motifs. In contrast, all the 3-node and 4-node size motifs show an obviously negative correlation between positive link ratio and the number of methylated genes in the motifs (Figure 2, Supplementary Table 2). These results suggest that cancer gene methylation preferentially occurs in negative regulatory motifs. A similar trend was found for the 15 known tumor suppressors (Supplementary Figure 4a–d), which is in agreement with the notion that cancer-associated methylated genes play roles as tumor suppressors. Collectively, these facts suggest that mutated and methylated genes have different regulatory mechanisms in cancer signaling and support the notion that gene mutations and methylations are strongly selected in tumor samples. Signaling information propagates through a series of built-in regulatory motifs to contribute to cellular phenotypic functions (Ma'ayan et al, 2005). The transition from a normal cellular state into a long-term deregulated state such as cancer is often driven by prolonged activation of downstream proteins, which are regulated by upstream proteins or regulatory motifs or circuits. Positive regulatory loops (Ferrell, 2002) could amplify signals, promote the persistence of signals, serve as information storage and evoke biological responses to generate phenotypes such as cancer. In cancer cells, constitutive activation of the oncogene signaling is necessary. Neutral mutations do not affect protein function, whereas missense mutations may have positive or negative effects on protein activity. The enrichment of gene mutations in positive regulatory loops suggests that the mutants in the motifs must have gain of function or increase their biochemical activities compared to the wild-type genes in order to constitutively activate downstream proteins. Indeed, a recent survey showed that 14 out of the 15 PI3K mutants in tumors have gain of function (Gymnopoulos et al, 2007). Gain-of-function mutants in a positive regulatory loop afford the amplification of weak input stimuli and serve as information storage to extend the duration of activation of the affected downstream proteins. This might allow the downstream signaling cascades to persistently hold and transfer information leading to tumor phenotypes. Promoter gene methylation is a known mechanism of inducing loss of function by inhibiting the expression of genes (Ohm et al, 2007; Widschwendter et al, 2007). Negative regulatory loops controlled by tumor-suppressor proteins repress positive signals and play an important role in maintaining cellular homeostasis and restraining the cellular state transitions (Ma'ayan et al, 2005). A loss of function of gene methylation in a negative regulatory loop could break the negative feedbacks, thereby releasing the restrained activation signals and promoting oncogenic state transitions. Homeostasis relies on the balance between positive and negative signals in crucial components of the network. Both the gain-of-function mutated genes in positive regulatory loops and the loss-of-function methylated genes in negative regulatory loops could break this delicate balance, thus promoting state transitions and generating tumor phenotypes. Therefore, both mutated and methylated genes and their regulatory loops (oncogenic regulatory loops) are critical components of the network where the oncogenic stimuli occur. An oncogene-signaling map emerges from the network In the language of networks, genes whose mutations or epigenetic silencing are crucial to trigger oncogene signaling might link together as components in the network. Identification of such components will help us to discover the relationships and structural organizations of the oncogenic proteins. To uncover the architecture of cancer signaling and to gain insights into the higher-order regulatory relationships among signaling proteins that govern oncogenic signal stimuli, we mapped all genetic mutations and epigenetically silenced genes onto the network. We found that most of these genes (67%) are connected together to form a giant, linked network component. Randomization tests confirmed that such a component is unlikely to be formed by chance (P<2 × 10−4). To build an oncogenic map, we included other mutated and methylated genes that are not present in the composition of the component into the giant network component based on node connectivity (see Materials and methods). The resulting oncogene-signaling map consists of approximately 20% of the signaling network nodes (326 nodes, 892 links) and includes almost 90% of the mutated and methylated genes (Figure 3). The map showed different network topological characteristics from the signaling network. For example, the average length of the map is less than that of the signaling network (5 versus 6, P<2 × 10−16, Wilcoxon test). On the other hand, the average clustering coefficient of the map is greater than that of the signaling network (0.08 versus 0.04, P=0.06, Wilcoxon test). These results suggest that oncogenic proteins tend to have more interactions and signaling regulatory relationships. The emerging oncogene-signaling map represents a ‘hot area' where extensive oncogene-signaling events might occur. As a proof of concept, we found that the MAPK kinase and TGFβ pathways, which are well-known cancer-signaling pathways, are embedded in the map. For example, 50 out of 87 proteins in the MAPK kinase pathway (Supplementary Table 3) and 22 out of 52 proteins in the TGFβ pathway (Supplementary Table 4), respectively, are included in the map. More importantly, in addition to known oncogenic pathways, there are many other novel candidate cancer-signaling cascades present in the map. For a given gene mutation in a tumor, one could use this map to generate testable hypotheses to discover the underlying oncogene-signaling cascades in that tumor. As mentioned above, oncogene-signaling-dependent events, which we define as the interactions between the cancer mutated or methylated genes, are frequently found in tumor samples and represent various oncogene-driving events that could play more critical roles in generating tumor phenotypes. To systematically identify such events and discover how they are organized in the map, we charted the gene mutation frequency onto the map and highlighted the signaling links between any two genes that have high mutation frequencies. Most genes have mutation frequencies lower than 2%, whereas a handful of genes have very high mutation frequencies, such as p53 (41%), PI3K (10%) and RAS (15%) (see Materials and methods). Therefore, a gene mutation frequency equal to or greater than 2% was considered as high. Interestingly, nearly 10% of the links in the map are oncogene-signaling-dependent events. Certain signaling events such as Pten-PI3K and RAS-PI3K in the map are well-known oncogene-signaling-dependent events/cascades that are frequently used in various cancers. As shown in Figure 3, most oncogene-signaling-dependent events are connected, and three major regions that contain densely connected oncogene-signaling-dependent events emerge in the map: the first region (p53 region) contains mainly tumor suppressors such as p53, Rb, BRCA1, BRCA2 and p14 (CDKN2A) etc.; the second region (RAS region) contains mainly well-known oncogenes such as RAS, EGFR and PI3K etc.; and the third region (TGFβ region) contains SMAD3, SMAD4 and a few other TGFβ-signaling proteins. Interestingly, genes in the p53 and TGFβ regions are also heavily methylated in cancer stem cells, suggesting that these regions are involved in the early stage of oncogenesis. Other methylated genes are intertwined with the mutated genes in the map, suggesting that they share some oncogene-signaling cascades and might be regulated to cooperate in cancer signaling via gene mutation and/or methylation. Notably, it seems that, in cancer stem cells, TGFβ-signaling pathway is shut down, supporting its known role as a tumor suppressor in the early stages of tumorigenesis (Hanahan and Weinberg, 2000; Siegel and Massague, 2003). These results suggest that the crucial players of oncogene signaling tend to be closely clustered and regionalized. This map uncovers the architectural structure of the basic oncogene signaling and highlights the signaling events that are highly conserved in generating tumor phenotypes. Functional collaboration of genes between oncogene-signaling blocks The oncogene-signaling map can be decomposed into several network communities or network themes (Zhang et al, 2005), in which each network community contains a set of more closely linked nodes and ties to particular biological functions. To discover such network communities, we implemented and applied an algorithm that detects network communities to the map. As a result, 12 network communities, ranging in size from 11–65 nodes (Supplementary Table 5), called ‘oncogene-signaling blocks', were found in the map. Structurally, the nodes within each block have more links and signaling regulatory relations among themselves than others. The genes in each block share similar biological functions such as cell proliferation, development and apoptosis (Supplementary Table 5). We further performed Gene Ontology (GO) enrichment analysis for each oncogene-signaling block using DAVID Tools (http://david.abcc.ncifcrf.gov/home.jsp). Most of the oncogene-signaling blocks are enriched with protein serine/threonine kinase activity (Supplementary Table 6), which is well known to take part in tumorigenesis. Notably, Block 1 is enriched with cell surface receptor-linked signaling, whereas Block 10 is enriched with intracellular signaling cascades. Block 11 is enriched with tumor suppressors and biological processes such as apoptosis and cell cycle. These results suggest that certain blocks are taking part in different parts/kinds of signaling, that is, cell surface receptor-related signaling, intracellular signaling, cascade signaling and apoptotic signaling. However, three oncogene-signaling blocks have no GO enrichment detected. One of the reasons is that a fraction of the genes in these blocks is not well annotated yet. For example, about one-third of genes in Block 6 have no GO term associated. We asked if the genes in each block could operate in a compensatory or concerted manner to govern a set of similar functions. Toward this end, we surveyed the gene mutations in tumor samples where at least two genes are screened for mutations. As a result, the co-occurrence in tumor samples of 25 mutated gene pairs is found to be statistically significant (Supplementary Table 7). Significantly, only three collaborative gene pairs came from the same block, whereas other collaborative gene pairs came from two different blocks, with predominantly one of them arising from Block 11 (defined as p53 block), which contains p53, Rb, p14, BRCA1, BRCA2 and several other genes involved in control of DNA damage repair and cell division. Collectively, these results suggest that the signaling genes from different blocks most likely work together in a complementary way to generate tumor phenotypes. We further asked which oncogene-signaling blocks work together to produce a tumor phenotype. To address this question, we surveyed the gene mutations in the tumor samples where at least two gene mutations are found. In total, 592 tumor samples fit this criterion. We used the 592 samples to build a matrix (M) where samples are rows and the signaling blocks are columns. If a gene of a particular signaling block (b) gets mutated in a tumor sample(s), we set Ms,b to 1, otherwise we set Ms,b to 0. A heatmap was generated using the matrix (Figure 4A). If a sample contains statistically significant co-occurring mutated gene pairs (see Supplementary Table 7), these pairs were highlighted in the heatmap. Samples were organized based on the cancer types they belong to. Several cancer types such as breast, central nervous system, blood, lung, pancreas and skin tumors that have relatively more samples were also highlighted in the heatmap. As shown in Figure 4A, two signaling blocks have statistically significant enrichment of gene mutations (P<2 × 10−4, randomization tests), suggesting that genes in these two signaling blocks are predominantly used to generate tumor phenotypes. One oncogene-signaling block (Block 1, defined as RAS block) contains genes like RAS, EGFR and PI3K etc., which share similar biological functions such as cell proliferation, cell survival and cell growth, whereas the other is the p53 block, which share similar biological functions such as cell cycle checkpoint control, apoptosis and affecting genomic instability (Supplementary Table 5). These two blocks also represent the two oncogene-signaling-dependent regions (p53 and RAS regions) in Figure 3, respectively. When a tumor sample has a mutation in a gene from the RAS-signaling block, it is also most likely to contain a mutation in a gene from the p53 block (P<2 × 10−4). To check if this phenomenon is primarily due to a particular pair of genes, we calculated the likelihood of co-occurrence for each pair of the genes, of which one gene is mutated in one block and the other gene is mutated in the other block. We found that the P-values for gene pairs are always significantly greater than that for the pair of Blocks RAS and p53. For example, the P-value of co-occurrence of RAS (in Block RAS) and p53 (in Block p53) mutations is 0.01, which is greater than that of the two blocks (P<2 × 10−4). This indicates that these two oncogene-signaling blocks collaborate to generate tumor phenotypes for most tumors. Experimental examples have shown similar gene collaboration in tumorigenesis: activation of RAS (RAS block) and inactivation of p53 (p53 block) induce lung tumors (Meuwissen and Berns, 2005), whereas activation of RAS (RAS block) and inactivation of p16 (p53 block) induce pancreatic tumors (Obata et al, 1998). In general, tumor cells exhibit either elevated cell proliferation or reduced differentiation or apoptosis relative to normal cells. The oncogenic blocks we have identified, especially the RAS and p53 blocks, encode functions that are tumor-related, such as cell cycle control, cell proliferation and apoptosis (Supplementary Figure 5). Activation of genes in the RAS block promotes the cell proliferation, whereas inactivation of genes in the p53 block prevents apoptosis. Thus, a functional collaboration between the genes in these two blocks would promote synergistic cancer signaling and foster tumorigenesis. Notably, we found that at least one gene mutation in the p53 block had occurred in the tumor samples we examined. In other words, the p53 block is involved in generating tumors for most cancers. This result suggests that the p53 block is a central oncogene-signaling player and essential in tumorigenesis. This finding is further supported by the following observations. (a) To become oncogenic, tumor suppressors require loss-of-function mutations, which occurs more often than gain-of-function mutations (Gymnopoulos et al, 2007). Indeed, the average gene mutation frequency in the p53 block is higher than that of other signaling blocks including the RAS block. (b) The methylation of genes in the cancer stem cells resulting in long-term loss of expression represents the early stage of the tumorigenesis. In fact, most of the members of the p53 block are methylated in cancer stem cells. These facts further support that the p53 block might play an important role in the earlier stages of oncogenesis. (c) Gene methylation or inactivating mutations of DNA damage checkpoint genes such as p53 induce genome instability and thus increase the chance of other gene mutations, including the genes of other oncogene-signaling blocks that could functionally collaborate with the p53 block genes to generate tumor phenotypes. Using the map as a framework, we benchmarked the mutated genes in the NCI-60 cell lines, which represent a panel of well characterized cancer cell lines and various cancer types. A systematic mutation analysis of 24 known cancer genes showed that most NCI-60 cell lines have at least two mutations among the cancer genes examined (Ikediobi et al, 2006). We built a matrix and constructed a heatmap using these cell lines and their mutated genes as described above (Figure 4B). Overall, the pattern obtained from the NCI-60 panel resembles that of the 592 tumor panel with both RAS and p53 blocks enriched with gene mutations and exhibiting statistically significant collaborations in these cell lines (Figure 4B, P<2 × 10−4), which is in agreement with the earlier observations. We also benchmarked the mutated genes derived from a genome-wide sequencing of 22 tumor samples (Sjoblom et al, 2006). Among these 22 samples, 10 breast and 10 colon tumor samples have at least two gene mutations in the map. As shown in Figure 4C and D, the p53 block is enriched with gene mutations. For the 10 colon tumor samples, collaboration between Block 6 and Block p53 is established, but for the 10 breast tumors, collaborative patterns between multiple blocks emerged. To further examine the block collaborative patterns in individual tumor types in higher resolutions, from the heatmap (Figure 4A) we extracted the sub-heatmap for several tumor types that are better represented among the 592 tumor samples, that is, they have relatively more samples within the 592 samples (Figure 5). As shown in Figure 5, signaling block collaborative patterns are tissue dependent and are classified into two groups. One group contains pancreas, skin, central nervous system and blood tumors that have simple block collaborative patterns. In these tumors, signaling collaborations are mainly between Block p53, Block RAS with some minor contributions from Blocks 5, 6 or 7, suggesting that they predominantly use these oncogene-signaling routes to generate tumors resulting in relatively homogenous cancer cell types. The other group contains breast and lung tumors that also contain large proportions of mutations from the p53 block, but also have complex patterns of collaborations between assortments of multiple blocks, suggesting that these tumors may have a larger variety of oncogene-signaling routes, which may explain, in part, the heterogeneous nature of the tumor subtypes in this category. These results might also explain why both lung and breast cancers are the most common types of human tumors. In this study, the cancer mutated genes were collected from a ‘directed approach' (i.e., mutational analysis of specific genes, such as p53) and a ‘large-scale approach' (i.e., large-scale sequencing of tumor samples). We tested whether the mutated genes from the directed approach introduce bias to our analysis. Literature-curated cancer mutated genes (directed approach) have been assembled in the Cancer Gene Census (Futreal et al, 2004), of which 115 genes were found in the human signaling network. As of November 30th, 2006, among the 115 Cancer Gene Census genes, mutations in 55 of them have been further validated by additional experimental evidence (i.e., other independent experiments confirming the mutation of these specific genes in cancer samples have been documented in the COSMIC database), whereas 60 of them have no such evidence in the COSMIC database (see Materials and methods). In fact, we included only these 55 literature-curated genes in the cancer mutated gene set (227 genes) used in all of our analyses above (see Materials and methods). Of the 55 literature-curated genes, only 9 were not already present in the output of large-scale sequencing of tumor samples (Figure 1A). We removed these 9 genes from the cancer mutated gene set (227 genes), mapped rest of the genes onto the human signaling network and rebuilt an oncogene-signaling map, oncogene-signaling blocks and a heatmap. On the other hand, we added the 60 literature-curated genes, which have no independent supporting evidence in the COSMIC database, to the 227-gene set and obtained 287 genes. Using these 287 genes, we reran the analyses mentioned above. In these two analyses, although the gene members of each oncogene-signaling block have some minor differences with those of the original blocks, the major collaboration patterns of oncogene-signaling blocks remain largely unchanged (Supplementary Figure 6a and b), suggesting that our findings are robust to addition or removal of the cancer mutated genes derived from the directed approach. The mutated genes in the network provide a predictive power A substantial number (∼20%) of mutated genes were found in the network. We asked if a gene that has more links to mutated genes in the network is most likely to be cancer associated. To answer this question, we extracted the nonmutated genes that have more than one link to the mutated genes and then grouped them based on their link numbers to the mutated genes. We interrogated a cancer-associated gene set (Supplementary Table 8) compiled from literature mining (see Materials and methods) to find out how many genes in each group are cancer associated. As shown in Figure 6, the more mutated genes a gene links to, the more probably it is cancer associated. When the link number of the network genes is more than six, ∼80% of them are cancer-associated genes. For example, SHC, a gene that has been implicated in cancer metastasis (Jackson et al, 2000), has numerous links to the mutated genes in the network. To further investigate the predictive power of the mutated genes in the network, we took the 14 network genes, which not only have at least four links to the mutated genes, but also are not implicated in cancer in the literature, to perform a survival analysis using a microarray data set that contains the gene expression profiles and survival information for 295 breast tumor samples. As a result, the expression profiles of 5 out of the 14 genes (36%) are able to discriminate ‘good' and ‘bad' tumors (i.e., patients having ‘bad' tumors have higher chance of tumor recurrence and short survival time). Therefore, these genes are potentially novel biomarkers. In contrast, less than 10% of the nonmutated network genes have similar discriminatory power. These results suggest that the network genes, which have more links to the mutated genes, have more chance to be perturbed in tumorigenesis and be associated with cancer. Practically, the mutated genes in the network provide a predictive power that can be used to discover novel biomarkers of tumors. Concluding remarks Although a wide variety of genetic and epigenetic events contribute to the signaling of tumorigenesis, it has been challenging to gain a global view of where and how they affect the signaling alterations to generate tumors on the entire signaling network. By integrative analysis of the human signaling network with cancer-associated mutated and methylated genes, we uncovered an overall picture of the network architecture where the oncogenic stimuli occur and the regulatory mechanisms involving mutated and methylated genes. Mutations, the majority of which are activating, preferentially occur in the signaling hub genes (but not neutral hubs) and the genes of the positive regulatory loops, whereas methylated alterations tend to occur in the genes of the negative regulatory loops. Cancer and cell signaling have been well established, and extensive efforts have been made to illustrate cancer signaling during the past few decades. However, it has been a struggle to get clues of how the oncogene signaling is structurally and functionally organized. In this analysis, we extracted an oncogene-signaling map, which provides a blueprint of the oncogene signaling in cancer cells. From the map, we discerned that the oncogene-signaling-dependent events form three highly connected regions that resemble oncogene-signaling superhighways frequently used in tumorigenesis. Topologically, the map has been divided into 12 oncogene-signaling blocks. Functional collaborations between subsets of these blocks are underlying tumorigenesis. In most tumors, genes in both p53 and RAS blocks often get mutated, although the combinations of p53 with other signaling blocks are also found in a small fraction of tumors. Analysis of the NCI-60 cell line panel mutations showed the enrichment of gene mutations in p53 and RAS blocks, which is similar to the patterns found in the 592 samples. Furthermore, we can dissect some of this functional collaboration among different tumor types. These facts indicate that at least two signaling gene mutations, one from the p53 block and the other from another block, are necessary for tumorigenesis. This fact supports the notion that both the prevention of cell death (p53 block) and the promotion of cell proliferation (RAS or other blocks) are necessary to generate most tumors. At present, a number of researchers doubt or even argue against the value of large-scale human cancer genome sequencing as a meaningful or efficient strategy in cancer research. Their arguments are based on the following observations (Chng, 2007): (a) previous large-scale human cancer genome sequencing revealed that each tumor has a different mutation pattern, and the prevalence and patterns of somatic mutations in human cancers are tremendously diverse and complex (Kaiser, 2006; Sjoblom et al, 2006; Greenman et al, 2007); (b) the interpretation of such complex somatic alterations is a formidable challenge (Chanock and Thomas, 2007; Thomas et al, 2007). We mapped the mutation data from the genome-wide sequencing tumor samples (Sjoblom et al, 2006) using the oncogene-signaling map as a framework. Although the number of mutated genes is impressive in toto, most signaling gene mutations are limited to 2–3 critical mutations, divided among several signaling blocks, per individual tumor. This result suggests that the mutations in the samples of the same tumor type might share a similar underlying signaling mechanism, because each oncogene-signaling block contains a set of genes linked together through shared regulatory relations and key input and/or output signaling nodes that are involved in tumorigenesis. These findings imply that although the mutations seem tremendously diverse and complex at the gene level, clear patterns emerge recurrently at the network level in most tumors. Therefore, with proper bioinformatics analysis, large-scale cancer genome sequencing efforts would be fruitful in finding appropriate combinations of biological targets for cancer diagnostic and therapeutics. In summary, this work revealed novel insights into the oncogenic regulatory mechanisms, oncogene-signaling network architecture and oncogene-signaling cooperative relationships that drive cancer development and progression. It also highlights the emergence of the central players in cancer signaling. Cancer studies have integrated microarray, knowledge, pathways and networks (Liu and Lemberger, 2007), but not genetic and epigenetic data yet. However, as the next generation of genome sequencing technology becomes more accessible and affordable, much more efforts involving genome-wide sequencing of large number of tumor genomes will be conducted. Our work provides a conceptual and technical framework for incorporating the genome sequencing outputs and other types of data such as microarray profiles to get more insights into the cancer-signaling mechanisms that will facilitate the identification of key genes for biomarkers and drug development. Materials and methods Data sets used in this study Human signaling network To build up the human signaling network, we manually curated the signaling molecules (most of them are proteins) and the interactions between these molecules from the most comprehensive signaling pathway database, BioCarta (http://www.biocarta.com/). The pathways in the database are illustrated as diagrams. We manually recorded the names, functions, cellular locations, biochemical classifications and the regulatory (including activating and inhibitory) and interaction relations of the signaling molecules for each signaling pathway. To ensure the accuracy of the curation, all the data have been crosschecked four times by different researchers. After combining the curated information with another literature-mined signaling network that contains ∼500 signaling molecules (Ma'ayan et al, 2005), we obtained a signaling network containing ∼1100 proteins (Awan et al, 2007). We further extended this network by extracting and adding the signaling molecules and their relations from the Cancer Cell Map (http://cancer.cellmap.org/cellmap/), a database that contains 10 manually curated signaling pathways for cancer. As a result, the network contains 1634 nodes and 5089 links that include 2403 activation links (positive links), 741 inhibitory links (negative links), 1915 physical links (neutral links) and 30 links whose types are unknown (Supplementary Table 9). To our knowledge, this network is the biggest cellular signaling network at present. Cancer mutated genes The cancer mutated genes were taken from the COSMIC database (http://www.sanger.ac.uk/genetics/CGP/cosmic/) and other large-scale or genome-wide sequencing of tumor samples (Sjoblom et al, 2006; Greenman et al, 2007; Thomas et al, 2007). COSMIC database contains manually curated cancer mutated genes and the information of tumor samples, mutated sequences from literature and the output from the CGP's large-scale sequencing of tumor samples (Davies et al, 2005; Stephens et al, 2005; Greenman et al, 2007). The literature-curated genes were compiled as the Cancer Gene Census (Futreal et al, 2004), which is accessible in COSMIC database. The CGP is using human genome sequences and high-throughput mutation detection techniques to identify somatically acquired sequence mutations and hence to identify genes critical in the development of human cancers. A few recent publications (Davies et al, 2005; Stephens et al, 2005; Greenman et al, 2007) represent a small fraction of the CGP output. In addition, COSMIC database has provided mutation frequencies for most of the cancer mutated genes. The cancer gene mutation frequency of a gene is defined as the ratio of samples containing the mutated gene to the total samples screened for that gene. In the database, about one-third of the literature-curated mutated genes (Cancer Gene Census genes) have nonzero mutation frequencies, suggesting that the literature curation of these genes (i.e., included them into the Cancer Gene Census) has been supported by one or more other independent experiments. For the network analysis in this study, we first intersected the network genes with the literature-curated mutated genes. As a result, we obtained 115 genes (Supplementary Table 10), of which 55 genes (Supplementary Table 10) have nonzero mutation frequencies. Meanwhile, we intersected the network genes with the mutated genes derived from the CGP large-scale sequencing output and several other genome-wide and high-throughput sequencing of tumor samples (Stephens et al, 2005; Sjoblom et al, 2006; Greenman et al, 2007; Thomas et al, 2007). As a result, we obtained another gene set containing 218 genes. Finally, we obtained 227 genes by merging the 55 genes and the 218 genes mentioned above. Among these 227 genes, 218 (96%) and 55 (24%) genes were collected from the large-scale sequencing of tumors and literature curation, respectively (Figure 1A). Notably, 46 genes (84%) of the literature-curated genes were overlapped with the mutated genes derived from the large-scale gene sequencing of tumors. The genes and their mutation frequencies from sequencing of tumors and literature were collected in Supplementary Table 10. Methylated genes in cancer stem cells We obtained 287 DNA-methylated genes from the three recent genome-wide determinations of the methylated genes from cancer stem cells (Ohm et al, 2007; Schlesinger et al, 2007; Widschwendter et al, 2007). Out of the 287 genes, 93 were mapped onto the human signaling network (Supplementary Table 11). Cancer-associated gene set The cancer-associated gene set contains the following data sources: (a) the cancer mutated genes we mentioned above; (b) a literature-mined breast cancer gene set from plasmID database (http://plasmid.hms.harvard.edu/GetCollectionList.do); (c) the genes extracted from the NCBI's Online Mendelian Inheritance in Man (OMIM) data set using the keywords such as ‘cancer', ‘tumor' and ‘onco' etc. The cancer-associated gene list contains 2128 genes (Supplementary Table 8). Microarray data Gene expression profiles and the patients' survival data for the 295 breast tumor samples were obtained from Chang et al (2005). Oncogenic map extraction To extract an oncogenic map from the human signaling network, we mapped all the mutated and methylated genes onto the network. As a result, 67% of these genes are connected together to form a giant, linked network component. To include the mutated and methylated genes that are not present in this network component, we first found one shortest path between such a gene and a component node. If the length of the shortest path is 2 (i.e., the gene reaches one of the component nodes via a nonmutated network node), we linked that gene and the node on the shortest path into the component. A Java program had been written to implement this procedure (Supplementary File 1). Network analysis To extract the members of the branching and convergent units and 3-node- and 4-node-size network motifs, mfinder program (Kashtan et al, 2004) was used. To detect the signaling network communities from the oncogene-signaling map, we applied a network community algorithm (Newman, 2006). Analyzing the enrichment of the mutated and methylated genes in the network motifs We mapped the mutated and methylated genes onto each type of the motifs. We then counted the number of mutated or methylated genes in each motif and classified each type of motif into several subgroups based on the number of nodes that are mutated or methylated genes. We then calculated the ratio (Ra) of the activation links to the total activation and inhibitory links in each subgroup. Randomization tests We performed randomization tests to evaluate the statistical significance of the observations. A more detailed explanation of the randomization tests was described previously by Wang and Purisima (2005). Survival analysis To evaluate the prognostic value of a gene based on the gene expression profiles and the survival information of the tumor samples, we performed Kaplan–Meier analysis by implementing the Cox–Mantel log-rank test using R, a statistical computing language (http://www.r-project.org/). If the P-value is less than 0.05, the gene was thought as statistically significant to classify the tumor samples into ‘good' and ‘bad' groups. Supplementary Material Supplementary Figures Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7 Supplementary Table 8 Supplementary Table 9 Supplementary Table 10 Supplementary Table 11 Supplementary Java File Supplementary Tables and Java File Legends

Mol_Neurobiol-3-1-2039784

Neuronal Chemokines: Versatile Messengers In Central Nervous System Cell Interaction

Whereas chemokines are well known for their ability to induce cell migration, only recently it became evident that chemokines also control a variety of other cell functions and are versatile messengers in the interaction between a diversity of cell types. In the central nervous system (CNS), chemokines are generally found under both physiological and pathological conditions. Whereas many reports describe chemokine expression in astrocytes and microglia and their role in the migration of leukocytes into the CNS, only few studies describe chemokine expression in neurons. Nevertheless, the expression of neuronal chemokines and the corresponding chemokine receptors in CNS cells under physiological and pathological conditions indicates that neuronal chemokines contribute to CNS cell interaction. In this study, we review recent studies describing neuronal chemokine expression and discuss potential roles of neuronal chemokines in neuron–astrocyte, neuron–microglia, and neuron–neuron interaction. Introduction Chemokines are small proteins that are able to induce a chemotactic response in cells expressing the corresponding chemokine receptors. Since the discovery of the first protein with chemotactic activity [1], the chemokine family has expanded to approximately 50 chemokines [2] and 20 receptors [3]. Chemokines have been divided into four groups based on the position of four conserved cysteine residues in the N-terminal region of the protein. The two largest groups are CXC and CC. The first two cysteines in the CXC group are separated by one amino acid residue, whereas the first two cysteines in the CC group are adjacent to each other [4, 5]. The two small groups are the C chemokines, with only one cysteine in the N-terminal region, and the CX3C chemokine, where the first two cysteines are separated by three amino acid residues [5]. Chemokine receptors are designated according to the chemokine group they preferentially bind. For example, CC chemokines bind to CC receptors and so on. There has yet only been one exception reported, namely CCL21, that, in addition to CCR7, also binds to CXCR3 [6–8]. All chemokine receptors belong to the family of G-protein coupled receptors (GPCRs). In general, GPCRs can bind many different G-proteins, allowing for a great variety of intracellular signaling pathways (for excellent review, see [9]). The majority of chemokine-induced responses are inhibited by pertussis toxin (PTX), indicating that mediate many effects [10]. Chemokine receptors can activate intracellular targets like adenylcyclase, phospholipases, GTPases like Rho, Rac, and Cdc42 and pathways of major kinases like mitogen-activated protein kinase (MAPK) and phosphatidyl inositol-3 kinase (PI3-K) [11, 12]. This diversity of intracellular signaling shows that chemokine receptors, in addition to pathways involved in cell migration, also activate other pathways and may, in that way, control a great spectrum of cellular functions [13, 14]. Chemokines are well-known regulators of peripheral immune cell trafficking under both physiological and pathological conditions (reviewed by [15–17]). In addition to chemo-attraction of immune cells, chemokines have been implicated in a variety of cell functions, such as early development, formation of secondary lymphoid organs, wound healing, angiogenesis and angiostasis, regulation of adhesion molecule expression, development of Th1/Th2 profiles, tumor growth, and metastasis [5, 14, 18–24]. Thus, from being molecules thought to solely orchestrate immune cell migration, chemokines are now considered versatile messengers with the ability to control the interaction between a wide diversity of cell types. In addition to their presence in the periphery, numerous studies have demonstrated that chemokines are also expressed in the central nervous system (CNS), where they play a crucial role in physiological and pathological conditions, such as development, synaptic transmission, homeostasis, injury, and disease-associated neuroinflammation [19, 25, 26]. Although astrocytes and microglia are the primary source of chemokines, there is evidence that neurons express and secrete chemokines as well, indicative of a neuronal contribution to chemokine signaling. In this paper, we review recent studies describing neuronal chemokine expression and discuss the potential roles of neuronal chemokines in neuron–astrocyte, neuron–microglia, and neuron–neuron interaction. Neuronal Chemokine Expression Approximately 60 studies describe chemokine expression in neurons under physiological and pathological conditions (see Table 1). These studies, of which the majority is published in the last 3 years, are reviewed in the following sections. Table 1Neuronal chemokine expressionChemokineSpeciesConditionRNAProteinReferencesCCL2hBrain+[39]Spinal cord, ALS↑[35]Monoculture↑+[41]Cell line+[38]rBrain+[37]Brain, cranial nerve injury↑↑[30]Brain, ischemia↑↑[28]Retina, ischemia↑[168]Spinal cord, peripheral nerve injury↑↑[31–34]Monoculture+[164]mBrain, ischemia↑[27, 29]Spinal cord, ALS model↑[36]Monoculture, West Nile virus↑[43]CCL3hBrain, AD≈[40]Monoculture↑+[41]rRetina, ischemia↑[168]Monoculture+[51]mMonoculture↑[42]CCL4hMonoculture↑+[41]rRetina, ischemia↑[168]mMonoculture↑[42]CCL5hMonoculture↑+[41]mMonoculture↑↑[43, 44]CCL20rMonoculture+/↑[51]Trigeminal neuron culture↓[169]CCL21mBrain, ischemia↑[45]Monoculture↑↑[45, 46]Neonatal hippocampal slice culture↑[46]CXCL1/2/3rMonoculture++[51]CXCL8hMonoculture+[41]CXCL9mMonoculture+[43]CXCL10hBrain, HIV↑[54]Mixed brain culture+[54]macBrain, HIV↑[54]rBrain, ischemia↑[52]mBrain, entorhinal cortex lesion↑[53]Brain, West Nile virus↑↑[43]Monoculture++[43]CXCL11mMonoculture+[43]CXCL12hBrain, HIV↑[59]Monoculture+[59]rBrain++[55, 60–62]Monoculture++[58, 60, 87]mBrain, ischemia≈/↓[56]Brain, LPS injection≈[56]Mixed brain culture+[59]CX3CL1hBrain, MS↑[113]Brain, HIV↑[170]Spinal cord+[66]Monoculture≈≈/↑/↓[74, 75, 138]Cell line≈↑/↓a[66, 75, 138]macBrain+[66]rBrain and spinal cord, EAE≈[63, 65, 66, 68]Spinal cord, peripheral nerve injury≈≈[171, 172]Brain, LPS injection≈[67]Brain, KA injection≈[67]Monoculture≈↑/↓a[64, 65, 73, 76, 78, 118]mBrain++[69]Brain, prion disease≈[67]Brain, LPS injection≈[67]Brain, KA injection≈[67]Brain, EAE≈[66]Monoculture≈≈[77]Cell line≈≈[77]h Human, mac macaque, r rat, m mouse, ALS amyotrophic lateral sclerosis, AD Alzheimer’s disease, HIV human immunodeficiency virus, MS multiple sclerosis, EAE experimental autoimmune encephalomyelitis, LPS lipopolysaccharide, KA kanaic acid; + present, ≈ present without change in mentioned conditions, ↑ present with increase in mentioned conditions, ↓ present with decrease in mentioned conditions,aIncrease in soluble CX3CL1 and decrease in membrane-bound CX3CL1 CC Chemokines CCL2 CCL2 is currently the most extensively described neuronal chemokine. The majority of reports describing neuronal CCL2 expression are focused on pathological conditions. An induction of neuronal CCL2 expression was described upon ischemia [27–29], after axonal injury [30–34] and in motoneurons of patients with amyotrophic lateral sclerosis (ALS), and in mouse models for ALS [35, 36]. Interestingly, neuronal CCL2 expression in response to ischemia was detectable within 2 h, whereas CCL2 expression in astrocytes was detected only after 2 days [27]. Although most reports show induction of neuronal CCL2 expression under pathological conditions, a recent study has shown constitutive CCL2 expression in neurons throughout the rat brain [37]. This study demonstrated that, depending on the brain region, up to 100% of the neurons were positive for CCL2 [37]. CCL2 was mainly detected in neuronal cell bodies and costaining-depicted colocalization with various neurotransmitters and neuropeptides, corroborating a population-specific expression of CCL2 [37]. Constitutive neuronal CCL2 expression was also shown in a human neuronal cell line [38] and during human CNS development [39]. CCL3, CCL4, and CCL5 At present, there is only one study describing neuronal CCL3 expression in situ, depicting protein expression in adult human brain [40]. Further, expression of CCL3, CCL4, and CCL5 was described in cultured forebrain neurons derived from human first trimester embryos. These chemokines showed increased expression after exposure to immunological stimuli [41]. CCL3 and CCL4 expression were induced in mouse cerebellar granule neurons after infection with Toxoplasma gondii [42], as was CCL5 expression after viral infections [43, 44]. CCL21 In a middle cerebral artery occlusion (MCAO) mouse model of brain ischemia, cortical neurons rapidly expressed CCL21 in the penumbra of the ischemic core. Because control brain tissue did not express CCL21, CCL21 was assumed to be specifically expressed in endangered neurons [45]. In accordance with the in vivo findings, CCL21 expression was induced in cortical neurons in vitro within 2 h after excitotoxicity [45, 46]. The CCL21 expression in endangered neurons was rather surprising, as CCL21 is well known for its constitutive expression in secondary lymphoid organs, controlling the homing of mature dendritic cells and naïve T cells [47] and is, therefore, generally considered a homeostatic chemokine linked to the development and maintenance of secondary lymphoid organs [48]. The rapid CCL21 expression in endangered neurons after injury indicates a brain specific role of CCL21. This assumption is corroborated by findings in transgenic mice in which CCL21 was expressed ectopically in various tissues. CCL21 expression in the brain induced a massive brain inflammation that killed the animals within 3 days after the expression onset [49], whereas CCL21 expression in the skin induced the formation of secondary lymphoid structures [50]. Other CC Chemokines A single study has demonstrated a constitutive and inducible expression of CCL20 in rat cerebellar granule neurons in vitro, which was suggested to play a role in neuronal apoptosis [51]. Expression of other CC chemokines has not yet been observed in neurons. CXC Chemokines CXCL10 CXCL10 expression was first described in cortical neurons in rat in response to MCAO-induced brain ischemia [52]. Remarkably, neuronal CXCL10 expression was transient and appeared rapidly after stroke (within 3–12 h), whereas CXCL10 expression in astrocytes was detectable later and persisted up to 15 days after MCAO [52]. Correspondingly, neurons also showed a rapid CXCL10 expression after entorhinal cortex lesion [53]. Further, neuronal CXCL10 expression and release was induced after viral infection in vitro and in vivo [43, 54]. CXCL12 The CXCL12 gene contains three splice variants, termed stromal cell-derived factor-1 (SDF-1) α, β, and γ. SDF-1γ was cloned from rat brain and showed constitutive neuronal mRNA expression with almost no change in level after peripheral nerve injury [55]. In addition, SDF-1α showed neuronal mRNA expression with almost no change in level after brain ischemia or intracerebral LPS injection [56]. In contrast, SDF-1 β mRNA expression was not detected in neurons [56]. As little is known about the role of SDF splice variants, and most studies did not specify the splice variants, CXCL12 is used for all SDF splice variants henceforth. Like CCL2, but in contrast to most of the other neuronal chemokines, CXCL12 is expressed constitutively in specific neuronal populations. Neuronal CXCL12 expression in vitro was observed in cultured cortical, hippocampal, and cerebellar neurons from human, rat, and mouse [57–60]. Neuronal CXCL12 expression in vivo was studied in detail in the adult rat brain, showing CXCL12 mRNA and protein expression in cholinergic, dopaminergic, and vasopressin containing neurons throughout the brain [61, 62]. Other CXC Chemokines Studies describing the expression of other CXC chemokines in neurons are limited. Most notably, in vitro neuronal mRNA expression of CXCL1 [51], CXCL8 [41], CXCL9, and CXCL11 [43] has been illustrated. Expression of other CXC chemokines has not yet been described in neurons. CX3CL1 CX3CL1 was the first chemokine shown to be expressed in neurons [63–66]. Because microglia were shown to express the corresponding receptor CX3CR1, a role of CX3CL1–CX3CR1 signaling in neuron–microglia interaction was suggested [63–65]. CX3CL1 is constitutively expressed in human, macaque, rat, and mouse neurons in vitro and in vivo, with high expression in cerebral cortex, hippocampus, caudate putamen, thalamus, and olfactory bulb [63, 65, 66, 68, 69]. CX3CL1 appears to be the only chemokine with a higher expression level in brain than in peripheral organs [70]. It is membrane bound and can be cleaved from the cell surface by proteases of the A Disintegrin and Metalloprotease (ADAM) family [71, 72]. The neuronal CX3CL1 mRNA expression remained relatively stable in response to both neuron-damaging stimuli in vitro [73–77] and during neuroinflammation in vivo [66], whereas in vitro neurons released CX3CL1 protein after glutamate-induced damage [73, 74, 78]. Furthermore, CX3CL1 concentrations higher than 300 pg/mg were described in aqueous extracts of the brain [79], indicating that CX3CL1 can be cleaved from the neuronal membrane and released into the extracellular environment. It is yet unknown which ADAM protease cleaves CX3CL1 in neurons and whether CX3CL1 protein expression changes during in vivo neuroinflammation or degeneration. Potential Roles of Neuronal Chemokines in Neuron–Astrocyte, Neuron–Microglia, and Neuron–Neuron Interaction Astrocytes, microglia, and neurons have been shown to express chemokine receptors in vitro under physiological and pathological conditions and in vivo. These would include CCR2 for CCL2, CXCR3 for CCL21 and CXCL10, CXCR4 for CXCL12, and CX3CR1 for CX3CL1. Studies describing the expression of these chemokine receptors on astrocytes, microglia, and neurons (see Table 2) and studies indicating a role for these chemokine–chemokine receptor pairs in CNS cell interaction are discussed in the following sections on neuron–astrocyte, neuron–microglia, and neuron–neuron interaction. Table 2Chemokine receptor expression in astrocytes, microglia, and neuronsChemokineReceptorCell typeSpeciesConditionRNAProteinReferencesCCL2CCR2astrocytehBrain, MS, HIV↑[96, 97, 173]Monoculture↑↑[83, 99, 174–176]macMonoculture↑[175]rBrain, EAE, LPS injection↑[177, 178]microgliahBrain, MS, HIV↑[96, 173, 179]Monoculture↓+[173, 180]Glia culture+[99]rBrain, tumor, LPS injection, NMDA injection↑[177, 181, 182]Monoculture↑[88]mSpinal cord, peripheral nerve injury↑[183]neuronhBrain, HIV+[184]Monoculture++[38]Cell line++[38]rBrain and spinal cord++[161, 177, 185]Monoculture++[161, 164]CXCL10/CCL21CXCR3astrocytehBrain, MS, HIV↑[95, 97, 179, 186]Astrocyte culture↑↑[84, 98, 175]Mixed glial culture+[95]macMonoculture+[175]mMonoculture+[84]microgliahMonoculture++[7, 84, 98, 114]Cell line↑↑[98]rCell line↑/↓[187]mBrain, various infectious agents, axotomy≈/↑≈/↑[125]Monoculture+↓[45, 84]Cell line↑/↓[188]neuronhBrain, AD+≈[95, 179, 189]Monoculture++[38]Cell line++[38]macBrain, HIV+[54]rMonoculture+[163]CXCL12CXCR4astrocytehBrain, HIV↑[173, 190, 191]Monoculture↑↑[85, 90, 98, 99, 175, 192–196]macMonoculture↑↑[85, 175]rBrain+[197]Monoculture↑/↓↑/↓[57, 58, 102, 198]mMonoculture↑/↓↑/↓[86, 101, 104, 199, 200]microgliahBrain and spinal cord, HIV++[173, 179, 190, 191, 201, 202]Monoculture+↓[98, 99, 191, 202–205]Cell line≈[98]babMonoculture↑[206]rBrain+[197]Monoculture≈/↑+[58, 207, 198]mCell line+[86]neuronhBrain, HIV+≈/↑[179, 184, 191, 197, 202, 204, 208]Monoculture++[38, 85]Mixed brain culture++[202]Cell line++[38, 191, 209]macBrain+[210, 211]Monoculture+[85]rBrain++[57]Monoculture++[64, 158]CX3CL1CX3CR1astrocytehBrain, MS≈[113]Monoculture↑+[113, 175]macMonoculture↑[175]rMonoculture≈/↑↑[76, 198, 212]mMonoculture↓↓[77, 81, 200]microgliahBrain, MS≈[113]Brain, HIV↑[170]Monoculture++[75, 113]rBrain, ischemia, prion disease, cranial nerve injury, EAE↑↑[63, 67, 68, 213]Brain, LPS injection, KA injection≈[67]Spinal cord, peripheral nerve injury↑↑[171, 172]Monoculture↑/↓↑[63, 65, 198, 212, 214]Cell line↓[187]Brain, LPS injection, KA injection≈[67]mMonoculture≈≈[77]neuronhMonoculture++[75]Cell line≈↑[75]rBrain, LPS injection, KA injection≈[67]Monoculture++[64, 139]mBrain, prion disease↓[67]Brain, LPS injection, KA injection≈[67]h human, mac macaque, bab baboon, r rat, m mouse, MS multiple sclerosis, HIV human immunodeficiency virus, EAE experimental autoimmune encephalomyelitis, LPS lipopolysaccharide, NMDAN-methyl-d-aspartatic acid, AD Alzheimer’s disease, KA kainic acid; + present, ≈ present without change in mentioned conditions, ↑ present with increase in mentioned conditions, ↓ present with decrease in mentioned conditions Neuron–Astrocyte Interaction Astrocytes comprise the largest group of CNS-residing cells and are not only essential in development, homeostasis, maintenance of the blood–brain barrier, and regulation of central blood flow but are also involved in the immune defense of the CNS. Furthermore, astrocytes are considered to be involved in neuronal information processing [80]. It is becoming clear that astrocytes play an active role in the intricate chemokine network of the CNS. Not only has it been shown that astrocytes express a wide variety of constitutive and inducible chemokines in vivo and in vitro, there is also extensive evidence that they express a repertoire of chemokine receptors under physiological and pathological conditions (see reviews [81, 82]). Neuronal Chemokines Induce Calcium Transients in Astrocytes The activation of intracellular calcium transients is a hallmark in chemokine receptor signaling, a mechanism that also holds true for astrocytes [57, 76, 83–86]. Activation of GPCRs, including chemokine receptors, results in a rapid release of calcium from the endoplasmatic reticulum (ER) through the activation of inositol-1,4,5-triphosphate receptors on the ER membrane. One of the first chemokines described to induce calcium transients in astrocytes is CXCL12 [57, 85–87]. CXCL12 concentrations ranging from 0.1 to 100 ng/ml [85, 86] or 10–100 nM [57, 87] induced calcium fluxes in in vitro human, rat, and mouse astrocytes. In all cases, CXCL12-induced calcium mobilization was PTX-sensitive, indicating that this process is Gαi-protein mediated. Similar results were found for CXCL10 [84], CCL2 [83, 88], and CX3CL1 [76]. In astrocytes, intracellular calcium transients not only function as a second messenger in multiple intracellular signaling pathways but are also implicated in astrocyte–astrocyte signal propagation, astrocyte–neuron synaptic transmission, and neurotransmitter release (see reviews [80, 89]). Recent findings corroborate that chemokines could also be involved in astrocyte-mediated neurotransmitter release. CXCL12 induced calcium-dependent release of glutamate from astrocytes in human and rat astrocyte cultures and rat hippocampal slice cultures [90]. Moreover, reports that investigated the effects of CXCL12 on the electrophysiological properties of neurons in brain slice cultures suggest that CXCL12-induced effects on neurons at least partly depend on astrocytic glutamate release [91–93]. Whether this astrocytic glutamate release was induced by CXCR4 activation or via other pathways was not investigated. Neuronal Chemokines Induce Astrocyte Proliferation and Migration in Vitro: Implications for Astrogliosis? Astrocytes respond to CNS injury or neuroinflammation by enhanced GFAP expression, proliferation, and possibly, migration, a process known as astrogliosis (see review [94]). In these reactive astrocytes, enhanced expression of chemokine receptors has been described under various pathological conditions, such as multiple sclerosis (MS), human immunodeficiency virus (HIV) infection, ischemia, and neoplasm [95–97]. Under these conditions, CXCR3 was mainly found in reactive astrocytes in the proximity of the lesion sites, suggesting that induction of CXCR3 expression in astrocytes is limited to damaged areas of the brain [95–97]. A comparable induction of CCR2 expression was found in reactive astrocytes in MS patients [97]. Interestingly, both CCL2 and CXCL10 are implicated in astrocyte proliferation in vitro [98, 99]. In addition, CXCL12 has been shown to induce astrocyte proliferation in vitro, a process that is dependent on activation of extracellular signal-regulated kinases ERK1 and ERK2 [87, 100–102]. Both CXCL12-induced astrocyte proliferation and ERK1/2 activation was inhibited by PTX and wortmannin, suggesting that they are dependent on upstream activation of Gαi proteins and PI3-K [87]. As chemokines are primarily known for their capacity to induce cell migration, migration assays have been used to determine chemokine receptor functionality in astrocytes [103]. Accordingly, astrocyte migration was demonstrated in vitro in response to CCL2, CXCL10, and CXCL12 [83, 84, 86, 103, 104]. Thus, reactive astrocytes express various chemokine receptors and activation of these receptors in vitro induces proliferation and migration, cellular reactions that are generally involved in astrogliosis. Therefore, it is tempting to speculate that chemokines are involved in the regulation of astrogliosis upon CNS injury or neuroinflammation. Whether neuronal chemokines are indeed responsible for either proliferation or migration of astrocytes in vivo is yet unknown. Neuron–Microglia Interaction Microglia in the healthy CNS are ramified cells that continually survey their environment by moving their processes. Upon injury, they quickly protrude their processes toward the damaged site and subsequently transform into amoeboid cells, reflecting a fast activation [105, 106]. Activated microglia form a first line of defense in CNS injury through their capacity to migrate, proliferate, secrete inflammatory and neurotrophic factors, phagocytose-damaged cells and debris, and present antigens [82, 107]. Although activated microglia were initially considered to be detrimental in CNS injury, recent findings indicate a prominent neuroprotective activity as well, suggesting a balance between neurotoxic and neuroprotective microglia activity (see for recent review, [108]). Therefore, it is of particular interest to gain insight into the process of microglia activation. Until now, it is largely unknown which environmental signals mediate microglia surveillance and activation. Almost 10 years ago, chemokines were indicated as promising candidates for neuron–microglia signaling [63–65]. Because then, various studies have described constitutive chemokine expression in neurons and rapid changes in expression levels upon injury. Parallel to this, corresponding chemokine receptors were described in resting and/or activated microglia. In addition, there is increasing evidence that neurons play an important role in microglia activity, which is at least partly mediated by chemokines. Microglia Activity Upon Neuronal Damage Upon CNS injury, activated microglia retract their protrusions, transforming into amoeboid cells with migratory and/or proliferative capacities [109–111]. It is known that damaged neurons are accompanied by prominent activated microglia within hours after injury, suggesting that neurons emit signals that attract microglia [111]. Several findings support the notion that these signals are primarily chemokines. Microglia express various chemokine receptors, and cell migration is induced upon exposure to chemokines in vitro [7, 45, 84, 112–115]. Moreover, damaged neurons in culture express and release chemokines like CX3CL1 [73, 74, 78], CCL21 [45, 46], and CXCL10 [43, 54], all of which are able to induce microglia migration [7, 45, 46, 73, 76, 84, 113, 115]. In accordance with this, inhibition of chemokine function diminished microglia migration in response to supernatants from damaged neurons [73]. Thus, in vitro results suggest a role of neuronal chemokines in neuron–microglia activation. The issue of chemokine-mediated neuron–microglia activation has been further investigated using genetically modified mice. Mice deficient for either CX3CR1 [116] and CX3CL1 [117] have been studied in various CNS injury and neuroinflammation models. Although CX3CR1 deficiency did not influence microglia activity in response to facial nerve lesion [116], CX3CR1 deficiency was uniformly associated with higher levels of microglia activity in LPS-induced neuroinflammation, 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridin-induced neurotoxicity, and in the SOD1G93A-model of motoneuronal death in the spinal cord [79]. Interestingly, enhanced microglia activity in the last three models was accompanied by increased neuronal death, indicating that, in wild-type mice, neurotoxic microglia activity is inhibited by CX3CL1–CX3CR1 signaling [79]. These findings are corroborated by several in vitro findings. Exposure of a neuron–microglia coculture to CX3CL1 reduced inflammation-related neuronal death, accompanied by suppressed nitric oxide and proinflammatory cytokine production [118]. In conjunction with these findings, inhibition of endogenous CX3CL1 increased neuronal cell death in cocultures [77]. Moreover, in vitro exposure to CX3CL1 supported microglia survival under basal culture conditions and reduced Fas-ligand induced apoptosis considerably [119]. Thus, exposure of microglia to CX3CL1 reduced microglia toxicity and protected microglia from apoptosis under inflammatory conditions. In contrast to these results, CX3CL1 deficiency reduced the infarct volume and mortality after transient focal cerebral ischemia [117]. However, microglia activity in CX3CL1-deficient and wild-type mice was not compared in this study, making it difficult to determine whether disturbed neuron–microglia signaling was responsible for the differences [117]. Increased microglia toxicity by CXCR3 and its ligands CXCL10 and/or CCL21 is suggested by findings derived from the entorhinal cortex lesion (ECL) model, in which CXCR3 deficiency was associated with reduced microglia activity and reduced loss of secondary neurons in the hippocampal formation [53]. An interesting aspect of chemokines in neuron–microglia signaling is acknowledged in the ECL model. In this paper, microglia activity is specifically found within the midmolecular layer of the dentate gyrus, which is the projection site of the transected neurons (see for review, [120]). The microglia activity at a site distant from the primary lesion indicates transport of the chemokine signal. Recent data reinforced this notion, showing that CCL2 that was induced in dorsal root ganglion (DRG) neurons after peripheral nerve injury was transported to afferent terminals in the spinal cord [34]. Moreover, in vitro neuronal CCL21 was sorted into vesicles, transported into neuronal processes, and even reached presynaptic terminals [46]. The finding of CCL21 protein in neuronal vesicles is a strong indication that neuronal chemokines may be the signals responsible for microglia activity at sites distant from the primary lesion, a phenomenon that has been observed also in humans [121, 122]. A role of CCL2 in microglia activity after neuronal death is suggested by a delayed microglia activity in the thalamus of CCL2-deficient mice in response to cortical injury [123]. The delayed microglia activity was accompanied by a transient improvement of neuron survival in the thalamus, which may indicate that CCL2 is involved in neurotoxic microglia activity [123]. However, it is not yet clear if this effect is due to disturbed neuron–microglia interaction. Damaged neurons are capable to express CCL2, as was found after axotomy in sympathetic ganglia [31] and facial nerve lesion [30], but cortical injury predominantly induced CCL2 mRNA expression in astrocytes [124]. Whether interference with CCL2 signaling would affect microglia activity in the first two models has not yet been investigated. It is clear that the assumption that neuronal chemokines are involved in neuron–microglia signaling is no longer based on the finding that damaged neurons rapidly alter chemokine expression patterns and that microglia express the corresponding receptors [63–65]. Studies now show that microglia activation is reduced in mice with genetically disturbed chemokine function, indicating an important role of chemokines in microglia activation. Recent data even suggest that neurosupportive and neurotoxic microglia activity are associated with chemokine receptor expression [125]. The ultimate effects of neuronal chemokines are likely dependent on injury type, brain region, and disruption of the blood–brain barrier [53, 79, 116]. Whereas the exact role of neuronal chemokines in neuron–microglia signaling remains obscure, their importance in regulating damage responses is becoming apparent. Neuron–Neuron Interaction Various reports indicate that chemokines influence neuronal development, differentiation [126, 127], survival [128–130], electrophysiological properties [93, 131, 132], and synaptic transmission [26, 92, 133]. Because neurons can express numerous chemokines, autocrine and paracrine contributions of neuronal chemokines are likely. Neuroprotection Neuronal cell death, the ultimate consequence of all neuroinflammatory conditions, has been studied extensively in vitro. However, there are relatively few in vitro models that can be extrapolated to pathological conditions leading to neuronal death in vivo. One of the most prominent models is glutamate or NMDA-induced neurotoxicity, a model for excitotoxicity, which is most likely involved in various neurodegenerative diseases. β-amyloid-induced neuronal death serves as a model that may explain the loss of neurons in Alzheimer’s disease, whereas exposure of neuronal cultures to HIV proteins gp120 or HIVtat are aimed to elucidate neuronal death in HIV-dependent neurodegeneration [134–137]. Several reports indicate that neuronal chemokines may protect neurons from these toxic conditions. In vitro, CX3CL1 is known to protect neurons from glutamate-induced toxicity [78, 138], gp120-induced neuronal death [64, 139], and death induced by deprivation of trophic support [140]. Similar to CX3CL1, CCL2 exposure is shown to protect neurons from glutamate- and HIV-tat-induced neurotoxicity [141, 142]. However, CCL2 exposure was not protective in β-amyloid-dependent neuronal death [141]. As exposure of neuronal cells to chemokines is known to activate the putatively neuroprotective MEK/ERK and PI3-K/Akt signaling pathways [78, 138–140], it is reasonable to argue that chemokine-dependent protection is mediated by these pathways. Indeed, inhibition of both pathways completely abolished the neuroprotective effects of CX3CL1 in gp120- and glutamate-induced neurotoxicity in hippocampal neurons [64, 78]. Interestingly, in case of glutamate-dependent neurotoxicity, the involvement of MEK/ERK and PI3-K/Akt signaling pathways was only evident when CX3CL1 was applied together with glutamate [78]. CX3CL1 exposure was shown to be protective even when the chemokine was applied up to 8 h after the glutamate stimulus. However, an inhibition of MEK/ERK and PI3-K/Akt pathways did not affect the protective activity of delayed CX3CL1 exposure, indicating that CX3CL1 may activate additional pathways in neurons that lead to neuroprotection [78]. The effect of CXCL12 on neuronal death is contradictory. Although several reports indicate that CXCL12 exposure may protect neurons from gp120-induced neuronal death, most papers describe a toxic effect of CXCL12 in neuronal cultures (see below) [64, 143]. Neurotoxicity Approximately 10% of HIV-infected patients develop HIV-1 associated dementia (HAD). It has been shown that the viral protein gp120 itself is neurotoxic [144], indicating that the neuronal loss in HAD is not only due to inflammation occurring after the virus enters the brain but also because of direct toxic effects of viral proteins (see for recent review, [145]). It was shown in 1998 that the neurotoxic effect of gp120 is mediated via the chemokine receptor CXCR4 [146], findings that have been corroborated in subsequent years by various groups [64, 102, 147, 148]. The viral protein gp120 binds and activates CXCR4, the main coreceptor utilized by HIV-1 to infect T cells. CXCR4 has subsequently become the best investigated chemokine receptor with respect to neurotoxicity, and its involvement in neurotoxic signaling has been demonstrated by use of the specific CXCR4 antagonist AMD31000 [102, 147]. The HIV-derived protein gp120 shows agonist activity on CXCR4, and therefore, it is not surprising that its ligand CXCL12 has also been described to be neurotoxic [102, 143, 146, 148, 149]. Currently, there is little information on intracellular signaling pathways that are activated by CXCR4 and subsequently lead to neuronal death. One recent report indicates the involvement of Src activity in CXCL12-induced apoptosis in a neuronal cell line, whereas gp120-induced apoptosis in these cells was independent of Src activity [149]. Interestingly, CXCL12 and gp120 had different effects on ERK activation in neurons and astrocytes [102], indicating that CXCR4 signaling exerts both ligand and cell-type specific effects. The effect of CXCL12 is further complicated by matrix metalloproteinase-2, which was shown to remove the first four amino acids of CXCL12, resulting in a truncated form of CXCL12 [150]. This truncated form was found to be highly neurotoxic compared to the full-length CXCL12, which remarkably was not mediated by CXCR4 but by a yet unknown PTX-sensitive receptor [150]. Because MMP-2 has also been described in HIV-infected patients, it is reasonable to assume that truncated CXCL12 may be a neurotoxic player in HAD [151]. CXCL12 is not the only neuronal chemokine that exerts neurotoxic effects. Neuronal cell lines and primary human neurons respond to high concentrations of CXCL10 with intracellular calcium transients, caspase activity, and apoptosis [54, 152, 153]. The direct involvement of CXCR3 was demonstrated by the use of an antibody that prevents the activation of CXCR3 and subsequently inhibited CXCL10-dependent neurotoxicity [153]. Chemokinergic Effects on Synaptic Transmission Recent data show that CXCR4 activation by either gp120 or CXCL12 significantly enhanced giant depolarizing potentials (GDP) in rat neonatal hippocampus [154]. These GDPs only occur in the developing hippocampus and are involved in growth and synapse formation. These data may explain why HIV infections have a greater impact in the developing brain than in adults [152] and show that neuronal chemokines may change the electrophysiological properties of neurons, thereby corroborating earlier findings [131, 132, 155, 156]. The electrophysiological properties of neuronal chemokine receptors have predominantly been studied in cultured primary neurons or neuronal cell lines and brain slice cultures [157]. Remarkably, in cultures of DRG, cerebellar granule or Purkinje neurons, and hippocampal pyramidal cells, chemokines induced changes in the electrophysiological properties of only 10–20% of the neurons [64, 140, 158, 159]. In addition, several effects of chemokines in neurons were not sensitive to PTX, in contrast to hematopoietic cells, suggesting that chemokine receptors in neurons, although generally accepted, are not solely coupled to Gαi proteins [128, 140, 149]. Whether these chemokinergic PTX-insensitive effects are mediated by neuronal Gz-subunits is yet unclear [160]. Cultured cerebellar and DRG neurons respond to various chemokines with intracellular calcium transients [140]. In DRG neurons, exposure to CX3CL1 and CXCL12 also increased their excitability [158]. Although chemokinergic effects of CX3CL1, CXCL12, CCL2, and CXCL10 in neurons have been reported to modulate the frequency of both spontaneous and activity-dependent neuronal firing, a direct effect on the induction of action potentials has not yet been described [159, 161–164]. Similar to the effects on isolated neurons, CX3CL1, CXCL12, and CXCL10 also affected neuronal signaling in brain slice cultures [26, 78, 91–93, 133, 134, 165, 166]. However, the presence of glia cells (astrocytes, microglia, and oligodendrocytes) in these slice cultures makes it difficult to determine whether the electrophysiological effects of chemokines are mediated by chemokine receptors on neurons and/or on glia cells, as glia cells may also induce electrophysiological changes in neurons [26, 133, 156]. Whether the effects of CXCL12 in brain slice cultures are mediated via chemokine receptors on neurons and/or on glia cells may depend on the concentration, as concentrations up to 1 nM caused a direct decrease in peak and discharge frequency of evoked action potentials in neurons and concentrations higher than 10 nM activated an indirect GABA-mediated hyperpolarization of neurons [92]. Future Directions As discussed here, neuronal chemokines appear to be versatile messengers in CNS cell interaction. However, several important issues need to be addressed in future studies. To begin with, neuronal CX3CR1 expression in vivo remains controversial. Immunohistochemical analysis revealed CX3CR1 positive neurons in mouse brain sections with little changes under pathological conditions [67], whereas neuronal CX3CR1 expression was never described in studies using genetically modified mice in which CX3CR1-expressing cells are also positive for EGFP [116]. Different microscopic techniques and models of neurodegeneration have been explored in these mice, demonstrating only CX3CR1 expression in resting and activated microglia [79, 105, 106, 116]. An explanation may be that neuronal CX3CR1 expression is at such a low level that detection is difficult to achieve with microscopic techniques. This may also be the case for CXCR3 expression in microglia. Although CXCR3 expression has yet not been described in microglia in vivo, functional evidence derived from CXCR3-deficient mice strongly indicates that microglia do express CXCR3 in vivo [53, 115]. Therefore, it seems appropriate that future experiments concerning the expression of chemokine receptors in CNS cells in vivo also include functional analysis. Another issue that needs to be addressed in more detail regards cellular localization. Neurons are highly polarized cells, as their function is largely dependent on their morphology and contacts with other cells (e.g. synapses with other neurons). Although neuronal signaling molecules, such as neurotransmitters, neuropeptides, and neurotrophins, are generally found at specific sites, most reports describing neuronal chemokine expression did not address this issue. Interestingly, a few recent publications do suggest a localized expression of chemokines comparable to other neuronal signaling molecules. Our group demonstrated that neuronal CCL21 is transported in vesicles, reaching presynaptic terminals in cortical neurons in vitro [46]. In subsequent studies, these vesicles appeared to be of the large-dense core type, in which other neuronal peptides are also found (e.g. neurotrophins; Stanulovic et al., manuscript in preparation). Moreover, it has been described for several neuronal populations in vivo that CCL2 and CXCL12 colocalize with other neurotransmitters and neuropeptides in synaptic regions [37, 62]. Like neuronal chemokine expression, a site-specific expression of chemokine receptors may exist, as is suggested by CXCR4 redistribution in the axonal and dendritic compartment of hippocampal neurons after prolonged CXCL12 exposure [167]. Because a localized expression of chemokines and their receptors may have a consequence for their role in cell interaction, future studies on neuronal chemokine expression may address this issue. At last, as all reports indicating that chemokine exposure alters the excitability of neurons used exogenous chemokines, it is yet unknown whether chemokines released from neurons have similar effects. Conclusion Knowledge on the spatial and temporal expression of neuronal chemokines and their regulation under physiological and pathological conditions is increasing rapidly. As CNS cells can express the corresponding chemokine receptors, contribution of these neuronal chemokines to CNS cell interaction is conceivable. This assumption is corroborated by various in vitro and in vivo studies. For example, the following effects of neuronal chemokines were observed in vitro: in astrocytes proliferation and migration, in microglia migration and neurotoxic and neuroprotective activity and in neurons electrophysiological changes, neurotoxicity, and neuroprotection. Further, the synaptic transmission between neurons seems to be influenced by the action of neuronal chemokines on neurons and/or glia cells. In vivo studies support the important role of chemokines in migration and neurotoxic and neuroprotective activity of microglia upon CNS injury and neuroinflammation. Further exploration of the roles of neuronal chemokines in CNS cell interaction is needful, as insight into the role of neuronal chemokines in CNS injury and neuroinflammation may contribute to the development of therapeutic strategies.

Skeletal_Radiol-4-1-2424184

Bone marrow edema in the knee in osteoarthrosis and association with total knee arthroplasty within a three-year follow-up

Objective The purpose of this study was to determine if a correlation exists between magnetic resonance imaging (MRI) findings of bone marrow edema (BME) in osteoarthrosis (OA) of the knee joint and need for total knee arthroplasty (TKA) within a follow-up period of 3 years. Introduction Osteoarthrosis (OA) is the most common cause of disability among the elderly population [1, 2]. OA is a condition where articular cartilage cannot maintain homeostasis in response to the forces acting on it. When homeostasis breaks down, there is a wide range of possible biologic responses based on the genetic background of the individual [3]. These responses may be anabolic or catabolic and involve cartilage, bone, and synovium. The exact pathogenesis of OA and the cause of the pain produced are unclear. OA is classically associated with degradation of the hyaline cartilage around the joint. However, cartilage contains no pain fibers. Other proposed sources of pain production in OA include periarticular structures and the underlying bone [3]. Several recent studies have examined bone marrow edema (BME) lesions and their association with OA [1, 4–10]. BME has been defined as an area of low signal intensity on T1-weighted images, associated with intermediate or high signal intensity findings on T2-weighted images. The literature suggests that there are two distinct types of BME. The first type occurs traumatically with injury to a joint and tends to resolve over a period of weeks to months [11–14]. This type of BME does not appear to have any long-term implications [15, 16]. The second type of BME occurs without trauma and may be associated with rapidly progressing osteoarthrosis. Zanetti et al. [9] suggest a possible link between atraumatic BME and increased pain in OA. Felson et al. [7] have suggested a possible link between BME lesions on magnetic resonance imaging (MRI) and progression of OA findings on MRI over time. Hunter et al. found that bone marrow lesions that enlarge with time are associated with more cartilage loss, compared with bone marrow lesions that stay the same size over time [17]. Despite the recent literature on BME, the implication of these findings on MRI and how they relate to clinical outcomes is unknown. Most studies have used the subjective complaint of pain as the only clinical correlate. There has been no measure of an objective clinical outcome related to the finding of BME. The purpose of this study was to determine if a correlation exists between MRI findings of BME of the knee joint and the incidence of total knee arthroplasty (TKA) within a follow-up period of 3 years. Materials and methods The entire database of knee MRI studies from 1995–1997 (over 4,000 studies) conducted at a large urban hospital system was used to select individuals with knee osteoarthrosis (OA). The study was reviewed and approved by the hospital’s institutional review board. An initial random search was conducted within this database to identify two distinct groups of patients. The first group had MRI reports containing the phrase “knee osteoarthrosis/osteoarthritis,” and the second group consisted of those containing the phrases “knee osteoarthrosis/osteoarthritis” and “bone edema.” The search for “osteoarthrosis/osteoarthritis” yielded 235 cases for review. The search for “osteoarthrosis/osteoarthritis” and “bone edema” yielded 146 cases for review. After these two initial groups were identified, a chart review was conducted on all 381 patients to identify and include only those patients who had at least a 3-year clinical follow-up appointment from the time of the MRI study. Subjects were also excluded if the reason for referral to MRI was post-traumatic or post-surgical. An initial review of the images was completed by an experienced musculoskeletal radiologist to exclude any subjects with evidence of recent post-surgical or post-traumatic changes not mentioned in the report. After this chart review and initial radiology review process, there were 38 patients in the OA-only group and 35 in the OA with BME group. The same musculoskeletal radiologist who provided the initial review was blinded to the original interpretation of the MRI studies and the patient outcome, and reviewed all 73 studies. The radiologist was aware of the study hypothesis at the time of interpretation. The radiologist assessed each knee for the presence or absence of BME and assigned each subject into either the BME group or the no-BME group for further evaluation. After this review of the images, there were a total of 25 patients with OA only and 48 with OA and BME. The OA-only group consisted of four males and 21 females, with an age range of 28–75 years, and an average age of 49.3 years. Conventional radiographs consisting of anteroposterior, lateral and sunrise views were available for review with 13 of the 25 patients, 52% of the group. The OA and BME group consisted of 23 men and 25 women, with an age range of 35–82 years and an average age of 53.5 years. Conventional radiographs, again consisting of anteroposterior, lateral, and sunrise views, were available for review with 33 of the 48 patients, 68.75% of the group. The conventional radiographs, when available, were reviewed by the musculoskeletal radiologist. Radiographic evidence of OA on the radiographs was assessed using the Kellgren–Lawrence scale [18] as follows: NoneDoubtfulMinimalModerateSevere According to this scale, evidence of OA includes (1) formation of osteophytes on joint margins or on the tibial spine, (2) periarticular ossicles, (3) narrowing of joint cartilage associated with sclerosis of the subchondral bone, (4) small pseudocystic areas with sclerotic walls situated usually in the subchondral bone, and (5) altered shape of the bone ends [18]. All patients had dedicated knee MR scans either on a GE 1.5-T magnet (23 in the OA-only group or 92% and 42 subjects in the OA with edema group or 87.5%) or a 1.0-T magnet (8% of the OA-only group, two subjects, and 12.5% of the OA with edema group, six patients). MRI films were reviewed and interpreted by the musculoskeletal radiologist for all patients in the study. For each patient, the sagittal proton density sequences (TR range, 2,666–4,100; TE range, 14–33; 44 conventional spin echo studies [ET = 0], 29 fast spin echo studies [ET = 4–8], 3- to 5-mm section thickness, 16 × 16–18 × 18 field of view [FOV], 256 × 128–256 image matrix, bandwidth range 16–32), coronal proton density sequences (TR range, 2,316–3,950; TE range, 14–33; 44 conventional spin echo studies [ET = 0]; 29 fast spin echo studies[ET = 4–8]; 3-mm section thickness; 16 × 16–18 × 18 FOV; 256 × 128–256 image matrix; bandwidth range, 16–32), and coronal fat suppressed T2-weighted images (TR range, 2,400–3,200; TE range, 56–80; 3- to 5-mm section thickness; 16 × 16–18 × 18 FOV; 256 × 128–256 image matrix; bandwidth range, 15.6–16) were reviewed. Axial fat suppressed T2-weighted images (TR range, 3,000–3,600; TE range, 56–76; 1.5- to 10-mm section thickness; 16 × 16–20 × 20 FOV; 256 × 128–256; bandwidth range, 15.6–16) were available and reviewed for 69 of the 73 knees. For the four subjects who had no axial T2 imaging available, BME in the patellofemoral joint was not assessed. The medial, lateral, and patello-femoral compartments of the knee joint were assessed. Evidence of OA on MRI was assessed using a modified Noyes arthroscopic classification [19] as follows: NormalInternal changes only1–49% loss of articular cartilage50–99% loss of articular cartilage100% loss of cartilage with subchondral cortex intact100% loss of cartilage with ulcerated subchondral cortex The most severe grade cartilage lesion within a compartment was utilized for the grading of that compartment. Also, the highest grade cartilage lesion on either side (femoral or tibial) of a compartment was used. The radiologist then assessed each knee MRI for the presence or absence of BME. BME was defined as an area of intermediate or high-signal intensity findings on T2-weighted images. All patients in the BME group were then further classified into three separate, groups based on patterns of BME. Type I was termed “focal” and contained a small area of edema adjacent to a cartilage defect (Figs. 1 and 2). Type II was termed “global” pattern with marrow edema occupying most or all of a femoral or tibial condyle (Figs. 3, 4, and 5), and type III was termed “cystic” and was associated with a subchondral geode with a small area of BME (Fig. 6). The worst area of edema in a compartment was used in the classification. The presence or absence of meniscal tear was also noted. Fig. 1Focal edema with a total knee joint replacement in a 51-year-old female patient. The radiograph demonstrated moderate loss of cartilage joint space medially and spurring of the tibial spines. Coronal fast fat suppressed T2-weighted image (TR 3000/TE 76/256 × 192) shows the loss of hyaline cartilage and a small focal area of edema in the medial aspect of the medial femoral condyle (arrow)Fig. 2Focal edema, no total knee joint replacement, in a 58-year-old male patient. Coronal fast fat suppressed T2-weighted image (TR 3000/TE 56/256 × 192) shows extrusion of the medial meniscus (curved arrow) and small adjacent areas of focal marrow edema (straight arrows). The medial meniscus has a degenerative horizontal cleavage tear approximating the tibial surface. Note also the moderate loss of hyaline cartilageFig. 3Global edema with a total knee joint replacement in a 65-year-old male patient. The radiograph demonstrated absence of joint space in the medial and lateral compartments, and irregularity of the subchondral bone plate of the lateral femoral condyle. Coronal fast fat-suppressed T2-weighted image (TR 3000/TE 72/256 × 192) through the knee shows extensive edema in the lateral femoral and tibial condyles, extending into the metaphyses (straight arrows). Note also the small extruded body of the degenerative and torn lateral meniscus (curved arrow)Fig. 4Global edema with a total knee joint replacement in a 72-year-old male patient. The radiograph demonstrated loss of cartilage joint space within the lateral compartment. Coronal fast fat-suppressed T2-weighted image (TR 3200/TE 72/256 × 192) shows extensive edema in the femoral and tibial condyles (arrows)Fig. 5Global edema, no total knee joint replacement, in a 58-year-old male patient. Coronal fast fat suppressed T2-weighted image (TR 3000/TE 76/256 × 256) through the knee shows extensive edema in the lateral tibial condyle and edema in some of the lateral femoral condyle (arrows). Note also the almost complete absence of hyaline cartilageFig. 6Cystic pattern with a total knee joint replacement in an 82-year-old female patient. The radiograph demonstrated subchondral lucency consistent with a geode and irregularity of the adjacent subchondral bone plate. Coronal fast fat-suppressed T2-weighted image (TR 3000/TE 56/256 × 192) shows complete loss of hyaline cartilage on the medial femoral condyle and adjacent moderate subchondral cystic areas consistent with geodes, and the adjacent bone marrow edema (arrows). Note also the extruded body of the medial meniscus containing a degenerative horizontal cleavage tear (curved arrow) Generalized estimating equations (GEE) were used to determine which factors were related to receiving a TKA. There were eight subjects who had bilateral knee MRIs. GEE takes into account the correlation within an individual who had both knees involved, whereas a general linear model assumes that all observations are independent and does not allow for multiple observations from one individual. A multivariable logistic regression model using generalized estimation was also used to adjust for the age of the patients and to determine if the results were still significant after accounting for the age differences. Results There were 65 subjects in this study, with a total of 73 knees. After the final classification by the radiologist, there were 48 knees in the group with BME. There were a total of 15 knees or 31.125% in this group that had a TKA. There were 34 knees with a focal pattern of BME, 12 with a global pattern, and two with a cystic pattern. Of those with a focal pattern, 20.5% [7] had a TKA (Fig. 1). For those with a global pattern, 58.3% [7] had a TKA (Figs. 3 and 4), and for those with a cystic pattern, 50% [1] had a TKA (Fig. 6). These results are detailed in Table 1. Table 1Patterns of edema and percentage of those who subsequently received TKAPattern of edemaTotal numberNumber with TKAPercentage with TKA (%)Focal34720.5Global12758.3Cyst2150All patterns of edema481531.2None2528 There were 25 knees in the group with no BME. There were a total of two knees or 8% of those without BME who had a TKA (Fig. 7; see Table 1). Fig. 7No bone marrow edema with a total joint replacement in a 66-year-old female patient. Coronal fast fat-suppressed T2-weighted image (TR 2500/TE 80/256 × 256) shows almost complete loss of hyaline cartilage in the medial compartment and an extruded, degenerative torn medial meniscus. Note the absence of marrow edema For statistical purposes, secondary to the small numbers of subjects in each subgroup, subjects were classified using the modified Noyes scale (0, 1, 2a, 2b, 3a, and 3b), and the data was grouped as cartilage damage less than 50% (0, 1, 2a) and cartilage damage greater than or equal to 50% (2b, 3a, 3b). For all three compartments of the knee, there were no statistical differences in rates of TKA between the groups that had less than 50% damage and the groups that had greater than 50% damage (see Table 2). Table 2Cartilage loss seen on MR, as graded by the modified Noyes scale, and its relationship to the occurrence of TKAVariableNo TKAYes TKAOR95% CIp ValueMedial cartilage loss <50%17 (30%)2 (12%)2.770.71–10.720.14Medial cartilage loss ≥50%39 (70%)15 (88%)Lateral cartilage loss <50%42 (75%)10 (59%)2.060.74–5.700.16Lateral cartilage loss ≥50%14 (25%)7 (41%)PF cartilage loss <50%15 (28%)3 (19%)1.680.45–6.290.44PF cartilage loss ≥50%38 (72%)13 (81%) For statistical analysis of the radiographs using the Kellgren–Lawrence scale, subjects who had scores of 0–2 (none, doubtful, or minimal) were compared with subjects who had a score of 3–4 (moderate, severe) to assess for statistical difference in the occurrence rate of TKA. Those subjects who had a Kellgren–Lawrence score of 3–4 were no more likely to have a TKA performed than those who had a score of 0–2 (see Table 3). Table 3Radiographic findings, as graded by the Kellgren–Lawrence scale, and their relationship to occurrence of TKAScoreNo knee replacement(N = 35)Knee replacement(N = 11)OR95% CIp Value0–28 (23%)1 (9%)1.00060.9996–1.00160.233–427 (77%)10 (91%) Table 4 shows the results of the subjects with BME, using the patterns of edema, compared to those without BME. Subjects who had BME of any pattern type were 5.48 times as likely to have a TKA when compared to subjects with no BME (p = 0.05). Upon further investigation, subjects with a global pattern of BME were 7.63 times as likely to have a TKA compared to subjects with focal, cyst, or no BME (p = 0.004). Subjects with global BME were 15.21 times as likely to get a TKA when compared to subjects without BME (p < 0.001). Table 4Comparison of bone marrow edema patterns and their relationship to occurrence of TKAVariableNo TKAYes TKAOR95% CIp ValueNo BME23 (41%)2 (12%)5.480.98–30.680.05BME of any pattern33 (59%)15 (88%)Global BME5 (9%)7 (41%)7.631.89–28.340.004All other patterns (focal, cyst, none)51 (91%)10 (59%)Global BME5 (18%)7 (78%)15.212.38–97.10<0.001No BME23 (82%)2 (22%)First, no bone marrow edema is compared with BME of any pattern. Next global BME is compared to all other patterns, including no edema. Finally, global BME is compared with no BME. Table 5 demonstrates the results of a univariate comparison between subjects with BME and without BME, and their scores on the modified Noyes and the Kellgren–Lawrence classifications. For the medial compartment only, subjects who had high scores on the Noyes and Kellgren–Lawrence scales tended to have BME, whereas those subjects with lower scores on the Noyes and Kellgren–Lawrence scales tended to have no BME. Table 5Univariate comparisons between those knees with BME and those without BME (t tests)VariableScoreNo BMEBMEp ValueModified Noyes score lateral compartment0–2a20 (80%)32 (67%)0.2322b–3b5 (20%)16 (33%)Modified Noyes score medial compartment0–2a11 (44%)8 (17%)0.0122b–3b14 (56%)40 (83%)Modified Noyes score patellofemoral compartment0–2a4 (17%)14 (31%)0.1932b–3b20 (83%)31 (69%)Kellgren–Lawrence score lateral compartment0–213 (100%)26 (79%)0.0713–40 (0%)7 (21%)Kellgren–Lawrence score medial compartment0–29 (69%)11 (33%)0.0273–44 (31%)22 (67%)Kellgren–Lawrence score patellofemoral compartment0–25 (42%)11 (33%)0.6063–47 (58%)22 (67%) The presence of a meniscal tear did not correlate with subsequent TKA. There was no significant difference in the gender between the group that received a TKA and those that did not. The group of subjects who had a TKA were 12.6 years older than those who did not have a TKA (p = <0.001). In this study, the odds of having a TKA increase by about 1.11 times for each year increase in age. A multivariable logistic regression model using generalized estimation was performed to account for the age difference in the groups and assess whether or not the results were still significant despite the age difference. Table 6 displays these results and demonstrates that, after accounting for the age difference, subjects who had BME of any pattern type were 8.95 times as likely to have a TKA when compared to subjects with no BME (p = 0.016). Subjects with global BME were 13.04 times as likely to get a TKA when compared to subjects without BME (p < 0.01). Subjects with a global pattern of BME were 5.45 times as likely to have a TKA compared to subjects with focal, cyst, or no BME (p < 0.05). Table 6Results of the multivariable logistic regression model using generalized estimation to account for the age difference in the groups that received a TKA versus those that did not Outcome: TKAVariableOR95% CIp-valueModel 1BME versus no BMEf8.951.49–53.680.0164Age1.131.07–1.20<0.0001Model 2Global BME versus No BME13.042.06–82.580.0064Age1.060.99–1.150.11Model 3Global BME versus all other patterns (focal, cyst, and no edema)5.451.02–28.960.0467Age1.1071.05–1.170.0002 Discussion BME has previously been noted in the musculoskeletal system in osteoarthrosis and other conditions [1, 4–11, 13, 14, 20]. Felson and colleagues have examined BME and its relation to progression of knee osteoarthrosis and noted that BME is a potent risk factor for structural deterioration in knee osteoarthrosis [1, 7, 17]. Its relation to progression is explained in part by its association with limb alignment. In two of his studies, medial bone marrow lesions were seen mostly in patients with varus limbs and lateral lesions were seen mostly in those with valgus limbs [7, 17]. Link and colleagues [4] reviewed MR findings in osteoarthrosis and noted that cartilage lesions, BME patterns, and meniscal and ligamentous lesions were frequently demonstrated as MR changes in patients with advanced osteoarthrosis. However, in this study, there was no significant correlation between MR and clinical findings. A recent study conducted by Kornaat et al. [21] examined multiple imaging findings and their association with clinical symptoms. Their results suggest that only findings of a large-joint effusion or the presence of an osteophyte in the patellofemoral compartment were associated with pain and/or stiffness. They found no association between BME and symptoms of pain or stiffness. Three studies have reviewed the MR appearances with histopathology findings [6, 9, 10]. Bergman and colleagues in their study found that subchondral bone marrow changes were present in seven of nine patients undergoing total knee replacement [10]. Histopathologically, those regions showed focal areas where fibrous tissue replaced fatty marrow in the subchondral trabecular space. Zanetti and colleagues reviewed the histopathological findings in 16 patients who had MR before total knee replacement. The BME pattern consisted of normal tissue and a smaller proportion of several abnormalities including bone marrow necrosis, bone marrow fibrosis, abnormal trabeculae, BME, and bone marrow bleeding. They concluded that a BME pattern in knees with osteoarthrosis represents a number of non-characteristic histological abnormalities [9]. Nolte–Ernstein and colleagues examined the correlation between MR and histological findings of degenerative bone marrow lesions in experimental osteoartrhosis models in canine knee joints [6]. In these experimental lesions, the histopathology revealed 21 osteosclerotic lesions and 5 intraosseous cysts. Histopathological findings showed different degrees of osteosclerosis associated with bone marrow degeneration. Cystic lesions were of two types: subchondral epiphyseal cysts and synovial cysts within a large tibial osteophyte. High signal intensity on T2-weighted images and decreased signal intensity on T1-weighted images indicated high fluid content. None of these prior studies specifically looked at the BME pattern on MR. Our classification of the patterns into global, focal, cystic, and absence of edema is an attempt to subdivide the presence or absence of edema in osteoarthrosis. However, this attempt is limited by the absence of histopathological findings. We were surprised by the significantly increased risk of knee joint replacement with the global pattern of BME in relation to the other patterns. It appears that the more extensive and intense the BME, the more likely it is for the patient to have symptoms. The global pattern of BME was the best predictor of risk of TKA within 3 years, as those subjects with the global pattern were over five times as likely to receive a TKA when compared to those with any of the other patterns and over 13 times as likely to have a TKA when compared with subjects with no BME, after accounting for the age difference. We were also surprised by the lack of association between cartilage loss and the likelihood of total knee replacement. Intuitively, one would think that the greater the cartilage loss, the more likely the possibility of total knee replacement. However, Link and colleagues [4] noted that clinical findings showed no significant correlation with the extent of cartilage loss on MR imaging. In this study, the lowest scores for pain and function loss were found in patients without cartilage lesions. However, the highest scores for pain, stiffness, and function loss were found in patients with less than 50% cartilage loss. The lowest scores for stiffness were found in patients with more than 50% cartilage loss and full-thickness lesions. These findings further support the theory that cartilage loss may not be the primary source of pain in patients with OA of the knee. We did find, in the medial compartment only, that subjects with high grades of cartilage loss or advanced degenerative changes on radiographs tended to have some degree of BME. The lack of association between cartilage loss and the likelihood of TKA in our study could be a reflection of the limitations of the Noyes classification system. However, other investigators have also found that radiographic signs of cartilage loss may not relate to the degree of clinical symptoms. As mentioned above, Link and colleagues found no correlation with the extent of cartilage loss on MR and clinical findings. In their study, they also used a modified Noyes classification system [4]. Kornaat et al. also found no correlation between the extent of cartilage damage seen on MR imaging and clinical symptoms of pain and stiffness. They used a grading scale based on the maximum diameter of cartilage defects and the depth of the lesion based on the percentage of cartilage loss [21]. In clinical practice, OA is associated with the radiographic findings of joint space loss and subchondral sclerosis, presumably secondary to cartilage loss that results in pain. However, as stated earlier, there are no pain fibers in hyaline cartilage [3]. Furthermore, many patients have pain out of proportion to their radiographic findings. Our study may provide an insight into another possible mechanism for pain production in OA of the knee. While not every patient with OA warrants an MR scan, those patients without the classic presentation may benefit. We did find that patients who had a TKA were 12.6 years older than those who did not have a TKA. Patients who are older are more likely to have OA and, in particular, more severe OA. Surgical replacement of the knee is more likely to occur in an older patient with OA than a younger patient with the same severity of OA. After further statistical analysis, we however found that our BME results were still significant despite the differences in age. This paper has several limitations. Most patients who have a TKA do not get a pre-operative MR, and therefore, we may be dealing with a pre-selected population. What we have called “BME” on MR may not be true edema but may relate to other histologic findings as noted earlier. Our numbers are also small, and we had to group some of our subsets together for statistical analysis. Only one musculoskeletal radiologist was involved in the review of the radiology. There was also no arthroscopic correlation for cartilage defects or presence of meniscal tears. Our study is limited by the fact that it is retrospective. To generate a group of MR scans to review, we had to use a keyword search to identify potential subjects. We would have obviously missed all scans where the specific terminology “bone edema” or knee “osteoarthrosis/osteoarthritis” was not utilized. However, we feel that this is a relatively minor limitation, as we were able to generate 381 cases for review. Another potential limitation in the retrospective study design is the possibility of missing subjects who received follow-up at a different institution or those whose symptoms resolved without treatment. In summary, we reviewed a series of knees in patients with osteoarthrosis and evaluated the pattern of BME, cartilage loss, radiographic findings when available, and the incidence of total knee joint replacement within a 3-year follow-up. Subjects with any bone marrow edema pattern were more likely to have a total knee joint replacement compared to subjects with no bone marrow edema. The worst prognostic pattern was the global pattern of bone marrow edema. Subjects who had a total knee replacement were also older than those who did not. However, even allowing for age, the global pattern of BME remained the variable with the highest statistically significant association with the incidence of total knee replacement.

Crit_Care-5-4-37409

Handheld computers in critical care

Background Computing technology has the potential to improve health care management but is often underutilized. Handheld computers are versatile and relatively inexpensive, bringing the benefits of computers to the bedside. We evaluated the role of this technology for managing patient data and accessing medical reference information, in an academic intensive-care unit (ICU). Introduction The rapid development of computing technology has had a major impact on health care, particularly in technology-oriented areas such as critical care. Electronic patient records require a major commitment by the institution, in hardware, software, training, and support. In many places, bedside care of patients still relies on paper records or nonintegrated computer systems that do not take full advantage of their data-management capabilities [1]. Even where there are advanced computerized systems, the bedside clinician may still rely on written notes for patient management and billing, and refer to pocket textbooks or printed management algorithms. For busy clinicians, the use of computers for hospital-based clinical care may be hampered by the computers' inaccessibility. Handheld computing technology is versatile and relatively inexpensive [2], combining many of the benefits of electronic patient records and paper charts. Handheld computers have been described in various medical situations; early reports describe programmable calculators used to make complex calculations in intensive-care units (ICUs) [3]. Handheld devices are increasingly being used by physicians for a variety of functions, such as scheduling, accessing drug reference information, patient data storage and billing. However, there are few published reports describing the benefits of this technology [4, 5, 6, 7]. In view of the potential advantages and increasing use of handheld computers in medicine, we evaluated the benefits and drawbacks associated with introducing this technology in an academic ICU. Materials and methods Hardware The Palm III series handheld device (Palm device, Palm Canada Inc, Toronto, Ontario) was used, as some of our staff were familiar with this equipment. It is a pocket-sized (8 × 12 cm; 165 g) computer with a 4-Mb (PalmIIIx) or 8-Mb (PalmIIIxe) memory. It has an infrared data association (IrDA) port that allows transmission of data between Palm devices and other IrDA-compatible devices such as printers, laptop computers and cellphones. The device has a monochrome 160 × 160 pixel liquid-crystal display screen (Fig. 1) and allows the user to input data either by writing on the touch-sensitive screen with a stylus or by tapping on an on-screen keyboard. Handwriting is deciphered by Graffiti handwriting-recognition software (Palm Inc, Santa Clara, CA, USA), which requires the user to learn specific characters. For users who preferred to enter data using a keyboard, two GoType keyboards (LandWare Inc, Oradell, NJ, USA) were found in the ICU. When the Palm device was placed in this keyboard, the user could type in the standard way. Software Each personal digital assistant (PDA) was installed with medical reference information as well as hospital and ICU specific guidelines (Table 1). This occupied approximately 2 Mb of memory. The applications that come with the PDA (Addressbook, Datebook, Memopad, To Do list) were used for essential telephone numbers as well as call and teaching schedules, but additional software was required for medical databases. The spreadsheet database program JFile (Land-J Technologies, Orlando, FL, USA) was used for reference information, such as drug doses and laboratory reference ranges. The text readers AvantGo (AvantGo Inc, San Mateo, CA, USA) and iSilo (www.isilo.com), which convert word-processing and HTML documents, were used for textual medical reference information. CbasPad, a Tiny BASIC programming language interpreter and editor, was used to develop software to perform common critical-care calculations, such as calculated creatinine clearance and intravenous infusion rates. Additional software for medical reference data was introduced during the study period, including ePocrates qRx [8], a drug information database. Patient data were entered into the Memopad using a customized template generated with MemoPlus (Hands High Software Inc, Palo Alto, CA, USA). The information entered included demographic data, medical history, current diagnoses, therapy, procedures performed, and management plan. Data was transferred between medical personnel using the PDA's infrared beaming ability. As hospital policy requires a paper record, daily notes were generated by Palmprint software (Stevens Creek Software, Cupertino, CA, USA) using infrared transmission to an HP Laserjet 6P printer (Hewlett Packard, Palo Alto, CA, USA). Various software packages for patient data management (shareware or commercially available software) were evaluated during the study period. In the light of focus-group feedback, a more comprehensive reference database was developed. The electronic files for the Critical Care Handbook of the Massachusetts General Hospital [9] were provided by the publishers, and converted to a PDA-readable (iSilo) format. This 1.4-Mb file contained the full text of the book, with multiple hyperlinks, and some of the images. Hard copies of the book were also obtained. Study subjects PDAs were given to the ICU attending physicians, the rotating resident trainees, and other medical staff. Four to six residents (postgraduate years 2 to 4) worked in the ICU at any one time. On the first day of their ICU rotation, residents were taught how to use the PDA in a 1-hour seminar. The principle investigator and research team were available for further help and troubleshooting throughout the study The research team was responsible for installing and updating software and schedules. Patient data was entered by residents, either during morning rounds or when patients were admitted to the ICU. The updated database was beamed to the on-call resident in the evening and transmitted back to the team in the morning, with new admissions added. Methodology An independent evaluation company (Smaller World Communications, Richmond Hill, Ontario) with experience in focus-group methodology was contracted to develop the qualitative methodology, collect data through focus-group meetings, and analyze the data [10, 11]. A preliminary moderator's guide was developed and tested on an expert panel, comprising two critical care physicians, an anaesthesiologist, three medical residents with experience in data management or PDAs, and a representative from Palm Canada Inc. The moderator's guide was designed to stimulate discussion about users' familiarity with the technology, the benefits to patient management, and the drawbacks encountered. Finally, ideas were generated for new applications for the technology and improvements to the hardware and software. Three focus groups were held with the residents and staff who used PDAs in the ICU. Tapes were transcribed verbatim and the notes were analyzed for themes by a research analyst [12]. Interim reports from the meetings were provided to the investigators. On the basis of this feedback, ongoing improvements were made to the medical databases and patient-management software. The PDA reference database was evaluated objectively using a crossover study. The trainees' rotation was split into two 3-week periods. One of the periods was allocated as a control (PDA-free) block and in the other the PDA was available. Two groups of trainees were studied: in one the PDA period preceded the PDA-free period, and in the other, the order was reversed. During the PDA period, trainees had access to the full PDA database as well as the electronic version of the Critical Care Handbook of the Massachusetts General Hospital [9]. The printed copy of the handbook was given to trainees during the PDA-free period. Objective evaluation was accomplished using a pair of standardized clinical scenario tests made up of 20 questions answered over 30 minutes. The questions were about common critical-care problems, drawn randomly from a pool of questions written by physicians in our ICU and at other teaching hospitals in the Toronto area. Trainees made use of the textbook (control period) or PDA database (study period) during the examination. To standardize for the possible difference in difficulty between the two tests, 11 General Internal Medicine trainees, not involved in the PDA study, wrote both the tests. This generated a mean and standard deviation for each test. Study trainees' results were expressed as the standard deviation above or below this control mean, and compared using a permutation test, with P < 0.05 considered significant. Results During the 6-month study period, PDAs were used by 20 physicians (4 attending physicians, 1 research fellow, and 15 rotating medical residents) and 6 paramedical staff (3 respiratory therapists, 2 pharmacists, and 1 nurse educator). The three focus groups had a total of 19 participants. Two residents who were unable to attend participated in telephone interviews. Each focus group had six or seven participants, a number within the recommended range [11]. Only five of the users (19%) had previous experience with the PDA computing format. Physical attributes Users found the PDA to be a convenient pocket size, allowing it to be available at all times. The screen was clear and easy to read, although not ideal for long text documents or large tables. Many users became proficient in text entry using Graffiti, while others preferred to use the GoType keyboards. Of the 19 PDA units used during the 6-month study period, only one had a technical malfunction requiring replacement. Two were damaged after being dropped and needed to have their screens replaced. No other problems were encountered. Medical reference databases Reference databases used regularly by medical residents included the critical-care drug dosing reference, ventilator weaning protocol, and electrolyte correction application. The calculation programs (creatinine clearance, ideal body weight) were found to be useful by the pharmacist and some residents. The ventilator weaning protocol was used by medical staff, as well as respiratory therapists, allowing regular assessment of whether patients met the criteria for extubation. Many databases were, however, not fully used. This appeared to relate more to inadequate training than to faults in the databases. In many cases, the PDA users were unaware that certain information was in their PDAs. This was because data were located on separate software programs (J-file, AvantGo, Cbas, Memopad) and may have been difficult to find. The PDA had a global 'Find' function to search for keywords, but this does not incorporate some of the added software programs, such as AvantGo. A unified database program with a search capability was suggested as a useful addition. Patient-management software Patient information was managed using the text-based MemoPlus software and a customized template. This required text entry on the PDA. Several modifications to the template were made during the study period. Residents responsible for patient data entry described difficulty entering data for new patients and keeping patient information updated during busy weekends. Attending staff found the patient data useful, particularly when they were taking over care of patients at the beginning of their on-call duties. Transferring the care of critically ill patients to a new physician is time-consuming and potentially stressful. The PDA patient database improved the staff's knowledge of patients, especially of previous medical problems in patients with complex conditions who had had a long stay in hospital. It also gave staff access to patient information when they were out of the ICU, aiding decision-making. During ICU rounds, the summarized chronological information was useful to find out how long intravenous lines had been in place and to review antibiotic therapy. Less benefit was noted in short-term patients. During night call, the patient summaries were of value when residents were called to see patients with whom they were not very familiar. In our ICU, a daily physician note is written in the patient record. The print function to create a daily note reduced duplication of work, but the process for entering patient data was found to be time-consuming initially. While residents did not feel that the patient-management application (MemoPlus) improved efficiency, it did increase their knowledge of the patients. During the study period, other commercially available patient-management software systems were evaluated. These had the advantage of easy data input using single keystrokes for date entry and 'pop-up' lists of drugs and diagnoses. While this simplified data inputting, no system was found to be ideal for the ICU. Many of these systems did not support the infrared data transfer or printing functions. Other uses of the software Study participants used a variety of other applications on a regular basis. Having the call and teaching schedules easily accessible was considered a benefit. The telephone list of hospital numbers was found to be valuable and the To Do list was used by most users to keep track of their work. Teaching rounds and morbidity and mortality rounds were facilitated by using archived patient data. Many participants used the Memopad to take notes in teaching seminars. Suggestions for change The focus-group discussions generated a number of suggestions for improvement. The hardware unit was considered suitable, but a more robust one may be needed in view of the two damaged screens. Because most of the users had had no previous experience with the PDA, additional teaching sessions and follow-up training were suggested to make optimal use of the technology. This would have helped users to become more aware of the many databases available on their PDA. In this regard, the medical information on the PDA would clearly benefit from integration into a single, searchable program. The patient-management software would be more user-friendly if the data could be entered with minimal effort, using customized pull-down lists of drugs, diagnoses, and procedures. The demographic data could be entered and updated daily by a ward clerk. Alarms were suggested - for example, to warn of prolonged intravenous line duration or the end of a course of antibiotic therapy. While transmission of data between staff by infrared was found to be useful, synchronization with the hospital electronic patient record was considered the optimal situation. Objective evaluation Two groups of four trainees took part in each crossover study. Half of the residents had prior experience with PDAs. No difference was noted in their subjective preference for the PDA or printed copy of the handbook, and the individual's preference did not correlate with previous PDA experience. Comparison of the test scores revealed no difference between the scores in the PDA-assisted test and the paper-assisted test, analyzed after correction for difficulty using the control mean and standard deviation. Discussion This study prospectively evaluated the benefits and drawbacks associated with the introduction of handheld computers in an academic-critical care environment. Who benefitted most? The introduction of handheld computers was well received by all users, despite differences in their familiarity with these devices. The most favourable response was from the more senior staff, namely, the attending physicians and fellows. This may be because of the longer time they were involved in the study, allowing more familiarity with the PDA platform. They were also more likely to benefit from having patient data available while on call outside the ICU. Furthermore, they were usually not responsible for entering patient data. Clearly, two conditions that might enhance the acceptance of these technological changes are adequate education and ease of data entry. Although an initial education session was held, it was when the junior medical staff in the study were beginning their rotation in an unfamiliar environment. Making the devices more user-friendly The patient data applications assessed were not ideal but did enable us to identify several criteria for a user-friendly system. These include ease of data entry using shortcuts and lists, limiting the range of data stored to that essential for patient management, and the ability to transmit data easily between staff. It is important that this computerized patient database should decrease workload and not cause duplication in work. In our study, enabling residents to print a daily note from their handheld computer offset the additional work of data entry. Ideally, the handheld system should be integrated with the hospital electronic patient record, allowing direct entry of demographic data as well as access to laboratory data. A wireless capability may also have significant benefits with respect to medical information databases. This would allow access to Medline searches and evidence-based guidelines. While internet access is available from desktop computers in the ICU, the ability to perform these searches on rounds or while consulting outside the ICU may be beneficial. Databases on paper or on screen? The comparison of paper and electronic databases did not reveal an advantage of one medium over the other. No significance difference was observed between the objective scenario test scores using the PDA or the paper database. The fact that equivalent results were obtained using this single database may suggest a potential benefit of using the PDA. The memory capability of the 8-Mb device would allow the trainees to carry five reference texts each of a size similar to that of the Critical Care Handbook of the Massachusetts General Hospital. What is needed Critical-care decision-making requires rapid access to strategic clinical data as well as to medical reference information. A patient in an ICU generates a large amount of data, and the number of information variables may exceed what clinicians can integrate and process [13]. Current information technology has the potential to realize the needs of the intensivist, but no customized product has been developed for this use. Handheld technology has a definite role to play, but systems need to be developed specifically for the critical-care environment to optimize real-time patient data management and communication between health care workers. Abbreviations HTML = hypertext markup language; ICU = intensive-care unit; IrDA = infrared data association; Mb = megabytes; PDA = personal digital assistant.

Brain_Struct_Funct-4-1-2248604

Sensory and cognitive mechanisms of change detection in the context of speech

The aim of this study was to dissociate the contributions of memory-based (cognitive) and adaptation-based (sensory) mechanisms underlying deviance detection in the context of natural speech. Twenty healthy right-handed native speakers of English participated in an event-related design scan in which natural speech stimuli, /de:/ (“deh”) and /deI/ (“day”); (/te:/ (“teh”) and /teI/ (“tay”) served as standards and deviants within functional magnetic resonance imaging event-related “oddball” paradigm designed to elicit the mismatch negativity component. Thus, “oddball” blocks could involve either a word deviant (“day”) resulting in a “word advantage” effect, or a non-word deviant (“deh” or “tay”). We utilized an experimental protocol controlling for refractoriness similar to that used previously when deviance detection was studied in the context of tones. Results showed that the cognitive and sensory mechanisms of deviance detection were located in the anterior and posterior auditory cortices, respectively, as was previously found in the context of tones. The cognitive effect, that was most robust for the word deviant, diminished in the “oddball” condition. In addition, the results indicated that the lexical status of the speech stimulus interacts with acoustic factors exerting a top-down modulation of the extent to which novel sounds gain access to the subject’s awareness through memory-based processes. Thus, the more salient the deviant stimulus is the more likely it is to be released from the effects of adaptation exerted by the posterior auditory cortex. Introduction The mismatch negativity (MMN) component of the event-related-potentials (ERPs) is assumed to reflect the operation of a pre-attentive memory-based comparison mechanism (e.g., Näätänen et al. 2005). For a wide range of stimuli, there is evidence that the MMN is elicited by a cognitive mechanism based on auditory sensory memory that compares between the incoming deviant stimulus and the standard template (Näätänen and Alho 1997; Näätänen et al. 2001). The contrasting view is that attention switch to novel sounds is based on the transient frequency-specific adaptation of posterior auditory-cortex feature-detector neurons (Desimone 1992; Näätänen 1990, 1992; Ulanovsky et al. 2003; Jääskeläinen et al. 2004). Specifically, it has been suggested that MMN arises because of selective adaptation of the N1 response by preceding standard stimuli (sensory component) leading to its attenuation. This attenuation is interpreted in terms of neural refractoriness (Näätänen 1992). Thus, this account of MMN elicitation does not rely on memory representations and is sensorial in nature. Previous research has shown that the MMN is sensitive to acoustic as well as to phonetic attributes of phonemes (e.g., Tampas et al. 2005). Furthermore, it was demonstrated that the MMN reflects the activation of memory networks for language sounds and spoken words (e.g., Pulvermüller et al. 2003; Shtyrov et al. 2005; Pulvermüller and Shtyrov 2006), the latter referred to as the “word advantage” effect (Pettigrew et al. 2004). The design of the current functional magnetic resonance imaging (fMRI) study was a combination of modified protocols utilized in previous studies (i.e., Pettigrew et al. 2004; Opitz et al. 2005) that enabled to disentangle cognitive and sensory contributions to change detection (Opitz et al. 2005; Maess et al. 2007) when the deviant stimulus could be either a word (“day”) resulting in the “word advantage” effect, or a non-word deviant (“deh” or “tay”). We hypothesized that the sensory component will be located bilaterally in regions posterior to Heschl’s gyrus (HG), including the posterior superior temporal gyrus (STG), regardless of the identity of the deviant stimulus (“deh”, “day” or “tay”). This component relies on frequency-specific neurons in the auditory cortex responsible for the repetition-related decrement of N1 and its counterpart, the N1m (Romani et al. 1982; Jääskelainen et al. 2004; Opitz et al. 2005). Moreover, it was shown that the region posterior to HG is broadly tuned with respect to phonetic features (Ahveninen et al. 2006). In contrast, it was shown that areas anterior to HG process sound-identity cues such as speech (Binder et al. 2000; Obleser et al. 2007; Ahveninen et al. 2006) and pitch (Warren and Griffiths 2003). Therefore, we assumed that the cognitive component which involves the representation of the current auditory event (Schröger 1997) will differentiate between the deviant word (“day”) and the non-words (“teh” and “tay”, i.e., the “word advantage” effect) and will be located anterior to HG. Methods Subjects Twenty right-handed adult healthy subjects, native speakers of English, 8 women and 12 men, participated in the study. Subjects ranged in age from 23 to 28. All subjects gave informed consent to participate in the study. Experimental protocol The procedure is a modification of the one used by Opitz et al. (2005). Each subject participated in 12 functional imaging runs. The speech stimuli were grouped into two types of blocks “oddball” and “control”. In the first type of block, the “oddball” block, deviant exemplars occurred quasi-randomly embedded within standard stimuli so that the frequency of occurrence throughout the block was balanced among exemplars. The interval between two successive deviants varied quasi-randomly with gaps of 4, 6, 8, 10 or 12 standards with the constraint that gap size was balanced throughout the block. In the “Oddball” blocks deviants appeared 42 times out of a total of 354 stimuli (probability of occurrence = 12%). In the “control” blocks the same physical deviants and standards as in the “oddball” blocks (deviant-counterparts and standard-counterparts, respectively), occurred quasi-randomly, while deviants were constrained by the same spacing rule mentioned above. However, each of the “control” blocks contained eight different equiprobable stimuli, including the deviant and standard counterparts. Thus, in the control runs the seven stimuli beside the deviant-counterpart served as “filler” or contextual stimuli which were added to the sequence so that the deviant-counterpart will appear at the same probability as any other stimulus in the sequence. Each of the stimuli in the “control” block (a total of eight different stimuli) repeated 42 times and appeared with equal probability which was identical to that of the deviants in the “oddball” block (12%). There were three blocks of each type (“oddball”/“control”). Each block was repeated twice. In total, 12 blocks of approximately 6 min each were randomly presented for each subject within a session. Three non-words, /de:/ “deh”; /te:/ “teh”; and /teI/ “tay”, and one word, /deI/ “day”, were recorded from a male native English speaker in a sound-proof chamber. These stimuli comprised the following standard-deviant pairs that resulted in the three “oddball” runs: (1) deh (standard)–day (deviant); (2) day–deh and (3) teh–tay. Thus, “deh” and “day” swapped their roles as standard and deviant in the second “oddball” block, whereas the third block controlled for the acoustic change associated with the transition from a monophthong (/e:/) to a diphthong (/eI/) occurring in the first standard-deviant pair. Additional four stimuli, /pe:/ “peh”; /peI/ “pay”; /be:/ “beh”; and /beI/ “bay”, together with the previous four mentioned above were embedded within three control runs, each containing a deviant-counterpart , either “day”, “deh” or “tay” appearing with the same probability and obeying the same spacing rule as in the oddball blocks. Three exemplars for each stimulus (e.g., deh1, deh2, deh3) were selected (out of a pool of 24 recordings per stimulus) on the basis of acoustic similarity (see Table 1 for the values of the lowest three formants). The parameters that were used to choose similar exemplars for each stimulus included the shape of the spectrogram at the voice onset, vowel durations, pitch and formant values (Hz) of the first three formants at the beginning (ca. 84 ms) and end (ca. 168 ms) of the /e/ segment (ca. 252 ms duration) of each stimulus. The stimuli were truncated to 280 ms and normalized to the same loudness level by using Adobe Audition 1.5 trial version software package. Spectral analysis of the stimuli was conducted by PRAAT software version 4.3.19 (http://www.praat.org). Table 1Pitch and frequencies of the main speech stimuli (in Hz)StimulusF0F1 onset of /e/F1 end of /e/MeanF2 onset of /e/F2 end of /e/MeanF3 onset of /e/F3 end of /e/Meandeh11015546346031,7141,6671,7022,5482,6172,636deh2995396406241,7261,5931,6582,5762,5832,636deh31015396306141,7301,6001,6662,5852,5902,635day11044774064112,0762,2302,1422,6932,7062,703day21024924124191,9832,1372,0892,6452,6692,668day31034974624512,0182,0632,0512,6842,7032,685tay11015815275371,8581,9111,9422,5812,4952,641tay21026105666001,7871,8871,9062,6712,6352,670tay31036005896221,7901,8391,9152,6602,6222,681teh11036366237091,7091,6621,7472,5872,6252,695teh21017836267071,7501,6501,7352,5842,5252,639teh31018576157471,8051,6081,7482,5632,5312,655F0 = pitch. F1, F2 and F3 indicate the frequencies (Hz) of the first, second and third formants, respectively. F1, F2 and F3 frequencies are given for the beginning and end of the /e/ segment common to the four main stimuli of the study. The mean frequency across the length of the common /e/ section is also indicated The reason for using three tokens for each consonant-vowel stimulus was to control as much as possible for acoustic factors, other than those inherent in the structure of the stimulus, which could confound the elicitation of a deviant response. Using three different exemplars for each consonant-vowel (CV) stimulus diminished the likelihood of a contingency developing between a specific deviant-standard pair because of an uncontrolled acoustic facet associated with either the deviant or the standard. As could be seen in Table 1 the frequency parameters across homogeneous exemplars (i.e., the three representatives of the same CV) were very similar to each other. In addition, the F1/F2 ratio in “teh” and “tay” is similar to the F1/F2 ratio in “deh” and “day”, respectively. Thus, the comparison between the responses to “teh” and “tay” could serve as a suitable control for the latter pair (Jacobsen et al. 2004). The speech stimuli appeared randomly within each block and their occurrence was balanced throughout the block for standards as well as deviants. Stimuli were presented with an SOA (stimulus onset asynchrony) of 1 s at 95 dB SPL via headphones to subjects with ear-plugs (see Table 2 for an example of a “control” and “oddball” blocks in case of the “deh”–“day” contrast). Stimulus presentation was carried out by E-Prime 1.1 (1.1.4.1) (Psychology Software Tools http://www.pstnet.com). Table 2“Oddball” and “control” sequences for the “deh” (standard)-“day” (deviant) pairOddballdeh1(A) deh1 deh3 deh2 day1(B) deh1 deh2 deh1 deh3 deh1 deh1 deh3 day3…Controlbeh1 deh3(C) tay2 teh3 bay2 bay3 peh2 deh2 day2(D) peh3 bay1 teh3 pay1 day2….A–D indicate the stimuli contrasted to evaluate: the deviance effect (B vs. A), the cognitive effect (B vs. D) and the sensory effect (A vs. C). The numbers attached to the stimuli indicate different exemplars of the same CV stimulus. Each contrast was computed across all exemplars of a specific CV. In the “oddball” sequence “day1” serves as a deviant and “deh1” functions as a standard. In the “control” condition “deh3” is a standard control counterpart and “day2” serves as a deviant control counterpart Contrasting “oddball” deviants and standards with the corresponding stimuli in “control” runs (deviant and standard counterparts) allowed disentangling the two mechanisms of deviance detection, namely, the sensory one based on refractoriness (Näätänen and Picton 1987) from the cognitive one based on memory-based processes (Näätänen 1990; Näätänen and Winkler 1999). The subjects were instructed to count every stimulus, press a button when they reached 100 and then reset their count and restart counting from zero again. This task was chosen to control for a possible contamination by N2b and P3b ERP-components which are elicited when deviants are being discriminated from standards (Donchin et al. 1997; Opitz et al. 2005). Behavioral assessment of discrimination between stimuli Thirteen subjects (out of the 20 participating in the study) rated the stimuli in a behavioral session performed outside of the magnet. The behavioral assessment was carried out in a different session. Stimuli were presented simultaneously with recorded MR scanner noise. The stimuli and the scanner noise were both presented in 62 dB SPL which was a convenient hearing level for both the stimuli and the noise presented together. Subjects were presented with “triplets” comprised of stimuli containing /e:/ and /eI/. In each block, 15 triplets were presented separated by a 2 s interval of silence allowing the subject to respond. The stimuli were the same as those presented in the fMRI experiment. There were several patterns of triplet presentation in each block, as follows. There were three triplets in which “day” was presented first followed by two “deh”-s; three triplets in which “day” was presented last and preceded by two “deh”-s; three triplets where “deh” was presented first followed by two “day”-s and three triplets were “deh” was last and preceded by two “day”-s. In addition, there were three triplets in which “day”, “deh” and “tay” each appeared in the middle position, respectively, while the neighboring stimuli (in the first and third positions) where either the /eI/ or /e:/ counterparts (i.e., “deh” “day” “deh”; “day” “deh” “day” and “teh “tay” “teh”). Overall, four blocks of 15 triplets each were presented to the subjects. The order of triplets was randomized within each block. The three different exemplars representing each CV stimulus were balanced across the four blocks. The subjects were required to indicate the outlier in each triplet by pressing the key (either “1”,”2” or “3” on the keyboard) that corresponded to the position of the outlier in the triplet. The outliers were assigned mainly to the extreme positions (1 or 3) in the triplet to simulate more closely the “oddball” design in which the deviant is surrounded by repetitive standards. The trials with the outlier appearing in the middle position were introduced to minimize the probability of guessing the identity of the third stimulus in the triplet after hearing the first two which were non-identical. Thus, subjects could guess the identity of the outlier with more confidence only after hearing two identical stimuli in a row. Data acquisition parameters Data was collected on a 3T Siemens Trio scanner. Each study began with two localizers: a three-plane localizer and a multiple-slice sagittal localizer. These were followed by the acquisition of twenty-five 6 mm T1-weighted axial slices (TR = 300 ms, TE = 2.47 ms, flip angle = 60°, FOV = 220 mm, 256 acquisition matrix). For each subject, 12 functional imaging scans were collected with slices in the same locations as the anatomical T1-weighted data. Functional images were recorded using a gradient-echo EPI sequence (TR = 1,550 ms, TE = 30 ms, flip angle = 80°, FOV = 220 mm, 64 acquisition matrix). Each functional run involved the acquisition of 245 volumes with twenty-five 6-mm axial slices. Images were converted to analyze format and the first ten volumes of each functional series were removed to account for the approach to steady-state magnetization, leaving 235 volumes for analysis. Image analysis Preprocessing First, using sinc interpolation, the data from each slice were adjusted for slice acquisition time and then all data were motion corrected using SPM99 for six rigid body motions (displacement in the x, y, z direction and rotation: for pitch, roll, yaw). Flags were set for de-correlation and masking so a pixel was set to zero for every time point if it moved outside the volume. Functional image data were motion corrected by realigning the time sequence imaging to the first image in the middle run of the sequence using SPM99. Individual subject data (responses to the deviants, standards and control stimuli) were analyzed using a General Linear Model on each voxel in the entire brain volume (Rajeevan et al. 2007). The data were normalized to a signal measure of 100 and were spatially smoothed with a 8.08 mm Gaussian kernel to account for variations in the location of activation across participants. The output maps are normalized beta-maps which are in the acquired space (3.44 mm × 3.44 mm × 6 mm). To take these data into a common reference space, three registrations were calculated within the Yale BioImage Suite software package (http://www.bioimagesuite.org/, Papademetris et al. 2006) using the intensity-only component of the method reported in Papademetris et al. (2004). The first registration performs a linear registration between the individual subject raw functional image and that subject’s 2D anatomical image. The 2D anatomical image was then linearally registered to the individual’s 3D anatomical image. The 3D differs from the 2D in that it has a 1 × 1 × 1 mm resolution whereas the 2D z-dimension is set by slice-thickness and its x–y dimensions are set by voxel size. Finally, a non-linear registration is computed between the individual 3D anatomical image and a reference 3D image. The reference brain used was the Colin27 Brain (Holmes et al. 1998) commonly applied in SPM and other software packages. All three registrations were applied sequentially to the individual normalized beta-maps to bring all data into the common reference space. Statistical analyses We used a two-stage random-effects model to analyze the data. In the first stage, statistical maps were calculated in the comparison of interest for each subject as described above. In the second stage, across subject analyses were conducted and the distribution of the individual subjects’ statistics were tested for significance. Within subject analyses For each of the three deviants (“deh”, “day” and “tay”) the following six types of maps were calculated: (1) Deviant maps computed from oddball runs extracting the response to low frequency stimuli (2) Standard maps computed from oddball runs extracting the response to high frequency stimuli (3) Deviant control maps computed from control runs extracting the response to the same physical stimuli that served as deviants in the oddball runs (deviant-counterparts) (4) Standard control maps computed from control runs extracting the response to the same physical stimuli that served as standards in the oddball runs (standard-counterparts). The contrasts extracting each deviant-counterpart and standard-counterpart stimuli were carried out against the same baseline that formed part of the “filler” stimuli, i.e., “beh”, “peh”, “bay” and “pay”. The stimuli that served as deviants and standards (i.e., “deh”, “day”, “teh”, “tay”) were taken out of the baseline so that their extraction could be conducted against a common baseline. (5) A Word map computed from control runs contrasting the response to the meaningful word (“day”) with the non-words that formed part of the “filler” stimuli, i.e., “deh”, “beh”, “peh”. (6) A Non-word map computed from control runs contrasting the response to the non-word (“tay”) with the non-words that formed part of the “filler” stimuli (i.e., “deh”, “beh”, “peh”.) The non-word “tay” was excluded from the baseline, since it was the only non-word with a diphthong (/eI/), and to avoid contrasting it with itself. Across subject analyses Composite maps. Two types of composites were created: three composites of deviant maps, one for each deviant (“deh”, “day”, “tay”) and three composites of the deviant-control maps, one composite for each control run that included either “deh”, “day” or “tay” as a deviant-counterpart, respectively. Paired t tests. The contrast between deviants and their respective counterparts isolates the cognitive component of deviance detection. For this contrast, a paired t test was calculated between the Deviant map and the Deviant control map. The contrast between standards and their respective counterparts accounts for the sensory component based on refractoriness. For this contrast, a paired t test was calculated between the Standard map and the Standard control map. Statistical images were corrected for multiple comparisons over the whole brain using a magnitude threshold derived from Monte–Carlo simulation that takes into account the number of contiguous activated voxels (Forman et al. 1995). Individual voxel thresholds were set at P < 0.001 for both the composites and paired t test maps. Data were corrected for multiple comparisons by spatial extent of contiguous supra-threshold individual voxels (experiment-wise P < 0.001 for a cluster). In a Monte–Carlo simulation within the AFNI software package, using a smoothing kernel of 8.08 mm, a connection radius of 7.72 mm on 3.48 mm × 3.48 mm × 6 mm voxels, it was determined that an activation volume of 1,278 μl satisfied the P < 0.001 threshold. Regions of interest analyses. The different regions of interest (ROIs) were identified on the basis of the results obtained by the paired t test maps prior to cluster-size correction (P < 0.001) reflecting the cognitive and sensory mechanisms of change-detection. Then, each participant’s model estimate of the percent signal change for each region of activation, averaged across voxels within the region was calculated for each of the six composite maps (three deviant maps and three deviant-control maps). The ROIs identified were analyzed in a 3 (ROI: anterior auditory cortex, posterior auditory cortex, superior/middle frontal gyrus) × 2 (task-type: “oddball”/“control”) × 3 (deviant: “deh”, “day”, “tay”) × 2 (laterality: left/right) repeated measures ANOVA with subjects (n = 20) as a random factor. ROI analysis for the Word/Non-word maps. To further investigate the “word effect”, each participant’s model estimate of the percent signal change for each region of activation, averaged across voxels within the region was calculated for the Word map and Non-word map. Both types of maps were calculated from control blocks. Specifically, to create the Word map “day” was contrasted with a baseline comprised of “filler” or contextual stimuli that were non-words (i.e., “beh”, “peh”, “deh”, “the”). To create the Non-Word map “tay” was contrasted with the same baseline. The delta blood-oxygen-level-dependent (BOLD) measures were subjected to a 2 (Laterality: left/right) × 3 (ROI: anterior auditory cortex, posterior auditory cortex, superior/middle frontal gyrus) × 2 [Lexical Status: word (“day”) vs. non-word (“tay”)] repeated measures ANOVA with subjects (n = 20) as a random factor. Results Behavioral results There were no differences in response accuracy scores, computed across the three possible positions of the outlier, among the three CVs. The response accuracy for “day” was 95.88% (±5.01), for “tay” 96.70 (±5.15) and for “deh” 95.33 (±4.22) (the number in brackets denotes the standard deviation). A paired t test that was run on the individual response accuracy data for each CV confirmed that there was no difference in accuracy levels between CVs (in all three possible comparisons P > 0.1). We have also examined the response accuracy scores for the first and third positions in the triplet for each of the CVs. The response accuracy scores ranged between 98 and 99% among the three different CV stimuli as well as between the first and third positions (since there were only four triplets per block in which the outlier appeared in the middle, the middle position was excluded from the latter calculation). Table 3 displays the reaction time data for each deviant and for each of the three possible positions of the outlier within a triplet. As expected, it could be seen that the reaction time for the third position in the triplet was the shortest. To evaluate this trend statistically a two-way ANOVA was run with Stimulus (“day”, “tay”, “deh”) and Position (first, second, third) as factors and with Greenhouse–Geisser adjustment to account for sphericity. Only the Position factor was significant [F(2,38) = 28.70, P < 0.001]. Paired comparisons (Bonferroni corrected, P < 0.05) confirmed that reaction time associated with stimuli in the first and second position in the triplet was prolonged relative to that associated with the third stimulus in the triplet. Taken together, there is dissociation between response accuracy and reaction time data. While no differences in response accuracy were found, reaction time was shorter when the outlier CV stimulus was in the third position in the triplet. Table 3Reaction time (ms) in the behavioral triplet testStimulusPosition123day787.58 ± 132.07907.01 ± 240.71638.55 ± 215.98tay858.59 ± 180.23961.06 ± 187.15650.98 ± 212.58deh864.03 ± 164.90847.17 ± 158.27628.33 ± 137.63Reaction time is given for each of the three positions in the triplet and for each stimulus that served as a deviant in the “oddball” blocks. ±Standard deviation This confirms our prediction that reaction time will be shorter in case of an outlier presented in the third position following two identical stimuli. However, the fact that the main effect of Stimulus as well as the interaction between Stimulus and Position did not reach significance confirms that the stimuli were equally discriminable in the context of the “oddball” paradigms used in this experiment. Composite maps Deviance effect The spread of activation associated with “tay” deviant was very similar to that in response to “day” (Fig. 1). To elucidate the positions and extents of the activated brain regions associated with the similar responses to the deviants “day” and “tay”, images in the coronal and sagittal planes were included (Fig. 1, bottom panels). Fig. 1Deviant maps: brain regions in response to deviant stimuli embedded within the “oddball” paradigms. BOLD contrasts were superimposed on a reference anatomical image (Holmes et al. 1998). Upper panels: axial sections displaying responses to the deviants “deh” (left), “day” (middle) and “tay” (right). Bottom panels: coronal (two left panels) and sagittal (two right panels) sections of the responses to the deviants “day” and “tay”. Axial sections range from z = −6 to 46 (by increments of 4 mm). Coronal sections range from y = −1 to 50 (by increments of 5 mm). Sagittal sections range from x = −57 to 59 (by increments of 8 mm). Display follows radiological convention (left side of the brain is displayed on the right side of the screen) The size of the right STG region activated in the word condition (i.e., “day”), was similar to that activated in the non-word condition (i.e., “tay”) (Table 4). In contrast, the size of the left STG region activated in the word condition was almost twofold larger than that activated by the “tay” condition. In both hemispheres, the brain regions activated in the “day” and the “tay” conditions were larger than those observed for the “deh” condition. Furthermore, the “day”, “tay” and “deh” stimuli appear to activate different regions of the STG. The “day” stimulus activated parts of the posterior STG (i.e., BA 22/42), whereas the “tay” and “deh” stimuli activated large parts of the superior and middle temporal cortices (i.e., BA 22/22) (Table 4). Table 4Brain regions activated in Deviant maps (Fig. 1)StimulusVolume size (mm3)Mean t valueTalairach coordinates (mm)SideIdentified regionBAxyzday10,8034.7256−224RTSTG22/42day11,3494.73−59−216LTSTG22/42deh3,5624.9155−224RTSTG21/22deh1,5544.4−62−234LTSTG22tay9,1074.6856−182RTSTG21/22tay6,0344.97−57−183LTSTG21/22Talairach coordinates (Talairach and Tournoux 1988) are given for the center of mass. The mean t value was computed across the voxels within an anatomical region. Maps were thresholded at P < 0.001, corrected for multiple comparisonsSTG superior temporal gyrus, RT right, LT left, BA Brodmann area(s) Similar brain regions to those implicated in the deviance response were also found in the study of Opitz et al. (2005) in the context of tones. Specifically, in that study the deviance effect also implicated HG (primary auditory cortex) and the superior temporal plane (Talairach coordinates of peak location: −49, −14, 9; 53, −21, 10, respectively). Deviant control maps While activity extended from the STG across the lateral sulcus to the central sulcus in response to “day”, “deh” had a similar but more focal response in the vicinity of the primary auditory cortex. To elucidate the positions and extents of the activated brain regions associated with the more similar responses to the deviants “day” and “deh”, images in the coronal and sagittal planes were included (Fig. 2, bottom panels). In response to “tay” activity in the STG did not reach significance and activation was only observed in the left precentral gyrus (Fig. 2, Table 5). Fig. 2Deviant control maps: brain regions in response to deviant-counterparts stimuli embedded within the “control” paradigms. Upper panels: axial sections displaying responses to the deviant controls “deh” (left), “day” (middle) and “tay” (right). Bottom panels: coronal (two left panels) and sagittal (two right panels) sections of the responses to the deviant controls “day” and “deh”. Additional display details as in Fig. 1Table 5Brain regions activated in Deviant control maps (Fig. 2)StimulusVolume size (mm3)Mean t valueTalairach coordinates (mm)SideIdentified regionBAxyzday21,5944.9353−1615RTPoCG/PrCG STG40/43/6day27,9194.93−52−1816LTPoCG/PrCG40/43/6day4,600−4.39−244830LTSFrG9day2,860−4.2−9−80−14LTLT OCCP18/19deh4,3904.6853−219RTSTG42deh5,0214.4−51−216LTSTG42tay3,4014.6−49−838LTPrCG6PoCG postcentral gyrus, PrCG precentral gyrus, SFrG superior frontal gyrus, STG superior temporal gyrus, RT right, LT left, OCCP occipital pole, BA Brodmann area(s). Other details as in Table 4 The size of STG activation was the most extensive over the left STG for deviant “day” in comparison to “tay” where it did not reach statistical significance, and “deh” where it was more than five times smaller (Fig. 2, Table 5). In addition, in response to “day” a negative differential BOLD was evident in the left superior frontal gyrus and left occipital pole (Fig. 2, Table 5). Paired t test maps Cognitive effect The cognitive effect was derived by contrasting the Deviant maps with the corresponding Deviant control maps. Only the contrast map for “day” showed activations at the significance level used throughout this study (P < 0.001, corrected for multiple comparisons), reflecting the “word-advantage” effect. The negative differential BOLD associated with the cognitive effect implicated bilaterally the region extending from the postcentral gyrus to HG including parts of the Sylvian fissure, lateral sulcus and insula (Table 6, Fig. 3). The location of this region is proximal to the one obtained by Opitz et al. (2005) that was associated with the cognitive mechanism and located bilaterally in a non-primary auditory area within the lateral temporal lobe in the anterior rim of HG (Talairach coordinates of peak location: −42, −13, 6 and 49, −12, 7). Table 6Brain regions implicated in the cognitive effect (Fig. 3)StimulusVolume size (mm3)Mean t valueTalairach coordinates (mm)SideIdentified regionBAxyzday4,179−4.3741−1616LTLS/INSa40/43day3,802−4.5−50−2019RTLS/INSa40/43LS lateral sulcus, INS insulaaRegion chosen for ROI analysis. Other details as in Table 4Fig. 3Brain regions implicated in the cognitive effect, showing less activation in the Deviant map than in the Deviant control map for “day”. BOLD contrasts were superimposed on a reference anatomical image (Holmes et al. 1998). Axial sections range from z = −6 to 46 (by increments of 4 mm) Refractoriness effect By looking at the t test images associated with refractoriness (Fig. 4) a negative differential BOLD could be observed posterior to HG in response to each of the standard-deviant pairs (Table 7). In case of the refractoriness effect when “deh” served as the standard activity also extended to the superior temporal sulcus (Fig. 4, left panel). Opitz et al. (2005) have also found activations in a proximal brain region in the lateral aspect of the posterior rim of HG bilaterally that was associated with the sensory mechanism of change detection (Talairach coordinates of peak location: −51, −18, 7 and 53, −19, 4). Fig. 4Brain regions implicated in the sensory (refractoriness) effect, showing less activation in Standard maps than in Standard control maps. Responses for the contrasts with “deh” (left) “day” (middle) and “teh” are shown. BOLD contrasts were superimposed on a reference anatomical image (Holmes et al. 1998). Axial sections range from z = −6 to 46 (by increments of 4 mm)Table 7Brain regions implicated in the sensory (refractoriness) effect (Fig. 4)StimulusVolume size (mm3)Mean t valueTalairach coordinates (mm)SideIdentified regionBAxyzday2,777−4.2759−187RTSTGa22/42day4,473−4.39−59−204LTSTGa22/42deh7724.08274330RTS/MFrGa8/9deh2,5104.29−334330LTS/MFrGa8/9deh3,1894.3131442MedialACC32deh2,726−4.4157−263RTSTG22deh3,030−4.39−57−183LTSTG22tay2,048−4.3559−226RTSTG22/42tay2,395−4.53−53−174LTSTG22STG superior temporal gyrus, S/MFrG superior/middle frontal gyrus, ACC anterior cingulate cortexaRegion chosen for ROI analysis; other details as in Table 4 From here on, we will refer to the region posterior to HG (Fig. 4, Table 7) as posterior auditory cortex. This region was defined functionally as showing a refractoriness effect in the current study. The region anterior to HG extending from the postcentral gyrus to HG (Fig. 3, Table 6) will be referred to as anterior auditory cortex. This region was defined functionally as showing a cognitive effect in the current study. This terminology is based on the one used by Opitz et al. (2005) that described the regions implicated in the sensory and cognitive effects as residing in the vicinity of the posterior and anterior rim of HG, respectively. In addition, when the sensory component was extracted for “deh” standard (Fig. 4, left panel and Table 7) positive differential BOLD was observed in the superior/middle frontal gyrus as well as in the anterior cingulate cortex (ACC). Thus, the positive differential BOLD observed in the ACC and in the frontal gyrus is a direct result of contrasting “deh” with a baseline of “day” deviants to create the map reflecting the sensory component for “deh” standard. ROI analysis Six ROIs were chosen to be included in the ROI analysis. These ROIs were defined on the basis of previous findings and the data obtained in the current study. We have chosen the ROIs which were directly linked to the mechanisms associated with change detection, namely, the cognitive and sensory effects, or those that might be related to both change-detection and to the processing of linguistic stimuli. This is described in more detail in the section “Discussion”. The following ROIs were identified on the basis of the paired t test maps. From the paired t test map reflecting the cognitive effect (Fig. 3) the anterior auditory cortex was chosen bilaterally. From the paired t test map reflecting the refractoriness effect when “day” served as a standard the posterior auditory cortex was chosen bilaterally (Fig. 4, middle panel). From the paired t test map reflecting the sensory component when “deh” served as the standard the superior/middle frontal gyrus was chosen bilaterally (Fig. 4, left panel, Table 7). Note that all the ROIs were defined based on the t-maps prior to cluster-size correction (P < 0.001). All except the right superior/middle frontal gyrus were apparent in the t test maps after cluster-size correction. Figure 5 displays the average delta BOLD for each condition (“oddball”/“control”) and hemisphere (left/right) as a function of deviant stimulus and ROI. From Fig. 5 it can be seen that the largest differences in delta BOLD between conditions (“oddball”/”control”) are evident over the left hemisphere, mainly the anterior auditory cortex and superior frontal gyrus. These were larger for “day” than “deh” and smallest for “tay”. The latter was associated with diminished levels of delta BOLD within the posterior auditory cortex in the control condition. Fig. 5Average delta BOLD for each condition (“oddball”/“control”) and hemisphere (left/right) as a function of ROI and deviant stimulus (“deh”, “day” or “tay”). Anter_Aud anterior auditory cortex, Post_Aud posterior auditory cortex, Sup/Mid_Frontal superior/middle frontal gyrus The ROI data was subjected to a four-way ANOVA with Laterality (Left, Right), ROI (anterior auditory cortex, posterior auditory cortex, superior/middle frontal gyrus), Deviant (“deh”, “day”, “tay”) and Task type (Control, Oddball) as independent factors, with an appropriate adjustment for sphericity (Greenhouse–Geisser correction). A main effect of ROI was found [F(2,38) = 49.17, P < 0.001], as well as a Laterality × ROI [F(2,38) = 5.39, P < 0.05], Deviant × ROI [F(4,76) = 4.23, P < 0.05], Task Type × ROI [F(2,38) = 7.25, P < 0.01], and Deviant × Task Type × ROI [F(4,76) = 4.60, P < 0.01] interactions. The Laterality × ROI interaction was due to enhanced delta BOLD over the left hemisphere associated with the anterior auditory cortex (P < 0.05, Bonferroni corrected). The Deviant × Task Type × ROI interaction is depicted in Fig. 6. A set of paired comparisons (Bonferroni corrected, P < 0.05) were conducted separately within each task-type (“oddball”/”control”). In the “oddball” condition the comparisons were conducted among the three deviants for the posterior auditory cortex only, where a gradient of delta BOLD could be observed as a function of deviant type. In the “control” condition comparisons were performed among the three deviants for each of the ROIs. In the “oddball” condition a significantly larger delta BOLD was found in response to “tay” than to “deh” in the posterior auditory cortex. In the “control” condition delta BOLD to the word “day” in the anterior and posterior auditory cortices was significantly larger than that associated with the non-word “tay”. Fig. 6ROI × stimulus ("deh", "day","tay") × condition (“oddball”/“control”) interaction. The interaction was obtained by calculating the individual percent signal change in each ROI (across hemispheres) for each of the six composite maps (three Deviant maps and three Deviant-control maps). See text for the results of the a priori tests. Abbreviations as in Fig. 5 In a separate set of paired comparisons (Bonferroni corrected, P < 0.05) the averages of delta BOLD between conditions (“oddball”/”control”) were compared for each of the three deviant stimuli and each of the ROIs (across hemispheres). A larger delta BOLD was found for the response to “day” in the “control” condition relative to the “oddball” condition implicating the anterior auditory cortex. ROI analysis for the Word and Non-word maps To further investigate the “word effect”, each participant’s model estimate of the percent signal change for each region of activation, was calculated separately for the Word map and Non-word map (Fig. 7, Table 8). The ROI data was subjected to a three-way ANOVA with Laterality (left, right), ROI (anterior auditory cortex, posterior auditory cortex, superior/middle frontal gyrus) and Lexical Status [word (“day”), non-word (“tay”)] as independent factors, with an appropriate adjustment for sphericity (Greenhouse–Geisser correction). A main effect of ROI was found [F(2,38) = 21.03, P < 0.001] as well as a Lexical Status × ROI interaction [F(2,38) = 6.05, P < 0.01] (Fig. 8). Paired comparisons (Bonferroni corrected, P < 0.05) found a significant difference between “day” and “tay” only within the anterior auditory cortex. Fig. 7Word and Non-Word maps. Brain regions in response to “day” deviant-counterpart (Word map) and “tay” deviant-counterpart (Non-word map). Stimuli were contrasted against a baseline comprised of the non-word “fillers” (i.e., “peh”, “beh”, “deh”, “teh”). Left panel: response to the Word-map. Right panel: response to the Non-word map. BOLD contrasts were superimposed on a reference anatomical image (Holmes et al. 1998). Axial sections range from z = −6 to 46 (by increments of 4 mm)Table 8Brain regions activated in Word and Non-word maps in response to “day” and “tay”, respectively (Figs. 7, 8)StimulusVolume size (mm3)Mean t valueTalairach coordinates (mm)SideIdentified regionBAxyzday (Word-map)12,5794.6355−1614RTPoCG/PrCG STG40/43/6day (Word-map)14,2604.65−53−1813LTSTG42tay (Non-Word-map)4,8464.92−52−834LTPrCG4PoCG postcentral gyrus, PrCG precentral gyrus, STG superior temporal gyrus, other details as in Table 4Fig. 8ROI × Lexical Status interaction. The interaction was obtained by calculating the individual percent signal change in each ROI for the Word and Non-word maps. Abbreviations as in Fig. 5 The distribution of percent signal change across subjects To learn more about the consistency and extent of activation of specific brain regions, data were analyzed for individual subjects. Specifically, the percent signal change of individual subjects was computed for each of the six ROIs chosen for the group level analysis as well as for additional two brain regions that might be implicated in speech perception, that is, the left occipital pole and the left precentral gyrus. The results showed that the majority of subjects within each statistical test showed the same pattern of activity (see Table 9). Table 9The distribution of percent signal change across subjectsBrain regionSideMap typeNo. of subj. displaying increases (+)/decreases (−) (n = 20)Mean ± SD % signal changeLS/INSRTCognitive (“day”), Fig. 317(−)−0.09 ± 0.07LS/INSLTCognitive (“day”), Fig. 318(−)−0.09 ± 0.10STGRTRefractoriness (“day”), Fig. 418(−)−0.15 ± 0.17STGLTRefractoriness (“day”), Fig. 417(−)−0.16 ± 0.14S/MFrGRTRefractoriness (“deh”), Fig. 413(+)0.06 ± 0.12S/MFrGLTRefractoriness (“deh”), Fig. 413(+)0.07 ± 0.10PrCGLTDeviant control (“tay”), Fig. 219(+)0.13 ± 0.13OCCPLTDeviant control (“day”), Fig. 218(−)−0.24 ± 0.20SD standard deviation. For each region and hemisphere the number of subjects presenting increases or decreases in percent signal change is indicated as well as the mean and SD across subjects. The stimulus in brackets denotes the specific map from which the brain region was selected. For the regions associated with the t test maps of the cognitive and refractoriness effects, data are reported for the difference percent signal change between the contrasted conditions in the t testLS lateral sulcus, INS insula, STG superior temporal gyrus, S/MFrG superior/middle frontal gyrus, PrCG precentral gyrus, OCCP occipital pole, RT right, LT left, No. of subj number of subjects, n total number of subjects that participated in the study Discussion Main findings Overall, the results of the current study confirmed our hypotheses. Specifically, the sensory mechanism was located in the posterior auditory cortex (Fig. 4, Table 7) and the cognitive mechanism in the anterior auditory cortex (Fig. 3, Table 6). These locations match, respectively, those found in the context of tones (Opitz et al. 2005). Moreover, as expected, the cognitive effect was statistically significant only when the deviant was a word (i.e., “day”). The similarity between the brain regions activated by speech in this study and the regions activated by tones (Opitz et al. 2005) could not be explained by the analysis of the physical sound features. First, the contribution of the acoustic component to the cognitive mechanism in this study was removed by contrasting deviants with their physical identical control counterparts. Second, our data suggest that the salience of the sounds, either tones or speech stimuli, accounts for the activation of the anterior auditory cortex (Jääskeläinen et al. 2004; Ahveninen et al. 2006) (see also ‘Support for the existence of “what” and “where” auditory streams’). Thus, the acoustic component could only account for the similar activation observed in the posterior HG. This is also in line with previous findings that link dorsal temporal brain areas including HG and planum temporale with the analysis of auditory features of complex sounds (Binder et al. 1996; Wessinger et al. 2001; Hall et al. 2002, 2003; Seifritz et al. 2002). Activation loci in superior and middle temporal areas that were found to be activated in the Deviant (Fig. 1) and Deviant control maps (Fig. 2) had also been earlier reported to contribute to lexical and semantic processing (Price 2000; Salmelin et al. 2000; Scott and Johnsrude 2003; Hickok and Poeppel 2004). Similarly, the left dominance for processing intelligible speech that was found in our study in the “oddball” condition match previous results that link the left posterior STG to higher level linguistic processes (Narain et al. 2003). Support for the existence of “what” and “where” auditory streams The results of this study are in agreement with the segregation of the auditory system into “what” and “where” processing streams (Jääskeläinen et al. 2004; Ahveninen et al. 2006) associated with the analysis of auditory object content and location features that reside in areas anterior and posterior to primary auditory cortex, respectively (Rauschecker and Tian 2000). It was suggested (Jääskeläinen et al. 2004) that the “where” system is responsible both for fast analysis of sound location which is important for attentional orienting, and for detecting the degree of sound novelty, which affects its degree of distractibility. Specifically, the degree to which unattended novel sounds distracted visual forced-choice task performance coincided with the extent that the posterior N1 response was released from inhibition. On the other hand, the anterior N1 response was associated with the processing of fine object features (Jääskeläinen et al. 2004). Similarly, the results in our study indicate that the deviant word was released from inhibition in the “control” condition because of its lexical status and that this “word advantage effect” was located anterior to HG. This is in agreement with the finding that neurons in the anterior auditory cortex are more sharply tuned to phonetic features of sounds (Ahveninen et al. 2006). The effect of speech perception Despite the general similarity to the results obtained in the context of tones, some of the results are specific to speech perception, as follows. In the “oddball” condition the brain regions activated in response to the deviants “day” and “tay” were larger than those observed for the “deh” deviant (Deviant maps, Fig. 1). In addition, in the control condition the precentral gyrus was activated in response to both “day” and “tay” (Fig. 2). These similarities between “day” and “tay” may reflect an extra processing effort associated with the diphthong vowel /eI/ shared by “day” and “tay” (Sonty et al. 2003; Bohland and Guenther 2006). However, the fact that the similarity between the responses to “day” and “tay” diminished in the control condition (Fig. 2) may indicate that the /eI/ diphthong associated with deviant “day” was more salient among repetitive “deh” standards (“oddball” condition) than among the variable filler stimuli (“control” condition) (Nordby et al. 1994; Sabri and Campbell 2000). The more similar responses between “day” and “deh” in the control condition (Fig. 2) may be explained by the shared /e:/ monophthong that might have elicited an expectation to hear a meaningful word (“day”) while presented with “deh”. This expectation could account for the more extensive response in the STG relative to “tay” (Fig. 2). The finding that activation in the left anterior auditory cortex dissociates the “oddball” and “control” conditions strengthens this interpretation (Scott et al. 2006) as well as the observed increase in the percent signal change in that region associated with both “deh” and “day” (Fig. 5, “Left” panels). Thus, in the “control” condition (Fig. 2), an interaction between lexical and acoustic features affected the results. There were additional findings indicating left hemisphere dominance, as follows. The size of STG activation was the most extensive over the left STG for “day” in comparison to “tay” where it did not reach statistical significance and “deh” where it was more than five times smaller (Deviant control maps, Fig. 2, Table 5). Furthermore, the “day” control map revealed negative differential BOLD in the left superior frontal gyrus as well as in the left occipital pole (Fig. 2, Table 5). Left hemisphere dominance was also evident in the middle/superior frontal gyrus in the sensory mechanism map for “deh” (Fig. 4, Table 7). In addition, the four-way ANOVA (Fig. 5) has shown. This pattern of left hemisphere dominance is in agreement with a number of imaging and clinical studies (for a review see Tervaniemi and Hugdahl 2003). The left middle/superior frontal gyrus activations are in line with previous findings implicating these brain regions with tasks engaging phonological working memory (Paulesu et al. 1993; Burton et al. 2000; Siok et al. 2003; LoCasto et al. 2004) and with the suggestion that a significant portion of active frontal areas is recruited for extracting acoustic information and maintaining it in memory (LoCasto et al. 2004). The negative differential activation evident in the left occipital pole associated with the “day” control map (Fig. 2) may indicate that meaningful words are more likely to elicit activity in visual processing regions (Billingsley-Marshall et al. 2007). It is noteworthy that the sensory mechanism map for “deh” revealed a positive differential BOLD in the ACC (Fig. 4, Table 7). The ACC is implicated in initiating or inhibiting action and is considered to be part of a larger network that includes medial/lateral frontal, prefrontal and temporal regions (Wang et al. 2005; Dias et al. 2006; Gold et al. 2006). Hence, it is possible that in our study inhibitory activations occurred in the ACC (as well as in the superior frontal gyrus) in response to “day” since it was a meaningful stimulus that interfered with the main counting task (see also Rinne et al. 2005). Since standard “deh” was contrasted with deviant “day” to create the “deh” standard map, the positive differential BOLD located at the ACC and middle/superior frontal gyrus (Fig. 4, “deh”) might have been caused by a reduced activation associated with the response to deviant “day”. A possible effect of the lexical status in the “oddball condition It appears that the lexical status of the deviant stimulus affected its processing not only in the “control” condition but also in the “oddball” condition in which the effects of adaptation caused by the repeating standards were more pronounced. Specifically, the “day” stimulus activated parts of the posterior STG (i.e., BA 22/42), whereas the “tay” and “deh” stimuli activated large parts of the superior and middle temporal cortices (i.e., BA 21/22) (Table 4). However, the size of the left STG region activated by “day” deviant was almost twofold larger than that activated by “tay” deviant (Deviant maps, Fig. 2, Table 4). This is in agreement with the assumption that the left posterior STG is the focus of a multi-modal network associated with language comprehension (Aboitiz and Garcia 1997; Narain et al. 2003). More strongly left-lateralized posterior superior-temporal activation is associated with analysis of speech sounds for mapping onto higher levels of language processing (e.g., syllable, word) (Price et al. 1992; Zatorre et al. 1996). Thus, although the word advantage effect was not salient in the “oddball” condition, the different spread of activation between “day” and the other non-words may indicate differential processing based on the lexical status of the deviant stimulus. Summary Taken together, the results of the current study corroborate the existence of two independent mechanisms contributing to the change-detection response (Opitz et al. 2005; Hoshiyama et al. 2007; Maess et al. 2007): a sensory mechanism reflected by different refractory states of those subpopulations activated by the standard and the deviant and a cognitive mechanism which relies on auditory sensory memory representations which gives rise to the word advantage effect. Thus, on the one hand, our results support the view that the MMN represents a change-detection mechanism functionally and spatially distinct from an afferent input population (N1 generators) (Näätänen et al. 2005). On the other hand, our results indicate that a release from the inhibitory effects of adaptation is a prerequisite for the full realization of the significance of the deviant stimulus. This is in line with the adaptation hypothesis that assumes that the posterior auditory cortex gates novel sounds to awareness (Jääskeläinen et al. 2004). Conclusion Our findings serve to unify the two opposing views suggested by Jääskeläinen et al. (2004) and Näätänen et al. (2005). Specifically, the gate to awareness for auditory deviation (Jääskeläinen et al. 2004; Näätänen et al. 2005) relies on adaptation that modulates the extent to which novel sounds will be accessible to memory-based processes. In case of the present study, the sensory component serves to modulate the salience of the speech stimulus by the degree to which it will be accessible to cortical memory traces for speech sounds (Pulvermüller and Shtyrov 2006). Furthermore, the lexical status of the speech stimulus interacts with acoustic factors exerting a top-down effect on the novelty value of the auditory object that affects, in turn, its degree of accessibility to the cognitive component.

J_Neurooncol-4-1-2174520

Lovastatin sensitized human glioblastoma cells to TRAIL-induced apoptosis

Synergy study with chemotherapeutic agents is a common in vitro strategy in the search for effective cancer therapy. For non-chemotherapeutic agents, efficacious synergistic effects are uncommon. Here, we have examined two non-chemotherapeutic agents for synergistic effects: lovastatin and Tumor Necrosis Factor (TNF)-related apoptosis-inducing ligand (TRAIL) for synergistic effects; on three human malignant glioblastoma cell lines, M059K, M59J, and A172. Cells treated with lovastatin plus TRAIL for 48 h showed 50% apoptotic cell death, whereas TRAIL alone (1,000 ng/ml) did not, suggesting that lovastatin sensitized the glioblastoma cells to TRAIL attack. Cell cycle analysis indicated that lovastatin increased G0–G1 arrest in these cells. Annexin V study demonstrated that apoptosis was the predominant mode of cell death. We conclude that the combination of lovastatin and TRAIL enhances apoptosis synergistically. Moreover, lovastatin sensitized glioblastoma cells to TRAIL, suggesting a new strategy to treat glioblastoma. Introduction Glioblastomas are the most common intracranial brain tumors. Its prognosis is usually poor, with survival times of less than 15 months from first diagnosis [1]. Surgical resection and chemotherapy are common treatments [2]. Despite recent advances in the understanding of the molecular mechanism of tumourogenesis, the outcome of malignant glioma remains poor [3]. Thus, new effective forms of therapy are needed. The Tumor Necrosis Factor-Related Apoptosis-Inducing Ligand (TRAIL) [4], a member of the TNF superfamily, can bind with death receptors, DR4 and DR5 [5, 6] and induces apoptosis in a wide range of cancer cells without harming normal cells. The specific property of TRAIL has attracted many researchers to look for new treatments by combining it with chemotherapeutic agents such as phenoxazine derivatives [7], doxorubicin and cisplatin [8]. Such combinations have shown synergistic effects on different types of cancer cells in vitro. Lovastatin, a 3-hydroxy-3-methlyglutaryl CoA (HMG CoA) reductase inhibitor is a commonly used cholesterol-lowering agent for prevention of atherosclerotic cardiovascular diseases [9, 10]. Lovastatin blocks the mevalonate pathway and reduces the formation of the downstream products, cholesterol, geranylgeranyl proteins, farnesylated [11]. Recently both in vitro and in vitro studies have found that lovastatin has antiproliferative, proapoptotic and anti-invasive properties in a wide range of cancer cell types [12]. Lovastatin is known to have an apoptotic effect on tumor cells and its combination with chemotherapeutics and cytokines often exert a synergistic effect against tumor growth [13–15]. The mechanism that leads to lovastatin-induced apoptosis is not yet clear but the main event is thought to be associated with the alteration of mitochondrial stress, which releases cytochrome C, activates pro-caspase cascade and finally leads to apoptotic cell death. Escape from apoptotic regulation is one of the major characteristics of cancer [16, 17], and many successful anti-cancer agents induce apoptosis by damaging DNA. Unfortunately such agents may also severely affect normal cells. Given the fact that both lovastatin and TRAIL are non-chemotherapeutic agents and capable of inducing apoptosis in different types of cancer cells, it is important to determine whether the combination of these two agents would produce synergistic effects that may be lighten for a novel therapeutic application in gliomas. We therefore hypothesized that the combination of TRAIL and lovastatin, neither of which alone has noxious effects on healthy cells, could generate a regime that was effective in killing cancer cells but caused minimal insult to normal healthy cells. In this study we report the effects of TRAIL in combination with a non-chemotherapeutic drug, lovastatin, on glioblastoma cells. Materials and methods Reagents 2-Methyl-1,2,3,7,8,8a-hexahdro-3,7-dimethyl-8-[2-(tetrahydro-4-hydroxy-6-oxo-2H-pyran-2-yl)ethyl]-1-naphthalenyl ester butanoic acid (Lovastatin), DL-Mevalonic acid lactone, and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) were purchased from Sigma (St Louis, MO). Lovastatin was dissolved in DMSO for stock and adjusted to final concentrations using complete medium or serum free medium. Soluble Human TRAIL (Apo2L) was affinity purified from lysates of bacteria transformed with pET plasmid containing TRAIL [18]. Cellular DNA fragmentation ELISA kit (Roche, Mannheim, Germany), RNeasy kit, DNA extraction kit (Qiagen, Germany) and RT-PCR kit (Promega, Madison, WI) were used. Three primary antibodies used were as follows: rabbit polyclonal antibody to DR4 (Chemicon International, 1:1,000 dilution), rabbit polyclonal antibody to DR5 (Cell Signaling Technology, 1:1,000 dilution), and rabbit polyclonal antibody to β-tubulin (Santa Cruz Biotechnology, 1:1,000 dilution). Goat anti-rabbit secondary antibody was obtained from Santa Cruz Biotechnology. Cell culture Three human glioblastoma cell lines, A172, M059J, and M059K were purchased from American Type Culture Collection (Rockville, MD). The glioblastoma cells were kept in Dulbecco’s modified Eagle’s medium (DMEM/F12) (GIBCO BRL, Grand Island, NY) with or without 10% fetal bovine serum, 1% penicillin and streptomycin at 37°C under 5% CO2. Media were changed every 3 days. Measurement of cell viability We measured cell viability using the MTT assay. MTT is a water-soluble tetrazolium salt that is metabolized by viable cells to a colored, water-insoluble formazan salt. Thus the salt allows cell viability measurements. In short, 1 × 104 cells were cultured in the serum free DMEM/F12 medium in the presence of lovastatin with or without TRAIL in a 96 well microtiter plate for designated time periods. The medium was aspirated and 100 μl MTT (0.5 mg/ml in PBS) were added to each well and the cells incubated for 3 h. After MTT medium was aspirated, the cells were solubilized in 200 μl DMSO. The optical density of each sample at 570 nm (reference 630 nm) was measured using a microplate reader. The optical density of the media was proportional to the degree of viable cells. Cell cycle analysis Propidium iodide (PI) staining and flow cytometry were used to determine the stage of the cell cycle. Cells were treated with 20 μM lovastatin for 48 h at 37°C and the control cells were treated with normal medium (DMSO) or serum free medium (DMSO). 2 × 106 treated cells were washed with 5 ml PBS and then were trypsinized at 37°C for 5–10 min. Cells were spun down and washed with 5 ml PBS. Finally cells were resuspended in 500 μl PBS and fixed with 4.5 ml 70% Ethanol with gentle vortexing. Cells were allowed to sit in −20°C for overnight. Fixed cells were spun down and washed with 5 ml PBS, and then cells were resuspended in 500 μl PI (2 μg/ml)/Triton X-100 (0.1% v/v) staining solution with RNase A (200 μg/ml) in dark and analyzed by a flow cytometer. The staining solution was purchased from Chemicon (Temecula, CA). Apoptosis assay Apoptotic cells were determined by two methods, Annexin-V and PI stained cells by flow cytometry and DNA fragmentation assay. During apoptosis, translocation of phosphatidylserine from inner membrane to outer membrane is a common phenomenon. Cells were stained with Annexin V for analysis of phosphoserine inversion, which was considered to be a sensitive marker of apoptosis. Using an Annexin V-FITC apoptosis detection kit (Molecular Probe Inc, Eugene, OR), the levels of binding of Annexin V and staining with PI were measured for the detection of early and late apoptosis respectively. All of the procedures were preformed under manufacturer’s guidelines. Cells were treated with lovastatin or/and TRAIL for 48 h and then stained with Annexin-V and PI. Viable cells were recognized as negative for both Annexin-V and PI; early apoptotic events were positive for Annexin-V but negative for PI staining. Late apoptotic events were positive to both Annexin V and PI. Necrotic cells were positive for PI staining only. DNA fragmentation determination, was carried out with the ELISA assay kit. After cells (1 × 104) were treated with lovastatin and/or TRAIL for 48 h, DNA fragmentation was detected using 96 wells microplate reader. RT-PCR Total RNA was isolated using Qiagen RNeasy extraction kit and performed according to the manufacturer’s protocol. Total RNA (5 μg) was reversely transcribed using Promega RT-PCR kit and thermal program was set at 42°C for 15 min and 95°C for 5 min. PCR reaction was performed using the following primers, which have previously been tested successfully: TRAIL-R1, 5′-CTG AGC AAC GCA GAC TCG CTG TCC AC-3′ and 5′-TCC AAG GAC ACG GCA GAG CCT GTG CCA T-3′; TRAIL-R2, 5′-GCC TCA TGG ACA ATG AGA TAA AGG TGG CT-3′ and 5′-CCA AAT CTC AAA GTA CGC ACA AAC GG-3′; TRAIL-R3, 5′-GAA GAA TTT GGT GCC AAT GCC ACT G-3′ and 5′-CTC TTG GAC TTG GCT GGG AGA TGT G-3′; TRAIL-R4, 5′-CTT TTC CGG CGG CGT TCA TGT CCT TC-3′ and 5′-GTT TCT TCC AGG CTG CTT CCC TTT GTA G-3′. The thermal program was set up as one cycle at 94°C for 5 min, 30 cycles at 94°C for 1 min, 55°C for 1 min, 72°C for 2 min, and one cycle at 72°C for 5 min. PCR products were resolved and visualized on a 2% agarose gel stained with ethidium bromide. Western blot analysis Western blot was performed according to previous description [19, 20]. Briefly, after 2 × 106 cells were treated for 48 h, total protein was isolated and reacted with the relevant antibodies. The probed proteins were visualized using the enhanced chemiluminescence Western blotting detection system (ECL Western Blotting Detection, Amersham Biosciences). Statistics The statistical significance was analyzed using one-way ANOVA analysis and Student’s t-test. All statistical work was carried out using the SPSS software for Windows (Release 11.0.1, Chicago, IL). Differences were considered to be significant when P < 0.05. Results Cell viability measured by MTT assay We first examined the anti-proliferation effect of TRAIL on three glioblastoma cell lines. M059K and M059J glioblastoma cells were resistant to TRAIL-induced cell death and remained 100% cell viable following treatment in both normal and serum free conditions (Fig. 1a, b). Only A172 showed minor cell death after TRAIL treatment (22%, Fig. 1a). Fig. 1Anti-proliferation effects on glioblastoma cell lines A172, M059K, and M059J. MTT assays were preformed after 48 h incubation time treated with 0, 500, 1,000 ng/ml TRAIL in normal medium (a) or serum free medium (b). A172 subjected to minor TRAIL-induced cell death while M059J and M059K cells were resistant to TRAIL. Same trend of data were obtained in both normal medium and serum free condition. Experiment set were repeated at least three times with triplicate wells for each condition (mean ± SD) Synergistic effects by TRAIL and lovastatin were then examined. A172, M059J, and M059K glioblastoma cells were incubated in 1, 5, 20, and 40 μM lovastatin alone or together with 500 ng/ml TRAIL for 48 h. In A172 glioblastoma cells, 5, 20, and 40 μM lovastatin with 500 ng/ml TRAIL promoted significant cell death when compared with lovastatin control (Fig. 2a). 5, 20 and 40 μM lovastatin alone induced 50%, 60% and 76% of cell death respectively, however in the presence of 500 ng/ml TRAIL, lovastatin induced 78%, 94%, and 92% of cell death respectively. 1, 5, 20, and 40 μM lovastatin with 500 ng/ml TRAIL also synergistically promoted cell death in M059J and M059K. In M059J cells, 1, 5, 20 and 40 μM lovastatin alone caused 40%, 55%, 40% and 60% cell death whereas in the presence of 500 ng/ml TRAIL, lovastatin induced 83%, 94%, 95% and 95% cell death respectively (Fig. 2b). There were 26%, 51%, 58%, and 71% cell death induced by 1, 5, 20, and 40 μM lovastatin alone in M059K, in the presence of 500 ng/ml TRAIL, lovastatin at all concentration tested caused about 98% of cell death (Fig. 2c). Collectively the combination of TRAIL and lovastatin was much more effective than lovastatin alone in the induction of cell death in all three glioblastoma cells tested and the result indicated that there was a synergistic effect when TRAIL and lovastatin were used together. Fig. 2Synergistic anti-proliferation effects on glioblastoma cell lines A172, M059J, and M059K. Glioblastoma cells were treated with 0, 5, 20 and 40 μM lovastatin alone or 0 μM lovastatin plus 500 ng/ml TRAIL, 5 μM lovastatin plus 500 ng/ml TRAIL, 20 μM lovastatin plus 500 ng/ml TRAIL and 40 μM lovastatin plus 500 ng/ml TRAIL for 48 h. Percentage of viable glioblastoma cells showed synergistic anti-proliferation effects by combined the two agents. In A172 glioblastoma cells (a), 500 ng/ml TRAIL with 5, 20 and 40 μM lovastatin induced cell death significantly when compared with lovastatin only groups. In M059J (b) and M059K (c), 500 ng/ml TRAIL with 1, 5, 20 and 40 μM lovastatin induced cell death significantly when compared with lovastatin only groups. ANOVA were used for statistics analysis and *P < 0.05, **P < 0.01. Experimental set were repeated for at least three times with triplicate wells for each condition (mean ± SD) Cell cycle determination by PI staining Propidium iodide staining and flow cytometry were used to determine the degree of cell synchronization by 20 μM lovastatin. Cells were incubated with normal serum medium, serum free medium and 20 μM lovastatin with normal serum medium for 48 h. PI-stained cells were analyzed using flow cytometry to quantify cells in certain cell cycle stages. A significant increase in cell population at G0/G1 phase was observed when the cells were treated with 20 μM lovastatin indicating that lovastatin was able to arrest the cells at G0/G1 stage. G0/G1 cell population was also increased by serum free medium conditions (positive control) and the increase reached a significant level except in M059J cells (Fig. 3). Fig. 3Lovastatin-induced glioblastoma cells arrested in G0–G1 Phase. Propidium Iodide staining for cell cycle analysis were performed after glioblastoma cells were treated with normal medium without lovastatin, serum free medium without lovastatin and 20 μM lovastatin for 48 h. Serum free condition was used as a positive control which is commonly known to induce cell arrest in G0–G1 phase. Serum free condition and lovastatin increased G0–G1 cell arrest in all glioblastoma cells and reached significant level (except serum free condition in M059J). ANOVA were used for statistics analysis and *P < 0.05, **P < 0.01. Experimental set were repeated for at least three times with triplicate wells for each condition (mean ± SD) Apoptosis is the major mode of cell death In order to determine the mode of cell death induced by lovastatin and TRAIL, Annexin V and PI staining was employed for this purpose. Cells were incubated with 5 or 20 μM of lovastatin in the presence or absence of 500 ng/ml TRAIL for 48 h. The percentage of cell death of glioblastoma cells treated with both lovastatin and TRAIL was significantly higher than that with either agent alone (Fig. 4a, b and c). A172 glioblastoma cells were vulnerable to TRAIL induced apoptosis (Fig. 4a) but significantly more apoptotic cells were observed following treatment with both TRAIL and lovastatin (P < 0.005) (Fig. 4a). TRAIL was able to induce apoptosis in 50% of A172 glioblastoma cells. However, TRAIL in combination with 5 and 20 μM could induce apoptosis in approximately 67% and 74% cells respectively, indicating a synergistic effect occurred. Similarly, a significant synergistic effect was also observed in M059K and M059J cells (P < 0.005) (Fig. 4b, c), with up to nearly a 10-fold increase in apoptotic cells. 5 μM lovastatin only induced apoptosis in 5.3% and 2.3% M059J and M059K respectively, which were not different from the control (without any treatment). 5 μM lovastatin plus 500 ng/ml TRAIL dramatically increased apoptotic cells to 47.8% and 61.4% in M059J and M059K respectively. M059J cells, which lack of DNA-dependent protein kinase expression [21], were less vulnerable to the two agents, implying that DNA-dependent protein kinase may play a role in apoptosis induced by TRAIL and lovastatin. Fig. 4The synergistic apoptotic effects were quantified by flow cytometry using Annexin V and PI staining. Glioblastoma cells were treated with DMSO (Control), 5 μM lovastatin, 20 μM lovastatin, 500 ng/ml TRAIL, 5 μM lovastatin plus 500 ng/ml TRAIL and 20 μM lovastatin plus 500 ng/ml TRAIL for 48 h. Then cells were stained with Annexin V and PI to determine percentage of apoptotic cell death using flow cytometry. Synergistic apoptotic effects were observed in three glioblastoma cell lines A172 (a), M059J (b) and M059K (c), reached significant level. The portion of apoptotic cell death was indicated in low-right quarter of the flow-cytometry scatter plot and the trend of apoptotic cells between groups. ANOVA were used for statistics analysis and *P < 0.05, **P < 0.01 (compared to control group), +P < 0.05, ++P < 0.01 (compared to TRAIL only group). Experimental set were repeated for at least three times (mean ± SD) DNA fragmentation in glioblastoma cells DNA fragmentation is one of the hallmarks when cells undergo apoptosis. We performed ELISA assay to confirm DNA fragmentation was induced in the cells treated with both lovastatin and TRAIL. A synergistic effect on DNA fragmentation by both agents was demonstrated in all three cell lines tested (Fig. 5a, b and c). The level of DNA fragmentation was increased in treated A172 glioblastoma cells, but it was only statistically significant in the cells treated with TRAIL plus 20 μM lovastatin. In M059J and M059K cells, however, there was a dramatic elevation of DNA fragmentation in a dose-dependent manner when both of the agents were applied. This observation supports the argument that the cell death induced by the two agents was apoptotic and that both agents were able to function in a synergistic manner. Fig. 5DNA fragmentation was detected in combination of lovastatin and TRAIL. Same treatment in apoptotic cell staining was preformed for DNA fragmentation ELISA detection. Significant level of DNA fragmentation was detected in all glioblastoma cells when combined with lovastatin and TRAIL. ANOVA were used for statistics analysis and *P < 0.05, **P < 0.01 (compared to control group), +P < 0.05, ++P < 0.01 (compared to TRAIL only group). Experimental set were repeated for at least three times (mean ± SD) The expression of TRAIL receptors in glioblastoma cells Resistance to cell death induced by TRAIL may be adapted by an altered level of TRAIL receptors. TRAIL-R3 (DcR1) and TRAIL-R4 (DcR2) are known to attenuate TRAIL-induced apoptosis whereas TRAIL-R1 (DR4) and TRAIL-R2 (DR5) promote TRAIL-induced apoptosis [5]. The effect of lovastatin treatment on the TRAIL receptors was not determined. RT-PCR was performed to investigate the expression profiles on three glioblastoma cell lines tested. Cells were treated with normal serum medium, serum free medium, serum medium with 5 or 20 μM lovastatin for 48 h. TRAIL-R3 and R4 were not detected in all conditions and cell types tested. TRAIL-R1 was only detected on M059J cells with serum free medium or lovastatin (Fig. 6a). TRAIL-R2 was expressed on all cell types of cells with or without lovastatin and its level was not significantly different between lovastatin-treated cells and the controls. Therefore, the expression of TRAIL-R2 did not contribute to the lovastatin-induced cell death in glioblastoma cells tested. Fig. 6The expression of TRAIL receptor mRNA in glioblastoma cells. The cells were treated with normal serum medium, serum free medium, 5 and 20 μM lovastatin with normal serum medium. At the end of the treatment, RNA was isolated for the detection of TRAIL-R1 (DR4) (a) and TRAIL-R2 (DR5) (b) by RT-PCR After we tested the mRNA expression of TRAIL receptors on glioblastoma cells, the further investigation was performed to determine the active protein expression of TRAIL-R1 and matured TRAIL-R2 by Western blot. The result showed that the protein expression profiles of TRAIL-R1 and TRAIL-R2 were found in three glioblastoma cells with or without lovastatin, suggesting the protein of these two receptors was underwent post-translational modifications. The significantly differences were only found on the TRAIL-R1 expression of M059K cells with 20 μM lovastatin (P = 0.05) and the expression of TRAIL-R2 in A172 cells with lovastatin (P < 0.01) (Fig. 7). Fig. 7The expression of TRAIL-R1 (DR4) and TRAIL-R2 (DR5) proteins in glioblastoma cells. The cells were treated with normal serum medium, serum free medium, 5 and 20 μM lovastatin with normal serum medium for 48 h. At the end of the treatment, proteins were isolated for the detection of TRAIL-R1 (a, b) and TRAIL-R2 (c, d) by Western blot. Representative Western blots were shown (a, c). The target bands were scanned and normalized to β-tubulin. The index of densities was calculated (b, d). *P = 0.05, **P < 0.01 Discussion Malignant glioblastoma is one of the major causes of brain tumors morbidity. Aggressive infiltration in the CNS ultimately leads to death in nearly all cases [2]. Malignant glioblastoma carries with aberrant biological and biochemical properties including several activating mutations that can lead to chemotherapeutics resistance [22]. Targeting apoptotic signaling machinery is thought to be a promising alternative for glioblastoma treatment [23]. The preliminary data presented in this report indicate that lovastatin, a blood cholesterol lowering medicine, sensitizes glioblastoma cells to TRAIL-mediated apoptosis. Earlier reports show statin-induced cell death through a mitochondrial-mediate pathway (intrinsic pathway) that is closely related to the Bcl-2 family protein Bid and activation of caspase 8, 9 and 3 [24–27]. However, the mechanism for statin activation of caspase 8 remains unknown because caspase 8 is normally activated by receptor-mediated signals, such as Fas ligand and TRAIL [5]. Lovastatin has been found to enhance TRAIL-induced cytotoxicity in a synergistic manner in glioblastoma cells. We demonstrated a synergistic effect produced by the combination of lovastatin and TRAIL on glioblastoma cells. Lovastatin was found to sensitize the cells to cell death induced by TRAIL. The mode of cell death induced by both agents in combination was apoptosis, as demonstrated by two different methods, Annexin V and PI staining and DNA fragmentation assay. We also demonstrated that two of the glioblastoma cell lines tested were resistant to TRAIL induced apoptosis. From this we inferred that lovastatin not only sensitized these glioblastoma cells through its effects on the TRAIL receptor pathway but also triggered an unknown mechanism: Lovastatin served as a cytostatic agent and turned on an unknown mechanism to support TRAIL-induced apoptosis in these glioblastoma cells. It has been shown that colon and lung tumor cells arrested in G0–G1 stages are vulnerable to TRAIL-induced cell death [28]. However, our G0–G1 arrested glioblastoma cell lines remained resistant to TRAIL-induced cell death in the serum free control. Our finding suggests that a combination of TRAIL and lovastatin together may form a new treatment for glioblastoma multiforme. The mechanism by which lovastatin sensitizes glioblastoma cells to TRAIL induced-apoptosis remains unknown. In human glioblastoma, lovastatin has been shown to induce or to enhance apoptosis by altering a number of apoptotic molecules. For example, it can induce apoptosis or downregulate cell proliferation by targeting Ras in primary cultured human glioblastoma cells [29] and increase pro-apoptotic Bim in U87 and U251 glioblastoma cells [30]. Lovastatin has also been shown to downregulate RhoA and increase iNOS in T98G and A172 glioblastoma cells [31]. Additionally, Lovastatin may induce apoptosis by increasing p21 and the apoptosis induced can be prevented by the overexpression of Bcl-2 [32], suggesting a mitochondrial-related apoptosis. It is noted that this study by Schmidt et al. fails to document that lovastatin can enhance death receptor (CD95)-mediated apoptosis in glioblastoma cells LN-18, LN-229, LN-308 and T98G [32]. The result from our study indicate that lovastatin in combination with TRAIL can synergistically induce apoptosis in A172, M059K and M059J glioblastoma cells. The apoptosis induced is associated with G0–G1 arrest but not with pro-apoptotic Bid (data not shown). It is well known that TRAIL-induced apoptosis takes place via a death receptor-mediated pathway [5, 33]. Our study showed that TRAIL-R1 mRNA expressed in M059J but was hardly detected in the other two cell lines. However, TRAIL-R1 protein was detectable in all three cell lines, suggesting that the TRAIL-R1 was modified by a post-translational mechanism in the cells tested. The level of TRAIL-R1 protein was higher in M059K cells treated with 20 μM of lovastatin. However, such an elevation of TRAIL-R1 is unlikely to be responsible for apoptosis induced by lovastatin because the apoptotic rate between M059K was not different from the other cell lines. TRAIL-R2 can be detected at both protein and RNA levels. The level of TRAIL-2 mRNA was not different in the cells treated with lovastatin but its protein level was much higher in A178 cells treated with either 5 or 10 mM lovastatin, the result of which indicated that TRAIL-R2 was modified by a post-translational mechanism in A172 cells. It is noted that the base level of TRAIL-2 in A172 cells does not differ from other two cell lines tested. Therefore, it is unlikely that TRAIL-2 could count for the relatively higher sensitivity of A172 cells to TRAIL stimulation than the other two cell lines. All three cell lines underwent a similar level of cell death when they were treated by lovastatin. Therefore, the contribution of lovastatin-induced TRAIL-2 protein to the cell death in A172 cells seems to be minimal. Such results appear to be in line with a study using mevastatin, a similar HMG-CoA reductase inhibitor. Mevastatin can significantly induce apoptosis of myeloma cells in a pathway independent of death receptors including TRAIL-R2 [34]. Nevertheless, further quantitative tests are needed in glioblastoma cells to verify the result obtained. Apoptosis induced by lovastatin is generally considered to be via the mitochondrial-mediated pathway [30; 32]. It is possible for both pathways to talk to each other to amplify apoptotic signals and this is indeed a case for TRAIL, since TRAIL is capable of inducing either mitochondrial-independent or -dependent apoptotic pathways in some types of cells [17, 33]. The Bid is a molecule functions as a bridge that links death receptor- and mitochondrial-mediated pathways. Therefore, without involvement of Bid, apoptosis induced by TRAIL and lovastatin in combination in the present study seem to be separately mediated by these two pathways. However, considering the fact that two out of three glioblastoma cells tested are insensitive to TRAIL treatment but they become responsive in the presence of lovastatin, it can be hypothesized that lovastatin treatment may remove an unknown inhibitory factor(s) that overcomes the TRAIL-mediated pathway or that lovastatin may “wake up” an activator(s) that normally remains in a resting condition. One of inhibitory factors known to involve TRAIL-mediated pathways is its decoys, TRAIL-R3 and TRAIL-R4 [5, 33]. However, it can be seem from our study that this inhibitory factor(s), if any, should not be the decoys TRAIL-R3 and TRAIL-R4 since both are not detectable in the cells tested. One possible explanation for this synergistic effect is the activation of caspase 8, an initiator caspase in death receptor-mediated pathways. Lovastatin has been shown to enhance caspase 8 activity [35]. It is possible that caspase-8 is the activator factors waken up by lovastatin. Whatever it is, the mechanism accounting for the synergistic effect of TRAIL and lovastatin against glioblastoma cells in the present study is complicated and remains to be uncovered. Mevalonate is a critical component of a complex biochemical pathway and its products are vital for a variety of important cellular functions including cell signaling, protein synthesis, and cell cycle regulation [11]. Little is known about the molecular events leading to apoptosis of cancer cells due to lovastatin exposure. It is likely that apoptosis is abrogated by mevalonate and GGPP and is partially reversed by FPP [12]. Our results also reinforced the conclusion of our previous study, that DNA-dependent protein kinase (DNA-PK) plays an important role in cell apoptosis. M059J cells that lack of DNA-PK activity are resistant not only to total cell death but also apoptosis [19, 20]. Experts disagree over the functions of DNA-PK in cell regulation. DNA-PK has been reported to promote cell death by interacting with telomeres, whereas other reports suggest that DNA-PK protects cells from cell death via caspase-independent or p53 independent pathways. The reason for this paradoxical finding remains unknown. This study demonstrated a synergistic interaction between lovastatin and TRAIL, but the mechanisms of action by which lovastatin sensitized glioblastoma cells remains unknown. Our results are in agreement with the concept of combined cancer therapeutic action via both intrinsic and extrinsic apoptotic cell death pathways. This combination of non-chemotherapeutic agents, TRAIL and lovastatin, may offer a potential regime for glioblastoma treatment.

Mol_Biochem_Parasitol-1-5-1964783

TbARF1 influences lysosomal function but not endocytosis in procyclic stage Trypanosoma brucei

The ADP ribosylation factors (Arfs) are a highly conserved subfamily of the Ras small GTPases with crucial roles in vesicle budding and membrane trafficking. Unlike in other eukaryotes, the orthologue of Arf1 in the host bloodstream form of Trypanosoma brucei is essential for the maintenance of endocytosis. In contrast, as shown in this study, knockdown of TbARF1 by RNA interference has no effect on fluid-phase endocytosis in the insect stage of the parasite. The protein remains essential for the viability of these procyclic cells but the major effect of TbARF1-depletion is enlargement of the lysosome. Our data indicate that protein trafficking and lysosomal function are differentially regulated by multiple factors, including TbARF1, during progression through the T. brucei lifecycle. 1 Introduction The secretory system plays a critical role in the efficient sorting and transport of a range of products to the surface of all eukaryotic cells. The small GTPase Arf1 is a vital component of this system, localizing to the Golgi apparatus in yeast and mammalian systems where it has highly conserved functions in vesicular assembly and the activation of phospholipid-modifying enzymes [1]. An exception is the orthologue of Arf1 (TbARF1) that has been recently characterized functionally in the bloodstream form (BSF) of the kinetoplastid parasite Trypanosoma brucei [2], a cell type which supports extremely rapid rates of internalisation and recycling of its major outer membrane protein, variant surface glycoprotein (VSG; [3]). Although TbARF1 localizes at or proximal to the Golgi apparatus in these parasites, the primary effect of TbARF1 knockdown is severe inhibition of the endocytic system, leading to cell death. In contrast, overexpression of a constitutively active GTP-bound form of TbARF1 inhibits protein trafficking from the Golgi apparatus to the lysosome [2]. Here we show that the vital role of TbARF1 in endocytosis is not conserved throughout the T. brucei lifecycle. In the more sedentary insect procyclic form (PCF) of the parasite, ARF1 is required for cell viability but depletion of the protein by RNA interference (RNAi) has no effect on the uptake of dextran by fluid-phase endocytosis. Instead, the lysosome becomes enlarged, although degradation of the protein p67 within this organelle is not significantly impaired. 2 Materials and methods 2.1 Disruption of TbARF1 expression by RNAi The plasmid p2T7ARF1 [2] contains a non-conserved region of the T. brucei open reading frame (residues 101–225) between two opposing T7 promoters under the control of tetracycline repressors. Mid-log phase parasites of T. brucei procyclic cell line 29–13 [4] were electroporated with 10 μg of NotI-digested p2T7ARF1 using methods described previously [5] to produce the stable cell line 29-13/p2T7ARF1. Expression of ARF1-specific dsRNA was induced by incubating parasites in 100 ng/ml tetracycline. Expression of TbARF1 was monitored by quantitative PCR, using SYBR Green Mastermix (Applied Biosystems). Total RNA was extracted from parasites using TRIzol Reagent (Invitrogen), treated with DNase I (Ambion) and reverse-transcribed using Omniscript RT (Qiagen). A 66 bp fragment of TbARF1 was amplified using SYBR Green Mastermix (Applied Biosystems) on a Prism7000 (Applied Biosystems) and compared to levels of a constitutively expressed control, α-tubulin. Oligonucleotides for amplification were: TbARF1 RTF1, 5′-GGCTTCCGCTTTCAAATCC-3′, TbARF1 RTR1, 5′-CATCAAGGCCGACCATAAGAA-3′, TbαTub RTF1, 5′-CGTGAGGCTATCTGCATCCA-3′ and TbαTub RTR1, 5′-CCCAGCAGGCGTTACCAA-3′. 2.2 Microscopy Indirect immunofluorescence assays on parasites were performed as described [6]. Primary antibodies were gifts as follows: rabbit anti-TbGRASP from Graham Warren (Department of Cell Biology, Yale University School of Medicine, New Haven, CT, USA), mouse anti-p67 from Jay Bangs (Department of Medical Microbiology and Immunology, Madison, WI, USA). Primary antibodies were detected using Alexa-Fluor 488 or 633-conjugated secondary antibodies (Invitrogen). Samples were visualized by confocal microscopy using a Zeiss LSM 510 meta with a Plan-Apochromat 63x/1.4 Oil DIC I objective lens and images acquired using LSM 510 version 3.2 software (Zeiss). For transmission electron microscopy, cells were sequentially treated in 1% (w/v) glutaraldehyde for 1 h, 1% (w/v) tannic acid for 10 min, 0.5% (w/v) osmium tetroxide for 45 min (all in 100 mM phosphate buffer), then in 1% (w/v) aqueous uranyl acetate for 1 h. After dehydration in an acetone series, cells were embedded in Spurrs resin. Sections were cut on a Leica Ultracut, stained with saturated uranyl acetate in 50% ethanol and Reynolds lead citrate, and viewed with a Tecnai 12 BioTwin (FEI) at 120 kV. Images were acquired with a SIS MegaView III digital camera. 2.3 Trafficking assays To monitor the uptake of dextran, cells were resuspended at 5 × 107/ml in 1 mg/ml Alexa-Fluor 488-conjugated dextran (Invitrogen) in SDM-79 medium and incubated for 20 min at 26 °C. For microscopy, parasites were then fixed in 4% paraformaldehyde and co-stained with DAPI, as described previously [7]. For FACS analysis, cells were fixed as above, then resuspended in PBS at a density of 5 × 107/ml. Samples were then divided into two equal aliquots and anti-488 antibody (1:100, Invitrogen) added to one aliquot for 30 min at RT, before washing in PBS. Fluorescence was measured on a Dako Cytomation CyAn flow cytometer using the FL1 detector and results analyzed with Summit v4.1 software. ER-to-lysosome trafficking and subsequent degradation of the protein p67 was analyzed by metabolic labelling of cells and immunoprecipitation as described previously [8]. 3 Results 3.1 TbARF1 is essential for viability in T. brucei procyclic cells We studied the functions of TbARF1 in PCF cells by knocking down expression using tetracycline-inducible RNAi. Detection of phenotypic effects arising from TbARF1 knockdown in these cells was considerably delayed in comparison to BSF parasites [2] in which cell death was evident by 24 h post-induction. This probably reflects a lower dependence on rapid protein trafficking in the PCF cell, as observed previously [9]. There was no significant decrease in PCF cell division until 90 h post-induction, when the cells started to become rounded and less motile, followed by cell death from 96 h (Fig. 1A, B). Quantitative PCR showed a significant loss of TbARF1-specific RNA by 16 h post-induction (Fig. 1C). Induced cells had no significant differences in nucleus and kinetoplast configurations between uninduced and induced cells over a 120 h time course (data not shown), indicating no major defects in the regulation of cell division. 3.2 Loss of ARF1 has no effect on fluid-phase endocytosis The effects on fluid-phase endocytosis in these TbARF1-depleted procyclic cells were determined by analyzing the uptake of fluorescently labelled dextran. Analysis by microscopy showed internalization of dextran in both uninduced parasites and cells induced with tetracycline for 96 h (data not shown). We also used a novel FACS-based method to distinguish between external binding and internalization of dextran in these cells (Fig. 2A). Cells were incubated in the presence of Alexa Fluor 488-conjugated dextran, before fixing and incubation in anti-488 antibody which quenches the fluorescence associated with externally bound dextran. No significance differences were observed between uninduced and induced cells, either in total dextran fluorescence or internalized dextran alone (i.e. following antibody treatment) (Fig. 2A). 3.3 Loss of ARF1 causes an enlargement of the lysosome Immunofluorescence assays using the lysosomal marker, p67, showed that TbARF1-depleted procyclic parasites have an enlargement of the lysosome (Fig. 2B). Densitometric analysis of 20 cells per experimental group revealed that p67 staining in cells induced for 96 h covered an area approximately three times as large as in uninduced cells. No differences were seen in the size or position of the Golgi apparatus, as indicated by localization of the GRASP protein (Fig. 2B). The main defect visualized by transmission electron microscopy was the presence of a single electron-dense rosette-like structure near the flagellar pocket in a subset (30%) of cells by 72 h post-induction, before detection of a growth defect in these cells. This structure (Fig. 2C–E) is composed of concentric layers of membrane and resembles that of membranous cytoplasmic bodies (MCBs) found in mammalian cells. These are lysosomal in origin and typically found in sphingolipidoses, a group of lysosomal storage diseases characterized by deficiencies in the lysosomal enzymes required for sphingolipid degradation [10]. Similar structures of mitochondrial origin have also been observed in T. brucei following depletion of the glycosomal PEX11-like protein GIM5B [11]. However, we found no obvious differences in mitochondrial structure in procyclic cells following the induction of TbARF1 RNAi, either by electron microscopy or fluorescent staining with Mitotracker (Invitrogen, data not shown). We therefore conclude that the structure detected in Fig. 2 is lysosomal in origin, although proof of this designation requires immuno-electron microscopy with a lysosomal probe. 3.4 Effects of ARF1 depletion on lysosomal trafficking and degradation of p67 The trafficking and degradation of p67 was analyzed by pulse-chase metabolic labelling of cells, followed by immunoprecipitation, as described previously [5,8]. In contrast to this process in BSF parasites, ER- synthesized p67 is not further modified in the Golgi in PCF cells, but proteolytic cleavage in the lysosome results in the generation of four glycosylated fragments: gp75, gp42, gp32 and gp28 [8]. In Fig. 2F, proteolytic fragments resulting from lysosomal degradation accumulated in uninduced and induced cells over the same time course, indicating that both trafficking of p67 and its subsequent degradation were unimpaired in cells depleted of TbARF1 protein (Fig. 2F). 4 Discussion The observations presented here and described previously [2] show that TbARF1 is essential for viability throughout the T. brucei lifecycle. However, the downstream effects of modulating expression of this protein differ between the two major life cycle stages. In the highly active bloodstream form, TbARF1 is required for both receptor-mediated and fluid-phase endocytosis, and depletion of the protein by RNAi causes a severe defect in these mechanisms, rapidly followed by cell death [2]. In contrast, procyclic cells significantly downregulate the expression of proteins associated with receptor-mediated endocytosis, including clathrin [12], and fluid-phase endocytosis is unaffected by the knockdown of TbARF1 expression. Rather than affecting the endocytic system, the absence of TbARF1 causes enlargement of the lysosome, indicating an accumulation of undegraded and/or mistargeted material. An interesting finding is that lysosomal targeting of p67 is not disrupted by knockdown of TbARF1 in either of the two life cycle stages but is inhibited by an increase in activated ARF1 protein in BSF parasites [2]. Lysosomal targeting in trypanosomes is not yet fully understood but must differ from higher eukaryotes in several respects. In mammals, most soluble lysosomal enzymes, including acid hydrolases, are sorted using mannose-6-phosphate signals [13] but there is no evidence to suggest that this mechanism is conserved in lower eukaryotes such as the kinetoplastids [8]. An alternative to the mannose-6 receptor, sortilin, has also been implicated in hydrolase sorting [13] but orthologues of this protein are not encoded by the T. brucei genome. The remaining class of identified signals, either tyrosine or dileucine-based motifs, are able to bind to GGA and AP adapter proteins, which then trigger the packaging of cargo for clathrin-mediated trafficking from the trans-Golgi network (TGN) to the lysosome. Both GGA and AP proteins are directly recruited to the Golgi apparatus by GTP-bound Arf1 [14]. However, genes encoding GGA proteins and AP-2 are absent from the T. brucei genome [2,15,16], implying either a greater dependence on the remaining AP-1 and AP-3 complexes or the operation of additional uncharacterized sorting mechanisms. Much of our current knowledge of lysosomal targeting in T. brucei stems from studies on p67, a protein of unknown function that shares structural similarity but not sequence identity with the mammalian LAMP proteins [17]. This protein is trafficked from the ER via the Golgi to the lysosome, utilizing distinct targeting signals in different stages of the parasite life cycle. In PCF cells, the C-terminal cytoplasmic domain of p67 is necessary and sufficient for lysosomal targeting, whereas this region is not required in BSF cells [8]. While the cytoplasmic domain contains two putative dileucine motifs [8], a recent study has analyzed these regions functionally in PCF and demonstrated that both are required to support maximal targeting [18]. It is likely that p67 is sorted in PCF cells via binding to either AP-1 or AP-3 at the TGN, a process that could be inhibited by knockdown of ARF1. While the data presented here suggest that p67 continues to be targeted efficiently for degradation in the lysosome, this does not preclude a defect in the sorting or trafficking of other essential lysosomal factors. In BSF parasites, the AP complexes may not be directly involved in p67 sorting, given the redundancy of the dileucine motifs in this life cycle stage [8]. As in PCF cells, knockdown of TbARF1 does not prevent p67 localization to the lysosome [2]. However, in contrast, the inducible expression of a constitutively active form of TbARF1 does not inhibit endocytosis but prevents p67 from reaching the lysosome, resulting in extremely rapid cell death [2]. In this situation, excess ARF1 may be sequestering effector proteins required for alternative pathways which would normally facilitate the correct localization of p67. Given that forward biosynthetic trafficking to the lysosome is not quantitatively impaired in the absence of ARF1, this protein may alternatively influence the recycling of membranes from the lysosome, a defect that might generate multivesicular-type structures such as those observed at higher resolution in Fig. 2. Further studies are required to pinpoint the relationship between TbARF1 and lysosomal function, analysis which may reveal insights into the secretory pathways of lower eukaryotes.

Behav_Genet-4-1-2226021

Evaluation of the Serotonergic Genes htr1A, htr1B, htr2A, and slc6A4 in Aggressive Behavior of Golden Retriever Dogs

Aggressive behavior displays a high heritability in our study group of Golden Retriever dogs. Alterations in brain serotonin metabolism have been described in aggressive dogs before. Here, we evaluate whether four genes of the canine serotonergic system, coding for the serotonin receptors 1A, 1B, and 2A, and the serotonin transporter, could play a major role in aggression in Golden Retrievers. We performed mutation screens, linkage analysis, an association study, and a quantitative genetic analysis. There was no systematic difference between the coding DNA sequence of the candidate genes in aggressive and non-aggressive Golden Retrievers. An affecteds-only parametric linkage analysis revealed no strong major locus effect on human-directed aggression related to the candidate genes. An analysis of 41 single nucleotide polymorphisms (SNPs) in the 1 Mb regions flanking the genes in 49 unrelated human-directed aggressive and 49 unrelated non-aggressive dogs did not show association of SNP alleles, genotypes, or haplotypes with aggression at the candidate loci. We completed our analyses with a study of the effect of variation in the candidate genes on a collection of aggression-related phenotypic measures. The effects of the candidate gene haplotypes were estimated using the Restricted Maximum Likelihood method, with the haplotypes included as fixed effects in a linear animal model. We observed no effect of the candidate gene haplotypes on a range of aggression-related phenotypes, thus extending our conclusions to several types of aggressive behavior. We conclude that it is unlikely that these genes play a major role in the variation in aggression in the Golden Retrievers that we studied. Smaller phenotypic effects of these loci could not be ruled out with our sample size. Introduction Dogs have been living in close proximity to humans for at least 15,000 years (Clutton-Brock 1995). Behavior has been a strong selective factor in the domestication and breeding of dogs. According to the breed standard, Golden Retriever dogs should have a friendly character (http://www.goldenretrieverclub.nl; link accessed March 2007). However, there are reports of very aggressive Golden Retrievers (Galac and Knol 1997; Heath 1991). We recently described the behavioral phenotype of 110 Golden Retrievers referred to our clinic for aggression problems and 118 Golden Retrievers that were recruited because they were related to one or more of the aggressive dogs (van den Berg et al. 2006). The phenotypes were based on mail questionnaires and on personal interviews with dog owners. In a quantitative genetic study including 325 Golden Retrievers, we found a heritability of 0.8 for the traits of human-directed aggression and dog-directed aggression (Liinamo et al. 2007). The influence of serotonin (5-hydroxytryptamine, 5-HT) on aggressive behavior has been studied extensively (reviewed by Berman and Coccaro 1998; Gingrich and Hen 2001; Lesch and Merschdorf 2000). There is evidence for a role of the 5-HT system in canine aggression as well. For instance, Reisner and colleagues (1996) reported decreased levels of 5-hydroxyindoleacetic acid (the major metabolite of 5-HT) in cerebrospinal fluid of dominant aggressive dogs. Badino et al. (2004) found modifications of 5-HT receptor concentrations in brains of aggressive dogs. Domestication of silver foxes, which are taxonomically close relatives of dogs, seems to cause modifications in the 5-HT system (see Trut 2001 for a review). The role of 5-HT in canine aggression is further supported by two small clinical studies, where pharmacological or dietary intervention in the 5-HT system was shown to modulate aggressive behavior (DeNapoli et al. 2000; Dodman et al. 1996). Four genes that code for factors involved in serotonergic neurotransmission are particularly good candidates for the regulation of aggressive behavior: the serotonin receptor genes 1A (htr1A), 1B (htr1B), and 2A (htr2A), and the serotonin transporter gene (slc6A4). Serotonin receptor 1A plays a role in anxiety, stress response, and aggression (Olivier et al. 1995). Htr1A knockout mice show increased anxiety and stress response and an antidepressant-like phenotype (Heisler et al. 1998; Ramboz et al. 1998). In the above-mentioned studies of silver foxes, the researchers observed a lower density of 5-HT1A receptors in the hypothalamus of tame foxes compared to their wild counterparts (Popova et al. 1991). Many studies have suggested involvement of htr1B in the etiology of mental disorders. For instance, Huang et al. (2003) and Sanders et al. (2002) reported an association between one of the polymorphisms in the human HTR1B gene and alcoholism, suicidality, and obsessive-compulsive disorder. Knockout mice lacking htr1B display increased aggression (Saudou et al. 1994). A mutation in the human HTR2A gene is associated with altered 5-HT binding, which has been implicated in schizophrenia, suicidal behavior, impaired impulse control, and aggression history (Abdolmaleky et al. 2004; Bjork et al. 2002; Khait et al. 2005). Peremans and colleagues (2003) found an increased binding index of serotonin 2A receptors in cortical brain regions of impulsive aggressive dogs. A polymorphism in the promoter region of SLC6A4 influences serotonin transporter density in the brain and is associated with mental disorders in humans (Anguelova et al. 2003; Hariri et al. 2002; Lesch et al. 1996). Slc6A4 knockout mice show reduced aggression (Holmes et al. 2003). In this paper, we test the hypothesis that there is a strong effect of variation in these genes on the variation in aggression in Golden Retrievers. We performed mutation screens of the coding DNA sequence in unrelated aggressive Golden Retrievers. In addition, we used linkage analysis to determine the likelihood of the presence of a strong aggression locus in or close to the genes in several dog families. Third, we used 50 unrelated aggressive Golden Retrievers and 50 unrelated non-aggressive Golden Retrievers to search for association of alleles of 41 SNPs flanking the candidate genes with the trait of human-directed aggression. To complete our analyses, we evaluated the effects of variation in the genes on a range of aggression-related phenotypes using the same models as in Liinamo et al. (2007), extended to include the effects of the most common candidate gene haplotypes. Materials and methods Animals, DNA isolation, and phenotyping We have collected behavioral information of 328 privately owned Golden Retrievers. This group includes 162 dogs that were referred to our clinic because of their aggressive behavior (“probands”) and 166 relatives of 36 probands. DNA samples were available for 281 of these dogs. In addition, we collected DNA of a cohort of random privately owned Goldens that were born between July 2002 and February 2003. No phenotypes were available for these random dogs. Genomic DNA was isolated from whole blood leucocytes using a standard protocol (Miller et al.1988). For each type of analysis (mutation analysis, linkage analysis, association analysis, and quantitative genetic analysis) we selected a study group that was suited for the study design. The study groups are described in the Supplementary Information I. We have collected various quantitative measures of aggressiveness for the dogs (van den Berg et al. 2003a, 2006). In the linkage and association analysis we focused on one of these measures: the dog owners impression on human-directed aggression. Owner impressions were collected in a personal interview. We asked the owners if their dog was aggressive towards humans and the status of the dog was coded in three classes: non-aggressive (score 1), threatens (score 2), or bites (score 3). We focused on human-directed aggression because the majority of the probands were referred to our clinic for human-directed aggressive behavior. Owner impressions were available for all dogs and the quantitative genetic analyses showed that the heritability of this trait was high in our population of dogs (Liinamo et al. 2007). Mutation screening We analyzed the coding DNA sequence (CDS) of the four candidate genes in seven (htr1A and htr1B) or eight (htr2A and slc6A4) probands. The CDS were amplified and sequenced using overlapping primer pairs as described previously (van den Berg et al. 2004, 2005). Possible functional effects of polymorphisms were predicted with POLYPHEN (http://www.genetics.bwh.harvard.edu/cgi-bin/pph/polyphen.cgi). Effects of polymorphisms close to splice sites were predicted with three splice prediction programs: NetGene2 (Brunak et al. 1991), Splice Prediction by Neural Network (Reese et al. 1997), and SpliceSiteFinder (Shapiro and Senapathy 1987). Linkage analysis We used nine families for linkage analysis (Figs. S1–S9 of Supplementary Information I). DNA samples were available for 31 affected and 65 unaffected dogs from these families. We converted the owner impression about human-directed aggression into a dichotomous variable for the linkage analysis (see Supplementary Information I). We have described polymorphic markers for the candidate genes before (van den Berg et al. 2003b, 2004, 2005). We selected three microsatellite markers and seven single nucleotide polymorphisms for linkage analysis (see Table 1). Microsatellite markers were genotyped after PCR on an ABI 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA). PCR conditions were described by van den Berg et al. (2004, 2005). GENESCAN 3.7 software was used for genotype assessment. Single nucleotide polymorphism genotyping was performed by DNA sequencing of PCR products on the ABI 3100 Genetic Analyzer (van den Berg et al. 2004, 2005). The DNA sequence chromatograms were inspected using LASERGENE software (DNASTAR, Inc., Madison, WI USA). We combined several markers into haplotypes for the genes htr1B, htr2A, and slc6A4. For dogs that were heterozygous for multiple markers, we deduced the haplotypes from the data of relatives. If this was not possible, we assigned the most frequently observed possible haplotypes to these dogs. Table 1Markers used for linkage analysisGeneaType of markerbPosition of markercAlleles observed (allele frequency)dHaplotypes observed (haplotype frequency)dhtr1A(CA)n(UU160O12)*7370297 (0.5)303 (0.5)–htr1BA/C SNPe157A (0.58)C (0.42)143-A-G-T-G (0.20)143-C-A-T-G (0.42)143-A-G-C-C (0.06)139-A-G-C-C (0.24)139-A-G-T-G (0.05)147-A-G-T-G (0.03) G/A SNP246G (0.58)A (0.42)T/C SNP955T (0.69)C (0.31)G/C SNP1146G (0.69)C (0.31)(GA)n(UU18L8)−68395139 (0.29)143 (0.68)147 (0.03)htr2AC/T SNPIVS 2-10C (0.85)T (0.15)128-C (0.07)130-C (0.41)132-C (0.37)132-T (0.15)(CA)n(UUHTR2AEX2)IVS2 + 1439128 (0.07)130 (0.41)132 (0.52)slc6A4C/T SNP411C (0.75)T (0.25)C-G (0.75)T-A (0.25)G/A SNPIVS9-12G (0.75)A (0.25)ahtr1A, htr1B, htr2A = respectively serotonin receptor 1A, 1B, and 2A gene; slc6A4 = serotonin transporter genebSNP = single nucleotide polymorphism. Names of microsatellite markers have been included in bracketscPosition refers to the coding sequence of the canine gene. We used the nomenclature recommended by den Dunnen and Antonarakis (2001): the A of the ATG start codon is designated number 1, the nucleotide 5′ to this A is numbered −1, and the nucleotide 3′ of the translation termination codon is *1. Positions in introns refer to the nearest exon. The nomenclature of the introns is based on the human gene structure. IVS = intervening sequencedAllele and haplotype frequencies were determined in a group of 27 (htr1A and htr2A), 31 (htr1B), or 26 (slc6A4) parentseThis polymorphism is nonsynonymous We performed a parametric affecteds-only linkage analysis to determine whether the candidate gene haplotypes were linked to aggressive behavior in the Golden Retriever families. Marker haplotype frequencies were determined in a group of 27 (htr1A and htr2A), 31 (htr1B), or 26 (slc6A4) parent dogs (see Supplementary Information I). The mode of inheritance of the aggressive phenotype in our families is unclear. We therefore analyzed the data under both autosomal dominant and autosomal recessive models. The penetrance of the genotype at risk was set at 0.01. In this way, affected dogs are assumed to have the risk allele and the software calculated likelihood that aggressive dogs share alleles by descent from a common ancestor. Unaffected dogs with the genotype at risk have no effect on the outcome of the calculation. We assumed that there were no phenocopies in the families and we assumed genetic homogeneity because all probands were related to each other within a limited number of generations (not shown). The frequency of the aggression allele was set at 0.1 to allow for multiple transmitting ancestors in the pedigrees. SUPERLINK software was used to calculate two-point logarithm of the odds (LOD) scores (Fishelson and Geiger 2002, 2004). In order to estimate the power of the pedigrees, we calculated the maximum obtainable LOD scores. Affected individuals were assigned haplotypes 2/2 in these calculations; unaffected parents were assigned haplotypes 1/2; and other unaffected individuals were assigned haplotypes 1/1. We assigned haplotypes 0/0 (unknown) to dogs from which we did not have a DNA sample. We assumed that there were four alleles of the hypothetical marker with equal allele frequencies. Association study To test for a more complex genetic effect of variants of the candidate genes, we performed an association study. Fifty aggressive Golden Retrievers were selected from our database. The main selection criteria were high estimated breeding values for human-directed aggression and as little interrelationship among the cases as possible. Non-aggressive dogs were selected for low estimated breeding values for both human- and dog-directed aggression. We avoided an excess of relationship within either the case or the control group. The non-aggressive group was completed with 25 dogs from the random group. A more detailed description of cases and controls is provided in Supplementary Information I. We genotyped a total number of 60,073 SNPs in these 100 Golden Retrievers using customized Affymetrix Genotyping Arrays. The SNPs were selected for the chip using a scoring system that optimized the SNPs accounting for low repeat content, low likelihood of SNPs in the assay probe sequences and their distribution over the genome as a whole (Lindblad-Toh 2007; personal communication). Twenty 32-mer probes interrogated each locus with genotyping calls made using the algorithm BRLMM (http://www.affymetrix.com) which analyses intensities for sets of probes that interrogate both forward and reverse sequences with perfect match and mismatched probes. Dogs with call rates of lower than 50% were discarded from the analysis. The total set of SNPs was filtered for genotype call probability, heterozygosity rate, and call rate across a large set of dogs, reducing the dataset to 26,625 SNPs. From this set, only SNPs within 1 Mb of the four candidate genes were used for the analysis described in this paper. There were 29 SNPs within 1 Mb of htr1A, 20 within 1 Mb of htr1B, 25 within 1 Mb of htr2A, and 20 within 1 Mb of slc6A4. From these SNPs, we selected 43 SNPs with a minor allele frequency >0.05 and call rates of >0.75. We used Haploview software version 4.0 (Barrett et al. 2005) for the analysis of the presence of Hardy-Weinberg equilibrium (HWE), the local association analysis, and the calculation of pairwise linkage disequilibrium (LD) between the SNPs. A HWE P-value cutoff of 0.001 was used. We used Bonferonni correction to account for multiple testing in the association analysis. Genotype frequencies in cases and controls were compared with Chi square tests using SPSS software. Two tailed Fisher’s exact tests were used when the number of expected cases was less than 5 in more than 20% of the categories. Haplotype blocks were formed using three methods in Haploview (confidence intervals, four gamete rule, and solid spine of LD). For all possible combinations we performed 10,000 permutations to obtain empirical P values for haplotype association tests. We used the genetic power calculator prepared by Purcell to estimate the power of the association analysis (http://www.pngu.mgh.harvard.edu/~purcell/gpc/). The following assumptions were made: high-risk allele frequency = 0.1; prevalence = 0.01. The mean pairwise D′ between the SNPs flanking a candidate gene was used as an estimate of the local D′ in these estimations. Calculations were performed for two different genotype relative risks: 2 (genotype relative risk Aa = 2; genotype relative risk AA = 4) and 5 (genotype relative risk Aa = 5; genotype relative risk AA = 10). Quantitative genetic analysis In addition to the owner impression on human-directed aggression and dog-directed aggression, we collected a variety of other aggression-related behavioral measures using the canine behavioral assessment and research questionnaire (CBARQ; Hsu and Serpell 2003). As described in Liinamo et al. (2007), these measures were of three types: original CBARQ items (27 items on the aggressiveness of the dog in various everyday situations), shortened CBARQ scores (scores based on questions that addressed stranger-directed, owner-directed, and strange dog-directed aggression), and CBARQ factors (scores based on questions about stranger-directed, owner-directed, and strange dog-directed, and familiar dog-directed aggression). For further explanation of the measures and the difference between shortened CBARQ scores and CBARQ factors, see Liinamo et al. (2007). The effects of the haplotypes of the serotonergic genes on the different aggression measures were estimated with Restricted Maximum Likelihood (REML) method (Patterson and Thompson 1971), using univariate analyses and an animal model with the VCE4.2.4 software (Groeneveld 1997). The analyses were an extension of the analyses outlined in Liinamo et al. (2007), using similar linear animal model methodology, but this time also including the haplotype classes of the dogs for the four studied loci as additional fixed effects in the model. For instance, the linear animal model that was assumed in the analyses for owner impression traits was: where yijklmnois the observed value for the owner impression score for animal o;μ the general mean in the population; sexithe fixed effect of the reproductive status (i = 1–4, with 1 = intact male, 2 = castrated male, 3 = intact female, and 4 = castrated female); agej the fixed effect of the age j (j = 1–11, with 1 = 0.5–1 year old, 2 = 1–2 years old,…,10 = 9–10 years old, and 11 = over 10 years old); htr1ak , htr1bl, htr2am, and slc6a4n the fixed effects of the respective haplotype classes, aothe random additive genetic effect (i.e., polygenic breeding value) of the animal o, and eo the random residual effect related to the animal o. The age and reproductive status of the dogs had been recorded at the same time as the owner impressions. The haplotype classes were formed so that the most common haplotypes formed separate classes, the very rare haplotypes were all combined in one class, and the unknown haplotypes were classified as a separate class (see Table 4). Results Mutation screening of the coding DNA sequence The coding DNA sequence of each candidate gene was scanned for mutations in seven (htr1A and htr1B) or eight (htr2A and slc6A4) aggressive Golden Retrievers. Analysis of the CDS in non-aggressive Golden Retrievers has been described by van den Berg et al. (2004; 2005). There was no variation in the CDS of htr1A and htr2A in the Golden Retrievers. We observed five SNPs in the CDS of htr1B and one SNP in the CDS of slc6A4. The allele distribution of these SNPs in the two groups of Golden Retrievers did not indicate a role in aggressive behavior. In conclusion, there seems to be no systematic difference between the CDS of the candidate genes in aggressive and non-aggressive Golden Retrievers. Linkage analysis We observed two alleles for htr1A marker UU160O12 (Table 1). The four SNPs in htr1B were fully in LD (D′ = 1). The SNPs displayed six haplotypes in the Goldens, three of which were rare (frequency <0.1). We detected four haplotypes of htr2A in the Golden Retrievers. The two SNPs in slc6A4 were fully in LD and formed two haplotypes in the dogs. In the nine families that we used for linkage analysis, haplotypes were deduced with certainty in 86% of the dogs for htr1B, 100% for htr2A, and 87% for slc6A4. We calculated the maximum achievable LOD score using hypothetical genotypes. The maximum LOD score generated by our pedigrees was 2.8 at recombination fraction θ = 0 assuming a dominant mode of inheritance (Table 2). Under a recessive model, the maximum LOD score was 5.3 at θ = 0. The families are therefore theoretically powerful enough to prove linkage under a recessive model and powerful enough to provide a good indication of the presence of linkage under a dominant model. There was no significant linkage of any of the candidate genes with the aggressive phenotype (Table 2). LOD scores varied from −1.0 to +0.26 assuming dominant inheritance and from −2.3 to −0.30 assuming recessive inheritance. The highest LOD scores were obtained for htr1A (+0.26 under a dominant model and −0.30 under a recessive model). Table 2Results from the ODDS (LOD) scoresaGeneAutosomal dominant Autosomal recessivehtr1A0.26−0.30htr1B−0.72 −2.3htr2A−1.0 −2.1slc6A40.030−1.2Maximum2.85.3aLOD scores were calculated with the following assumptions: frequency of the “aggression allele” = 0.1; penetrance of the “aggression allele” = 0.01; θ = 0. Marker haplotype frequencies were deduced from a group of parents Association study Two dogs (one case and one control) were discarded from the association analysis because they had call rates lower than 50%. Mean call rates in the other 98 dogs were 92% for SNPs flanking htr1A, 94% for htr1B, 93% for htr2A, and 93% for slc6A4. There were 43 SNPs with a minor allele frequency >0.05 and call rates of >0.75 that occurred within 1 Mb of the candidate genes. The genotype frequencies of BICF2P1093362 for htr1B and BICF2P969902 for slc6A4 deviated from HWE in control dogs (P = 4.42E-14 and P = 5.242E-12, respectively). All but one dog in both case and control group had heterozygous genotypes for these SNPs. We concluded that the data for these two SNPs was artefactual and they were excluded from further analyses. The observed genotype frequencies of the other 41 SNPs were in HWE in controls (P values are listed in Table S2 in Supplementary Information II). The final SNP set used for the association analysis consisted of 12 SNPs flanking htr1A, 11 flanking htr1B, 8 flanking htr2A, and 10 flanking slc6A4 (Table 3). Table 3Single nucleotide polymorphisms (SNPs) used for the association study, their allele frequencies in 49 aggressive cases and 49 control dogs, and results of chi-square tests for comparisons of case and control allele frequenciesSNP nameaChromosomal locationChromosomal position (Mb)bMinor allele frequency in controlsCorresponding allele frequency in casesχ2P-valueBICF2P546848251.800.290.300.0440.83BICF2P1051894251.960.0730.0281.60.21BICF2P1398268252.010.330.330.0110.92BICF2S23127755252.140.230.210.0780.78BICF2P1200391252.220.300.330.210.64BICF2P590055252.350.240.270.190.66BICF2S22939125252.440.480.530.540.46BICF2S23215863252.480.300.310.0250.87BICF2P25993252.730.490.450.190.66BICF2S23442706252.760.190.281.80.18Htr1A252.88–52.88––––BICF2P519607253.220.270.240.170.68BICF2P1341930253.620.490.540.520.47BICF2P11592411241.110.120.130.0120.91BICF2P5551301241.590.130.212.10.15BICF2S233262291241.630.240.321.30.25Htr1B1241.65–41.66–––BICF2P6703311241.800.150.232.20.14BICF2S231537601241.840.0390.123.20.073BICF2P14265221242.290.120.120.0130.91BICF2P275711242.340.120.100.210.65BICF2S234440661242.390.120.110.0320.86TIGRP2P164447_rs88059861242.510.0380.123.60.059BICF2P8121531242.570.0330.124.80.029BICF2P8554021242.620.0210.126.90.0086BICF2G630315581226.3830.100.140.830.36BICF2G630315746226.6110.100.140.450.50BICF2P1168502226.9730.100.110.00200.96Htr2A227.395–7.453––––BICF2P164280227.5090.430.542.00.16BICF2G630316047227.7090.440.531.70.19BICF2S23125159227.8660.190.210.0470.83BICF2S23661838228.2070.310.360.590.44BICF2S22954191228.4700.230.270.350.55BICF2P813837946.930.390.301.50.22BICF2S23018060947.060.330.370.220.64BICF2S23551918947.460.340.380.240.62Slc6A4947.55–47.57–BICF2S23325050947.640.110.150.530.47BICF2S23124809947.790.180.220.380.54BICF2P950384947.880.110.120.0320.86BICF2S245135948.020.270.250.0740.79BICF2S2347312948.090.220.280.840.36BICF2S23154457948.130.310.241.20.28BICF2S23141984948.260.360.360.00700.93ahtr1A, htr1B, htr2A = respectively serotonin receptor 1A, 1B, and 2A gene; slc6A4 = serotonin transporter gene. The genes are included in the table to show their position relative to the SNPsbSNP positions are based on the second version of the dog genome assembly, released in May 2005 (CanFam2.0) as can be viewed on http://www.broad.mit.edu/ftp/pub/papers/dog_genome/snps_canfam2/Positions of the genes are based on the second version of the dog genome assembly (CanFam2.0) as displayed in NCBI Map Viewer for Canis familiarishttp://www.ncbi.nlm.nih.gov/mapview/map_search.cgi?taxid=9615 The power of our association analysis depends on the local extent of LD. The mean r2 between the SNPs flanking htr1A was 0.36 and the mean D′ between these SNPs was 0.89. Mean r2 values were 0.23, 0.25, and 0.21 for htr1B, htr2A, and slc6A4, respectively. Mean D′ values were 0.80, 0.72, and 0.88 for htr1B, htr2A, and slc6A4, respectively. When the marker allele frequency is 0.1, the power to detect a variant with a relative risk of 5 with 49 cases and 49 controls would be 0.91 for htr1A, 0.85 for htr1B, 0.78 for htr2A, and 0.90 for slc6A4. Additional power estimations are provided in Fig. S10 in Supplementary Information II. The allele frequencies of the SNPs did not differ significantly between cases and controls after correction for multiple testing (Table 3). Genotype frequencies also did not display significant differences between cases and controls (see Table S2 in Supplementary Information II). We also analyzed the association of haplotypes with the phenotype. No significant associations were found (data not shown). In conclusion, there seemed to be no association between alleles, genotypes or haplotypes of SNPs that flank the candidate genes and human-directed aggression in the Golden Retrievers. Quantitative genetic analysis We completed our analyses with a study of the effect of variation in the candidate genes on a collection of aggression-related phenotypic measures. The haplotype effects were studied on owner impressions on human- and dog-directed aggression, the original CBARQ items related to stranger- and owner-directed aggression, the shortened CBARQ scores and the CBARQ factors. The haplotypes did not have a significant effect on any of the studied measures, i.e. the heritability estimates of the measures remained similar to the results presented in Liinamo et al. (2007) in spite of incorporation of the haplotypes in the mixed model. The results for owner impressions on human- and dog-directed aggression, which are the most reliable estimates due to the largest number of observations, are presented in Table 4. In conclusion, the large genetic variability between the dogs could not be explained by the serotonergic genes studied in this paper. Table 4The effects of the studied genotypes (htr1A) or haplotypes (htr1B, htr2A, and slc6A4) on owner impressions of human-directed aggression and dog-directed aggression in 320 dogs Genotype or haplotype classaNumber of animalsEffect relative to class ‘unknown’Human-directed aggressionDog-directed aggressionhtr1AUnknown440.000.00297/297740.0280.11297/3031110.240.054297/305150.0120.089303/303620.180.085303/30510−0.11−0.067Other4−0.046−0.17htr1BUnknown1660.000.00143-A-G-T-G/143-A-G-T-G20−0.730.28143-A-G-T-G /143-C-A-T-G8−0.63−0.19143-A-G-T-G /139-A-G-C-C190.110.19143-C-A-T-G/143-C-A-T-G41−0.170.13143-C-A-T-G /143-A-G-C-C10−0.0210.059143-C-A-T-G /139-A-G-C-C300.00060.32139-A-G-C-C/139-A-G-C-C90.240.11Other170.150.36htr2AUnknown1140.000.00132-C/132-C40−0.15−0.19132-C/130-C60−0.27−0.023132-C/132-T28−0.220.073132-C/128-C15−0.38−0.43130-C/130-C21−0.33−0.072130-C/132-T220.10−0.19130-C/128-C9−0.52−0.29Other110.33−0.036slc6A4Unknown1030.000.00C-G/C-G1160.025−0.23C-G/T-A940.14−0.23Other70.079−0.10None of the effects was significantaNote that there are additional alleles compared to Table 1 as a result of the larger study group. The htr1A class “other” contains genotypes 295/297 (n = 2 dogs) and 305/305 (n = 1 dog). For htr1B, the class “other” contains haplotypes 143-A-G-T-G/143-A-G-C-C (n = 3 dogs), 143-A-G-T-G/139-A-G-T-G (n = 1 dog), 143-A-G-T-G/147-A-G-T-G (n = 2 dogs), 143-A-G-C-C /139-A-G-C-C (n = 3 dogs), 143-A-G-C-C/147-A-G-T-G (n = 1 dog), 139-A-G-C-C /139-A-G-T-G (n = 3 dogs), and 143-C-A-T-G/139-A-G-T-G (n = 4 dogs). For htr2A, “other” contains haplotypes 0/130-C (n = 2 dogs), 0/132-T (n = 1 dog), 132-T /132-T (n = 6 dogs), and 132-T/128-C (n = 2 dogs). For slc6A4, “other” contains C-G/T-G (n = 1 dog) and T-A/T-A (n = 6 dogs). Unknown genotypes and haplotypes are the result of failure of genotyping or of the absence of a DNA sample (n = 44 dogs) Discussion We collected behavioral information and DNA samples of 281 dogs over a period of 10 years. Dogs were selected from this collection to evaluate four genes involved in serotonin metabolism by four methods: DNA sequence analysis of the coding region of the genes, genetic linkage analysis, genetic association analysis, and quantitative genetic analysis. The results indicate that it is unlikely that there is a major locus effect of one of the genes on aggression in the Golden Retrievers that we studied. The genetic study of the variation of aggression in Golden Retrievers is a promising tool to identify the molecular systems involved in aggression. The relative ease to find disease loci in the dog genome compared to the human genome is the result of the population structure of dog breeds. Within a breed, Lindblad-Toh and colleagues (2005) observed a limited number of common haplotypes per genomic region. In addition, LD in dog breeds extends over at least 50-fold greater distances than in human populations. These characteristics make the dog highly suited for molecular genetic studies of complex traits (Sutter and Ostrander 2004). We did not detect mutations in the CDS of the genes specific for aggressive dogs. All SNPs except A157C in htr1B were synonymous. This A157C variation, resulting in an isoleucine/leucine polymorphism of amino acid 53, was predicted to be functionally insignificant by POLYPHEN. The genes htr2A and slc6A4 contain SNPs close to splice sites (at position IVS2-10 and IVS9-12, respectively) that could theoretically affect splicing. However, the polymorphisms did not have a large effect on splice site prediction by three software programs. We performed mutation screening in a limited number of dogs and it is possible that we have missed rare alterations in the genes. Apart from this limitation, we conclude that there is no common variant acting on protein structure that contributes to the variation in aggression in our Golden Retriever sample. We used linkage analysis to evaluate the likelihood that there is a major aggression-influencing variant in the chromosomal regions surrounding the coding exons. The affecteds only parameters that we used in the calculations are a simplification with the assumption that all affected dogs of a family have the genotype at risk but unaffected dogs can have any genotype at the aggression locus. A LOD score of 3 is usually considered as evidence for linkage, whereas LOD scores below −2 exclude the gene. A power calculation with simulated genotypes was not feasible in this study due to the complexity of some pedigrees with multiple loops. Instead, we calculated the maximum achievable LOD scores to get an impression of the power embodied in the pedigrees. We assumed full informativeness of the markers in these calculations, but in reality, we expect the markers to have limited informativeness. Realistic obtainable LOD scores would then be lower than the maximum values that we presented in Table 2. These LOD scores are too low to obtain significant results, but they provide a means to set the obtained results into perspective. The LOD scores for htr1B and htr2A were low compared to the maximum obtainable scores. A major role of these genes is unlikely. The results for htr1A and slc6A4 are less conclusive. This is probably the result of the poor informativeness of the markers. For both genes, we observed only two alleles or haplotypes with high frequencies. Typing of additional markers might help to definitively exclude the genes. However, in the light of the observed low level of variation it is unlikely that htr1A and slc6A4 have a strong effect on aggression in the Golden Retriever families. Our linkage analysis does not account for genetic heterogeneity or phenocopies. We have thus only tested for a very strong major locus effect. In reality, the aggressiveness in the Golden Retrievers may be more complex. We therefore used a third study design to investigate the candidate genes: association analysis. For this analysis, we used data from a large-scale genotyping project in 100 Golden Retrievers. Our power calculations demonstrate that this sample size is expected to be sufficient to detect variants that confer a high relative risk for a range of marker allele frequencies. From the total set of 60,073 SNP genotypes, we selected 41 SNPs that flank the candidate genes. We found no association between alleles, genotypes or haplotypes of these SNPs flanking the candidate genes and human-directed aggression of the Golden Retrievers. We focused on human-directed aggression in the affecteds-only linkage analysis and the association study. In our quantitative genetic analysis, we studied additional types of aggression. There is no consensus in the literature on how aggression should be subdivided (Houpt and Willis 2001; Jacobs et al. 2003; Serpell and Jagoe 1995). There are indications that various types of aggression have a distinct genetic basis. For instance, selection of rats and silver foxes for reduced fear-induced aggression towards humans did not change predatory or inter-male aggression (Naumenko et al. 1989; Popova et al. 1993). This suggests that molecular genetic studies of aggressive behavior should focus on specific classes of aggression. However, reduced aggressiveness towards man in the rats and foxes was accompanied by reduced fear of novelties and irritable aggression, indicating that there is overlap between classes. As long as the genetic roots of aggressive behavior are poorly understood, it will remain impossible to design a classification that reflects the genetic basis. In conclusion, none of the four methods of analysis provided evidence for a strong effect of variants of the candidate genes on aggression in the Golden Retrievers that we studied. These results seem to contradict reports of the involvement of the candidate genes in the regulation of aggressive behavior. However, the study designs that we used are not powerful enough to detect variants of small effect. We can therefore not rule out the possibility that variation in the candidate genes has a smaller genetic effect on aggression. In addition, our results cannot be construed as evidence against a major role for these genes in aggression in other dog breeds. Possibly, other genes in the serotonin pathway play a role. With the completion of the dog genome project, genome-wide association studies have become feasible in dogs (Lindblad-Toh et al. 2005). This opens the opportunity for finding genes that have not been associated with aggression up to date. Such studies are in progress. Electronic supplementary material Below is the link to the electronic supplementary material. (DOC 174 kb)

J_Urban_Health-2-2-1705546

Application of Respondent Driven Sampling to Collect Baseline Data on FSWs and MSM for HIV Risk Reduction Interventions in Two Urban Centres in Papua New Guinea

The need to obtain unbiased information among hard–to-reach and hidden populations for behavioural and biological surveillance, epidemiological studies, and intervention program evaluations has led researchers to search for a suitable sampling method. One method that has been tested among IDU and MSM recently is respondent-driven sampling (RDS). We used RDS to conduct a behavioural survey among FSWs and MSM in two urban centres in Papua New Guinea (PNG). In this paper we present the lessons learned implementing RDS in a developing country setting. We also present comparisons of RDSAT-adjusted versus unadjusted crude estimates of some key socio-demographic indicators as well as comparisons between the estimates from RDS and a hypothetical time–location sample (TLS). Overall, the use of RDS among the MSM and FSWs in PNG had numerous advantages in terms of collecting a required sample size in a short time period, minimizing costs and maximising security for staff and respondents. Although there were a few problems these were easily remedied and we would recommend RDS for other similar studies in PNG and other developing countries. Introduction Papua New Guinea is currently experiencing a major generalized HIV epidemic driven by high rates of unprotected sex, multiple sexual partners, early age of sexual debut, and high rates of STIs. Sexual violence and aggression is also widespread.1 PNG's HIV surveillance system has limited capacity, and valid scientific information is needed to guide community prevention efforts.2 Female sex workers (FSWs) and men who have sex with men (MSM) have been identified as groups at high–risk for HIV throughout the world. There have been few published studies about FSWs in PNG.3,4 No previous quantitative research has been conducted with MSM in PNG, but qualitative research and unpublished reports indicate that MSM behaviour is prevalent in some areas of the country.5 Moreover, numerous anthropological studies indicate that many tribes in PNG have practised male–male sexual behaviour in traditional rites and rituals.6–8 The major challenge in studying HIV–risk behaviours and developing prevention efforts among high risk populations is gathering information from non–biased samples.9 Members of population groups such as men who have sex with men (MSM), injecting drug users (IDU) and female sex workers (FSWs) are involved in stigmatized or illegal behaviour, leading individuals to hide their identity and to be reluctant to participate in research studies. No sampling frame exists for these groups, and the fact that creating one may be difficult and very costly makes them hard to sample with traditional probability–based sampling methods. Respondent driven sampling (RDS) is a relatively new adaptation of chain–referral sampling, where subsequent respondents are recruited by previous respondents through their network of acquaintances.9–11 It has several features that overcome the limitations of other sampling methods that allow it to provide unbiased and representative population–based estimates. RDS has therefore been recommended as an alternative sampling approach for “hidden” populations that do not congregate in identifiable or accessible locations.12–14 Because RDS relies on participants from the target population to recruit subsequent respondents, it has the advantages that it requires less detailed formative research, is less costly, can be completed in a shorter amount of time, and has greater external validity than other methods. This paper presents lessons learned from applying RDS in a study to collect baseline behavioural data for HIV prevention programming among two high risk populations in two urban centres in Papua New Guinea. The research among FSWs in Goroka and Port Moresby and MSM in Port Moresby was linked to the Poro Sapot Project (PSP), which is an HIV prevention initiative of Save the Children in PNG (SCiPNG). The Papua New Guinea Institute of Medical Research (PNGIMR) partnered with SCiPNG and Family Health International (FHI) to conduct the research. Reasons for Choosing RDS The first phase of the research project was a qualitative evaluation of the two target populations. Qualitative research methods included field observations, focus group discussions, in–depth interviews, and key informant interviews. In addition, a literature review of research studies on these populations in PNG was carried out. Our formative research showed that there was a relatively large and socially well connected population of MSM in Port Moresby, the capital and largest city of PNG. Although there were a number of commercial and public spaces where MSM met and socialised, none of them catered to a primarily MSM clientele. There were no formal community–based or social organizations of gay or homosexual men. Due to social stigma and discrimination, the vast majority of MSM kept their sexual orientation hidden from their family, friends and co–workers. The lack of MSM–identified sites and the hidden nature of homosexuality in PNG meant that it would be difficult to use time–location sampling (TLS) as the recruitment methodology.13,14 The decision to use RDS for the survey of FSWs came about following formative assessments in both locations. FSWs in PNG were found to be very mobile. Although a few illegal brothels did exist, only a small number of women worked in those locations. Most FSWs frequented formal establishments such as disco bars, hotels, motels, and guest houses in search of potential clients. In most locations FSWs mixed with the general population in a way that would have made it difficult to differentiate sex workers from other females. In addition, security for the research staff and participants was a major concern. MSM informants had complained of physical and verbal abuse due to social stigmatization. The crime rate in Port Moresby was very high and had steadily deteriorated in the time period before the research was conducted. Since most MSM and FSWs went out at night to socialise or meet clients, it would have been difficult to ensure the security of research staff members if they had to contact potential respondents in public areas. RDS provided a safer alternative in this regard because respondents were not actively recruited by study interviewers but were instead referred by their peers to be interviewed at a safe study facility during a convenient time of the day. Research Methods The field work in Port Moresby began in February 2005 and was completed at the beginning of April 2005. Data collection in Goroka began in April 2005 and was completed in 2 weeks. The interviewers were all FSWs and MSM who received 1 and 1/2 weeks of intensive training prior to the starting of field work. The research teams at each site were comprised of a coupon manager, a research assistant and a group of peer interviewers. In Port Moresby we trained and used six MSM interviewers and four FSWs interviewers. In Goroka we trained and used five FSWs interviewers. The Coupon Management System A coupon management system was developed in Microsoft FoxPro 2.6 and was used to track the relationships between the recruiters and their recruits. The same program was used to record biometric measurements (circumference of both wrists and length of both forearms), which were used to check for and prevent duplication of recruitment. The latter was done through the calculation of the index of differences between the candidate and all of the other biometric measures already entered.15 Recruitment of Seeds and Study Participants The FSWs seeds for the RDS sample were drawn from distinct geographic areas within Port Moresby and Goroka. The areas chosen in each location were the coverage areas that were mapped out and targeted for the intervention programs. Therefore eight seeds were chosen who lived and worked in five different areas of the city of Port Moresby, and six seeds were chosen from six different areas in Goroka. These areas and the number of seeds drawn for Port Moresby and Goroka are shown in Tables 1 and 2, respectively. Table 1Socio–demographic characteristics of FSWs in Port Moresby SeedsSample(n=8)Percent(n=245)Crude (%)RDSAT–adjusted (%)Absol. diff (%)Age16–19112.54819.625.05.420–24112.53815.518.63.125–29112.55422.022.70.730+562.510542.933.59.4Marital statusEver married675.018575.566.98.8Currently married112.55823.717.26.5Currently married cohabiting with spouse00.072.93.91.0Education levelNo formal education00.05823.726.67.9Primary337.512149.447.51.9Secondary562.56124.922.22.7Other (eg. vocational)00.052.03.21.2Region of birthHighlands225.011346.142.73.4Southern562.511647.351.23.9Others112.5166.55.90.6Place of residenceTown/Koki area225.06225.318.46.9Boroko area112.510040.823.617.27–mile/airport225.0156.14.61.5Gordons/Hohola225.0197.88.81.0Waigani/Gerehu112.54618.843.224.4Others00.031.21.11.2Table 2Socio–demographic characteristics of FSWs in Goroka SeedsSample(n=6)Percent(n=249)Crude (%)RDSAT–adjusted (%)Absol. diff (%)Age16–19116.74518.116.61.520–24116.78032.135.33.225–29116.76224.923.91.030+350.06224.924.00.9Marital statusEver married350.016465.965.80.1Currently married00.02710.812.41.6Currently married cohabiting with spouse00.000.00.00.0Education levelNo formal education233.38734.935.80.9Primary233.310843.443.20.2Secondary233.35321.322.00.7Other (eg. vocational)00.020.81.00.2Region of birthHighlands6100.023895.696.40.8Momase00.083.22.70.5Others00.031.20.80.4Place of residenceSeigu/fish Wara116.77329.331.21.9Asariyufa116.76224.926.31.4North Goroka116.73514.111.13.0Kama116.7124.82.62.2Lofi/Faniyufa116.7239.27.12.1Waterise116.73212.915.52.6Others00.0124.85.70.9 In some areas of Port Moresby two seeds were chosen: one seed was a younger FSWs while the other was an older one. Formative research had shown that the younger and older FSWs in some locations did not mix socially and therefore could be considered as separate populations. In other areas only one seed was used as in these areas the younger and older FSWs mixed together socially. The eligibility criteria for the selection of FSWs seeds as well as their subsequent recruits were defined as “a female over the age of 16 who has exchanged sex for money or other goods and services in the last 12 months and is not drunk at the time of the interview.” For the MSM study in Port Moresby we also attempted to recruit a diverse set of seeds based on self–identified sexual orientation and region of origin. Seeds were recruited through SCiPNG, which had recently initiated a peer education program with MSM. The peer educators in the project also worked as interviewers for the research. These men tended to recruit MSM seeds from their own social group. As a result, eight of the ten seeds used were self–identified as homosexual, and seven of the ten were from the southern region, which was the site of the research. As we will report later, the biases inherent within the initial group of MSM seeds were not present in the final sample population, in line with RDS theory. The eligibility criteria for the selection of the MSM seeds as well as their subsequent recruits were defined as “a man over the age of 16 who has had sex with another man in the last 12 months and is not drunk at the time of the interview.” In none of the three study groups was it found necessary to add additional seeds during the course of the study. Incentives Seeds and their recruits were offered monetary incentives as well as gift bags containing pamphlets, posters, free condoms and lubricant. Participants in Port Moresby were given 20 PNG Kina (K20, equal to U.S.$6.00) for successfully participating in the study and K10 (U.S.$3.00) each for every successfully recruited peer. The FSWs in Goroka were given K10 (U.S.$3.00) for participating and K5 (U.S.$1.50) for every successful recruitment made. The amount for incentive was less in Goroka because of the lower cost of living there compared to Port Moresby. Interviews and Setting As part of the RDS field work in Port Moresby, the research team rented two private properties situated in secluded locations of the city. The FSWs study in Goroka was also carried out at a private property in a residential part of town. Data was collected by trained interviewers who were MSM and FSWs themselves and was directly supervised by the principal investigators. All interviews took place in private rooms with only the interviewer and subject present. The secure nature of the locations used also allowed the research staff to exclude non–participants from the sites while recruitment and interviews were taking place. Participants gave witnessed verbal consent to be part of the study. Interviews were conducted using a structured questionnaire. Information obtained included demographics, sexual behaviour, drug and alcohol use, knowledge and use of male and female condoms, knowledge and attitudes towards HIV/AIDS and STI, history and treatment seeking behaviour of STI, experience with stigma and discrimination, and exposure to intervention programs. The MSM interviews included information about their sexual orientation. Additional information was collected that was specifically required for RDS methodology: personal network size, relationships to recruiters, and the number of recruitment refusals encountered. Sample Sizes A total of 245 and 249 FSWs were interviewed in Port Moresby and Goroka, respectively. For the MSM study in Port Moresby, a total of 225 MSM participants were interviewed. A total of 471 and 621 coupons were given out to 157 and 207 FSWs in Port Moresby and Goroka, respectively, while 603 coupons were given out to 201 MSM. Participants in all sites were told in advance that recruitment would stop when the desired sample size was reached. Participants who had lost a coupon were not given any additional ones. Failure to meet the eligibility criteria resulted in 15 females (four from Goroka) and 35 males being refused participation in the study. Half of the women who were excluded in Port Moresby were under 16 years of age, while the other half were excluded because they reported that they had never sold sex. The four FSWs that were excluded in Goroka were all under 16 years of age. The bulk of the men that were excluded from the MSM study (33) did not meet the criteria of having had sex with another man in the previous 12 months, while two others were excluded for trying to repeat the interview. Data Analysis Methods Data was double entered using Microsoft FoxPro 2.6 and analysed using Stata 8.0 (Stata Corporation, Texas, USA), RDS Analysis Tool (RDSAT, Cornell University, NY, USA) and EpiInfo 6.04 (CDC, USA and WHO, Geneva). Adjusted analysis was carried out using RDSAT on single categorical variables whereas bivariate and multivariate analysis was done using EpiInfo and Stata. RDSAT adjusts the analysis of population estimates by taking into consideration the links between the recruiters and their recruits as well as their reported personal network sizes.11,12 A detailed description of the methods of calculation is beyond the scope of this paper. Some Key Results Of the 157 and 207 FSWs in Port Moresby and Goroka that were given coupons, 56.7 and 42.0%, respectively, recruited one or more of their peers into the study. For the MSM study, 57.2% of the 201 participants that were given coupons recruited one or more peers. Two questions were asked with regards to the number of coupons given out and the number of refusals encountered by each recruiter to determine refusal rates. However, the research assistants did not consistently record the answers to these questions. Nevertheless, the refusal rates appeared to be very low as indicated by the fact that 74.2 and 85.1% of the FSWs recruiters who successfully recruited other participants in Port Moresby and Goroka returned all three coupons. Of the successful MSM recruiters, 52.2% of them had all three coupons returned, 35.6% had two coupons returned and 12.2% had one coupon returned. Half of the MSM seeds (five) did not recruit any peers while three FSWs seeds (two from Port Moresby) did not recruit any peer. It took a median of 4 days (Range=1–12) and 3.5 days (Range=1–10) for a coupon to be returned by FSWs in Port Moresby and Goroka, respectively. MSM returned the coupon at a median of 3 days (range=1–18). For both the FSWs and MSM, however, recruitment occurred at a much faster rate than had been anticipated and recruits presented to the study locations faster than they could be interviewed. It became necessary to slow down recruitment by scheduling interviews each day on a first–come first–served basis. Those that could not be interviewed the day they first presented were scheduled for the next day. Eventually the recruiters were told to bring their recruits in on specified dates in order to control the flow of new subjects to the study sites. The fast rate of recruitment for the three subpopulations enabled the sample sizes for FSWs in Port Moresby and Goroka to be reached in 17 and 20 days, respectively, while the sample size for MSM was reached in 22 days. The main sociodemographic characteristics of the seeds and the overall samples for the three studies are presented in Tables 1, 2 and 3. Comparisons between the RDSAT–adjusted and crude analyses are given as well. Table 3Socio–demographic characteristics of MSM in Port Moresby SeedsSample(n=10)Percent(n=225)Crude (%)RDSAT–adjusted (%)Absol. diff (%)Age16–19220.04921.820.81.020–24110.08136.038.12.125–29660.06026.727.71.030+110.03515.613.22.4Education levelNo formal education00.0125.36.71.4Primary00.08738.749.510.8Secondary660.06930.740.49.7Secondary440.0156.73.13.6Employment statusUnemployed220.012054.052.02.0Employed330.02913.011.02.0Self–employed550.06931.035.04.0Students00.042.02.00.0Region of birthHighlands220.03314.723.99.2Southern—gulf only110.011450.754.13.4Southern—others660.04821.310.710.6New Guinea islands00.0104.44.80.4Momase110.0208.96.32.6Sexual orientationHeterosexual110.02912.910.02.9Gay/homosexual880.06528.923.05.9Bisexual110.013158.267.08.8Paid by man for sexYes990.016975.170.74.4No110.05624.929.24.3 The results for the FSWs study in Port Moresby show that, while the seeds where roughly comparable to the overall sample, there were still some major differences with regard to educational status and region of birth. For the FSWs study in Goroka, there was less variation between the seeds and final sample (Table 2), although the seeds tended to be older than overall sample. The most striking differences between the initially selected seeds and the final sample were found for the MSM study in Port Moresby (Table 3). The final sample was younger, less educated, and more often unemployed. In addition, while the majority of the seeds (80%) were gay identified, less than a third of the final sample self–identified as homosexual and the majority of the final sample self–identified as bisexual. For all three studies (Tables 1, 2 and 3), a common pattern in the findings that emerged with regard to comparisons of the characteristics of seeds and final samples, was that the overall sample was of lower socioeconomic status (especially for the MSM study). This suggests that RDS is able to penetrate “hidden” populations that are often very different from the initially selected seeds. In terms of the comparisons between the RDSAT–adjusted and crude analysis in Tables 1, 2 and 3, the greatest variations between the two analyses (as measured by the percent of absolute difference) were found for FSWs in Port Moresby. This was especially true with regard to place of residence, where up to 24.4% differences were found for the ones that reside in one part of the city (Table 1). For FSWs in Goroka, considerably less variation was found with absolute differences ranging from 0 to 3.2% (Table 2). For the MSM study, some considerable differences were also seen and absolute differences ranged from 0 to 10.8%. The results presented in Tables 4 and 5 were an attempt to understand the potential differences in respondent characteristics that would have existed between RDS and TLS, had this latter sampling method been used instead of RDS. The hypothetical TLS group for this analysis included FSWs and MSM respondents who reported that they ‘always’ looked for and met their sexual partners in visible locations that would have been likely included in a TLS sampling frame (i.e., nightclubs, bars, hotels, guesthouses and identifiable public venues). The results shown in Table 4 indicate that the RDS samples for FSWs in both Port Moresby and Goroka were younger and less educated than the hypothetical TLS group. While reported condom use with clients was similar between the RDS and TLS groups, reported condom use with non–paying partners was substantially lower among the RDS group. On the other hand, forced sex, knowledge, and reported STI symptoms were higher in the TLS group. Analysis of the key program exposure variable, contacted by peer educator, also revealed that program exposure was substantially lower among the RDS sample than hypothetical TLS sample. The analysis of the MSM study population in Port Moresby (Table 5) showed striking differences in almost all variables examined. In comparison to the hypothetical TLS group, the RDS group was older, less educated, more likely to be unemployed, more likely to identify their sexual orientation as heterosexual or bisexual and less likely to use condoms for partnerships examined. This group was also less likely to use lubricant during anal sex and less likely to have been exposed to HIV prevention programming. On the other hand, the hypothetical TLS group was more likely to report forced sex, incidents of discrimination and symptoms of STIs. Table 4RDS versus hypothesized TLS estimates of key indicators for FSWsKey indicatorsPort Moresby (%)Goroka (%)RDSa (n=235)TLS (n=103)RDSa (n=227)TLS (n=165)Age16–2443.636.951.949.725+56.263.147.950.3Education levelNo formal education26.621.435.830.1Primary47.553.343.244.6Secondary22.224.322.024.1Used condom with client at last sex86.088.279.077.7Used condom consistently with clientc62.661.231.331.3Used condom with non–paying partner at last sex48.375.743.062.0Used condom consistently with non–paying partnerc33.548.716.026.0Forced to have sexb58.466.766.570.5Thinks mosquitos can transmit HIV37.150.030.736.1Had vaginal dischargeb43.253.428.637.4Had sore in/around vaginab20.524.312.119.4Never contacted by peerb educator24.418.540.939.6aRDSAT adjustedbIn the last 12 monthscIn the last 4 weeksTable 5RDS versus hypothesized TLS estimates of key indicators for MSM It is important to note that the observed differences between the two groups in Tables 4 and 5 would have been even more pronounced if the comparisons had been made between the mutually exclusive categories—the hypothetical TLS group and the RDS group restricted to respondents who would likely not have been sampled using TLS. Discussion of the Lessons Learned Sampling procedures should be capable of reaching all members of the population or subpopulation under surveillance in order to produce unbiased estimates of trends in HIV behavioural risks.12 In this respect, RDS was chosen to study the three subpopulations of interest in PNG after carefully considering its advantages and disadvantages for each target group. Many factors prompted us to opt for RDS instead of other sampling methods. For MSM, the main reasons why other sampling methods would not be adequate were that most of them were hidden due to social stigma and that no public venues existed to use as a sampling frame for TLS. For the FSWs, creation of a sampling frame for random sampling or TLS would be difficult and costly given their very high mobility and the illegal nature of sex work in PNG. For all three target groups in both locations, security for both the subjects and the research staff was a major concern. This is an issue specific to PNG, but may also be an important issue in other developing countries with unstable economic and political situations. The ability to conduct interviews in a private and secure location during the daytime and have potential subjects come directly to the study location after being recruited by previous respondents was a major advantage of RDS in this study. In PNG, it would have been difficult to have ensured the safety of the research staff if they had to go out into the field at night time or may have required hiring security guards at considerable expense. RDS has been described by those studying more the difficult sub–population of illicit drug users as a flexible and robust method that can produce a sample representative of the heterogeneity of the target population.16 We had a similar experience in implementing RDS among FSWs and MSM in a developing country. Although our original seeds were not as diverse as we intended them to be, a comparison of the seeds versus the final sample (Tables 1, 2 and 3) shows that by using RDS, the characteristics of the final samples were often quite different from the initially selected seeds. Particularly striking was our finding that the overall samples in all three subpopulations were mainly lower socio–economic status. A comparison of the RDSAT–adjusted versus the crude estimates of the various socio–demographic characteristics (Tables 1, 2 and 3) showed mostly little variation. The rapid rate of recruitment among all subpopulations suggests that the dual incentive system did work exceptionally well. We observed that the majority of the recruiters accompanied their recruits to the interview site. The reason for this was most likely that the recruiters wanted to collect their recruitment incentives immediately, but it also ensured that many of the recruited persons did in fact show up at the study sites in a relatively short amount of time. Not one FSWs or MSM subject used the phone number printed on the coupon to make an appointment. Telephone charges are relatively expensive, and mobile phones are not widespread among the general population in PNG. Therefore, it may have been easier and less expensive for recruits to simply show up at the research site than to have called ahead for an appointment. Thus the use of a telephone call to make appointments may not be necessary in this country in any future application of RDS for these or similar populations. The results of the comparisons between the RDS and hypothetical TLS samples (Tables 4 and 5) suggest that RDS allowed for greater penetration and representation of more vulnerable segments of the FSWs and MSM target populations. The use of dual financial incentives associated with RDS is also believed to have been important for reaching less educated and unemployed segments of the FSWs and MSM populations because they were more likely to find this a motivating factor for participation. During the RDS field work, it was observed anecdotally that the more economically desperate participants with ample disposable time were more likely to try to recruit individuals who did not meet the study eligibility criteria. Careful planning, staff training, and preparation for the field work enabled us to handle the onslaught of recruits and ensure the smooth implementation of RDS among FSWs and MSM in Papua New Guinea. The procedures checklist13 and practicing with the coupon management system were particularly important in preparing for the research. We agree with the view expressed by others that RDS is relatively easy to implement and less costly in that it did not require exhaustive mapping exercise to construct sampling frames.2,16 In our experience, the dual incentive system of RDS helped to reduce non–response rates and fuelled recruitment.13 Finally, in this study we relied on others to help select the original seeds. Some of the seeds failed to recruit any additional subjects. Half of the MSM seeds recommended to us simply did not recruit any other MSM, indicating that the recruitment of initial seeds in this study was not very effective. Problems were also encountered with potential recruits who misrepresented themselves in attempts to participate in the study in order to earn money. This was especially true in the MSM study, where many men who were not really MSM tried to participate. Similarly, a number of underage girls were refused entry into the study. This problem most likely resulted from the dual incentive system, in which a participant could potentially earn up to U.S. $15.00, a significant sum in a poor country like PNG. However, the screening and interview process in place was sufficient to track and preclude repeat participation or recruitment of individuals who did not meet the inclusion criteria. For this study among FSWs and MSM in PNG, using RDS as the sampling methodology had numerous advantages in terms of collecting the required sample size in a short time period, minimizing study costs, and resulting in representative estimates of the target populations. Although there were a few problems in implementing a new research method in a resource–poor setting, these were easily remedied, and we would recommend RDS for similar studies in PNG and other developing nations.

Ann_Surg_Oncol-3-1-1914254

Better Survival in Patients with Esophageal Cancer After Surgical Treatment in University Hospitals: A Plea for Performance by Surgical Oncologists

Background In primary esophageal cancer, studies have frequently focused on surgical patients in an effort to link outcome to hospital- or surgeon-related experience, with operative mortality used as the main outcome measure. Many studies have found an inverse relationship between operative mortality and hospital volume and surgical expertise. This study aims to assess the influence of surgeon-related expertise and hospital volume on the relative survival of operated esophageal cancer patients. With an incidence of 6.3 per 100,000 (European standardized rate) in the period 1994–1998, esophageal cancer ranked 13th among malignancies in men and 12th in women in the Netherlands. Incidence increased in both male and female patients, with an estimated annual percentage change of 3.1% and 2.0%, respectively. The mortality for men and women was, respectively, 9.4 per 100,000 and 3.1 per 100,000 in 2004, indicating that the prognosis of patients with esophageal cancer remained poor in this period (http://www.kankerregistratie.nl). The only curative option for esophageal cancer is surgery, which implies that improving the outcome of surgery is the best means of reducing mortality.1 Esophageal cancer is one of the most challenging pathological conditions confronting the surgeon. It therefore seems reasonable to assume that concentration of esophageal surgery could improve outcome. Several studies have shown that various characteristics, including surgeon subspecialty certification, hospital setting, and the number of procedures performed, are associated with practice variation, complication rates, and even outcome.2–4 Evidence of improved outcome associated with specialist care exists for breast cancer,2,4 ovarian cancer,3,5 and malignant teratoma.6 However, there is no evidence of comparable quality for esophageal cancers. Several studies on esophageal cancer have focused on surgical patients in an effort to link outcome to experience, either on the part of the institution or the surgeon, by using postoperative mortality as the main outcome measure.7–9 A study on esophagectomies performed in England during the late 1980s found no independent association between operative mortality and hospital surgical volume,10 whereas two North American studies did demonstrate a lower mortality rate for esophagectomy when high-volume and low-volume hospitals were compared.11,12 Various smaller studies also found lower operative mortality rates among high-volume surgeons.13,14 Few studies, if any, have attempted to relate esophageal cancer patient survival to surgeon expertise or hospital volume. To remedy this, in this study, we assessed the effect of surgeon-related expertise and hospital volume on the relative survival of operated esophageal cancer patients. We compared data from university, teaching nonuniversity, and nonteaching hospitals. PATIENTS AND METHODS Patients All patients diagnosed with a primary invasive esophageal cancer in the region of the Comprehensive Cancer Centre North-Netherlands (CCCN) between January 1994 and January 2002 were eligible for entry onto the study. Patients with a history of cancer other than nonmelanoma skin cancer were excluded. The patients were selected through the population-based Regional Cancer Registry of the CCCN, which covers the northern part of the Netherlands, a mainly rural area with a population of approximately 2.1 million. The area is served by 17 community hospitals, 3 of which are teaching hospitals and 1 of which is a university hospital; the hospitals include four radiotherapy departments and seven pathology laboratories. Data Collection by the Regional Cancer Registry PALGA, a Dutch nationwide network and registry of histopathology and cytopathology, regularly submits reports of all diagnosed malignancies to the cancer registry. The national hospital discharge data bank, which receives discharge diagnoses of admitted patients from all hospitals, completes case ascertainment. The cancer registry has no access to death certificates. After notification, trained registry personnel collect data on diagnosis and staging from the medical records, including pathology and radiology reports, in the hospitals. The cancer registry collected all data regarding the diagnosis and staging, but collected no data on specific surgical treatment in this patient population before 2002. The data collection occurs at least 4 months after diagnosis to comprehensively document all aspects of preoperative staging. All patients are staged according to the tumor, node, metastasis system (TNM) system for esophageal cancer in use during that period.15,16 In the Netherlands, the population registries of the municipality contain information on the vital status of their inhabitants. Vital status was established either through information derived from the patient’s medical records or through linkage of cancer registry data with information from the population registries of the municipality within the registry areas or through linkage with the national death registry of the Central Bureau of Genealogy. The regional cancer registry of the CCCN checked vital status by active record linkage with municipal population registries in 2002–2003 and 2005 and with the national death registry of the Central Bureau of Genealogy in 2004. Guidelines for Staging and Treatment By establishing multidisciplinary teams and cancer networks, the CCCN strives to improve the quality of cancer care. Within the CCCN area tumor working groups, comprising delegated specialists representing all regional hospitals, that have been developing and revising guidelines on diagnosis and treatment. The regional guidelines for esophageal cancer were based on the international TNM classification according to the International Union Against Cancer in use at that time. Statistics and Definitions of Variables The χ2 test was used to compare the distribution over the patient population for categorical variables. For continuous variables, analysis of variance was used. Relative survival analysis was performed to estimate the effect of university, teaching nonuniversity, and nonteaching hospitals on the prognosis of operated patients with esophageal cancer. The a priori hypothesis was that patient volume and hospital expertise would increase from nonteaching to university teaching hospitals. Survival time was calculated from the date of diagnosis and ended at the date of death, including perioperative death, or the date of most recent linkage with the municipal population registries and/or national death registry. The overall survival probability was estimated by the Kaplan-Meier method. The expected survival probability was calculated by using age-, sex-, and period-matched mortality rates that were based on life expectancy tables in the Netherlands (http://www.statline.cbs.nl/StatWeb/) and was estimated by the Ederer 2 method.17 The cumulative relative survival (the ratio of the overall survival to the expected survival) was estimated by Stata (version 8.0) software and the strs function. The relative survival, which estimates the net esophageal cancer survival in the hypothetical situation that esophageal cancer is the only possible cause of death, has been shown to be a good estimator of disease-specific survival in the absence of information on the cause of death or in case information on the cause of death is inaccurate. The excess mortality rate was calculated by subtracting the expected number of deaths, estimated from the expected survival probability, from the observed number of deaths in a subgroup or stratum and dividing the resulting excess number of deaths by the number of accumulated person-years, taking censoring into account. The relative excess risks (RER) of death were estimated as the ratio of excess mortality rates. RERs were estimated by a multivariate generalized linear model with a Poisson error structure, which was based on collapsed relative survival data, by using exact survival times.18 By use of this model, the effect of the type of hospital (university, teaching nonuniversity, and nonteaching) was studied, adjusting for the effect of various covariables on the excess mortality experienced by our cohort. Variables included in the final model were age (<50, 50–59, 60–69, >70), stage based on collapsed TNM data (stage 1,2a, 2b, 3/4, unknown), hospital volume (<20 patients operated, ≥20 patients operated), frequency of referral (high referral [>33.3%], low referral [≤33.3%]), and time since diagnosis (1-year intervals). The pathological stage was used whenever possible; in the absence of information about the pathological stage, the clinical stage was used. RESULTS TABLE 1. Characteristics of operated and nonoperated patients with esophageal cancer diagnosed 1994–2002 Characteristic Total Operated Nonoperated P value N % n % n % Sex .465 Male 796 69.1 152 71.4 644 68.6 Female 353 30.9 61 28.6 292 31.2 Histology <.001 Squamous cell carcinoma 415 36.1 62 29.1 353 37.7 Adenocarcinoma 592 51.6 140 65.7 453 48.4 Other 141 12.3 11 5.2 130 13.9 Tumor location <.001 Upper thoracic 82 7.1 3 1.4 79 8.4 Middle thoracic 210 18.3 36 16.9 174 18.6 Lower thoracic 770 67.0 169 79.3 601 64.2 Overlapping and unspecified 87 7.6 5 2.3 82 8.8 Age at diagnosis (y) <.001 <50 86 7.5 24 11.3 62 6.6 50–59 221 19.2 67 31.5 154 16.5 60–69 319 27.8 80 37.6 239 25.5 70+ 523 45.5 42 19.7 481 51.4 Stage <.001 1 52 4.5 32 15.0 20 2.1 2A 174 15.1 65 30.5 109 11.7 2B 69 6.0 26 12.2 43 4.6 3 207 18.0 74 34.7 133 14.2 4 316 27.5 8 3.8 308 32.9 Unknown 331 28.8 8 3.8 323 34.5 Total 1149 100.0 213 100.0 935 100.0 In the period 1994–2002, a total of 1149 patients were diagnosed with esophageal cancer, comprising 796 men and 353 (69.3%) women (30.7%). The median age was 68 years (range, 17–103 years). Patient characteristics are described in Table 1. A large proportion of the patients was diagnosed at an advanced stage of disease; 45.5% were stage III or higher. A further 28.8% were insufficiently staged. Of the 1149 patients, 85 patients (7.4%) were initially diagnosed in the university hospital, 428 patients (37.2%) in teaching nonuniversity hospitals. The remaining 636 patients (55.4%) were diagnosed in the nonteaching hospitals. Patients who were referred from the hospital of initial diagnosis for treatment may have subsequently undergone further diagnostic testing, and previously performed tests determined to be insufficient were repeated. The 5-year relative survival rate for men with esophageal cancer was 12.8% vs. 9.8% for women (P = .496). The 5-year relative survival markedly decreased as the stage advanced. The 5-year relative survival was 71.5% in stage I, 26.5% in stage IIA, 13.3% in stage IIB, 9.2% in stage III, 1.3% in stage IV, and 6.7% for patients with unknown stage (P < .0001). As Table 1 shows, only 213 patients (18.5%) underwent surgery. Older age (P < .001), advanced or unknown stage (P < .001), and proximal tumor location (P < .001) resulted in a lower probability of tumor resection. Squamous cell carcinoma was also associated with less surgery, but it was highly correlated with the tumor location. In all, 21.4% of the patients diagnosed in nonteaching hospitals underwent surgery, compared with 15.9% and 10.6% for teaching nonuniversity and university hospitals, respectively (P = .011). Adjusted for age, stage, and tumor location, the odds of operation was 1.89 (95% confidence interval [95% CI], 1.26–2.82) for patients diagnosed in a nonteaching hospital compared with patients diagnosed in a teaching nonuniversity hospital. Of all operated patients 45.1% were referred for surgery after diagnosis in the hospital of presentation (Table 2). The nonteaching hospitals referred 57.4% of patients diagnosed in their hospitals for an operation elsewhere. Of the 14 nonteaching hospitals in which esophageal cancer patients were diagnosed, 12 referred nearly all (75%–100%) patients, and two rarely (.0%–6.9%) referred patients. The teaching nonuniversity hospitals referred 26.5% of the patients diagnosed in their hospitals to a larger institution for therapy, with one hospital referring 63.2% of their patients. TABLE 2.Referral pattern for esophageal cancer surgery per hospital in the North-Netherlands, 1994–2002HospitalOperated in hospital of diagnosis, n (%)Referred for surgery, n (%)Total (n)High-referral nonteaching hospitals 20 (20.8)76 (79.2)96Hospital A–2 (100.0)2Hospital B–3 (100.0)3Hospital C–3 (100.0)3Hospital D–9 (100.0)9Hospital E2 (11.8)15 (88.2)17Hospital F1 (14.3)6 (85.7)7Hospital G3 (18.8)13 (81.3)16Hospital H3(25.0)9 (75.0)12Hospital I3 (30.0)7 (70.0)10Hospital J1 (33.3)2 (66.7)3Hospital K1 (33.3)2 (66.7)3Hospital L6 (54.5)5 (45.5)11Low-referral nonteaching hospitals 38 (95.0)2 (5.0)40Hospital M27 (93.1)2 (6.9)29Hospital N11 (100.0)–11High-referral teaching, nonuniversity hospitals7 (36.8)12 (63.2)19Hospital O7 (36.8)12 (63.2)19Low-referral teaching, nonuniversity hospitals43 (87.8)6 (12.2)49Hospital P24 (82.8)5 (17.2)29Hospital Q19 (95.0)1 (5.0)20Low-referral university hospital9 (100.0)–9Hospital R9 (100.0)–9117 (54.9)96 (45.1)213 Table 3 compares characteristics of the operated patients in the three hospital types. Of the 213 operated patients, 95 underwent surgery in the university hospital; 86 of these patients were referrals. The three teaching nonuniversity hospitals provided surgery to 60 patients, including 10 referrals: one hospital performed surgery on 7 patients, with the two other hospitals operating on more than 20. The remaining 58 patients underwent surgery at one of the 14 nonteaching hospital after all were initially diagnosed in the same hospital; two low-referral hospitals operated on more than 10 patients, eight hospitals operated on 5 or fewer patients, and the remaining four high-referral hospitals did not perform any esophageal cancer surgery. There were no statistically significant differences in the distribution of age (P = .230), stage (P = .299), or tumor location (P = .130) between the hospital types, again showing little evidence of selective referral. However, of all operated patients per hospital, a slightly larger proportion of stage III/IV tumors were operated on in the nonteaching (46.6%) and university hospitals (41.1%) compared with the teaching nonuniversity hospitals (26.7%), and the portion of stage IIA tumors was somewhat higher in nonteaching (31.0%) and teaching nonuniversity hospitals (38.3%) compared with the university hospital (25.3%). TABLE 3.Characteristics for operated esophageal cancer patients diagnosed 1994–2002, according to hospital of surgeryCharacteristicTotalTeaching, nonuniversityUniversityNonteachingP valueN%n%n%n%StageStage 13215.51220.01414.7610.3.299aStage 2A6526.81838.32325.32431.0Stage 2B2612.2813.31212.6610.3Stage 3 + 48238.51626.73941.12746.6Unknown87.011.766.311.7Age at diagnosis (y)<502411.3813.31313.735.2.23050–596731.51626.73637.91525.960–698037.62541.73031.62543.170+4219.71118.31616.81525.9HistologySquamous cell carcinoma6229.12135.02728.41424.1.606Adenocarcinoma14065.73558.36366.34272.4Other115.246.755.323.4Tumor locationUpper and middle thoracic3918.31321.71717.9915.5.130Lower thoracic16979.34778.37376.84984.5Not stated52.3––55.3––Total213100.060100.095100.058100.0a Excluding stage unknown. The cumulative relative survival for university, teaching nonuniversity, and nonteaching hospitals is shown in Fig. 1. Surprisingly, relative survival was markedly better in the university hospital compared with teaching nonuniversity and nonteaching hospitals. The 5-year relative survival was 49.2% for the university hospital versus 32.6% and 27.3% for teaching nonuniversity and nonteaching hospitals, respectively (P = .0039, Table 4). In univariate analysis, the RER for university and teaching nonuniversity was, respectively, .48 (95% CI, .30–.74) and .88 (95% CI, .55–1.42) compared with nonteaching hospitals. The proportion of operated patients who died within 3 months after diagnosis differed between university, teaching nonuniversity and nonteaching hospitals (4.2%, 13.3% and 19.0% respectively, P = .013). Excluding patients who died within the first 3 months, the RER for university and teaching nonuniversity was, respectively, .59 (95% CI, .35–1.00) and .97 (95% CI, .56–1.69) compared with nonteaching hospitals. FIG. 1.Cumulative relative survival of patients operated for esophageal cancer diagnosed during 1994–2002 according to hospital type.TABLE 4.Overall and relative 5-year survival and estimated excess risk (RER) of death with 95% confidence intervals (95% CI) for operated esophageal cancer patients diagnosed 1994–2002CharacteristicNOS (5 y)RS (5 y)OD (5 y)UnivariateMultivariate95% CIP valueED (5 y)RERa95% CIRERaStage<.0001Stage 1 3287.5%99.1%43.1.06.01–.32.05.01–.22Stage 2A6546.1%52.4%333.9.44.27–.70.39.24–.63Stage 2B2621.8%24.7%191.1.75.43–1.29.72.40–1.27Stage 3+4 (reference)8214.2%14.5%651.61.001.00Unknown816.7%18.0%6.2.99.41–2.351.62.65–4.01Hospital type.0126Nonteaching (reference)5824.9%27.3%402.31.001.00Teaching nonuniversity6029.7%32.6%391.9.89.55–1.421.32.79–2.22University9544.3%49.2%485.6.48.30–.77.57.29–1.12Age (y).0467<50 (reference)2447.6%48.6%11.21.001.0050–596732.1%33.6%421.21.73.87–3.441.51.74–3.0460–698030.9%35.1%523.52.121.07–4.182.361.18–4.7070+4241.9%54.5%224.91.42.64–3.142.05.94–4.46Hospital volume.1125<20 patients operated3819.0%22.3%271.31.001.00≥20 patients operated17537.8%41.7%1008.5.53.33–.83.62.34–1.12Referral rate .8080High (>33.3%)11535.4%39.2%665.21.001.00Low (≤ 33.3%) 9834.1%37.8%614.61.16.79–1.69.94.57–1.54OS, overall survival; RS, relative survival; OD, observed deaths; ED, expected deaths.a Adjusted for time since diagnosis. In a multivariate analysis to adjust for the prognostic effect of patient age, tumor stage, tumor location, hospital volume, frequency of referral, and time since diagnosis, we found that stage, age, hospital type, and time since diagnosis were independently associated with the RER (Table 4). The RER increased with more advanced stage. Patients aged <50 and patients aged ≥70 had a lower RER compared with patients aged 50–69 years. Adjusted for age, stage, and time since diagnosis, the RER for the university hospital was still considerably lower, at .57 (95% CI, .29–1.12), compared with nonteaching hospitals and .43 (95% CI, .24–.76) compared with teaching nonuniversity hospitals (P = .0126). There was some evidence in the data for an independent effect of hospital volume, with a lower RER (.62; 95% CI, .34–1.12) if a hospital operated on ≥20 patients during the study period. In our study 8.9% of the operated patients received some form of adjuvant therapy (Table 5). Of these patients, 6.1% received preoperative chemotherapy and 2.8% received postoperative radiotherapy. Patients receiving chemotherapy were all treated in the university hospital. Comparing clinical and pathological stage, we found no evidence that chemotherapy led to a marked downstaging. TABLE 5.Adjuvant therapies for patients operated for esophageal cancer diagnosed 1994–2002, according to hospital of treatmentTreatmentNonteachingTeaching, nonuniversityUniversityTotaln%n%N%N%Surgery5798.35896.77983.219491.1Surgery + radiotherapy11.7––55.362.8Surgery + chemotherapy––23.31111.6136.1Total58100.060100.095100.0213100.0 DISCUSSION Our study shows that patients who underwent surgery for esophageal cancer in the university hospital had a markedly better relative survival, with a 50% lower risk of death compared with patients treated in nonuniversity hospitals. The risk of death did not differ for patients operated in teaching nonuniversity or nonteaching hospitals. In our study, we found that a higher hospital volume was weakly associated with better survival. Although hospital volume seems to influence better survival, it is unlikely that the difference between the three hospital types can be completely explained by hospital volume; other factors likely play a role. A recent British study found no effect of hospital volume on survival for the operated patients; further, this study did not find teaching hospital status independently associated with survival, but it is unclear whether this reflects the operated patients or the population as a whole.10 Our study is one of the first studies to research the effect of a marker for hospital/surgeon experience on patient survival. Most studies to date have evaluated 30-day postoperative or in-hospital mortality, and several demonstrated an inverse relationship between volume or surgical experience and operative mortality after esophageal resection.8,11,19 A Dutch study found that hospitals performing 1 to 10 operations per year for esophageal cancer and cancer of the gastroesophageal junction had an operative mortality of 12.2%, compared with 4.9% for hospitals performing >50 procedures per year.19 An American study had a similar result, with a mortality of 3.0% among high-volume hospitals and 12.2% among low-volume hospitals for both distal and proximal esophageal cancer,12 but this study used a cutoff point of five procedures per year to differentiate between high- and low-volume hospitals. What threshold distinguishes high-volume from low-volume hospitals remains matter of discussion. The results of these studies do suggest that centralization of esophageal surgery, so that only a few hospitals per region operate on esophageal cancer patients, may improve survival. The results of our study support the recommendation for referral of esophageal cancer patients to a center where there is a specific focus on esophageal cancer treatment. Combined with the fact that the surgical literature is increasingly advocating the need for centralization, we think that further research into the advantages of centralization of esophageal cancer treatment is warranted. One of the possible pitfalls in our study remains selective referral. We showed that the likelihood of being operated on was 47% higher for patients diagnosed in a nonteaching hospital than for those diagnosed in teaching nonuniversity hospitals. Adjusted for age, stage, and tumor location, this was even higher, 73%, indicating that nonteaching hospitals considered patients with a worse prognosis possible candidates for surgery, or at least for referral to evaluate resectability. However, adding hospital referral frequency to our multivariate analyses showed no influence of referral on overall and relative survival or RER. The referral pattern showed little evidence for selective referral. Those nonteaching hospitals that referred patients referred almost all, and the teaching nonuniversity hospitals referred only few patients for surgery (Table 2). The referral pattern of the nonteaching hospitals implies that the university hospital operated on a priori prognostically worse patients. This is a likely explanation for the higher number of stage III patients in the university hospital. To minimize the eventual effect of any residual selection referral, the relative survival rate was adjusted for case mix, despite there being no statistically significant differences in the distribution of age, stage, and sex between the different types of hospital. One may suggest that patients who underwent esophagectomy in the university hospital were mostly referred for treatment, thus adding a delay before surgery. A consequence could be that these patients had thus a slightly longer preoperative survival time, estimated to be between 2 and 4 weeks in our study. This short delay could mean that prognostically worse patients scheduled for surgery eventually fall out of the surgery category through disease progression during the delay period. However, little is known in the literature about the effect of longer preoperative delays on surgical outcome or eligibility in esophageal cancer. Although patients with advanced disease may miss surgery through stage progression, patients who do end up having surgery also progress, meaning the university hospital operates on patients with more advanced disease. This should negatively influence the survival outcome and would not explain the better performance by the university hospital. Furthermore, even if we were very conservative and excluded all patients who died in the first 3 months of our study, the university hospital still performs far better than teaching nonuniversity and nonteaching hospitals. So although we cannot discount early mortality as a factor in survival, we think that it is unlikely that the difference in performance can be fully explained by this. We had no information about the operative procedure that had been performed. Treatment guidelines indicated a curative surgical approach for tumors encompassing ≤5 cm of the length of the esophagus, as based on ultrasonographic or radiological examination. Surgical resection could be attempted for tumors 5–8 cm in length. For this last group, neoadjuvant chemotherapy, which was provided in the university hospital after proof of locally advanced disease, could be attempted to improve resectability. For adenocarcinomas, a transthoracoabdominal approach with two-field lymphadenectomy was advised, combining a midline laparotomy and a right-sided thoracotomy. Alternatively a transhiatal blind esophagectomy could be performed with a cervical esophagogastrostomy. For distal adenocarcinomas without Barrett dysplasia, a left-sided thoracotomy with intrathoracic anastomosis was an alternative approach. In general, patients in our population who were operated on in the university hospital underwent a transthoracic esophagectomy, with the exception of superficial T1 tumors, whereas patients treated in regional hospitals frequently underwent surgery with a transhiatal approach. Therefore, the two main operative strategies we encountered were transhiatal resection and transthoracoabdominal resection with a two-field lymph node dissection. There is no evidence in literature that the outcome differs for these two procedures,14,20 except for a tendency toward an improved long-term survival in the extended transthoracic group in the study of Hulscher et al.21 So the clinical outcome in our population is likely uninfluenced by differences in surgical procedure. However, there is possibly a stage migration effect between the more thorough pathological staging in operations with lymphadenectomy, as in the university hospital, and understaging in patients undergoing a transhiatal esophageal resection. A few of the operated patients received neoadjuvant therapy in our study, mostly in the university setting, which might account for a small part of the better survival in the university hospital. Separate analysis, however, showed that patients in our study, who received neoadjuvant therapy with surgical removal of the tumor, did not perform better than patients who underwent surgical removal alone in the university hospital. In other studies, the preoperative effect of cisplatin-based chemotherapy on both adenocarcinoma and squamous cell carcinoma showed no increase in overall survival.22–25 Preoperative chemotherapy or radiotherapy can result in downstaging and thus lead to a better resectability, but no clear downstaging was seen in our study when comparing pre- and postoperative clinical and pathological stage. In several randomized, controlled studies, postoperative radiotherapy demonstrated either no increase26,27 or a decrease28 in survival compared with resection alone. Postoperative chemotherapy has also been compared with surgical management alone in several randomized controlled trials, without demonstrating an improvement in survival.29 According to these results, the effect of perioperative treatment is not likely to influence our data. We found a tentative relationship between higher volume and a better relative survival. However, this issue still is a debatable problem in determining treatment guidelines. Therefore, we suggest that guidelines concerning specific referral of esophageal cancer patients should be based on hospital outcomes, preferably in experienced centers, rather than on annual numbers of procedures as long as the factor that is determining patient survival is still unknown. The individual surgeon could be an important parameter in determining the hospital outcome. Although the implications of the assertion that some surgeons have better outcomes than others make clinicians uncomfortable, there should be little doubt that it is true. Variation in performance has been shown to be related to surgeon characteristics, including surgical volume, subspecialization, and the hospital setting in which they operate.7,30 Individual surgical experience has been associated with the postoperative mortality of esophageal cancer. Sutton et al.9 showed a reduction in mortality from 6% to 3% after 150 procedures. Miller et al.14 published results in one center demonstrating an operative mortality of 22% among esophageal resections performed by a surgeon who performed fewer than six procedures a year. These results are widely quoted in the surgical literature as proving that surgeons without the necessary expertise should not perform esophageal resections. There is some evidence that subspecialization improves outcomes. Herr et al.31 found that patients who underwent radical cystectomy by urology oncologists had substantially lower rates of local tumor recurrence than those who were operated on by general urologists (6% vs. 23% (P = .006). Dueck et al.32 reported that patients who underwent surgery for a ruptured abdominal aortic aneurysm had markedly better outcomes when the surgery was performed by a vascular surgeon rather than a general surgeon. The effect of subspecialization of the surgeon on the outcome of esophageal cancer has not yet been studied, but it may be a promising factor for decisions with regard to centralization. It has been suggested in previous reports that the skill of the anesthesia and nursing staff affects morbidity and relative survival of esophagectomy patients and that it confounds the surgeon’s personal outcome.7,9 Better critical-care experience of the support staff may explain a higher relative survival in university hospitals; staff may be more adept at caring for esophagectomy patients. Some authors have suggested that the expertise of the anesthesia and nurses in a hospital is directly correlated to the hospital and surgical load in that hospital.9 However, expertise can be acquired elsewhere, and expertise only develops through effective feedback, not only by number of patients. Although referral to dedicated centers possibly results in improved relative survival, the focus entails some disadvantages, which should be considered. Referral to centers means that many patients have to travel to distant sites, which can create hardship for the patient and his or her family. In-hospital family support and postoperative follow-up are more difficult when the hospital is farther away. Finlayson et al.33 demonstrated that 45% of the patients prefer to stay in their local area even if the projected operative mortality is doubled. However, that study represents the American situation, and it is questionable whether distance is perceived to be a problem in the Dutch situation. In our region, which has a relatively high density of hospitals, a recent patient survey showed that traveling distance was not considered a critical issue. In conclusion, we demonstrated that in our region, the relative survival for patients operated on for esophageal cancer is better in the university hospital compared with teaching nonuniversity and nonteaching hospitals, emphasizing the need for referral to centers focused on the treatment of esophageal cancer. The underlying parameter for the observed difference remains unclear. We suggest that centers at least periodically review the morbidity and mortality rates of esophageal resections to assess their outcome and the possibility of referral. Eligibility for centers focused on esophageal cancer treatment should therefore be based on patient outcomes rather than on patient numbers.

Purinergic_Signal-2-3-2096647

Excitatory effect of ATP on rat area postrema neurons

ATP-induced inward currents and increases in the cytosolic Ca2+ concentration ([Ca]in) were investigated in neurons acutely dissociated from rat area postrema using whole-cell patch-clamp recordings and fura-2 microfluorometry, respectively. The ATP-induced current (IATP) and [Ca]in increases were mimicked by 2-methylthio-ATP and ATP-γS, and were inhibited by P2X receptor (P2XR) antagonists. The current–voltage relationship of the IATP exhibited a strong inward rectification, and the amplitude of the IATP was concentration-dependent. The IATP was markedly reduced in the absence of external Na+, and the addition of Ca2+ to Na+-free saline increased the IATP. ATP did not increase [Ca]in in the absence of external Ca2+, and Ca2+ channel antagonists partially inhibited the ATP-induced [Ca]in increase, indicating that ATP increases [Ca]in by Ca2+ influx through both P2XR channels and voltage-dependent Ca2+ channels. There was a negative interaction between P2XR- and nicotinic ACh receptor (nAChR)-channels, which depended on the amplitude and direction of current flow through either channel. Current occlusion was observed at Vhs between −70 and −10 mV when the IATP and ACh-induced current (IACh) were inward, but no occlusion was observed when these currents were outward at a Vh of +40 mV. The IATP was not inhibited by co-application of ACh when the IACh was markedly decreased either by removal of permeant cations, by setting Vh close to the equilibrium potential of IACh, or by the addition of d-tubocurarine or serotonin. These results suggest that the inhibitory interaction is attributable to inward current flow of cations through the activated P2XR- and nAChR-channels. Introduction The rat area postrema (AP) is a medullary circum-ventricular organ located in the hindbrain at the level of the obex, with a dense vascular supply devoid of a blood-brain barrier. The lack of a blood-brain barrier makes the AP ideally placed to act as a chemosensitive trigger zone involved in the control of vomiting in response to circulating emetic substances (Borison [1, 2]). Neurons in the AP are also responsive to changes in osmolarity or sodium concentration (Franchini et al. [3]), and can be activated by circulating vasoactive peptides such as angiotensin II (Fink et al. [4]), and arg-vasopressin (Undesser et al. [5]). Anatomical studies have revealed that the AP sends dense efferent projections to the nucleus tractus solitarius, parabrachial nucleus, nucleus ambiguus, and the dorsal motor nucleus of the vagus, and receives afferent inputs from the hypothalamic paraventricular and dorsomedial nucleii, and from the caudal nucleus tractus solitarius (Morest [6]; Kooy and Koda [7]; Shapiro and Miselis [8]). Thus, the AP is not only capable of responding to circulating hormones, but is also anatomically well situated to regulate a range of other central neurons, including those important in cardiovascular control (Sun and Spyer [9]). The low intrinsic firing rates of AP neurons in vivo (Papas et al. [10]) suggests that understanding and modulating excitatory inputs to AP could be particularly important in the functional output of AP neurons. A number of transmitters can evoke excitatory currents in the AP. Inward currents and increases in cytosolic Ca2+ ([Ca]in) via non-NMDA-glutamate receptors have been reported in rabbit and rat AP neurons, respectively (Jahn et al. [11]; Hay and Lindsley [12]). Our preliminary reports indicated that ATP also induces inward currents and [Ca]in increases via the activation of P2X receptor (P2XR) in isolated rat AP neurons (Sorimachi et al. [13, 14]). In addition, pre- and post-synaptic nicotinic ACh receptors (nAChR) have been demonstrated in the AP in rat brain slices (Funahashi et al. [15]), and we have recently reported the presence of nAChRs in dissociated rat AP neurons (Sorimachi and Wakamori [16]). During that study, we also noticed that many of these AP neurons also responded to ATP, which has prompted us to further investigate ATP responses in AP neurons, and potential interactions between nAChR and P2XR responses. In a variety of different peripheral neurons, including sympathetic neurons of bullfrog (Akasu and Koketsu [17]), rat (Nakazawa et al. [18]), and guinea-pig (Searl et al. [19]), cultured guinea-pig enteric and submucosal neurons (Zhou and Galligan [20] Barajas-Lopez et al. [21]), a negative interaction between P2XR- and nAChR responses has been reported. Such an interaction has also been observed for recombinant P2X2 and α3β4 nAChR channels in Xenopus oocytes and HEK cells (Khakh et al. [22, 23]; Boue-Grabot et al. [24]), where it has been recently suggested that this results from direct physical interactions between co-localized receptors (Khakh et al. [23]). In this study we more fully describe P2XR responses in AP neurons, demonstrate cross-inhibition between P2XRs and nAChRs and characterize some of the features of this cross-inhibition. Experimental procedures Preparation of AP neurons The study was approved by the Committee on Animal Experimentation, Kagoshima University. Wistar rats (13–18 days-old) were anaesthetized with ether and decapitated. The brain was quickly removed from the skull and placed in ice-cold HEPES-buffered saline containing 150 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM HEPES and 5.5 mM glucose. The pH of the saline solution was adjusted to 7.4 by adding tris (hydroxymethyl) aminomethane. The brain was sliced at a thickness of 400 µm with a microslicer (DTK-1000, Dosaka, Kyoto, Japan), and the slices were kept in bicarbonate-buffered saline bubbled continuously with 95% O2–5% CO2 at room temperature (21–26 °C). The bicarbonate-buffered saline contained 120 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 20 mM NaHCO3, 2 mM KH2PO4 and 5.5 mM glucose. Following 30–60 min incubation, the slices were treated with pronase (0.2 mg/ml) in bicarbonate-buffered saline for 20 min at 33 °C, followed by incubation for 30 min with bicarbonate-buffered saline containing thermolysin (0.12–0.15 mg/ml). The bilateral AP regions were identified with a binocular microscope (Zeiss, Germany), and were cut out using the tip of an injection needle, and subsequently mechanically triturated with fire-polished glass pipettes of decreasing diameters. Dissociated neurons were placed on to the bottom of 35 mm culture dishes (Falcon, USA) for electrophysiological recordings, or on to glass coverslips (Matsunami, Japan) coated with poly-l-lysine for [Ca]in measurements. Electrophysiological recordings Electrical measurements were done using the whole-cell patch-clamp recording configuration under voltage-clamp conditions. Patch pipettes were fabricated from borosilicate glass tubes in five or six stages using a pipette puller (Model P-97, Sutter Instrument, San Rafael, CA, USA). The resistance between the recording electrode filled with the internal solution, and the reference electrode in the external solution, was 3–6 MΩ. Ionic currents were measured, and voltages controlled, using a patch-clamp amplifier (EPC-9, Heka, Darmstadt, Germany). All experiments were carried out at 24–26 °C. Culture dishes were placed on an inverted microscope (TE200, Nikon, Japan), and drugs were rapidly applied to single cells using a Y-tube perfusion device. The internal solution (patch pipette solution) contained 70 mM K-gluconate, 50 mM KCl, 10 mM NaCl, 0.5 mM CaCl2, 3 mM MgCl2, 10 mM HEPES, 10 mM EGTA, and 2 mM ATP. The pH was adjusted to 7.2 with KOH. For measuring current-voltage (I–V) relationships, a Cs-based internal solution was employed, which contained 98 mM CsOH, 40 mM CsCl, 98 mM aspartate, 2 mM MgCl2, 5 mM HEPES, 5 mM EGTA, and 2 mM ATP. pH was adjusted to 7.2 with CsOH. The external solution was the HEPES-buffered saline, pH7.4, described above. When N-methyl-d-glucamine (NMDG+) or sucrose was substituted for external Na+ or NaCl, respectively, the osmolarity of the solution was kept constant as measured using an osmometer (Vogel OM801, Germany). To obtain the Ca2+/Na+ permeability ratio (PCa/PNa), we measured the reversal potential of the ATP-activated current (EATP) in the presence of 1 or 110 mM of external Ca2+ by stepping the holding potential (Vh) between 10 and 30 mV, using increments of 10 mV. The 1 and 110 mM Ca2+ solutions contained 155 or 0 mM NaCl, respectively, in addition to 10 mM HEPES with pH adjusted to 7.4 with NaOH or Ca(OH)2, respectively. The values of PCa/PCs and PNa/PCs were calculated using the constant field equation as described by Lewis ([25]), taking the activity coefficients of Na+, Cs+, and Ca2+ as 0.75, 0.75, and 0.3, respectively. Measurement of [Ca]in Dissociated AP neurons on a glass coverslip were incubated with 1–2 µM fura-2 acetoxymethyl ester (fura-2/AM), 0.1% dimethyl sulfoxide, and 1% bovine serum albumin for 45 min at 37 °C. The coverslips were then mounted in a superfusion chamber and placed on the stage of an inverted microscope (Diaphot-TMD, Nikon, Japan). Cells were continuously superfused at a rate of 1 ml/min with HEPES-buffered saline at 24–26 °C via a polyethylene tube placed 1–2 mm away from the cells. Cells were viewed with a 40× Fluor objective lens (Nikon), and a single cell (10–12 µm diameter) was fixed in a window positioned between the photomultiplier and the microscope. The changes in fluorescence ratios at 340 and 380 nm excitation wavelengths were measured using a CAM-200 spectrometer (Jasco, Japan). The absolute value of [Ca]in was calculated using the formulae as given by Grynkiewicz et al. [26]: . Calibration constants were determined in separate experiments with the same experimental set-up, as described previously (Sorimachi et al. [27]). Student’s paired t test was used to evaluate differences between mean values obtained from the same cells, and Student’s unpaired t test was used for data obtained from different groups of cells. Reagents The following reagents were used: Fura-2/AM [Dojindo, Kumamoto, Japan], ACh [Horai Chem. Co., Japan], PPADS (pyridoxal-phosphate-6-azophenyl-2′, 4′-disulphonic acid) [RBI, USA.], pronase [Calbiochem., USA], thermolysin, ATP, 2-methylthio-ATP, ATPγS, α,β-methylene-ATP, β,γ-methylene-ATP, ADP, nitrendipine, nicardipine, suramin [all from Sigma, Aldrich, Tokyo, Japan], ω-conotoxin-MVIIA, ω-conotoxin-MVIIC and ω-agatoxin-IVA [all from Peptide Institute, Osaka, Japan]. Results ATP-induced current Rapid application of ATP (100 µM) to isolated AP neurons voltage clamped at a Vh of −70 mV induced an inward current, which desensitized slowly (Fig. 1A). The ATP-induced current (IATP) was recorded at various Vhs between −70 and +50 mV (Fig. 1: N = 6), and the relative current amplitude was plotted against Vh exhibiting a strong inward rectification. The EATP, estimated from the intersection of the current response and the zero voltage-axis (Fig. 1B), was 22.7 ± 0.9 mV (N = 18). Fig. 1Current-voltage relationship for ATP-induced currents. A) Representative currents in response to 200 µM ATP recorded at Vhs between −70 and +50 mV. B) Current-voltage relationship for the IATP. All responses are normalized to the peak current amplitude obtained at a Vh of −70 mV, and each point is the mean ± S.E.M. of five neurons. C) Averaged concentration-current relationships for ATP in the presence of 0.1, 2, and 10 mM external Ca2+. All responses were normalized to the mean of two control responses induced by 100 µM ATP in saline containing 2 mM external Ca2+ before and after the test response. Each point is the average ± S.E.M of responses from four to six neurons. To investigate the ATP concentration-response relationship, care was taken to adjust the pH of the solutions containing higher concentrations of ATP to 7.3, since addition of ATP reduced pH and a small decrease in pH to 7.1 increased the IATP to 189 ± 8% (N = 6) of control (pH7.3), as previously reported for recombinant P2X2 receptors (King et al. [28]). The responses to different ATP concentrations were flanked by responses to the control concentration of 100 µM, and normalized responses were expressed relative to the average of these control responses (Fig. 1C; Ca2mM). We also examined the effects of 0.1 and 10 mM Ca2+ on IATP. As summarized in Fig. 1C, an increase in the external concentration of Ca2+ shifted the concentration-response curve for ATP to the right, with the half-maximum effective concentration (EC50) values at 0.1, 2 and 10 mM Ca2+ being 30, 70, and 190 µM, respectively. The purinergic agonists, 2-methylthio-ATP (50 µM) and ATPγS (50 µM) induced currents that were 41 ± 3% (N = 4), and 41 ± 4% (N = 5), respectively, of the IATP in response to 100 µM ATP. ADP (0.5 mM) did not produce any response. The IATP in response to 50 µM ATP was inhibited by the P2 antagonists; suramin (10 µM and 20 µM) and PPADS (50 µM), with the response being 35 ± 2% (N = 5), 25 ± 2% (N = 4), and 56 ± 5% (N = 5), of the control IATP, respectively. ATP-induced [Ca]in increase ATP also increased [Ca]in in a dose-dependent manner (Fig. 2A). These responses were all recorded in the presence of 2 mM external Ca2+, and the [Ca]in increases by various concentrations of ATP were normalized to that induced by 100 µM ATP. The normalized and averaged responses to ATP, 2-methylthio-ATP and ATPγS are shown in Fig. 2B. There was no ATP-induced [Ca]in increase in the absence of external Ca2+ (N = 6; data not shown). Furthermore, neither ADP, α,β-methylene-ATP nor β,γ-methylene-ATP increased [Ca]in when tested at concentrations of 200 µM(data not shown). The ATP-induced [Ca]in increases were inhibited by suramin and PPADS (Table 1). Fig. 2Concentration-response relationships for the [Ca]in increases in response to ATP and ATP analogues. A) Representative [Ca]in increases induced by various concentrations of ATP in the presence of 2 mM external Ca2+. B) Averaged concentration-response relationships for ATP and ATP analogues. All responses were normalized to the mean of two control responses induced by 100 µM ATP before and after the test response. Each point is the mean ± S.E.M. of data from five to nine neurons. MeS-ATP: 2-methylthio-ATP.Table 1Effects of P2 receptor antagonists and Ca2+ channel blockers on the [Ca]in increases induced by ATP and high K+-salineStimulusBlockersPercent of controlATP100 MSuramin 10 M55 ± 7 (N = 5)20 M30 ± 5(N = 6)50 M13 ± 3 (N = 5)PPADS 10 M96 ± 6 (N = 3)20 M70 ± 8 (N = 9)50 M25 ± 5 (N = 6)100 M11 ± 5 (N = 4)110 mMKClNitrendipine 2 M44 ± 4 (N = 16)−CT.M C 2 M61 ± 4 (N = 16)−CT.M A 2 M62 ± 6 (N = 16)−CT.M A 2 M + −CT.M C 2 M58 ± 5 (N = 10)ATP100 MNitrendipine 2 M56 ± 4 (N = 21)Nicardipine 2 M47 ± 9 (N = 10)Cd2+ 50 M27 ± 3 (N = 14)−CT.M A 2 M + −CT.M C 2 M77 ± 9 (N = 4)The mean [Ca]in increase induced by the first and third applications of ATP (100 μM) or 110 μM KCl were averaged and referred to as 100%, and the response in the presence of the P2 antagonist or Ca2+ channel blocker was expressed as a percentage of this control value. The cell was pre-treated for 30 s with the indicated agent before the second stimulation with 100 μM ATP or 110 mM KCl. Number of experiments is shown in parentheses. ω-CT ω-conotoxin. Effects of Ca2+ channel antagonists on ATP-induced [Ca]in increase Previous results, using cultured rabbit AP neurons, demonstrated the presence of a ω-conotoxin-M Csensitive Ca2+ response, but that did not involve L- or N-type Ca2+ channels (Hay et al. [29]). We also investigated the effects of various Ca2+ channel antagonists on [Ca]in increases induced by high KCl (110 mM) and ATP (100 µM). Antagonist for the L-type (nitrendipine, nicardipine), N-type (ω-conotoxin-M A), or P/Q-type (ω-conotoxin-M C) Ca2+ channels each substantially inhibited these [Ca]in increases (Table 1). In contrast, the selective P-type Ca2+ channel antagonist, ω-agatoxin IVA (1 µM), did not have any inhibitory effect on the [Ca]in increases (98 ± 5% of control, N = 4). ATP-induced current in the absence of external Na+ When external Na+ was completely replaced by NMDG+, and in the absence of external Ca2+ (0 Ca2+ plus 0.5 mM EGTA), the IATP was markedly reduced to 8 ± 1% of the control IATP recorded in the presence of external Ca2+ and 150 mM Na+ (N = 9, Fig. 3A). This current was further reduced to 4 ± 1% (N = 3) in the presence of 20 µM suramin. The small suramin-sensitive, Na+-independent current supports the previous suggestion that the P2XR channel is permeable to glucosamine (Nakazawa [30]). To confirm this, ATP did not induce a current at all when sucrose (0.25 M) was substituted for NaCl (again in the absence of external Ca2+, N = 5). Addition of Ca2+ to the NMDG+-substituted saline further increased IATP. The currents induced by 500 µM ATP in the presence of 2 and 10 mM Ca2+, but in the absence of external Na+ (replaced by NMDG+) were 18 ± 2% (N = 11) and 29 ± 2% (N = 14, Fig. 3B) of the control IATP, respectively. Similarly, the currents induced by 500 µM ATP in the presence of 2 and 10 mM Ca2+ added to the sucrose-substituted saline were 6 ± 1% (N = 6), and 13 ± 2% (N = 16; Fig. 3C) of the control IATP, respectively. Fig. 3ATP-induced current in Na+- and NaCl-free saline containing 0 and 10 mM Ca2+. A) A representative IATP recorded in Na+-free (Na+ replaced by NMDG+) and Ca2+-free saline (containing 0.5 mM EGTA) at a Vh of −70 mV. B) A representative IATP in Na+-free (NMDG+) saline containing 10 mM Ca2+. C) A representative IATP in NaCl-free (NaCl replaced by sucrose) saline containing 10 mM Ca2+. We also measured the EATP of IATP in the presence of 1 mM external Ca2+ and 150 mM NaCl or 110 mM external Ca2+ and 0 mM NaCl. The EATP at 110 mM Ca2+ was 21.1 ± 1.0 mV (N = 18), from which we calculated a PCa/PCs of 6.3. The EATP at 1 mM Ca2+ was 20.2 ± 1.0 mV (N = 13), from which we calculated a PNa/PCs ratio of 2.1. From these, we obtained a PCa/PNa ratio of 3.0, confirming the substantial Ca2+ permeability of P2XRs. Negative interaction between P2XR and nAChR channels Requirement of actual current flow through receptor channels for cross-inhibition It has previously been reported that there is mutual occlusion between P2XR and nAChR in some neurons (Nakazawa et al. [18]; Nakazawa [30]; Searl et al. [19]; Zhou and Galligan [20]; Barajas-Lopez et al. [21]; Khakh et al. [22, 23]; Boue-Grabot et al. [24]). To examine whether there were negative interactions between the ACh-activated current (IACh) and IATP, one receptor agonist was added in the presence of the other. As shown in Fig. 4A, the IACh (100 µM) was markedly reduced when activated in the presence of ATP (50 µM). We next examined IACh (100 µM) in the presence of various concentrations of ATP. With 2, 10, 20, and 100 µM ATP, IACh was reduced to 95 ± 1% (N = 20), 84 ± 2% (N = 24), 60 ± 4% (N = 15), and 46 ± 6% (N = 7), respectively, of control (P < 0.01 except at 2 µM ATP). Thus, the IACh inhibition became stronger as the concentration of ATP was increased and with a higher agonist-receptor occupancy. In fact, when IATP at 100 µM ATP was markedly inhibited in the presence of 200 mM PPADS (8 ± 2% of control, N = 6), IACh was not occluded (97 ± 1% of control; N = 6). When the peak amplitude of IATP, evoked by various concentrations of ATP, was plotted against the ratio of IACh in the presence and absence of ATP, there was an inverse correlation between the amplitudes of these responses (Fig. 4B). Conversely, when ATP (50 µM) was applied in the presence of ACh (100 µM), IATP was also occluded (Fig. 4C). Again, when IACh at 200 mM ACh was nullified in the presence of 1 mM hexamethonium, a nAChR antagonist (0.2 ± 0.2% of control, N = 6), IATP was 96 ± 2% (N = 6) of control. IACh at 10, 20, 50, and 200 µM ACh were 6 ± 2% (N = 5), 20 ± 4% (N = 5), 61 ± 8% (N = 6), and 139 ± 8% (N = 6) of IACh at 100 µM. There was also an inverse correlation between the peak amplitude of IACh, evoked by various concentrations of ACh, and the ratio of IATP in the presence and absence of ACh (Fig. 4D). However, the inhibition of IATP by nAChR activation was weaker than that of IACh by P2XR activation. As the IACh desensitized faster than the IATP, the weaker inhibition of IATP by ACh could be due to the reduced amplitude of the IACh at the time of ATP application. Fig. 4Inhibition of IACh and IATP, respectively, in the presence of ATP and ACh. A) Sample currents in response to ACh (100 µM) in the presence of ATP (50 µM). The responses to ACh applied in the presence of ATP were reduced. The extent of occlusion was normalized to the control ACh responses recorded before and after ATP application. This sequence was routinely repeated twice or three times, and the averaged amount of occlusion was plotted in B. B) Relative amplitude of the ACh current amplitude in the presence of ATP (ordinate) plotted against the peak amplitude of IATP obtained in the same neuron (abscissae). Note the negative correlation between these two variables. The line was drawn using KaleidaGraph (Synergy Software). Vh: −70 mV. C) The IATP was recorded in the presence of 100 µM ACh and this response was flanked by control ATP alone responses. This sequence was routinely obtained twice or three times in the same cell, and the averaged amount of occlusion was plotted in D. D) The ratio of the ATP current amplitude in the presence and absence of ACh (ordinate) is negatively correlated with the peak amplitude of IACh (abscissae). Vh : −70 mV. To examine this possibility, IATP was measured in the presence of both ACh and d-tubocurarine (dTc, 10 µM) or serotonin (5HT, 50 µM), two experimental conditions which change the time course of the IACh response. The former agent, dTc, has been shown to be both an open channel blocker and a competitive antagonist of nAChR (Manalis [31]), and, as expected, a low concentration (10 µM) slightly reduced the peak amplitude of IACh (79 ± 4% of control, N = 11) and caused the IACh to be terminated within 0.5 s (Fig. 5A, insert). Once the IACh response had returned to baseline (in the presence of ACh and 10 µM dTc) and ATP was applied, there was no inhibition of IATP (95 ± 1% of control, N = 5; Fig. 5A). The IATP in the presence of ACh alone was 77 ± 4% of control (N = 5, P < 0.01; Fig. 5A). When we measured the current (IACh+ATP+dTc) induced by a combination of ACh, ATP and dTc, the peak amplitude of IACh+ATP+dTc was occluded under these conditions, being 74 ± 2% (N = 10) of the predicted sum of IATP and IACh+ dTc (Fig. 5B). The level of IACh+ATP+dTc at 0.5 s following ligand application, which corresponds to a time when the nAChR channels are completely blocked (Fig. 5A, insert), was 96 ± 3% (N = 8) of the control IATP, measured immediately before and after (Fig. 5C). These results suggest that the inhibition of IATP by IACh disappears immediately after the IACh is abolished, although ACh is still present and bound to the nAChRs. 5HT has been shown to also accelerate the decay of IACh (Grassi et al. [32]; Garcia-Colunga and Miledi [33]; Sorimachi and Wakamori [34]), but not to the same extent as dTc. 5HT at 50 µM decreased the peak amplitude of IACh slightly, to 88 ± 4% (N = 6) of control, and reduced the time constant (tau) of IACh decay from 2.02 ± 0.18 to 1.06 ± 0.19 s (N = 6). As shown in Fig. 5D, the extent of inhibition of IATP by ACh was smaller (to 92 ± 2% of control, N = 6) in the presence of ACh and 5HT compared to that in the presence of ACh alone (to 73 ± 3% of control, N = 6; P < 0.01). Fig. 5Loss of the inhibition of IATP by ACh in the presence of dTc and 5HT. A) Representative IATP in control conditions (applied alone), in the presence of ACh and in the presence of ACh and dTc. This sequence was routinely repeated twice, and the averaged response was obtained. Note the very transient nature of IACh in the presence of dTc, with the current returning to baseline levels within 0.5 s (insert). Note also the lack of occlusion under these conditions. Vh: −70 mV. B) Effect of dTc on IACh+ATP. The sequence of applications was repeated multiple times in each neuron. The peak amplitude of IACh+ATP+dTc was smaller than the predicted sum of IATP and IACh+dTc. C) Rapid recovery of IATP from inhibition by ACh in the presence of dTc. The amplitude of IACh+ATP+dTc 0.5 s following application was equal to the mean of two control IATP at an equivalent time. Inset shows the rapid decline of IACh+ATP+dTc on an expanded time scale. D) Representative IATP in control conditions, in the presence of ACh, and in the presence of both ACh and 5HT. In this example, the tau of IACh in response to the first application was 2.7 s, while the tau of IACh+5HT was 0.9 s, showing the acceleration of the IACh decay by 5HT. Note the reduced occlusion under these conditions. Vh : −70 mV. Next, we measured the current induced by the concomitant applications of ACh and ATP (IACh+ATP) at a Vh of −70 mV. As shown in Fig. 6A, IACh+ATP was 75 ± 1% (N = 44; P < 0.01) of the predicted sum of IACh and IATP. The decay phase of IACh, IATP and IACh+ATP were fit to a single exponential function, giving tau of 3.0 ± 0.2, 7.5 ± 0.6 and 3.7 ± 0.2 s, respectively. When the decay of IACh was accelerated by the presence of 100 µM 5HT, the tau of IACh, IATP, and IACh+ATP were 0.8 ± 0.3 s (N = 7), 7.4 ± 1.4 s (N = 7) and 2.6 ± 0.8 s (N = 7), respectively (the amplitude of IACh+ATP under these conditions was 76 ± 2% of the predicted sum of IACh and IATP). Thus, IACh+ATP desensitization occurs with kinetics that cannot be explained by the desensitization kinetics of IACh or IATP alone, suggesting that IACh+ATP is mediated via both receptors. Fig. 6Voltage dependence of IACh and of IATP occlusion. A) At a Vh of −70 mV, co-application of ACh (200 µM) and ATP (100 µM) induced an inward current that was smaller in amplitude than the predicted sum of the individual IACh and IATP. B) At a Vh of +40 mV, the amplitude of the outward current induced by the co-application of ACh (500 µM) and ATP (200 µM) was equal to the predicted sum of the individual IACh and IATP. C) Co-application of ATP and ACh does not markedly change EACh or EATP. Representative currents induced by ACh, ATP, and by co-application of ACh and ATP at a Vh of +10 mV. Note that IACh+ATP is composed of early outward and delayed inward current, corresponding to the ACh component and the ATP component, respectively. When Ca2+ was omitted from extracellular solution, the negative interaction between the two receptor channels was still obtained; with IACh+ATP being 77 ± 1% (N = 5; P < 0.01) of the predicted sum of IACh and IATP. Hence negative interaction is not mediated by a Ca2+ influx-dependent mechanism, although both nAChR and P2XR are Ca2+-permeable cation channels (Fieber and Adams [35]; Rogers and Dani [36]). This result, however, does not necessarily rule out the possible involvement of [Ca]in in negative interaction. Occlusion was not only observed at a Vh of −70 mV; IACh+ATP at Vhs of −20, and −10 mV were 80 ± 2% (N = 12; P < 0.01), and 76 ± 2% (N = 8; P < 0.01), respectively, of the predicted sum of IACh and IATP. In sharp contrast, such occlusion was not observed at a positive potential; the outward current caused by coapplication of ACh and ATP at a Vh of +40 mV was 100 ± 1% (N = 10) of the predicted sum of individual current (Fig. 6B). We next investigated the possibility that the combined addition of ACh and ATP alters the driving force for Na+, by attempting to measure EACh+ATP. In these studies, currents in response to ACh, ATP and both ligands were measured at various fixed potentials close to the reversal potential for IACh and IATP. The EACh and EATP were 8.7 ± 1.5 mV (N = 5) and 21.0 ± 1.6 mV (N = 7), respectively, and IACh+ATP measured at a potential between these two reversal potentials (e.g.10 mV, Fig. 6C) showed a combination of both the ACh-induced outward current and the ATP-induced inward current. These results suggest that the driving force for permeant ions during combined ACh and ATP is similar to that during the application of each ligand separately. The question arises as to whether occlusion is specific just for Na+ ions or whether inward currents carried by cations other than Na+ can also contribute to occlusion during co-activation of nAChRs and P2XRs. To address this question, we measured IACh, IATP, and IACh +ATP in the presence of Na+-free (NMDG+) saline containing 10 mM Ca2+ (Fig. 7). NMDG+ does not permeate through nAChR (Sorimachi and Wakamori [16]), and hence IACh would be mediated by only Ca2+ influx, whereas IATP would be mediated via both NMDG+ and Ca2+ influx (Fig. 3). Under these conditions, IACh+ATP was again occluded, being 70 ± 3% of the predicted sum of IACh and IATP (N = 11; P < 0.01; Fig. 7). Fig. 7Occlusion of IACh+ATP in the presence of Na+-free saline containing 10 mM Ca2+. IACh, IATP or IACh+ATP was recorded 20 s after switching to Na+-free (replaced by NMDG+) saline containing 10 mM Ca2+. The sequence of applications was repeated multiple times. Following each ligand application, the external solution was changed back to normal saline containing 150 mM Na+ before the next ligand application. Vh: −70 mV. In a further attempt to distinguish whether channel activation or ion permeation was primarily responsible for occlusion of one channel by the other, we compared IATP with IACh+ATP under two experimental conditions, in which a higher concentration of ACh should activate nAChR but produce very little current. In the absence of Na+ and Ca2+ (replaced by NMDG+) application of ACh induces a negligible current (Sorimachi and Wakamori [16]), while ATP induces a substantial current under the same conditions (Fig. 3A). As shown in Fig. 8B, there was no occlusion in these conditions; the amplitude of IATP+ACh was 105 ± 1% (N = 10) of the predicted sum of IATP and IACh. We also investigated currents in response to the ligands at a potential close to the EACh (7 ± 2 mV in this experiment, N = 13), taking advantage of the more positive EATP (22.7 ± 0.9 mV), so that at this potential only the IATP was observed. The amplitude of IATP+ACh was 105 ± 3% (N = 13) of that of IATP, and IACh was close to zero (Fig. 8D). These results are, however, in contrast to that obtained in sympathetic neurons (Nakazawa[30]), in which IATP in the absence of both Na+ (replaced by glucosamine) and Ca2+ was inhibited in the presence of ACh, which caused no current (Nakazawa [30]). In these experiments, we only included data in which IATP was large enough to be clearly resolved, greater than 50 pA (corresponding to a control IATP at a Vh of −70 mV larger than 1.5 nA). These results suggest that actual current flow through both nAChR and P2XR is responsible for occlusion. Fig. 8Lack of inhibition in response to co-application of ACh and ATP under two experimental conditions designed to virtually eliminate IACh. A) Control IACh recorded at a Vh of −70 mV in standard saline. B) Current responses recorded at −70 mV from 20 s after exchanging the standard saline for a Na+-free (replaced by NMDG+), Ca2+-free saline. This sequence was repeated twice or three times. Note the small amplitude of IACh, and the lack of occlusion of IACh+ATP. C) Control IACh recorded at a Vh of −70 mV in another neuron. D) Representative traces of IACh, IATP, and IACh+ATP recorded at a Vh of +7 mV, which is very close to EACh. Note the lack of occlusion at this potential, when current flow through the AChR is negligible. Discussion In this study, we found that ATP induced an inward current and increased [Ca]in in isolated rat AP neurons. The ATP-induced current and [Ca]in increase were mimicked by ATPγS and 2-methylthio-ATP, but not by α,β-methylene-ATP, β, γ-methylene-ATP nor ADP, and was inhibited by suramin and PPADS, suggesting that it was mediated by P2XRs. These results are in good agreement with previous histochemical findings, which demonstrated the presence of P2X2, P2X4, and P2X6 receptor mRNAs (Collo et al. [37]), and of P2X2 receptor immunoreactivity in AP neurons (Kanjhan et al. [38]; Atkinson et al. [39]). The IATP showed strong inward rectification and the EATP was 22.7 ± 0.9 mV (Fig. 1B). The amplitude of IATP varied inversely with the extracellular Ca2+ concentrations (Fig. 1C). The inhibitory effect of increasing extracellular Ca2+ has also been reported in rat sensory neurons (Krishtal et al. [40]), PC-12 cells (Nakazawa et al. [18]), ventromedial hypothalamic neurons (Sorimachi et al. [41]), and in cells expressing recombinant P2X2 receptors (Evans et al. [42]; Virginio et al. [43]), where an allosteric alteration of the ATP binding sites has been suggested to be the underlying mechanism. The IATP was markedly reduced, but still persisted when NMDG+ was substituted for external Na+ even in the absence of Ca2+ (Fig. 3A). Since ATP did not induce a current when sucrose was substituted for external NaCl, our result suggests that NMDG+ could permeate through P2XR. A substantial IATP has been similarly reported in PC-12 cells and sympathetic neurons when glucosamine was substituted for Na+ (Nakazawa et al. [18]; Nakazawa [30]). The addition of Ca2+ to NMDG-Cl- and sucrose-substituted saline increased the IATP (Fig. 3B and C, respectively), indicating that Ca2+ also permeates through P2XR channel. We quantified the relatively high Ca2+ permeability, obtaining a PCa/PNa ratio of 3.0. A relatively high permeability of P2XR to Ca2+ has been reported in previous studies (Nakazawa et al. [18]; Sorimachi et al. [41]; Evans et al. [42]; Virginio et al. [43]). Although direct influx of extracellular Ca2+ through P2XR channel may thus contribute to the ATP-induced [Ca]in increase, membrane depolarization and the secondary activation of voltage-dependent Ca2+ channels could also make a significant contribution to the [Ca]in increase. In fact, the high K+- and ATP-induced [Ca]in increases were substantially inhibited by a range of Ca2+ channel antagonists, including those which block L- and N-type Ca2+ channels (Table 1). This is in contrast to a previous study, which reported an absence of the L- and N-type Ca2+ channels in rabbit AP neurons (Hay et al. [12]). The discrepancies between their results and ours could be, at least in part, accounted for by the use of different AP preparations (cultured rabbit neurons vs. acutely dissociated rat neurons). The present results, in combination with our previous demonstration of nAChR on AP neurons (Sorimachi and Wakamori [16]), suggest that both ATP and ACh may act as excitatory neurotransmitters in AP neurons, although their release from presynaptic nerve terminals has not yet been reported. We also report a negative functional interaction between P2XRs and nAChRs in AP neurons, as has been observed in a variety of peripheral neurons and in recombinant expression systems (Nakazawa et al. [18]; Nakazawa [30]; Zhou and Galligan [20]; Barajas-Lopez et al. [21]; Searl et al. [19]; Khakh et al. [22, 23]; Boue-Grabot et al. [24]). This is the first report, that we are aware of, of such interactions occurring in central neurons. When ACh was applied in the presence of ATP, there was a positive correlation between the peak amplitude of inward IATP and the amount of occlusion of inward IACh (Fig. 4A and B). The converse was also true when ATP was applied in the presence of ACh (Fig. 4C and D). Non-additivity of the IACh+ATP was observed even when the inward current was carried by NMDG+ and/or Ca2+ (Fig. 7). Co-application of two agonists did not seem to change the driving force for Na+, because at a Vh between EACh and EATP an outward current due to the activation of nAChR, followed by an inward current due to the activation of P2XR, was observed (Fig. 6C). Occlusion was also observed at a Vh of −10 mV, which is closer to EACh or EATP. Furthermore, the removal of external Ca2+ did not alter the occlusion, and thus a Ca2+-mediated mechanism does not contribute to the current occlusion. The IACh+ATP occlusion was observed at all negative holding potentials when the current was inward but was not observed for outward currents at a Vh of +40 mV. Here the outward IACh+ATP was not different from the predicted sum of IACh and IATP (Fig. 6B). The same dependence on current direction has been reported in guinea-pig enteric and submucosal neurons (Zhou and Galligan [20]; Barajas-Lopez et al. [21]), suggesting that the current occlusion was triggered by the inward movement of cations through two channels. Some investigators have shown that the amplitude of IACh+ATP is equal to or even smaller than that of either IACh or IATP. For instance, the concentration of one agonist causing little or no inward current produced dramatic occlusion of the inward current generated by the other agonist (Searl et al. [19]). Khakh et al. [22] using Xenopus oocytes co-expressing P2X2 and α3β4 nAChR channels, provided several lines of evidence which indicated that occlusion was mostly mediated by the inhibition of the nAChR channel by activation of the P2XR. However, we found that concomitant application of two agonists always caused a larger current than either agonist (Fig. 6A), and that the amount of occlusion of one channel current in the presence of the other channel agonist was correlated with the amplitude of current through the other channel (Fig. 4). Furthermore, we found that IACh+ATP desensitizes faster than IATP, but more slowly than IACh, and IACh+ATP at a Vh between EACh and EATP showed a combination pattern of early outward IACh and delayed inward IATP (Fig. 6C). These results both strongly suggest that in AP neurons, IACh+ATP are carried through both nAChR and P2XR channels, and that inhibition between these channels is reciprocal. The IATP was not inhibited by co-application of ACh when the inward IACh was markedly reduced in the absence of permeant cations (Fig. 8B) or at a Vh very close to EACh (Fig. 8D), and the inhibited IATP in the presence of ACh (Fig. 5B) recovered as soon as the nAChR channel closed in the presence of dTc (Fig. 5C). These results thus suggest that the inhibitory interaction not only requires the activation of both receptor channels, but also requires a substantial current to flow through these channels. These results are in contrast to that obtained in sympathetic neurons, in which the ATP-induced glucosamine influx was inhibited in the presence of ACh (Nakazawa [30]). The cause of this inconsistency remained unknown, but may reflect a real difference in the underlying mechanisms of occlusion in these two types of neurons. Altogether, our results have characterized the P2XR responses in AP neurons and the cross-inhibition between P2XRs and nAChRs in AP neurons. We show that the current flow through one receptor channel hinders the current flow through the other channel. These interactions support the notion that these channels are positioned very close to each other, as has been previously considered (Zhou and Galligan [20]; Barajas-Lopez et al. [21]; Khakh et al. [22]; Boue-Grabot et al. [24]) and more recently demonstrated for recombinant channels (Khakh et al. [23]). These results will be important to consider when designing ligands to modify excitability of ATP neurons and may have some physiological function during co-activation of P2XRs and nAChRs by synaptically released transmitters.

Breast_Cancer_Res_Treat-3-1-2096638

Loss of expression of FANCD2 protein in sporadic and hereditary breast cancer

Fanconi anemia (FA) is a recessive disorder associated with progressive pancytopenia, multiple developmental defects, and marked predisposition to malignancies. FA is genetically heterogeneous, comprising at least 12 complementation groups (A–M). Activation of one of the FA proteins (FANCD2) by mono-ubiquitination is an essential step in DNA damage response. As FANCD2 interacts with BRCA1, is expressed in proliferating normal breast cells, and FANCD2 knockout mice develop breast tumors, we investigated the expression of FANCD2 in sporadic and hereditary invasive breast cancer patients to evaluate its possible role in breast carcinogenesis. Two tissue microarrays of 129 and 220 sporadic breast cancers and a tissue microarray containing 25 BRCA1 germline mutation-related invasive breast cancers were stained for FANCD2. Expression results were compared with several clinicopathological variables and tested for prognostic value. Eighteen of 96 (19%) sporadic breast cancers and two of 21 (10%) BRCA1-related breast cancers were completely FANCD2-negative, which, however, still showed proliferation. In the remaining cases, the percentage of FANCD2-expressing cells correlated strongly with mitotic index and percentage of cells positive for the proliferation markers Ki-67 and Cyclin A. In immunofluorescence double staining, coexpression of FANCD2 and Ki-67 was apparent. In survival analysis, high FANCD2 expression appeared to be prognostically unfavorable for overall survival (p = 0.03), independent from other major prognosticators (p = 0.026). In conclusion, FANCD2 expression is absent in 10–20% of sporadic and BRCA1-related breast cancers, indicating that somatic inactivating (epi)genetic events in FANCD2 may be important in both sporadic and hereditary breast carcinogenesis. FANCD2 is of independent prognostic value in sporadic breast cancer. Introduction Fanconi anemia (FA) is a recessive disease with both autosomal and X-linked inheritance. FA is associated with progressive pancytopenia, developmental defects, and marked predisposition to malignancies, especially acute myeloid leukemia and squamous cell carcinoma of the head and neck [1, 2]. FA cells are characterized by spontaneous chromosomal instability and hypersensitivity to DNA cross-linking agents such as mitomycin C (MMC). FA is genetically heterogeneous and comprises at least 12 complementation groups (A–M). Eleven of the FA genes have been identified so far: FANCA, FANCB, FANCC, FANCD1/BRCA2, FANCD2, FANCE, FANCF, FANCG, FANCJ, FANCL, and FANCM [3–18]. Eight FA proteins (A, B, C, E, F, G, L, and M) form a nuclear protein complex which is required for mono-ubiquitination of the downstream FA protein, FANCD2. Activation of FANCD2 by mono-ubiquitination is an essential step in the DNA damage response induced by MMC or ionizing irradiation [2, 10]. This DNA damage response pathway also includes the breast cancer susceptibility genes BRCA1 and BRCA2, also referred to as the FA-BRCA pathway. Following ionizing radiation, FANCD2 and BRCA1 accumulate and colocalize in nuclear foci, which reflect sites of DNA damage and repair [10, 19]. Like FA cells, cells lacking BRCA1/2 proteins are hypersensitive to DNA cross-linking agents. D’Andrea et al. showed that FANCD1 and BRCA2 are the same proteins. BRCA2 is a direct regulator of RAD51, a protein essential for homologous recombination repair [20]. Although BRCA1 is mainly involved in hereditary breast cancer [21], it has also been implicated in sporadic breast cancer [22]. In an immunohistochemical analysis, we have previously shown that FANCD2 is expressed in proliferating cells of different organs, including the premenopausal breast duct epithelium [23]. This is in line with the role of FANCD2 in DNA repair which is important to guarantee the integrity of the genome during cell replication [10]. As deregulation of proliferation is one of the crucial processes of carcinogenesis, these observations imply a potential role for FANCD2 in the pathogenesis of breast cancer. Indeed, FANCD2 knockout mice develop breast tumors [24]. These considerations prompted us to investigate the expression of FANCD2 in sporadic and hereditary invasive breast cancers by immunohistochemistry in relation to several other proliferation-related biomarkers and survival. Materials and methods Tissue microarray Paraffin blocks containing formaldehyde-fixed breast cancer tissues of 129 cases of invasive breast cancer not selected for family history (further denoted “sporadic”) were obtained from the archives of the Department of Pathology of the VU University Medical Center, Amsterdam. For all breast cancer cases, age, lymph node status, and tumor size were documented. A second array block was constructed containing 24 cases with a proven BRCA1 germline mutation identified through the Family Cancer Clinic of the VU University Medical Center as previously described [25]. Patient characteristics are shown in Table 1. The hematoxylin–eosin stainings were used to identify representative areas of tumor tissue in the blocks. A tissue microarray was then constructed by transferring tissue cylinders of 4–5 mm from the representative tumor area of each donor block to the recipient block using a tissue arrayer (Beecher Instruments, Sun Prairie, WI, USA) as described before [26]. A third tissue array block of 220 sporadic breast cancer patients with long-term follow-up was obtained from the archives of the Gerhard-Domagk Institute of Pathology, University of Muenster, as previously described [27]. Table 1Patient characteristics and histology of 120 sporadic and hereditary breast cancersSporadicBRCA1 mutationTotalNo. of patients9624120Age   Mean654262Lymph node status   Negative58967   Positive381149Tumor size   Mean2.342.782.40Histological type   Ductal8517102   Lobular718   Medullary123   Tubular1–1   Cribriform1–1   Apocrine1–1   Metaplastic–44 Sections of 4 μm were cut and transferred on SuperFrost+ (Menzel&Glaeser, Germany) slides for immunohistochemistry. Use of anonymous or coded left over material for scientific purposes is part of the standard treatment contract with patients in our hospital [28]. Immunohistochemistry Immunohistochemical analysis had been previously performed on conventional sections for the following markers: Ki-67, Cyclin A, p21, p27, p53, estrogen receptor (ER), progesterone receptor (PR), HER-2/neu, and EGF-receptor [29]. Rabbit polyclonal antiserum against FANCD2 was generated as previously described [23]. Tissue sections were deparaffinized and rehydrated. Endogenous peroxidase was blocked with methanol/0.3% H2O2 (20 min). Sections were heated (30 min, 120°C) in 0.1 M citrate buffer pH 6. Unspecific binding was blocked with a 1:50 normal goat serum in PBS pH 7.4/1% BSA. Polyclonal rabbit anti-FANCD2 (200 μg/ml) was diluted 1:500 in PBS/1% BSA, and sections were incubated overnight (4°C) in a humidified chamber. Subsequently, sections were incubated with HRP-conjugated secondary antibodies (EnVision, DAKO) and diaminobenzidin (10 min), counterstained with hematoxylin (20 s), dehydrated, and cover-slipped. Appropriate positive controls were used throughout, and negative controls were obtained by omission of the primary antibodies. Percentages of positively stained nuclei were estimated by an experienced observer (P.J.v.D.), except for HER-2/neu and EGF-receptor where membrane staining was scored as positive. In addition, FANCD2 intensity was scored semiquantitatively as 0–3, and an FANCD2 score was calculated for each case by multiplying the % FANCD2-positive cells by the staining intensity. For FANCD2/Ki-67 double staining, anti-FANCD2 was diluted 1:150, incubated overnight, followed by incubation with swine anti-Rabbit HRP 1:200 (Dako, Glostrup, Denmark), and detected with the TSA™ Tetramethylrhodamine system (PerkinElmer Life Sciences, Boston, USA). This was immediately followed by incubation with mouse anti-Ki-67 1:50 (MIB1, Immunotech, Marseille, France) followed by rabbit anti-mouse FITC 1:40 (Dako, Glostrup, Denmark). Nuclei were counterstained by incubation with TO-PRO-3 (Molecular Probes, Eugene, OR, USA) 1:5,000 as previously described [30]. Statistics Bivariate scatter plots were generated between the percentage of FANCD2-expressing cells and the other continuous features. For the proliferation-associated features Ki-67, Cyclin A, and MAI, the cases with no FANCD2 expression were excluded from the analysis, assuming that FANCD2 is, by some mechanisms, no longer expressed in these cases. By linear regression analysis, the correlation coefficient R and related p values were calculated. Student’s t test was used to compare FANCD2 expression levels between the low-level vs. high-level groups for HER-2/neu, the EGF-receptor, ER (cut off 10%) and PR (cut off 10%), p53 (cut off 10%), and Cyclin D1 (cut off 5%). Prognostic value of FANCD2 (Muenster cases) was assessed by computing Kaplan–Meier curves, and differences between the curves were evaluated with the log-rank test. Multivariate survival analysis was performed by Cox regression. Results Successful FANCD2 staining was performed in 96 of the 129 cases of the VUmc sporadic array block. The drop outs were caused by damaged or detached cores during cutting, mounting, or staining, or did not contain tumor. Eighteen of these cases (19%) were completely FANCD2-negative. The other cases showed variable staining from 1 to 85% of the nuclei. In 21 of the 24 BRCA1 cases, FANCD2 staining was performed successfully. In two of these (9.5%), FANCD2 expression was completely negative. Figure 1 shows some representative examples of FANCD2 staining. Fig. 1Examples of FANCD2 staining in sporadic (A/B, A = negative control) and BRCA1-related breast cancers (C/D, C = negative control) The mean percentage of FANCD2-expressing cells was significantly higher in ER-negative patients (p = 0.019), PR-negative patients (p = 0.016), EGFR-positive patients (p = 0.002), Cyclin-D1-negative patients (p = 0.002), and p53-positive patients (p < 0.001) (Table 2). No statistically significant difference was seen for HER-2/neu. When analyzing the sporadic and hereditary subgroups, similar associations were seen. FANCD2 staining intensity yielded no useful correlations and no prognostic value, and the FANCD2 score yielded essentially the same correlations and prognostic value as the % FANCD2-positive cells (data not shown). Table 2Mean percentage of FANCD2-positive cells in high- and low-level expression groups of hormone receptors, growth factor receptors, Cyclin D1, and p53 in sporadic and hereditary breast cancersNumber (%)Mean (SE) of FANCD2%p Value (t test)Total117 (100)ER   Low56 (48)14 (2.5)0.019   High61 (52)8 (1.6)PR   Low81 (69)17 (1.9)0.016   High36 (31)12 (1.9)HER-2/neu   Negative93 (79)16 (1.6)0.797   Positive24 (21)17 (3.5)EGF receptor   Negative78 (66)13 (1.5)0.002   Positive39 (33)19 (3.1)Cyclin D1   Low85 (73)12 (1.9)0.002   High32 (27)7 (2.6)p53   Low88 (75)8 (1.3)<0.001   High29 (25)19 (4.3)ER estrogen receptor, PR progesterone receptor In linear regression analysis (Table 3), the percentage of FANCD2-expressing cells was significantly positively correlated to Ki-67 (R = 0.502, p < 0.0001), Cyclin A (R = 0.482, p < 0.0001), MAI (R = 0.506, p < 0.001), and p53 (R = 0.379, p < 0.0001), and significantly negatively correlated to age (R = 0.197, p = 0.033), ER (R = 0.221, p = 0.017), and PR (R = 0.204, p = 0.028). There was no correlation between FANCD2 and the other continuous features. Figure 2 shows examples of FANCD2/Ki-67 immunofluorescence double staining, underlining the coexpression of FANCD2 and Ki-67 in invasive breast cancers cells. In the Muenster cases, most of these correlations could be reproduced (age: p = 0.063, Ki-67: p = 0.001, p53: p = 0.003, ER: p = 0.034). Only PR was not significant here, and Cyclin A was not performed. Table 3Correlation between mean percentage of FANCD2-expressing cells and other continuous clinicopathological variables in sporadic and hereditary breast cancersFeatureRp ValueMAIa0.506<0.001Ki-67a0.502<0.001Cyclin Aa0.482<0.001Age−0.1970.033ER−0.2210.017PR−0.2040.028Tumor size0.0180.852p27−0.0740.430p21−0.1710.065p530.379<0.001Cyclin D1−0.1260.176ER estrogen receptor, PR progesterone receptoraFANCD2-negative cases excludedFig. 2FANCD2/Ki-67 immunofluorescence double staining in a representative case of invasive breast cancer. Top left: Ki-67 staining. Top right: FANCD2 staining. Bottom left: TO-PRO staining. D: Triple exposure showing coexpression of FANCD2 and Ki-67 In the regression analysis between the percentages of FANCD2-expressing cells and Ki-67 and Cyclin A, the completely negative FANCD2 cases were excluded. These FANCD2-negative cases had Ki-67 values between 1 and 65 (mean 16%), Cyclin A values between 0 and 50 (mean 10%), and MAI values between 0 and 37 (mean 11), indicating that these cases had (sometimes even high) FANCD2-independent proliferation. In survival analysis (sporadic Muenster cases), high FANCD2 expression appeared to be prognostically unfavorable (p = 0.03). Figure 3 shows the survival curves. In Cox regression including tumor size, lymph node status, ER, and grade, FANCD2 staining appeared to have independent prognostic value for overall survival (p = 0.026). Fig. 3Prognostic value of FANCD2 expression in sporadic invasive breast cancer. Low expressors have a better survival than high expressors (p = 0.03, N = 122) Discussion The aim of this study was to investigate the expression of FANCD2 in sporadic and hereditary breast cancers. This was inspired by several observations. First, FANCD2 and BRCA1/2 are functionally closely linked in the DNA repair response, and BRCA1 and BRCA2 are implicated in hereditary and sporadic breast cancers [21, 22]. Second, targeted deletion of FANCD2 in mice resulted in an increased rate of breast tumors [24]. Third, we have shown that FANCD2 is expressed in proliferating cells in the duct epithelium of the normal breast [23]. It appeared that 19% of sporadic breast cancers completely lacked FANCD2 expression. Yet, these FANCD2-negative cases had high mean Ki-67, Cyclin A, and MAI values, indicating that the low FANCD2 levels in these cases cannot be explained by low proliferation. The fact that these FANCD2-negative cases stained for other proteins makes it quite unlikely that the FANCD2 negativity is due to fixation problems. In the FANCD2-negative cases, FANCD2 inactivation may, in view of its important function, have been a hit in carcinogenesis. BRCA1 germline mutation-related breast cancers showed lack of FANCD2 expression in only 9.5% of cases, which fits with the concept that a major hit in an important pathway (in these cases the BRCA1 germline mutation) is usually not associated with further hits in this pathway. It is yet unclear what the mechanism behind the lack of FANCD2 expression in these cases is. It needs to be further studied whether there are inactivating somatic mutations in these cases or whether promoter methylation plays a role. FANCD2 expression was strongly correlated with expression of the proliferation-associated features Ki-67, Cyclin A, and MAI, and FANCD2 and Ki-67 were coexpressed in invasive cancer cells. This is likely a reflection of the physiological function of FANCD2 in DNA repair of proliferating cells, rather than an independent overexpression of an altered gene, in line with our previous study where we found a coexpression of FANCD2 and Ki-67 in proliferating cells of various normal human tissues [23]. The observation that high FANCD2 expression indicated poor prognosis fits within the same concept, as rate of proliferation (and thereby FANCD2 expression) is a major cell biological phenomenon determining prognosis [31–33]. FANCD2 had prognostic value independent of stage and grade, which can be explained by the fact that proliferation and stage are not strongly correlated, and that grade includes nuclear atypia and tubule formation besides rate of proliferation as measured by mitotic index. Although heterozygosity for p53 was shown to accelerate epithelial tumor formation in Fancd2 knockout mice [34], a functional link between p53 and Fancd2 has not been described to explain the association found in the present study, which is likely caused by the fact that p53 mutated and thereby p53 protein accumulated tumors show, in general, higher proliferation and therefore more proliferating FANCD2-expressing cells. The same may also hold for the relation between FANCD2 and EGFR expressions, for which also no functional relationship has been described. The negative relation between FANCD2 and age can likely be explained by the fact that BRCA1-related patients that have higher FANCD2 expression are younger. Within the light of the above observations, the question remains why FANCD2 patients do not seem to be predisposed to breast cancer in clinical practice. FA itself is a rare genetic disease where the complementation group D2 constitutes only 1–2% of all FA cases and these patients generally have a more severe clinical course. They may therefore simply not live to get breast cancer in an apparent increased frequency. Our results do not indicate that somatic (epi)genetic changes in FANCD2 are a frequent secondary carcinogenetic event in BRCA1 germline-mutated patients, although this needs to be confirmed in a larger study group. In conclusion, FANCD2 expression is absent in 10–20% of sporadic and BRCA1-related breast cancers, indicating that somatic inactivating (epi)genetic events in FANCD2 may be important in both sporadic and hereditary breast carcinogenesis. FANCD2 is of independent prognostic value in sporadic breast cancer.

Ann_Surg_Oncol-3-1-1914262

The Impact on Post-surgical Treatment of Sentinel Lymph Node Biopsy of Internal Mammary Lymph Nodes in Patients with Breast Cancer

Background Since the introduction of the sentinel lymph node (SLN) biopsy in breast cancer patients there is a renewed interest in lymphatic drainage to the internal mammary (IM) chain nodes. We evaluated the frequency of lymphatic drainage to the IM chain, the rate of SLNs that contain metastases and the clinical implications of IM LN metastases. The internal mammary (IM) chain is well known as a site of metastases in patients with breast cancer. The presence of IM lymph node (LN) metastases is a poor prognostic sign as reflected by the UICC-TNM staging system that classifies IM LN metastases as stage N3.1 In earlier times dissection of the IM LNs, as part of an extended mastectomy with IM node dissection, was advocated by some authors but randomized trials failed to show any beneficial effect of the procedure that was accompanied by significant morbidity.2–4 Since then, interest in the IM LN chain has waned. Since the introduction of the sentinel lymph node (SLN) biopsy in breast cancer patients there has been a renewed interest in the IM LNs. Drainage to this basin is frequently seen on preoperative lymphoscintigraphy, while the localization of the primary tumor and the nanocolloid injection technique appear to influence the frequency of visualized IM nodes.5–7 Various authors advocate removal of these IM SLNs,8,9 but the clinical implications of IM LN biopsy are still unclear. In this study we evaluated the frequency of lymphatic drainage to the IM chain, the rate of metastases in the IM SLN and the clinical implications of IM LN metastases. PATIENTS AND METHODS Between June 1999 and April 2005, 523 consecutive patients underwent surgical treatment including SLN biopsy as a staging procedure for clinically stage T1-2N0 breast cancer. Data regarding all procedures were collected prospectively in a database of the nuclear medicine department. A diagnosis of invasive breast cancer was established preoperatively by fine-needle aspiration or image-guided large core needle biopsy. Sixteen patients who underwent SLN biopsy as a secondary operative procedure following previous excisional biopsy of the primary tumor and one patient who eventually turned out to have multicentric breast cancer were excluded from the study. The study cohort consisted of 506 patients. At the time of the introduction of the sentinel node procedure the ethical committee of the hospital approved the routine use of the SLN biopsy as a staging procedure. All patients received written information regarding the SLN procedure and the possibility of SLNs in the IM chain. The unknown clinical implications of surgically removing IM LNs were discussed with the patients. On the day of the operation, all patients received a combination of peritumoral intraparenchymal and subcutaneous injections of an average of 77.7 MBq (53–150 MBq) of 99mTc nanocolloid in a total volume of 0.6 ml of physiologic saline, given in two to three equal doses. In case of a nonpalpable breast tumor the tracer was injected around the tumor using a 7.5-MHz ultrasound probe (Aloka). After injection the area was massaged until the appearance of the SLN. Continuous visualization was done and imaging started as soon as lymphatic drainage was visualized on the persistence scope and at 2–3 h after injection in both the anterior and lateral direction. Images were obtained during a 2-min imaging time on the Toshiba 901 HG single-head gamma camera, using low-energy high-resolution collimators. A skin marker was placed on the projection of the SN using a handheld γ-ray detection probe (Europrobe, PI Medical diagnostic equipment BV). The operative procedure was carried out in the afternoon of the same day of the 99mTc nanocolloid injection. A γ-ray detection probe and a peritumoral injection of patent blue dye (Bleu patenté V, Laboratoire Guerbet, Aulnay-sous-Bois, France) were intraoperatively used for SLN identification. During the operation axillary sentinel nodes were retrieved first and frozen section analysis was done to enable axillary dissection during the same operative procedure in case of LN metastases. When no axillary SLN was visualized on preoperative lymphoscintigraphy, the axilla was nonetheless explored in search for a blue dye containing SLN. When no axillary SLN was identified by either means, axillary LN dissection was performed. Following retrieval of axillary SLNs, surgical exploration for IM SLNs was done. When an IM SLN was visualized on preoperative lymphoscintigraphy, the γ-ray detection probe was used to guide a parasternal intercostal incision through the best-suited intercostal space to harvest the visualized node. A partial rib resection was not routinely done for sentinel IM nodes that were localized behind the ribs. In addition to frozen section analysis of the axillary SLNs, all collected SLNs were formalin-fixed, paraffin embedded and cut at five levels of 250 μm. Pathological evaluation followed hematoxylin-eosin and immunohistochemical cytokeratin-8 staining. The presence of axillary and IM LN metastases was classified according to the 2002 version of the UICC-TNM-classification.1 To assess the additional operative time of IM SLN biopsy we compared the time between incision and skin closure needed for lumpectomy and mastectomy with and without IM SLN biopsy. We selected four groups of ten patients: those who underwent lumpectomy with or without IM SLN biopsy and those who had had mastectomy with or without IM SLN biopsy. In all patients the operative procedure had started with axillary SLN biopsy. We only evaluated patients who did not have axillary dissection to avoid the potential bias of extra time awaiting the result of frozen section analysis. For each category we selected the last ten patients to avoid the learning curve effect. To evaluate the impact of IM LN metastases on subsequent systemic treatment we assessed the proposed adjuvant treatment strategy in the absence and presence of metastases in these nodes, applying the Dutch national guidelines for treatment of breast cancer (version 2005, http//www.oncoline.nl, summarized in Table 1). Locoregional radiotherapy, comprising the affected breast and/or thoracic wall, and the ipsilateral axillary, periclavicular and parasternal fields was indicated in patients with four or more metastatic axillary LNs. In patients with IM LN metastases with 0–3 tumor-positive axillary LNs a parasternal irradiation was given: in combination with irradiation of the breast in patients who had breast-conserving therapy or as parasternal radiotherapy following mastectomy. TABLE 1.Indications for adjuvant chemo- and hormonal systemic therapy according to the Dutch national guidelines 2005Axillary lymph node metastasesPrimary tumor characteristicsTumor >3 cmTumor >2 cm and BR grade IITumor >1 cm and BR grade IIIOther conditionsAge <35 years: always systemic therapy 60-69 years: chemotherapy when ER- or ≥4 axillary lymph node metastases ≥70 years: no chemotherapyER statusHormonal therapy for the aforementioned indications at all ages if the tumor is ER receptor positive.BR, Bloom-Richardson; ER, estrogen receptor. Chi-square analysis was performed to evaluate differences in IM SLN visualization rates between groups of patients with various clinicopathological variables and to explore the relation between IM and axillary LN metastases. The ANOVA test was used to explore the relation between age and the visualization of IM SLNs, as well as for the analysis of operative time differences. RESULTS The median age of the 506 patients was 60 years (range 24–92 years) and there were three male patients (0.8%). The distribution of primary tumor characteristics is summarized in Table 2. TABLE 2.Comparison of characteristics of patients who had visualized internal mammary sentinel lymph nodes (IM SLNs) versus those who had not visualized IM SLNs on preoperative lymphoscintigraphyAll patientsIM SLN visualizedIM SLN not visualizedPn = 506n = 109n = 397Median age (years) 60 (range 24–92) 57 (range 30–91)61 (range 24–92)0.016*Gender0.36  Male 3(0.6)0(0)3(0.8)  Female 503(99.4)109(100)394(99.2)T-stage0.47  T1316(62.5)73(67.0)243(61.2)  T2184(36.4)36(33.0) 148(37.3)  T34(0.8)0(0)4(1.0)  Tx2(0.4)0(0)2(0.5)Tumor localization <0.001  Cranial47(9.3)11(10.1)36(9.1)  Craniolateral230(45.5)30(27.5)200(50.4)  Lateral25(4.9)2(1.8)23(5.8)  Caudolateral40(7.9)10(9.2)30(7.6)  Caudal18(3.6)1(3.6)17(4.3)  Caudomedial30(5.9)13(11.9)17(4.3)  Medial11(2.2)4(3.7)7(1.8)  Craniomedial80(15.8)34(31.2)46(11.6)  Central25(4.9)4(3.7)21(5.3)Malignancy grade0.63  BRI206(40.7)48(44.0)158(39.8)  BRII196(38.7)38(34.9)158(39.8)  BRIII104(20.6)23(21.1)81(20.4)Estrogen receptor status 0.24  Positive427(84.4)88(80.7)339(85.4)  Negative79(15.6)21(19.3)58(14.6)Axillary lymph node involvement0.58  No axillary metastases296(58.5)68(62.4)228(57.4)  1–3 lymph node metastases174(34.4)35(32.1)139(35.0)  *4 lymph node metastases 36(7.1)6(5.5)30(7.6)Values in parentheses are percentages.BR, Bloom-Richardson grade.* Age difference between the groups was compared by ANOVA. Visualization and Retrieval of Sentinel Lymph Nodes in the IM Chain (Fig. 1) SLNs were visualized on preoperative lymphoscintigraphy in 502 of the 506 patients (99%). Axillary SLNs were visualized in 499 of 502 patients (99%), while one or more IM SLNs were found in 109 patients (22%). Three of the 109 patients with visualized IM SLNs had IM sentinel nodes only (3%); in one of these latter patients, an axillary SN was removed following patent blue dye injection and axillary exploration. Location of the tumor in the craniomedial and caudomedial aspect of the breast was associated with drainage to the IM nodes (P < 0.001), and patients with visualized internal mammary nodes were younger (Table 2). FIG. 1.Summary of search for internal mammary sentinel lymph nodes (IM SLNs). IM SLNs could be retrieved through a parasternal intercostal incision in 85 of the 109 patients (78%). Parasternal exploration for IM SLNs added 16 min to the mean operative time in patients who underwent breast conservative therapy (P = 0.02). For patients who underwent mastectomy the operative time was not significantly prolonged (Table 3). Complications (pleural breeching, internal mammary vessel damage) of the intercostal surgical exploration were rare; in particular there were no pneumothoraxes or bleeding complications necessitating drainage or reoperation (Table 4).TABLE 3.Operative time of surgical exploration for internal mammary sentinel lymph nodes (IM SLNs) (n = 40)Mean operative time in minutes (range)PAxillary SLN biopsy/lumpectomy/IM SLN biopsy60(27–76)0.02Axillary SLN biopsy/lumpectomy/no IM SLN biopsy44(32–83)Axillary SLN biopsy/mastectomy/IM SLN biopsy72(42–104)0.8Axillary SLN biopsy/mastecomy/no IM SLN biopsy69(30–100)TABLE 4.Complications of surgical exploration for internal mammary sentinel lymph nodes (IM SLNs) (n = 109)nIntraoperative complicationsPleural breeching4(4)Internal mammary vessel damage6(6)Postoperative complicationsPneumothorax−Bleeding necessitating reoperation−Values in parentheses are percentages. IM LN Metastases Metastases in the IM SLN were observed in 20 of the 85 patients who underwent successful IM SLN exploration (24%), and 42% of the patients with axillary SLNs (210/499). There was a correlation between IM metastases and axillary metastases (P = 0.002). In 16 of the 20 patients with IM LN metastases, the axillary SLN contained metastases too (80%). Conversely, IM metastases were found in 1% of patients without axillary metastases (4/297), in 7% of patients with 1–3 axillary metastases (13/174) and in 8 % of patients with ≥4 axillary metastases (3/36). Clinical Implications of Metastases in the IM SLN (Table 5) Adjuvant systemic therapy was already indicated in 14 of the 20 patients with IM metastases due to concomitant axillary LN metastases or unfavorable primary tumor characteristics: ten patients would be candidates for postoperative chemotherapy and 14 would receive hormonal treatment. Based on tumor-positive IM SLNs six additional patients would receive systemic treatment. This proportion reflects 7% of the patients in whom IM SLNs were removed. In four patients chemotherapy was indicated and in five patients hormonal therapy. TABLE 5.Clinical postsurgical implications of internal mammary lymph nodes IM LN metastases (n = 20)No.Tumor characteristicsAxillary metastasesPostsurgical treatmentTreatment changed due to IM metastasesIM SLN not consideredIM SLN consideredSize (cm)Grade (BR)ER(n)CTHTRTCTHTRT Axilla N4+/age<701) 63, BCT1.5I+7++LR++LRNo2) 45, BCT1.6III+4++LR++LRNo3) 47, mastectomy1.8I+4++LR++LRNoAxilla N1-3+/age<704) 39, mastectomy2.5II+3++BCT++BCT+PSRT5) 46, BCT3.5III−3+−BCT+−BCT+PSRT6) 58, BCT1.8II+2++BCT++BCT+PSRT7) 45, BCT2.4I+2++BCT++BCT+PSRTAxilla N1-N1a or unfavorable primary tumor characteristics/age<708) 43, BCT1.8I+1mi−−BCT++BCT+PSCT/HT/RT9) 54, BCT2.2II+1mi++BCT++BCT+PSRT10) 66, BCT2.5II+1mi−+BCT−+BCT+PSRT11) 50, BCT2.5III+1mi++BCT++BCT+PSRT12) 42, mastectomy3.0I+1++No++PSRTAxilla N0/favorable tumor characteristics/age<7012) 54, BCT2.1I+0−−BCT++BCT+PSCT/HT/RT14) 60, BCT2.5I+0−−BCT++BCT+PS CT/HT/RT15) 67, mastectomy0.8I+0−−No−+PSHT/RT16) 61, BCT1.1I−0−−BCT+−BCT+PSCT/RTAge>7017) 82, mastectomy2.8II+2−+No−+PS?RT?18) 71, BCT2.1II+1−+BCT−+BCT+PSRT19) 72, BCT2.1II+1mi−+BCT−+BCT+PSRT20) 85, mastectomy0.9I+1mi−−No−+PS?HT/RT?BCT, breast-conserving therapy; BR, Bloom-Richardson grade; ER, estrogen receptor status; Nax+, number of positive axillary lymph nodes; IM SLN, internal mammary sentinel lymph node; ST, systemic therapy; CT, chemotherapy; HT, hormonal therapy; RT, radiotherapy; PS, parasternal radiotherapy; 1mi, micrometastases. In 3 of the 20 patients axillary tumor load (≥4 tumor-positive lymph nodes) was a reason for locoregional radiotherapy including the IM lymphatic chain, leaving 17 patients in whom the radiotherapy field was adjusted because of metastases in the IM SLN. These 17 patients in whom the radiotherapeutic strategy was changed made up 20% of the patients in whom IM sentinel nodes were visualized. Conversely, there were three patients with ≥4 axillary metastases and IM SLNs without metastases, and in these patients parasternal irradiation was omitted. In addition, parasternal irradiation could also be omitted in 30 patients with ≥4 axillary metastases who had no IM lymphatic drainage on preoperative lymphoscintigraphy. DISCUSSION In the present study, SLNs in the IM chain were visualized in approximately one-fifth of the patients who underwent surgery for primary breast cancer. Retrieving these nodes by parasternal intercostal exploration was feasible in the majority of patients. One-fifth of the retrieved IM LNs contained metastases and radiation treatment was adjusted in most of these patients. In this prospective cohort of patients who underwent SLN biopsy for clinically T1-2N0 breast cancer, IM SLNs were seen on the preoperative lymphoscintigraphy in 22% of the patients.5,6,8,10,11 The visualization rate in our study was rather high and the likely result of the tracer injection technique (Table 6). We used a combination of a parenchymatous peritumoral and a subcutaneous injection technique of the 99mTc nanocolloid, and this technique is associated with a higher visualization rate of IM SLNs than the subcutaneous or periareolar injection of the radiotracer.5–7,12,13 Apart from the effect of the injection technique we observed a higher frequency of lymphatic drainage to the IM LNs in patients with cranio- and caudomedially located tumors, as reported by others,8,14,15 as well as an effect of age: IM SLNs were more common in young patients. TABLE 6.Visualization and surgical extirpation rate of internal mammary sentinel lymph nodes (IM SLNs) in breast cancer patientsAuthorYear nMethod of tracer injectionVisualized IM SLNs (%)Surgically removed IM SLNs (%)*Madsen et al.2006506PT and SC2278Leidenius et al.102005984IT1488Paredes et al.62005383SC, later IT/PT 0–1773Farrus et al.52004225SC, later IT/PT 11–1769Estourgie et al.82003681IT2287van der Ent et al.112002256PT2563* Proportion of the visualized IM SLNs. SC, subcutaneous; PT, peritumoral; IT, intratumoral. Retrieving SLNs from the IM chain does not appear to be very troublesome. Approximately 15 min extraoperative time is needed during breast conservative surgery, while extra time is negligible in patients who undergo mastectomy. Although the success rate was lower than for SLNs in the axilla, 78% of the visualized IM nodes could be harvested by the described parasternal intercostal exploration. Others reported similar ‘surgical identification’ rates (69–88%, see Table 6).5,6,8,10,11 It can be difficult to retrieve the usually very small IM SLNs when they are localized directly behind one of the ribs. ‘Blind’ retrocostal dissection is not without risks, and we consider a rib resection not justified as long as the place of parasternal lymph node exploration is not well defined. Although the pleural cavity was breeched occasionally during the procedure and a number of patients had postoperative hematomas, postoperative drainage of a pneumothorax or a reoperation for a bleeding complication of the parasternal wound was never necessary. Other studies also reported low morbidity rates.8,9,11,16,17 In the present study IM LN metastases were found in 24% of the patients who underwent surgical extirpation of these SLNs, while others usually reported lower rates of metastases containing IM SLNs (9–26%).8–11,18 In patients with IM metastases axillary metastases were common: 80% had concomitant axillary metastases.19 Conversely, axillary metastases were accompanied by IM metastases in 8% of the cases. Isolated IM metastases were rare affecting 5% of the patients with harvested IM SLNs and approximately 1% of patients without axillary LN metastases.8,10,19 Successful exploration of SLNs from the IM chain had a substantial impact on subsequent radiotherapeutic treatment. In 20% of these patients adjustment of radiotherapy was considered necessary. Although randomized data about the use of radiotherapy in patients with IM metastases are lacking, adding a parasternal irradiation in these patients appears conceivable. After all radiotherapy does affect locoregional control in other high-risk patient groups such as patients with ≥4 axillary metastases and node-negative patients with young age, poor tumor differentation and large tumor size.20 We consider patients with IM metastases at risk for locoregional recurrence and therefore feel that parasternal irradiation is indicated in patients with IM metastases. Conversely, absence of IM lymphatic drainage or metastases in retrieved IM SLNs justified omission of parasternal radiotherapy in 7% of all the patients. Systemic treatment strategy was rarely influenced by IM metastases. Due to axillary metastases and unfavorable primary tumor characteristics, half of the patients already would have received chemotherapy and even more would have had hormonal therapy. In the remaining group of patients, old age and negative estrogen receptor status further limited the proportion of patients that would receive adjuvant systemic therapy based on IM LN metastases. However, since prognosis of patients with both axillary and IM metastases is poor when compared with axillary or IM metastases alone,3 patients with IM and axillary metastases might need different chemotherapy schedules. That would increase the proportion of patients in whom adjuvant systemic treatment would be adjusted. Is IM SLN biopsy worthwhile? One may argue that more extensive radiotherapy was indicated in less than 5% of all patients and additional systemic treatment in only 1%, and thus consider IM SLN biopsy hardly “worth the effort.”10 We feel that the group of patients with visualized SLNs should be taken as a reference, and that the proportion of patients in whom less radiotherapy was given should also be taken into account. As a consequence, we do consider the clinical implications of IM SLN biopsy substantial. In conclusion, lymphatic drainage of breast cancer to IM LNs is a common feature and retrieving these nodes is relatively easy. The clinical impact of metastases in IM lymph nodes is substantial and justifies surgical exploration for these nodes. We advocate routine parasternal intercostal exploration for IM SLNs whenever preoperative lymphoscintigraphy visualizes IM SLNs. For that purpose we also advocate the (additional) intraparenchymatous tracer injection to optimize the visualization of IM SLNs.

Bioprocess_Biosyst_Eng-2-2-1705518

Development of a kinetic metabolic model: application to Catharanthus roseus hairy root

A kinetic metabolic model describing Catharanthus roseus hairy root growth and nutrition was developed. The metabolic network includes glycolysis, pentose-phosphate pathway, TCA cycle and the catabolic reactions leading to cell building blocks such as amino acids, organic acids, organic phosphates, lipids and structural hexoses. The central primary metabolic network was taken at pseudo-steady state and metabolic flux analysis technique allowed reducing from 31 metabolic fluxes to 20 independent pathways. Hairy root specific growth rate was described as a function of intracellular concentration in cell building blocks. Intracellular transport and accumulation kinetics for major nutrients were included. The model uses intracellular nutrients as well as energy shuttles to describe metabolic regulation. Model calibration was performed using experimental data obtained from batch and medium exchange liquid cultures of C. roseus hairy root using a minimal medium in Petri dish. The model is efficient in estimating the growth rate. Introduction In vitro plant biotechnology offers a controlled environment and has been widely studied for phyto-pharmaceuticals and recombinant proteins production. However, the low productivity and the poor reproducibility of the cultures are still limiting the economical feasibility of such in vitro bioprocesses. Moreover, the lack of reproducibility of the cultures significantly complicates process validation and acceptance by the regulatory agencies, and thus the potential to rapidly put a product to market. Different approaches have succeeded in decreasing the technological risk associated with in vitro plant biotechnology. The introduction of elicitors such as chitin [1] and jasmonic acid [2] has shown to enhance significantly the production level for many plant species cultured as cell suspensions and hairy roots. The use of an extractive phase allowed the simplification of harvesting procedures [3, 4]. More recently, cell engineering studies have shown its potential in improving cell catalytic capacity towards the production of secondary metabolites [5–7] and in in vitro molecular farming for recombinant human proteins [8]. However, high variability levels in cell growth and production of biomolecules of interest are still observed. The genetic flexibility of plant cells may partly explain these phenomena [9] and inadequate culture management may also be involved. A plant cell has the ability to accumulate nutrients and metabolites which are involved in the regulation of its metabolic pathways. The key for improving plant cell culture reproducibility may thus rely on a better understanding of the links that are exerting between a plant cell physiological state and its potential towards growth and production of a biomolecule of interest. This understanding (following that of Bailey [10]) could take the form of a descriptive and predictive metabolic model. Such structured model may then be either useful to enhance our understanding of cell behaviour, in identifying possible regulatory roles [11], as well as being a tool for defining adequate controlled culture conditions. Metabolic modelling has been applied to plant cells for studying specific metabolic sub-networks such as photosynthesis [12, 13], respiration [14], cellulose biosynthesis [15] and lipid biosynthesis [16]. These studies have clearly showed the importance of the energy shuttles on the control of the metabolic pathways. In addition, some nutrients are known to be involved in the regulation of the cell metabolism. Intracellular inorganic phosphate (Pi) plays a central role in the regulation of enzymes activity through phosphorylation/dephosphorylation processes, ATP/ADP concentration ratio, starch synthesis and storage, and in the flux distribution between the glycolysis and the pentose-phosphate pathway (PPP). Intracellular nitrate and ammonium are known to affect amino acid (AA) production. The large capacity of plant cells for nutrient and metabolite accumulation plays a crucial role in cell growth and biomolecule production, as observed for Daucus carota hairy root [17] and Eschscholtzia californica suspension cells [18]. Recently, we have developed a kinetic model based on intracellular nutrients such as inorganic phosphate, nitrate and sugars which showed to be efficient in simulating carrot hairy root growth for different culture media composition [17]. Therefore, the aim of this work was to include the description of metabolic pathways to the nutritional model in order to describe plant cells behaviour from the estimation of the cells physiological state, including nutritional and metabolic states. Catharanthus roseus was studied as a model biological system. Cell nutritional state in Pi, nitrogen (NO3− and NH4+) and carbohydrates (sucrose, fructose, glucose and starch) is described. The hypothesis of a central primary metabolism at steady state has been proposed based on literature [19, 20]. Using the metabolic flux analysis (MFA) approach, a model reduction [21] was applied on the central primary metabolism network and resulted in independent pathways. A second network includes transient fluxes such as for nutrient uptake and storage, energy shuttles management and root cells growth. Metabolic regulation of the fluxes from energy shuttles and nutrients is included. The hairy root specific growth rate is described as a function of the content in cell building blocks such as amino acids (including proteins), lipids (LIP), organic acids (ORA), organic phosphates (OP) (including nucleic acids) and structural hexoses (STH). Batch and medium exchange cultures of C. roseus hairy root were performed and the experimental data were used for model calibration. Model general structure The model has been first developed by Tikhomiroff [22]. The cell metabolic network (Fig. 1) is divided into two interlinked sub-networks as the stationary (SPMP) (Fig. 2) and the transient (TPMP) primary metabolic pathways (Fig. 1). The SPMP includes glycolysis, PPP, the TCA cycle and the catabolic reactions leading to the cell building blocks. The cell building blocks are amino acids and peptides which were taken as a unique pool of AA, ORA, OP, LIP and STH. The TPMP network is linked to the SPMP network and describes cells growth and nutrient transport between medium and intracellular volumes. Compartmentalization of nutrients and metabolites among the cytosol, the vacuole and other organelles is not included in the model and a single cell population was considered to describe the hairy root cells pools. This simplification already showed to be efficient to model hairy root growth and nutrition [17]. The secondary metabolism is simplified to fluxes leading to two pools, one accounting for the global pool in secondary metabolites derived from tryptamin (TRYSM), and one accounting for the global pool in secondary metabolites derived from secologanin (SECSM). The model is thus composed of interlinked metabolic networks that are at steady (SPMP) and transient (TPMP) states, and is described by a mass balance using the stoichiometric matrix and the hairy root specific growth rate: where S is the stoichiometric matrix, Φ is a vector containing reaction fluxes, μ is the hairy root specific growth rate and M is a vector containing the concentration in cellular metabolites and nutrients. Root mass with time can then be estimated both kinetically and from a mass balance on all cell constituents.Fig. 1The metabolic model global structure. Fluxes in the transient primary metabolic pathways (TPMP). Flux numbers refer to the stoichiometric biochemical reactions of Table 3. Kinetic description of the resulting fluxes is presented in Table 4Fig. 2Fluxes in the stationary primary metabolic pathways (SPMP). Flux numbers refer to the stoichiometric biochemical reactions of Table 1 Stationary primary metabolic pathways The pseudo-steady-state assumption for the central metabolism was based on observations from Rontein et al. [19] and Stitt and Fernie [20] and proposed to simplify the model development. The original SPMP metabolic network has 31 fluxes (Fig. 2; Table 1), which were reduced to 20 independent pathways (Fig. 3; Table 2) using the method proposed by Simpson et al. [21] and Stephanopoulos et al. [23] and the following simplifications. Briefly, the minimal number of independent feasible metabolic pathways is determined with a group of metabolites assumed at steady state: G6P, F6P, R5P, G3P, E4P, CHO, PEP, PYR, A-CoA, OXO and OAA. The “N” matrix (11×31) is first constructed from the 31 biochemical reactions and the 11 metabolites at steady state. The determination of the independent pathways then requires the knowledge of the kernel matrix “K” (31×number of independent pathways), which describe each pathway as a linear combination of the 31 biochemical reactions of the SPMP. This matrix is the non-trivial solution of the equation: Since this equation has an infinite number of solutions, both “N” and “K” matrices are decomposed each into two matrices: with “ N1” and “ K2” selected as square invertible matrices. Starting using identity matrix for “K2”, “K1” can then be determined by modifying “K2” such that possible independent pathways can be identified: Using the complete matrix K, we can build the independent pathways presented in Fig. 3. In this work, the size of this matrix is 31×20. The complete method to obtain the pathways can be found in Ref. [22].Table 1Biochemical reactions of the stationary primary metabolic pathwaysFluxν(1)a–cν(2)a–cν(3)aν(4)aν(5)bν(6)bν(7)a,cν(8)aν(9)aν(10)bν(11)aν(12)bν(13)aν(14)a–cν(15)aν(16)a–cν(17)c,dν(18)a–cν(19)c,dν(20)a–cν(21)eν(22)bν(23)bν(24)aν(25)bν(26)eν(27)ν(28)ν(29)dν(30)aν(31)da[39]b[24]c[40]d[27]e[41]f[26]Fig. 3Reduced independent biochemical pathways from the SPMP. Pathway numbers refer to the stoichiometric biochemical reactions of Table 2. Kinetic description of the resulting fluxes is presented in Table 4. The 20 independent biochemical pathways were obtained from the 31 biochemical reactions of the SPMP described in Fig. 2 and Table 1, using the reduction method of Simpson et al. [21] and Stephanopoulos et al. [23], as described in Sect. 3Table 2Independent fluxes of the SPMPs after pathways reduction using MFA approachFluxaν(r1)ν(r2)ν(r3)ν(r4)ν(r5)ν(r6)ν(r7)ν(r8)ν(r9)ν(r10)ν(r11)ν(r12)ν(r13)ν(r14)ν(r15)ν(r16)ν(r17)ν(r18)ν(r19)ν(r20)aThe biochemical reactions are denoted with an "r" to indicate that they are obtained from pathways reduction The nutrient concentrations described are glucose (GLC), fructose (FRU), sucrose (SUC), ammonium (NH4) and inorganic phosphate (Pi). The anaplerotic pathways are simplified to the transformation of phosphoenolpyruvate (PEP) into oxaloacetate (OAA). The metabolite pools resulting from the SPMP are starch (STA), STH (cell wall and membrane constituents), OP (nucleotides, phospholipids, nucleic acids), total AA including that of peptides, ORA, pools from tryptamin (TRYSM) and from secologanin (SECSM). Starch biosynthesis (flux v(r14) in Table 2) and catabolism (v(r13)) are included. AA (v(r5)) are assumed to be synthesized from oxoglutarate, since it is the site for ammonium fixation and thus the initiation step in the AA biosynthesis [24], even though there are AA synthesized from other pathways. Tryptamin (TRYSM) biosynthesis (v(r1)) has been explicitly described because this AA is a precursor, with secologanin (SECSM) (v(r2) and v(r10)), to the formation of the secondary metabolites. OP biosynthesis (v(r11)) has been simplified as illustrated in Fig. 3 and is accounting for the four nucleotides. The biosynthesis of STH results from multiple biochemical reactions [25] but it is simplified to the condensation of G6P with R5P (v(r12)). The biosynthesis (v(r7)) and catabolism (v(r15)) of LIP is linked to acetyl-coenzyme A [26] as for ORA biosynthesis (v(r16)) and catabolism (v(r17)). The kinetic expressions of the fluxes are described in the next section. Transient primary metabolic pathways The TPMP network includes the biochemical reactions that cannot satisfy the pseudo-steady-state hypothesis. These include nutrients transport and accumulation, root cells growth and energy shuttles dynamics. The general structure of the TPMP and its interactions with the SPMP are illustrated in Figs. 1 and 2. The stoichiometric mass balances are presented in Table 3 and the kinetic terms used in fluxes’ regulation are described in Table 4. Extracellular nutrients such as sucrose (ESUC), glucose (EGLC), fructose (EFRU), ammonium (ENH4), nitrate (ENO3) and Pi (EPi) are included. Sucrose is hydrolysed into fructose and glucose extracellularly by apoplastic invertases (Table 3, v(28)), or intracellularly as described in the SPMP (Table 2, v(r19)). It should be noted that sucrose biosynthesis (Table 2, v(r20)) is not included into the model because there are no evidence that hairy roots have active chloroplasts. v(r19) then represents the net flux of sucrose hydrolysis. Alternative glycolytic pathways [27] are described and the carbohydrate sources include sucrose, glucose, fructose and starch. However, since all the biosynthesis pathways of the SPMP are defined with glucose as precursor, all available intracellular carbon sources have first to be converted into glucose. Fructose conversion (v(r3)), starch catabolism (v(r13)) and sucrose hydrolysis (v(r19)) processes are thus feeding the intracellular glucose pool. Plant cell adaptation mechanism to Pi deficiency is described with the fluxes for the degradation of pyrophosphate (v(26)) and OP (v(23)) into Pi. The plant cell storage capacity for sugars, NO3−, NH4+, Pi and cell building blocks is described. Respiration is described (v(30)), however, oxygen is assumed to be non-limiting as discussed below. Energy loss associated with maintenance and other reactions that are not included into the SPMP is contained in flux v(29).Table 3Biochemical reactions of the transient primary metabolic pathwaysFluxaν(21)ν(22)ν(23)ν(24)ν(25)ν(26)ν(27)ν(28)ν(29)ν(30)ν(31)aStoichiometric coefficients were taken from Ref. [24]Table 4Kinetic expression of the fluxesFluxa–c is fixed at 0 and no sucrose synthesis is consideredaReaction number refers to pathways of SPMP in Fig. 3, and of TPMP in Fig. 1bNot shown fluxes are related to exchange between the cytoplasm and the vacuole and were not included in this modelcUse of NTP, NDP, NADH, NAD, NADPH and NADP in flux kinetics was taken from Ref. [24]d[42]e[43]f[44] Metabolic regulation Metabolic regulation of two kinds is integrated into the model: that associated with the energetic status of the cells and that related to the nutritional state of the cells, variable denoted as “θi” below. Each flux (ν) is regulated as follows, according to a multiplicative kinetic of each mechanism involved. The order of the regulation kinetics can also be adjusted from the term αi to account for multi-steps mechanisms. Since we established the independent pathways for the SPMP with arbitrary conditions, it is not possible to compare the structure of the resulting kinetics with literature. However, the identified independent pathways have been selected for their feasibility, and their combination is mathematically equivalent to the whole metabolic system (31 fluxes), given that the pseudo-steady-state hypothesis is acceptable. Furthermore, a sigmoid function is used for nutrients acting like switches to avoid large discontinuities when solving the differential equations, which would have been the case using a simple on/off switch type. Continuous functions are also more representative of biological behaviour [28]. The sigmoid function is also used for imposing maximum accumulation levels (see Table 4): The parameter “a” defines the steepness of the function, “Mi” is the concentration of the nutrient involved in the regulation of a flux and “Mi,t” is the nutrient concentration threshold. The role of energy shuttles in the regulatory mechanisms is described in Table 4. NADH, NADPH and NTP (the sum of ATP, GTP, CTP and UTP) as well as their reduced forms (NAD, NADP and NDP) are used. The sum in energy shuttles per g DW of roots in both oxidized and reduced forms (NADH + NAD; NDP + NTP; NADPH + NADP) are taken constant with time. However, the energetic status of the cells [NADH/(NADH + NAD); NTP/(NDP + NTP); NADPH/(NADPH + NADP)] was not assumed constant. ATPase proton pumps that are linked to H+/nutrient co-transport and involved in the control of intracellular pH are thus included. ATP consumption for transmembrane transport of glucose (v(21)), fructose (v(22)), sucrose (v(24)), nitrate (v(27)) and Pi (v(31)) is thus included. A unique value for the affinity constant for NTP (KmNTP) is used. The same strategy of using unique affinity constants for NAD, NADH, NADP and NADPH is also applied. Root cells nutritional state in Pi, NH4 and in sugars are involved in metabolic flux regulation. Intracellular Pi is involved in enzyme activation/deactivation processes through dephosphorylation/phosphorylation, in the biosynthesis of NTP (v(30)) and of many other metabolites as described below. Intracellular NH4 is involved in AA biosynthesis (v(r5)) (Fig. 3; Table 4). Michaelis–Menten kinetics with NH4 and Pi are at power two since a second-order mechanism is involved. NH4 has first to be transformed into NH2 radical, then into an AA which is finally integrated into a protein structure. Management of carbon source is crucial for plant cells since glucose, fructose, sucrose and starch can be used. However, since the model is based only on glucose, biochemical reactions converting other sugars into glucose are described (Fig. 3; Table 4). Starch biosynthesis (v(r14)) is controlled by the total intracellular available sugar concentration, and the sum in intracellular GLC, FRU and SUC has to be above 0.2 mmol/g DW for the STA biosynthesis to occur. A maximal STA storage capacity of 1.5 mmol/g DW was observed experimentally (see Sect. 7). Starch is degraded (v(r13)) when the GLC concentration is below 0.2 mmol/g DW. PPP (v(r4)) is regulated by NADP/(NADP+NAPH) ratio, which has to be above 0.5 for the pathway to be active. Pyruvate kinase regulation (v(r6)) is under the control of intracellular Pi concentration with a flux increase at 1 mmol Pi/g DW and below. G3P conversion into PEP (v(r8)) is regulated by Pi concentration with a flux increase at 1 mmol Pi/g DW and below. TCA cycle (v(r9)) is regulated by the NAD/(NAD+NADH) ratio, and a ratio that is above 0.5 will induce a flux increase. GLC (v(21)) and FRU (v(22)) uptake require a STA concentration that is below 0.07 and 0.02 mmol/g DW, respectively. OP synthesis (v(r11)) is regulated by intracellular Pi concentration with a flux increase at above 0.1 mmol Pi/g DW. OP degradation (v(23)) is controlled by intracellular Pi concentration with a flux increase at below 0.045 mmol Pi/g DW. LIP (v(r15)) and ORA (v(r17)) degradation into A-CoA is controlled by NADH concentration with fluxes increase at below 1 mmol NADH/g DW. Phosphofructokinase regulation (v(r18)) is controlled by intracellular Pi with a flux increase below 1 mmol Pi/g DW. Respiration rate (v(30)) is controlled by NDP/(NTP+NDP) ratio with a flux increase below a ratio of 0.5. Finally, regulation of root cells growth rate (v(32) is controlled by the intracellular concentration in the cell building blocks such as OP, LIP, amino acids and peptides (AA), ORA and STH. Monod model was used for LIP, AA and ORA. A hybrid Moser–Monod model was used for OP and STH with each kinetic term at power 4 and 1.25, respectively (Table 4). Since OP and STH are crucial to cell growth, the steepness of the affinity for both cell building blocks was increased. Materials and methods Culture conditions Liquid cultures of hairy roots were performed in Petri dish as described in Jolicoeur et al. [17]. The major problem in culturing hairy roots is the difficulty to obtain a representative sample of the root network since the roots develop a highly dense interlinked bed. Petri dish culture allowed distributing the roots inoculum in a series of dishes with each dish taken as a single sample. Root growth was also limited by the use of a low salt medium, thus preventing the occurrence of dense root network. Previous results in Petri dish [17] and in bioreactor [3] suggest that there was no oxygen limitation in the cultures of this work. Hairy roots of C. roseus L. G. Don were established as described by Bhadra et al. [29], with Agrobacterium rhizogenes strain A4. Hairy roots were transferred every month in Petri dishes in 25 mL of minimum medium [30] supplemented with 3% (w/v) sucrose and with a threefold KH2PO4 (0.352 mM). In the batch culture experiments, approximately 0.125 g fresh weight (FW) of hairy roots was inoculated in each Petri dish containing 25 mL of minimum medium. In the medium exchange cultures, approximately 0.125 g FW of hairy roots was inoculated in Petri dish containing 25 mL of minimum medium. The medium was renewed at 3- or 2-day intervals to avoid depletion of Pi in the medium. Whole Petri dishes were harvested in triplicates (n=3) periodically and taken as distinct samples. Liquid from each dish was filtered at 0.45 μm (Millipore, Billerica, Massachusetts) and stored at −20°C for further analysis. Roots were filtered under vacuum on a glass fiber filter (47 mm diameter Glass Microfiber filters GF/D, Whatman, #1823 047) and rinsed three times with 20 mL of de-ionized water. The filtered roots were removed from the filter and weighed for FW in a disposable aluminum dish (Fisher Scientific, # 08–732) on a precision balance (Sartorius). Fresh roots were immediately frozen into liquid N2 and stored in liquid N2 for further analysis. Root samples were then freeze dried (Duratop and Duradry, FTS Systems Inc., Stone Ridge, NY, USA), weighted for DW measurement and grinded (mortar and pestle) for further analysis. All further analyses were performed using freeze-dried roots. Amino acids analysis Approximately 2 mg of freeze-dried roots were extracted in 1 mL of 2% w/v 5-sulfosalicylic acid. The samples were sonicated for 15 min and then centrifuged at 12,000g for 5 min. The supernatant was analysed for AA by HPLC using a modified method from Gombert et al. [31] as described in Benslimane et al. [32]. The column is a high-efficiency Nova-Pak TM (C18, 4 μm). Precolumn derivatization of the AA was performed using AccQ.Fluor (6-aminoquinolyl-N-hydroxysuccinimidyl carbamate, or AQC), which is an N-hydroxysuccinimide-activated heterocyclic carbamate, and the derivatized AA were quantified via a fluorescence detector. α-Aminobutyric acid was used as an internal standard. The standard error of measurement using this method was routinely below 5%. AA concentration was then calculated since the peak for each individual AA accounted for the combination of free and that from peptides and proteins. Extraction of sugars Approximately 10 mg of freeze-dried roots was washed with 80% ethanol and then centrifuged at 16,000g for 5 min. This wash was done three times and each time the supernatant was kept for glucose, fructose and sucrose analyses. After the third wash the pellet was kept for starch analysis. Analysis of sugars Glucose, fructose and sucrose analysis was performed with a Beckman Coulter HPLC (Beckman Coulter Canada Inc, Mississauga, Canada; Pump model 126, automatic injector model 508) equipped with a Gilson model 132 refractive index detector. A Prevail Carbohydrate ES column 4.6 mm×250 mm (Alltech Canada, Guelph, Ontario, Canada), coupled with a Prevail Carbohydrate ES All-Guard 4.6 mm×7.5 mm guard column (Alltech Canada, Guelph, Ontario, Canada), was used at a column temperature of 35°C. The injection volume was 20 μL. The mobile phase consisted of acetonitrile and water 75:25 (v/v) at 1.0 mL/min. The cell pellet obtained from the soluble sugars extraction was re-suspended in 1 mL de-ionized water and sterilized at 121°C for 15 min together with a 1 mL 6 g/L starch solution in de-ionized water. Samples were allowed to reach room temperature and their volume was readjusted to 1 mL with de-ionized water if necessary. Starch calibration standards from 0 to 6 g/L were prepared by serial dilutions of the starch solution in de-ionized water. Calibration standards and samples were diluted 1:1 with an amyloglucosidase solution (Sigma, St. Louis, Missouri, USA, Cat. #S9144) and incubated for 15 min in an ultrasound bath at 60°C. Samples and standards were centrifuged for 10 min at 16,000g. A 10 μL aliquot of the supernatant was directly transferred into a spectrometric cuvette with 500 μL of “Glucose Infinity” reagent (Sigma, Cat. #17-25). After incubation (15 min) at room temperature, 500 μL of 100 mM KH2PO4 buffer pH 7.5 was added followed by spectrophotometric reading at 340 nm. Ions extraction Ions were extracted from approximately 10 mg of freeze-dried roots in 1.5 mL of 5% (w/v) trichloroacetic acid. This mixture was sonicated at 40°C for 30 min and then centrifuged at 16,000g for 10 min. The supernatant was filtered at 0.45 μm and analysed by HPLC. Ions analysis Culture medium and intracellular contents in majors ions (Cl−, NO3−, H2PO4−, SO42−, NH4+, K+, Na+, Ca2+) were analysed using a Dionex HPLC system (Dionex Canada Ltd., Oakville, Canada) equipped with an isocratic pump, an automated sampler AS-3500 and a pulsed electrochemical detector in the conductivity mode, controlled by the Dionex A1-450 software for cations and the Dionex Peaknet software for anions. Anions were separated using a 4×250 mm IONPAC AS14A-SC analytical column, an IONPACAG14A-SC guard column and a ASRS-1 anion self-regeneration suppressor to improve the signal-to-noise ratio. The mobile phase consisted of an aqueous buffer of 2 mM Na2CO3/1 mM NaHCO3 solution flowing at a rate of 1.0 mL/min. Cations were separated using a 4×250 IONPAC CS-12 analytical column, a IONPAC CG-12 guard column and a CSRS-1 cation self-regenerating suppressor. The mobile phase was an aqueous 20 mM methanesulphonic acid solution flowing at a rate of 0.9 mL/min. Model simulations The model simulations were done using the Matlab software (The MathWorks Inc., Natick, MA, USA). The differential equations system was integrated through the Ordinary Differential Equation solver ode15s.m. The model error minimization was performed by means of manual and algorithm-based methods. The former was done to find good initial estimates based on literature (when available). The latter was then used to reduce the global error between the estimates and the experimental data. The algorithm used was the lsqcurvefit.m subroutine (Optimisation toolbox, Matlab) based on the Levenberg–Marquardt algorithm [33]. Results and discussion Model calibration and determination of the kinetic parameters Model calibration was performed using experimental data from C. roseus hairy root batch and medium exchange cultures. Each experiment (one batch and one medium exchange) was performed in triplicate. A total of 14 samples were taken during the batch experiment and 17 during the medium exchange. Parameter values were obtained from experimental data (maximum uptake rates, maximum growth rate, sucrose maximum hydrolysis rate and some of the maximal accumulation levels) and from literature, taking values from other plant species when unavailable for C. roseus (Tables 5, 6, 7). Adjustment of the unknown parameters as well the values taken from literature was performed manually in parallel using a non-linear least-square algorithm (lsqcurvefit) from Matlab software. The least-squares criterion was applied on all the data points (triplicates taken as mean in (Mi)exp). The sum of squares were weighted as described below. whereTable 5Affinity constants (Km)ComponentValueLiteratureSpeciesUnitsReferencesAA0.0136Catharanthus roseusmmol/g DWModel calibrationFRU0.120C. roseusmmol/g DWModel calibrationGLC0.120C. roseusmmol/g DWModel calibrationLIP0.00254C. roseusmmol/g DWModel calibrationNAD0.00023C. roseusmmol/g DWModel calibrationNADH0.00030C. roseusmmol/g DWModel calibrationNADP0.000585C. roseusmmol/g DWModel calibrationNADPH0.00037C. roseusmmol/g DWModel calibrationNDP0.010C. roseusmmol/g DWModel calibrationNH40.2279C. roseusmmol/g DWModel calibrationNTP0.00625C. roseusmmol/g DWModel calibrationORA0.00807C. roseusmmol/g DWModel calibrationPi0.1997C. roseusmmol/g DWModel calibrationPPi0.0737C. roseusmmol/g DWModel calibrationSUC1.00C. roseusmmol/g DWModel calibrationSTA1.00C. roseusmmol/g DWModel calibrationSTH0.2C. roseusmmol/g DWModel calibrationOP0.0206C. roseusmmol/g DWModel calibrationEFRU0.04616Daucus carotamM[17]EGLC0.05356D. carotamM[17]ENO3_HA0.06770.281Citrus reticulatamM[45]0.005–0.2Arabidopsis thaliana[46]0.006–0.02Zea mays[43]ENO3_LA1.430.5A. thalianamM[46]0.02–0.1Z. mays[43]0.16D. carota[17]EPI_HA0.00260.007C. roseusmM[47]0.0056C. roseus[48]0.0035C. roseus[49]0.003C. roseus[49]0.0025Nicotinana tabacum[50]0.0079Lemma gibba[51]0.00049Z. mays[52]0.0018Z. mays[52]0.007Chara corallina[37]0.004C. corallina[37]0.025Nitella translucens[53]0.0024Neurospora crassa[54]0.0029N. crassa[54]0.0026D. carota[17]EPI_LA0.0800.900C. roseusmM[49]0.0463C. roseus[49]0.058Nicotinana glutinosa[55]0.076L. gibba[51]0.190C. corallina[37]0.220C. corallina[37]0.370N. crassa[54]1.029N. crassa[54]0.47D. carota[17]ESUC26.69C. roseusmMModel calibrationNO30.5181C. roseusmmol/g DWModel calibrationTable 6Maximum reaction rates (Vmax)Reaction in TPMPCurrent value (mmol/g DW/d)Literature (mmol/g DW/d)SpeciesReferencer174.9934C. roseusModel calibrationr22C. roseusModel calibrationr30.764C. roseusModel calibrationr410,000C. roseusModel calibrationr5150C. roseusModel calibrationr61C. roseusModel calibrationr70.1C. roseusModel calibrationr83.0C. roseusModel calibrationr930.0C. roseusModel calibrationr102.00C. roseusModel calibrationr116.00C. roseusModel calibrationr120.05C. roseusModel calibrationr1315.0C. roseusModel calibrationr142.00C. roseusModel calibrationr165C. roseusModel calibrationr180.764C. roseusModel calibrationr190.1C. roseusModel calibrationr200.01C. roseusModel calibration2115.1C. roseusModel calibration224.6929C. roseusModel calibration230.4986C. roseusModel calibration240.100C. roseusModel calibration251.1377C. roseusModel calibration2633.5326C. roseusModel calibration27 HA0.00150.1555C. reticulata[45]0.072–0.1968Z. mays[43]27 LA0.1670.72–1.92Z. mays[43]281.001.69D. carota[17]290.382C. roseusModel calibration30900C. roseusModel calibration31 HA0.001753.80C. roseus[47]0.415C. roseus[48]0.127C. roseus[49]0.196C. roseus[49]1.96N. tabacum[50]0.0369N. tabacum[50]0.1728L. gibba[51]0.13824Z. mays[52]0.0092C. corallina[37]0.0576N. translucens[53]3.80N. crassa[54]0.0030D. carota[17]31 LA0.0401.268C. roseus[49]0.090C. roseus[49]0.046N. glutinosa[55]0.265L. gibba[51]0.023L. gibba[51]0.104C. corallina[37]0.0576C. corallina[37]10.4N. crassa[54]5.64N. crassa[54]0.045D. carota[17]GLC0.764C. roseusModel calibration320.1400.27D. carota[17]Table 7State variables and initial values (t=0) used for model simulationsComponentValuesUnitsSpeciesReferencesAA0.818/0.779mmol/g DWC. roseusBatch/medium exchangeFRU0.0336/0.0556mmol/g DWC. roseusBatch/medium exchangeGLC0.1625/0.1744mmol/g DWC. roseusBatch/medium exchangeSEC0.00515mmol/g DWC. roseusUnpublished resultsLIP0.1mmol/g DWN. tabacum suspension cells[34]NAD7×10−5mmol/g DWYoung photosynthetic tissue[56]NADH1.25×10−5mmol/g DWYoung photosynthetic tissue[56]NADP8×10−6mmol/g DWYoung photosynthetic tissue[56]NADPH1.12×10−4mmol/g DWYoung photosynthetic tissue[56]NDP2×10−3mmol/g DWYoung photosynthetic tissue[56]NH40.053/0.041mmol/g DWC. roseusBatch/medium exchangeNTP6.8×10−3mmol/g DWYoung photosynthetic tissue[56]ORA0.16mmol/g DWN. tabacum suspension cells[34]Pi0.117/0.100mmol/g DWC. roseusBatch/medium exchangePPi0.0015mmol/g DWPotato tubers[34]SUC0.403/0.330mmol/g DWC. roseusBatch/medium exchangeSTA1.705/1.137mmol/g DWC. roseusBatch/medium exchangeSTH0.081mmol/g DWN. tabacum suspension cells[34]TRY0.001428mmol/g DWC. roseusUnpublished resultsOP0.01mmol/g DWC. roseusModel calibrationEFRU8.33/29.6mMC. roseusBatch/medium exchangeEGLC7.22/23.0mMC. roseusBatch/medium exchangeENH40mMC. roseusCulture mediumENO33.27/3.27mMC. roseusBatch/medium exchangeEPi0.100/0.100mMC. roseusBatch/medium exchangeESUC114.0/149.0mMC. roseusBatch/medium exchangeNO30.385/0.345mmol/g DWC. roseusBatchX0.0197/0.0180gDWC. roseusBatch/medium exchange Error terms were weighted using the maximum (max(Mi)exp) and the minimum (min(Mi)exp) experimental values (exp) from either batch or medium exchange cultures. This method was preferred to a method using experimental error because the experimental error was very high for some data points and varied with experimental data (Table 8). This method has also showed previously to perform adequately to calibrate a nutritional model on hairy root data [17]. All the experimental data points from batch and medium exchange culture were used for the non-linear least-square regression. The total number of experimental data was then of 411 with 13 different measurements for each of the 31 samples plus eight measurements for intracellular AA. Many set of parameters were tested as initial guesses to improve the fitting. A combination of parameter values (Tables 5, 6) showed to minimize the global error for combined data from batch and from medium exchange cultures, using a convergence criterion of 1×10−6 on the variation of the global error. From the calculation of R2 values (Table 8) for each of the 28 state variables that were fitted (14 state variables in two experiments), it seems that the best fit is obtained for cells growth and nutrient transport and storage (NO3, Pi, NH4). However, the sensitivity analysis also suggests that some parameters can still be optimized (Fig. 4). A critical problem with such a descriptive model is a high number of parameters. The model has 35 maximum reaction rates and 26 affinity constants and the least-square regression was only performed on only 14 of the 31 state variables of the model. It is then possible that the identified solution corresponds to a local minimum. More measurements on metabolite concentration are thus necessary to improve the performance of the model and for being able to perform an adequate identifiability analysis on the parameters. In this context, the proposed model is probably over-parameterized. However, the goal of this work was not to establish a minimal model but a descriptive model that can be used as a tool useful at improving our understanding of plant cell culture.Table 8Mean standard deviations on measured variables and correlation coefficients (R2)ComponentMean SD (%) (batch)Mean SD (%) (medium exchange)R2 (batch)R2 (medium exchange)AA11.8813.830.05210.0413FRU9.9234.960.00120.0415GLC10.8148.290.00580.1059NH49.8626.100.60500.8628Pi35.8512.010.75680.8303SUC25.7945.750.00320.0781STA9.0112.440.00100.0013EFRU15.0618.620.09130.0612EGLC13.2613.910.08380.0785ENO340.766.050.94560.9459EPi26.2743.200.98950.7737ESUC11.2610.830.91230.9461NO313.2624.960.65410.4103X14.7820.460.97500.9708Fig. 4Sensitivity analysis on model parameters for batch (solid bar) and medium exchange (empty bar) cultures. Relative change in error is calculated as described in Sect. 7.1. Parameter adjustment of +50% (a). Parameter adjustment of −50% (b). Parameters not shown have absolute relative error changes that are below 0.05 Simulation of the cells physiological state Hairy root growth The model allows estimating root growth from flux ν(32) as well as from the total estimated mass of the root cells constituents per dish (Σ Mi on a mass basis), taking average molecular weights for the cell building blocks (Fig. 5). However, the total mass calculated at initial condition (t=0) was 25% lower than that of the inoculum with 0.0148 g DW per dish as compared to 0.0197 g DW per dish, respectively. The difference of −0.0049 g DW per dish can be explained from the molecular species that are present in the cells but which are not included in the model. Therefore, the estimation of root mass with time obtained from ν(32) was compared with that calculated from total cell mass, but adding a constant correction factor of +25%. Model simulations then describe hairy root growth for both batch and medium exchange cultures. Both root mass estimates from ν(32) and ΣMi are superimposed (Fig. 5). However, the sum of the molecular species is overestimating measured biomass at the end of batch culture because of cell accumulation in diverse compounds while growth slowed down. Surprisingly, both batch and medium exchange cultures behaved similarly with a maximum specific growth rate of 0.035 d−1 and a growth cessation at around 40 d. Periodic medium renewal has avoided nutrient depletion in macronutrients and sugars (Fig. 6). However, the plateau observed for the medium exchange culture may have resulted from a limitation in micronutrients, which were not measured. Therefore, only experimental data obtained before the occurrence of the plateau for the medium exchange culture were used for model calibration and other analyses. The strategy of describing cell growth as a function of intracellular content in building blocks thus showed to perform adequately. This result may support the global model structure as well as the hypothesis of pseudo-steady state for the SPMP. The central primary metabolic network, described here as SPMP, was observed at pseudo-steady state by Rontein et al. [19] for most of in vitro culture duration for tomato suspension cells [20]. However, this hypothesis has to be further investigated.Fig. 5Model simulation for Catharanthus roseus hairy root batch (filled square) and medium exchange (triangle) liquid cultures in Petri dish. a–e. Intracellular concentration in cell building blocks with time. Batch (f) and medium exchange (g) cultures. The specific growth rate is described from ν(32) as a function of intracellular concentration in cell building blocks. Model simulations for batch (solid lines) and medium exchange (dashed lines) cultures, and total mass for the molecular species accounted in the model increased of 0.0049 g DW accounting for non-estimated molecular species (dashed line). Root mass was calculated using average MW for STA (180.15, based on glucose units); SEC (400); TRY (187); NAD/NADH (712/713); NADP/NADPH (744.4/745.4); NDP/NTP (476/507); ORA (809.75); STH (180.15, based on glucose units); OP (average of 305 from nucleotides); LIP (810, based on that of A-CoA); IPP (246) and AA (average of 136.75)Fig. 6Extracellular concentration in nutrients with time. Experimental data for C. roseus hairy root batch (filled square) and medium exchange (triangle) liquid cultures. Concentration in sugars (a–c), inorganic phosphate (d) and nitrate (e) in the culture media with time. Model simulations for batch (solid lines) and medium exchange (dashed lines) cultures The cells building blocks The measured total AA concentration decreased with time for the batch culture and reached a plateau for the medium exchange culture (Fig. 5a). Simulated AA concentrations seem to follow the trend for experimental data but were overestimated for both batch and medium exchange cultures. The deviations cannot be explained, however, by exuded AA and proteins since no AA were detected in the used culture media. Values of 0.63±0.10 to 0.82±0.10 mmol AA/g DW were measured for the batch culture which corresponds to 0.088±0.014 and 0.12±0.014 g AA/g DW, respectively (taking an average amino acid MW of 140 g/mol). Higher values were measured in the medium exchange culture with 0.63±0.15 to 1.1±0.14 mmol AA/g DW, which represents 0.12±0.021 and 0.16±0.020 g AA/g DW, respectively. Free AA content of 0.21 mmol AA/g DW (estimated at 0.029 g AA/g DW) is reported in literature for tobacco suspension cells using B5 medium [34] and 0.15 mmol AA/g DW (estimated at 0.021 g AA/g DW) in potato tuber [35, 36]. In tomato cell culture, the total protein content was of 0.1–0.4 g proteins/g DW [19]. The total AA mass per root dry weight that have been measured, which represents the sum of free AA and that of peptides and proteins, is thus within the range reported in literature. The use of a minimal medium may explain a low value for total AA. The other cell building blocks were not measured but the model estimations were close to literature data. Total ORA concentration was estimated at 0.15 mmol/g DW as compared to 0.42 mmol/g DW reported by Farré et al. [35]. Total STH were estimated at 0.14 mmol/g DW as compared to a range of 0.3–1.6 mmol/g DW (on the basis of glucose units). Maximal simulated rates of biosynthesis were compared with that from literature. The rate of AA biosynthesis (v(r5)) was estimated at 0.12 mmol/g DW/d, which was within the range of 0.094–0.36 mmol/g DW/d reported by Rontein et al. [19] for tomato suspension cells. STH biosynthesis (v(r12)) was slow with 0.009 mmol/g DW/d as compared to 0.36 mmol/g DW/d [19]. Maximal net rate of starch (STA) biosynthesis (v(r14)) was 0.20 mmol/g DW/d as compared to 0.15 mmol/g DW/d [19]. Differences in metabolic fluxes may be explained from the differences in the specific growth rate. Rontein et al. [19] have reported a specific growth rate of 0.4 d−1 for tomato suspension cells in B5 medium as compared to 0.035 d−1 in this study for C. roseus using a minimal medium. The same minimal medium has been reported previously to induce lower growth rate as compared to standard media for carrot hairy root [17]. The cells nutritional state Extracellular sucrose (ESUC) concentration profile was simulated by the model for both the batch and the medium exchange cultures (Fig. 6), thus suggesting that the sucrose hydrolysis modelling strategy was adequate. In the case of glucose (EGLC) and fructose (EFRU), simulations for the batch culture followed the trend of experimental data but before growth cessation. Then, estimated glucose and fructose levels stayed high as compared to experimental data. This may suggest that the model is underestimating glucose and fructose uptake at reduced growth. However, simulated concentrations in intracellular sucrose (SUC) and glucose (GLC) were following similar trends than experimental data (Fig. 7). However, model simulated constant starch (STA) concentration and overestimated fructose (FRU) concentration. In the case of the medium exchange culture the model simulated adequately experimental data on extracellular and intracellular sucrose and intracellular glucose. Extracellular glucose and fructose concentrations were underestimated and starch and fructose concentrations were overestimated. A maximal sucrose rate of hydrolysis (v(28)) of 0.72 mmol/g DW/d was estimated as compared to 1.7 mmol/g DW/d for D. carota hairy root using same minimal medium [17]. Contents of 0.25 mmol SUC/g DW (batch) and 0.7 mmol SUC/g DW (medium exchange) were obtained as compared to 0.05 mmol SUC/g DW [19], 0.26 mmol SUC/g DW [36] and 0.4 mmol SUC/g DW [34]. For free glucose, cell contents of 0.3–0.05 mmol GLC/g DW were obtained as compared to 0.6 mmol GLC/g DW [19], 0.24 mmol GLC/g DW [36] and 0.18 mmol GLC/g DW [34]. For fructose, cell contents of 0.1–0.25 mmol FRU/g DW were observed as compared to 0.009 mmol FRU/g DW [35], 0.21 mmol GLC/g DW [34] and 0.5 mmol GLC/g DW [19]. Starch accumulation reached 1.25 mmol STA (based on glucose)/g DW as compared to 0.055 [34] and 0.1–0.6 mmol STA (based on glucose)/g DW [19].Fig. 7Intracellular concentration in nutrients and metabolites with time. Experimental data for C. roseus hairy root batch (filled square) and medium exchange (triangle) liquid cultures. Intracellular concentration in sugars (a–d), inorganic phosphate (e), nitrate (f), ammonium (g), and in secondary metabolites issued from tryptamin (h) and secologanin (i) with time. Model simulations for batch (solid lines) and medium exchange (dashed lines) cultures Model simulations of extracellular Pi and NO3− (ENO3) concentrations followed experimental data for the complete duration of the batch and the medium exchange cultures (Fig. 6). Hairy root growth (simulated and experimental) has ceased concurrently to ENO3 depletion suggesting this ion to be limiting in the batch culture. Intracellular Pi and NH4 were simulated (Fig. 7). It is interesting to note that the final intracellular Pi concentration in the batch culture seemed to reach a plateau at 0.05 mmol/g DW which was simulated. This plateau may be close to the Pi level that is essential to maintain the endogeneous metabolism [37]. A maximum Pi accumulation plateau was also observed for the medium exchange culture around 0.2 mmol/g DW. This value is close to the maximum value of 0.23 mmol/g DW measured for carrot hairy root [17, and references therein]. Maximum intracellular nitrate of 0.6 mmol NO3/g DW was measured for both cultures. Intracellular nitrate was kept constant for the medium exchange and decreased to its initial value for the batch. Intracellular ammonium reached a value of 0.12 mmol NH4/g DW for the medium exchange culture and decreased to 0.09 mmol NH4/g DW after 30 d. For the batch culture, intracellular ammonium reached a maximum value (0.08 mmol NH4/g DW) at day 5 and then decreased closely to its value at inoculation. The model overestimated intracellular NO3 concentration and simulated experimental intracellular concentrations in NH4 for both the batch and medium exchange cultures. A more precise description of the secondary metabolism and of the AA biosynthesis may be required to improve model simulation of nitrogenous compounds. The cells energetic state Model has simulated energy rich cells with high levels in NTP and NADH at exponential growth (Fig. 8). However, estimated cell content in NADPH showed an opposite trend with lower values at exponential growth. This result is interesting since NADPH is produced through the pentose-phosphate pathway, which is mostly active at secondary metabolism. The simulated cell energetic level decreased after 20 d at the initial value (inoculation) in batch culture and was constant for the medium exchange culture. Since the medium was renewed every 3 days for the medium exchange experiment, a pseudo-steady state was expected. NTP/(NTP+NDP) and NADH/(NAD+NADH) ratios of ∼75% have been reported for tobacco suspension cells [34]. Farré et al. [35] measured ratios of ∼73% for NTP/(NTP+NDP) in potato tubers. These authors have measured an NTP level of 0.00013 mmol/g DW as compared to 0.00143 and 0.00074 mmol/g DW that was obtained from simulations at exponential growth and stationary growth, respectively. We have recently measured by in vivo NMR ATP levels of 0.009 mmol/g DW at day 1 and 5, and of 0.0065 mmol/g DW at day 10 for suspension cells of E. californica cultured using B5 medium [38]. Energetic levels simulated by the model were then similar to that found in literature. However, variation of energy shuttles concentration has to be further investigated.Fig. 8Energetic level of the cells with time. Intracellular NTP (a), NADH (b) and NADPH (c) relative levels with time. Model simulations for C. roseus hairy root batch (solid lines) and medium exchange (dashed lines) liquid cultures Conclusion A metabolic model capable of describing hairy root growth from the estimation of the cells physiological state was developed. The model includes the central metabolism, the primary metabolic pathways (SPMP) assumed at steady state and a network for the TPMP. At this point, the model is simplified and intracellular compartmentalization processes into the different cell compartments and organelles are not included. Nevertheless, the metabolic model showed to perform efficiently in simulating hairy root growth and nutrition. The use of intracellular concentrations in nutrients and co-substrates as well as the cells energetic state seems an efficient strategy in describing regulation of the metabolic fluxes. However, more experimental is required for improving the model structure as well as parameter values which may be regulated with cell physiological state. Finally, the model will be applied to other plant species as well as cell suspensions, and it will be studied as a tool to describe transient processes such as metabolic regulation.

J_Biomol_NMR-3-1-2048825

A global analysis of NMR distance constraints from the PDB

Information obtained from Nuclear Magnetic Resonance (NMR) experiments is encoded as a set of constraint lists when calculating three-dimensional structures for a protein. With the amount of constraint data from the world wide Protein Data Bank (wwPDB) that is now available, it is possible to do a global, large-scale analysis using only information from the constraints, without taking the coordinate information into account. This article describes such an analysis of distance constraints from NOE data based on a set of 1834 NMR PDB entries containing 1909 protein chains. In order to best represent the quality and extent of the data that is currently deposited at the wwPDB, only the original data as deposited by the authors was used, and no attempt was made to ‘clean up’ and further interpret this information. Because the constraint lists provide a single set of data, and not an ensemble of structural solutions, they are easier to analyse and provide a reduced form of structural information that is relevant for NMR analysis only. The online resource resulting from this analysis (http://www.ebi.ac.uk/msd/srv/docs/NMR/analysis/results/html/comparison.html) makes it possible to check, for example, how often a particular contact occurs when assigning NOESY spectra, or to find out whether a particular sequence fragment is likely to be difficult to assign. In this respect it formalises information that scientists with experience in spectrum analysis are aware of but cannot necessarily quantify. The analysis described here illustrates the importance of depositing constraints (and all other possible NMR derived information) along with the structure coordinates, as this type of information can greatly assist the NMR community. Introduction The information contained in the world wide Protein Data Bank (wwPDB) (Berman et al. 2007) is growing steadily, with increasing numbers of structures being deposited from both traditional single laboratory sources and recent structural genomics efforts. The two main methods to determine these structures are X-ray crystallography and Nuclear Magnetic Resonance (NMR). NMR structures account for around 15% of all entries in the wwPDB. While inherent size restrictions limit the method to molecules of lower molecular weight, NMR has still made a significant contribution to protein folds and molecular interaction data. However, NMR structures are less straightforward to use than structures determined by X-ray because they are often represented as ‘ensembles’ of structures, where the whole ensemble (and not individual structures by themselves) represents the solution of the structure determination problem based on the experimental data. The reason for this is that information derived from NMR is insufficient with respect to the structure calculation process and thus cannot lead to a single exact solution. For example, measured distance-related NOE data is ensemble and time averaged, so that the final observed NOE data for a set of multiple distinct conformations in fast exchange will be a degenerate mix of the distance information in each of those conformations. All calculated structures that conform to a set of criteria based on the fit to the experimental data (NMR derived constraints) and physical characteristics encoded in the calculation process (overall energy minimum, inter-atomic packing, …) are therefore in principle equally valid (Spronk et al. 2003). However, there is no ‘standard’ set of criteria, and different programs and researchers use different sets when selecting the ‘best’ structures out of the calculated ones. The quality of the NMR structures deposited at the wwPDB therefore varies widely (Nabuurs et al. 2006). This problem is being addressed by novel structure determination methods like Inferential Structure Determination (ISD) (Rieping et al. 2005) that perform more objective structure calculations and select statistically relevant ensembles as a representative solution to the experimental data. These methods are, however, computationally expensive. The observed NMR data is used in structure calculations to constrain inter-atomic distances, relative bond orientations and/or dihedral angles. Only just over half of the NMR entries in the wwPDB were deposited together with the constraint lists used in the structure calculation process. These constraint lists provide very valuable information, as, for example, they enable recalculation of the structures with better or different protocols, and a globally consistent comparison of the constraints to the original coordinates for validation purposes. However, a major problem in using this information is that constraint files come in many file formats, and that the atom naming and residue numbering in the coordinate and the constraint files often differs. To handle the problem of different file formats, constraint lists are now continuously converted into the NMR-STAR format at the BioMagResBank (BMRB) (Doreleijers et al. 2003; Ulrich et al. 1989). A further step, as part of the DOCR project (Doreleijers et al. 2005), addresses the atom naming and residue numbering problem by directly relating the constraint atoms to the coordinate atoms and molecular system using the FormatConverter and CCPN data model (Fogh et al. 2002; Vranken et al. 2005). This data is then checked for consistency and redundant constraints are removed using the WATTOS software as part of the FRED project (Doreleijers et al. 2005). In the original implementation the DOCR/FRED project resulted in a set of more than 500 internally consistent constraint lists, molecular system information and structure coordinates. This data led to the construction of the RECOORD (Nederveen et al. 2005) and DRESS (Nabuurs et al. 2004) databanks, where structures from the PDB were respectively recalculated and re-refined based on the cleaned up constraint lists. These efforts, as well as validation software like AQUA (Doreleijers et al. 1998) or Procheck-NMR (Laskowski et al. 1996), are applied on a per-entry basis, where the original or recalculated coordinates are related to the original or standardised constraint lists. With the amount of constraint data that is now available it has become possible to do a global, large-scale analysis using only the information from the constraints, without taking the coordinate information into account. The constraint lists are the ‘final product’ of the usually human analysis of NMR spectra, and as such they represent the experimental NMR-derived information that is relevant for the structure calculation. The constraint lists also provide a single set of data, whereas NMR structures are usually represented as ensembles that can be calculated in many different ways, which complicates their interpretation. Analysing the constraint data on a large scale can thus provide insights into the NMR data analysis process (e.g. which type of inter-atomic contact is often derived from the spectra), and the relation of the constraints to the coordinate data (e.g. does the structure calculation process add any distance information that is relevant for NMR). The analysis described here relates only to distance constraints derived from NOE data, with a base set of 1834 NMR PDB entries containing 1909 protein chains. Only constraints between protons in protein chains were retained for analysis, and for validation purposes the base set was further divided into subsets for entries that contain intra-residue constraints and entries where all the original constraint and coordinate information was recognised and linked to each other. A coordinate data set based on the original coordinate files was also generated and used for comparison. This article explores some of the issues surrounding distance constraints and the NMR data they are derived from, and hopes to highlight the importance of depositing the constraint lists used for structure calculations along with the molecular coordinates. Materials and methods The data was obtained from two sources: the molecular system, coordinate, secondary structure and author information from the wwPDB, and the original constraint information as NMR-STAR files from the NMR Restraints Grid at the BioMagResBank. Each file was directly parsed and combined into the CCPN data model via the FormatConverter software (Vranken et al. 2005), in a process that extends the procedure used in the original DOCR/FRED project. To handle the larger number of entries an automatic mapping procedure was developed that maps the molecule sequence as derived from the PDB file to the atom information and sequence code numbering used within the constraint lists. For some entries manual mapping between the information from the PDB file and the constraint file was required. The original 409 mappings from the RECOORD project (by Aart Nederveen) and a further mappings 124 by the author and 17 by Jurgen Doreleijers were used to correctly set the PDB-constraint file mapping for the entries in the base set. Using the automatic or manual mapping the atom information from the constraints was then connected to that for the molecule. During this process the dependence on string-based atom names for the assignment (as used in all constraint files) was also removed, since the CCPN data model is object-based. The final CCPN project, in which all the information from the original files is now highly organised, was then written out. This process was completed for 2643 PDB entries. Three subsets of the base set were generated for comparison and validation purposes (Table 1). Sets HIP and AHIP contain only entries with intra-residue constraints, sets AHP and AHIP only include entries where over 99% of relevant constraint information was assigned to atoms. In addition a set was generated based on the coordinates for the entries included in the HP set. Only the HP, HPC and AHIP sets are further referred to in this article, the HIP and AHP sets are available online for reference purposes. Table 1Overview of the available data sets used in the analysisSet NameData typeDetails1HPConstraintsBase set2HIPConstraintsIntra-residue3AHPConstraintsHigh assignment4AHIPConstraintsIntra-residue, high assignment5HPCCoordinatesOriginal coordinate data A workflow based on Python (van Rossum 2003) scripts and dictionaries was then developed to handle the information from the CCPN project files for the 2643 entries (Fig. 1). This workflow was run separately for the base set and each subset. In the first step, the original information was filtered so that only valid constraints between protons in protein chains were retained (Fig. 1). Entries were removed for the following reasons (Table 2): (1) No valid protein chains: the entry contained only chains that are sequence fragments or duplicates of other chains (only the chain with the highest number of valid constraints linked to it is retained in the data set), (2) No valid constraint lists: the originally deposited constraint lists could not be parsed or handled correctly, or (for sets HIP and AHIP), did not contain any intra-residue constraints, (3) Insufficient linking: less than 80% (sets HP, HIP) or 99.9% (sets AHP, AHIP) of the constraint information could be linked to the atom information (constraint information belonging to non-protein chains was ignored for this purpose), (4) Insufficient valid constraints: less then 20% of constraints remained for all lists after removing invalid constraints. Constraints were considered invalid if they did not have any valid items, no upper distance limit, an upper distance limit of larger than 10 Å, or had only items between invalid atoms (non-proton, unlinked or non-protein). Note that some entries had specific distance constraint lists and/or constraints removed for these reasons, but were still included in the analysed set. Fig. 1Overview of the workflow employed in the analysis. Grey boxes indicate files, white boxes Python scriptsTable 2Overview, for each data set, of the number of removed and analysed entriesHPHIPAHPAHIPNo valid protein chains396396396396No valid constraint lists310409310409Insufficient linking4837677635Insufficient valid constraints55555555Total included entries1834174612031146Total included chains1909181712521192 In the second step, the data from all relevant entries was collated (Fig. 1). After reading in the CCPN projects, non-relevant information was removed based on the filtering information, and the data reorganised into a set of Python dictionaries that contain the overall information on the entry, residue and constraint levels. In the final analysis step these dictionaries were read in, areas and data of interest were marked, and organised HTML output was produced for browsing. A Python dictionary with highly reduced information that can be further used for validation or constraint filtering purposes was also generated. The same 1834 entries from the HP set were used to generate the HPC set. The original coordinates were analysed using r−6 distance averaging for equivalent and prochiral atom sets over all structures in the ensemble (Nilges 1995). Only individual distances of less then 7 Å were retained, and final averaged distances of less than 5 Å were considered to be equivalent to an observable constraint contact. Atoms without coordinates were ignored in the analysis. Highlights from the analysis are described in the results section, and complete details are available from a web site (http://www.ebi.ac.uk/msd-srv/docs/NMR/analysis/results/html/comparison.html). For all statistical operations, the R package (Bates et al. 2007) was accessed via the RPy Python module (Moreira and Warnes 2006). Since a contact is either observed or not observed, it was possible to use a binomial analysis to determine, for example, which secondary structure specific contacts were significantly less or more likely to be observed. Binomial analysis was used throughout with a confidence level of 0.99, meaning that only 1 out of every 100 determined outliers is a false positive. The correlations between the coordinate and the constraint data within a data set and correlations between data sets were plotted via RPy, with linear correlation coefficients determined by both the Spearman and Pearson methods. The Spearman method (Spearman 1904) is a non-parametric measure of correlation, which does not make assumptions about the frequency distribution of the variables, and does not require a linear relationship between the variables. The Pearson method (Pearson 1896) on the other hand assesses if the relationship between the variables is linear. In the analysis, a residue is marked as ‘assigned’ when at least one proton belonging to it is linked to a constraint. The total number of times a particular inter-atomic contact is observed can be a fraction, as for ambiguous constraints each constraint item contributes a fraction of 1 to the total: The total number of relevant distances for a particular inter-atomic contact (ndist) is always an integer. The occurrence for a particular inter-atomic contact is calculated as: where nactual is the number of times the contact occurs (with ambiguity taken into account), and npossible the total number of times this contact could occur between the relevant residues, either for all residues or only between ‘assigned’ residues. The occurrence is given in percent or as a fraction. The ‘ambiguity’ of an inter-atomic contact is defined as: If the ambiguity is 0, this means that all contributing contacts are unambiguously assigned. The more highly ambiguous the contributing contacts are, the closer this number will be to 1, which also means it is less dependable. Within secondary structure combinations, the same definitions are used, except that npossible and nactual are now determined within that secondary structure combination as npossible,ss and nactual,ss. The uniqueness of a contact for a secondary structure combination, which indicates how unique the contact is within that combination, is defined as: For prochiral and possible non-equivalent atom sets, the contacts are divided into the individual atom names (e.g. Hβ2 and Hβ3, or Hδ1 and Hδ2 for Phe) if they occur by themselves in a constraint. If both the atom names occur in different items within the same constraint to the same other atom, then they are grouped (e.g. Hβ*, or Hδ* for Phe). If they occur as a group in one constraint item they are used as is. In order to obtain total statistics for these contacts, they were also combined (e.g. (Hβ2 + Hβ3), or (Hδ1 + Hδ2) for Phe). To get the combined occurrence, the nactual for the group is added to the individual contribution with the highest nactual to obtain the nactual used to calculate the occurrence for the combined contact. The combined distance information is obtained from all contributions, and the ambiguity is then calculated from the total nactual of all contributions. Results Information from analysis The analysis of the data for each set is divided into separate categories (visible on the left-hand side menu in Fig. 2): (1) Contact analysis: Arranges inter-atomic distances by residue–residue combination, secondary structure of those residues, and contact type (intra-residue (i–i), sequential (i–i + 1), medium-range defined here as up to 6 residues separation (i–i + n), and long range (i–i + 6<)). (2) Protein secondary structure analysis: Groups inter-atomic distances by secondary structure combinations. (3) Residue atom analysis: Shows assignment percentages for the atoms in each amino acid. (4) Unassigned fragments breakdown: Lists unassigned sequence fragments. (5) Fragment analysis by residue: Analyses tripeptide fragments based on the assignment status of the central residue. Fig. 2Example web page for Ala → Ala i → i + 2 contact information from http://www.ebi.ac.uk/msd-srv/docs/NMR/analysis/results/html/comparison.html. The different data sets, comparisons between them, and analysis categories can be accessed via the left hand side menu. In this contact information page, overall statistics for the individual Ala residues involved in this contact are shown in the ‘Residue 1’ and ‘Residue 2’ tables at the top, statistics for the combined Ala–Ala residues in the ‘Residue combination’ table. The information on an atom level is listed in the ‘Breakdown per contact’ table. Outliers per secondary structure combination based on a binomial analysis are highlighted in red (higher than expected) and blue (lower than expected). This colour coding is used throughout the web pages for other types of analyses Validation of data sets The relevance of the base set (HP) was validated by comparison with the most restricted AHIP set. This is necessary because the HP set contains entries without intra-residue constraints and entries where less constraint information is linked to atoms. The correlation between the occurrences of contacts between both sets is very high (0.97) (Fig. 3), and in a breakdown per contact type no correlation is less than 0.91 (results on web pages). A further detailed analysis shows that out of 43,984 compared contacts only 5 differ significantly at a confidence level of 0.999. Of these, 3 are His → His contacts where the differences are most likely introduced by changes in the number of His residues that are assigned in the data sets (from 71.66% in HP to 68.47% in AHIP). The other 2 are long-range contacts involving side chain protons between Leu → Phe and Phe → Val, which are uncommon, and these differences can be attributed to accidental variations in the number of entries that have these contacts and are included in the data set. Overall, this data shows that including the additional 688 entries does not introduce major changes in the occurrence levels. The HP set was therefore chosen as the reference set, as estimates of, for example, significant differences between contacts in secondary structure elements tend to become more reliable as more data is included. Information on correlations between all data sets is available from the web pages (Fig. 2, the ‘Comparison’ link in the left-hand side menu). Fig. 3Correlation between the foccurrence for the HP and AHIP data sets (correlation Spearman 0.971, Pearson 0.996) A further way to validate the constraint information is that it should reproduce the typical contacts that are observed in secondary structure elements. The secondary structure definitions as determined by DSSP (Kabsch and Sander 1983) were taken from the PDB file header for each entry. Because of the size of the data that is generated by the analysis, only the Ala residue is used here as an example. The complete information can be accessed via the web pages (Fig. 2). The numbers that are relevant in order to determine whether a contact is significant for a particular secondary structure element are the occurrence within that particular secondary structure element (which has to be significantly higher than expected from the overall occurrence), the ambiguity of the contact (if it is high it comes, by definition, from highly ambiguous constraints and is therefore unreliable), and the uniqueness of the contact (which indicates how often the contact is seen within that secondary structure element compared to the total number of times it is observed). For intra-residue contacts there are not many significant differences in the observed occurrence within different secondary structure elements, although overall fewer contacts are observed when secondary structure is absent (Table 3). This is likely due to higher signal overlap for atoms in ‘random coil’ fragments, which complicates assignment. The information from the coordinate-derived HPC set is provided throughout for comparison: there are, as expected when using a distance cutoff of 5 Å, no differences for this HPC set between the occurrence for different secondary structure elements, and only minor differences in the average and mean distance are observed (data available on web pages). Table 3foccurrence for selected backbone Ala–Ala contactsTypeAtom 1Atom 2α-helixβ-sheetNo secondary structurei → iHHα0.50/1.000.51/1.000.43/1.00HHβ*0.73/1.000.68/1.000.57/1.00HαHβ*0.32/1.000.33/1.000.29/1.00i → i + 1HH0.84/1.000.50/1.000.57/1.00HHβ*0.22/0.990.07/0.690.10/0.83HαH0.52/1.000.90/1.000.67/1.00HαHβ*0.03/0.420.30/1.000.11/0.93i → i + 2HH0.47/0.990.00/0.000.16/0.45HHβ*0.09/0.790.00/0.000.03/0.25HαH0.28/0.970.04/0.040.17/0.60HαHβ*0.02/0.630.00/0.000.02/0.27Hβ*H0.22/0.980.09/0.450.17/0.77Hβ*Hβ*0.02/0.120.05/0.700.01/0.51i → i + 3HH0.09/0.860.04/0.120.04/0.16HHβ*0.12/0.840.04/0.120.04/0.22HαH0.66/0.990.00/0.040.14/0.28HαHβ*0.58/0.990.00/0.040.14/0.33Hβ*H0.14/0.910.04/0.160.06/0.34Hβ*Hβ*0.11/0.970.04/0.120.07/0.46i → i + 4 HH0.02/0.010.00/0.000.02/0.03HHβ*0.01/0.010.05/0.090.02/0.07HαH0.32/0.940.00/0.000.07/0.16HαHβ*0.02/0.790.00/0.000.02/0.23Hβ*H0.09/0.850.00/0.090.02/0.22Hβ*Hβ*0.03/0.930.00/0.090.01/0.37In the secondary structure columns, the first value is from the HP set, the second from the HPC set. A bold value indicates a contact that occurs significantly more than average, a italic value significantly less For sequential Ala–Ala contacts, the H–H contact is observed significantly more in α-helices than in β-sheets (Table 3), and is also highly unique (0.65). This situation is, as expected, reversed for the Hα–H contact, which is observed significantly less in α-helices and more in β-sheets. Overall, however, this contact is more uniquely observed in α-helices (0.49 compared to 0.08 for β-sheets), which is due to the prevalence of Ala–Ala fragments in α-helices (56.2% compared to 5.35%). Interestingly, however, an H–Hβ* contact is also observed significantly more in α-helices than in β-sheets or when no secondary structure is present, and has a uniqueness of 0.71, so within a sequential Ala–Ala fragment this type of contact is highly predictive of α-helical structure. The sequential Hα–Hβ* contact is, on the other hand, highly predictive of β-sheet with a uniqueness of 0.20, although it is on average mostly observed when no secondary structure is present (0.38). This illustrates that the uniqueness of a contact is strongly related to the prevalence of a particular sequence combination in a particular secondary structure element, and is not necessarily indicative of the kind of secondary structure element a particular contact usually occurs in. The information from the constraints and the coordinates show clear differences, as illustrated by comparison with some of the contacts described earlier. The H–Hβ* contact occurs quite often in β-sheets based on the coordinate data, with a median distance that is slightly higher than in α-helices (4.63 Å compared to 4.31 Å). The Hα–Hβ* contact, which would generally be difficult to identify due to overlap in the aliphatic region of a NOESY spectrum, is always present based on the coordinate data, but is only seen in 0.30 cases based on the constraints. These differences illustrate that the NOE constraint data not only incorporates distance information, but also encodes NMR-specific information such as the difficulty with which a particular contact can be assigned. For i–i + 2 and i–i + 3 contacts, as expected, the typical H–H, Hα–H, Hα–Hβ* and Hβ*–H contacts are highly prevalent for α-helices (Table 3), with generally high uniqueness and low ambiguity. An H–Hβ* contact is also more often observed than average (respectively in 0.09 and 0.12 cases), while Hβ*–Hβ* contacts are very rarely seen for i–i + 2 contacts in an α-helix. Note that based on the coordinate data the i–i + 2 Hβ*–Hβ* contact is highly relevant for β-sheets (0.70), but it is in practice rarely observed (0.05), probably because it falls in a densely populated region of a typical NOESY spectrum. For 310 helices the percentages are often similar to the α-helical ones, but there are often not enough data to determine whether a difference is significant (data on web pages). Also of interest is the i–i + 3 contact between Hβ*–H, which occurs in 0.04 cases for β-sheets and 0.06 cases where secondary structure is absent. The binomial analysis indicates that the 0.06 fraction is observed significantly less, while it does not mark the 0.04 fraction. This is because this type of analysis is dependent on sample size (only 1 sample for the 0.04 fraction, 12 for the 0.06 fraction). For the i–i + 4 contacts, Hα–H and Hβ*–H connections are again highly prevalent for α-helices, while Hα–Hβ* and Hβ*–Hβ* contacts are present in the coordinate set but are seldom observed in practice (Table 3). All contacts with a separation of 5 residues or higher are very rarely observed in α-helices, but become highly prevalent for β-sheets (data not shown). The only exception to this are long range Hα–Hβ* and Hβ*–Hβ* contacts, which are seen in significantly higher percentages between α-helices as compared to the average. Contact data highlights Traditionally, and as described above, identifying secondary structure contacts is based on the commonly observed contacts between protons from the backbone and the β position (Wüthrich 1986). However, with experience in assigning NOE peaks comes the knowledge that other contacts are also often observed (e.g. i → i + 2 contacts between side chain protons in a β-sheet). In this analysis, such contacts are readily observed (Table 4). For example, the sequential Trp Hε3–Gly H contact is quite common based on the coordinate data, but is in practice particularly observed in a β-sheet. The sequential Trp Hε3–Phe Hα contact is seemingly more often observed in an α-helix, but the differences are not significant. This is a case that could be clarified if more relevant data were available. The Thr Hγ2*–Tyr Hε* i–i + 2 contact is clearly observed in mostly β-sheet. Interesting in this case is that according to the constraint data it can occur in an α-helix, while this is not the case based on the coordinate data, even though the ambiguity of the contact is 0.00. This is possible because this data point is based on one contact with an upper distance limit of 6 Å, whereas the cutoff used for coordinates is 5 Å. Generally speaking this type of situation can occur for highly ambiguous constraints, where other constraint items satisfy the upper distance limit. An Hα–Hδ1* i–i + 3 contact between Ala and Ile is observed almost exclusively in an α-helix (and never in a β-sheet), and is very unique (0.78). To be able to discern whether this contact is very common from the Hα of any amino acid to the Hδ1* of an Ile, the data was joined as Xxx residues to produce generic information to and from each amino acid (data on web pages). The specific information for the i–i + 3 Hα–Hδ1* contact from all residues to Ile is shown in Table 4 and shows that this contact is common in α-helices. Finally, some other interesting i–i + 4 side chain contacts that often occur in helices are listed in Table 4. Table 4foccurrence for selected secondary structure related contactsTypeAtom 1Atom 2α-helixβ-sheetNo secondary structurei → i + 1Trp Hε3Gly H0.05/0.450.56/0.880.24/0.63Trp Hε3Phe Hα0.55/0.680.11/0.160.13/0.47i → i + 2Thr Hγ2*Tyr Hε*0.02/0.000.46/0.800.12/0.41i → i + 3Ala HαIle Hδ10.66/0.930.00/0.000.11/0.21Xxx HαIle Hδ10.57/0.920.00/0.020.09/0.22i → i + 4Trp Hζ2Thr Hγ20.52/0.560.00/0.000.00/0.00Tyr Hε1 + Hε2Val Hγ20.47/0.740.00/0.050.06/0.21In the secondary structure columns, the first value is from the HP set, the second from the HPC set. A bold value indicates a contact that occurs significantly more than average, an italic value significantly less It is not possible to describe all information in detail in this article, and the web pages serve as the reference resource for any investigations. However, to provide a better overview of the overall trends in secondary structure elements, all backbone contacts were grouped by atom type (H, Hα, Hβ) and secondary structure combinations (Table 5, full data from web site). For intra-residue contacts, more contacts involving Hβ protons are defined in β sheets. Sequential contacts in α helices originating from the H and Hβ protons are observed significantly more often, whereas ones originating from the Hα proton are less common. This situation is reversed in β sheets (except for Hβ–H contacts). There are, only for sequential contacts, some discrepancies between the data from the HP and HPC sets, with, for example, the rate at which sequential H–Hα contacts are observed being reversed in the HPC set as compared to the HP set. The reasons for this are not immediately clear, but are likely related to overlap. Most i–i + 2 contacts are more common in α helices, except for ones to the Hα proton and between Hβ protons. The latter contact is more often observed in β sheets. As expected most i–i + 3 contacts are commonly observed in α helices, except for Hβ–Hα and H–Hα. The latter is again more frequently seen for β sheets, but in this case this is expected to be between different strands in the hairpin area. The trends are not as clear for i–i + 4 contacts, where Hα–H, Hα–Hβ and Hβ–H contacts are more frequent in α helices, and all other ones more frequent in β sheets. This is again likely related to hairpin contacts. All contacts from i → i + 5 and more are very infrequent for α helices but relatively very frequent for β sheets. Table 5Brief overview of general trends in joint secondary structure information for all contactsTypeSSH (i)Hα (i)Hβ (i)HHαHβHHαHβHHαHβi–iHelix. . + .. .− .− .+ .− .− .Sheet. .+ .. .. .+ .+ .+ .+ .i–i + 1Helix+ .+ −+ +− .− +−−+ .+ ++ +Sheet− .− +−−+ .+ ++ ++ .−−−−i–i + 2Helix+ +−−+ ++ +- -. ++ +−−−−Sheet−−. −−−−−. −. −−−+ .+ +i–i + 3Helix+ +. −+ ++ ++ ++ ++ +−−+ +Sheet−−+ +−−−−. −−−−−. −−−i–i + 4Helix. .−−−−++. .+ ++ +−−. +Sheet+++ ++ .−−+ −. −. −+ .+ −i–i + 4<Helix−−−−−−−−−−−−−−−−−−Sheet+ ++ ++ ++ ++ ++ ++ ++ ++ ++ Indicates that signals observed more than average, − less than average, . signifies that there is no trend. The first character in each cell contains the constraint HP set information, the second the coordinate HPC set An analysis of the percentage of atoms that were assigned within each residue type (Residue atom analysis on the website) shows that generally atoms are significantly higher assigned in α-helices and β-sheets, and lower when no secondary structure is present. This finding is not surprising as the secondary structure elements are defined by constraints, and the atoms have to be assigned to obtain those. Another general trend is that prochiral methylenes are within α-helices significantly more degenerated, i.e. the HB2 and HB3 atoms exists as a QB pseudoatom or an HB* type atom set. This type of analysis can be significantly improved by cleaning up the stereospecific assignment status based on the coordinates. The unique sequence fragments for which no assignments were found are also listed on the website (Unassigned fragments breakdown). An example from this data is that often no constraints are found for His tags. The LEHHHHH fragment, for example, was not assigned in 13 entries. A general overview of the percentage of residues that are assigned confirms that His residues, for example, are assigned in only 71.66% of cases, while Trp is assigned in 97.87% of cases (Fragment analysis by residue). To examine this in more detail, all tripeptide fragments where a particular amino acid is the central residue are listed (for the N- and C-terminus these are dipeptide fragments). Listed for each fragment are (1) the total number of times the tripeptide fragment occurs in the data set, (2) the percentage of times it is unassigned compared to the total number of times the amino acid occurs, (3) the number of times the amino acid is not assigned, (4) the number of times the tripeptide fragment occurs overall, and (5) the assignment percentages per secondary structure element. The entries in which the fragments are unassigned are also listed. Continuing with His as an example, it is not assigned when part of the EHH fragment in 69% of cases, and when part of a C-terminal HH fragment it is not assigned in 96% of all times the fragment occurs. To get a better overall view of the sequence fragments that are difficult to assign, they were grouped by joining, respectively, the i − 1 and i + 1 residues. The results for His and some selected other fragments are shown in Table 6. A Ser residue preceded or followed by a Gly, for example, is often unassigned. This type of information could be useful for predicting which areas of a protein sequence are difficult to assign from an NMR perspective. Table 6Selected sequence fragments where the central residue is often unassigned. The unassigned percentages are relative to the total number of times the fragment occursFragmentUnassigned (%)TotalXxx–His–His62756Xxx–His–Met583Gly–His–Xxx234His–His–Xxx63740Ser–His–Xxx8114Xxx–Ser–Gly18365Gly–Ser–Xxx19479Xxx–Pro–Ser8157 Discussion In this analysis only the original data as deposited by the authors was used, and no attempt was made to ‘clean up’ and further interpret this information, except for linking the constraint with the coordinate data and removing identical sequences from the data set pool (where only the entry with the highest number of constraints linked to atoms was kept). This approach is intentional, as it best represents the quality and extent of the data that is currently deposited at the PDB. Only the distance constraint information was included in the analysis, and the information from dihedral, H-bond and RDC constraints was ignored. Even though these constraints contain important structural information, they were, as experimental data, recorded independently from the NOE data. They are used in the structure determination process, however, and it was not investigated whether their presence influences the quality of the final distance constraint lists. There are several other issues that can still be addressed, and although these are likely to improve some of the aspects of this type of study, it is also important to start with the original information so that a comparison point is available. The first issue is that stereospecific assignments can be swapped or deassigned based on the original coordinates, similar to the approach in the RECOORD project. This could in principle reveal preferences related to stereospecifically assigned atoms in secondary structure elements. The second issue concerns the distances that were provided with the constraints. These are often ‘binned’ in weak/medium/strong classes with fixed distance cutoffs, so that the resulting distance distributions often show spikes at these distances. In Fig. 4, for example, it is clear that spikes occur at 3.0, 3.5 and 5.0 Å. Recalibrating the distances based on the deposited coordinates should improve the quality of the resulting information, and reveal relationships between distance and occurrence. The third issue concerns the sequences that are included: the current data sets include protein sequences with a high homology. This is clearly not an ideal situation, but there is currently not enough data available for a cleaner analysis. It is therefore important to check whether a particular contact appears in a large amount of entries, or the observed occurrence might be due to systematic error from homologous proteins produced by the same laboratory. The fourth problem is the identification of the secondary structure fragments. This is now based on the PDB DSSP analysis from the original coordinates for only the first or representative model so that secondary structure elements are not always identified properly. If chemical shifts were available an identification based on CSI (Wishart et al. 1992) would become possible so secondary structure stretches can be included that are more flexible and less defined on the coordinate level. More refined secondary structure identifications from the coordinates could also reveal patterns related to, for example, turns. Overall, the best way to improve this analysis remains to increase the sample size by encouraging deposition of constraint lists and all related NMR information (peak lists, chemical shifts, spectra), and ensuring that the data is consistent when deposited by the authors. Efforts like the CCPN project (Fogh et al. 2002, 2005), which allow data harvesting from NMR data collection to structure calculation, should provide this kind of data without requiring any additional effort by the scientists who produce the data. Fig. 4Distance distribution from the constraint information for sequential Ala–Ala contacts between backbone H protons In this analysis a particular inter-atomic contact between two residues from one PDB entry is either observed or not observed. The reason why a contact is observed (or not) implicitly includes distance information, peak overlap, water exchange line broadening, and all other factors that can lead to not observing or assigning a contact during analysis of a spectrum. This is different from the traditional meaning of an ‘assigned atom’ on the chemical shift level, where it means that the chemical shift value for the resonance that arises from the atom is known. However, this does not necessarily mean that these assigned atoms produce any valid inter-atomic distance information. Thus, an ‘assigned atom’ (or residue) on the constraint level means that a chemical shift assignment also produced useful and valid information related to the inter-atomic distances within the molecule. The original study that defined the inter-atomic contacts relevant for assigning secondary structure elements with NMR used a set of 19 high-resolution protein crystal structures comprising about 3,200 residues (Wüthrich et al. 1984). In this study the extent of identification was defined as the percentage of times a distance was smaller than a particular cutoff value within a secondary structure element, while the uniqueness of identification is the percentage of times the distance is observed within a particular secondary structure element out of the total number of times it is observed. The extent is equivalent to the foccurence used in this study for the constraint sets, with the exception that no distance cutoff is used (although some distance information is, as mentioned, implicitly included because the reason a constraint is observed or not is very dependent on distance), and that values are labelled as significant based on a binomial analysis. The uniqueness has the same definition, except that again it is based on the amount of constraints that are observed. Also of interest is the relationship between the information that comes directly from the deposited constraints and the information that comes from the deposited coordinates. Here, the constraint information is compared to the distances from the originally deposited coordinates. Although a set of recalculated coordinates (as in RECOORD) or X-ray structures could have equally well been used, the originally deposited coordinates were chosen because they should best match the content of the constraint lists. All comparisons between constraint and coordinate information are intended for informative purposes only: the constraints represent the experimental NMR side of the information contained in the coordinates, and are in effect only a subset of the information contained therein. However, a dependable determination of whether a particular NOE contact is observed or not is not possible based on an NMR structural ensemble, but is trivial based on the constraints because they inherently contain NMR-specific information like signal overlap, dynamics, etc. From Fig. 5 it is clear that the contact occurrence is almost always higher for the HPC set compared to the HP set. This is related to the use of a direct distance cutoff of 5.0 Å in the HPC set: contacts with long distances could give rise to peaks that are too weak to be seen in a real spectrum but are still included. Also, many contacts have an foccurrence of 1.0 in the HPC set because of conformational constraints from covalent bonds. Not all of these contacts are seen in real spectra because of, for example, peak overlap or line broadening. The correlations between the occurrences overall are not very high (Spearman 0.770, Pearson 0.694), with especially the intra-residue contacts giving bad correlations (Spearman 0.451, Pearson 0.356, see web site), and i–i + 2 (Spearman 0.720, Pearson 0.783) and i + 3 (Spearman 0.711, Pearson 0.659) contacts giving the best results. An indication that the main reason for the bad correlation between the occurrences is distance related comes from the large improvement that is observed in the overall correlations if only coordinate distances of less than 3.6 Å are considered (Spearman 0.906, Pearson 0.916). However, results from using both the HP and HPC information to filter ambiguous constraints lists show that both sets essentially give the same results (personal data), even though the constraints are available in a much more ‘compressed’ form than the coordinates, and no force field information was used. Fig. 5Correlation between the foccurrence from the HPC (Coordinates) and HP (Constraint) sets In the KNOWNOE (Gronwald et al. 2002) X-ray structure based approach to obtain probabilities for assignments, the distance distributions for inter-atomic contacts are used to generate volume-based probabilities in addition to the atom identity based probabilities. This approach improves the probabilities that are generated, but it does require that the original peak list with volumes is available. This is not possible within the current analysis, although this will be pursued if a meaningful way to recalibrate the distance constraint bounds is available. This would also allow a better comparison between the NMR constraint data and any coordinate data (from NMR or X-ray structures). Conclusion A resource is now available where it is possible to check how likely a particular contact is when assigning NOESY spectra, or if a particular sequence fragment is likely to be difficult to assign. In this respect it formalises information that scientists with experience in spectrum analysis are aware of but cannot quantify. The amount of information provided here is extensive, however, and is even more useful when used as ‘knowledge based’ probabilities in automatic assignment strategies, to filter and/or validate ambiguous constraint possibilities, and as a tool to rank assignment possibilities in spectrum analysis programs. These are being implemented as part of the CCPN framework. Finally, the NMR constraint lists encompass the experimental NMR data encoded in the NMR structural ensembles, and comprise a single set of data that is much easier to analyse than an ensemble of solutions. As such, they provide a reduced form of structural information that is relevant for NMR analysis only. For this reason, and to allow a basic level of scientific reproducibility and validation, it is important that constraints, and all other possible NMR derived information, are deposited along with the structure coordinates. It is very likely that a lot more information than described in this article can be gained from it, which in turn can assist the NMR community and can help to understand the relationships between NMR and structure.

Anal_Bioanal_Chem-4-1-2259236

High-precision frequency measurements: indispensable tools at the core of the molecular-level analysis of complex systems

This perspective article provides an assessment of the state-of-the-art in the molecular-resolution analysis of complex organic materials. These materials can be divided into biomolecules in complex mixtures (which are amenable to successful separation into unambiguously defined molecular fractions) and complex nonrepetitive materials (which cannot be purified in the conventional sense because they are even more intricate). Molecular-level analyses of these complex systems critically depend on the integrated use of high-performance separation, high-resolution organic structural spectroscopy and mathematical data treatment. At present, only high-precision frequency-derived data exhibit sufficient resolution to overcome the otherwise common and detrimental effects of intrinsic averaging, which deteriorate spectral resolution to the degree of bulk-level rather than molecular-resolution analysis. High-precision frequency measurements are integral to the two most influential organic structural spectroscopic methods for the investigation of complex materials—NMR spectroscopy (which provides unsurpassed detail on close-range molecular order) and FTICR mass spectrometry (which provides unrivalled resolution)—and they can be translated into isotope-specific molecular-resolution data of unprecedented significance and richness. The quality of this standalone de novo molecular-level resolution data is of unparalleled mechanistic relevance and is sufficient to fundamentally advance our understanding of the structures and functions of complex biomolecular mixtures and nonrepetitive complex materials, such as natural organic matter (NOM), aerosols, and soil, plant and microbial extracts, all of which are currently poorly amenable to meaningful target analysis. The discrete analytical volumetric pixel space that is presently available to describe complex systems (defined by NMR, FT mass spectrometry and separation technologies) is in the range of 108–14 voxels, and is therefore capable of providing the necessary detail for a meaningful molecular-level analysis of very complex mixtures. Nonrepetitive complex materials exhibit mass spectral signatures in which the signal intensity often follows the number of chemically feasible isomers. This suggests that even the most strongly resolved FTICR mass spectra of complex materials represent simplified (e.g. isomer-filtered) projections of structural space. Introduction Status of the molecular-level analysis of complex materials Natural molecules can be classified into two main groups according to their functions and how they are synthesized (Fig. 1). In living organisms, a genetic code initiates and controls the synthesis of functional, discrete molecules, which range in size from multienzyme complexes and molecular machines through mid-sized natural products to small molecules (e.g. NO). These molecules are often assigned to well-established distinct classes, such as proteins/peptides, carbohydrates, lipids, and (less structurally related), natural products and metabolites. Groups of these molecules frequently act together in regulatory networks [1–4] so as to enable critical life functions. Typical biofluids and tissues are very complex mixtures which can be resolved into defined molecular fractions using current high-performance separation technologies. Fig. 1Natural complex organic materials divide into either functional biomolecules which eventually derive from a genetic code or complex biogeochemical nonrepetitive materials which are formed according to the general constraints of thermodynamics and kinetics from geochemical or ultimately biogenic molecules. While biomaterials are amenable to successful separation into unambiguously defined molecular fractions, complex nonrepetitive materials cannot be purified in the conventional meaning of purity due to their extreme intricacy; in fact, the molecular signatures of these supermixtures often approach the limitations imposed by the laws of chemical binding. Improvements in the resolution and sensitivity of analytical techniques combined with the use of minimal (non)invasive sampling techniques have enabled environmental and living systems to be observed to a degree of molecular resolution that was considered unthinkable only a few years ago. A molecular-level understanding of biogeochemical and life processes implies a key role for de novo structural analysis, which depends on the combined use of separation technology hyphenated to organic structural spectroscopy and integrated mathematical data analysis. Analyses of supermixtures depend even more on the mathematical analysis of correlated data obtained from complementary molecular-level precision analytical methods. The formation of NOM on the Earth preceded the evolution of life; the binding of NOM-derived prebiotic molecules to borate contributed to the synthesis of ribose, a crucial precursor of nucleotides, in good yield [5]. Later on in the Earth’s history, coevolution occurred between prebiotic/abiotic molecules, NOM and primitive and higher forms of life. The near-continuum of binding sites available to ions and organic molecules acts to buffer against environmental and chemical extremes in the geo- and biosphere, which could damage life because of their potent reactivity. This key supportive role of (for example) natural organic matter (NOM) in life processes is sustained by strong interactions between biological and geochemical cycles (Fig. 2). Hence, plant and animal residues are key ingredients of NOM synthesis, while NOM itself, which defines the bioavailability of crucial organic and inorganic nutrients, is indispensable for the sustenance of the microbial life at the bottom of the food chainFig. 2Natural organic matter (NOM) continuously interacts with a broad range of terrestrial, limnic and marine ecosystems. Common to all of these environments are the fundamental molecular aspects of life, and an availability of extended mineral surfaces for interactions with and binding of NOM. The dynamic equilibrium of NOM generation and decomposition spans timescales of many different orders of magnitude (from microseconds to hundreds of thousands of years), and it results from a combined action of biotic and abiotic reactions. NOM may be intrinsically recalcitrant because of the chemical structures of its organic molecules; alternatively, strong NOM–mineral interactions could alter the reactivity of these organic molecules towards increased resistance to degradation. The physical protection of organic matter at interior mineral surfaces provides alternative pathways that enable the recalcitrance of NOM. Photochemical degradation, one of the most significant abiotic reactions of NOM, often results in small molecules like CO2, which are mobile and are easily distributed within various ecosystems. Biomolecules derived from photosynthesis or otherwise originating from a genetic code are eventually decomposed according to the general laws and constraints of thermodynamics and kinetics (Fig. 1). Over very long timescales, the interactions of NOM with minerals at elevated temperatures result in the formation of geopolymers, like kerogen, coal, and oil shales. These ancient materials participate in bio- and geochemical cycling through natural and anthropogenic combustion and through weathering [6]Fig. 3Hierarchical order of intricacy associated with the structural analysis of materials, in terms of polydispersity and molecular heterogeneity (see the main text). The structures (connectivities and stereochemistry) of monodisperse molecules are readily accessible (provided that sufficient amounts of the material are available) by organic structural spectroscopy [7–9]. Supramolecular structures [10–12] require an adequate definition of the covalently bonded molecules and of their noncovalent interactions [13–16]. The structural analysis of nonrepetitive complex unknowns, which feature substantial levels of both polydispersity and molecular heterogeneity [17, 18], is most demanding in terms of methodology and concepts [19–22]. Any highly resolved three-dimensional structure of a monodisperse biomolecule is based on a precise description of the unique chemical environment of any single atom [23, 24]. Currently, the molecular-level structural analysis of complex systems is primarily focused on covalent bond definition. Future high-quality structural analyses of these materials will have to assess both the (classes of) individual molecules and their interaction mechanisms [25] The fate of the second vast group of molecules in the bio- and geosphere is governed by the rather fundamental restraints of thermodynamics and kinetics (Figs. 1 and 2). In these intricate materials, the “classical” signatures of the (geogenic or ultimately biogenic) precursor molecules, like lipids, glycans and proteins, have been attenuated [26, 27], often beyond recognition, during a succession of biotic and abiotic (e.g. photo- and redox chemistry) reactions. Because of this loss of a biochemical signature, these materials can be designated nonrepetitive complex systems. The quantity of molecules in the Earth’s crust that can be attributed to these nonrepetitive complex materials, in the form of kerogens and natural organic matter (NOM) alone, exceeds the quantity of functional biomolecules by several orders of magnitude [28, 29]. Examples include freshwater, marine, and soil organic matter, kerogens and aerosols, among others. These materials typically exhibit an extremely complex array of chemical structures and interactions across a large range of size- and timescales, resulting in molecular signatures that reflect the fundamentals of chemical binding rather than those of their precursors. These novel signatures may in fact cover a sizable proportion of the theoretically feasible molecular composition space (Fig. 9). This extraordinary heterogeneity of molecularly diverse species renders these materials refractory and also implies a limited probability of detecting identical molecules [18]. This contrasts sharply with even the most complex mixtures of biomolecules extracted from any living organism, from which molecularly pure fractions can be readily obtained. Given these unique features, nonrepetitive complex systems epitomize supermixtures. The purification of a supermixture would, in the ultimate sense, approach a molecule-by-molecule separation—a feat beyond our reach, both conceptually and practically. Therefore, these complex nonrepetitive systems are operationally defined according to their properties rather than according to their chemical structures, and their purification (in the conventional sense of the word) remains elusive [18]. While the analysis of complex biomolecules has advanced to the degree that it is possible to obtain well-resolved three-dimensional molecular structures and even meaningful descriptions of dynamics and interactions [30–35], the molecular-level precision analysis of complex nonrepetitive materials remains rather rudimentary in comparison [20, 36–40]. First of all, theoretically well-founded approaches to numerically describe the complex, polydisperse and nonstoichiometric characteristics of nonrepetitive unknowns are missing at present, limiting our understanding of molecular structures and any application of quantitative structure–activity relationships (QSAR) when modelling their properties. Novel approaches suitable for a quantitative description of various hierarchical levels of molecular organisation (e.g. elements, fragments, molecules) must be developed. Secondly, a meaningful molecular-level analysis of nonrepetitive systems—such as aerosols, natural organic matter and native cell extracts—obviously cannot rely on target analysis, as most of the chemical environments and linkages present are simply not known (Fig. 3). Consequently, any comparative analysis of nonrepetitive unknowns with reference materials is very unlikely to provide satisfactory molecular resolution, because rather tiny variations in chemical binding may strongly and often unpredictably affect the properties commonly used for detection, such as retention times and spectral signatures. These fundamental restrictions that are intrinsic to comparative and target analysis are not easily circumvented and they necessitate an independent, spectroscopic “bottom-up” approach to the molecular-resolution characterisation of these complex unknowns. Information transfer in organic structural spectroscopy and separation technologies Interestingly, this “bottom-up” approach to the molecular characterisation of complex systems and materials necessarily relies upon spectroscopic methods that translate high-precision frequency measurements into important molecular-level information. Frequencies can be measured to an accuracy of 15 digits, and recently (2005) a Nobel Prize in Physics was awarded for improvements of laser-based precision spectroscopy that enable even more exact frequency measurements to be obtained [41]. This high accuracy of frequency measurement translates directly into high resolution, itself a very useful and even indispensable feature when producing information-rich data. Well-resolved signatures are less susceptible to the detrimental consequences of intrinsic averaging, which is an often overlooked key feature of any low-resolution (spectroscopic and separation) method, when they are employed in the characterisation of complex systems (Fig. 4). Fig. 4Information transfer in organic structural spectroscopy. These images define a three-dimensional space composed of the area of the image (pixel resolution along the x- and y-axes) and the depth of the color space. This three-dimensional space offers n options for depicting dissimilarity (n = x × y × color depth). Significant resolution is only attained if there is sufficient information to enable a meaningful assessment of data [here, the apparent differentiation of (a) a fish, (b) a mountain, (c) a human, and (d) a beetle is only possible at panels C and D]. In analogy with these considerations, a three-dimensional analytical volumetric pixel space comprising NMR spectroscopy, mass spectrometry and separation is developed in Fig. 12. The current expansion of this analytical volumetric pixel space (log ) is sufficient to elaborate meaningful detail at molecular resolution from the most complex biological and biogeochemical mixtures Bulk data of complex systems, like physical parameters, total acidity and elemental analyses, seem to be more precisely defined [42], but exhibit limited resolution. However, any sound structural model of these materials must conform to the constraints defined by these “hard” bulk data. High-energy methods of organic structural organic spectroscopy, like XANES, UV/VIS and infrared spectroscopy, exhibit intermediate structural resolution, which is sufficient, for example, for the characterisation of specific chemical environments [43]; for instance, functional group analysis (carbonyl derivatives, aromatics, heterocycles) in intricate materials. In general, the degree of significant detail generated by a certain analytical technique will depend on both the intrinsic resolution of the respective method and the characteristics of the analysed material. Any inadequate relationship between the resolving power of the technique and intrinsic analyte properties will be wasteful. Investigations of near-featureless materials with methods of supreme resolution could result in unnecessary effort and expenditure. Insufficient resolution of any analytical method with respect to the properties of the analyte will inevitably result in intrinsic averaging, which typically results in poorly resolved properties (which affect the separation) and/or poorly resolved chemical environments (which affect the spectra). Intrinsic averaging is visualized in Fig. 4 in the form of images of ever-degrading resolution. Similarly, insufficient resolution deteriorates detail in spectra and chromatograms of complex nonrepetitive materials, producing low-resolution signatures and limited bandwidths of variance in bulk and spectral properties. Hence, any organic structural spectroscopy with a limited peak capacity (Fig. 5) will inevitably lead to a summary bulk-type description of complex materials and considerable averaging, rather than to a meaningful molecular-level resolution analysis. In the case of NOM, this inevitable relationship has been observed in many spectroscopic, separation and chemical experiments, resulting in data with a remarkably limited bandwidth of variance, even when advanced techniques (e.g. at the level of one-dimensional solid-state 13C NMR spectroscopy) are used [44, 45]. Analogously, the widespread use of the idiom HULIS (or humic-like substances) in the fields of, for instance, aerosol and remediation research [46–53] reflects the operational definition of humic materials as well as our current inability to perform a meaningful molecular-level analysis of complex unknowns, as materials currently denoted HULIS or humic-like substances undoubtedly encompass a wide range of very different species. Due to the huge peak capacity of FTICR mass spectrometry, FT mass spectra provide the most convincing direct experimental evidence for the extraordinary molecular diversity of complex materials at present. In these, the molecular-level intricacy of the complex unknowns is most adequately converted into very highly resolved and, consequently, extremely information-rich signatures. Analogous considerations to those given here for spectroscopic characterisation also apply to the separation of complex materials [54, 55, 56]. Instrumentation and methods The 13C NMR spectrum of Suwannee river fulvic acid (SuwFA) shown in Fig. 7 was acquired at the GSF with a Bruker (Bremen, Germany) AC 400 NMR spectrometer, operating at 100 MHz for 13C. FTICR mass spectra were acquired at Bruker’s facilities with a 9.4-T APEXq FT mass spectrometer (data in Fig. 9) and a 12-T APEXq FT mass spectrometer at the GSF (Fig. 8). Here, FTMS spectra were acquired with a time domain size of 1 MWord (Fig. 8a; Fig. 9, typical resolution 3 × 105) or 4 MWord (Fig. 8b, typical resolution 7 × 105). For Figs. 8a and 8b, elemental compositions were computed with the DataAnalysis software, version 3.4 (Bruker), using the following restrictions: C, H, N, O, unlimited; S, P, 0–5; H/C ratio < 3, mass error ≤ 0.5 ppm; observance of the nitrogen rule. Exactly one elemental formula was obtained for each peak. The elemental formulae of Fig. 9 were batch-calculated using a software tool written in-house, as described elsewhere [36]. Fig. 5 Characteristic resolutions (peak capacity: total range / half width [57]) of various separation technologies and organic structural spectroscopic methods (see the main text). This diagram represents a two-dimensional projection of the analytical volumetric pixel space—comprising NMR spectroscopy, mass spectrometry and separation technologies—that defines our current capacity to depict variance in complex systems with molecular resolution (see Fig. 12) The numbers of isomers, as displayed in Fig. 11, were calculated by constructive enumeration using the software MOLGEN (Department of Mathematics, University of Bayreuth, Bayreuth, Germany). For this computation [58], certain restrictions were applied in order to exclude structures that are mathematically possible but are not likely to occur in materials of biogeochemical origin; this means that there are (1) no peroxides; i.e. no –O–O– connectivities; (2) no triple bonds (–C≡C–); (3) no three- or four-membered rings; and (4) no carbon with cumulative double bonds (=C=). The 1H and 13C NMR spectra of cholesterolacetate (section of 1H NMR spectrum in Fig. 6, top panels) and 2-carboxypyrene (Fig. 6, middle panels) were computed with ACD software (ACD/HNMR and ACD/CNMR Predictor, v. 5.0); the mass spectral isotope peaks of Ciguatoxin C60H86O19 (Fig. 6, middle panels) were computed with Bruker Compass FTMS software. Fig. 6The two most significant methods of organic structural spectroscopy, nuclear magnetic resonance (NMR) and FTICR mass spectrometry, are based on high-precision frequency measurements. The top panels illustrate NMR and FTICR MS for atomic and molecular processes. The precessions of atomic magnetic moments in molecules are defined by the chemical environment, and this means that NMR yields unsurpassed resolution of short-range molecular order (in noncrystalline materials, for which X-ray crystallography is not available). In FT mass spectrometry, the orbital frequencies of ions depend on their molecular masses. Mid-size molecules (see middle panels) provide information-rich signatures in NMR and only single peaks in mass spectra (under conditions of non-fragmentation), while complex non-repetitive materials (see bottom panels) produce low-resolution signatures in NMR because of extensive peak overlap. High FT mass spectra resolution is retained for complex molecules, however, because of the extensive peak capacity of the technique (see Fig. 5) High-precision frequency-derived organic structural spectroscopy The two most influential organic structural spectroscopic methods for the investigation of complex materials, which depend upon high-precision frequency measurements, are NMR spectroscopy and FTICR mass spectrometry (Table 1, Figs. 6, 7). In NMR, the precession frequencies of individual atomic nuclei in an external magnetic field B0 are influenced by their respective chemical surroundings; in FTICR mass spectrometry, the orbital frequencies of ions in an ion trap cell depend on the mass and charge of the molecule of interest [59]. Both methods are isotope-specific, and the combination of NMR and FTICR mass spectral data provides more useful spectral information on complex unknowns at the molecular level than any other spectroscopic method at present. When studying typical organic molecules, NMR spectra provide more information-rich data than mass spectra, because any single atom within can produce an individual NMR signature (Figs. 6 and 7) [7, 60–62]. In the absence of fragmentation, the mass spectrum of any particular molecule will contain only a single peak (in conjunction with its corresponding isotopic pattern [63, 64]). NMR spectra of ever more complex materials will eventually become near featureless because of the extensive overlap between individual NMR resonances [65]; however, the significance of the information is maintained because of the quantification reliability and insightful relationships between NMR chemical shift and extended substructures (at least for the “small” NMR-active nuclei, e.g. 1H, 13C, 15N, 31P, which represent the key players in organic structural spectroscopy) [66]. Mass spectrometry retains its supreme resolution for extremely complex systems [63, 67–70], but will eventually become limited by the inability of mass spectrometry to (easily) discriminate between isomers [71]. Hence, mass spectra will at best represent isomer-filtered projections of the entire structural space of molecules (see Figs. 7, 8, 9, 10, 11). Fig. 7Molecular-level resolution spectroscopic data represent projections of the vast total structural space of molecules, for which count estimates range from 1060 to 10200 [72]. The complementarity of NMR spectroscopy and mass spectrometry for the spectral characterisation of intricate materials is caused by the entirely different atomic and molecular processes that these methods rely upon (Fig. 6). Mass spectra reflect the isomer-filtered complement of the entire space of molecular structures. The compositional space of molecules can be probed with ultrahigh-resolution FTICR mass spectroscopy, resulting in single peaks for molecules (in the absence of fragmentation). NMR spectra represent site- and isotope-specific projections of the molecular environments (the projected NMR spectrum given here shows a 13C NMR spectrum for an aquatic NOM). Accordingly, typical organic molecules exhibit single mass peaks (molecular ions given here: C9H9NO3) and more elaborate NMR signatures (13C NMR data are shown here for C9H9NO3; see also Fig. 6). Because these atomic and molecular signatures are not entirely orthogonal, the data provided by NMR and MS show correlations that can be used to reconstruct chemical structures by empirical and mathematical back-projection. In mixture analysis, separation offers a (near-)orthogonal means of expanding the two-dimensional spectroscopic projection area into a three-dimensional analytical volumetric pixel space (see Fig. 12). Following separation, molecules that often exhibit wide concentration variances in mixtures can be investigated by combined NMR/MS, with acquisition parameters specifically adapted in order to maintain the optimum dynamic range, which is necessary for good-quality spectraFig. 8FT mass spectral resolution and the C,H,O-compositional space (see the main text and Fig. 9)Fig. 9Van Krevelen diagram (which illustrates the C,H,O-compositional space) for consolidated ESI, APPI, APCI positive and negative ions of Suwannee river fulvic acid (SuwFA) in the mass range 200–700 Da (see the main text and Fig. 8)Fig. 10Elucidating mass spectra of complex materials requires advanced means of data analysis, such as van Krevelen diagrams, which are based upon assigned molecular formulae, and fragment- or molecule-specific Kendrick mass defect analyses [73]. Any dot in the van Krevelen diagram of a complex material represents a projection of the elemental ratios derived from assigned molecular formulae, irrespective of molecular mass. Accordingly, any of these dots could represent an intrinsic superposition of all feasible isomers from possibly different molecular compositions, sharing only their respective elemental ratios. The numbers of chemically reasonable isomers easily account for many millions of isomers seen for moderately sized molecules (a few hundred Daltons), even when only a few double bond equivalents and heteroatoms (e.g. oxygen) are present (see Fig. 11)Fig. 11The relationship between the FTICR mass spectral intensity and the number of feasible isomers in complex materials. The distribution of (chemically relevant) C,H,O-isomer counts within the C,H,O-compositional space is visualized here in a van Krevelen diagram (left) for the eleven feasible molecular compositions CnHmOq that have an IUPAC nominal mass of 178 Da (see the main text). These molecules are arranged into three series (series 1a–1c with three members; series 2a–2f with six members; series 3a–3b with two members) of isobaric molecules, which are related by a formal exchange of CH4 against oxygen. For any molecular composition, the number of chemically relevant isomers is given in units of 104. Carboxyl-rich alicyclic molecules (CRAM), which represent a complex mixture of molecules with near-absent olefinic and aromatic unsaturation, have recently been identified in marine organic matter [18], and they occupy an area of C,H,O-compositional space for which the largest number of feasible C,H,O-isomers is expected (see the main text), suggesting that the actual number of different molecules in the mass spectra of CRAM for any given mass may correlate with mass spectral intensity patterns. The availability of aromatic moieties in Suwannee river fulvic acid (SuwFA) also allows chemically relevant structures at lower H/C ratios (see the main text). The green circumfenced area denotes the van Krevelen compositional space of SuwFA, as provided in Fig. 9 (note that the SuwFA here represents consolidated positive and negative ion APCI+APPI+ESI FTICR mass spectra, as opposed to the negative ion ESI FTICR mass spectra given in the case of CRAM; see the main text). The panel on the right shows the CnHmOq isomers computed for an IUPAC nominal mass of 178 Da in an intensity versus mass display (analogous to a mass spectrum), denoted according to series 1–3 (see the main text). This pattern resembles the clusters of peaks observed in good-quality FT mass spectra of NOM (see Fig. 6, bottom panels, and Fig. 8), suggesting that the intensities of the C,H,O-derived mass spectral peaks of NOM follow the number of chemically relevant C,H,O-isomers computed Table 1 Characteristics and significance of key molecular-level resolution techniques Molecular-level resolution technique Advantages Current weaknesses and future developments NMR spectroscopy NMR spectroscopy provides isotope-specific information, in unsurpassed detail, on short-range molecular order (the arrangement of chemical bonds, including connectivities, stereochemistry and spatial proximity), dynamics [33, 74] and reactivity [35] Relative insensitivity compared with other analytical techniques Nondestructive and isotope-specific [32] analysis across almost the entire periodic table combines with the most accurately defined near-quantitative relationship between the spin number and the area of the NMR signal. This key feature of NMR when applied to the analysis of complex systems implies the use of NMR spectrometry as a quantitative reference for other, complementary analytical methods [75–81] Intricate physics and chemistry of intra- and intermolecular interactions in complex mixtures may interfere with the direct relationship between chemical shift and molecular structure and, because of relaxation-induced variable line widths, quantification The unique ability to generate and analyse data from multiple 1,2,3-D NMR experiments performed on a single sample enables the significance and authenticity of individual spectra to be assessed [82–84] Near-identical chemical shifts do not necessarily imply similar chemical structures [85] Extensive and far-reaching information can be obtained, even from ill-resolved NMR spectra, for “small” nuclei (e.g. 1H, 13C, 15N, 31P) because of the plausible correspondence between chemical shift and extended substructures Sensitivity and resolution increased by high-field magnets [86, 87], cryogenic [88] and micro- [89–92] probes, and by changing (to nanoliter) sample size [93, 94] Throughput increased by using fast higher dimensional spectroscopy with superior sensitivity and through the parallel acquisition of NMR spectra [95–99] FTICR mass spectrometry Best combination of spectral resolution and sensitivity, which allows miniaturisation [100] and hyphenation of mass spectrometry with high-performance separation techniques like capillary electrophoresis and UPLC (CE/UPLC with mass-selective detection) Molecular-level structural information is mainly restricted to ionizable compounds Due to its supreme mass accuracy and resolution, molecular formulae from thousands of compounds can be obtained in a single experiment directly from mixtures [63, 69, 101–104] Isomer differentiation is a nontrivial task [71] Fragmentation provides further molecular-level structural information beyond molecular composition Quantification is difficult, even for identical molecules in mixtures, because of the variable ionization efficiencies of individual compounds, which strongly depend on the experimental conditions and mixture composition Column adsorption and fractionation as well as electrochemical and redox reactions associated with the spray conditions may interfere with authentic sample representation A wide range of ionization techniques (electrospray, ESI; chemical ionization, CI; photoionization, PI; desorption ionization, DESI; field ionization, FI, among others, all performed in either positive or/and negative modes) for mixtures is available under specifically adapted conditions [105–110] Mass-selective imaging is feasible with high spatial and mass resolution; qTOF mass spectrometry allows for very fast scan rates, and is perfectly suited for hyphenation with high-performance separation techniques (CE and UPLC) as well as mass-selective imaging Further miniaturisation of separation and detection devices in conjunction with ultrahigh-resolution FTICR mass spectrometry will permit highly resolved and information-rich data to be obtained from tiny amounts of sample (chip-MS) [100] High-performance separation techniques (UPLC/HPLC and capillary electrophoresis) Large separation capacity and extensive miniaturisation; is cost-effective; can be highly automated Gives only limited structure-specific information about the short-range molecular order [111, 112] Electrophoretic mobility and chromatographic retention time carry structure-specific information, which can be adapted to a wide range of experimental conditions in order to probe size, shape, charge characteristics and reactivity Sensitive and versatile suite of separation methods and of structure-specific (and nondestructive) detection systems, such as (laser-induced) fluorescence, UV/VIS, radioisotope or mass-selective detection CE complements NMR information about primary chemical structures (covalent bonds) by providing data on the corresponding secondary and tertiary structure Feasibility of up-scaling from capillary zone electrophoresis (CZE) to a preparative level by means of free flow electrophoresis (FFE) and from UPLC to any preparative LC method Further miniaturisation offers hyphenation options down to single-cell analysis and compartments within Molecular-level resolution spectroscopic data represent projections of the vast total structural space of molecules, for which count estimates range from 1060 to 10200 [72]. The complementarity of NMR and mass spectrometry for the spectral characterisation of intricate materials is caused by the entirely different atomic and molecular processes these methods rely upon (Fig. 6). Mass spectra reflect the isomer-filtered complement of the entire space of molecular structures. The compositional space of molecules can be probed with ultrahigh-resolution FTICR mass spectroscopy, resulting in single peaks for molecules (in the absence of fragmentation). Two-dimensional projections of the structural space, like van Krevelen diagrams and Kendrick mass defect analyses, are indispensable tools for the evaluation of mass spectra of complex materials (Figs. 8, 9, 10, 11) [73, 113, 114]. NMR spectra represent site- and isotope-specific projections of the molecular environments. Therefore, typical organic molecules exhibit single mass peaks (molecule ions) in mass spectra and more elaborate NMR signatures (Figs. 6 and 7). Because these atomic and molecular signatures are not entirely orthogonal, the data provided by NMR and MS exhibit correlations that can be used to reconstruct chemical structures by empirical and mathematical back-projection. The NMR and mass spectral data can be acquired via direct hyphenation or in separate experiments [115–117]. In mixture analysis, separation offers a (near-)orthogonal means of expanding the two-dimensional spectroscopic projection area into a three-dimensional analytical volumetric pixel space (see Fig. 12). Following separation, molecules that often exhibit wide concentration variances in mixtures can be investigated by combined NMR/MS, with acquisition parameters specifically adapted to maintain the optimum dynamic range necessary for good quality spectra. Particularly in the case of mixture analysis, any joint mathematical evaluation of these correlated data will reveal hidden detail and will considerably enhance the resolution as well as the significance of molecular-level information (see also Figs. 4 and 12). Fig. 12The currently accessible discrete volumetric pixel (voxel) space for the characterisation of complex materials is in the range of 108–14 voxels. Its expansion is defined by the significant resolution of the complementary techniques of nuclear magnetic resonance (102–5 buckets, depicting the short-range order of molecules), ultrahigh-resolution FTICR mass spectrometry (104–5 buckets, depicting molecular masses and formulae of gas-phase ions) and high-performance separation (102–4 buckets, depicting both ions and molecules; this provides a way to validate NMR against MS data); see also Fig. 5 for even wider expansion. The various projections of this voxel space, like separation/MS, separation/NMR and NMR/MS, can be realised in the form of direct hyphenation [104, 115, 117–120] and via mathematical analysis [121, 122] FTICR mass spectra show supreme resolution, as indicated by the 12-T negative ionization ESI FT mass spectra of a barley extract (Fig. 8a) and IHSS Suwannee River Natural Organic Matter (International Humic Substances Society NOM; Fig. 8b). Here, CnHmOq molecules contribute most to the total ion count. These molecules can be arranged into series, which are related by the formal exchange of CH4 against oxygen. Figure 8c denotes the mass peaks corresponding to the 37 theoretically possible and chemically reasonable C,H,O-compositions depicted in Fig. 8d that have a nominal mass of 301 Da. Note that negative M–H+ ions (i.e. [M−H+e]−) are observed in the FTICR mass spectra (Fig. 8b), and the C,H,O-compositions of molecules M are denoted in Fig. 8c. M and M–H+ differ in mass by one hydrogen (1.007825032 Da) minus an electron (0.000548625 Da); in Fig. 8c this difference is decomposed into a mass shift of one (see the shift between the mass axes) and an additional small mass spacing Δm = 0.000233878 Da. The molecules in the barley extract exhibit mass peaks outside of the range accessible for any C,H,O-composition (dotted purple box in Fig. 8a), indicating the presence of additional heteroatoms (e.g. N, P, S) in these ions. Figure 8d denotes a van Krevelen diagram of the 37 chemically reasonable CnHmOq molecules, in which the 16 C,H,O-ions observed in Fig. 8b are highlighted. The number of peaks identified corresponds to a coverage of 43% of the entire C,H,O-compositional space. These ions occupy an area for which the largest number of feasible C,H,O-isomers is expected (see Fig. 11). Molecularly intricate materials, like natural organic matter (NOM), exhibit molecular signatures approaching the theoretical limits defined by the laws of chemical binding. In Fig. 9, a van Krevelen diagram of Suwannee River fulvic acid (SuwFA) depicts the elemental ratios of CnHmOq ions (the ions shown represent a consolidation of the ions obtained by ESI, APCI and APPI positive and negative ionization from 9.4-T FTICR mass spectra; unpublished data). The peaks observed in the negative/positive ionization mode only are coloured green/orange; peaks observed in both positive and negative modes are depicted in black. The lack of signatures from biochemical precursor molecules [123] indicates the considerable level of processing typical of NOM. Within a mass range of 200–700 Da and the given limits of the H/C and O/C ratios, the minimum consolidated number of individual C,H,O-molecular compositions (4270) represents a sizable fraction (23%) of the entire feasible compositional space of CnHmOq molecules (18414 in total; small grey dots). To further appreciate the remarkable intricacy of natural organic matter, it should be noted that any dot in the van Krevelen diagrams of these complex materials represents a projection of the elemental ratios derived from assigned molecular formulae, irrespective of molecular mass. Hence, the dots in the van Krevelen diagrams can represent multiple molecular formulae (Figs. 10 and 11), while any identified molecular composition reflects an intrinsic superposition of all feasible isomers (Fig. 11). Considering typical molecular weights of several hundreds of Daltons in the mass spectra of NOM (Fig. 6, bottom panels), it is readily anticipated that the mass spectra of such systems represent simplified (e.g. isomer-filtered) projections of a still hugely more expansive structural space (Fig. 7). Key trends relating molecular composition to the number of feasible isomers For any exceedingly complex material, it is logical to postulate that many isomers will contribute to any given molecular formula. Analogously, the intensities of the mass spectral peaks, which superimpose all of the isomers present, will be a function of the abundances of these isomers in these materials and the ionization efficiency of each isomer under the given experimental conditions. For molecules of a given mass composed of carbon, hydrogen, and oxygen, two major and independent trends are expected to define the number of feasible isomers. First, decreasing the H/C ratio from fully saturated molecules (CnH2n+2) means removing hydrogen atoms, which is equivalent to introducing double bonds or (ali)cyclic structures (double bond equivalents, DBEs). Molecules with large H/C ratios are structurally fairly uniform, consisting mainly of various branched chains of single bonds. Introducing large numbers of DBEs will lead to many new structures with double bonds and or (ali)cyclic structures in various positions. For an H/C ratio of close to one, on average two carbons carry one DBE, and the introduction of further DBEs will lead to a lack of single bonds. Hence, the maximum number of feasible C,H,O-isomers is expected to occur for intermediate numbers of DBEs in a molecule, because the occurrence of a DBE (which solely depends on the H/C ratio) enables double-bond displacement and the formation of (ali)cyclic structures, both of which greatly enlarge the number of feasible isomers. In contrast, only highly condensed structures can be assembled at very low H/C ratios [37], and this constraint severely diminishes the number of feasible C,H,O-isomers (if mathematically possible but chemically unlikely isomers are excluded; see Fig. 11). Second, the insertion of oxygen into potentially any carbon–carbon (creating C–O–C units) or carbon–hydrogen bond (creating C–OH functionalities) will result in many more feasible isomers at low O/C ratios; in the presence of DBEs, “terminal” carbonyl derivatives (C=O) can also be constructed. At higher O/C ratios, however, further insertion of oxygen decreases the number of feasible isomers for two reasons: oxygen provides fewer (two) options for forming (single) bonds with other partners than carbon (four); in addition, the higher mass of oxygen (16 Da) compared with that of carbon (12 Da) decreases the total number of “heavy” atoms available for the construction of CnHmOq molecules of a given mass. These considerations imply that the number of feasible C,H,O-isomers for a given mass will reach maximum values at intermediate H/C and O/C ratios, and that these numbers will (sharply) decline at extreme (high and low) H/C and O/C ratios, respectively. These dependencies are displayed in a van Krevelen diagram (Fig. 11), in which the numbers of chemically relevant isomers for any given molecular composition CnHmOq of a single nominal IUPAC mass are provided. For any given nominal mass, the mathematically possible and chemically relevant structures composed solely of carbon, hydrogen and oxygen atoms can be constructively enumerated for each composition (molecular formula) [58]. By “chemically relevant isomers”, we mean all mathematically possible isomers (not counting stereoisomers) except for those containing O–O bonds, C≡C bonds, three- or four-membered rings, or =C= fragments (cumulated double bonds), which are not assumed to occur in the materials of interest (natural organic matter here). These data are displayed in the right panel of Fig. 11, where they are arranged according to actual mass. For practical reasons, we have selected CnHmOq compositions with a nominal IUPAC mass of 178, for which the number of isomers can be computed within a reasonable time on a desktop computer; within the given limits of H/C and O/C elemental ratios, eleven feasible C,H,O-molecules are found, which are grouped into three series of isobaric molecules, related by a formal exchange of CH4 for oxygen (Fig. 11). Series 1 represents highly unsaturated molecules in which the number of isomers declines sharply with decreasing H/C ratio. Series 2 presents the maximum number of isomers at intermediate H/C (and O/C) ratios and the decline in the number of isomers at both high and low H/C (and O/C) ratios, as anticipated (see above). The maximum H/C ratio found for a series 2 molecule amounts to almost 1.7, and the corresponding molecule C13H22 (2a) features three DBEs, thereby allowing for a much larger array of unsaturation-related isomers than obtained for a fully saturated parent molecule. This is demonstrated by the variance in the isomer count when the fully saturated analogue C13H28 (184 Da, 802 isomers) is compared with the series 2 “endmember” C13H22 (2a; 178 Da, 1.7 × 105 isomers); analogous relationships are found for the series 1 “endmember” C14H10 (1a; 178 Da, 5.3 × 106 isomers) in comparison with its fully saturated parent molecule C14H30 (198 Da, 1858 isomers). A considerable fraction of the 16.6-fold increase in the isomer count observed when comparing C13H22 (2a; three DBEs) and C12H18O (2b; four DBEs) results from the ability to produce novel isomers with singly bonded oxygen and those with a C=O bond (carbonyl derivative). The maximum number of isomers (~1.1 × 107 each) is attained for the molecules C11H14O2 (2c; five DBEs) and C10H10O3 (2d; six DBEs), respectively. Further exchange of CH4 against oxygen again sharply decreases the number of feasible isomers [by a factor of 5.5 when proceeding from C10H10O3 (2d) to C9H6O4 (2e), and by a factor of 78 when changing from C9H6O4 (2e) to C8H2O5 (2f)]. The series 3 molecules C7H14O5 (3a) and C6H10O6 (3b) feature rather limited numbers of isomers because of their large O/C ratios (see above). A comparison of C8H2O5 (2f) and C7H14O5 (3a) indicates that extreme hydrogen deficiency restricts the feasible number of isomers more severely than almost full saturation. Molecules C14H10 (1a; ten DBEs, 5.3 × 106 isomers; series 1), C10H10O3 (3b; six DBEs, 1.1 × 107 isomers; series 2) and C6H10O6 (3b; two DBEs, 6 × 104 isomers; series 3) are all related by a formal exchange of four carbons for three oxygen atoms. The introduction of oxygen initially outweighs the decrease in the number of carbon atoms and DBEs available because of (i) the reduced severity of unsaturation and (ii) the availability of oxygen to construct isomers (see above). Upon the transition from C10H10O3 (2d) to C6H10O6 (3b), however, both the lesser ability of oxygen to participate in chemical bonding (two bonds for any oxygen instead of four for any carbon) and the decline in available DBEs lead to a drastic decrease in the number of accessible isomers. Mass spectral intensities and number of feasible isomers in marine and terrestrial NOM In a highly processed and supposedly exceedingly complex material such as deep sea marine organic matter, most of the molecules of formula CnHmOq will contain an intermediate amount of unsaturation and numerous oxygen atoms [18, 20]. This flexibility to generate a potentially huge number of isomers implies that (in the absence of severe ion suppression) mass spectral intensities should correlate roughly with the number of feasible isomers for any given molecular composition. Recently, carboxyl-rich alicyclic molecules (CRAM) have been identified as prominent constituents of marine (and possibly freshwater and terrestrial) organic matter [18]. CRAM likely represent highly processed products of ultimately terpenoid origin and are expected to represent an extremely complex mixture of molecules. Based on the molecules of formulae CnHmOq and a recognition of FT mass spectral intensities, the CRAM that occur in deep ocean marine ultrafiltered organic matter comform mainly to the region inside the dotted ellipsoid in the van Krevelen diagram of Fig. 11, which appears to coincide with the maximum number of feasible C,H,O-isomers. The availability of aromatic structures in terrestrially and freshwater-derived NOM, such as that in Suwannee river fulvic acid (SuwFA; Fig. 9), opens up the compositional space of chemically relevant NOM molecules (see above) to significantly lower H/C ratios than accessible solely on the basis of open-chain unsaturation (e.g. olefinic and carbonyl) and alicyclic double-bond equivalents (DBE). These dependencies are nicely illustrated by comparing the van Krevelen diagrams of marine ultrafiltered dissolved organic matter (UDOM) [18], a blackwater NOM [18, 124], and that of SuwFA (Figs. 9 and 11). While mass spectra of marine UDOM are dominated by carboxyl-rich alicyclic molecules (CRAM), composed mainly of carboxylic groups and alicyclic rings with only negligible aromatic and olefinic unsaturation [18], the significant terrestrial, aromatic-rich signature present in both blackwater NOM and SuwFA populates the compositional space with notably lower elemental H/C ratios than feasible in marine UDOM [18]. It should be noted that oxygen-depleted molecules of formula CnHm are less likely to be ionized in standard ESI FTICR mass spectra in comparison with oxygenated molecules of formula CnHmOq. Carbohydrates, which are oxygen-rich, also are less efficiently ionized under standard ESI-FTICR mass spectral conditions than carboxyl-rich molecules like CRAM. CRAM therefore represent the most likely constituents of NOM to produce strong signals in ESI-FTICR mass spectra. The compositional space of Suwannee river fulvic acid (SuwFA) given in Figs. 9 and 11 is derived from consolidated positive and negative ion FTICR mass spectra, obtained via APCI+APPI+ESI ionization modes, thereby facilitating the observation of oxygen-depleted molecules (Fig. 9). The discrete analytical volumetric pixel space defines our current capacity to depict molecular-level variance in complex systems The current capacity to describe complex materials at molecular resolution can be visualized in the form of an analytical space comprising individual volumetric pixels (voxels). The range of this discrete and quantized space is 108–14 voxels, as defined by the significant resolution of the complementary techniques of nuclear magnetic resonance (102–5 buckets, depicting the short-range order of molecules), ultrahigh-resolution FTICR mass spectrometry (104–5 buckets, depicting molecular masses and formulae of gas-phase ions) and high-performance separation (102–4 buckets, has the capacity to investigate both ions and molecules, and so provides a way to validate NMR against MS data). An investigation of these correlated data is feasible at the level of the direct hyphenation of separation and spectroscopy [e.g. LC/NMR and LC or CE/MS; corresponding to the rear faces of the voxel space [119, 125–128]) and by means of statistical heterospectroscopy (SHY) [129, 130]; corresponding to the top face (or any two faces) of the voxel space]. Any joint mathematical analysis of these correlated data will enhance the effective resolution of the data and the significance of the molecular-level analysis of complex unknowns [119, 121, 129]. This voxel space can be readily expanded to higher dimensions by including complementary data, like those derived from genomic and proteomic analyses [84, 131–134] or by recognising selective chemical reaction products [135–140]. Degradative approaches to the characterisation of complex systems produce limited amounts of unambiguously identifiable small molecules but lose crucial linkage information. Soft and selective biochemical and chemical reactions like mild hydrolysis, reduction, oxidation and derivatisation [141, 142] of complex systems will often result in larger fragments with valuable positional and stereochemical information for the assessment of synthesis and degradation pathways. The chemical transformation of functional groups with NMR- and MS-recognisable labels enables isotope-specific functional group analysis based on structural rather than behavioural characteristics [143, 144]. Information concerning stereochemistry and stable isotope composition will become more important when assessing the origins and diagenesis of complex natural materials. Any progress in the determination of position-specific stable isotope composition (e.g. by NMR and MS methods) will be useful for advancing this field. Physical and chemical fractionation will greatly assist in these studies; further miniaturisation will enhance separation capacity and thereby improve the resolution of the analytical voxel space (Figs. 5 and 12). Integrated biomarker profiling approaches [145–147] with higher resolutions, significances and accuracies will substantially improve the quality and relevance of current systems biology approaches in the health and environmental sciences. The great progress made in the molecular-level characterisation of complex systems over the last few years and foreseeable improvements in nascent technology and concepts will lead to strong synergetic effects that will further advance our understanding of any complex natural and living system whose properties and functioning depend on both strong (covalent) and weak (noncovalent) interactions. Electronic supplementary material Below is the link to the electronic supplementary material. ESM 1 (PDF 12.9 MB)

Support_Care_Cancer-4-1-2206248

Quality of life of parents with children living at home: when one parent has cancer

Goals of work This study examined the quality of life (QoL) of cancer patients diagnosed 1–5 years previously and their spouses, with children 4–18 years living at home. Relationships between parents’ QoL and the children’s functioning were explored. Introduction The impact of cancer on a patient’s psychosocial functioning has received much attention in the literature over the past decades. It is generally acknowledged that a patient’s quality of life (QoL) decreases while they are battling their illness. Problems in QoL tend to diminish over time as a patient responds to treatment. However, a significant percentage of patients continues to experience clinically elevated levels of problems requiring professional treatment [40]. Furthermore, a growing number of studies provides evidence that cancer patients’ spouses develop problems affecting their QoL [14, 17, 21, 23, 25, 27]. Spouses have reported psychological and physical distress, which have been found to correlate with the patient’s dysfunction [17]. However, the majority of studies examine patients and spouses in later adulthood [5, 6, 13, 25, 29]. Most incidences of cancer occur in older adults. According to the Dutch Comprehensive Cancer Center’s database, 70% of all new incidences of cancer in The Netherlands in 2003 occurred in patients more than the age of 60 [11]. Parental cancer in younger families may be a more serious stressor than in later phases of life, as illness and death in later adulthood could be considered more natural or normative [37, 39]. It has been shown that older cancer patients report less anxiety and depression and better QoL than younger adult patients [26]. Couples in an earlier life phase with young children and adolescents at home lead busy lives. They juggle child-rearing responsibilities and the demands of sustaining a healthy marriage with trying to meet the individual (career) needs of each partner [28]. In the case of families with parental cancer, parents have the added strain of their loved one being ill. Additionally, their stress may be confounded when the patient may not be able to work or take care of the children as well as before the onset of the disease. Only a few studies have examined the impact of cancer on the functioning of all members in families in this phase of life. In our previous investigations of families confronted with parental cancer, the main focus has been on the children’s psychological well being [20, 36]. Results have shown clinically elevated levels of distress in 35% of adolescent daughters and 21% of adolescent sons [20] and higher levels of emotional problems in adolescent daughters and elementary school-aged sons of cancer patients [36]. It is possible that the children’s emotional and behavioral functioning, which may be affected by their parent’s cancer, affects how the parents function. The current study had four main aims. The first aim was to examine health-related QoL of patients and spouses in families with young children 1–5 years after diagnosis. We hypothesized that cancer patients and their spouses’ QoL would be lower than that of a norm group and that QoL would differ as a function of health status and gender (i.e., which parent was ill: the mother or father). Based on literature [30], we expected to find that female patients reported the lowest QoL, followed by female partners, male patients, and male partners. The second aim was to examine the effect of illness-related variables (time since diagnosis, recurrence, treatment intensity, and type of cancer) on parents’ QoL. We expected that patients and spouses would report more problems more shortly after diagnosis, when confronted with recurrence of cancer and when treatment was intense and, also, that QoL would differ depending on type of cancer. Our third aim was to examine relationships between patient’s and spouse’s QoL; we hypothesized that they would be significantly related. Finally, the fourth aim was to explore possible relationships between the parents’ QoL and the children’s emotional and behavioral functioning. Materials and methods Procedure Both hospitalized cancer patients and those being treated or seen for follow-up visits at the outpatient clinic of the University Medical Center Groningen between January 2001 and February 2003 were approached by surgical, medical, and radiation oncologists and oncology nurses. Patients were eligible for the study if they were diagnosed with cancer 1–5 years before study entry and had children between the ages of 4 and 18 years at the time of parent’s diagnosis who resided or had frequent contact with the diagnosed parent. Furthermore, both parents and children needed to be fluent in Dutch. A family could participate in the study if the patient and at least one child agreed to participate. Physicians and oncology nurses offered all eligible patients and spouses written information plus an information brochure adapted for the children. Informed consent was obtained from each family member separately, as regulated by the Medical Ethical Committee of the University Medical Center Groningen. After obtaining informed consent, researchers mailed packages with questionnaires and prepaid return envelopes to each family member. Cancer patients, spouses, and children were instructed to fill in questionnaires independently of each other and to not discuss their answers. Instruments Demographic data was gathered on: age, gender, level of education, length of relationship, number of children living at home, and who the primary caregiver in the family is. Patients provided information about when they were diagnosed, their type of cancer, treatment intensity, and recurrence. Education was measured on a seven-point scale from (1) elementary school only to (7) university degree. The eight subscales and two component summary scores of the Dutch translation of the RAND-36 were used to measure QoL [8, 31, 38]. Scores on the subscales can range from 0 to 100, with higher scores indicating better functioning. When comparing patients and spouses to a reference group, norm scores were used from the Dutch manual for the RAND-36 [31]. The manual provides mean scores from a random sample of 1,063 people between the ages of 18 and 89 (65% women, age M = 44.1) from the population register of a municipality in The Netherlands. Norm scores are provided for each subscale for the entire sample and for men and women separately. Summary scores report physical functioning and psychosocial functioning and are standardized with a mean of 50 and a standard deviation of 10. The reliability and validity of the RAND-36 has been supported in a wide number of international and national studies [8, 32, 33]. Cronbach’s alphas in this study on the eight subscales ranged between 0.72 and 0.90 for patients and 0.78 and 0.90 for partners. To investigate how parents’ QoL related to the children’s functioning, parents were asked to complete the 120-item Child Behavior Checklist (CBCL) [1, 34]. Adolescents also completed the 102-item Youth Self-Report (YSR) [2, 35]. In this study, the internalizing (emotional functioning), externalizing (behavioral functioning), and total problem (total of internalizing, externalizing, and cognitive problems) scales were used to provide a picture of the problems occurring in the children of parents diagnosed with cancer. The CBCL’s and YSR’s reliability and validity have been supported in a great number of studies. In this study, Cronbach’s alphas for the internalizing, externalizing, and total problem scales ranged from 0.84 to 0.94 for reports from patients, spouses, and adolescent children. Analysis In our analyses, a single variable was created to define treatment intensity. Patients were grouped into two categories based on the clinical expectation that surgical treatment alone (nonintense treatment) would be less distressing to the family because of less time away from home and fewer visible side effects. Other single-modal (chemo-, radiotherapy) and multimodal treatments (combination of surgery, chemotherapy, radiotherapy, hormone, or immunotherapy) were defined as intense. Time since diagnosis was calculated by subtracting the date on which the patient filled in the questionnaires from the date they were diagnosed. Descriptive analyses were performed on demographic information. Comparisons between patients, spouses, and the reference group were investigated using independent t tests. Results were corrected for the number of tests performed with a Bonferroni correction for multiple comparisons; p values less than or equal to 0.0015 were considered significant (0.05 of 32 tests). To assess clinical relevance, effect sizes (ESs) were calculated by dividing the difference between means by the square root of the average of the squared standard deviations [10]. ESs between 0.20 and 0.49 were considered small, 0.50 and 0.79 medium, and ≥0.80 large [10]. We considered that ESs greater than or equal to 0.50 indicated a clinically relevant difference [24]. A single variable was created to describe parents’ gender and health status (ill mothers, ill fathers, healthy mothers, and healthy fathers). One-way analyses of variance (ANOVAs) with a Bonferroni post-hoc test were used to compare QoL of patient and spouse as a function of gender and health status. After the correction for multiple tests was applied, p values less than or equal to 0.005 were considered significant (p = 0.05 of ten tests). In analyses of QoL as a function of type of cancer, only groups of ten or more patients were included. An ANOVA and a Kruskal–Wallis test (some patient groups were small) were performed to compare patients’ and spouses’ QoL as a function of type of cancer. After the correction for multiple tests was applied, p values less than or equal to 0.005 were considered significant (p = 0.05 of ten tests). The two QoL composite scores (physical summary score and psychosocial summary score) were used in correlational analyses to explore relationships between parents’ QoL and between each parent’s QoL and the children’s problems, with separate analyses for elementary school-aged (4–10 years) and adolescent (11–18 years) children. Pearson’s product-moment correlation coefficients were calculated; correlations with a coefficient less than 0.30 were considered weak, 0.30–0.50, moderately strong, and greater than 0.50, strong [10]. Results Participants A total of 476 families were approached for the study. Two hundred and nine agreed to participate (44%), including 336 children, with an average of 2.3 children per family. Because this study focused on couples, we selected data from families where both patient and spouse filled in questionnaires, resulting in a database of 166 couples and 304 children. Table 1 summarizes the demographic information. The majority of cancer patients in this study were women (78%). Patients were diagnosed with various types of cancer (Table 3), including breast (52%), gynecological (10%), hematological (9%), skin (9%), urological (6%), soft tissue and bone tumors (6%), head/neck (4%), gastrointestinal (3%), and central nervous system (1%). Fourteen percent of patients had undergone nonintense treatment (N = 23), and 86% had undergone intense treatment (N = 143). Table 1Demographics PatientSpouseChildrenMean age in years (SD)44.7 (4.9)45.8 (6.4)14.5 (4.2)Range32.8–57.831.2–65.74–23Gender, N (%)—female129 (78)37 (22)167 (55)Gender, N (%)—male37 (22)129 (78)137 (45)Mean length of relationship in years (SD), range 4–4121.3 (6.6)Mean education level (SD), range 1–73.8 (1.6)3.9 (1.8)Mean number of children per couple, range = 1–62.3Primary caregiver in family, N (%) Mother105 (63) Father5 (3) Both parents54 (33) Other2 (1)Mean time since diagnosis in years (SD)2.76 (1.2)Treatment intensity—nonintense, N (%)23 (14)Treatment intensity—intense, N (%)143 (86)Recurrence of cancer N (%)36 (22) There were no significant differences between participating and nonparticipating parents regarding patient’s gender, type of cancer, or time since diagnosis. Of the nonparticipants, 22% (N = 59) declined because of reasons directly related to the parents (e.g., parents had moved on with their lives or were too emotionally distressed). Reasons related to the children (e.g., children were not interested, children had not been informed of parent’s illness) were given as the explanation for not wanting to be included in 20% (N = 53) of nonparticipating families. Twenty-five percent (N = 67) mentioned a variety of reasons, including another illness in the family or the parents’ or children’s busy-ness. The remaining 33% (N = 88) did not provide an explanation. Patients’ and spouses’ QoL in comparison with the norm Results are summarized in Table 2. Clinically relevant ESs were found for male patients’ social functioning, role limitations because of physical problems, and vitality. Both male and female patients scored statistically significantly lower QoL than the norm on one subscale: vitality. Female patients additionally scored lower on two subscales: social functioning and role limitations because of physical problems. Table 2Descriptives of the RAND-36 scores by parents’ gender and health status and comparison between groups and with the norm group GenderPatientPartnerHealth status and genderNormMeanSDt (v. norm)Effect sizeMeanSDt (v. norm)Effect sizeANOVA FMeanSDPhysical functioningfemale76.122.2−2.120.2091.914.94.290.7512.9980.723.6male77.322.0−1.960.3389.116.92.440.2784.522.3Social functioningfemale78.524.4−3.160.3174.715.9−4.170.714.2886.120.9male72.628.4−3.310.5685.621.71.290.1388.419.6Role limitations—physicalfemale64.941.2−3.440.3384.229.11.190.207.2078.336.5male57.445.6−3.130.5381.635.00.030.0081.533.6Role limitations—emotionalfemale79.335.60.980.0981.933.9−0.090.021.4582.533.5male83.832.9−0.620.1187.828.90.160.0287.329.3Mental healthfemale72.516.8−1.820.1874.614.8−0.360.062.9675.518.9male78.514.5−0.370.0677.916.0−0.890.0979.417.3Vitalityfemale58.821.8−3.630.3463.817.6−0.840.146.5466.319.6male56.921.4−3.070.5969.118.7−0.230.0269.520.5Painfemale81.621.5−0.780.0785.318.3−1.690.292.3480.025.4male80.424.60.660.1187.719.22.140.2383.223.8General healthfemale66.321.9−2.250.2474.815.3−1.270.228.7171.521.8male58.525.9−2.920.4975.318.7−1.920.2171.423.3Physical summaryfemale47.29.8––49.710.2––11.38Not availablemale44.911.4––50.39.8–Not availablePsychosocial summaryfemale49.710.2––53.68.6––2.31Not availablemale44.911.4––52.28.5––Not availablep values were corrected for multiple comparisons.Significant values are shown in italics: p ≤ 0.0015 for comparisons with the norm; p ≤ 0.005 for comparisons of health status and gender. Clinically relevant ESs were found for female spouses’ physical functioning and social functioning. Female spouses reported statistically significantly higher scores than the norm on physical functioning. They also scored statistically significantly lower on social functioning than the norm. Effects of health status and gender, time since diagnosis, recurrence, treatment intensity, and type of cancer on QoL Parents’ health status and gender had a statistically significant effect on parents’ physical summary score. A statistically significant effect was also found on parents’ physical functioning, role limitations because of physical problems, vitality, and general health perception (Table 2). A Bonferroni post-hoc test revealed the following. On the physical summary score, female patients scored lower than male spouses (p < 0.001), and male patients scored lower than female and male spouses (p < 0.001 for both). On physical functioning, female patients scored lower than female and male spouses (p < 0.001 for both). On role limitations because of physical problems, female and male patients scored significantly lower than male spouses (p = 0.003 and p = 0.005, respectively). On vitality, female patients scored significantly lower than male spouses (p < 0.001). Finally, on general health perception, female patients scored significantly lower than male spouses (p = 0.003), and male patients scored significantly lower than female and male spouses (p = 0.004, p < 0.001, respectively). Time since diagnosis did not correlate significantly with patients’ QoL. Time since diagnosis correlated weakly with spouses’ physical summary score (r = 0.19, p = 0.031). Patients with a recurrence scored clinically relevantly lower than patients without recurrence on the physical summary score (t = 3.44, ES = 0.82, p = 0.001), social functioning (t = 3.12, ES = 0.79, p = 0.003), role limitations because of physical problems (t = 3.72, ES = 0.72, p < 0.001), and general health perception (t = 3.69, ES = 0.82, p = 0.001). Spouses of patients with a recurrence scored clinically relevantly lower than spouses of patients without a recurrence on the psychosocial summary score (t = 2.23, ES = 0.94, p = 0.027) and the subscales: role limitations because of emotional problems (t = 2.72, ES = 0.78, p = 0.009), vitality (t = 3.56, ES = 0.69, p < 0.001), and general health perception (t = 2.19, ES = 0.52, p = 0.033). Treatment intensity was significantly related to patients’ QoL. Patients who received intense treatment scored clinically relevantly lower than patients who received nonintense treatment on the physical summary score (t = 2.81, ES = 0.79, p = 0.004) and the following six subscales: physical functioning (t = 2.32, ES = 0.69, p = 0.02), social functioning (t = 2.33, ES = 0.76, p = 0.02), role limitations because of physical problems (t = 2.95, ES = 0.68, p = 0.006), vitality (t = 3.03, ES = 0.81, p = 0.003), pain (t = 3.19, ES = 0.78, p = 0.016), and general health perception (t = 2.60, ES = 0.69, p = 0.01). Treatment intensity was not significantly related to spouses’ QoL. Type of cancer had a statistically significant effect on patients’ QoL as summarized in Table 3. Statistically significant differences between patient groups were found on the physical summary scale and five subscales. Patients with skin cancer reported the highest QoL. One third of skin cancer patients received intense treatment. Patient with hematological cancer reported the lowest QoL levels. All 16 hematological cancer patients received intense treatment. A Kruskall–Wallis test confirmed these findings. With regard to the spouses, QoL did not vary as a function of the patient’s type of cancer. Table 3Descriptives of the patients’ RAND 36 scores by type of cancer and comparisons between groups Gynecological tumorsBreast cancerUrological tumorSoft tissue/boneHematologicalDermatologicalANOVAN = 16N = 86N = 10N = 10N = 16N = 15FMSDMSDMSDMSDMSDMSDPhysical functioning78.117.178.520.481.014.759.515.757.832.188.315.85.21Social functioning77.317.279.223.372.522.770.025.854.735.996.79.95.34Role limitations—physical64.043.868.339.962.511.532.533.431.243.386.728.14.61Role limitations—emotional91.725.879.534.796.710.576.741.760.442.5100.000.03.16Mental health73.015.073.616.577.616.469.525.974.517.776.013.80.29Vitality60.617.658.920.758.524.352.016.943.428.777.318.84.18Pain86.917.681.721.975.919.160.620.973.629.292.813.93.44Gen. health perception72.816.064.923.161.021.654.517.947.228.285.79.85.79Physical summary78.117.1178.520.481.014.759.515.757.832.188.315.86.48Psychosocial summary77.317.279.223.372.522.770.025.854.735.996.79.92.95p-values were corrected for multiple comparisons. Significant values are shown in italics: p ≤ 0.005 Functionality of patient in relation to functionality of spouse Spouses’ psychosocial functioning was moderately strongly related to patients’ psychosocial functioning (r = 0.44, p < 0.001) and weakly related to patients’ physical functioning (r = 0.29, p < 0.001). No significant relationship was found between the spouse’s physical functioning and the patient’s psychosocial (r = 0.06) or physical functioning (r = 0.07). Relationship between patient and spouse’s functioning and emotional and behavioral functioning of the children Patients’ physical functioning was not significantly related to their elementary school-aged children’s functioning but was weakly related to the adolescents’ total problems and internalizing as reported by the patients (Table 4). Patients’ psychosocial functioning was moderately strongly to weakly related to their elementary school-aged children’s internalizing, externalizing, and total problems and weakly related to the adolescents’ internalizing, externalizing, and total problems as reported by the patient. The patients’ physical and psychosocial functioning was moderately strongly to weakly related to the adolescents’ internalizing and total problems according to adolescents’ self-reports; their physical functioning was weakly related to adolescents’ externalizing. Table 4Correlations between parents QoL, as measured by RAND-36 composite scores, and children’s functioning by age group as reported by parents (CBCL) and adolescents themselves (YSR) Patient (N = 162)Spouse (N = 150)Physical functioningPsychosocial functioningPhysical functioningPsychosocial functioningElementary school-aged children (CBCL)Total−0.05−0.33**−0.19−0.11Internalization−0.03−0.36**−0.19−0.17Externalization−0.03−0.26**−0.16−0.05Adolescents (CBCL)Total−0.18**−0.24**−0.11−0.22**Internalization−0.19**−0.23**−0.15−0.27**Externalization−0.10−0.21**−0.09−0.09Adolescents (YSR)Total−0.32**−0.29**−0.11−0.15Internalization−0.31**−0.32**−0.11−0.19*Externalization−0.20*−0.13−0.09−0.02*p < 0.05**p < 0.01 Spouses’ psychosocial and physical functioning was not found to be significantly related to their elementary school-aged children’s functioning as reported by the spouses. Additionally, their physical functioning was not significantly related to the adolescents’ functioning. Spouses’ psychosocial functioning was weakly related to the adolescents’ internalization and total problems as reported by the spouses. Finally, the spouses’ psychosocial functioning was weakly related to the adolescents’ internalizing as reported by the adolescents (Table 4). Discussion The aim of this study was to gain insight into the QoL of cancer patients in the child-rearing stage and their spouses. We found that cancer patients 1–5 years after diagnosis evaluate their QoL as clinically relevantly and/or statistically significantly lower than the normal population on three of the eight domains, partially supporting our first hypothesis. This is in line with studies that have reported a decrease in patients’ social and physical domains [7, 9]. However, our study did not find a decrease in patient’s emotional functioning, in contrast to some reports [3, 4, 16]. Our study focused on a subgroup of cancer patients, namely those who are relatively young. The finding that our patient group scored similarly to the norm seems to indicate that these patients’ lives have gotten fairly back to normal. These parents seem to be handling the unexpectedness of a cancer diagnosis during this life phase, the treatment, or the confrontation with a possible death at an early age, coupled with the responsibility of raising children fairly well. However, they still seem to experience problems in some areas, specifically social functioning, role limitations because of physical problems, and vitality. Spouses of cancer patients reported a QoL comparable to the norm group, and female spouses reported even better physical functioning than the norm. Only on one subscale (social functioning) did spouses (and then only women) report decreased functioning. Our findings largely negate our hypothesis and are in contrast with other studies where spouses have reported a decreased QoL [27]. An explanation for our findings may be that spouses in other studies, where the average age is higher, may have had a lower QoL simply because of their older age. As age increases, QoL scores decrease [33]. A second explanation could be that spouses viewed their QoL relative to their ill partner and therefore regarded their own health as good as or better than people generally may. It is also possible that we did not find a decreased QoL in the spouses because our patient group ranged from 1 to 5 years after diagnosis; time since diagnosis correlated significantly positively with spouses’ physical functioning. The threat that the patient might not survive may be less prominent for spouses at this point in time. Our study found significant effects of gender and health status on the physical summary score and four of the eight subscales. To summarize our findings, patients scored lower than spouses, male spouses reported the highest QoL, and, on some scales, the female patient reported the lowest QoL and, on others, the male patient. In a study on gastrointestinal cancer, female patients and female spouses were both reported to suffer overall more distress than male patients [30], which our study did not find. They also reported that female spouses reported lower QoL than male patients and spouses. However, that study examined patients within 6 months after surgery. They also focused specifically on patients with gastrointestinal cancer, which is equally prevalent in men and women. That allowed them to more easily generalize that the differences they found were due to gender. As our study sample was diagnostically heterogeneous, it is difficult to differentiate whether the observed differences were due to gender or cancer site and consequent treatment received. Our study sample consisted of 78% women, which could seem skewed. However, according to the Dutch Comprehensive Cancer Center’s database, cancer occurs more frequently in women during this age range; approximately 65% of cancer patients in this region of The Netherlands with invasive tumors are women [12]. To further explore our second aim, we analyzed QoL as a function of type of cancer and found differences depending on the type of tumor the patient had, supporting our hypothesis. Skin cancer patients in our study reported the highest QoL, while patients with hematological cancer reported the lowest QoL. These differences may be related to the kind of treatment the patient received. We found that patients who received nonintense treatment reported a better QoL than those who received intense treatment. The removal of a localized melanoma that only requires outpatient excision may affect QoL less than frequent and long hospital stays for courses of chemotherapy for hematological malignancies. Our findings are similar to another study reporting patient distress levels [40] that found that distress levels varied depending on cancer site. With regard to how spouses function, no significant QoL differences were found because of the type of cancer or treatment intensity. It would seem that tumor type or treatment intensity do not affect spouses’ functioning. Whether the patient had a recurrence does seem to affect spouses’ QoL; the ES of the psychosocial summary score was strikingly large. This may be due to, for those spouses, the still current threat that the patient may not survive. It is interesting to note that recurrence seems to have affected patients’ physical and social functioning but not their mental functioning. Patients with recurrence do not report more mental distress than patients who have not had a recurrence. Thirdly, we investigated relationships between patient and spouse QoL and found weak to moderate positive relationships between cancer patient’s physical and psychosocial functioning and spouse’s psychosocial functioning. These findings are in line with two studies reporting a moderate positive relationship between patients’ and spouses’ psychological distress [9, 18]. Finally, the hypothesis that we would find a significant relationship between the parents’ and the children’s’ functioning was supported, although the relationships were not strong. The patient’s QoL related more often significantly with the children’s functioning than the spouse’s. Given that four of five of the patients were women, it could be that mothers were more often alert to possible problems in the children than fathers. This may be due to the large number of families in our sample where the mother, sick or healthy, is the primary caregiver. Mothers tend to orient themselves more toward others, whereas fathers tend be more self-oriented [18]; this could enable mothers to judge problems better. It is possible that illness plays a role in the patients’ reports as distressed parents are likely to rate more behavioral and emotional problems in their children [22]. However, we found significant relationships between the adolescents’ self-reports and parents’ functioning. Unfortunately, the cross-sectional design of this study limits us in understanding the causal nature of this finding. We found more significant, negative correlations with the children’s internalizing and total problem scores than externalizing. It would seem that parents’ QoL negatively relates to the children’s emotional functioning, more than to their acting up. It is interesting to note that the parents’ functioning related differently to adolescents and school-aged children. The patient’s physical functioning correlated weakly with the adolescents’ functioning but not significantly with the younger children’s functioning. This was not found for spouses—their physical functioning did not relate significantly to how children from either age group function. Patient’s psychosocial functioning correlated weakly to moderately strongly with children in both age groups, while spouses’ psychosocial functioning correlated weakly with the adolescents’ functioning. Adolescent self-reports provided a similar picture; we found weak to moderately strong correlations between patients’ physical and psychosocial functioning but only one weak correlation between spouses’ psychosocial functioning and adolescents’ internalization. Our finding that parents’ functioning related more often significantly with the adolescents may be due to adolescents’ stage of cognitive development. Adolescents are more able to understand the patient’s illness and may pick up on physical problems more than elementary school-aged children [15]. Research considerations In our investigation, being the first to evaluate family functioning in families in the child-rearing age, we purposely did not choose a homogenous cancer group (i.e., only breast cancer). Our results are statements for the general group of child-rearing families with cancer. Comparisons between patients in our research group with varying diagnoses showed differences between groups; however, the small subgroups make it difficult to generalize our findings. Future research will be required to investigate the effect of varying forms of cancer, the stage of cancer, its prognosis, and treatment on the functioning of individual family members. Additionally, our response rate of 44% could mean that despite the fact that no differences were found between participants and nonparticipants in gender, type of cancer, or time since diagnosis, a sample bias may exist. We cannot be sure whether psychological problems were over- or under-reported; some nonrespondents stated still being overwhelmed by the illness as a reason for not participating, while others indicated they had moved on. Additionally, the cross-sectional nature of this study makes it impossible to accurately capture the dynamic processes present in family relationships or determine whether parents’ QoL effects the children’s functioning or vice versa. Furthermore, a longitudinal study could provide insight into the QoL of cancer patients and their family members over a period of time. This study found significant relationships between patient and spouse functioning and between parent’s functioning and children’s behavior. However, these relationships appeared to have modest predictive power. The patients’ and spouses’ functioning is likely more influenced by other factors not measured in this study, such as personality, social support, or family environment [19, 26]. In conclusion, cancer patients’ QoL 1–5 years after diagnosis seems to be returning to normal, except in three domains. Their spouses seem to be doing well. The patients’ QoL varied according to the type of cancer, how intense their treatment had been, and whether they had experienced a recurrence. Spouses’ QoL seemed to be unaffected by the type of cancer and treatment intensity but did vary depending on whether the patient experienced a recurrence. A moderate positive relationship was found between the patient’s functioning and his/her spouse’s. Parents’ physical and psychosocial functioning was weakly to moderately strongly related to their children’s functioning. The patients’ functioning related more strongly to the children’s functioning than the spouses’ did. How cancer patients’ families function may have an impact on the patient’s functioning, up to 5 years after diagnosis. This is something that should be taken into account by clinicians.

Mol_Syst_Biol-4-_-2387231

A map of human protein interactions derived from co-expression of human mRNAs and their orthologs

The human protein interaction network will offer global insights into the molecular organization of cells and provide a framework for modeling human disease, but the network's large scale demands new approaches. We report a set of 7000 physical associations among human proteins inferred from indirect evidence: the comparison of human mRNA co-expression patterns with those of orthologous genes in five other eukaryotes, which we demonstrate identifies proteins in the same physical complexes. To evaluate the accuracy of the predicted physical associations, we apply quantitative mass spectrometry shotgun proteomics to measure elution profiles of 3013 human proteins during native biochemical fractionation, demonstrating systematically that putative interaction partners tend to co-sediment. We further validate uncharacterized proteins implicated by the associations in ribosome biogenesis, including WBSCR20C, associated with Williams–Beuren syndrome. This meta-analysis therefore exploits non-protein-based data, but successfully predicts associations, including 5589 novel human physical protein associations, with measured accuracies of 54±10%, comparable to direct large-scale interaction assays. The new associations' derivation from conserved in vivo phenomena argues strongly for their biological relevance. Introduction Although considerable progress has been made in mapping the protein interaction network of yeast (Ito et al, 2000, 2001; Uetz et al, 2000; Ho et al, 2002; Gavin et al, 2006; Krogan et al, 2006), only minimal progress has been made on the interaction networks of higher eukaryotes, due primarily to their scale: for the ∼20 000–25 000 human proteins, we expect a network of roughly 1–400 000 interactions (Hart et al, 2006). Among the few methods scaleable to this size, the yeast two-hybrid assay has proven the most successful, with maps of ∼20 000 interactions in fly (Giot et al, 2003), ∼4000 in worm (Li et al, 2004), and more recently, assays of ∼2800 and ∼3200 human protein interactions (Rual et al, 2005; Stelzl et al, 2005). Direct mapping of protein complexes by mass spectrometry has also contributed another ∼5000 interactions (Ewing et al, 2007). After including previously known human protein interactions (Bader et al, 2003; Lehner and Fraser, 2004; Peri et al, 2004; Joshi-Tope et al, 2005; Ramani et al, 2005), the human protein interaction map is currently perhaps 10–30% complete (Hart et al, 2006). It is therefore important to identify and employ methods for discovering interacting proteins without exhaustive experimental measurement of all pairs of proteins under each relevant condition or assay. Proteins are evolved to interact under specific conditions in the cell, with the cell correspondingly optimized to facilitate these events, e.g. by expressing mRNAs before proteins are required, coordinating the expression of interacting partners, directing proteins to appropriate locations for their interactions, and so on. In this way, in vivo protein interactions are accompanied by corollary events that can be used to identify biologically relevant physical interaction partners. We took advantage of two such corollary data types, the tendency for interacting proteins to have correlated mRNA expression patterns and the evolutionary conservation of such patterns, to identify new human protein interactions. It is well established that genes whose mRNA expression patterns are correlated across many diverse conditions can often be inferred to ‘work together', i.e. to be functionally coupled (Eisen et al, 1998; Marcotte et al, 1999; Stuart et al, 2003; Lee et al, 2004a; Segal et al, 2004). Analyses of co-expression patterns of orthologous genes have shown that the conserved correlation in expression can also be used to transfer functional information across species (Teichmann and Babu, 2002; Stuart et al, 2003; van Noort et al, 2003; Bergmann et al, 2004; Snel et al, 2004). Transcriptional co-expression patterns have proved useful for inferring physical protein interactions (e.g. Deane et al, 2002; Jansen et al, 2003), with strongly co-expressed mRNAs more likely to indicate long-lived interactions (Ge et al, 2001; Jansen et al, 2002; Simonis et al, 2006). In general, we do not expect transcriptional data to distinguish between direct protein binding and membership in the same protein complex, and we term all such cases physically associated proteins. To exploit these trends, we applied a supervised algorithm to discover physical associations among human proteins based upon the co-expression of their mRNAs and that of their orthologs in five organisms (the mustard plant Arabidopsis thaliana, the mouse Mus musculus, the fly Drosophila melanogaster, the nematode Caenorhabditis elegans, and the yeast Saccharomyces cerevisiae). By this approach, we mapped 7000 predicted human protein physical associations, of which 5589 are new to this analysis. Results Predicting physically associated proteins from patterns of conserved co-expression Figure 1 illustrates the overall method. We first identified orthologs for human genes in five other organisms using the InParanoid algorithm (Remm et al, 2001). We then compared the correlation in mRNA expression of each pair of human genes with the correlations in expression of each of their corresponding ortholog pairs from five organisms, in all calculating mRNA expression correlations for 5 708 925 human gene pairs on the basis of 3977 DNA microarrays. After removing 105 140 gene pairs likely to cross-hybridize on the microarrays (see Materials and methods) and filtering pairs with nonsignificant correlations, we employed a supervised algorithm on these data to identify those patterns of conserved co-expression (CCE) diagnostic of physical protein associations, based upon the correlations observed for known protein interactions versus random protein pairs. By searching for additional gene pairs exhibiting these patterns, we identified new associations. Figure 2 plots the derivation of the relationship between CCE and the tendency to be in the same physical complex, relying in this case on the comparison of human and C. elegans mRNA expression data. Briefly, the distribution of mRNA co-expression relationships was measured for 1769 gene pairs whose corresponding proteins are known to physically associate (Ramani et al, 2005), serving as positive training examples (Figure 2A); these 1769 pairs represent the subset of known human protein associations in the training set that also occur in the human–worm co-expression data sets. Likewise, the distribution was measured for 642 295 gene pairs that are in the physical interaction training set but are not known to physically associate, serving as a negative training set (Figure 2B). Therefore, the log ratio of these two plots, corrected by prior expectation, represents the log likelihood for protein pairs to physically associate given any particular pattern of co-expression conservation (Figure 2C): where P(I∣D) and P(∼I∣D) are the frequencies of positive (I) and negative (∼I) training associations observed in the data set (D), respectively, while P(I) and P(∼I) represent the overall frequencies of positive and negative training associations, respectively. This score indicates how likely two proteins are to physically associate given their specific mRNA co-expression conservation in these data. The training set includes both direct interactions and protein pairs belonging to the same complex; we therefore consider this approach to support the more general case, i.e. proteins belonging to the same complex whether or not directly interacting. Note that the highest scores do not necessarily occur in the extreme top right corner of Figure 2C; lower counts of both positive and negative examples in the extreme corner, as well as filtration of highly correlated gene pairs where they may suffer from DNA microarray cross-hybridization (see Materials and methods), results in the highest scores occurring at correlation coefficients less than one. We similarly analyzed co-expression patterns of human gene pairs with orthologs from four other organisms (A. thaliana, M. musculus, D. melanogaster, and S. cerevisiae), analyzing 3977 DNA microarray experiments in all. From each analysis (Figure 2C and Supplementary Figure 1) strongly co-expressing human genes with co-expressing orthologs are generally likely to encode physically associated proteins. The highest likelihoods of associating occur when the mRNA expression patterns of both human gene pairs and their orthologs are positively correlated, with odds of associating approaching 460:1 for C. elegans, 200:1 for A. thaliana, 100:1 for M. musculus, 25:1 for D. melanogaster, and 400:1 for S. cerevisiae. These learned relationships between mRNA expression profiles and physical associations were then applied to protein pairs not in the training set, thereby assigning a likelihood of physically associating to each untested protein pair. Each human gene pair discovered has at least one log likelihood score, to a maximum of five, from which the highest score was identified; pairs were ranked based on this score, then evaluated as a function of their rank. Validation of predicted physical protein associations using known interactions As this assay relies upon indirect evidence, it is critical that putative physical associations discovered by this approach be carefully evaluated. We devised six tests for the enrichment of true physical associations, including direct experimental assay of physical association and of four proteins' functions suggested by the associations. First and most critically, to verify the accuracy of the co-expression-derived associations, we measured their likelihood to physically associate using an independent test set of 15 810 known physical associations, including both direct interactions and complexes (Ramani et al, 2005). Figure 3A shows that the CCE associations are highly enriched for true physical associations, varying from a likelihood ratio of ∼60:1 to as high as ∼400:1 of correctly capturing true physical associations. Importantly, the CCE pairs score ∼25–200 times higher than randomized pairs of the same proteins, as well as associations derived in the same manner but using only human (not ortholog) DNA microarray data (Figure 3A). Therefore, the data from orthologs enriched the signal for human physical protein associations considerably beyond the human data alone. Second, we examined the functional relationships between the putative interaction partners. For this test, we compared the Gene Ontology (GO) and KEGG pathway database annotations of interacting partners, using a log likelihood framework (Lee et al, 2004b; Ramani et al, 2005) and testing the performance of the mapped associations with that of literature physical interactions (Figure 3B) (Bader et al, 2003; Peri et al, 2004; Joshi-Tope et al, 2005; Ramani et al, 2005). Literature associations score in the range of log likelihood ratio (LLR)=2.6–3.6, indicating high consistency with GO/KEGG annotation. As expected, randomized interactions score near zero, and interactions derived from human-only co-expression data score lower (LLR=0.59–1.09). The CCE associations are comparable to the literature associations. Using interactions transferred from other organisms (orthology core set (Lehner and Fraser, 2004); LLR=2.2) to define a threshold of minimum acceptable quality and choosing associations (in bins of 1000) exceeding this threshold, we obtain 7000 associations from the present analysis, and all subsequent tests were performed on this set. These associations have a minimum likelihood of 9:1 (90%) of belonging to the same GO/KEGG pathways. For consistency, subsequent tests include comparisons to the top-scoring 7000 associations derived from human-only mRNA co-expression, as well as to networks generated from the same proteins found in the CCE associations, but connected by 7000 random interactions (N=10 random networks). Validation of predicted physical protein associations by mass spectrometry We next used quantitative mass spectrometry to test for physical associations between CCE partners (Figure 4A). Performing purifications under native conditions known to keep protein complexes intact (Dignam et al, 1983), HeLa cells were lysed, the cytoplasmic and nuclear/mitochondrial fractions were separated, and their respective contents were fractionated biochemically on two sucrose density gradients. Proteins were quantified in each fraction by mass spectrometry. In all, 3013 proteins were quantified across 14 cytoplasmic and 14 nuclear/mitochondrial sucrose density gradient fractions (Figure 4B). As proteins in the same complex should generally co-sediment, we expected physically associated proteins to often have correlated elution profiles. Analysis of known protein complexes verified that components of a complex tended to co-elute (Figure 4C; additional controls in Supplementary Figures 3–5). For example, components of the TCP1 chaperone complex show strongly correlated elution profiles, as do core components of RNA polymerase II; the latter profiles are distinct from the former. Likewise, components of the NADH dehydrogenase 1b complex show strongly correlated elution profiles, eluting entirely in the nuclear/mitochondrial fraction (Figure 4C). As an example of the utility of this approach, the protein GRIM-19, initially identified as a regulator of cell death induced by interferon-beta and retinoic acid, was later identified to be a subunit of the NADH dehydrogenase complex 1 (Fearnley et al, 2001); this association is clearly evident in the co-elution of GRIM-19 with other components of this complex. More systematically, positive control human protein interaction partners known from literature (Joshi-Tope et al, 2005) show highly correlated elution profiles (Figure 5), unlike negative control random pairs (see histograms in Figure 6A). For cases in which both interaction partners were observed in the mass spectrometry experiment, 63% of the positive control pairs exhibited Pearson correlation coefficients >0.4, indicating a false-negative rate for identifying physical associations using the mass spectrometry-based elution profiles of 37% at this correlation threshold (28% if considering correlation coefficients >0.2). This agrees with the expectation that not all interacting proteins will co-sediment, with a probable bias toward stable complexes. Similarly, for proteins in the positive control set, protein pairs with the most correlated elution profiles showed ∼40% probability of being in the same physical complex (Figure 5). Thus, correlated elution across these fractions is a strong indicator of direct physical association. As this assay is not used independently for discovery, but is confirmatory in nature, the false-negative and true-positive rates are sufficient for evaluating CCE associations. Although individual associations could be validated in this manner, by instead examining the aggregate distribution of elution profile correlation coefficients, we could directly estimate the error rate of the CCE associations. We calculated histograms of Pearson correlation coefficients from pairwise comparisons of elution profiles for CCE protein pairs, for protein pairs known from literature (Joshi-Tope et al, 2005) to be in the same complexes, and from random pairs of proteins (Figure 6A). We then fit the CCE histogram as a linear mixture of the positive and negative control histograms; the proportions that give the best fit thereby provided an estimate of the relative proportions of true and false associations in the CCE set. From this analysis, we estimate that 49–59% of the CCE associations correspond to true physical associations (Figure 6A, inset). Comparing the CCE associations and the shotgun proteomics elution profiles reveals many interesting associations. For example, known complexes are correctly recovered, as for the DNA replication licensing factors MCM3, 5, 6, and 7, or for components of the proteasome. Figure 6B shows the example of the proteins prohibitin and prohibitin-2, known to form a large complex on the mitochondrial membrane that acts to suppress apoptosis, but which also shuttles to the nucleus in an estrogen-receptor-dependent manner and acts to repress transcription (Kasashima et al, 2006). New associations are also revealed: we observed a predicted physical association between MCM3 and MCM6 with the retinoblastoma-binding protein 4 (RBBP4). RBBP4 is known to participate in several chromosome replication and chromatin remodeling complexes, among them the chromatin assembly factor CAF-1 and a DNA replication-dependent chromatin assembly complex (Verreault et al, 1996). The CCE associations, supported by mass spectrometry, suggest direct physical association of RBBP4 with the replication initiation complex as well. Figure 6B illustrates two other such examples: first, we predict the ras oncogene-related small GTPase RAB5A, an essential component of receptor-mediated endocytosis (Bucci et al, 1992), to associate with the clathrin assembly lymphoid–myeloid leukemia gene (CALM), a protein that helps recruit clathrin to endocytic vesicles. The CALM gene is a recurring site for chromosomal translocations in acute myeloid and mixed lineage leukemias (Wechsler et al, 2003). Physical association with RAB5A highlights a possible functional connection between these two endocytic components and is interesting in light of the leukemogenesis potential of chromosomal translocations involving CALM. Likewise, we predict the A-kinase anchor protein AKAP1 to be associated with the splicing factor SFRS9. AKAP1 is primarily involved in anchoring protein kinases, phosphatases, and a phosphodiesterase to specific cellular locations, but also contains KH and Tudor domains, motifs for single-strand RNA binding (Trendelenburg et al, 1996), which help target AKAP1 to well-defined nuclear foci in an RNA-dependent manner (Rogne et al, 2006). The association with SFRS9, which among other functions is involved in both constitutive and alternative splicing and can be specifically localized with other RNA processing proteins to nuclear stress bodies (Denegri et al, 2001), suggests that AKAP1 may also have a role in these processes or in mRNA localization, perhaps integrating RNA processing with signaling. Quantitative estimates of interaction accuracy We further validated predicted physical associations by additional approaches. For each of these tests, we defined a standard curve relating a quantitative property of a protein pair with the tendency of the pair to physically associate, constructing the curve from control mixtures of literature (positive) and randomized (negative) physically associating proteins (Figure 7A). A set of all true-positive physical associations typical of those used to construct the curve scores high on these tests (∼100% for each standard curve); the addition of random interactions degrades the performance. The relationship between each test's performance and inclusion of false positives is unambiguous and well behaved, as judged by the quality of the standard curves, agreement between the tests, and performance curves from mixtures of known accuracy (Supplementary Figure 6). It is important to note that there are possible sources of bias for each test; however, taken as a whole, the tests are strong indicators for the enrichment of true physical associations. First, true human physical protein associations should be enriched for physical interactions among orthologous proteins in model organisms. We generated control association sets with known error rates by randomly selecting sets of 7000 interactions with varying proportions of true-positive and true-negative associations, ranging from 0% true positive (all 7000 interactions chosen from the true-negative set) to 100% true positive (all 7000 interactions chosen from the literature set), repeating the randomization 10 times. We measured the overlap of each control set with a benchmark set of 19 119 human protein pairs whose worm, yeast, or fly orthologs have been observed to interact by yeast two-hybrid (Ito et al, 2000, 2001; Uetz et al, 2000; Giot et al, 2003; Li et al, 2004) or affinity purification/mass spectrometry assays (Ho et al, 2002; Gavin et al, 2006; Krogan et al, 2006). Figure 7B shows the resulting standard curve that relates enrichment for orthologous interactions to percentage of true physical associations. On the basis of standard curve, we estimated that ∼37–41% of the CCE pairs correspond to true physical associations, somewhat lower than the value by mass spectrometry co-sedimentation but considerably higher than both randomized pairs and pairs derived from only human co-expression data. As physically associated proteins often share similar functional annotation (von Mering et al, 2002), we also created a standard curve based upon the agreement of interaction partners' functional annotation, relating agreement of SwissProt keywords (Wu et al, 2006) to the percentage of true physical associations (Figure 7C). For each control set, we measured the average keyword overlap across 882 SwissProt keywords between the interaction partners. Keyword overlap varied from ∼5% for the true-negative set to ∼42% for the true-positive set. From this curve and measurements of the keyword overlap by the 7000 interactions, we estimate ∼59–68% of the CCE set represent physical associations. Measurements of the percentage overlap of GO ‘biological process' and KEGG pathway annotations result in comparable values (∼50–53%). Finally, legitimately associated proteins should be closer in a gene network (i.e. separated by fewer interactions) than random pairs. For each putative physical association, we calculated the distance between the genes' yeast orthologs in a functional gene network (Lee et al, 2007). We compared the distribution of path lengths to distributions from positive and negative control sets. Figure 7D shows that the interactions from CCE have a path length distribution more similar to the positive control set than the negative set, indicating strong enrichment for true-positive associations among the 7000 interactions. We fit the distribution of path lengths from the 7000 CCE associations as a linear combination of the positive and negative control distributions (as in Deane et al, 2002). The proportions from the best fit (Figure 7D) provide an estimate of the percentage of true physical associations. This approach estimates the CCE set at ∼63±3% true physical associations. Table I summarizes the measurements of physical association, along with comparisons to the randomized and human-only co-expression control sets. Estimates vary only minimally with changes in parameters (e.g. using percentage keyword overlap for Figure 7B versus Jaccard coefficient) or choice of control sets (e.g. employing alternative literature positive control sets for Figures 3 or 7; see Supplementary Table 2). Although individual tests may show some bias, we expect these biases to average out across the five tests; in fact, the estimates are similar across the five tests. These measures demonstrate that CCE associations are, on average, reasonably accurate (54±10% true physical associations) and biologically relevant, are comparable in accuracy to direct large-scale experimental assays (Rual et al, 2005; Stelzl et al, 2005), and are significantly more enriched for physical associations than random controls. To summarize benchmark support for individual CCE associations, we calculated a ‘Binary Interaction Overlap Score (BIOS)' (Stelzl et al, 2005) for each association (Supplementary Figure 10). By this measure, 4354 (62%) of the 7000 associations have at least one line of additional evidence supporting them. Scores are reported in the supporting data file. Detailed evaluation of ribosome biogenesis proteins To experimentally evaluate the quality of hypotheses arising from the CCE associations, and given a statistical enrichment for proteins of ribosome biogenesis (see below), we analyzed proteins predominantly linked to proteins of ribosome biogenesis, a pathway involving several hundred proteins yet still incomplete (Granneman and Baserga, 2004). We chose four proteins with yeast orthologs for direct validation: (i) WBSCR20C, named for Williams–Beuren syndrome chromosome region 20C, shares high sequence similarity with duplicate genes WBSCR20A and WBSCR20B. This gene is deleted in Williams–Beuren syndrome, a multi-system developmental disorder caused by deletion of genes at the 7q11.23 locus (Doll and Grzeschik, 2001). WBSCR20C encodes a conserved Nol1/Nop2/Sun family protein domain and is also a member of the COG0144 protein family, other members of which are tRNA and rRNA cytosine-C5-methylases involved in translation, ribosomal structure, and biogenesis. YNL022C, the yeast ortholog of WBSCR20C, is also uncharacterized. (ii) BCCIP, or ‘BRCA2 and CDKN1A interacting protein', is an evolutionarily conserved nuclear protein with multiple protein interaction domains. This protein may be an important cofactor for BRCA2 in tumor suppression (Lu et al, 2005) and a modulator of CDK2 kinase activity via p21 (Meng et al, 2004). The yeast ortholog of this protein (Bcp1p) is an essential nuclear protein involved in nuclear export of lipid kinase Mss4p (Audhya and Emr, 2003). (iii) EPRS is predicted by sequence to be a multi-functional aminoacyl-tRNA synthetase. Its yeast ortholog YHR020W is essential, with sequence similarity to proline-tRNA ligase, but otherwise uncharacterized. (iv) LYAR is a nucleolar zinc-finger-containing protein (Su et al, 1993) whose yeast ortholog YCR087C-A is nucleolar (Huh et al, 2003), but uncharacterized. We tested the ribosomal processing phenotypes of yeast strains with tetracycline-controlled downregulatable alleles of the genes (Mnaimneh et al, 2004). Two of the strains (TetO7-BCP1 and TetO7-YHR020W) show clear ribosomal processing defects upon downregulation of the genes (Figure 8B and C). From polysome profiles, Bcp1p (corresponding to human protein BCCIP) appears to participate in 60S ribosomal subunit biogenesis; loss of the protein results in the reduction in the 60S peak relative to the 40S peak. YHR020W (corresponding to human protein EPRS) appears to participate in 40S biogenesis, resulting in a decreased 40S/60S ratio when depleted. We also tested each of the four proteins for co-sedimentation with the 40S, 60S, or 80S monoribosomes, which would provide additional support for the proteins' participation in ribosome processing. From crude cell lysates of yeast strains expressing TAP-tagged versions of each protein (Ghaemmaghami et al, 2003), we size-fractionated ribosomal subunits, free ribosomes, and polysomes using sucrose gradients. Three of the proteins (YCR087C-A, YNL022C, and Bcp1p) showed clear association with 40S and 60S subunits (Figure 8A), with Bcp1p and YCR087C-A associated with both 40S and 60S, and YNL022C showing preferential co-sedimentation with the 60S subunits. Mass spectrometry of untagged YHR020W, analyzing yeast lysate with the approach of Figure 4 (data not shown), indicates that YHR020W also co-sediments with 60S ribosomal subunits (Z Li and EM Marcotte, unpublished data). The Bcp1p-TAP co-sedimentation is less definitive than the controls; however, the polysome profile of the TAP-tagged Bcp1p strain indicates that the TAP tag itself disrupts Bcp1p activity (Figure 8B), causing a 60S ribosomal biogenesis defect and definitively implicating the protein in this process. The human BCCIP protein was also found by mass spectrometry to co-sediment with free cytosolic 40S and 60S ribosomal subunits (Supplementary Figure 4), raising the possibility of a role in ribosome recycling or nuclear export. All four genes assayed could therefore be implicated in ribosomal biogenesis. Discussion Characteristics of the newly mapped associations We have described the prediction of 7000 human protein physical associations from indirect transcriptional evidence, as well as measurement of overall error rates and validation of specific associations. We further examined the associations for novelty, functional bias, and evidence for stable protein complexes. First, we compared the predicted interaction set directly to the existing human protein interaction data sets. Roughly 20% of the CCE associations can be directly verified from previously known interactions, while ∼80% are new. Our analysis bears some relation to one reported by Stuart et al, which analyzed CCE, although not for the purpose of discovering physical interactions. However, we obtain a largely non-overlapping set of associations, sharing only 12% of associations (Supplementary Table 1). Differences arise primarily because we are explicitly learning physical associations using a supervised training framework; other differences include the choice of expression data, the methods for defining orthologs, the criterion used to define co-expression (we set a statistical significance threshold on the correlation coefficient; Stuart et al use correlation coefficients >0.2), and our removal of potential cross-hybridization artifacts, all of which contribute to producing largely distinct sets of associations. Only three CCE interactions are shared with large-scale yeast two-hybrid analyses of human proteins (Rual et al, 2005; Stelzl et al, 2005), 15 with affinity purification/mass spectrometry analysis (Ewing et al, 2007), 195 with a previous computational analysis (Rhodes et al, 2005), and 211 with interactions inferred from other organisms in the OPHID database (Brown and Jurisica, 2005). These comparisons are summarized in Supplementary Table 1. In all, 5589 of the 7000 associations predicted in this analysis were not identified in previous high-throughput human protein interaction screens. Besides simply being novel associations, 80% of the interaction partners (66% of annotated interaction partners) share neither KEGG nor GO annotations. While this extent of annotation sharing across the set of partners is sufficient to imply high confidence associations (Table I), these results indicate that the inferred associations extend beyond trivial identification of new associations among proteins already known to be in the same pathways. We examined the functions of the 2348 proteins in the set of 7000 associations (Supplementary Figure 7), using for this purpose the proteins' KOG annotations (Tatusov et al, 2003). We find the dominant class of proteins to be those for whom only general function is known (224 proteins); followed by 180 proteins participating in post-translational modifications, protein turnover, and chaperones; 163 in signal transduction; 141 in translation, ribosomal structure, and biogenesis; 141 in transcription; 117 of RNA processing and modification; and 87 of unknown function. Therefore, the proteins are not dominated by a single structure (e.g. the ribosome), but are generally informative for major cellular systems and uncharacterized proteins. Nonetheless, some specific functional biases occur among the CCE proteins (Supplementary Figures 8 and 9), mostly notable a statistical enrichment for proteins of membrane-bound organelles (e.g. nucleus, mitochondria, nucleolus, etc.), presumably reflecting evolutionary conservation of these proteins and their regulation, favoring detection by the CCE method. Likewise, proteins of RNA metabolism are enriched, especially ribosome biogenesis, with accompanying enrichment for nucleotide-binding protein domains. The overall trends among the proteins can be seen in a plot of the CCE associations, clustered by the spectrum of associations per protein (Figure 9), which shows that although several major clusters exist, many associations are binary protein pairs that are not otherwise seen to exist in larger assemblies, and thus lie far from the diagonal of the clustergram. This clustering, along with the 1411 associations overlapping other data sets, provides some insight into the nature of the CCE associations. Both direct interactions, especially among members of larger complexes, as well as co-complex physical associations are observed. For example, interactions are observed between alpha- and beta-tubulin, which assemble into a heterodimer; SNRPE and SNRPF, known to bind directly in the core complex of spliceosomal U1, U2, U4, and U5 snRNPs (Camasses et al, 1998); and the E2 and E3 subunits of pyruvate dehydrogenase, which interact directly. A comparison of the 7000 CCE associations with experimentally determined protein interactions among components of the 20S proteasome, as determined from the X-ray crystal structure (Groll et al, 1997), reveals four interactions between proteins that directly contact each other in the proteasome (PSMA2–PSMA6, PSMB3–PSMB2, PSMA4–PSMB2, and PSMA1–PSMB1), and nine interactions between proteins that assemble into the same physical complex, but do not directly contact each other (PSMA3–PSMA4, PSMA6–PSMB2, PSMA4–PSMA6, PSMA6–PSMB3, PSMA2–PSMB1, PSMA2–PSMA1, PSMA2–PSMB2, PSMA6–PSMA5, and PSMA5–PSMB2). Therefore, as expected, both direct binding and co-complex interactions can be found among the 7000 associations. In fact, the CCE pairs are strongly statistically enriched for co-complex associations typical of affinity purification/mass spectrometry interaction assays. If we consider only the subset of 2138 CCE pairs (out of the 7000) in which both proteins have yeast orthologs, 118 of these can be verified by the MIPS database as belonging to the same yeast complex (using the ‘hand-curated' set of protein complexes; Guldener et al, 2005). This value is 21 standard deviations (P<10−98) above the mean of random trials: randomizing the 7000 interactions and repeating the comparison gives a mean of 23±4.5 confirmed interactions for N=100 random trials. Similar enrichment can be seen by comparison with yeast protein complexes defined by affinity purification/mass spectrometry (Gavin et al, 2006; Krogan et al, 2006): 74 of the 2138 CCE pairs can be confirmed, 8 standard deviations (P<10−15) above the mean of random trials (29±5.5, N=100). Likewise, 392 of the 2138 CCE pairs can be confirmed by comparison to the full-matrix form of these data (i.e. considering both bait–prey and prey–prey interactions), 10 standard deviations (P<10−25) above the mean of random trials (214±17, N=100). Finally, we looked for organismal bias among the CCE pairs, examining which model organism contributed the top LLR score for each interaction (Supplementary Table 3). The most associations were contributed from the comparison of human and C. elegans expression, accounting for 2949 of the 7000 associations, and the least (158) from mouse. The low number contributed by comparison to mouse may suggest the importance of employing more distant orthologs, especially to non-mammalian animals, in identifying interactions by this approach, but more probably stems from characteristics of the data employed, such as the smaller number of mouse microarray experiments analyzed (Supplementary Table 4). One interesting aspect of the CCE assay is that it intrinsically samples all pairs of genes that are measured on the DNA microarrays. This has the effect of increasing the numbers of proteins for which interactions are observed, and thereby decreasing the number of interactions per protein (7000 interactions for 2348 proteins ∼3 interactions per protein, somewhat lower than the 5–15 interactions per protein observed in other data sets (Ramani et al, 2005)). Limitations, false positives, and potential improvements Given the derivation of CCE pairs from transcriptional evidence, there are important features and limitations to note. First, strong co-expression tends to coincide with stable, rather than transient, physical association (Jansen et al, 2002), and we expect CCE pairs to reflect this trend, with a correspondingly higher false-negative rate for transient interactions. Second, based on our measured error rates, there are still appreciable false-positive associations, although the false-positive and -negative error rates are comparable to the only direct experimental approaches—yeast two-hybrid assays and mass spectrometry of cloned, epitope-tagged human proteins—that have been applied to map physical associations on this scale. However, CCE false positives have unusual properties. As the CCE pairs were the highest scoring (top 0.1%) of >5 million tested gene pairs, the association partners are strongly co-regulated in an evolutionarily conserved manner, and thus are highly likely to function together, even if not physically associated. Finally, algorithmic improvements, such as better orthology assignment and alternative supervised learning frameworks, and application to additional DNA microarray data, e.g. tissue- and cell-type-specific data to learn tissue- and cell-type-specific associations, are certain to reveal new associations when applied in the general framework we have described. Thus, we expect new CCE associations can be identified by modifications to this method. Similarly, the mass spectrometry data used to test the CCE associations have some important features and limitations. Primarily, co-sedimentation alone is not proof of physical association—it is possible for unrelated complexes to co-sediment—as reflected in the measured true-positive and false-negative rates for associations inferred solely from these data. These sedimentation-derived associations should thus not be viewed as standalone. However, as a benchmark applied in the manner we present (e.g. analyzed in aggregate form), or when considered in combination with other data, such as incorporated into the BIOS scores of the CCE associations, we find the mass spectrometry data to be extremely valuable. We suggest that benchmarks of this sort could be of great utility for evaluating physical complexes determined by other methods, and could be generally adopted for measuring assay accuracy. Conclusions The scale of the human interactome appears to be beyond any individual technique; a combination of complementary approaches will be needed to map the complete human protein–protein interaction network. Although current methods for mapping interactions focus largely on direct experimental observations, sufficient functional genomics data exist that physical protein associations can also be indirectly identified from these data. We demonstrate that these approaches can be comparable in scale and quality, both in terms of false-positive and false-negative rates, to the current largest scale experimental screens. Finally, as CCE-based physical protein association mapping is based on conserved in vivo phenomena, this approach is likely to specifically discover associations relevant to in vivo biology. Materials and methods Mapping of orthologs Orthologs were obtained from the InParanoid database (Remm et al, 2001) as SwissProt identifiers for human proteins and their orthologs from five other organisms (A. thaliana, C. elegans, D. melanogaster, M. musculus and S. cerevisiae). Using ID-Serve (http://bioinformatics.icmb.utexas.edu/idserve) and organism-specific databases, the SwissProt identifier for each gene was mapped to alternate identifiers: LocusLink identifiers (human), common names (M. musculus), WormBase identifiers (C. elegans), Locus codes (A. thaliana), Flybase gene identifiers (D. melanogaster), and standardized gene names (S. cerevisiae). Supplementary Table 5 lists the numbers of orthologous genes analyzed. mRNA expression data All mRNA expression data (Supplementary Table 4) were obtained from the Stanford Microarray Database (Ball et al, 2005). It has previously been shown that extraction of co-expression relationships is improved by restricting comparisons to similar conditions and experiments (Lee et al, 2004a, 2004b; Segal et al, 2004). We therefore divided the available 1922 human DNA microarray experiments into 11 categories of experiments, as assigned by the Stanford Microarray Database, and restricted comparisons to experiments in the same category. Expression data for other organisms were treated as single categories. Each of the microarray expression vectors was mean centered (row and column) and normalized before carrying out correlation analysis. Calculation of co-expression For each pair of human genes, as well as for their corresponding orthologs, the Pearson correlation coefficient was computed between the mRNA expression vectors. For each gene pair, this gives 11 measurements of correlation corresponding to the 11 categories of human expression data sets and up to 5 for the correlation between the orthologs in the other organisms. Paralogs (as defined by InParanoid) were excluded from being compared to each other, as they tend to have similar expression profiles and thus high correlation, which we empirically observe to substantially increase the false-positive rate. The significance of each correlation was computed based on t-test statistics as where r is the minimum significant correlation for n values in the two expression vectors being compared and t is the t-test value at a probability of P⩽0.01 from a t-test table. Only statistically significant correlation coefficients were retained, thereby accounting for variability in the sparseness of expression vectors. For example, using expression vectors of 100 experiments with only 50 data points available for both genes being compared, the absolute value of correlation must be >0.36 for the comparison to be statistically significant at P⩽0.01. Removal of cross-hybridization artifacts Cross-hybridization occurs when an mRNA probe binds to a non-cognate spot on the microarray instead of its perfect complement spot. This creates both false positives (due to additional signal at incorrect positions on the array) and false negatives (due to reduced signal in correct positions). Although cross-hybridization is well established in spotted cDNA-based DNA microarray experiments (Kane et al, 2000; Murray et al, 2001; Xu et al, 2001), there are no universal standards for filtering such effects. In this analysis, we expected that cross-hybridizing gene pairs would appear to have similar expression patterns and therefore contribute false positives to our analysis. To filter out these potentially spurious interactions arising from cross-hybridization, we established a threshold for excluding cross-hybridization based upon analysis of the hybridization of four yeast genes (YPL274W, YLR467W, YIR039C, and YKL224) to their homologs on a yeast DNA microarray. The four genes were chosen such that BLAST-based comparisons of the genes' DNA sequences to other genes in the yeast genome yielded hits with percent identities to the query sequence in the range of 50–100% and BLAST E-values ⩽10−4. The four query genes were amplified using standard PCR techniques and primers to flanking DNA, labeled with Cy5, mixed with Cy3-labeled reference DNA (Carlson, 2002), and hybridized to a yeast cDNA microarray containing ∼12 000 spots comprising all the yeast genes and intergenic regions (Carlson, 2002; Hahn et al, 2004; Kim and Iyer, 2004). Standard microarray analysis was carried out to quantify hybridization strength as the mean of ratios of Cy5/Cy3 fluorescence intensities across spots. By plotting hybridization strength against the DNA sequence identity of the genes (Supplementary Figure 2), we identified an operational threshold of BLAST E-value ⩽10−4 and DNA sequence identity ⩾70% within the aligned regions. Gene pairs that exceed this threshold (with either the human or model organism gene pair DNA sequences) were likely to cross-hybridize and were excluded from further analysis. This filter removes 47 145 protein pairs from the plant–human analysis, 37 519 from the worm–human, 26 724 from the fly–human, 39 286 from the mouse–human, and 2193 from the yeast–human analysis. This filtration preferentially removes many false-positive interactions, as the average expression correlation of the filtered pairs was significantly higher than for the remaining pairs (e.g. the average expression correlation in the human–plant analysis was 0.28, while the average for the filtered pairs was 0.56), with the maximum expression correlation among the removed pairs equal to 1.0 for all comparisons. Training to extract physical protein associations We used the 31 609 human protein interactions from Ramani et al (2005) as the physical association benchmark. The associations were randomly separated into testing and training data sets (15 810 and 15 799 associations, respectively). For each of the five human gene pair/ortholog gene pair sets, the maximum expression correlation of the human genes from the 11 data sets was plotted along the x axis and the correlation of the orthologous genes plotted along the y axis (as in Figure 2). The fraction of gene pairs that showed a particular expression pattern was measured in bins of 0.1 × 0.1 units. Two-dimensional histograms were calculated for interacting proteins and for non-interacting proteins in the training set. The logarithm of the ratio of the histograms at a given position in the plot, corrected by the background likelihood of physical associations in the training set, gives the log likelihood estimate of physical association conditioned on the degree of co-expression of the human genes and their orthologs in that organism. To minimize possible errors due to orthology assignments, we further considered only counts in the upper right-hand quadrant of each analysis, corresponding to gene pairs for which the human and other organismal experiments describe similar co-expression trends. Protein pairs outside of the training set were then assigned log likelihood scores according to their expression patterns in these data sets. Similar analyses were performed for associations derived from comparison of human expression data with each of the four other organism-specific data sets, associating the maximum score from these five analyses as each protein pairs' estimated likelihood of associating physically. (The maximum score outperformed the naïve Bayes sum of scores, suggesting that the five scores are not independent.) The 7000 top-scoring associations are listed in Supplementary information. The human-only co-expression control set was generated by considering only the human DNA microarray data, ignoring contributions from other organisms and lifting the requirement for each member of a gene pair to have orthologs in the same second organism. Putative associations were identified as for the CCE case, but instead using the log likelihood framework to relate the correlation coefficients across only the human DNA microarray experiments to the likelihood of physically associating. All other calculations were performed identically to the CCE case, including calculation of correlation coefficients, significance testing of correlations, calculation of likelihood values, selection of priors, and filtration for cross-hybridization. Testing for enrichment of known physical associations We measured enrichment for known physical associations using the independent test set of 15 810 physical associations and the same LLR framework used to initially derive the CCE associations. The 15 810 associations formed the positive test set; the negative test set was defined as all pairs of proteins chosen from the 15 810 associations set, omitting the 15 810 associations themselves. The prior odds ratio of interacting (P(I)/P(∼I)) equaled the ratio of positive to negative test set examples (0.00085). For each query association network being tested (or for a given bin of 1000 associations selected from a rank-ordered list), we measured the fraction of query set associations shared with the positive test set (P(I∣D)), as well as the fraction shared with the negative test set (P(∼I∣D)). The posterior odds ratio was calculated as P(I∣D)/P(∼I∣D), and the LLR calculated as indicated in the main text, equal to the posterior odds ratio divided by the prior odds ratio. For the purposes of Figure 3A, the log likelihood was calculated in a cumulative manner (i.e. aggregating successive bins of 1000 associations for analysis.). Testing for functional similarity We measured functional similarity of interacting protein pairs by using the gene annotation information obtained from GO (Ashburner et al, 2000) process level 8 annotation and KEGG pathway annotation (Kanehisa et al, 2004). These databases provide specific pathway and biological process annotations for 7390 human genes, assigning them into 155 KEGG pathways (at the lowest level of KEGG) and 1356 GO pathways (at level 8 of the GO biological process annotation). Interactions were first rank-ordered by confidence scores. For each successive bin of 1000 interactions, the functional similarity was calculated in a cumulative manner by counting the number of pairs in that bin or previous bins that shared a functional annotation, dividing this by the number of pairs that did not share functional annotation, and correcting by the prior probability of annotated pairs sharing annotation (0.0589). Construction of standard curves for estimating percentages of physical associations Standard curves were constructed as described in the main text. Positive control sets for Figure 7B and C were selected from the hand-curated protein complex assignments of Reactome (Joshi-Tope et al, 2005). For the analysis of Figure 7B, we restricted the analysis to the portion of each data set for which both interacting proteins have orthologs among the yeast, worm, or fly proteins sampled by the benchmark assays (i.e. considering only the subspace of interactions spanned by the assay bait–prey pairs). For the standard curves of both Figure 7B and C, the derived percentages of physical associations do not strongly depend upon the sizes of the data sets or control sets, only upon their tendencies to share orthologous interactions or functional annotations (data not shown). Ranges of accuracies were derived directly from the standard curve (i.e. as empirically measured from replicate analysis of control mixtures of true- and false-positive interactions). For the linear mixture model of Figure 7D, positive control associations were taken from Joshi-Tope et al (2005), only considering genes with yeast orthologs, and negative control associations taken as pairs of human genes from the positive control set that have yeast orthologs but do not have recorded interactions. To minimize possible circularity, we removed all functional linkages from the yeast network that were derived only from mRNA co-expression data. The variances associated with accuracy estimates in Figure 7D were derived from 10 replicate analyses of mixtures of known proportions of true- and false-positive interactions (Supplementary Figure 6). Binary interaction overlap score To further assign confidence to each association, we have adopted the BIOS of Stelzl et al (2005): based upon the benchmark sets (Table I), we assign each association +1 if the protein pair is observed in the physical interaction benchmark, +1 for sharing GO/KEGG keywords, +1 for sharing SwissProt keywords, +1 for sharing KOG annotation, +1 for being observed in the orthologous interaction benchmark, +1 for having a correlation coefficient >0.4 in the mass spectrometry elution profile experiments, and +1 for having yeast orthologs that are either directly connected or one link separated in the yeast functional network benchmark (with expression-only and orthology-derived links omitted). Each association is thus scored from 0 to 7 based on additional support for that association; the BIOS scores generally correlate with the LLR scores (Supplementary Figure 10) and are reported in the supporting data file. Human cell culture and mass spectrometry HeLa S3 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C with 5% CO2. At about 80% confluency, cells were treated with 100 μg/ml emetine for 10 min and harvested by scraping. Cells were centrifuged at 500 g for 10 min, washed three times with cold PBS buffer, and resuspended in five packed cell volumes of cold lysis buffer (10 mM Tris pH 7.4, 20 mM KCl, 5 mM MgCl2). After swelling on ice for 10 min, cells were centrifuged at 500 g for 10 min and resuspended in one packed cell volume cold lysis buffer supplemented with 1 × protease inhibitor cocktail (Roche) and 100 μg/ml emetine. After lysing the cells with a dounce homogenizer, nuclei were collected by centrifuging at 1000 g for 10 min. The supernatant was centrifuged at 15 000 g for 10 min to obtain the cytosolic fraction. Nuclei were suspended in lysis buffer and lysed by sonication, collecting the clarified supernatant after centrifugation at 15 000 g for 10 min. The cytosolic and the nuclear fractions were each loaded onto continuous 7–47% sucrose gradients in lysis buffer. After a 2.5-h spin at 40 000 r.p.m. in a Beckman SW40 rotor, the sucrose gradient was fractionated using an ISCO gradient fractionation system. Proteins from each fraction were precipitated with 10% cold trichloroacetic acid (TCA) and washed with 100% cold acetone. The protein pellets were suspended in 100 mM pH 8.0 Tris buffer and digested with sequencing grade trypsin (Sigma). For each fraction, tryptic peptides were loaded onto a reverse-phase C18 column and washed with 95% water, 5% acetonitrile, and 0.1% formic acid. Peptides were eluted with a 240-min gradient from 5 to 40% acetonitrile and analyzed online with a nanoelectrospray ionization (300 nl/min flow rate) LTQ-Orbitrap hybrid mass spectrometer (Thermo Electron) using data-dependent precursor ion selection. Each parent ion mass spectrum (MS) was analyzed at high resolution (100 000) with the Orbitrap; the top seven MS peaks were fragmented by helium collision-induced dissociation at 35 eV, analyzing the resulting MS/MS spectra with the LTQ. Approximately 35 000 MS/MS spectra were collected per fraction. Spectra were searched against the set of NCBI human protein sequences using TurboSequest (Bioworks v.3.2, Thermo Electron). Proteins from each fraction were identified at a 5% false detection rate using Peptide/ProteinProphet (Keller et al, 2002; Nesvizhskii et al, 2003). The spectral count (number of total observations of MS/MS spectra from a given protein in a given fraction) was used as an estimate of protein abundance (Liu et al, 2004), dividing the spectral count of a protein ( × 10 000) by the sum of spectral counts for all proteins identified in that fraction. Protein elution profiles are provided as Supplementary information. Yeast media and strains All yeast strains were cultured in YPD (1% yeast extract, 2% peptone, and 2% dextrose) at 30°C. Tetracycline promoter-controlled essential gene haploid MATa strains (Mnaimneh et al, 2004) and TAP-tagged haploid MATa strains (Ghaemmaghami et al, 2003) were obtained from Open Biosystems. Polysome profile analysis All yeast strains were cultured to OD600 0.3–0.5. For tetracycline promoter-controlled alleles, overnight cultures were diluted to OD600 0.01, 10 μg/ml doxycycline (Fisher Scientific) was added into the media, and cells were grown to OD600 0.3–0.5. Cycloheximide (100 μg/ml) (Sigma) was added to each culture. Cultures were immediately cooled with ice, and all subsequent steps were performed on ice or at 4°C. Each cell pellet was washed once with lysis buffer (20 mM Tris pH 7.4, 20 mM KCl, 5 mM MgCl2, 100 μg/ml cycloheximide, 12 mM β-mercaptoethanol, 2 μg/ml leupeptin, 2 μg/ml aprotinin, 1 μg/ml bestatin, and 1 μg/ml pepstatin A) without protease inhibitors (MP Biomedicals Inc.). The cells were pelleted, resuspended in one volume lysis buffer, and lysed with glass beads. Crude lysates were centrifuged at 15 000 g for 10 min. Fifteen OD260 units of each supernatant were loaded onto continuous 12 ml 7–47% sucrose gradients in polysome lysis buffer without protease inhibitors, as in Baim et al (1985). After a 2.5-h spin at 40 000 r.p.m. in a Beckman SW40 rotor, the sucrose gradient was fractionated and absorbance at 254 nm was measured. For TAP-tagged strains, fractions were collected, and proteins were precipitated with 10% cold TCA and washed with 100% cold acetone. Immunoblotting Precipitated proteins were resuspended in 20 μl Laemmli buffer and 2 μl of each sample was deposited onto nitrocellulose membrane. TAP-tagged proteins were detected with PAP antibody (Rockland Immunochemicals Inc.) and chemiluminescence (ECL; Amersham Biosciences). Supplementary Material Supplementary Figures Supplementary Tables Supplementary Information 1 Supplementary Information 2

Environ_Health_Perspect-115-3-1849909

Skin Exposure to Isocyanates: Reasons for Concern

Objective Isocyanates (di- and poly-), important chemicals used worldwide to produce polyurethane products, are a leading cause of occupational asthma. Respiratory exposures have been reduced through improved hygiene controls and the use of less-volatile isocyanates. Yet isocyanate asthma continues to occur, not uncommonly in settings with minimal inhalation exposure but opportunity for skin exposure. In this review we evaluate the potential role of skin exposure in the development of isocyanate asthma. Isocyanates, a group of reactive chemicals [with the functional group N = C = O (NCO)] used extensively in the production of numerous polyurethane foams, coatings, and a wide array of consumer products, are a major cause of occupational asthma worldwide. The polyurethane industry has expanded dramatically, along with the number of workers and consumers at risk for exposure. Inhalation has long been considered the primary route of isocyanate exposure, induction of sensitization, and asthma; research, practice, and regulation have focused almost exclusively on understanding and preventing inhalation exposures. Airborne isocyanate exposures have been reduced through improved controls and use of less-volatile isocyanates. Yet isocyanate asthma continues to occur, not uncommonly in work settings where measured isocyanate respiratory exposures are very low or nondetectable, but where there is opportunity for skin exposure. It has been > 25 years since Karol et al. (1981) demonstrated in guinea pigs that skin contact with isocyanates could lead to sensitization and subsequent asthmatic responses following inhalation exposure. However, knowledge and awareness remain limited regarding the potential for isocyanate skin exposure to contribute to the development of isocyanate asthma. For example, the literature on occupational asthma rarely mentions isocyanate skin exposure as a potential risk factor or target for prevention (Nicholson et al. 2005; Tarlo and Liss 2005). Over the past several years there has been a growing, but largely unrecognized, collection of animal, industrial hygiene, clinical, and epidemiologic data related to isocyanate skin exposure and its role in the development of isocyanate sensitization and asthma. Our primary purpose in this article is to review and synthesize this multi-disciplinary literature to address several key unresolved issues, including the extent of isocyanate skin exposures in the workplace, the effectiveness of personal protective equipment, and most importantly, whether human skin exposure contributes to the development of isocyanate asthma. The findings may be relevant to the larger issue of the role of skin as an important underrecognized site of exposure and sensitization for other environmental allergens. Methods Definition of terms The terms “skin exposure” and “dermal exposure” are used interchangeably to indicate exposure to the outermost layer of the epidermis or epicutaneous exposure, as is commonly done in the occupational and environmental literature. “Isocyanates” refers to diisocyanate monomers (two NCO groups) and their related polyisocyanates, which have similar health effects (Bello et al. 2004). The term “sensitization” can generate misunderstanding. “Sensitization” generally refers to priming of the immune system in response to a specific non-self antigen, a condition that involves immune memory, typically antigen-specific T cells and/or antibodies. Subsequent reexposure to the antigen can result in an immunopathologic adverse reaction, such as a Th2 (T-helper 2)-type acute allergic reaction or asthmatic response, or contact hypersensitivity-type reaction, such as allergic contact dermatitis (Sheaarer and Fleisher 2003). Others consider sensitization any immunologic memory of exposure, regardless of its pathogenic potential (i.e., a specific IgG response) (Abbas et al. 2000). Skin or respiratory sensitization typically refers to the route of exposure that results in systemic sensitization, rather than localized immune responses at those sites. Literature search We have been active in the field of isocyanate research and have collected > 800 published and unpublished articles and documents from 1951 to the present; these articles and documents are related to isocyanates and span many disciplines. In addition, we performed computerized searches of the literature on Medline [1966 to the present (National Library of Medicine, Bethesda, MD)], National Institute for Occupational Safety and Health (NIOSH), Occupational Safety and Health Administration (OSHA) and U.S. Environmental Protection Agency (EPA) databases and Google (google.com) using the key words “isocyanate,” “diisocyanate,” “asthma,” “sensitization,” “exposure,” “dermal,” “skin,” “occupational,” “methylene diphenyl diisocyanate,” and other synonymous terms. Additional articles were identified from the reference lists of the selected relevant articles. We reviewed primarily English-language articles, as well as selected articles in German, Danish, and French. Human and animal articles that addressed isocyanate skin exposure, sensitization, and health effects were retained for further analysis. We also included clinical, epidemiologic, and biomarker studies and case reports that mentioned skin as a potential route of exposure or had low isocyanate air levels based on exposure data or work processes. Results and Discussion Health effects of isocyanate exposure Isocyanates are considered potent respiratory allergens. Isocyanate asthma is the major health problem in isocyanate-exposed workers, affecting approximately 1–25% of the exposed population. The most important risk factor is isocyanate exposure, but the exposure characteristics and host factors involved remain unclear (Bernstein 1996; Wisnewski et al. 2006). Isocyanates can also cause hypersensitivity pneumonitis, contact dermatitis, and rhinitis, but these outcomes are less commonly reported (Baur et al. 1994; Musk et al. 1988). Clinically isocyanate asthma presents similar to other types of allergic Th2-like asthma: Isocyanate asthma typically develops after repeated exposure for months to years, during which time sensitization to isocyanates is presumed to occur. Once sensitized, extremely low respiratory levels of isocyanate can elicit asthmatic responses. However, unlike for typical high molecular weight (MW) occupational allergens or environmental aeroallergens, research has failed to identify an isocyanate-specific immune response in most isocyanate asthmatics that indicates isocyanate Th2-like sensitization, such as a radioallergosorbent test (RAST) or skin prick test, thus hindering diagnosis. This difference may reflect involvement of non-IgE mechanisms in isocyanate asthma pathogenesis, and/or may be related to the NCO functional group common to all isocyanates that renders them ideal cross-linking agents (Wisnewski et al. 2006). Unlike high-MW allergens, isocyanates can react with amino (NH2, NH), hydroxyl, and sulfhydryl groups of various proteins and peptides, including albumin and keratin, to form a number of different hapten–protein complexes or antigens (Wisnewski et al. 1999, 2000). Human isocyanate skin-patch testing has been used to confirm sensitization in the uncommon person who develops contact hypersensitivity (allergic contact dermatitis) due to isocyanate skin exposure, but has not been helpful for detecting the Th2-like sensitization presumed to lead to isocyanate asthma (Kanerva et al. 2001). The mechanisms by which isocyanates cause asthma remain poorly defined, and lack of a good immunologic marker and unclear dose–response relationships have hindered diagnosis and prevention (Wisnewski et al. 2006). The primary focus here is on the potential for skin exposure to contribute to the development of isocyanate asthma rather than pathogenic mechanisms. However, a better understanding of the role of skin exposure may help address these key problems. Several observations suggest that skin may also be an important site of exposure and sensitization. Isocyanate respiratory exposure alone, without any skin exposure, seems unlikely in most work settings. Isocyanate asthma occurs in settings with minimal documented respiratory exposures but clear potential for skin exposure, and splashes and spills have been reported by workers who subsequently develop isocyanate asthma (Bernstein et al. 1993; Lenaerts-Langanke 1992; Liss et al. 1988; Nemery and Lenaerts 1993; Zammit-Tabona et al. 1983). Workplace and environmental isocyanate exposures The chemical structures and important physicochemical properties of selected isocyanates are shown in Figure 1 and Table 1. The major commercial isocyanates are methylene diphenyl diisocyanate (MDI), toluene diisocyanate (TDI), and their nonvolatile polymeric forms pMDI and pTDI, followed by polymeric hexamethylene diisocyanate (pHDI) and isophorone diisocyanate (pIPDI). Isocyanates are reactive chemicals used extensively to make numerous polyurethane and other commercial products, such as poly-urethane foams, adhesives, and coatings. They are found in a wide range of industries, from construction to medical care. The increasing use of nonvolatile polyisocyanates has raised issues related to their measurement, exposure metrics, and regulation (Bello et al. 2004). Occupational exposures to isocyanates occur primarily in the many end-use settings, as well as in primary production facilities where exposures are generally better controlled. The total number of workers currently exposed to isocyanates is not known. NIOSH’s estimate of 280,000 U.S. workers exposed or potentially exposed to isocyanates (NIOSH 1996) is undoubtedly higher today, given industry growth and new applications. Most commercial isocyanate products are complex isocyanate mixtures of variable molecular mass, volatility, and isocyanate content. Workplace exposures can occur in the form of isocyanate vapors, aerosols, or both, depending on the isocyanate type as well as the application method and other factors. Isocyanates are commonly mixed with various solvents, polyols, and other substances, such as catalysts and blowing agents, which may affect isocyanate reactivity, skin absorption, and health effects. Although data are limited, environmental exposures to unreacted isocyanates are also possible. A number of consumer products contain unreacted isocyanates, such as glues, polyurethane coatings, and foam insulation; domestic use of such products on rare occasion has been reported to trigger asthma symptoms in individuals previously sensitized (Carroll et al. 1976; Dietemann-Molard et al. 1991; Peters and Murphy 1971). Based on patch testing, a few cases of allergic contact dermatitis have been reported with the use of consumer products made from isocyanates and polyurethanes (Alomar 1986; Morgan and Haworth 2003; Vilaplana et al. 1987). Certain biomedical products, such as standard orthopedic casting material, may be a potential source of skin exposure to unreacted isocyanates for patients and cast technicians (Legris et al. 1995). Polyurethane foams and packaging have been reported to contain very small amounts of unreacted isocyanates (Gagne et al. 2003; Krone et al. 2003), but it is unclear whether these products have any potential to result in skin exposure. Environmental isocyanate exposures can also potentially occur from the release of isocyanates into the environment from primary or secondary production facilities (Darcey et al. 2002; De Zotti et al. 2000; Orloff et al. 1998), during the transport or storage of isocyanates or polyurethanes (Allport et al. 2003), or during the thermal degradation of overheated polyurethane products (Boutin et al. 2006), but these exposures are rarely reported. Isocyanate releases from end-users, such as auto body shops located in or near residential neighborhoods, are also possible but rarely reported. It has been hypothesized that environmental skin exposure to polyurethanes in childhood has contributed to the increased prevalence of childhood asthma (Krone and Klingner 2005), but human isocyanate exposure from such products has not been documented, and there are numerous other likely causative factors. The primary isocyanate exposure routes are through the respiratory tract and the skin. Historically, the focus has been on inhalation exposures. Increased use of less-volatile MDI and polymeric isocyanates, as well as improved hygiene practices, have resulted in reductions in inhalation exposures to volatile monomer (Bello et al. 2004), thus potentially increasing the relative importance of skin exposure. Isocyanate skin exposure could contribute a significant part of the total body burden. For example, 1% skin absorption of a small MDI droplet (10 mg) would deliver a dose approximately 4.5-fold (450%) higher than the inhalation exposure at the current short-term UK occupational exposure limit (15-min, 70 μg NCO/m3) or approximately 50% of the corresponding 8-hr (20 μg NCO/m3) standard, assuming 100% lung retention and a ventilation rate of 7 L/min (Bello et al. 2004). Measuring isocyanate skin exposure Quantification of isocyanate skin exposure is important for research, prevention, and control. Assessment of skin exposure, in general, is much less developed than that of inhalation exposure. Skin exposure sampling methods typically are nonstandardized, have undergone limited validation, and can be technically challenging. Isocyanate skin sampling is further complicated by several factors, including the reactivity of NCO groups toward skin proteins, water, or other compounds, and the complexity of most isocyanate exposures. Biomarkers such as urinary metabolites, if available, could potentially be used to assess internal dose, but would not distinguish between skin and respiratory exposure. Techniques that have been used to detect isocyanate skin exposure include SWYPE pads (Colormetric Laboratories Inc., Des Plains, IL) (Liu et al. 2000), wipes (Bello et al., in press), and tape stripping (Fent et al. 2006). These methods rely on removal of isocyanates from the skin—usually a period of time after initial exposure, and can underestimate exposure as a result of losses due to absorption, chemical reactions, and/or poor removal efficiency (Bello et al. 2005; Wester et al. 1999). Techniques that quantitate isocyanate deposited on the skin, such as reagent-impregnated patches, may overcome some of these limitations. Recently, extraction of isocyanates from contaminated gloves was used to measure hand skin exposure to isocyanates (Pronk et al. 2006). Application of these skin exposure methods in the workplace has been limited, and findings have not been compared or validated. Workplace isocyanate skin exposure assessment is further complicated by the frequently sporadic nature of such exposures. Workplace isocyanate skin exposure Numerous isocyanate end uses, such as spraying and application of foams and adhesives, provide opportunity for isocyanate skin exposure from deposition of aerosols and/or absorption of vapors. Typical workplace isocyanate exposure levels are not irritating and give few warning signs, and skin protective equipment may not be worn, even when respiratory protection is used (NIOSH 1999). Skin exposure may result from direct contact of unprotected skin or the failure of personal protective equipment, such as gloves. Opportunities for isocyanate skin exposure, such as spills, cleanup, and contact with contaminated equipment, are well known to workers and field researchers. For example, NIOSH Health Hazard Evaluations of a variety of work settings have described workers with skin contact to isocyanates or uncured polyurethane products (NIOSH 1998, 1999). Isocyanate products that are not fully cured are another potential source of isocyanate skin exposure (Bello et al., in press). It is commonly believed that isocyanate-containing products polymerize rapidly, and once the product appears hardened, no unbound isocyanate species remain on the surface. However, there are few published data confirming this. A recent study demonstrated that curing proceeded more slowly than expected, with unbound isocyanate species detected on painted surfaces for prolonged periods of time (days to weeks) (Bello et al., in press). Thus, handling recently cured isocyanate products could be a source of isocyanate skin exposure. In addition, release of free isocyanates has been reported after heating of cured isocyanate products, as can occur when grinding, cutting, or sanding such products (Alliance for the Polyurethanes Industry 2005; Boutin et al. 2006; Littorin et al. 2002). Such tasks could represent another possible source of isocyanate skin exposure. In spite of observational documentation of workplace isocyanate skin exposures, published data documenting such exposure are surprisingly limited. Isocyanate skin exposure has been documented qualitatively with colorimetric techniques in several work settings (Liu et al. 2000; NIOSH 1998). Recent studies have begun to quantitate such exposures. Fent et al. (2006) detected HDI monomer on tape strips from an auto-body painter’s unprotected skin. Pronk et al. (2006) recently reported isocyanate exposure under gloves on both hands of auto body and industrial spray painters, with the highest median hand exposure detected during paint mixing (207 and 63 μg NCO, respectively). Effectiveness of personal protective equipment Gloves and protective clothing remain a primary means of preventing skin exposure in the workplace, in addition to engineering and work practice controls. Gloves and protective clothing are presumed to protect against isocyanate skin exposure, with nitrile gloves considered preferable to latex. However, data on the workplace performance of protective gloves and clothing are limited, and there is evidence that isocyanates (Liu et al. 2000; Pronk et al. 2006) and solvents (Collin-Hansen et al. 2006) can be detected underneath gloves. Skin absorption of isocyanates Animal studies employing radiolabeled 14C-MDI have demonstrated absorption of MDI after skin exposure (Leibold et al. 1999; Vock and Lutz 1997). However, quantitative data are limited and may have underestimated absorption due to technical issues, such as isocyanate binding to the dressing and skin. Studies documenting the disappearance of isocyanates from guinea pig skin with infrared spectroscopy also support isocyanate skin absorption (Bello et al. 2006). We are not aware of published data directly confirming isocyanate skin absorption in humans; however, a number of studies indirectly demonstrate isocyanate skin absorption. For example, HDI-conjugated keratins have been identified from human skin biopsies obtained after epicutaneous application of HDI (Wisnewski et al. 2000). Patch testing with isocyanate (0.1–1%) can elicit an isocyanate-specific skin contact hypersensitivity reaction that implies isocyanate skin absorption. Studies investigating urinary biomarkers of isocyanate exposure (the corresponding diamines) have provided additional indirect evidence for isocyanate skin uptake (Creely et al. 2006; Kääriä et al. 2001; Maitre et al. 1996). Elevated levels of these urinary biomarkers have been detected in workers, in spite of very low or nondetectable documented inhalation exposures. Greater than 2-fold higher urinary metabolite concentrations have been reported for operators with likely skin contamination compared to those without (Creely et al. 2006). Maitre et al. (1996) measured urinary hexamethylene diamine (HDA) of two HDI-exposed coworkers. Both workers had similar inhalation exposure as confirmed by air measurements, but one had considerably greater HDI skin contact and a 10-fold increase in urinary levels of HDA. The authors concluded that HDI seemed to be readily absorbed through the skin. Human isocyanate skin absorption likely depends on a number of factors, in addition to the extent of isocyanate skin exposure, and may vary between isocyanates because of differences in their physical and chemical properties, including molecular mass, fat solubility, and chemical reactivity. Skin absorption can be enhanced if the barrier properties of skin have been damaged, such as can occur with eczema, cuts, hand washing, cosmetics applications (shaving, waxing), and other conditions (Moody and Maibach 2006; Smith Pease et al. 2002). In addition, coexposures such as solvents used in the production of polyurethane foams, coatings, and spray applications can be absorbed through the skin (Boman and Maibach 2000). Such solvents may enhance isocyanate absorption and also break through the gloves. Skin sensitization Numerous substances—primarily low-MW haptens, such as metals and chemicals, and less commonly proteins—are known to initiate immune responses in the skin, most commonly hapten-induced contact hypersensitivity (Kimber and Dearman 2002). Contact hypersensitivity (allergic contact dermatitis) following skin exposure to isocyanates is well documented in animals and in the clinical dermatologic literature, with sensitization confirmed with patch testing (Goossens et al. 2002; Herrick 2002). Allergic contact dermatitis has been reported following skin exposure to isocyanates and polyurethane products in a number of different workplace and non-occupational settings, but has not been considered common, and is rarely reported in workers with isocyanate asthma (Alomar 1986; Frick et al. 2003; Goossens et al. 2002; Wilkinson et al. 1991). However, allergic contact dermatitis may be more common than suspected because symptoms can be mild, workers being evaluated for asthma are frequently not asked about skin problems, and patch testing can be falsely negative (Frick et al. 2004; Goossens et al. 2002). Much less is known about the role of skin exposure in inducing Th2-type immune responses seen in asthma. Recent animal studies have documented that skin exposure to proteins such as ovalbumin or peanuts can induce systemic Th2-type sensitization and subsequent asthmatic responses (Herrick et al. 2003; Strid et al. 2005). Limited clinical and epidemiologic studies also support a role for skin exposure to allergens in the development of Th2-type sensitization and asthma, or of other immunologic lung diseases such as chronic beryllium disease (Cummings et al. 2006; Lack et al. 2003; Saloga and Knop 2000; Smith Pease et al. 2002; Tinkle et al. 2003). Isocyanate skin exposure, sensitization, and asthma Animal models Animal models using the three major isocyanates MDI, TDI, and HDI have all employed skin exposure to induce sensitization, with subsequent inhalation challenge, to create an asthmatic response in the lungs (Table 2). Skin sensitization with other chemicals, such as trimellitic anhydride, followed by inhalational challenge has also produced asthmatic responses (Vanoirbeek et al. 2006; Zhang et al. 2004). Different sensitization and challenge protocols (variable doses, frequency, timing, formulation, and route of exposure) using different isocyanates have resulted in variable pulmonary responses. In spite of these differences, several common themes emerge from these animal studies. First, skin has been a very effective route of inducing sensitization, sometimes more effective than inhalation (Ban et al. 2006; Rattray et al. 1994). For example, Ban et al. (2006) recently evaluated several different TDI sensitization scenarios (inhalation, subcutaneous, topical) and challenge protocols (vapor, tracheal instillation) in mice. Topical sensitization followed by tracheal instillation of TDI most closely reproduced the Th2-type lung inflammatory response seen in human asthma, whereas sensitization with vapor TDI was not effective. Of note, these isocyanate animal models also demonstrate that a single one-time or two-time skin exposure with relatively low concentrations of isocyanates can be sufficient to induce sensitization (Herrick et al. 2002; Karol et al. 1981; Rattray et al. 1994). Additionally, several models have shown paradoxical dose–response relationships, such that a lower skin sensitization dose can result in greater lung inflammation upon inhalation challenge than a higher sensitization dose, and that isocyanate-specific antibody responses may not correlate with asthmatic responses (Herrick et al. 2002; Vanoirbeek et al. 2004). The effective skin sensitizing doses in these studies, typically in the order of 1–100 μmol NCO (Table 2) delivered as diluted 1–10% isocyanate solution, represent a few droplets of a diluted isocyanate product. Comparable exposures likely occur in the workplace. Thus, several different animal models in more than one species clearly demonstrate that isocyanate skin exposure can induce systemic sensitization, which with subsequent inhalation exposure can lead to asthma. Control experiments in which animals received only skin exposure demonstrate that isocyanate skin exposure alone does not cause asthmatic responses (Herrick et al. 2002; Karol et al. 1981; Vanoirbeek et al. 2004). Although issues of comparability with humans are inevitable, such as differences in skin permeability, much can be learned from animal models regarding exposure–response relationships and the mechanisms that lead to skin sensitization and asthma. Human studies Although patch testing has confirmed contact hypersensitivity following human isocyanate skin exposure, direct evidence that skin exposure leads to Th2-type sensitization and the subsequent development of asthma is limited. As noted above, there is no good test to identify isocyanate sensitization in humans. Isocyanate-specific IgE is present in less than half of isocyanate asthmatics (Mapp et al. 2005; Wisnewski et al. 2006), and isocyanate-specific IgG appears to be indicative primarily of exposure (Liss et al. 1988; Welinder et al. 1988; Wisnewski et al. 2006). Isocyanates are considered a potent respiratory sensitizer in humans, based largely on circumstantial evidence rather than on clear demonstration that respiratory exposure alone (without associated skin exposure) leads to isocyanate asthma. Indirect evidence from a growing number of case reports and clinical and epidemiologic studies suggests that isocyanate skin exposure occurs in the workplace and can increase the risk for sensitization and isocyanate asthma. Isocyanate asthma and/or sensitization (MDI-IgE) have been reported in several case studies of workers who applied MDI-based orthopedic casts (Donnelly et al. 2004; Legris et al. 1995; Sommer et al. 2000). Allergic contact urticaria and asthma following direct hand contact with MDI glue has been documented based on a positive MDI-IgE, MDI patch test, and MDI inhalation challenge (Valks et al. 2003). Consistent across these case reports is the development of MDI asthma in settings where skin exposure to MDI occurred and where MDI air levels, if measured, were non-detectable or extremely low and opportunity for MDI respiratory exposure was very limited. These cases strongly suggest that skin exposure was the predominant exposure route and contributed to the development of isocyanate sensitization and asthma. They also demonstrate that isocyanate asthma can occur in settings where measured isocyanate respiratory exposures are below the level of detection, even when sensitive analytical methods are used. Skin exposure has also been implicated as a risk factor for isocyanate asthma in several epidemiologic studies of MDI-exposed workers, in which measured inhalation exposures were all very low or nondetectable. Petsonk et al. (2000) investigated a group of new workers exposed to MDI resins in a facility designed for minimal airborne exposures and noted new asthma-like respiratory symptoms in 27% of the workers with the highest potential for MDI exposure. These symptoms were associated with liquid MDI skin exposure, as evaluated by worker questionnaires and workplace observations. Leanerts-Langanke (1992) described an investigation of about 500 coal mine workers who used MDI for rock consolidation. Quantitative air sampling results were at or below the detection limit (1 ppb). MDI skin exposure was reported in about half the workers; MDI-IgE was detected in several workers; and 2 were diagnosed with isocyanate asthma based on a positive specific challenge. Elevated levels of MDI metabolites in urine were found in 6 of 8 workers seen after “massive skin contamination.” The authors concluded that skin could be an important route of exposure, leading to sensitization and asthma. Bernstein et al. (1993) studied a cohort of 243 workers in a urethane molding plant that consistently maintained low MDI airborne exposures (< 5 ppb). Isocyanate asthma and/or sensitization (MDI-IgE) was diagnosed in several workers, most of whom were reported likely to have had MDI skin exposure. Our group has characterized workplace isocyanate exposures in a population of > 200 auto body shop workers with isocyanate-specific immune responses but without documented isocyanate asthma (Redlich et al. 2001; Woskie et al. 2004). Respiratory and skin exposures were estimated for each of the workers, based on exposure algorithms. HDI-specific IgG, present in 21% of the workers, was strongly associated with inhalation exposure, but skin exposure also contributed (Stowe et al. 2006). We have also detected MDI-specific IgG in > 30% of about 100 workers in a factory that uses MDI to produce polyurethane coated fabrics. MDI air monitoring data have consistently been very low, and MDI skin exposure has been documented by worker questionnaires and direct observation (Liss et al. 2006). In both work settings, isocyanate skin exposure appeared to contribute to the development of isocyanate-specific IgG, which has been associated with isocyanate exposure (Welinder et al. 1988; Wisnewski et al. 2004; Ye et al. 2006). Studies investigating the respiratory exposure conditions that lead to isocyanate asthma also suggest a potential role for skin exposure. Fewer cases of isocyanate asthma have generally been reported in settings with lower respiratory exposures, but cases continue to occur in settings with consistently low reported air levels (Baur 2003; Tarlo et al. 1997). Skin exposure (Bernstein et al. 1993; Leanerts-Langanke 1992) or intermittent peak exposures, which could also entail both respiratory and skin exposure (Tarlo et al. 1997), have been considered important contributing factors in such cases. Typically, such exposures are unpredictable and frequently accidental, making them difficult to investigate and quantify. For cutaneous allergen exposure in general, the likelihood of sensitization in a susceptible person is thought to depend on several factors including the total dose and concentration of allergen, skin surface area, and the frequency of repeated contact with the skin, with determinants of individual susceptibility remaining poorly defined (Basketter et al. 2006; Boukhman and Maibach 2001). There are insufficient data to address isocyanate skin exposure–response relationships in humans; however, as noted above, in animal models relatively small skin doses can induce sensitization, and dose–response relationships may be variable. Regulatory standards for isocyanate skin exposure The regulatory framework for skin exposure to chemicals is underdeveloped compared with inhalation exposures. Chemicals that pose a health hazard through skin exposure are commonly assigned two qualitative descriptors: a “skin” notation, referring to absorption of the chemical through the skin, and/or a “sensitizer” notation for an agent with the potential to produce sensitization regardless of the exposure route (respiratory, skin, or conjunctiva) [American Conference of Governmental Industrial Hygienists (ACGIH) 2006]. In the United States, although NIOSH recommends prevention of isocyanate skin exposure, skin exposure is not regulated. No “skin” notation exists for diisocyanates or polyisocyanates, except for the NIOSH (2005) recommendation for IPDI. Although data confirming the risks of human isocyanate skin exposure remain limited, there is sufficient information to recommend prevention of skin exposure. Such recommendations are now being made in material safety data sheets and are beginning to appear in the medical literature (Bakke et al. 2006; Baur 2003). Wider dissemination and improved hazard communication of this information by occupational and environmental health professionals, as well as better personal protection among workers, are needed. Inclusion of “skin” notation may encourage such protection for all isocyanates (diisocyanates and polyisocyanates). Research needs In this review we highlight several important areas for further research regarding isocyanate skin exposure, ideally using multidisciplinary approaches involving animal models and clinical and epidemiologic investigations. Such approaches should lead to a better understanding of the mechanistic pathways that result in isocyanate asthma and the role of skin exposure in this process. A key research need not unique to issues regarding skin exposure is the development of a good marker of isocyanate sensitization or “pre-clinical” asthma that correlates well with the subsequent development of asthma. Such a marker would greatly enhance isocyanate research, including elucidation of exposure–response relationships, and facilitate diagnosis and prevention. There is a need to better assess isocyanate skin exposures in the workplace and other environments and to incorporate these exposure data into epidemiologic and clinical studies. The typically more sporadic nature of such exposures further complicates real-world exposure assessment and requires algorithms that employ a combination of daily activities (diaries), questionnaires, and task-based exposure data. Development of route-specific bio-markers, such as those specific for skin or lung, would greatly facilitate isocyanate exposure assessment. Skin exposure methodologies and biomarkers of exposure can be further developed and validated using integrated animal models. Numerous host and environmental determinants of isocyanate skin exposure have barely been investigated. Isocyanate skin absorption likely depends on various factors including molecular size and coexposures (e.g., polyols, solvents, and other additives), which could enhance absorption. The role of host factors, such as history of eczema and hand washing, warrant further investigation, as does the effectiveness of gloves, protective clothing, and other preventive strategies. Given the widespread and growing use of polyurethane consumer products, research is also needed to investigate potential environmental exposures from these consumer products, as well as exposures that may occur when isocyanate products are manipulated (ground, cut, drilled) or undergo thermal degradation as in fires. Further research investigating the curing process is also warranted because incompletely cured products could be a potential and unexpected source of skin exposure (Bello et al., in press). Conclusion Isocyanates, primarily diisocyanate monomers and polyisocyanates, are a leading cause of occupational asthma. Human skin exposure to isocyanates has been underrecognized and can occur in various workplace and environmental settings. Multiple lines of evidence from animal studies and clinical, epidemiologic, and biomarker studies, as well as anecdotal evidence, indicate that in certain exposure settings human skin likely is an important route of isocyanate exposure and can contribute to the development of isocyanate asthma. This presumably occurs by isocyanate skin exposure inducing systemic sensitization, which then leads to isocyanate asthma after inhalation exposure; however, the mechanistic pathways involved remain poorly defined. Further research is needed to address issues regarding isocyanate skin exposure in the growing polyurethane industry. In spite of substantial research needs, sufficient evidence already exists to justify greater emphasis on the potential risks of isocyanate skin exposure and the need to prevent such exposures. We conclude with the visionary statement of Munn (1965): The more one knows about these fascinating compounds [isocyanates] the more fascinated one becomes. So diverse are their uses, it is obvious that they are here to stay, and that their use will increase. So numerous have been the accounts of their effects, it is obvious not merely they are hazardous but that the nature and extent of their hazard has not always been fully appreciated.

Int_Ophthalmol-4-1-2359829

Eye bank issues: II. Preservation techniques: warm versus cold storage

Most of the tissue used for penetrating keratoplasty is issued through eye banks that store the corneoscleral button either in hypothermic storage at 2–6°C or in organ culture at 31–37°C. Introduction Throughout the world post-mortem eye tissue is used for keratoplasty. The generally accepted storage method for the whole globe is the “moist chamber”; a moistened pot at 2–6°C introduced in 1935 [1]. The corneoscleral button is stored in tissue culture medium, either in the hypothermic storage method at 2–6°C introduced in 1974 [2] or in the organ culture method at 31–37°C introduced in 1976 [3]. Corneas cannot reliably be frozen. The prevailing storing technique today is the storage of the corneoscleral button. The storage time can be extended by removing the corneoscleral button from the globe. A longer storage time permits greater flexibility in the use of the donor tissue and prevents wastage. In addition in many countries, for different reasons, legal or ethical, corneoscleral disc excision in situ is preferred to the removal of the whole globe. Moreover, the preparation of lamellae and mushroom-shaped grafts from a corneoscleral button is nowadays possible with help of an artificial anterior chamber. With the introduction of the preservation media, corneal surgeons have to rely on highly skilled technicians employed in eye banks selecting and storing the donor tissue. The original hypothermic storage solution, the M–K medium, has been succeeded by other solutions claiming better and longer maximum storage results. The hypothermic method is common all over the world. Although organ culture originates from the United States [3, 4] it has been strongly promoted by the Eye Bank of Århus in Denmark [5, 6] and is now widely applied in Western Europe but not commonly used elsewhere. After some modifications shortly after its introduction the organ culture storage technique has stayed the same. The storage media became commercially available although some eye banks still prefer to prepare them themselves. They differ slightly in composition between banks and countries [7]. Both storage methods, hypothermic and organ culture will be compared considering technical aspects, tissue evaluation possibilities, storage time, microbiological safety, graft survival and future applications with regard to the increased interest in lamellar grafting. Technical aspects General Procurement and storage techniques have to be performed under aseptic conditions. Increasingly, eye banks have a formally established quality assurance program. This and the increasing regulation may affect banks working on a smaller scale. For example in France the number of operating eye banks reduced form 226 in 1993 to 43 in 2004. Hypothermic storage The technique is simple: refrigerator storage with minimal handling. It requires no complex or expensive equipment. The storage solutions are commercially available and manufacturer recommendations should be followed for temperature, maximal storage time, expiry date and other factors. The vials may allow inspection of the endothelium by specular microscopy (Fig. 1 left). During storage the cornea remains thin and, provided donor screening permits release of the tissue, it is directly available for surgery.Fig. 1Corneoscleral buttons in different vials for hypothermic storage (left) and in the incubator during organ culture (right) The medium consists of a tissue culture medium, supplemented with antibiotics, deturgescent agents like dextran and chondroitin sulphate to prevent corneal swelling in vitro, and other additives such as energy sources, antioxidants, membrane stabilizing and growth factors to improve the storage capacity. Inspection of the tissue by slit-lamp and/or specular microscope can be performed in a closed system. Organ culture The technique is relatively complicated, despite the fact that nowadays the storage solutions are commercially available. The corneas are stored in an incubator at 30–37° C in a tissue culture medium, supplemented with fetal or newborn calf serum, antibiotics and antimycotics (Fig. 1 right). Dehydrating macromolecules, necessary to maintain normal hydration in vitro, are ingested by the corneal cells at a physiological temperature and found accumulated in vacuoles in the cells and layers of the cornea [8, 9]. Therefore they are omitted from the storage solution. As a result the cornea swells to about twice its normal thickness during storage. The swelling should be reversed before transplantation. This is performed by placing the cornea in the storage medium supplemented with dextran. This so-called transport medium is also used for the transport of the cornea at room temperature. The extent of deswelling depends on the dextran concentration, varying from 4–8% in the different banks, resulting in a thickness of about 0.5 to 0.7 mm, depending on the preference of the surgeon(s) using the tissue from a certain bank. The maximum time the cornea spends in the transport medium also varies between banks; from less than one up to seven days [7, 8]. The risk of the ingested dextran in relation to the export area and transport time is judged differently. Depending on the media used, renewal of the medium occurs after 10–14 days of storage [7]. For inspection of the endothelium, mandatory after storage, an invasive technique has to be used, which has to be performed under strict aseptic conditions. The necessary transfer of the cornea from the storage to the transport solution averts a stored cornea from being directly available for use. Besides, a minimal storage period is required for microbiological testing. All this makes the technique more complicated than the hypothermic storage method. Tissue evaluation General Irrespective of the storage method(s) used the donor should be adequately screened. Tissue that is potentially hazardous to eye bank personnel and the recipient should be excluded in addition to tissue that poses a risk for the success of the surgery. Physical assessment, serologic testing and evaluation of medical and social history of the donor are mandatory. Routine inspection of the endothelium is also part of the donor evaluation. Moreover, it can play an important role in setting higher and more-uniform quality standards for tissue acceptance. It may also help to increase the donor supply by assessing corneas that may otherwise be arbitrarily excluded for transplantation on the basis of age or time post mortem. The possibilities for evaluation of the endothelium are dependent of the storage method used. Studies linking graft outcome with morphometric parameters of solely the endothelium are still lacking. A model has been presented to calculate endothelial cell loss in the long run of 10–20 years after penetrating keratoplasty [10]. In this way it is possible to predict when cell density would reach levels that are incompatible with maintenance of transparency and graft function. The model provides a rationale for the setting of minimum donor cell densities. However, the definite cut-off points are still at the discretion of the bank and surgeon. Hypothermic storage Determination of endothelial cell density is a standard method of corneal tissue evaluation according to the Medical Standards of the Eye Bank Association of America, effective since December 2001. In general this will be a pre-storage evaluation of the endothelium by specular microscopy (Fig. 2 left). Because the appearance of the endothelial cells varies with temperature, type and time of preservation and media, evaluation at room temperature is recommended. When it is impossible to obtain an endothelial cell count, this requirement may be waived on a case-by-case basis by the Medical Director.Fig. 2Evaluation of the endothelium by specualr microscopy (left) and light microscopy after artificial swelling of the intercellular space (right) Selection criteria and cut-off points for the morphology of the corneal endothelium are not defined. In most cases only descriptions of the endothelial mosaic are used: swollen cells, dark spots, guttae, lysed cells, mild to severe polymegethism and pleomorphism [11]. Most specular microscopes are equipped with software programs to determine parameters describing the endothelial mosaic in terms of variation in cell shape andthe percentage of hexagonal cells in addition to the assessment of the cell density. The obtained morphometric results may help to standardize descriptions, but only provided the microscopes are well calibrated and the necessary interactive manipulation is performed by an experienced observer. Organ culture After organ culture specular microscopy is not suitable for visualization of the endothelium. Therefore light microscopy is applied, bright field or phase contrast. It is necessary to visualize the endothelial cells by swelling the intercellular space with a hypotonic solution. This allows inspection over the entire endothelial surface (Fig. 2 right). The mechanism has been described by Kirk and Hassard [12] and was worked out for the corneal endothelium by Sperling [13]. Because it is an invasive technique, it must be performed under aseptic conditions. The swelling is transient, it disappears after a couple of minutes and is dependent on the integrity of the cellular membranes. In dead and necrotic cells and in their direct neighborhood swelling will not occur. Before and after storage balanced salt solution (BSS), phosphate buffered saline (PBS), 1.8% sucrose—PBS mixture or hypo-osmotic BSS may induce swelling, while for tissue in solutions with dehydrating agents a stronger stimulus with 1.8% sucrose might be necessary [7]. Induction of swelling and the swelling pattern are dependent on storage time and medium [14]. Therefore the interpretation of images requires experience and constant working conditions. The application of a vital stain such as trypan blue [15, 16] preceding the artificial swelling of the intercellular space may help to recognize dead or necrotic cells or denuded Descemet’s membrane. Microscopes should be well calibrated both for manual counting as for evaluation by image analysis programs [17]. These software programs are commercially available, either specifically designed for endothelial evaluation, or as general programs adapted to do this. All programs aim for automated cell analysis that is independent of the observer and experience. However, in general, reliable parameters for the endothelial mosaic are only obtained interactively. This still requires experienced observers. Manual counting by Gunderson’s method [18] can provide reliable cell counts [19], but for parameters such as variation in cell size and the percentage of hexagonals image analysis is necessary. The quality of the corneal stroma can be evaluated by light microscopy. The significance of the presence of lysed keratocytes for the survival of the graft has not yet been investigated [20]. Storage time General As the endothelium is essential for graft clarity and survival, the maximum allowed storage time is predominantly determined by maintenance of the endothelial function and integrity [21]. Regression lines, coefficients and equations have been published for storage time and damage or loss of endothelial cells. In Fig. 3 the regression lines are collected for different storage methods [22–24]. The referred studies have in common that the vital stain with trypan blue is used to assess endothelial damage. The figure clearly demonstrates the differences in endothelial viability and explains the differences in the allowed maximum storage period.Fig. 3Regression lines showing endothelial cell damage and loss in different storage solutions assessed after staining with trypan. The regression formulas for the M–K medium, the Optisol GS and the organ cultures are respectively: y = 11.8x + 11.6, y  = 0.19x + 4.2, and y = 0.11x − 1.8 Hypothermic storage The original M–K medium claimed a storage period of up to 10 days. Solutions introduced later, such as the modified M–K medium, K-sol, Dexsol, Likorol, Optisol (Plus, GS) claimed better storage capabilities and a maximum storage period of 14–16 days. Changes in the endothelium as a result of post-mortem time and other variables, such as cause of death, donor age, circumstances of death etc., might result in the transplantation of corneas of inferior quality after storage. Because degenerative changes will progress during hypothermic storage, endothelial cell loss has to be taken into account [25], and might even lead to complete cell death. The need for methods to detect corneas not tolerating the prolonged hypothermic storage has therefore been suggested [26]. To reduce the risk of primary graft failure, the recommended storage periods are kept far below the claimed maxima. The applied period increased from 2–3 days for the M–K medium [27, 28] up to 7–10 days for Optisol [7, 25]. In addition the time interval from death of the donor to storage of the cornea is generally kept relatively short, within 12 h or shorter [7]. Organ culture With organ culture the allowed storage period is longer. In addition the time interval between death and storage is generally extended to 24–48 h, because significant wound healing can still occur during storage [29]. The loss of endothelial cells during storage may differ between individual corneas (Fig. 4) and is supposed to reflect differences in vitality as a result of post-mortem time and other variables such as cause of death, donor age, circumstances of death etc. In this respect, prolonged hypothermic storage might not be so different from organ culture. Severe endothelial cell loss during organ culture may also be caused by herpes simplex virus infection of the donor tissue [30, 31]. Organ culture is therefore considered a stress test [32, 33]. A storage period of up to 4–5 weeks is possible [32–35]. However, in order to detect tissue not tolerating the storage, according to European Eye Bank Rules, inspection of the endothelium after storage is mandatory.Fig. 4Percentage endothelial cell loss (endothelial cell density before storage minus the cell density after divided by the cell density before multiplying by 100) during routine storage by organ culture is plotted for a given year Microbiological safety General Donor eye tissue is usually contaminated [36] and each eye bank has to cope with this. As a first step decontamination procedures are applied before enucleation or excision of the corneoscleral button. When properly performed these procedures are very effective at reducing the risk of contamination [37]. As a next step, antibiotics, and in the case of organ culture antimycotics, are present in the storage solutions. These are more effective if the contaminating microbes are metabolically active, which means that they are more effective in organ culture than during hypothermic storage. In addition the vulnerability of organ culture to microbial contamination can be exploited to detect microorganisms remaining from the donor and/or introduced into the culture medium by the environment or personnel. Theoretically, in organ culture the risk of contamination is therefore lower than in hypothermic storage. Hypothermic storage Antibiotics have little effect during hypothermic storage. Preoperative warming of the storage media to room temperature is important to enhance the decontamination effect. The optimal time period of room-temperature storage has yet to be established [38]. In addition, antibiotics accumulate in the tissue during storage [39] and become active in the eye after grafting as the temperature rises. Donor rims tested after grafting are positive for bacteria and fungi in 12–28% of the cases [40]. However, the value of routine donor rim cultures in clinical use is debated. On the one hand no relevance of infectious complications after keratoplasty has been reported [41, 42]. On the other hand a 22 times increased incidence of endophthalmitis in the case of a positive rim culture [43] to a fully associated fungal infection following penetrating keratoplasty has been described [44]. The overall incidence of postoperative keratitis and endophthalmitis caused by microbes transferred with the donor cornea is low and varies from 0.2% [43] to 0.41% [45] and to 1.3% [46]. The addition of alternative antibiotics to the current hypothermic storage solutions has been suggested, because donor corneas are often removed in hospitals and other settings where resistant bacteria may be thriving [38]. These may cause an increased incidence of endophthalmitis. From a microbiological point of view, however, it is disputable whether the addition of the latest antibiotics with the widest spectra is the best solution. Organ culture Contamination detected during organ culture varies between eye banks [7]. This may be dependent on the antibiotic cocktail present in the medium (narrow- to wide-spectrum antibiotics), collection procedures, in situ excision or enucleation and the post-mortem time of collection and storage [47]. Microbiological testing of medium samples before surgery is mandatory as well as a quarantine period before issuing corneas. Microbiological safety of the tissue stored by organ culture is obtained by discarding contaminated tissue before grafting. The incidence of endophthalmitis reported after a properly performed organ culture procedure is 0–0.1% [7]. Sterility may be better with organ culture since microbial contamination will become more readily evident. Organ culture might be considered as the method of choice in circumstances where corneas are suspected of being at a higher risk of contamination. The mandatory reporting of adverse reactions and the central registration of these within the EU area where both the hypothermic and the organ culture techniques are used will demonstrate in time whether the incidence of endophthalmitis really differs between organ culture and hypothermic storage. Graft survival Only one prospective study is available comparing clinical results after grafting between the hypothermic storage and organ culture [48]. During the 1–2 year study period no statistically significant differences in visual acuity, corneal thickness or endothelial cell density were found in keratoconus patients. Bourne [49], however, found fewer endothelial cells after grafting on corneas stored by organ culture. At that time, M–K medium was used to reverse the swelling, a method differing from the method used in Europe. Retrospective studies claim improved [33, 50] or comparable results with organ culture [34, 35, 51]. The improvement is not ascribed to the storage itself but to the extra selections routinely included in organ culture preservation. In addition most of the studies were performed at a time when old-fashioned hypothermic storage methods did not always include inspection of the endothelium. Future aspects With the advent of new surgical techniques such as lamellar grafting, the issues for eyebanks are also changing. The risk for immunological graft rejection is theoretically lower in lamellar grafting. This may reduce the interest in HLA matching of donor and recipient, an important reason to prefer organ culture as a storage method. Organ culture provides sufficient time for typing and matching and is also thought to immunologically modify the tissue through the loss of passenger leucocytes and epithelium [52]. For a long time stored corneosclereal buttons were used for penetrating keratoplasty whereas lamellar grafting was performed with tissue from whole globes stored in a moist chamber. With the introduction of the artificial anterior chamber, the microkeratome and the intralase or femtosecond laser, this is changing quickly. Corneoscleral buttons which have been precut in order to reduce the manipulation of the donor tissue at the time of surgery will be new products of eye banks. For a reliable production of these precut buttons the intralase or femtosecond laser technique seems very promising [53–55]. Whether the storage technique (hypothermic of organ culture) might affect the postoperative fate of the corneal lamella is still open for discussion. The revival of the mushroom technique is another development, where the eye bank would be expected to prepare the corneal button [56, 57]. Hypothermic storage, where the thickness of the cornea is maintained and where the risk of epithelial ingrowth is low because there is no epithelial growth, seems to be more suitable for these indications. Conclusions With the more widespread use of the specular microcope and the introduction of hypothermic storage solutions claiming an intermediate storage period some of the advantages ascribed to organ culture [58], such as scheduling operations and minimizing wastage of donor tissue, have become less important. Others still remain: allowing time for tissue typing and matching and extensive testing, detection of residual micro-organisms before surgery, and selecting and dispatching corneas with a well-defined endothelial quality assessed after storage. In addition the pool of possible donors can be enlarged thanks to the possibility of wound healing during organ culture and the inspection of the tissue after storage. Because organ cullture involves extensive testing of the quality of the donor cornea, there are theoretically no preset limits on donor age and postmortem time. To permit these possibilities the organ culture procedure is more complicated than the hypothermic storage. Well qualified personnel, advice of microbiological laboratory staff, and a well suited facility are essential for a proper organ culture procedure but may also be valuable for hypothermic storage. The higher costs of organ culture have to be balanced against the offered advantages and possibilities not permitted by the hypothermic storage. Whether organ culture will also be the proper storage technique for surgically manipulated corneoscleral buttons remains to be investigated.

Eur_Biophys_J-3-1-2082062

Organisation of nucleosomal arrays reconstituted with repetitive African green monkey α-satellite DNA as analysed by atomic force microscopy

Alpha-satellite DNA (AS) is part of centromeric DNA and could be relevant for centromeric chromatin structure: its repetitive character may generate a specifically ordered nucleosomal arrangement and thereby facilitate kinetochore protein binding and chromatin condensation. Although nucleosomal positioning on some satellite sequences had been shown, including AS from African green monkey (AGM), the sequence-dependent nucleosomal organisation of repetitive AS of this species has so far not been analysed. We therefore studied the positioning of reconstituted nucleosomes on AGM AS tandemly repeated DNA. Enzymatic analysis of nucleosome arrays formed on an AS heptamer as well as the localisation of mononucleosomes on an AS dimer by atomic force microscopy (AFM) showed one major positioning frame, in agreement with earlier results. The occupancy of this site was in the range of 45–50%, in quite good agreement with published in vivo observations. AFM measurements of internucleosomal distances formed on the heptamer indicated that the nucleosomal arrangement is governed by sequence-specific DNA-histone interactions yielding defined internucleosomal distances, which, nevertheless, are not compatible with a uniform phasing of the nucleosomes with the AGM AS repeats. Introduction Satellite DNA consists of tandemly repeated sequences with monomer lengths of up to 500 bp and copy numbers of up to millions. It is found in all eukaryotic species where it is concentrated in the centromeric regions of the chromosomes, as shown for example in humans (Schueler et al. 2001). The association of satellite DNA with essential processes ensuring genome stability, such as the assembly of the kinetochore (Willard 1998), is still unclear since these processes are not strictly sequence-dependent (Amor and Choo 2002). α-satellite DNA (AS) might contribute to the formation of a centromere-specific chromatin structure. The presence of oligo-adenine tracts and intrinsic curvature are universal features of satellites, which therefore may have a role in positioning nucleosomes along DNA and in chromatin condensation (Fitzgerald et al. 1994). Indeed, sequence-dependent nucleosome positioning on satellite DNA fragments was observed in vitro (Linxweller and Horz 1985; Neubauer et al. 1986; Tanaka et al. 2005; Yoda et al. 1998). Also, satellite chromatin isolated from cell nuclei was found to differ from bulk chromatin by a higher compaction level which may be supported by the presence of a particularly ordered nucleosomal arrangement (Gilbert and Allan 2001). Alpha-satellite DNA is characterised by a monomer length of ∼171 bp and is primate specific. This DNA allows for a simple phase relationship between nucleosome binding and the repeat length. In fact, AS chromatin of African green monkey (AGM) (Cercopithecus aethiops) was used earlier to verify this phase relationship in vivo. The initial finding of a simple phase, based on micrococcal nuclease (MNase) digestions of cell nuclei (Brown et al. 1979; Musich et al. 1982), was challenged by the finding that the MNase cleaves the AGM AS non-randomly (Fittler and Zachau 1979; Horz et al. 1983). More detailed studies showed the occupation of different nucleosomal positions relative to the AS sequence in AGM-cells where, however, one highly preferred site consistently emerged in different approaches (Smith and Lieberman 1984; Wu et al. 1983; Zhang et al. 1983). Neubauer et al. (1986) identified nucleosomal positioning sites similar to those occurring in vivo by in vitro reconstitution experiments using 250 bp AGM AS fragments. The major site was occupied by 80% of the nucleosomes on 250 bp DNA in vitro and by 35% in vivo (Neubauer et al. 1986; Zhang et al. 1983). These results argued for a strong contribution of sequence-dependent DNA–histone interactions to the nucleosomal organisation in AS chromatin of this species, although nucleosomal positions and internucleosome spacing generally depend on many different factors in vivo. For instance, the sequence-dependent positioning signal contained in the 5S rDNA of Lytechinus variegatus can be overruled when nucleosomes are reconstituted in cell extracts (Blank and Becker 1996). Recently, Segal et al. (2006) found a high predicted occupancy of nucleosomes over centromeres indicating that centromere function requires enhanced stability of histone–DNA interactions that are encoded in the genomic sequence. The importance of a correct nucleosomal positioning at centromeric chromatin is also indicated by the observation that the foundation centromeric protein CENP-B influences the position of nucleosomes along AS in vitro (Tanaka et al. 2005). CENP-B is the only known inner kinetochore protein to bind to a specific centromere sequence, the 17 bp Cenp-B box, which occurs in a subset of α-satellite monomers in humans (Masumoto et al. 1989). AS arrays of humans are composed of a number of divergent AS monomers organised into higher order repeat units, whereas arrays of AGM exhibit a simple and homogeneous monomer organisation and, furthermore, lack Cenp-B-boxes (Alexandrov et al. 2001; Goldberg et al. 1996). Intriguingly, CENP-B was found to be absent from the centromeres of AGM although the protein is expressed (Goldberg et al. 1996). One interpretation for this result could be that CENP-B is dispensable in the centromeres of AGM, since the AS by itself efficiently defines nucleosomal positions (Goldberg et al. 1996), even though earlier the existence of a simple nucleosomal phase in vivo had been argued against, as mentioned above. The motivation of the present work was to extend the earlier studies of nucleosome positioning on AS of this species. So far, the sequence-dependent assembly of nucleosome arrays has not been studied using repetitive AGM-AS, and nucleosomal positioning on DNA fragments containing the uninterrupted AGM-AS sequence has not been quantified. We therefore reconstituted oligonucleosome arrays with a heptamer and mononucleosomes with a dimer of the AGM AS and took advantage of atomic force microscopy (AFM) as single molecule approach to determine the nucleosomal organisation along the DNA. AFM can measure internucleosome distances with high precision (Allen et al. 1993; Mechelli et al. 2004; Yodh et al. 2002; Zlatanova et al. 1998) and also the quality of the reconstituted material can be characterised at the single molecule level. The height information, available through AFM imaging, gives molecular structural information (Nikova et al. 2004). Thus, the AFM approach circumvents drawbacks of enzymatic assays. Yet, we additionally used MNase and restriction enzyme digestion in order to find out whether chromatin reconstitution would give results similar to the earlier in vivo findings (Musich et al. 1982), taking into account that this approach may overestimate the extent of evenly spaced nucleosomes. Our AFM measurements of internucleosome distances show that the in vitro generated nucleosomal arrangement along the AGM AS heptamer is clearly defined by the sequence, yet deviates from a simple phasing. The preferred occupation of the known major positioning site was unambiguously detected by (1) MNase and EcoRI digestion and (2) by localising mononucleosomes reconstituted on the AGM AS dimer by AFM. The AS sequence repeat thus has a profound influence on the nucleosomal arrangement formed by reconstitution on this DNA and also the estimated occupancy of the major site was in a quite good agreement with the earlier in vivo findings. Material and methods Preparation of DNA African green monkey α-satellite DNA was prepared from CV1-cells, a permanent line derived from C. aethiops kidney cells. Whole genomic DNA was isolated, deproteinised and treated with EcoRI restriction endonuclease essentially as described (Gruss and Sauer 1975). The DNA samples were run on 1.0% agarose gels and stained with ethidium. EcoRI produces one large smear in the upper region of the gel and a number of discrete bands below, corresponding to multiples of the AS-monomer (Gruss and Sauer 1975). Our attempts to clone repetitive AGM AS embarked on two strategies. One was to directly excise EcoRI-fragments containing different repeat numbers from low melting agarose gels (FMC) and ligate these with the cloning vector pBluescript-IIKS+. Escherichia Coli XL10 Gold cells (Stratagene) were transformed with the ligation products and cloned. By this approach, vectors containing AS-monomers and AS-dimers but no larger repeat numbers could be successfully cloned. It was verified by DNA-sequencing that the cloned DNA matches AGM AS: an isolated AS-dimer proved to be a tandem repeat of identical sequence. This yielded the plasmid pBluescript-IIKS-α2. The second approach to obtain longer AS-arrays was a stepwise self-ligation of the already available α2-insert. pBluescript-IIKS-α2 was subjected to a partial EcoRI-digestion and the cleavage products were separated on a 1.4% low melting agarose gel to isolate the α2-fragments. These were used for self-ligations at 4°C for 2–4 days using 10 U/μl of T4 ligase. Afterwards, the DNA was ligated with pBluescript-KSII+ and we could obtain clones containing an AS-tetramer (pBluescript-KSII-α4). The α4-insert was used for the next self-ligation step carried out accordingly. This resulted in a clone containing an AS-heptamer (pBluescript-KSII-α7). DNA-sequencing was used to verify the tandem orientation of the AS-repeats. A 208-6 construct was produced starting from the vector pPolI-5S-208-12 (a gift of Peter Becker) which contains a dodecamer of the 5S rDNA nucleosome positioning sequence (Blank and Becker 1996; Simpson et al. 1985). This DNA was subjected to a partial digestion with AvaI to produce linearised vector-DNA with different repeat numbers of the 208-sequence. The cleaved vector was re-ligated and cloned. A clone containing a hexamer of the 208-sequence was identified by 1.2% agarose gel electrophoresis after cleavage of the purified vector with PstI. To prepare the target DNA-fragments for the chromatin reconstitution experiments, bacteria were grown in Luria Bertani medium at 37°C for 24 h and the cells were harvested by centrifugation at 6,000 rpm for 20 min at 4°C. The plasmids were isolated using the Nucleobond-kit (Machery-Nagel). AS-vectors were cleaved with NotI and XhoI to separate a fragment containing the AS-insert flanked on both sides by 33 and 37 bp, respectively (172-α2, 415 bp and 172-α7, 1,275 bp). The 208-6 DNA was separated from pPolI-5S-208-6 by cleavage with PstI and XbaI. The cleavage products were run on 1.4% low melting agarose gels, the target fragments were excised from the gels and purified using the gel purification-kit (Qiagen). Chromatin reconstitution Recombinant histone proteins of Xenopus laevis were overexpressed and purified as described (Kepert et al. 2003; Luger et al. 1999). Histone octamers were formed by incubation of a stoichiometric mix of the histones for 30 min in unfolding buffer (7 M Urea, 20 mM Tris–HCl, pH 7.4) and subsequent dialysis against refolding buffer (20 mM Tris–HCl, pH 7.4; 0.1 mM EDTA; 0.5 mM β-mercaptoethanol; 2 M NaCl). An electrophoretic analysis of the used octamer preparation is shown elsewhere (Bussiek et al. 2005). Nucleosome arrays with 172-α7 and mononucleosomes with 172-α2 were reconstituted by the salt dialysis method. Octamers were mixed with the DNA in 10 mM Tris–HCl, pH 7.5; 2 M NaCl and kept at 30°C for 30 min. The mixture was transferred to mini dialysis tubes (Pierce, Rockford, USA) and dialysed against buffers with stepwise decreasing NaCl concentrations at 4°C (1.8, 1.4, 1.0, 0.8, 0.6, 0.4, 0.2 and 0.1 M for 1 h each and 0.005 M over night). The reconstitutions took place at different weight ratios of histone-octamer to DNA, ranging between values of 0.3–1.0. The DNA mass was 0.02 μg/μl. Analytical digestions with micrococcal nuclease and EcoRI restriction endonuclease 172-α7 DNA, either naked or reconstituted with histones, was digested with micrococcal nuclease (MNase). Reactions took place in 10 mM Tris–HCl, pH 7.5; 5 mM NaCl; 0.5 mM CaCl2 using 0.3 mU/μl MNase and 0.02 μg/μl of the DNA. The reactions were stopped by mixing 50 μl aliquotes of the reaction with 5 mM EDTA after varying reaction times. Then, the DNA was isolated with phenol/chloroform/isoamylalcohol, precipitated with ethanol and resuspended in 15 μl Tris–HCl, pH 7.5, 0.1 mM EDTA. The DNA was analysed on 1.2% agarose gels run in 1 × TBE-buffer using ethidium staining. For analytical digestion with EcoRI, another fraction of the reconstituted chromatin was first digested with MNase to produce mononucleosomal DNA. The reaction conditions and purification of the DNA was carried out as described above, where a total of 1.5 μg of DNA were digested. An aliquot corresponding to 1.0 μg of the purified DNA was subjected to EcoRI digestion for 2 h. The two DNA-samples (MNase and MNase + EcoRI) were run on a 1.7% agarose gel in 1 × TBE, together with a 20 bp-ladder and stained with ethidium. Atomic force microscopy Atomic force microscopy was performed in air and in solution using a Multimode™ (Digital Instruments) operated in tapping mode essentially as described (Bussiek et al. 2005). Mica modified with poly-L-lysine (PL) served as support for sample immobilisation. Freshly cleaved mica was pre-treated with 30 μl of an aqueous solution of PL (Sigma) at a concentration of 10 μg/ml. After incubation for 30 s, unbound PL was removed by rinsing the mica disc with 4.0 ml of Millipore purified water followed by drying under a nitrogen stream. The chromatin was diluted to concentrations ranging between 0.5 and 2.0 nM (depending on the DNA size) in 10 mM Hepes-NaOH, pH 8.0, supplemented with salt as indicated in Sect. “Results”. Thirty microlitres of the dissolved chromatin solution were placed on the PL-mica. For scanning directly in the adsorption buffer, silicon nitride probes (type NP-S20, Veeco Instruments) were used at drive frequencies of 8.0–9.5 kHz and a set point of 0.3–0.4 V. For scanning in air, the mica disc was washed carefully with 2.0 ml of Millipore water after incubating with the chromatin solution for 1 min. Then, the disc was dried with nitrogen and subsequently scanned using etched silicon probes (type NCH, Nanosensors) at drive frequencies of 280–320 kHz and set point of 2.0–2.2 V. The images were recorded both in solution and in air at a scan diameter of 2 × 2 μm, a scan rate of 1–2 Hz and a resolution of 512 × 512 pixels. Image analysis Atomic force microscopy images were flattened and zooms of individual complexes were produced using the Nanoscope IIIa software (version 5.12r3, Veeco Instruments). Measurements of nucleosomal heights were done with the section analysis tool available in the same software package. This was also used to decide in some critical cases, whether a pair of two closely localised nucleosomes or a single one was present. When the cross-section showed two peaks or clearly a plateau next to a peak, two closely localised nucleosomes were assumed. Length measurements along DNA or chromatin contours were done using the freehand line tool in the program ImageJ (version 3.13, National Institute of Health). To measure the lengths of DNA fragments complexed into nucleosomes (Lcomplex), contours were traced along the entry and exit sites of the DNA strands into the nucleosomes. The nucleosome diameter measured by AFM is usually larger than the real diameter due to tip convolution, which means that the actual entry/exit sites cannot be unambiguously localised. To account for this error, the apparent entry/exit sites were always connected through the nucleosomal centres. In this way, the length of free DNA outside the nucleosomes (Lfree; free linker DNA and DNA termini) could be estimated by subtracting the known diameter of the nucleosome core particle (NCP) of 11 nm: where nncp is the number of NCPs per complex. Internucleosomal spacings between consecutive nucleosomes in the arrays were measured as centre-to-centre (cc) distances. The cc-distances were measured along DNA, when two nucleosomes were separated by a distance sufficiently long to make the intervening DNA visible. Otherwise, when two nucleosomes appeared close to one another, a straight line between the nucleosomal centres was assumed. The resolution limit that allows distinguishing nearby nucleosomes closely coincides with the NCP diameter. Since a correlation between the cc-distance and the internucleosomal separation along the DNA is not justified in this range of small distances, a minimum value of 11 nm was defined for the data analysis. The accuracy of the measurement as given by the pixel-size of the zooms used here is 0.5 nm. Pairs of nucleosomes had to fulfil some criteria in order to be included in the cc-distance measurements: it had to be unambiguous that the nucleosomes are ordered one after another along the DNA and the nucleosomes had to feature a certain minimal height (see below). Finally, nucleosomes localised directly at DNA termini were excluded from the cc-distance measurements. Results We studied the positioning of reconstituted nucleosomes on AGM α -satellite (AS) tandemly repeated DNA. The cloned AGM AS We constructed vectors containing AGM AS-DNA monomers as tandem repeats of up to seven. Figure 1a shows the DNA-sequence of the isolated AS (Fig. 1a, 1) aligned with the sequence (Figs. 1a, 2) analysed earlier in nucleosome reconstitution experiments (Neubauer et al. 1986). This sequence did not contain the HindIII site that is present in the major fraction of AGM AS (Musich et al. 1982; Neubauer et al. 1986; Wu et al. 1983). The vectors containing the AS-fragments used for the reconstitution experiments (172-α2 and 172-α7) were NotI/XhoI digested and analysed on the agarose gel in Fig. 1b. For the reconstitution experiments, the target fragments were isolated from the agarose gels. Fig. 1Characterisation of the satellite DNA fragments used for chromatin reconstitutions. a The cloned AGM AS (1) is shown in upper case letters and aligned with the sequence (2) analysed earlier in reconstitutions experiments [adapted from Neubauer et al. 1986]. Lower case letters show flanking vector DNA. The EcoRI recognition sites are underlined. Numbering of sequence positions is according to Rosenberg et al. (1978). The sequence in brackets indicates the repetition of the 172 bp AS monomer starting and ending at the EcoRI site in our case (n = 2, 172-α2 DNA; n = 7, 172-α7 DNA). b Agarose-gel electrophoresis of the AS fragments after separation from the cloning vector with NotI and XhoI (lane1 172-α2, 415 bp, lane 2 172-α7, 1,275 bp). Note, that the target fragments were isolated from agarose gels for the reconstitution experiments. Here, the NotI/XhoI digestion was directly loaded on the gel to also show the occurrence of a distribution of different AS repeat numbers around the major repeat length of n = 7 in the cell line containing pBluescript-KSII-α7Fig. 2Micrococcal nuclease (MNase) and EcoRI cleavage of reconstituted 172-α7 nucleosome arrays. a Arrays (chrom) were treated with MNase for 1–5 min. Samples were loaded on the 1.2% agarose gel together with untreated arrays and naked 172-α7 DNA, also either treated with MNase for 5 min or untreated. Untreated arrays and naked DNA migrate similarly in the agarose gel (DNA compaction and the increase in molecular weight by complexing with histones approximately compensate). First two lanes marker DNA. A photograph of the gel taken at higher light intensity is included underneath. b Magnified area of the same gel, marked by the frame in a.Arrows are shown to suggest a biphasic fragment pattern produced by the MNase. c An aliquot of the same arrays was digested with MNase for 7 min to obtain pure core particle length DNA. An aliquot of the cleavage product was digested with EcoRI and both samples were loaded on the 1.7% agarose gel (lane1 core particle DNA, lane2 EcoRI digestion) together with a 20 bp ladder. Both gels were stained with ethidium. d Densitometric trace of lane2 in c calculated with NIHimage software (version 1.63) after background subtraction DNA-sequencing verified the tandem orientation of the repetitive AS-monomers. Interestingly, all sequenced constructs so far obtained by self-ligation of AS-dimers or of AS-tetramers, contained tandemly oriented repeats (a total of 12 clones with various repeat numbers were sequenced), potentially due to an intolerance of the bacteria to inverted repeats. The formation of odd or even numbers of repeated monomers after self-ligation indicates that entire AS-monomer units can be excised during bacterial growth. The gel in Fig. 1b also demonstrates that bacterial cell lines can contain a distribution of different AS repeat numbers centred around one prevailing length. Growing the bacteria for several days showed the stable maintenance of this distribution as well as the prevailing repeat number (data not shown). Preliminary results show that this distribution can be also used to generate successively longer repeats by isolating the largest fragments from gels (not shown). Gel electrophoretic characterisation of chromatin reconstituted with 172-α7 A proper reconstitution of nucleosomes as well as a regular nucleosomal periodicity can be identified by a digestion with MNase. The cloned AS heptamer (172-α7) was reconstituted with histones at an octamer to DNA weight ratio of 1.0. The chromatin was treated with MNase for three different periods of time, deproteinised and electrophoresed. Figure 2a shows the expected protection of the nucleosomal core particle DNA, which is not observed when naked DNA was digested. Larger fragments were observed with lengths that are multiples of the AS-monomer length, suggesting the existence of regularly spaced nucleosomes. This ladder successively disappears with digestion time. On the other hand, a sequence-specific cleavage of the AGM AS may likewise have produced the ladder (Horz et al. 1983). The magnified area of the gel (Fig. 2b) indeed suggested the presence of two phases and also a considerable smear superimposed the bands. In order to analyse nucleosomal binding further, pure core particle length DNA produced by MNase cleavage of the same reconstituted material was digested with EcoRI restriction endonuclease. This enzyme was chosen on the basis of earlier results that have localised the EcoRI cleavage site near the centre of a major nucleosomal positioning frame in vivo and in vitro (Musich et al. 1982; Neubauer et al. 1986; Zhang et al. 1983). The agarose gel in Fig. 2c and the densitometric trace in Fig. 2d demonstrate that EcoRI has produced discrete bands, indicating that the nucleosomes have occupied preferred sites along the 172-α7 DNA. The largest fragment (Fig. 2c, I) corresponds to core particle length (uncleaved) DNA, which must have originated from nucleosomes formed on one or different positions that do not incorporate the EcoRI site. Additionally, three smaller fragments were produced, one of which with size ∼85 bp prevailed (Fig. 2c, III). The size ∼65 bp of another fragment (IV), when added to the 85 bp, approximately yields core particle length DNA. We therefore assume that these two fragments were produced by cleavage of the same nucleosomal core DNA. In support of this, the intensities of these two bands (Fig. 2d) differ by a factor of 1.5, which roughly would be expected based on the molecular weight difference (factor of 1.3, assuming the 85 and 65 bp). Thus, the appearance of these fragments indicates a preferred nucleosomal binding frame, whose centre is located about 10 bp away from the EcoRI site, in good agreement with the earlier results (Musich et al. 1982; Neubauer et al. 1986; Zhang et al. 1983). Next, the gel indicates a second weaker positioning site by the presence of a ∼120 bp fragment (Fig. 2c, II). Here, however, the corresponding smaller fragment (expected size 27 bp to yield core particle length DNA) was faint. Based on the densitometric analysis, 50% of the nucleosomes were bound at the most preferred site. Note that the major site occurs six times in the 172-α7 DNA and is interrupted at the fragment ends (since, in our case, monomers start and end at the EcoRI site). AFM imaging of 172-α7 nucleosome arrays The gel results support an at least partially regular nucleosomal organisation on the 172-α7 template. It remains unclear, however, to what extent a fraction of the nucleosomes were positioned randomly. Therefore, AFM was used to determine distances between consecutive nucleosomes assembled along the 172-α7 AS heptamer DNA. As a control, nucleosome arrays were reconstituted on a hexamer repeat of 208 5S rDNA to obtain a reference with well-documented properties. Tandem repeats of a 208 bp long DNA containing 5S rDNA of L. variegatus are often used as a model system for evenly spaced nucleosome arrays (Carruthers et al. 1998; Simpson et al. 1985) and their distribution of internucleosomal distances had been determined by AFM (Allen et al. 1993; Mechelli et al. 2004; Nikova et al. 2004; Yodh et al. 2002). Scanning was done on air dried samples which ensured that all nucleosomes in the arrays were clearly distinguishable. The chromatin was immobilised at conditions that trap DNA on the surface irreversibly (Bussiek et al. 2003). We believe (and provide evidence) that this prevents the sample from extensive structural distortions once bound. Figure 3a gives an example of the scans obtained and Fig. 3b–i shows zooms of individual nucleosome arrays reconstituted with 172-α7 (B-G) and with 208-6 DNA (H, I). Adsorption of the arrays to the surface under low salt conditions yielded the expected extended conformations (Fig. 3b–e; H and I) and more compact conformations were observed in the presence of MgCl2 (Fig. 3f, g). Visual inspection suggested spaced nucleosomes on both sequences, although frequently nucleosomes appeared closely stacked on each other (arrows in Fig. 3). Height measurements of those particles identified as nucleosomes yielded 2.6(±0.7 SD) nm, in agreement with previous measurements under the same scanning conditions in air (Bussiek et al. 2005). All particles with heights below 1.9 nm were rejected in order to ensure that only properly folded nucleosomes were considered in the data analysis. The DNA was not fully saturated with nucleosomes since we expected to identify ‘gaps’ corresponding to unoccupied repeats, if preferred nucleosomal binding sites existed. For comparison, arrays were also scanned in liquid, which yielded the same extended shapes in low salt buffer (not shown). A quantitative analysis after scanning in liquid was, however, not useful because nucleosomes and DNA were not imaged with the required sharpness. On contrary, scanning of mononucleosomes in liquid yielded high quality images, as shown below. We explain this difference by a more unstable scanning process of nucleosome arrays, because a local conglomeration of nucleosomes has the effect of a locally higher surface roughness. Fig. 3a–i Atomic force microscopy imaging in air of nucleosome arrays. The chromatin was reconstituted at a nominal octamer to DNA weight ratio of 0.5 and adsorbed onto the surface in Hepes-NaOH buffer (10 mM, pH 8.0). a Overview scan of 172-α7 arrays adsorbed in the buffer additionally containing 10 mM NaCl. b–g Zooms of individual 172-α7 arrays [adsorption buffer + 10 mM NaCl (b, c), no additional salt (d, e) and 10 mM NaCl, 2 mM MgCl2 (f, g)]. The arrows mark nucleosomes in a close contact. h, i Examples of 208-6 nucleosome arrays adsorbed in buffer + 10 mM NaCl. k Linear compaction of the DNA with number of NCPs per array (nncp). Blue, closed symbols, continuous line indicate 172-α7 arrays in buffer + 10 mM NaCl, blue, open symbols, dashed line indicate 172-α7 arrays in buffer without additional NaCl, green indicates 208-6 arrays in buffer + 10 mM NaCl, red indicates 172-α7 arrays in buffer + 10 mM NaCl, 2 mM MgCl2. Error bars correspond to standard errors Internucleosomal distances were measured in a way that allowed a direct comparison with the expected nucleosomal repeat lengths of 172 and 208 bp. This was accomplished by determining the wrapped nucleosomal DNA length for the given experimental conditions. Low salt concentrations during the surface adsorption were selected so that nucleosomes formed extended conformations. As a consequence, linker DNAs could be well identified allowing for precise length measurements. However, it has to be taken into account that, due to repulsive electrostatic forces under low salt conditions, core particle DNA may partially unwrap from the histone surface to become linker DNA (Hamiche et al. 1996), a phenomenon that we also observed under the present experimental conditions (Bussiek et al. 2005). We therefore estimated the average length of wrapped DNA (Lncp) from the linear compaction of the DNA by complexing into nucleosomes. The 172-α7 and the 208-6 nucleosome arrays were deposited onto the mica surface under the same low salt conditions (10 mM Hepes-NaOH, pH 8.0, 10 mM NaCl). All arrays with an unambiguous number and order of nucleosomes were included in these length measurements (a total of 106). Note that Fig. 3f and g give examples of excluded arrays. These are shown to demonstrate the expected compaction in the presence of Mg2+. Figure 3k shows that the nucleosome arrays formed on the DNA templates indeed shortened linearly with the number of nucleosomes nncp. Lfree measured as described in Sect. “Materials and methods” rather than the entire array length was plotted versus nncp to be able to include the length of the naked DNA fragments as data points for the calculation of linear regressions (433 nm for 172-α7 DNA and 442 nm for 208-6 DNA). The slopes of the calculated regression lines, giving the wrapped DNA length, were about identical for the two nucleosomal arrays and removal of the NaCl from the buffer of one other sample had no effect, thus all three values obtained under these low salt conditions were averaged. This yielded a wrapped DNA length Lncp = 42 nm (or 123 bp for 0.34 nm/bp), significantly less than the known core particle length of 147 bp. A wrapped DNA length of ∼120 bp could be confirmed independently by measurements of mononucleosomes (see below). To further control the measurement accuracy, the wrapped DNA length was determined for a sample prepared in the presence of MgCl2, which yielded 48 nm (or 141 bp), very close to the core particle length. Next we determined the distribution of the centre-to-centre (cc) internucleosomal distances (Fig. 4). The data reveal a non-random nucleosomal positioning along the 172-α7 DNA. This is indicated by a clearly separated initial main peak followed by a recurring preference for a larger distance, consistent with the existence of preferred binding sites, a fraction of which was unoccupied (Yodh et al. 2002). The distribution includes three different samples with a combined average number of nucleosomes nncp-av = 4.2. Next, the cc-distances were plotted for only those complexes carrying less than five nucleosomes to obtain a subset sample with more unoccupied binding sites. Figure 4b shows that the first recurring peak now is more accentuated and sharper compared to the distribution in Fig. 4a. Even a second and a third recurring peak are suggested, but not statistically secure. Fig. 4Distributions of cc-distances in 172-α7 and 208-6 nucleosome arrays, measured by AFM after adsorption under low salt conditions. acc-distances in 172-α7 arrays measured in three samples with a combined average number of NCPs per array nncp-av = 4.2. bcc-distances for exclusively 172-α7 nucleosome arrays with number of NCPs nncp <5, corresponding to nncp-av =3.2. ccc-distances measured in 208-6 nucleosome arrays (nncp-av =3.2). Numbers and arrow heads indicate cc-distances expected for a uniform nucleosomal phasing with the monomer lengths of 172 bp (a, b) and 208 bp (c), as calculated assuming Eq. (2). a is the number of unoccupied repeats between successive nucleosomes. The distance distribution begins at a value of 11 nm, the diameter of the NCP, meaning that a few measured distance values smaller than 11 nm were added to the first bin. The bin width of 12 nm corresponds to the length difference between monomer lengths of 172 and 208 bp (36 bp). Error bars indicate the square root of the number of observations in each bin For comparison, nucleosome arrays were reconstituted with 208-6 DNA, expected to produce defined regular nucleosome arrays due to positioning in the 5S rDNA sequence (Carruthers et al. 1998; Meersseman et al. 1991; Yodh et al. 2002) and thus also serving as a control for appropriate experimental conditions. The reconstitution with the 208-6 DNA yielded an average number of nucleosomes that exactly matched the subset sample in Fig. 4b (nncp-av = 3.2). Figure 4c shows a sharply defined preference for cc-distances around the third bin (35–47 nm) for this sample. The peak, as also seen in Fig. 4a, is differentiated from a recurring peak and additionally from a distinguishable preference for very small cc-distances (those found in the first bin). These correspond to the many contacting nucleosome pairs. Close contacts in the 172-α7 arrays are as well reflected in a high number of counts in the first bin in Fig. 4a although not as a separated peak. Examples can be seen in Fig. 3b–d. Importantly, positioning in the 5S rDNA led to a sharp initial peak, consistent with other reported AFM data (Mechelli et al. 2004; Nikova et al. 2004; Yodh et al. 2002). On the basis of the average wrapped DNA length (Lncp ∼120 bp), expected cc-distances (ccexp) for monomer lengths (Lmono) of 172 and 208 bp (58 and 71 nm, respectively) can be calculated assuming that where a is the number of unoccupied repeats between successive nucleosomes. These distances and their positions along the x-axes are indicated by numbers and arrowheads in Fig. 4, indicating a good agreement with measured (and assignable) peak maxima. In the case of a uniform phasing along the AGM AS repeats, the initial peak for the 172-α7 nucleosome arrays should be centred at about 30 nm (in the second bin) and not exceed values of ∼40 nm, taking into account that variation in the wrapped DNA length likely broadens the cc-distance distribution. For an estimate of this variation we can assume a standard deviation of ±28 bp (corresponding to ±9 nm), as determined in 172-α2 mononucleosomes (see below). The observed distances clearly exceed this boundary and actually show a high occurrence of internucleosomal separations of up to three times that expected for uniformly phased nucleosomes. This is different in the 208-6 nucleosome arrays: here, the width of the initial main peak suggests that most nucleosomes are bound at or near the same positioning site in the 5S rDNA repeats (neglecting close contacts). It is known for the the 5S rDNA that alternative positioning sites around the main site exist (e.g. Meersseman et al. 1991). The comparison of both sequences shows that multiple positioning in the 172-α7 DNA leads to a higher variation between internucleosomal spacings, which can be due to a higher proportion of nucleosomes positioned out of the main phase and by larger distances between the alternative sites. Apart from this observation, the clear separation of the main and first recurring peaks observed for both sequences indicates clearly the strong sequence-dependence of the nucleosome array formation on repetitive AGM-AS. AFM imaging of 172-α2 mononucleosomes To further quantify the nucleosomal positioning along the AGM AS, mononucleosomes were reconstituted with the AS dimer containing 172-α2 DNA (415 bp). Here, the complexes were imaged in liquid (10 mM Hepes-NaOH, pH 8.0, 10 mM NaCl) and the mica surface was modified with the same amount of polylysine that was used for immobilising the nucleosome arrays to ensure constant adsorption conditions. Figure 5a shows a representative overview image displaying different kinds of complexes: mononucleosomes, a few dinucleosomes and uncomplexed DNA. Figure 5b–e shows zooms of individual complexes with different nucleosomal positions along the 172-α2 fragment. Fig. 5Atomic force microscopy imaging in liquid of mononucleosomes reconstituted with 172-α2 DNA. Scanning conditions were 10 mM Hepes-NaOH, pH 8.0, 10 mM NaCl. a Overview scan of a sample reconstituted at a nominal octamer to DNA weight ratio of 0.5. b–e Zooms of individual mononucleosomes. f Height distribution of nucleosomal particles. Frequencies were fitted with the sum of two Gaussian functions to derive mean height values (see text). g Distribution of wrapped nucleosomal DNA lengths Lncp, analysed separately for particles <3.0 nm height (white bars) and ≥3.0 nm height (black bars). Mean values and standard deviation of Lncp were calculated by fitting Gaussian functions to both datasets (see text). For comparison with Fig. 3, note that measured height values are larger in liquid than in air, presumably due to reduced sample–probe interactions We first characterised the heights of the mononucleosomes. The height values were distributed into two discrete peaks centred at 2.7 and 4.5 nm (Fig. 5f). The larger peak value is only slightly smaller than the expected height of 5.0 nm for the NCP and also agrees with our previous height measurements under the same scanning conditions (Bussiek et al. 2005). Seventy-four percent of the small 2.7 nm particles were found at DNA termini (an example is given in Fig. 5e). The sample was divided into two subsets, one containing the particles with less than 3.0 nm, the other those above 3.0 nm height. The wrapped DNA lengths (Lncp) of both subsets were determined after contour length measurements of the mononucleosomal fragments: the lengths of free DNA outside the nucleosomes were estimated according to Eq. (1). These lengths were converted to base pairs, taking into account that measured DNA lengths on AFM images can deviate slightly from the expected length of 0.34 nm/bp on PL-mica (Bussiek et al. 2003). A measurement of uncomplexed 172-α2 DNAs, present in the same sample preparation, gave a mean contour length of 134.4 ± 4.3 (SD) nm or 0.32 nm/bp. For each individual complex, the free DNA length in base pairs was subtracted from the total fragment length of 415 bp to obtain Lncp. The small and the regular particles can be as well discriminated in terms of this length, since both samples yielded two distinct distributions of Lncp (Fig. 5g). The peak centres of these distributions were at 119 ± 28 (SD) bp for the larger particles (≥3.0 nm height) and only 84 ± 37 (SD) bp for the smaller ones (<3.0 nm height) (P < 0.0001). Importantly, the length of 119 bp is in close agreement with the 123 bp determined for the 172-α7 and the 208-6 nucleosome arrays under the same salt conditions. This also provides some evidence that drying the arrays did not extensively distort the nucleosome conformation which we attribute to the irreversible trapping of the DNA to the PL-mica so that the nucleosomes become “fixed” after the binding to the surface. Nucleosomal positions along the 172-α2 AS dimer DNA were determined exclusively for mononucleosomes with particle heights larger than 3.0 nm. Positions were quantified as the ratio r, i.e. the distance from the nucleosomal centre to the nearest DNA terminus divided by the total mononucleosomal fragment length (Kepert et al. 2003). The distribution of r shown in Fig. 6 clearly reveals that nucleosomes are preferably localised around the fragment centre, indicated by a defined peak at r = 0.4–0.5. An r-value of 0.5 ± 0.05 corresponds to a nucleosomal frame whose centre is located next to the EcoRI cleavage site at sequence position 37 (±20) bp (see Fig. 1). Thus, taking account the error of r, the AFM measurement reveals the same prevailing binding site that was estimated by EcoRI cleavage (nucleosomal centre about 10 bp away from the EcoRI site, see above). The remaining nucleosomes (55%) were rather evenly distributed over the eight other bins (with some additional weak minor preferences) and there was no preference for regular size nucleosomes to be aligned with DNA termini. Fig. 6Positioning of nucleosomes along 172-α2 DNA. Different nucleosomal positions near the termini that would correspond to r-values between 0 and 0.1 could not be distinguished due to the resolution limit. Therefore, the occurrences of r-values equal to 0 were evenly distributed over the first two bins (white bars). The error of r in the peak range approximately equals the bin size, since an error Δr ∼ 10% could be expected based on the measured standard deviation of the total mononucleosomal fragment length (± 7.7%) Discussion Nucleosomal positioning along repetitive AGM AS We reconstituted nucleosomes with dimers (172-α2) and heptamers (172-α7) of repetitive AGM AS by the salt dialysis method. Enzymatic digestions and atomic force microscopy were used to characterise the nucleosomal organisation along this DNA. The results demonstrate the formation of a well-defined (sequence-dependent) nucleosomal arrangement, essentially similar to the 5S rDNA model system. In the AGM AS arrays, however, the variation in the separation between neighbouring nucleosomes was rather large, and evenly spaced nucleosomes occurred only to a limited extent. A partial digestion of 172-α7 nucleosome arrays with micrococcal nuclease (MNase) yielded a fragment ladder documenting the presence of phased rather than random nucleosome positions on 172-α7 DNA. This result is very similar to MNase digestions of AGM AS chromatin in cell nuclei (Musich et al. 1982). Core particle length DNA derived from 172-α7 nucleosome arrays was further digested with EcoRI. This produced discrete secondary fragments and thus verified the occupation of preferred nucleosomal binding sites. One of these sites was prevailing with a position consistent with published results (Musich et al. 1982; Neubauer et al. 1986; Zhang et al. 1983). One additional minor frame could be detected, which may explain a suggested biphasic cleavage pattern produced by the MNase. Musich et al. (1982) concluded that the MNase digestion of AGM AS chromatin depends on the nucleosomal organisation which includes the possibility that a sequence-specific cleavage coincidentally occurs in the internucleosomal linkers between nucleosomes bound at the mostly preferred site (Horz et al. 1983; Musich et al. 1982). This may lead to a selective representation of the fraction of phased nucleosomes and obscure deviations thereof. Although this source of error must be taken into account, the similar results for cellular AGM AS chromatin and reconstituted nucleosome arrays indicate similar properties, i.e. at least fractions of the nucleosomes are regularly arranged. Atomic force microscopy imaging was used to detect the spacing between all neighbouring nucleosome pairs in a sample. This circumvents possible artefacts caused by enzymatic analyses, although the precision of localising the nucleosomes is lower. Centre-to-centre (cc) internucleosomal distances measured by AFM imaging not only depend on nucleosomal positions relative to the DNA sequence but also on internucleosome interactions (which should be reduced under the low salt conditions in our measurements) and variations in the amount of wrapped nucleosomal DNA. An example is given by increasing cc-distances due to a release of nucleosomal DNA from the histone surface after removal of linker histones (Zlatanova et al. 1998). Especially the wrapped DNA length is susceptible to experimental conditions: a prevailing cc-distance of ∼20 nm was previously measured in 5S positioning sequence repeat (208-12) nucleosome arrays that were adsorbed to naked mica in the presence of Mg2+ (Nikova et al. 2004). These measurements used arrays not fixed with glutaraldehyde. Fixed 208-12 nucleosome arrays adsorbed to either spermidine- or AP-mica under low salt conditions gave cc-distances of ∼30 nm (Mechelli et al. 2004; Yodh et al. 2002). Our measurements in low salt conditions using PL-mica and unfixed material yielded a larger cc-distance for the 208-6 nucleosome arrays of 35–47 nm. An estimation of the wrapped DNA length at our conditions yielded ∼120 bp, corresponding to ∼1.4 superhelical turns around the octamer [full core particle length of 147 bp equals 1.7 turns (Davey et al. 2002; Furrer et al. 1995; Hamiche et al. 1996; Luger et al. 1997)]. The expected cc-distance for a nucleosomal repeat length of 208 bp on the basis of the 120 bp of wrapped DNA is ∼40 nm, in agreement with the results obtained for the 208-6 arrays. The expected distances for additional unoccupied repeats matched the position of recurring peaks in the cc-distance distributions measured for the 208-6 and the 172-α7 nucleosome arrays. Essentially, the shapes of the cc-distance distributions for the 208-6 and the 172-α7 DNA shown in Fig. 4b and c are very similar, which points to a sequence dependence of the nucleosomal organisation along repetitive AGM AS similarly strong as on the 5S sequence. However, in the case of the 172-α7 DNA, the initial peak of the cc-distance distribution, corresponding to the distance between a nucleosome and its nearest neighbour located on an adjacent repeat, is significantly broader and shifted to larger distances than could be expected for evenly spaced nucleosomes. The distributions still show a high frequency of distances up to ∼60 nm (corresponding to ∼145 bp of linker DNA length), almost three times the linker length expected for a uniform phasing with the AGM AS repeats (18 nm or 52 bp, based on the 120 bp of wrapped nucleosomal DNA). Thus, many nucleosomal positions were separated by up to about 1.5 AS monomer lengths. The loading with nucleosomes did not significantly affect this result (compare Fig. 4a, b). 208-6 nucleosome arrays showed the expected distance distribution for nucleosomes predominantly bound at and near the main phase, showing that the broad initial peak seen in the 172-α7 arrays was not simply due to a limited measurement accuracy or inappropriate reconstitution or sample preparation conditions. Thus, a strictly regular nucleosomal arrangement occurred in the AGM AS only to limited extent and variation in possible internucleosomal spacings is rather large. We conclude that the number of alternative sites is nevertheless small, because the gelelectrophoresis showed two identifiable sites and the initial peak in the cc-distance distribution is sharply separated from the first recurring peak. The mapping of nucleosomal positions by AFM imaging of the 415 bp long 172-α2 mononucleosomes was fully consistent with the occurrence of the known major positioning site, in addition to the identification of the same site in the nucleosome arrays by EcoRI digestion. The frequency of nucleosomes found at or near this major frame (45%) is lower than found by in vitro reconstitutions using a 250 bp AGM AS fragment (80%) (Neubauer et al. 1986), which is explained by the length difference of the DNA fragments used. The frequency is closer to the observed occupancy of the major site in vivo (35 %) (Zhang et al. 1983) and also closer to that estimated by gelelectrophoretic quantification of the EcoRI digested mononucleosomal DNA derived from the 172-α7 arrays (upper limit of 50%). The DNA fragment used here for the reconstitution of the mononucleosomes is the minimal length that contains the AGM AS sequence uninterrupted, such that all possible binding sites are present. However, when comparing the positioning observed here and in vivo, it must be taken into account that a few possible binding sites occur twice in the 172-α2 (i.e. those which do not incorporate the EcoRI site) and that some additional sites are present that include vector DNA at the ends (33 and 37 bp, respectively). In general, positioning on small DNA fragments can be influenced by a preference for DNA-ends (which was not observed here, as shown in Fig. 6) and the length of the fragment used. A surprising feature was a bimodal height distribution of mononucleosomal particles reconstituted with the 172-α2 DNA. The two height values were correlated with distinguishable wrapped DNA lengths. Nikova et al. (2004) obtained a very similar height distribution by AFM imaging of subsaturated 208-12 nucleosome arrays and presented evidence that particles with only ∼75 bp of wrapped DNA and ∼2.5 nm height correspond to sub-nucleosomal particles that do not contain the full histone octamer. In the present study, small particles of similar dimension (wrapped length of 85 bp and 2.7 nm height) were predominantly located at the termini of the 172-α2 DNA (74%). In contrast, no preferential binding to termini was found for nucleosomes of regular size (4.5 nm height). We may speculate that this high frequency of the small particles at the fragment termini is as well associated with sequence properties of the AGM AS. In our case, the AS sequence starts and ends at the EcoRI site, which is located near the centre of the major binding site. This implies that the end regions of the fragment contain only parts of this binding site. Assuming that multiple DNA–histone interaction along the entire core particle length are responsible for the binding specificity (Neubauer et al. 1986), it may be possible that histones interact preferentially even with parts of the major site. Then, a complete nucleosome assembly could be impeded at the fragment termini due to a lack of sufficient amounts of DNA in the required direction, leading to smaller particles. Relevance of AGM AS for centromeric chromatin structures The kinetochore protein CENP-B is supposed to be involved in nucleosomal positioning in human AS (Tanaka et al. 2005). A structural motive consisting of the AS monomer, nucleosomal positioning and founding kinetochore protein binding may be involved in specifying the location of centromeric function. CENP-B is not present at the centromeres of AGM and therefore AS arrays of this species might compensate for the absence of CENP-B by influencing nucleosomal positioning by their sequence alone (Goldberg et al. 1996). We show here that the AGM AS indeed has a profound influence on the arrangement of oligonucleosomal arrays. This arrangement has similarities to the 5S rDNA model system, although multiple positioning leads to larger variation of internucleosome spacings in the AGM AS. The spacing of nucleosomes is expected to contribute to the higher order structuring of chromatin, therefore it is of interest, for example with regard to the design of particular synthetic chromatin structures, to what extent the in vitro produced nucleosomal arrangement along a DNA template resembles that occurring in vivo. Our MNase digestions suggested a high fraction of evenly spaced nucleosomes, similar to digestions of AGM AS in cell nuclei (Musich et al. 1982) whereas the AFM results are consistent with more detailed studies in vivo which demonstrated the occupation of different minor binding sites in addition to the major site (Zhang et al. 1983). The occupancy of the major binding site was in the range of 45–50% in our analysis which is also in a quite good agreement with the result of the cited in vivo work (35 %). This may further support that the AS sequence contributes significantly to the determination of AS chromatin and probably to centromeric chromatin structures. The DNA constructs built in this study will give the opportunity to investigate further the structure and biophysical properties of AS chromatin and centromeric chromatin. For example, it will be important to compare the nucleosomal arrangement on repetitive AGM and human AS (with and without CENP-B boxes). Nucleosomal phasing also occurs on human AS sequences (Tanaka et al. 2005; Yoda et al. 1998) and a preferred positioning site found in immunoprecipitated human centromeric AS indeed corresponded to that occurring in AGM AS (Vafa and Sullivan 1997).

Oecologia-3-1-1915600

Soil microorganisms control plant ectoparasitic nematodes in natural coastal foredunes

Belowground herbivores can exert important controls on the composition of natural plant communities. Until now, relatively few studies have investigated which factors may control the abundance of belowground herbivores. In Dutch coastal foredunes, the root-feeding nematode Tylenchorhynchus ventralis is capable of reducing the performance of the dominant grass Ammophila arenaria (Marram grass). However, field surveys show that populations of this nematode usually are controlled to nondamaging densities, but the control mechanism is unknown. In the present study, we first established that T. ventralis populations are top-down controlled by soil biota. Then, selective removal of soil fauna suggested that soil microorganisms play an important role in controlling T. ventralis. This result was confirmed by an experiment where selective inoculation of microarthropods, nematodes and microbes together with T. ventralis into sterilized dune soil resulted in nematode control when microbes were present. Adding nematodes had some effect, whereas microarthropods did not have a significant effect on T. ventralis. Our results have important implications for the appreciation of herbivore controls in natural soils. Soil food web models assume that herbivorous nematodes are controlled by predaceous invertebrates, whereas many biological control studies focus on managing nematode abundance by soil microorganisms. We propose that soil microorganisms play a more important role than do carnivorous soil invertebrates in the top-down control of herbivorous ectoparasitic nematodes in natural ecosystems. This is opposite to many studies on factors controlling root-feeding insects, which are supposed to be controlled by carnivorous invertebrates, parasitoids, or entomopathogenic nematodes. Our conclusion is that the ectoparasitic nematode T. ventralis is potentially able to limit productivity of the dune grass A. arenaria but that soil organisms, mostly microorganisms, usually prevent the development of growth-reducing population densities. Introduction Root herbivores play an important role in shaping the composition of natural plant communities (Brown and Gange 1990). Nematodes and insects represent the vast majority of the belowground herbivores (Brown and Gange 1990; Stanton 1988). Nematodes are more abundant than soil insects, and in some grassland ecosystems, nematodes are the dominant herbivores (Ingham and Detling 1986). Root-feeding nematodes have been estimated to take up as much as one quarter of the net primary production of grassland vegetation (Stanton 1988), and they affect plant quality (Davis et al. 1994; Troelstra et al. 2001), plant diversity, and vegetation succession (de Deyn et al. 2003). Root-feeding nematodes can also indirectly affect plant performance by their influence on bottom-up and top-down control of aboveground invertebrate herbivores (Bezemer et al. 2005). However, in spite of the increasing knowledge on the significant role of belowground herbivores in the control of plant abundance and plant community composition, relatively few studies have investigated which factors control the abundance of the belowground herbivores in natural ecosystems (Strong et al. 1996, 1999). Herbivore abundance can be influenced by natural enemies (top-down), by the host plant (bottom-up), and by competition with other herbivores (horizontal control). In (semi) natural ecosystems, most studies on the control of root-feeding nematodes have focused on plant quality (Yeates 1987), interspecific competition (Brinkman et al. 2004, 2005), plant community composition (de Deyn et al. 2004), plant succession and soil conditions (Verschoor et al. 2002), and mycorrhizal fungi (de la Peña et al. 2006). Soil food web models assume root-feeding nematodes to be controlled by carnivorous nematodes and microarthropods (Hunt et al. 1987; Neutel et al. 2002). However, most biological control studies in agricultural systems focus on managing nematode abundance by parasitic soil microorganisms (Kerry 2000; Sikora 1992) or mycorrhizal fungi (Hol and Cook 2005), suggesting that root-feeding nematodes are mainly controlled by microorganisms. Therefore, previous studies show little agreement and do not clearly predict how root-feeding nematodes will be controlled in natural ecosystems. Empirical data for top-down mechanisms are rare for terrestrial ecosystems relative to the many studies in aquatic systems (Walker and Jones 2001). In general, trophic cascades have been argued to be less common on land than in water (Polis and Strong 1996). Nevertheless, there is empirical evidence supporting the existence of trophic cascades in terrestrial plant–predator–prey systems (Schmitz et al. 2004). Tritrophic systems of plants, aboveground insect herbivores, and their natural aboveground enemies are the best-studied terrestrial examples of top-down and bottom-up herbivore controls (Carson and Root 1999; Rosenheim 1998). Below ground, tritrophic interactions may not essentially differ from what is known above ground (Bezemer and van Dam 2005), although rates of dispersal of organisms and chemical compounds will be lower than is mostly the case above ground (Rasmann et al. 2005; van der Putten 2003). Therefore, the challenge is, similar to that above ground (Schmitz et al. 2004), to assess what controls the abundance of root herbivores. This knowledge will enhance our understanding of belowground multitrophic interactions and their influences on plant performance and plant community composition. In the present study, the role of microarthropods, nematodes, and microorganisms in controlling the abundance of the root-feeding nematode Tylenchorhynchus ventralis (Loof 1963) Fortuner and Luc (synonym Telotylenchus ventralis) was experimentally compared. This nematode is a polyphagous ectoparasite, which means that it is a quite generalistic root feeder that penetrates outer cortical cells with its stylet to collect and ingest cell contents (Yeates et al. 1993). T. ventralis is a root parasite of the dominant coastal foredune grass Ammophila arenaria (Marram grass). In field soil, T. ventralis reaches densities that are 80 times lower than achieved when inoculated into sterilized dune soil (de Rooij van der Goes 1995). Whereas T. ventralis can strongly reduce growth of A. arenaria in sterilized soil, field densities in nonsterilized soil are too low to directly influence plant performance (de Rooij van der Goes 1995). The roots of A. arenaria are parasitized by an array of herbivorous nematodes ranging from ectoparasites to sedentary endoparasites (de Rooij van der Goes et al. 1995). The control mechanisms of root herbivorous nematodes in dunes appear to highly depend on the feeding type of the nematode, and even on the species of nematode. Whereas the sedentary root knot nematode Meloidogyne maritima (Jepson 1987) Karssen, van Aelst and Cook is controlled by competition (Brinkman et al. 2005), the sedentary cyst nematode Heterodera arenaria (Cooper 1955) Robinson, Stone, Hooper and Rowe appears to be controlled by bottom-up processes (van der Stoel et al. 2006). The migratory endoparasitic root lesion nematode Pratylenchus penetrans (Cobb 1917) is controlled by arbuscular mycorrhizal fungi (de la Peña et al. 2006). Thus far, the factors that control the ectoparasitic nematode T. ventralis associated with A. arenaria are unknown. Previous studies showed bottom-up control of A. arenaria to occur only when the plants were severely growth reduced (de Rooij van der Goes et al. 1995). Alternatively, competition with cyst and root lesion nematodes is a potential factor controlling ectoparasitic nematodes (Eisenback 1993). However, endoparasitic nematodes did not control abundance of T. ventralis (Brinkman et al. 2004). In the present study, the top-down factors that may be involved in the control of T. ventralis populations were investigated in order to determine how belowground trophic interactions might influence plant performance and vegetation composition. To assess the top-down control of T. ventralis, three experiments were performed. The aim of experiment 1 was to elucidate the potential top-down control of T. ventralis by the dune soil community. In experiment 2, the particular role of microorganisms was investigated by selective elimination of soil fauna (nematodes and microarthropods). In experiment 3, the hypothesis that emerged from experiment 2, that soil microorganisms are the main cause of top-down control of T. ventralis, was tested. Here, we applied Koch’s postulates by collecting microorganisms, nematodes, and microarthropods from dune soil and adding them to sterilized soil inoculated with T. ventralis. New evidence that top-down control by soil microorganisms is the most important factor controlling the abundance of ectoparasitic nematodes in dune soil is presented and discussed. Materials and methods Soil In summer 2003, soil samples were collected from mobile and stable foredunes at Voorne, The Netherlands (Latitude 51°55′N to Longitude 04°05′E). The samples were collected along six transects parallel to the beach and 50 m apart. At each sampling station in the mobile and stable dune, 60 kg of soil was collected from the youngest root zone of A. arenaria. The soil was sieved (0.5-cm mesh size) to remove plant parts and debris and stored in plastic bags at 4°C until used (van der Stoel et al. 2002). Plants Seeds of A. arenaria were collected from the same foredune area and stored dry until used. In order to obtain seedlings, the seeds were germinated for 2 weeks on moist glass beads in a climate room at a 16/8 h light/dark regime at a temperature of 25/15°C, respectively. When the first leaf was 2–3 cm long, the seedlings were transplanted to 1.5-l plastic pots filled with 1,500 g of dune soil. In each pot, four seedlings of A. arenaria were planted, and the soil surface was covered with aluminum foil to protect the soil from desiccation. The soil moisture was adjusted to 10% w/w and maintained at this level throughout the experiment by weighing the pots twice a week and resetting their initial weight using demineralized water. Once a week, full-strength Hoagland nutrient solution was added at a weekly rate of 12.5 ml pot−1 for the first 3 weeks and then 25 ml pot−1, subsequently (Brinkman et al. 2004). This nutrient supply rate was effective to compensate for effects of nutrient release as a result of soil sterilization in dune soil (Troelstra et al. 2001; van der Putten et al. 1988). The experiments were carried out in a greenhouse at a day temperature of 21°C ± 2°C (day length 16 h) with additional light (to maintain a minimum of 225 μmol m−2 s−1 PAR with SON-T Agro lamps) and a night temperature of 16°C. These temperatures are comparable with summer conditions in the field and are optimal for both plant and nematode development (Troelstra and Wagenaar unpublished results). Experiments Experiment 1: multiplication of T. ventralis in sterilized and nonsterilized dune soil In this experiment, the effect of soil origin (mobile and stable dunes) and soil organisms on multiplication of the ectoparasitic nematode T. ventralis was tested. Half of the soil was sterilized by gamma irradiation at an average dose of 25 kGray, which eliminates microorganisms and nematodes effectively from dune soil (de Rooij van der Goes et al. 1998). One week after the seedlings of A. arenaria had been transplanted, half the pots were inoculated with 50 T. ventralis pot−1. The noninoculated pots served as controls for effects of T. ventralis on plant biomass production. There were six replicates of each treatment. Experiment 2: reproduction of T. ventralis in partially sterilized soil Multiplication of T. ventralis was studied in soils from which microarthropods and nematodes had been selectively removed by stirring the soil for 15 min at 1,500 rpm. This method has proven to effectively kill the soil fauna (de Rooij van der Goes et al. 1998). We confirmed this by inspecting the soil following stirring and found no live nematodes or microarthropods. The experiment was carried out as described above, but now the soil was completely sterilized by gamma irradiation (average 25 kGray), partially sterilized by stirring to remove the soil fauna, or nonstirred in order to have a nonsterilized control soil. Each soil was inoculated with 0, 25, or 250 T. ventralis pot−1 in order to examine any interaction between the effect of type of soil sterilization and nematode inoculation density. There were six replicates of each treatment. Experiment 3: reinoculation of microorganisms, nematodes, and microarthropods into sterilized soil with T. ventralis In order to completely apply Koch’s postulates, microarthropods, nematodes, and microorganisms were extracted from the soil of mobile and stable coastal foredunes and inoculated alone and in all factorial combinations into sterilized dune soil. Then, seedling plants of A. arenaria were grown as in the previous experiment, and every pot was inoculated with 50 T. ventralis. All treatments were carried out in six replicates. The microorganisms were obtained by shaking soil samples of 100 g with demineralized water (1:1 w/w) for 10 min and filtering the supernatant through a 20-μm mesh (Klironomos 2002). Prepared microbial filtrate contained no nematodes, but bacteria and fungi had readily passed through the filter. The pots with microorganisms were inoculated with 10 ml of the filtrate, which was 1/15 of the original soil density. For each pot, nematodes had been extracted from 1,500 g of nonsterile soil by Cobb’s method (Oostenbrink 1960) and added in a suspension of 10 ml pot−1, so that nematode inoculation density corresponded with the density of nematodes in field soil. The nematode community added to the pots was analyzed microscopically (magnification 200×) and consisted of plant parasites (T. ventralis, T. microphasmis, Pratylenchus spp, Paratylenchus spp., Meloidogyne spp., Rotylenchus spp., Criconematidae), bacterivores (Acrobeles spp., Acrobeloides spp., Chiloplacus spp., Cephalobidae, Plectus spp.), omnivores (Aporcelaimellus spp., Microdorylaimus spp.), and carnivores (Choanolaimus spp.). Microarthropods were collected from nonsterile dune soil by wet sieving through 180-μm mesh and added as 10 ml of suspension pot−1, which corresponded with the field density of microarthropods. Demineralized water was added to all pots in equal amounts. Assessing the presence of microbial enemies on nematodes in field soil In order to confirm whether microbial enemies may occur on T. ventralis in the field, we extracted mobile nematodes from 100 cm3 of the field soil from each of the sampling sites using an adaptation of the Tray method (Whitehead and Hemming 1965). Half of the resulting nematode suspension was inspected using an inverted microscope (magnification 200×), and the nematodes were checked for symptoms of infection by bacteria or fungi. Nematodes infected by fungi were picked from the suspension and transferred to a corn-meal agar plate with antibiotics to encourage sporulation (Smith and Onions 1994), making possible identification of fungi that were previously found in a vegetative state. Identification of fungal natural enemies was done by observing mycelia and spore structure morphology and comparing this with the descriptions of Barron (1977). Endospores of the parasitic bacterium Pasteuria spp. were recorded when observed attached to the nematode cuticle. Symptoms of infection by a nonlethal bacterial parasite Microbacterium nematophilum were assessed according to Sulston and Hodgkin (1988). To detect whether nematode natural enemies may occur as dormant forms in the soil, nematode-baited sprinkle plates were used. Soil (1 g) from each of the samples was sprinkled on water agar (1%) in a 9-cm-diameter Petri dish. A concentrated suspension of an estimated 500 Caenorhabditis elegans synchronized in the young adult stage (Sulston and Hodgkin 1988) was added to the plates. A negative control containing nematodes only in water agar (1%) was used. The plates were sealed, kept at room temperature, and observed after 2 weeks and subsequently at weekly intervals up to 5 weeks (Barron 1977). Identification of fungal natural enemies was done as described above. Harvest All three experiments were harvested 12 weeks after inoculation of T. ventralis, allowing this nematode to complete two reproductive cycles (de Rooij van der Goes 1995). The nematodes were extracted from soil by Cobb’s decantation method and from the roots using a mistifier (Oostenbrink 1960). The numbers of T. ventralis were counted using a microscope (magnification 200×) and expressed as numbers 100 g−1 of dry soil. The roots and shoots of A. arenaria were dried for 48 h at 75°C and weighed. Data analysis Normal distribution of data and homogeneity of variance were checked by inspection of the residuals after model fit (using the package Statistica 7). To obtain the normal distribution of data and homogeneity of variances, numbers of T. ventralis were log transformed in experiment 1 and square-root transformed in experiment 2. In all three experiments, the soil origin (stable or mobile dune) did not affect significantly (P > 0.05) the measured variables. Therefore, all data from treatments with those two soil origins was pooled, resulting in 12 replicates per treatment. Numbers of T. ventralis of experiment 1 were analyzed using one-way analysis of variance (ANOVA) with main the factor “soil treatment.” Two-way ANOVA with the main factors “soil sterilization” and “nematode inoculation” were performed for root and shoot biomass. Three-way ANOVA with the main factors “stirring,” “sterilization,” and “inoculation density” were performed for analyzing the numbers of T. ventralis shoot and root biomass in experiment 2. Experiment 3 was analyzed by three-way ANOVA with the main factors “invertebrates,” “nematodes,” and “microorganisms.” Treatments were compared by posthoc analysis using Tukey honestly significant difference (HSD) tests (P < 0.05). Results Experiment 1 The numbers of T. ventralis at harvest differed significantly between sterilized and nonsterilized soils (F2,33 = 77.9 and P < 0.001). In the nonsterilized soil, addition of T. ventralis resulted in a significant increase of numbers at the end of the experiment compared with nonsterilized, noninoculated soil (Fig. 1). However, there were five times more T. ventralis in the inoculated sterilized soil than in the inoculated nonsterilized soil (Fig. 1; P < 0.05; sterilized soil without T. ventralis added was not included because the nematodes were absent). These results show that multiplication of T. ventralis in nonsterilized soil was significantly enhanced by inoculation but that T. ventralis multiplication was significantly reduced by some factor in the nonsterilized soil that could be excluded by soil sterilization. Fig. 1Numbers of Tylenchorhynchus ventralis in 100 g of nonsterilized and sterilized dune soil 12 weeks after inoculation with T. ventralis. Error bars indicate standard error, and different letters above the bars indicate significant difference at P < 0.05 (experiment 1) Soil sterilization influenced shoot biomass more than did T. ventralis inoculation (F1,44 = 117, P < 0.001 for soil sterilization and F1,44 = 4.17, P < 0.05 for inoculation, Fig. 2), and the effect of T. ventralis inoculation depended on soil sterilization (F1,44 = 7.06, P < 0.05). Most shoot biomass was produced in sterilized soil, whereas T. ventralis inoculation significantly reduced shoot biomass (Fig. 2). As expected, the least shoot biomass was produced in nonsterile soil; however, addition of T. ventralis caused no further reduction in growth (Fig. 2). As expected, root biomass was also strongly influenced by soil sterilization (F1,44 = 56.1 and P < 0.001), whereas the effect of T. ventralis addition was greater than for shoot biomass (F1,44 = 16.8 and P < 0.001). As for shoot biomass, the effect of T. ventralis inoculation on root biomass depended on soil sterilization (F1,44 = 9.87 and P < 0.005), which reflects that shoot biomass was significantly reduced by T. ventralis inoculation in the sterilized soil only (Fig. 2). Fig. 2Shoot and root biomass of Ammophila arenaria in sterilized and nonsterilized soil after 12 weeks from inoculation with Tylenchorhynchus ventralis. Error bars and letters above indicate significant differences at P < 0.05 (experiment 1) Experiment 2 Significantly greater populations of T. ventralis developed in sterilized than in nonsterlilized soil at both inoculation densities (Tables 1 and 2). At the low-inoculation density, the number of the nematodes in nonsterliized soil was 30 times less than in sterilized soil and 15 times less at the high-inoculation density. There was no significant effect of soil stirring on the numbers of T. ventralis, although there was a trend (P = 0.06) that stirring reduced T. ventralis multiplication. Table 1Three-way analysis of variance (ANOVA) of the numbers of Tylenchorhynchus ventralis in nonsterilized and sterilized, and stirred and nonstirred dune soil at three inoculation rates (0, 25, 250 pot−1) after 12 weeks from inoculation to Ammophila arenaria. The data has been square-root transformed to achieve normal error distributiondfFPStirring (1)13.4810.06Sterilization (2)1137.41<0.001Inoculation density (3)295.55<0.0011 × 210.8580.361 × 321.2220.302 × 3243.94<0.0011 × 2 × 321.4990.23Error125Table 2 Effects of soil sterilization on numbers of Tylenchorhynchus ventralis in 100 g of soil [±1 standard error (SE)] 12 weeks from inoculation. Letters indicate significant differences at P < 0.05 (experiment 2)NonsterilizedSterilized0 T. ventralis added0.77 ± 0.22a0.08 ± 0.03a25 T. ventralis added12.3 ± 2.62b390 ± 73.03c250 T. ventralis added69.2 ± 16.76b,c1,162 ± 203.7d As expected, both soil stirring and sterilization influenced shoot biomass (Fig. 3; Table 3). Shoot biomass was greater in sterilized than in nonsterilized soil and in stirred than in nonstirred soil; however, inoculation density of T. ventralis did not influence the shoot biomass (Fig. 3). Root biomass was affected by stirring, soil sterilization, and inoculation density of T. ventralis, whereas effects of inoculation density depended on soil stirring as well as on soil sterilization (Table 3). If no nematodes were inoculated to the pots, soil sterilization almost doubled the root biomass. However, if inoculated with 25 or 250 T. ventralis per pot, the roots in sterilized soil with T. ventralis did not produce more biomass than those in nonsterilized soil (Fig. 3). Root biomass was significantly increased by soil stirring when no or few (25 pot−1) T. ventralis were added to the pots, but there was no increase in root weight at the high inoculation rate. Fig. 3Effects of soil stirring, soil sterilization, and addition of Tylenchorhynchus ventralis on shoot and root biomass of Ammophila arenaria. Error bars and letters above indicate significant differences at P < 0.05 (experiment 2)Table 3Shoot and root biomass of Ammophila arenaria 12 weeks after inoculation with nematodesdfShoot biomassRoot biomassFPFPStirring (1)114.13<0.0018.69<0.01Sterilization (2)130.49<0.000126.8<0.001Inoculation density (3)21.9980.13910.53<0.0011 × 213.2880.0721.910.1691 × 321.9160.1513.622<0.052 × 322.1450.1216.452<0.011 × 2 × 321.9210.1511.9110.152Error132The results of a factorial analysis of variance (ANOVA) with factors “stirring,” “sterilization,” and “inoculation density” (experiment 2) Experiment 3 The multiplication of T. ventralis numbers was significantly reduced by adding a mixture of soil nematodes; however, the effect of adding microorganisms was far greater (Fig. 4; Table 4). If microorganisms were present alone or in combination with other soil organisms, the number of T. ventralis was always less than when microorganisms were absent. On average, adding a suspension of nematodes reduced final numbers of T. ventralis by 15%, whereas adding microorganisms reduced numbers of T. ventralis by 55% (Fig. 4). Therefore, the effect of adding nematodes on T. ventralis multiplication was substantially weaker than the effect of microorganisms. Microarthropods did not have a significant effect on the numbers of T. ventralis. Adding microarthropods, nematodes, and microorganisms did not influence shoot or root biomass (P > 0.05; data not shown). Fig. 4The effects of mixed nematode inoculum and microorganisms on Tylenchorhynchus ventralis multiplication. Error bars and letters above indicate significant differences at P < 0.05 (experiment 3)Table 4The numbers of Tylenchorhynchus ventralis after 12 weeks from inoculationDegrees of freedom (df)F valueP valueMicroarthropods (1)10.00060.981Nematodes (2)15.167<0.05Microorganisms (3)187.84<0.00011 × 211.7360.1911 × 310.6180.4342 × 310.6720.4151 × 2 × 310.0470.828Error88The results of a three-way analysis of variance (ANOVA) with factors “microarthropods,” “nematodes,” and “microorganisms” (experiment 3) In the suspension of nematodes obtained from the pots to which microorganisms had been added, 47.3% showed signs of infection by culturable microbial enemies (Table 5). The fungal parasite Catenaria spp. was found infecting 16 out of 110 T. ventralis inspected, and an unidentified fungus was detected inside 30 destroyed of 110 T. ventralis checked. Bacterial attachment to the cuticle was also observed for six nematodes, a Paenibacillus-like organism was found on four nematodes, and Pasteuria spp. on two out of 110 nematodes. Table 5Microbial enemies in or attached to Tylenchorhynchus ventralis in a suspension obtained from microorganism treatment pots in experiment 3Microbial enemy detectedFraction of affected nematodes (%)Unidentified assimilative hyphae27.3Paenibacillus-like3.6Catenaria spp.14.5Pasteuria spp.1.8Total47.3(Healthy)52.7Total nematodes examined100 (n = 110) Assessing the presence of microbial enemies on nematodes in field soil The fungal natural enemies Catenaria spp., Harposporium spp., and Myzocytium spp. were found infecting nematodes extracted by the Tray method (Whitehead and Hemming 1965). The bacterium P.penetrans was attached to root knot nematodes (M. maritima). The fungal genera were also detected using nematode-baited sprinkle plates. Some of the nematodes on the plates had a swollen region behind the anus not observed in the original culture or in the negative control. This is a symptom of infection by a nonlethal bacterial parasite, M. nematophilum. An unidentified trapping fungus with nonconstricting rings was also detected in the sprinkle plates but could not be identified, as it did not sporulate. These identifications may not have been exhaustive, but they confirmed the presence of antagonistic microorganism species on T. ventralis as well as in the soil from the field. Discussion In coastal foredunes, the root-feeding ectoparasitic nematode T. ventralis would significantly influence the pioneer grass A. arenaria if the density of this nematode was not controlled naturally. Our study strongly suggests that the natural control of T. ventralis in coastal foredune soil is mostly due to soil microorganisms. When inoculated into sterilized soil, numbers of T. ventralis were more than five times greater than when inoculated into nonsterilized soil, whereas selective elimination of soil fauna by stirring did not affect nematode numbers. These results from selective elimination studies were confirmed by isolating microarthropods, nematodes, and microbes and adding these together with T. ventralis to sterilized soil. Inoculation with soil microorganisms reduced T. ventralis more strongly than did inoculation with a nematode community consisting of other plant parasites, bacterivores, omnivores, and carnivores. The negative effect of the nematode community on T. ventralis density might have been due to competition with other root feeders. However, competition between T. ventralis and endoparasitic nematodes (i.e. H. arenaria, M. maritima and P. penetrans) occurred only if numbers of the competitors strongly exceeded present field densities (Brinkman et al. 2004). Therefore, the observed effect of adding nematodes on reducing T. ventralis was due to soil microorganisms cointroduced with the nematode suspension, carnivorous nematodes (Jairajpuri and Bilgrami 1990), or by effects of other non-plant-feeding nematodes. Our results highlight an important discrepancy in thinking about control mechanisms of plant-feeding (also called plant-parasitic) nematodes between biocontrol studies on the one hand and soil food web studies on the other. The majority of studies on the biocontrol of nematodes in agricultural ecosystems mostly focus on parasitic bacteria, such as P. penetrans and fungi (Kerry 2000), for example Arthrobotrys spp. and Pochonia spp. (Stirling and Smith 1998). Biological control practice usually does not consider microarthropods to be relevant for parasitic nematode control (Kerry and Gowen 1995), whereas the role of carnivorous nematodes has been considered (Mankau 1980; Yeates and Wardle 1996) but not successfully used. According to the food web model used by Neutel et al. (2002), root-feeding nematodes in coastal ecosystems are affected by predaceous mites and carnivorous nematodes. The role of microorganisms in nematode control is generally ignored in prominent soil food-web models (de Ruiter et al. 1993; Hunt et al. 1987). In our study system, however, soil microorganisms appeared to play a more important role than soil fauna in the control of plant ectoparasitic nematodes. In dune grasslands, the densities of soil fauna are usually rather low, perhaps too low to control significantly the abundance of nematodes (Petersen and Luxton 1982). Our results, therefore, support the view of biological control studies more than it supports soil food-web models for the control of the ectoparasitic nematode T. ventralis. Our results suggest that top-down control by natural enemies is more important for ectoparasitic feeding generalists, such as T. ventralis, than are competition (Brinkman et al. 2004) or mycorrhizal fungi (Piśkiewicz and Hol unpublished results), although these control mechanisms have been suggested for the other endoparasitic root-feeding nematode species in the same study system (Brinkman et al. 2005; de la Peña et al. 2006; van der Stoel et al. 2006). The control of T. ventralis in nonsterilized soil to which T. ventralis was added was not as good as in nonsterile soil, with only the background population of T. ventralis present. Earlier studies have shown that multiplication of T. ventralis is density and time dependent (de Rooij van der Goes 1995). Over time, a low inoculation density of nematodes may result in population increase, whereas a high inoculation density may result in population decline (de Rooij van der Goes 1995). However, we have no information as to how the microbial control shown in the present study may depend on nematode density, and subsequent long-term studies are required to further explore density dependence of nematode top-down control by microbes. The two screening methods used to detect microbial enemies of nematodes in the soil yielded a diversity of fungal and bacterial antagonists. The Tray method (Whitehead and Hemming 1965) permitted the extraction of mobile stages of nematodes only, and therefore, dead or dying nematodes could not be screened. The nematode-baited sprinkle plate method, although useful for detecting microbial enemies in dormant forms in the soil, produced biased results that reflected the choice of nematode added, the bacterial-feeding nematode C. elegans. The fungal endoparasite Harposporium spp. infects nematodes that ingest its spores and therefore would not be able to infect plant parasites due to the narrow lumen of the stylet (Barron 1977). The bacterium M. nematophilum is thought to be a specialist parasite of C. elegans (Hodgkin et al. 2000). T. ventralis was attacked by a subset of the microbial enemy genera found in the soil, which reflects some specificity in the action of these microbes. Some microorganisms, namely Pasteuria spp. and the Paenibacillus-like organism, have not been detected in the dune soil, although they are widespread in the dune sites. This may be due to their absence or to presence in small numbers that are below the level of detection. The unidentified assimilative hyphae could be the result of attack by generalist trapping fungi. This was a first assessment to confirm the presence of nematode antagonistic microorganisms on T. ventralis, and further studies are needed to isolate culture and inoculate those antagonists in order to evaluate their contribution in root-feeding nematode control. Soil sterilization always led to increased root and shoot biomass. Previous studies have already shown that soil biota may reduce performance of A. arenaria (van der Putten et al. 1988; van der Stoel et al. 2002). The effect of soil sterilization was greater for root biomass than for shoot biomass. In sterilized soil, enhanced plant growth can be caused by nutrient release as a result of the soil sterilization process (Troelstra et al. 2001). We avoided different nutrient status of the sterilized and nonsterilized treatments by adding nutrient solution (van der Putten et al. 1988). When added to sterilized soil, the root-feeding nematode T. ventralis reduced root and shoot biomass of A. arenaria, and the effect of nematodes on root biomass increased with increasing inoculation density. Addition of T. ventralis to the nonsterile soil did not change root biomass, showing that the contribution of this nematode species to growth reduction of A. arenaria is limited (de Rooij van der Goes 1995). Our study shows that the dune soil not only contains biotic factors that reduce growth of A. arenaria but that there are also (micro) organisms that control population abundance of T. ventralis. In prairie grassland ecosystems, root-feeding nematodes have been assessed to account for reducing 58% of aboveground biomass (Stanton 1988). These estimates are based on elimination trials using soil biocides. However, these studies did not verify the biocide effects by inoculation trials, and they also did not account for natural top-down control of the root-feeding nematodes. Our results suggest that when assessing the effects of root feeders on plant production, interactions of the root feeders with their natural predators needs to be taken into account as well. We conclude that soil microorganisms contribute to controlling the plant ectoparasitic nematode T. ventralis at a low population density in natural coastal foredunes. When not controlled, T. ventralis would be a key control factor for A. arenaria performance. Addition of other soil fauna, i.e., nematodes and microarthropods, did not influence the abundance of T. ventralis as much as microorganism addition did, which confirms the marginal effects of soil fauna removal on reproduction of T. ventralis. Our results suggest that belowground multitrophic interactions can be crucial for plant performance. Revealing the precise identity of the microorganisms that have negative effects on the T. ventralis population, as well as mechanisms and involvement of the host plant, need further studies.

Arch_Dermatol_Res-3-1-1910889

Tolerance to coxibs in patients with intolerance to non-steroidal anti-inflammatory drugs (NSAIDs): a systematic structured review of the literature

Adverse events triggered by non-steroidal anti-inflammatory drugs (NSAIDs) are among the most common drug-related intolerance reactions in medicine; they are possibly related to inhibition of cyclooxygenase-1. Coxibs, preferentially inhibiting cyclooxygenase-2, may therefore represent safe alternatives in patients with NSAID intolerance. We reviewed the literature in a systematic and structured manner to identify and evaluate studies on the tolerance of coxibs in patients with NSAID intolerance. We searched MEDLINE (1966–2006), the COCHRANE LIBRARY (4th Issue 2006) and EMBASE (1966–2006) up to December 9, 2006, and analysed all publications included using a predefined evaluation sheet. Symptoms and severity of adverse events to coxibs were analysed based on all articles comprising such information. Subsequently, the probability for adverse events triggered by coxibs was determined on analyses of double-blind prospective trials only. Among 3,304 patients with NSAID intolerance, 119 adverse events occurred under coxib medication. All adverse events, except two, have been allergic/urticarial in nature; none was lethal, but two were graded as life-threatening (grade 4). The two non-allergic adverse events were described as a grade 1 upper respiratory tract haemorrhage, and a grade 1 gastrointestinal symptom, respectively. In 13 double-blind prospective studies comprising a total of 591 patients with NSAID intolerance, only 13 adverse reactions to coxib provocations were observed. The triggering coxibs were rofecoxib (2/286), celecoxib (6/208), etoricoxib (4/56), and valdecoxib (1/41). This review documents the good tolerability of coxibs in patients with NSAID intolerance, for whom access to this class of drugs for short-term treatment of pain and inflammation is advantageous. Introduction Non-steroidal anti-inflammatory drugs (NSAIDs) are the most commonly used therapeutics in the outpatient management of pain and inflammation in a wide spectrum of diseases. Their primary mode of action is the blockade of prostaglandin synthesis by cyclooxygenases (COX): Constitutively expressed COX-1 is involved in fundamental mechanisms of homeostasis, whereas the inducible COX-2 mediates inflammation. Therapeutic effects of NSAIDs are primarily related to their ability to inhibit COX-2, whereas some of their most frequent adverse effects may be caused by COX-1 inhibition (Fig. 1). In contrast to most “classic” NSAIDs which block both isoforms, the so-called coxibs preferentially inhibit COX-2. This may result in better tolerability, namely reduction of gastrointestinal side effects [29, 85]. Fig. 1Cyclooxygenase isoforms. Non-steroidal anti-inflammatory drugs (NSAIDs) block prostaglandin synthesis by cyclooxygenases (COX). Two isoforms exist of this enzyme: constitutively expressed COX-1 is involved in fundamental mechanisms of homeostasis, whereas the inducible isoform COX-2 mediates inflammation Respiratory and cutaneous adverse events triggered by NSAIDs are among the most common drug-related intolerance reactions in medicine. Typically, these manifest as asthma attacks or urticaria. Pathogenesis of these symptoms seems to be related to COX-1 inhibition [76]. Therefore, the hypothesis was put forward that coxibs may safely be used in patients with known NSAID intolerance. However, serious intolerance reactions to coxibs have also been observed, thus cautioning too euphoric expectations [71]. We therefore reviewed the relevant literature in a systematic and structured approach for evidence of coxib tolerance in patients with NSAID intolerance. Methods Search strategy We searched MEDLINE (1966–2006), the COCHRANE LIBRARY (4th Issue 2006) and EMBASE (1966–2006) up to December 9, 2006. The following search terms were used: rofecoxib OR celecoxib OR valdecoxib OR parecoxib OR etoricoxib, combined with hypersensitivity OR intolerance. Those coxibs used in veterinary medicine (deracoxib, tiracoxib and cimicoxib) were not included in the literature search. No language or publication restriction was predefined. All publications reporting individual patients with NSAID intolerance and subsequent exposure to a COX-2 Inhibitor were identified and the reference lists of these articles were hand-searched for further publications. If articles could not be retrieved in full text, a copy was requested from the corresponding author and/or journal editor. Inclusion criteria Each publication was appraised for inclusion in a stepwise approach (Fig. 2). Only publications describing individual patients and providing a rational medication scheme as well as a sufficiently specific outcome report were included in this review. In a first step, evaluation focused on clinical symptoms and severity of adverse events in patients with NSAID intolerance. Therefore, all articles on this topic were included for this analysis. Subsequently, the probability for adverse events was analysed based exclusively on publications of double-blind prospective trials. Fig. 2Flow chart documenting the handling data (NSAID non-steroidal anti-inflammatory drugs) Data extraction and synthesis All articles were analysed using a predefined evaluation sheet. Uncertainties were resolved by consensus decisions among the investigators. Data synthesis was qualitative and descriptive. The Common Terminology Criteria for Adverse Events version 3.0 (CTCAE) was used to categorize adverse events. Results We identified 230 publications on coxibs and/or NSAID intolerance. Hundred and forty-six references not focusing on individual patients with NSAID intolerance were excluded. Unclear medication schemes or outcome reports led to exclusion of two articles. Eighty-four publications were evaluated for severity and type of adverse reactions to coxibs. Thirteen publications on double-blind studies were used to determine the probability of adverse reactions to coxibs. Severity and type of adverse events to coxibs A total of 119 adverse events among 3,304 patients exposed to coxibs were identified [1–18, 20–24, 26–28, 30–35, 37–58, 60–75, 77–84, 86–89]. Adverse events were described for rofecoxib (48/1,732), celecoxib (59/1,148), etoricoxib (10/328), and valdecoxib (2/69), but not parecoxib (0/27) (Table 1). All adverse events, except two, have been allergic/urticarial in nature, and none was lethal. One adverse event after administration of rofecoxib and one after celecoxib were reported as life-threatening (grade 4). The two non-allergic adverse events were described as a grade 1 upper respiratory tract haemorrhage, namely blood in nasal discharge after provocation with rofecoxib, and a grade 1 gastrointestinal symptom, namely nausea and abdominal pain. Table 1Mode and severity of adverse event (CTCAE common terminology criteria for adverse events)PatientsEvents (total)Urticaria (grade)Allergic reactions (grade) Allergic rhinitis (grade)Airway obstruction (grade)Gastroin-testinal symtpms (grade)Pruritus (grade)Hemorrhage, upper respiratory (grade)Rofecoxib1,732489 (1)8 (2)1 (3)25 (3)1 (4)1 (2)2 (1)001 (1)Celecoxib1,1485919 (1)1 (2)1 (2)33 (3)1 (4)01 (1)03 (2)0Etoricoxib328103 (1)5 (2)1 (3)001 (1)00Valdecoxib6921 (1)1 (2)000000Parecoxib2700000000Total3,304119486213131Numbers and grade of adverse events (in parentheses) are indicated Evaluation of double-blind studies We identified 13 double-blind studies evaluating tolerability of coxibs in a total of 591 NSAID intolerant patients (Table 2). The coxibs tested in the context of these studies were rofecoxib (n = 286), celecoxib (n = 208), etoricoxib (n = 56) and valdecoxib (n = 41). Among the 591 patients described, 13 exhibited adverse reactions upon provocation with a coxib. The triggering coxibs were rofecoxib (2/286), celecoxib (6/208), etoricoxib (4/56), and valdecoxib (1/41) [13, 28, 48, 60, 69, 74, 75, 77, 78, 83, 84, 87, 88]. One of these reactions was clearly non-allergic, this one being the above-mentioned individual with grade 1 upper respiratory tract haemorrhage. Table 2Snopsis of double blind studies (NSAID non-steroidal anti-inflammatory drugs)ReferencesSubstance (number of patients)DoseOral provocation with NSAIDFollow-upAdverse events (number of patients/grade)[87]Celecoxib (17)Celecoxib 200 mgYesClinical examinations and forced expiratory volume hourly for 8 h.0[74]Celecoxib (12)/rofecoxib (3)Celecoxib 200 mg/rofecoxib 12.5 mgYesNo information0[75]Rofecoxib (60)Rofecoxib 37.5 mg total doseYesClinical examinations and flow/volume recording hourly0[77]Rofecoxib (12)Rofecoxib 25 mgYesClinical observation every 30 min for 8 h and finally after 24 h.Traces of blood in nasal discharge on days 3–5 of rofecoxib administration (1/1), transient urticaria for 30 min after administration of 5 mg rofecoxib, but not after further administration of higher doses of the drug (1/1)[78]Celecoxib (14)Rofecoxib 10Celecoxib 200–400 mg total dose Rofecoxib 37.5 mg total doseYesForced expiratory volume was recorded each hour for 3 h0[60]Rofecoxib (104)Rofecoxib 25 mg total doseYesClinical control for at least 6 h after each challenge and the next day.0[83]Celecoxib (60)Celecoxib 200 mgYesClinical examinations were conducted hourly0[28]Celecoxib (33)Celecoxib 130 mg total doseYesSpirometry and vital signs at 15- to 30-min intervals after each dose.0[88]Celecoxib (18)/rofecoxib (18)Celecoxib 300 mg total dose/rofecoxib 37.5 mg total dose1YesVital signs and forced expiratory volume every 15 min for 6 h.0[84]Rofecoxib (60)Rofecoxib 50 mgYesClinical examinations and forced expiratory volume hourly.0[13]Valdecoxib (41)Valedcoxib 30 mg total doseYesClinical examination and monitoring of pulse/blood pressare until 16 h past last exposureGeneralised urticaria (1/2)[69, 70]Celecoxib (54)Etoricoxib (56)Celecoxib 200 mgEtoricoxib 120 mgYesVital signs and pulmonary function was monitored at baseline and hourly for 3 h and the skin, nose, thorax were examined at the same intervalsCelecoxib: urticaria (3/1), lip angioedema (1/3), eyelid wheals and angioedema (1/3), urticaria, rhinorrhea, and conjunctival injection (1/1)etoricoxib: urticaria (3/1), tongue, eyelid and lip angioedema (1/3)[48]Rofecoxib (19)Rofecoxib 37.5 mg total doseYesSerial measurement of forced expiratory volume after 30, 60 and 120 min0 Discussion This systematic review documents the low probability of allergic/pseudo-allergic reactions induced by coxibs in patients with NSAID intolerance. To our knowledge, this is the first comprehensive analysis of data published on this topic. We have searched all three major medical databases available, namely MEDLINE, COCHRANE LIBRARY, and EMBASE using very broad and general search terms. Following identification of relevant publications, these were evaluated by means of a pre-defined evaluation form. The data available are described in the form of a structured review [19]. NSAIDs are among the most commonly prescribed therapeutics in the world. Although generally considered safe, their wide and frequent use results in these drugs being among the most common causes of drug-related intolerance reactions. This may at least in part be due to their non-selective inhibition of both cyclooxygenase isoforms. In line with this hypothesis, NSAIDs characterized by pronounced COX-1 inhibition bear a high risk to trigger asthma attacks in patients with aspirin-sensitive asthma bronchiale, whereas preferentially COX-2 inhibiting NSAIDS are better tolerated by these patients [36]. Our analysis of published studies on this issue further supports this notion, as only 13 of 591 NSAID-intolerant patients showed adverse reactions upon provocation with coxibs in double-blind clinical studies; all of these were grade 3 or milder. Still, relatively selective COX-2 inhibitors have been identified as triggers of serious intolerance reactions [71]. This implies that our current understanding of NSAID-triggered intolerance is still imperfect, and its pathogenesis cannot be reduced to cyclooxygenase-mediated effects alone, but needs to take into account clinically relevant additional NSAID-mediated effects such as secretion of leukotrienes from mast cells and other leukocytes. It has been suggested that coxibs may confer an elevated risk for acute myocardial infarction and sudden cardiac death, namely after long-term therapy [25]. As a reaction, several coxibs are no longer available despite a recommendation by the participants of an expert meeting organized by the Food and Drug Administration to grant further prescription of rofecoxib, celecoxib and valdecoxib in the US [59]. On the other hand, substantial evidence described here points towards a good tolerability of coxibs in patients with NSAID intolerance. Given the wide use of NSAIDs in the short-term treatment of trivial signs and symptoms, the availability of coxibs for these indications would be advantageous for this relevant subpopulation of patients, since this type of application is unlikely to increase cardiac risk.

Eur_Radiol-4-1-2220024

CT colonography: optimisation, diagnostic performance and patient acceptability of reduced-laxative regimens using barium-based faecal tagging

To establish the optimum barium-based reduced-laxative tagging regimen prior to CT colonography (CTC). Ninety-five subjects underwent reduced-laxative (13 g senna/18 g magnesium citrate) CTC prior to same-day colonoscopy and were randomised to one of four tagging regimens using 20 ml 40%w/v barium sulphate: regimen A: four doses, B: three doses, C: three doses plus 220 ml 2.1% barium sulphate, or D: three doses plus 15 ml diatriazoate megluamine. Patient experience was assessed immediately after CTC and 1 week later. Two radiologists graded residual stool (1: none/scattered to 4: >50% circumference) and tagging efficacy for stool (1: untagged to 5: 100% tagged) and fluid (1: untagged, 2: layered, 3: tagged), noting the HU of tagged fluid. Preparation was good (76–94% segments graded 1), although best for regimen D (P = 0.02). Across all regimens, stool tagging quality was high (mean 3.7–4.5) and not significantly different among regimens. The HU of layered tagged fluid was higher for regimens C/D than A/B (P = 0.002). Detection of cancer (n = 2), polyps ≥6 mm (n = 21), and ≤5 mm (n = 72) was 100, 81 and 32% respectively, with only four false positives ≥6 mm. Reduced preparation was tolerated better than full endoscopic preparation by 61%. Reduced-laxative CTC with three doses of 20 ml 40% barium sulphate is as effective as more complex regimens, retaining adequate diagnostic accuracy. Introduction Full bowel purgation remains a major cause of discomfort prior to any colonic investigation [1–3]. Furthermore, fluid and electrolyte imbalance may occur following aggressive cleansing [4]. A potential advantage of computed tomography colonography (CTC) over colonoscopy is the ability to reduce laxative requirements while maintaining diagnostic accuracy [5–7]. Reduced laxative regimens often incorporate orally ingested contrast agents to “tag” or “label” residual fluid and faecal residue. The ideal tagging regimen remains controversial but must be safe, effective, simple and well tolerated. In many respects, barium is an ideal tagging agent: it has an established safety profile, is relatively palatable, produces minimal side effects, and is effective for tagging solid residue [8]. Previous work has shown adequate tagging may be achieved using low volumes of 40% w/v barium, although the ideal volume and dosing regimen has not been fully established [9]. Furthermore efficacy for fluid tagging has been questioned [9], with some investigators preferring iodine-based contrast either alone or in combination, claiming a more homogeneous fluid opacification better suited to digital subtraction [10]. We aimed to establish the optimum barium-based reduced-laxative faecal- and fluid-tagging regimen, to assess patient acceptability, and to document diagnostic accuracy compared to an enhanced colonoscopic reference standard. Materials and methods Full ethical committee approval was obtained, and all subjects gave written informed consent. Consecutive patients were recruited from those scheduled to undergo afternoon diagnostic colonoscopy for symptoms suggestive of colorectal neoplasia (change in bowel habit, rectal bleeding, unexplained weight loss or palpable abdominal mass) from one of three institutions. Patients were excluded if below age 50 or if they had a known diagnosis of inflammatory bowel disease. Eligible patients were invited to undergo CTC at 8 am, prior to same day colonoscopy and were randomised via a computer random number generator to one of four reduced-laxative regimens using barium-based faecal tagging [20 ml 40% w/v barium sulphate suspension (Tagitol V, EZEM, Lake Success, NY)] (Fig. 1). For each of the regimens, patients followed a low-residue diet 2 days before the CTC (avoiding fatty food, milk and vegetables). The day prior to CTC, all patients were allowed a low-residue meal kit (Nutraprep, EZEM), and ingested the same reduced-laxative protocol [13 g sachet of senna granules (Reckitt Benckiser Healthcare, Hull, UK) at 7 am and 18 g magnesium citrate (Lo-So prep, EZEM) at 7 pm]. This laxative regimen was defined as “reduced” given the normal laxative preparation prior to colonoscopy at the recruiting institutions includes an additional 18 g of magnesium citrate (see below). Overall fluid intake was restricted to 2.1 l the day prior to CTC for all four regimens. Fig. 1Details of reduced-laxative tagging regimens. Taken with meals from low residue meal kit (single star). Diluted in 250 mls of water (double stars) In an attempt to improve fluid tagging, two regimens (C and D) included either an additional 250 ml 2.1% w/v barium sulphate (Readi-Cat Banana Smoothie, EZEM) or 15 ml meglumine amidotrizoate (10 g sodium amidotrizoate/66 g meglumine amidotrizoate per 100 ml; Gastrograffin, Schering) diluted in 250 ml of water, taken 2 h prior to CTC [11]. Patients otherwise took nothing by mouth the day of the CTC. Study power The study was powered to detect a 20% difference in tagging quality (see below) across the four regimens. Based on pilot data, the interclass correlation coefficient between colonic segments was calculated to be 0.30, and it was calculated that a sample size of 22 per group was required (alpha 0.05 at 80% power). CT colonography Prone and supine CTC was performed with automated CO2 insufflation (Protocol pump, EZEM) [12], using either a 4-detector-row (GE Lightspeed Plus, GE, Milwaukee, WI, USA; 120 kV, 50 mA, 2.5 mm collimation, slice reconstruction 1.25 mm, pitch 1.5, n = 82 patients) or 64-detector-row scanner (Siemens Somatom Sensation 64, SEMS, Germany; 120 kV, 50 mA, 0.6 mm collimation, pitch 0.24, n = 13 patients). One of three experienced radiologists (each with experience of at least 300 CTC cases with endoscopic validation) evaluated the datasets immediately after the examination using a dedicated workstation with proprietary software (Vitrea 3.8, Vital Images, Minnetonka, MN, USA), and noted the segmental location of any polyps or cancers on a study report sheet, together with lesion size measured using electronic callipers applied to the 2D MPR best showing the maximum diameter. The choice of reporting radiologist for each patient was dependent upon the particular recruiting institution and availability on the day of the scan. All three radiologists used a primary 2D approach with 3D reserved for problem solving. Colonoscopy Immediately following CTC patients ingested a further 18 g of magnesium citrate in order to complete the normal endoscopic cleansing regimen prior to afternoon colonoscopy. Colonoscopy was performed as per usual practice by one of five experienced endoscopists, on average 6 h (range 4–8 h) after the CTC scan. Segmental unblinding was used as described previously [13, 14], using the sealed CTC report. In brief, the CTC report was handed to the endoscopy nurse accompanying the patient. CTC findings were revealed to the colonoscopist by the nurse on a segmental basis (caecum to rectum) once examination of each colonic segment was deemed complete during extubation of the colon. If CTC suggested a lesion had been missed, the segment was re-intubated and a second segmental examination performed. There was no time limit imposed on the colonoscopist for this second look. All polyps were photographed, their sizes estimated by direct comparison to adjacent open biopsy forceps, and then excised for histology where possible. Polyp correlation Polyps found at CTC were deemed true positive if a corresponding polyp was found in the same or adjacent segment at endoscopy and if the estimated size of the polyp was within 50% of the endoscopic measurement. Patient experience Immediately following CTC a questionnaire was administered [15] investigating patient experience of the reduced-laxative tagging regimen (Table 1). Patient responses were compared to a historical cohort of 69 symptomatic patients recruited from the same endoscopy lists during a prior study comparing CTC to colonoscopy [16]. As part of this prior study, patients had undergone full bowel preparation [13 g senna granules and two doses of 18 g magnesium citrate (total 36 g) prior to CTC] and had completed the same questionnaire under similar circumstances. Table 1Bowel-tolerance questionnaire questions and responses in comparison to historical controls undergoing full bowel preparationVariableResponseReduced preparationa, n (%)Historical controls [full preparation]b, n (%)P valueHow did you find understanding prep sheet?Easy56 (63)48 (70)0.39Fairly easy/difficult33 (37)21 (30)How did you find swallowing medicine?Easy43 (48)44 (64)0.36Fairly easy36 (40)32 (32)Quite difficult/difficult10 (11)3 (4)How did you find coping with special diet?No problem59 (66)42 (61)0.49Bit difficult24 (27)24 (35)Very difficult6 (7)3 (4)How did you feel after medicine?Fine65 (73)48 (70)0.96Unwell/very unwell24 (27)21 (30)Did you have any abdominal pain?None31 (36)25 (36)0.37Mild39 (45)27 (39)Moderate/severe16 (19)17 (25)Did you have any nausea/vomiting?None58 (67)46 (67)0.92Mild22 (26)18 (26)Moderate/severe6 (7)5 (7)Did you experience any faintness or dizziness?None70 (81)54 (78)0.63Mild/moderate/severe16 (29)15 (22)Did you experience any wind?None25 (29)26 (38)0.27Mild41 (48)32 (46)Moderate/severe20 (23)11 (16)Did you experience any soreness?None37 (43)24 (35)0.15Mild37 (43)29 (42)Moderate/severe12 (14)16 (23)Did you experience any incontinence?None68 (80)49 (71)0.2Mild8 (9)13 (19)Moderate/severe9 (11)7 (10)Did you experience any sleep disturbance?None33 (39)41 (59)0.01Mild29 (34)20 (29)Moderate/severe23 (27)8 (12)How many times did you open your bowel after starting the preparation?1–31 (1)N/A3–520 (22)>569 (77)N/A Not applicable (not asked)a13 g senna plus 18 g magnesium citrateb13 g senna plus 36 g magnesium citrate One week later a follow-up questionnaire (Table 2) was mailed to the current study cohort investigating tolerance of the preparation prior to CT and preference, if any, over the full preparation required for subsequent colonoscopy. Table 2Questions and responses to follow-up questionnaire pertaining to patient tolerance and preferencesVariableResponsePatient number (%)How did you find taking the low-residue diet?No problem46 (67%)Moderately inconvenient20 (29%)Very inconvenient3 (4%)How did you find drinking the tagging liquid?No problem57 (83%)Moderately inconvenient11 (16%)Very inconvenient1 (1%)How did you tolerate the preparation before CT?Well37 (53%)Fairly well29 (42%)Poorly3 (4%)How did you tolerate the additional preparation prior to colonoscopy, compared to that before the CT?No problem51 (74%)More uncomfortable12 (17%)Much worse6 (9%)How did you find the preparation before CTC compared to the full colonoscopy preparation?Much better18 (26%)Better24 (35%)No better26 (38%) Grading of bowel preparation Two experienced radiologists (both with experience of over 700 validated CTC cases) in consensus retrospectively reviewed all CTC examinations, grading the quality of preparation and success of tagging. Observers were unaware of the tagging regimen used and divided the colon into six segments (rectum to caecum) for analysis [17]. Grading of residual stool (irrespective of tagging status) was based on the percentage of total mucosal circumference coated on an axial image. For each colonic segment, the slice with the most residual stool was used to assign the score. Scores were as follows: 1: no residue or scattered residue, 2: coating of <25% or thin circumferential “film” less than 2 mm in depth, 3: coating of ≥25 to 50%, 4: >50% coating. Grading of residual fluid (irrespective of tagging status) was based on the maximum anteroposterior (AP) diameter of the colonic lumen submerged. For each colonic segment, the slice with the most residual fluid was used to assign the score. Scores were as follows: 1: no fluid, 2: <25% AP diameter, 3: ≥25 to 50% AP diameter, 4: >50% AP diameter. The quality of tagging for solid residue and fluid was scored using a system adapted from Lefere at al. [9] and assigned for supine and prone positions combined for each colonic segment. Residual stool was divided into that measuring ≤5 mm and ≥6 mm (based on 2D measurement using electronic calipers), and the number of stool balls ≥6 mm was counted for each patient. Readers assessed the percentage of total residual stool volume (for each size category) that had been tagged for each colonic segment. Scores were assigned as follows: 1: all residual stool untagged, 2: 1 to <25% tagged, 3: 25 to <50% tagged, 4: 50 to <75% tagged, 5: 75 to 100% tagged. Tagging of residual fluid was scored as follows: 1: fluid untagged, 2: layered tagging (a mix of tagging densities in one fluid level with a denser dependent layer and visibly less dense non-dependent layer (Fig. 2), 3: fluid homogeneously tagged (single tagging density). The density of tagged fluid (in HU) was also recorded by taking the average of three regions of interest drawn in the deepest fluid pool for each colonic segment. If fluid was layered (Fig. 2), the minimum and maximum HU was recorded. Finally observers recorded their confidence using a percentage score that, based on the bowel preparation, they would be able to exclude a polyp ≥6 mm in day-to-day clinical practice for each colonic segment. Fig. 2A 68-year-old female with change in bowel habit. Axial CT colonographic image demonstrating layering of contrast (arrow) within tagged fluid Statistical analysis Preliminary data analysis revealed skewed data for most bowel preparation and tagging variables, and thus scores were combined for subsequent analysis. For residue, scores were grouped into either a score of 1, or a score of 2 or more. For residual fluid, scores were grouped into scores of 1/2, or scores of 3/4. Scores for solid residue tagging were grouped into those of ≤4 or 5. Fluid tagging scores were grouped into scores of 1/2, or score of 3. Logistic regression was then applied, adjusting for colonic segment, to compare the four regimens overall, and to compare the distal (rectum, sigmoid, descending) and proximal colon (transverse, ascending and caecum). In addition, the prevalence of layering within tagged fluid was compared on a per-patient basis using the chi-squared test. Tagged fluid attenuation was compared using linear regression following log transformation of the data. Confidence scores for excluding a polyp ≥6 mm were grouped as either <100% or 100%, and compared using logistic regression. For all regression analyses, robust standard errors were employed to account for interdependency between colonic segments and scan position (supine/prone) in the same patient. Results were expressed as the odds of the outcome under consideration compared to regimen A. Questionnaire responses were compared using Fischer’s exact test. Overall categorical per-polyp and per-patient data are presented using descriptive statistics. Given the relatively low polyp incidence (and thus low statistical power), comparative statistics for polyp detection were not performed across regimens. False-positive numbers were compared using one-way ANOVA. Results A total of 95 patients were recruited (50 female, mean age 64 years, range 50–85 years), with 24, 25, 24 and 22 randomised to regimens A to D respectively. Seventy-seven, 13 and 5 patients were recruited from institutions 1, 2 and 3 respectively. Overall, 67 (71%) had a change in bowel habit, 18 (19%) had rectal bleeding, 7 (7%) had non-specific weight loss, and 3 (3%) had a clinically palpable abdominal mass. Bowel preparation The percentage of segments (supine and prone combined) assigned a score of 1 for residual solid residue (i.e. no residue or scattered) was 76% (218/288), 82% (246/300), 81% (232/288) and 94% (247/264), for regimens A to D respectively, and on average there were 2.5, 2.0, 2.8 and 1.1 stool balls ≥6 mm per patient. The improved quality of preparation in regimen D reached statistical significance [odds of score 2–4: 0.19 (95% CI: 0.07, 0.56), P = 0.02]. Across all regimens, the distal colon was significantly better prepared then the proximal colon (P = 0.006). For example the number of segments assigned at least a score of 2 for solid residue in the caecum was 42% (20/48), 48% (24/50), 35% (17/48) and 23% (10/44) for regimens A to D respectively, compared to 25% (12/48), 16% (8/50), 21% (10/48) and 0% (0/44) for the rectum. The percentage of segments (supine and prone in total) assigned a score of 1 or 2 for residual fluid was 51% (148/288), 59% (178/300), 53% (153/288) and 43% (114/264) for regimens A to D respectively. There was no significance difference between groups either overall or between the proximal and distal colon (P = 0.22–0.37). Tagging quality Tagging quality was generally good (mean tagging score 3.7–4.5) (Fig. 3). While there was weak evidence of improved proximal colonic tagging for residue ≤5 mm for regimens C and D (P = 0.08), overall there was no difference in the efficacy of solid residue tagging across regimens (Table 3). Fig. 3A 54-year-old male with unexplained rectal bleeding. Axial CT colonographic image showing homogeneous tagging of stool ≤5 mm (arrow) and ≥6 mm (arrowhead) in sizeTable 3Efficacy of tagging of solid residue according to size and regimenSolid residue sizeColon segmentsRegimenMean tagging score (SD)Odds ratio (95% CI)aP value≤5 mmAll segmentsA4.3 (1.2)1B4.3 (1.2)0.94 (0.39, 2.32)C4.5 (1.1)1.47 (0.56, 3.87)D4.5 (1.2)1.67 (0.68, 4.14)0.56Distal colonbA1B0.97 (0.37, 2.52)C0.76 (0.26, 2.27)D1.00 (0.38, 2.61)0.97Proximal coloncA1B0.85 (0.26, 2.80)C4.06 (0.74, 22.1)D3.22 (0.95, 10.9)0.08≥6 mmAll segmentsA4.1 (1.6)1B4.3 (1.5)1.82 (0.44, 7.62)C4.1 (1.7)1.56 (0.32, 7.76)D3.7 (1.8)0.43 (0.11, 1.65)0.24Distal colonbA1B1.30 (0.26, 2.52)C3.84 (0.37, 39.2)D0.17 (0.02, 1.27)0.12Proximal coloncA1B3.22 (0.26, 39.8)C0.94 (0.11, 7.87)D0.91 (0.11, 7.54)0.78SD Standard deviationaOdds of tagging score of 5 (best) compared to regimen AbRectum, sigmoid and descending colon combinedcTransverse, ascending colon and caecum combined The average per-segment fluid tagging score was 2.5 (SD 0.8), 2.3 (SD 1.0), 2.5 (SD 0.9) and 2.4 (SD 1.2) for regimens A to D respectively. The odds of homogeneous tagging (i.e score 3) did not differ across the four regimens either overall or between the proximal and distal colon (P = 0.65–0.95). In total, 4% (6/144), 1% (2/150), 2% (3/144) and 5% (6/132) of segments respectively contained non-tagged fluid (score 1) (Fig. 4). Fig. 4A 75-year-old female with change in bowel habit. Axial CT colonographic image showing failure of fluid tagging (arrow) (grade 1) Overall 33% (8/24), 44% (11/25), 38% (9/24) and 36% (8/22) of patients receiving regimens A to D respectively had at least one segment with layering of tagged fluid (P = 0.89) (Fig. 2). Fluid tagging density In terms of mean HU of the tagged fluid (excluding the least tagged layer if tagging was layered), there was no significant difference among the four regimens, either overall or between the proximal/distal colon (P = 0.14–0.86). The mean HU of tagged fluid was 522 (SD 283), 504 (SD 395), 430 (SD 268), and 491 (SD 269) for regimens A to D respectively. However in those segments with layering of contrast, the maximum attenuation (HU) of the least tagged layer was significantly higher for regimens C and D than for regimens A and B (P = 0.002) [77 (SD 105), 119 (SD 139), 155 (SD 118) and 174 (SD 122), for regimens A to D respectively]. Patient experience A total of 89 of 95 (94%) patients completed the questionnaire (Table 1), although responses were incomplete in 4. Demographic data for the present study cohort (50 female, mean age 64 years, range 50–85 years) were not significantly different from the 69 historical controls (36 female, mean 63 years, range 35–85 years) [16]. Other than sleep disturbance (worse after reduced preparation) (P = 0.01), there was no significant difference in reported tolerance of the reduced and full bowel preparation regimens for any of the factors tested. Follow-up questionnaire Sixty-nine (73%) patients returned the follow-up questionnaire (Table 2). Overall 95% tolerated the reduced preparation regimen well or fairly well, and most (83%) found drinking the tagging agents acceptable. Although most [51/69 (74%)] found the additional preparation required for colonoscopy “no problem”, a majority (61%) found the reduced preparation regimen “better” or “much better” than the full preparation required for colonoscopy. Diagnostic performance Five patients (one each from regimens A, B, and C and two from regimen D) were excluded from the performance analysis due to a new diagnosis of inflammatory bowel disease (all presenting with rectal bleeding). A further patient (from regimen B) was excluded after refusing colonoscopy. Of the remaining 89 patients, 68 had either normal colonoscopy or diminutive polyps (≤5 mm) only, and 21 had at least one polyp ≥6 mm or cancer. Colonoscopy was incomplete in 10/89 (11%) reaching the transverse colon in 6, sigmoid in 3 and hepatic flexure in 1. Reasons for failure were obstructing stricture (1), severe diverticulosis (1), tortuous colon (5), and pain (3). Only those segments visualised at colonoscopy were included in the assessment of diagnostic performance. Per polyp analysis In total there were 9 polyps ≥10 mm (all adenomatous), 12 polyps 6–9 mm (10 adenomatous, 2 hyperplastic) and 72 polyps ≤5 mm (46 were recovered for histology, of which 26 were adenomatous, 15 hyperplastic and 5 normal mucosa). One 6-mm polyp and two 5-mm polyps were found only on re-look endoscopy after segmental unblinding of the CTC report. No CTC-detected polyps were classified as false positives due to segmental or size mismatching with colonoscopy. Summed across regimens A–D, detection of cancer, polyps ≥10 mm, 6–9 mm and ≤5 mm was 2/2 (100%), 8/9 (89%), 9/12 (75%) and 23/72 (32%) respectively (Fig. 5) (Table 4). In total there were only four false positives ≥6 mm (Figs. 6 and 7), with no significant difference between regimens (P = 0.15). Fig. 5A 71-year-old male with rectal bleeding. Axial CT colonographic image demonstrating a 30-mm rectal cancer (arrow). Note the adjacent well-tagged residue (arrowheads)Table 4Polyp detection overall and according to tagging regimenRegimenPatient numberDetection cancer, n (%)Detection 10 mm+, n (%)Detection 6–9 mm, n (%)Detection 1–5 mm, n (%)False positive 10 mm+, nFalse positive 6–9 mm, nFalse positive 1–5 mm, nP valueaA232/2 (100%)1/2 (50%)N/A3/10 (30%)0170.15B23N/A3/3 (100%)5/6 (83%)10/30 (33%)01b9C23N/A1/1 (100%)2/2 (100%)6/20 (30%)001D20N/A3/3 (100%)2/4 (50%)4/12 (33%)1c12Overall892/2 (100%)8/9 (89%)9/12 (75%)23/72 (32%)1319N/A Not applicableaComparison of false positives across regimens using one-way ANOVAbIn patient with confirmed 8-mm polypcIn patient with confirmed 10-mm polypFig. 6A 54-year-old male with change in bowel habit. Axial CT colonographic image of a 6-mm filling defect in the rectum reported as a polyp. No lesion was found at colonoscopy with segmental unblinding suggesting the lesion was untagged faecal residueFig. 7Overall number of false positives according to size and regimen (n = 89) Per patient analysis Across all regimens sensitivity, specificity, positive and negative predictive values for detection of patients with any lesion ≥6 mm were 96%, 97%, 0.9 and 0.96 respectively (Table 5). Table 5Per-patient performance overall and according to tagging regimenRegimenPatient numberPolyp ≥6 mm incl. cancer, n (%) [95% confidence limits]Polyp ≥10 mm incl. cancer, n (%) [95% confidence limits]SensitivitySpecificityPPVNPVSensitivitySpecificityPPVNPVA233/4 (75%) [33–100%]18/19 (95%) [85–100%]0.750.953/4 (75%) [33–100%]19/19 (100%) [100–100%]1.00.95B237/7 (100%) [100–100%]16/16 (100%) [100–100%]1.01.03/3 (100%) [100–100%]20/20 (100%) [100–100%]1.01.0C233/3 (100%) [100–100%]20/20 (100%) [100–100%]1.01.01/1 (100%) [100–100%]22/22 (100%) [100–100%]1.01.0D205/7 (71%) [38–100%]12/13 (93%) [78–100%]0.830.863/3 (100%) [100–100%]17/17 (100%) [100–100%]1.01.0Overall8918/21 (96%) [93–100%]66/68 (97%) [93–100%]0.9a0.9610/11 (91%) [74–100%]78/78 (100%) [100–100%]1.0b0.99NPV Negative predictive value, PPV positive predictive valueaPrevalence of abnormality = 0.24bPrevalence of abnormality = 0.12 Reader confidence The mean segmental observer confidence for excluding a polyp ≥6 mm was 90, 89, 97 and 93% for regimens A to D respectively. There was no significant difference among the regimens, either overall or in the proximal/distal colon (P = 0.15–0.57). Discussion In accordance with previous studies [6, 8, 10], we found patients in general preferred reduced preparation to the full purgation required for colonoscopy, although this preference was much less than expected (only 61% preferred the reduced preparation). We defined the laxation used as “reduced” in comparison to the endoscopic preparation used at our institutions which includes an additional 18 g of magnesium citrate. We also used senna rather than bisacodyl as used by Lefere et al. [9]. Both are stimulant laxatives with a similar mode of action and effect at the doses administrated, although senna was preferred as it is widely used at our institutions. However a combination of 13 g of senna and 18 g magnesium citrate clearly produces relatively strong purgation, reflected in the induced symptoms (77% of patients opened their bowels over five times) and marginal patient preference. It would therefore seem reasonable to study reduced laxation further, perhaps omitting the senna and/or reducing the dose of magnesium citrate. It is however important to realise that whereas normally patients took no solids by mouth for 24 h prior to colonoscopy, in the present study they were permitted to eat from a low-residue diet kit, an issue we did not specifically address with our questionnaires, thereby potentially underestimating this benefit. Recent data have questioned whether CTC is actually better tolerated than conventional colonoscopy [18], but because full bowel purgation is often cited by patients as the worst part of any colonic examination [1, 2], it is assumed that reduced laxative CTC will improve patient compliance (notably in a screening setting). However this assumption has not been proven in prospective trials, and we cannot extrapolate the preferences we found into increased compliance with CTC. Indeed it could be argued the laxative regimen we used would have relative little impact on compliance in a screening setting, given the side-effect profile. The overall quality of bowel preparation was good, with all four regimens producing at least 76% of segments with either no or scattered residue only. This is similar to data reported by Lefere et al. [9], who combined 16.5 g of magnesium citrate with biscodyl tablets and suppository. Interestingly, regimen D, which included 15 ml meglumine amidotrizoate, resulted in significantly better preparation than the other regimens, possibility due to a “washing effect” of the iodine-based contrast, incorporating solid residue into a more fluid solution [10]. Although we used a small dose, meglumine amidotrizoate is also known to have a laxative effect that may also have added to the superior cleansing. Tagging of solid residue was in general good, with no significant difference between the regimens tested. This suggests barium-based tagging can be simplified to a 1-day regimen only, when combined with reduced cathartic preparation. Similarly all regimens produced an average tagged fluid density of around 500 HU. Recent phantom data suggest that although submerged polyp conspicuity is optimised at 700 HU, it remains good at 500 HU [19]. We did however demonstrate that fluid tagging could be manipulated by additional oral agents on the morning of CTC. As it is non-miscible with water, barium often produces a layering effect with lower attenuation fluid sitting above denser barium. Although there was no significant difference in layering among the regimens, the addition of morning 2.1% barium sulphate or meglumine amidotrizoate significantly increased the attenuation of the non-dependent fluid layer. It is arguable whether this will have significant impact clinically, not least because fluid moves between supine and prone positions, but it may have implications for subtraction software. Zalis et al. have recently demonstrated suboptimal subtraction when high-density barium is used as the sole tagging agent [10]. It is interesting to speculate whether manipulation of fluid tagging could optimise subtraction. It should also be noted our regimen mildly restricted fluid intake the day prior to CTC, which may have influenced our fluid tagging results. One major advantage of barium as a tagging agent is its inert nature and safety. Although iodinated contrast medium produces more homogeneous fluid tagging, this is mainly of clinical importance if subtraction software is being used. While the risks of allergy to oral iodinated contrast are minimal, adverse events do occur [20, 21]. Our data suggest good results can be achieved using only barium. Ultimately, diagnostic performance is the best measure of the success of any reduced-laxative protocol. Reassuringly overall, we prospectively found high diagnostic performance when compared to colonoscopy. Our data confirm again that full bowel preparation is not required to maintain diagnostic accuracy for CTC, assuming adequate reader training [22]. Importantly, there were only four false positive polyps over 6 mm, and across all regimens the positive predictive value for patients with any lesion ≥6 mm and ≥10 mm was 0.9 and 1.0 respectively (Table 5). Our study does have weaknesses. It could be argued that we used a relatively harsh laxative regime given that good results have been reported without any laxative at all [5, 6]. However the vast majority of CTC worldwide is still performed using full bowel preparation, and in the authors’ opinion, significant changes in practice are most likely to follow gradual reduction of purgation, which allows radiologists to become comfortable with this approach. However we acknowledge that if there was good evidence of acceptable performance data using CTC without prior laxation, this may well rapidly become widely implemented by the CTC community. We compared patient symptomatology with a historical cohort, and although well-matched in demographic terms, we cannot exclude bias completely. We did not instruct colonoscopists to grade the quality of bowel preparation and, in particular, the influence of retained barium on mucosal visibility, although no endoscopy failed due to incomplete bowel preparation. It could be argued that reduced preparation regimens are best suited to increase compliance in asymptomatic screening populations [23]. However for pragmatic reasons we used symptomatic patients, as we do not have access to a large screening population. Finally, although we were able to show high diagnostic performance for reduced-preparation CTC overall, given the relatively low number of polyps, we were unable to meaningfully compare across regimens. In conclusion, a combination of reduced laxatives and a simple tagging regimen based on 40% barium sulphate the day prior to CTC maintains acceptable diagnostic accuracy. Three doses of 20 ml 40%w/v barium sulphate are as effective as more complex regimens, but fluid tagging can be manipulated by addition of dilute barium or meglumine amidotrizoate on the morning of CTC, the latter also reducing the volume of residual stool.

Eur_J_Pediatr-4-1-2292480

The efficacy of anakinra in an adolescent with colchicine-resistant familial Mediterranean fever

Colchicine is the treatment of choice in familial Mediterranean fever (FMF) for the prevention of both attacks and secondary amyloidosis. The overall nonresponder rate is about 5–10%. Anakinra is known to have good effectiveness in a severe autoinflammatory syndrome [chronic infantile neurological cutaneous and articular (CINCA) syndrome] and other recurrent hereditary periodic fevers. Pyrin—the protein involved in FMF—has a role in activating the proinflammatory cytokine interleukin (IL)-1β. We report the effectiveness of the addition of an IL-1-receptor inhibitor (anakinra) to colchicine in controlling the febrile attacks and acute phase response in an adolescent with FMF resistant to colchicine. Case report A 15-year-old girl with familial Mediterranean fever (FMF) was admitted to our hospital because of the persistence of recurrent fever attacks (twice a month), accompanied by chest and abdominal pain with headache, despite taking colchicine at a high dose (2 mg/kg per day). She suffered severe discomfort and complained of several school absences. Acute phase reactants [erythrocyte sedimentation rate (ESR) 110, C-reactive protein (CRP) 43 mg%] and serum interferon-γ (125.4 pg/ml) were high during the crises. There was no proteinuria, but given the severity of the clinical picture, we performed a rectal biopsy to definitely exclude the presence of amyloidosis and it was negative. Anakinra was started at 50 mg/day (1 mg/kg per day) subcutaneously, without stopping colchicine. Approval was obtained from an Ethics Committee (Burlo Garofolo Health Authority). In the following 3 months the patient developed only three mild episodes of abdominal pain without fever, which resolved in a few hours. During these episodes acute phase reactants were negative. Moreover, serum interferon-γ was low (23 pg/ml). After 3 months, anakinra was discontinued to confirm its real efficacy, without stopping colchicine. Following this interruption, the girl complained of eight fever attacks in 3 months accompanied by abdominal pain, vomiting, chest pain, and headache. Anakinra was reintroduced. Since then (15 months of follow-up), the patient has been in good health, without any fever attacks or abdominal or chest pain. There have been no adverse effects due to anakinra, apart from minor local stinging and erythema at the injection sites, which gradually diminished. Discussion Colchicine is the recommended treatment of FMF since it can prevent the febrile attacks and the development of amyloidosis, even though a few patients develop amyloidosis also when on colchicine treatment [3, 5]. However, approximately one third of the patients have a partial remission and about 5–10% are nonresponders [3]; another 2–5% do not tolerate the drug well [3]. In single case reports several drugs (interferon-α, infliximab, thalidomide, etanercept) have been suggested as a possible alternative treatment for these cases [3, 6]. Anakinra is a recombinant form of the human interleukin (IL)-1-receptor antagonist that targets type I IL-1-receptor which is expressed in many tissues. There are some studies and case reports that show a good efficacy of anakinra in a severe autoinflammatory syndrome (CINCA, chronic infantile neurological cutaneous and articular syndrome) [2] and other recurrent hereditary periodic fevers [4]. Despite the elevated costs (approximately 50 € per 100 mg) its use appears justified in some inflammatory diseases that lack alternative treatments. Since pyrin—the protein involved in FMF—activates the precursor form of IL-1β into its biologically active fragments there is a rationale for testing anakinra in FMF [1, 5]. Moreover, anakinra did appear to be effective in suppressing acute phase reactants in a patient with FMF (a nonresponder to the treatment with colchicine) and amyloidosis [1]. In our case, the addition of anakinra to colchicine had a dramatic therapeutic benefit without relevant side effects. As the febrile attacks recurred after discontinuing anakinra and disappeared again after reintroduction, we believe in the real therapeutic efficacy of this drug. In light of the present evidence [3, 5], colchicine should be continued in order to prevent further developments of amyloidosis.

Mol_Imaging_Biol-3-1-1915656

Attenuation-Corrected vs. Nonattenuation-Corrected 2-Deoxy-2-[F-18]fluoro-d-glucose-Positron Emission Tomography in Oncology, A Systematic Review

Purpose To perform a systematic review and meta-analysis to determine the diagnostic accuracy of attenuation-corrected (AC) vs. nonattenuation-corrected (NAC) 2-deoxy-2-[F-18]fluoro-d-glucose-positron emission tomography (FDG-PET) in oncological patients. Introduction The attenuation of photons originating from the subject before they are detected by the camera is a generic limitation of nuclear medicine imaging. This attenuation can lead to image distortion and impairs adequate quantification. Attenuation correction has been commonly employed in 2-deoxy-2-[F-18]fluoro-d-glucose (FDG) imaging in an attempt to correct for these effects. With positron emission tomography (PET) scanners, this is accomplished by transmission scanning using a radionuclide source, such as germanium-68 or cesium-137, and with PET/computed tomography (CT) using CT. With respect to visual interpretation of the images, the added value of attenuation correction has been controversial. Whereas attenuation correction provides a more realistic image of FDG distribution, its application significantly increases acquisition times on standard full-ring (FR) PET scanners. In addition, the performance of attenuation correction can introduce noise and even artifact. Paradoxically, even if the nuclear medicine community sees attenuation correction, or the lack of it, as a potential effect-modifier of test accuracy, its impact is rarely accounted for in systematic reviews on the diagnostic accuracy of PET. As a result, the impact of attenuation correction on lesion detectability and interpretation of PET for oncological purposes is not well established. With PET/CT scanning, it is customary to evaluate either modality (primarily to account for artifacts), but one needs to know how to deal with discrepancies. The objective of this systematic review and meta-analysis was to determine the diagnostic accuracy of nonattenuation-corrected (NAC) and attenuation-corrected (AC) FDG-PET in oncological patients. We studied the effects of attenuation correction for both FR-PET and dual-head coincidence PET (DH-PET), and as a function of different body locations (head/neck, chest, abdomen/pelvis). Materials and Methods Literature Search A computer-aided literature search was performed in both Medline and Embase databases without time range or language restrictions, applying controlled vocabulary (MeSH and EMTREE keywords, respectively) as well as free text words. The search date was February 10, 2006. The search strategy (Appendix) included terms for PET with FDG, modified from Mijnhout et al. [1] as well as search terms identifying both radionuclide and X-ray transmission, emission, attenuation correction, and oncological studies in humans. In addition, the reference lists of the eligible articles were reviewed to ensure that relevant articles had not been missed. Study Selection From the list of retrieved articles, articles were initially evaluated for eligibility on the basis of title and abstract by two independent reviewers (UJ, PR). If there was uncertainty as to whether an article was eligible for inclusion, the entire article was reviewed. Inclusion criteria were (1) clinical studies evaluating FDG imaging with and without attenuation correction in oncology patients; (2) study population of at least ten patients; (3) sufficient detail to reconstruct a 2 × 2 contingency table expressing FDG imaging results by disease status, or sufficient detail to reconstruct relative lesion detection measurement of AC vs. NAC imaging; and (4) studies utilizing FR-PET and/or DH-PET. We excluded abstracts, editorials, and reviews, although the latter two categories were used for cross-referencing. Methodological Quality Assessment The methodological quality of each article was independently assessed by each reviewer in terms of internal and external validity (Table 1), based on the Cochrane Methods Group in Screening and Diagnostic Tests, modified for our area of interest [2]. Internal validity items focus on whether a valid reference test was used and whether this reference test was uniformly and independently applied and interpreted as well the type of study design. The external validity items evaluate the applicability of the results in terms of the type of patient population and spectrum, demographics, the inclusion/exclusion criteria, the knowledge of previous test/clinical information that could influence interpretation, and the index test characteristics. Items were scored as positive, negative, or unclear. Table 1Methodological assessment of individual diagnostic studies: criteriaTestCriteriaA. Internal study validity Al. Valid reference testHistology, AC FR or DH coincidence PET A2. Blind measurement of reference test(s) without knowledge of index testAssessment of reference test independent of index test(s) results A3. Avoidance of verification biasChoice of patients assessed by reference test independent of index test result A4. Index test(s) interpreted independently of all clinical informationMentioned in publication A5. Prospective studyMentioned in publicationB. External study validity B1. Spectrum of diseasesLocalization of disease described (selected or general) B2. Demographic informationAge and sex given B3. Inclusion criteria describedMentioned in publication B4. Exclusion criteria describedMentioned in publication B5. Avoidance of selection biasConsecutive series of patients B6. Standardized execution of index test(s)Described technical aspects of index test(s)C. Reproducibility describedMentioned in publication Data Extraction and Quantitative Analysis In addition to methodological quality assessment, data related to the type of camera, the FDG dose, the time interval between injection and imaging, the transmission and emission acquisition protocols, the reconstruction protocol, and the interpretation protocol were independently extracted from each study by each reviewer. For studies where it was possible, a contingency 2 × 2 table was constructed. Disagreements were solved by consensus. For studies using an independent gold standard (histopathology), we determined the sensitivity and specificity of the index tests using the number of true positive, false positive, true negative, and false negative results from the 2 × 2 contingency table. Furthermore, we calculated the “relative lesion detection,” defined as the percentage of lesions scoring equally positive or negative with NAC vs. AC images. We performed a subgroup analysis for different locations of lesions and analyzed sensitivity, specificity, or relative lesion detection of NAC vs. AC for lesion location in the head and neck region, the chest, and the abdominopelvic region. In cases of discrepancy of relative lesion detection between NAC and AC, we extracted data to analyze whether this was related to lesion size and/or intensity. The statistical diagnostic heterogeneity of the sensitivity and specificity per index test across studies was tested by the chi-square test. In case of statistical heterogeneity of DH- or FR-FDG-PET imaging, a random effect model for pooling was used, whereas in case of statistical homogeneity, a fixed-effect model was used. Sensitivity, specificity, and relative lesion detection were pooled independently, all pooled estimates are presented with 95% confidence intervals (95% CI). The logit transformed sensitivity, specificity, relative lesion detection, and corresponding 95% CI of the index tests were compared using z-test statistics. A p value of less than 0.05 was considered significant. All statistical analyses were performed with the SPSS 11.0.01 program for Windows (version 11.0.1., SPSS, Chicago, IL, USA). Results The search strategy yielded 2,202 references, 1,477 in Medline and 725 in Embase on February 10, 2006. Of the Embase references, 370 were also included in Medline, leaving a total of 1,832 unduplicated references. On the basis of title and abstract alone, 1,806 references were excluded. After review of the full text of the remaining 26 articles, an additional 11 studies [3–12] proved to be ineligible because they did not perform a direct comparison of the yield of NAC vs. AC images in oncological patients. One study [13] was excluded because it was published in abstract form only. Another study [14] was excluded because it was published in Japanese and was not readily translatable. Finally, the study of Hustinx et al. [15], who evaluated the effect of attenuation correction in abdominal tumors for a sodium iodide crystal (NaI) PET scanner, was excluded because no 2 × 2 contingency tables could be constructed. Eventually, we included 12 studies for review [16–27]. A summary of the methodological quality assessment results can be found in Table 2. Methodological quality was scored as negative when quality items were unclear or absent in the original article. In a minority (4/12 = 33%) of studies, histology served as the reference test. However, nine of the 12 studies provided a direct comparison of AC and NAC PET. In three studies, blind measurement of reference test was performed without knowledge of index test. All but one study avoided verification bias. In four studies, the index test(s) was evaluated independently of all clinical information. All studies provided information about the spectrum of diseases being evaluated and standardized the execution of the index test(s). Almost all studies (11/12 = 92%) described the demographics of the study population and inclusion criteria. However, only one study mentioned specific exclusion criteria. Six studies were prospective. Only two of the 12 studies specifically mentioned including consecutive patients, and only three studies specifically described the reproducibility of their results. Table 2Quality assessment of included studiesStudyYearA1A2A3A4A5B1B2B3B4B5B6CBleckmann1999++++−+−+−−+−Chan2001+−+−−+++−−+−Delbeke2001+−+−++++−−+−Even-Sapir2004+−+−++++−++−Kotzerke1999+−+−−+++−+++Lonneux1999+−+−+++−−+++Nakamoto2002++++++++−−++Reinhardt2005+−+−−++++−+−Schauwecker2003+−−−−+++−−++Weber1999+−+−++++−−+−Zimny1999++++−+++−−+−Zimny2003+−++++++−−+− Meta-Analysis Three FR-PET studies were eligible for pooling of sensitivity on a patient basis [22, 24, 25]. The pooled sensitivities for AC and NAC FR-PET were 64% (95% CI 52–74%) and 62% (95% CI 51–73%), respectively (n = 182 patients). Two FR-PET studies provided data that allowed pooled analysis of specificity [22, 24]. Weber et al. [25] could not be included as there were no patients without disease. The pooled specificities for AC and NAC FR-PET were 97% (95% CI 92–99%) and 99% (95% CI 95–100%), respectively (n = 155 patients). For DH-PET, only one study provided data on sensitivity and specificity [27]. Relative lesion detection for NAC vs. AC PET was pooled for 11 studies, which are demonstrated in Fig. 1. Lesion detection of NAC FR-PET vs. AC FR-PET was 98% (95% CI: 96–99%) for n = 1,012 lesions (pooling of n = 7 studies); 79% of which were classified FDG positive at AC FR-PET. Lesion detection of NAC DH-PET vs. AC DH-PET was 88% (95% CI: 81–94%) for n = 288 lesions (pooling of n = 4 studies); 74% of which were classified as FDG-positive at AC DH-PET. Fig. 1Pooled lesion detection of NAC vs. AC images for FR-PET and DH-PET. In addition, we evaluated the relative lesion detection depending on body location (head/neck, chest, abdomen/pelvis) in the four FR-PET [20, 21, 23, 26] and in the three DH-PET studies that provided sufficiently detailed information [17, 18, 26]. The relative sensitivity and specificity based on body location could not be calculated due to an insufficient number of studies. For FR-PET, we found similar relative lesion detection for the three body locations: 95% for head/neck (95% CI 84–98%, n = 61 lesions), 97% for the chest (95% CI 94–99%, n = 396 lesions), and 97% for the abdomen/pelvis (95% CI 93–0.99%, n = 205 lesions). For DH-PET, relative detection rates for NAC were not significantly different for the various body sites: 78% in the abdomen/pelvis (95% CI 65–88%; 53 lesions), 84% in the chest (95% CI 74–91%; 136 lesions), and 90% in the head/neck area (95% CI 73–97%; 38 lesions). However, in chest (p = 0.000089) and abdomen/pelvis (p = 0.0037), the relative detection rates with NAC (vs. AC) for DH-PET were significantly lower than those obtained with FR-PET. A comprehensive analysis of the potential association of relative detection and lesion size and/or intensity, for lesions with discrepant AC and NAC results, was not possible due to a lack of detailed information. We summarized the results in Table 3: findings of Bleckmann et al. and Reinhardt et al. [16, 23] suggest that AC and NAC discrepancies may relate to (intrapulmonary) lesion size with more discrepancies occurring with smaller lesions at the subcentimeter level (an average of 3% of lesions were correctly detected with NAC and not with AC). The single discrepant lesion in the study of Weber et al. concerned a <1-cm lesion in the mediastinum [25]. However, the discrepant lesions in the studies of Nakamoto et al., Schauwecker et al., and Delbeke et al. included both small- and moderate-sized lesions (in relation to lesions included each study) [18, 22, 24]. In the study of Schauwecker et al., the discrepant lesions demonstrated SUVmax values ranging from 1.8 to 2.6, whereas, in the study of Delbeke et al., the two discrepant lesions demonstrated only mildly enhanced uptake on AC images and equivocal uptake on NAC images. Table 3Evaluation of discrepant lesions between AC and NAC images with respect to lesion size and intensityStudyCamera typeNumber of discrepant lesionsSize rangeIntensity (semiquantitative or qualitative)Bleckmann et al.FR-PET5<1 cmNot givenNakamoto et al.aFR-PET11.8 cmNot givenReinhardt et al.FR-PET60.5–1.1 cm79/174 lesions demonstrated discrepancy in qualitative lesion intensity: 72/174 lesions demonstrating higher intensity (i.e., better visibility) on NAC imagesSchauwecker et al.aFR-PET40.8–3.9 cm31.8–2.6 (SUVmax)Weber et al.FR-PET10.8 cmNot givenDelbeke et al.DH-PET21.0–3.0 cm“Mild uptake” at AC, “equivocal uptake” at NACaHistology used as gold standard and detailed information given only for true positive lesions Discussion The cumulated evidence summarized in this systematic review of oncological FDG imaging studies suggests that the accuracy of attenuation and nonattenuation corrected FR-PET are similar. However, with DH coincidence imaging NAC images detect 12% less lesions than AC images, without prominent differences between body areas. Although in the nuclear medicine field attenuation correction is generally seen as an important issue, it is surprising to find that several large systematic reviews did not thoroughly consider this as a potential effect-modifier. Gould et al. performed systematic reviews on FDG PET in pulmonary lesions [28] and mediastinal lymph node staging in non-small cell lung cancer [29]: in the former review, the item was not mentioned, and in the latter, attenuation correction was an item of study quality, but no analysis of potential impact was performed. The choice of the reference test is obviously relevant in studies on test accuracy. In oncology, histopathology is the typical endpoint. Of the 12 eligible studies, four used histology as an independent gold standard. Meta-analysis of sensitivity and specificity was only possible for FR-PET, and we found no significant difference for either measure. We chose to use the AC detection rate as an alternative reference test, which defines the relative lesion detection of NAC vs. AC images. This choice theoretically biases in favor of AC: Bleckmann et al. and Reinhardt et al. [16, 23] reported an average of 3% more true positive lesions with NAC FR-PET. However, we expect that the resulting error is small because, in the comparison with histopathology, false positive rates were quite low for either modality. Despite this theoretical negative bias towards NAC, similar lesion detection rates were observed with both AC and NAC for FR-PET. Hence, attenuation correction may not contribute to the detection of malignancy using FR-PET. Conversely, with DH-PET, AC images demonstrated a significantly higher detection rate as opposed to NAC images, which is surprising given that AC images are usually significantly noisier than NAC images. We postulate that this may be secondary to differences in reconstruction/filtering algorithms. In addition, there are limitations associated with performing a meta-analysis and data pooling, such as the homogeneity of the data and the quality of the published studies. Homogeneous data have higher statistical strength than heterogeneous data. The data in our study were heterogeneous so that we used a random effect model for pooling. In addition, the statistical strength of the meta-analysis is limited by the quality of the published studies included in it. As mentioned earlier and summarized in Table 2, the studies had several quality limitations. Finally, meta-analyses are limited by publication bias, which biases towards the publication of favorable results or popular subjects. We were surprised by the limited number of good comparative studies evaluating the value of attenuation correction. It appears that attenuation correction has been accepted as the standard of practice without sound scientific evidence to support it. The introduction of PET/CT machines has made the time constraints associated with transmission scanning less of an issue. However, PET/CT is not a panacea; X-ray transmission scanning has its own problems and numerous PET/CT publications have demonstrated artifact that can be introduced with X-ray transmission scanning [5, 30–45]. Furthermore, in the study of Reinhardt et al. [23], a significantly improved visibility was demonstrated for 41% of lung metastases with NAC images as opposed to CT-AC images. This higher visibility for NAC images was even more pronounced for lesions smaller than 1 cm. These findings underline that even as PET/CT use becomes more widespread, evaluation of both NAC and AC images should remain an integral part of image interpretation, and not just to recognize image artifacts. At the same time, NAC vs. AC discrepancies at PET/CT offer an obvious opportunity for further investigation. Conclusions In this meta-analysis, we found no significant difference in sensitivity, specificity, or relative lesion detectability between AC and NAC FR FDG PET. However, attenuation correction improved lesion detection for DH coincidence imaging.

Eur_J_Pediatr-3-1-2151774

Pneumomediastinum: a rare complication of anorexia nervosa in children and adolescents. A case study and review of the literature

Spontaneous pneumomediastinum is uncommon in paediatric practice. We describe two cases of spontaneous pneumomediastinum in a child and an adolescent with anorexia nervosa. Thorough investigation failed to reveal any underlying cause for secondary pneumomediastinum. Pneumomediastinum in anorexia nervosa can be caused by not only elevated intrathoracic pressures, but also by the poor quality of the alveolar walls due to malnutrition. The incidence of spontaneous pneumomediastinum in anorexia nervosa is probably higher than that recorded, since it resolves spontaneously and, therefore, it can remain undetected. We conclude that it is our considered opinion that malnutrition associated with anorexia nervosa predisposes for spontaneous pneumomediastinum due to weakness of the alveolar wall and the loss of connective tissue. Introduction Pneumomediastinum is rarely associated with anorexia nervosa (AN). Approximately 20 cases of pneumomediastinum in anorexia patients are described in the literature. Vomiting, a common symptom in AN, is a known cause of pneumomediastinum [13]. But of the cases reported in the literature, only a few (three) were preceded by vomiting [3, 4, 12]. So, there has to be another cause placing patients with AN at risk for spontaneous pneumomediastinum. We report one case with spontaneous pneumomediastinum as the presenting symptom of AN. In the other case, spontaneous pneumomediastinum is a complication of AN. A review of the literature is given and we discuss the possible pathophysiology of spontaneous pneumomediastinum in patients with AN. Case report Case 1 A 13-year-old girl presented with increasing complaints of an unusual crackling sensation and sound in her neck during three days. At that time, she complained of neck pain and headache. When these complaints started, she had a sore throat with painful swallowing. Further medical history mentioned a Cooper test, a test of physical fitness, at school two weeks earlier. There had been no trauma or injury and no coughing or vomiting. Physical examination showed a respiratory rate of 16 per minute, cardiac rate of 51 per minute, blood pressure of 87/55 mmHg and temperature of 36.5°C. Her height was 150 cm (SD −2), weight 34 kg (SD −1.5) and body mass index (BMI) 15 kg/m2 (SD −2.5). There was a slight bilateral bulging of the skin of her neck from the mandible to both clavicles, with, on palpation, a crackling sensation as a diagnostic sign of subcutaneous emphysema. Apart from being extremely slim, the skin was normal, with no signs of injury. Cardiac evaluation showed normal heart sounds without cardiac murmur. There was no dyspnoea. On auscultation, there was normal bilateral vesicular breathing. Blood tests revealed no abnormalities. Chest X-ray and computed tomography (CT) scan revealed mediastinal emphysema with extension to the neck region (Fig. 1). There was no pneumothorax and no extension to the retroperitoneum. An oesophageal gastric passage X-ray was normal, without signs of perforation. Laryngopharyngoscopy did not reveal mucosal lesions or signs of submucosal swelling. Electrocardiography was normal, except bradycardia. After admission to the hospital, she showed loss of appetite and further history revealed an anorectic episode, for which, she had been referred by her general practitioner to a psychologist. At the age of 13 years, she weighed 46 kg, diminishing to 37 kg 4 months later. At the time of admission to the hospital, her weight was 34 kg at the age of 13 years and 6 months. There was absolutely no history of vomiting. Fig. 1Coronal computed tomography (CT) reconstruction with mediastinal emphysema and subcutaneous emphysema in both axillas and lower neck regions in patient 1 with anorexia nervosa (AN) Within a week, the subcutaneous emphysema disappeared. She was referred to an eating disorders clinic to manage her eating disorder. She never showed any signs of purging during treatment and recovered from her AN. Case 2 A 17-year-old girl with known AN and vomiting presented with extreme malnutrition requiring refeeding. She complained of unusual crackling sensation in her neck during eight days, starting after a choking incident while drinking, followed by coughing. The same day, she felt pressure on her chest. During the following days, her neck became swollen and her voice became hoarse. The swelling had already subsided considerably at the time of presentation. Physical examination showed an extreme cachectic girl of 17 years with a respiratory rate of 16 per minute, cardiac rate of 64 per minute, blood pressure of 85/55 mmHg and temperature of 36.0°C. Her height was 170 cm (SD=0), weight 34.4 kg (SD<<−2) and BMI 12 kg/m2 (SD<<−2.5). There were signs of subcutaneous emphysema in the neck and in both arms. She had dry skin without signs of injury. Cardiac auscultation was normal without cardiac murmur. There was no dyspnoea. Pulmonary auscultation was also normal. Blood tests revealed no abnormalities. Chest X-ray and CT scan revealed subcutaneous emphysema in the neck and mediastinal emphysema with extension to the intra-abdominal, retroperitoneal and epidural regions (Fig. 2). There was no pneumothorax. Fig. 2Chest CT showing epidural pneumatosis (arrow) in patient 2 with pneumomediastinum and AN Laryngopharyngoscopy did not reveal mucosal lesions. The laryngeal mucosa was slightly swollen. After refeeding, she was referred to an eating disorders clinic to manage her eating disorder. Within two weeks, the subcutaneous emphysema further resolved spontaneously. Discussion Pneumomediastinum, the presence of free air contained within the mediastinum, usually results from spontaneous alveolar wall rupture and, far less commonly, from disruption of the upper airways or gastrointestinal tract [8]. Spontaneous pneumomediastinum can be distinguished from pneumomediastinum secondary to traumatic events, such as chest trauma or thoracic surgery, endobronchial or oesophageal procedures, mechanical ventilation or other invasive procedures [13]. A large and diverse group of factors has been implicated in the development of spontaneous pneumomediastinum. Various respiratory manoeuvres that have in common the development of high intrathoracic pressures may lead to pneumomediastinum. These include Valsalva manoeuvres, coughing, vigorous crying and vomiting. The mechanism of development of spontaneous pneumomediastinum was originally described by Macklin and Macklin as increased intra-alveolar pressure causing the rupture of alveoli with the escape of air into the mediastinum. The mediastinum is subsequently decompressed by further passage of the air into the retroperitoneal and subcutaneous spaces (the Macklin effect) [1, 9]. Whereas primary pneumomediastinum is a relatively benign and self-limited disorder, secondary pneumomediastinum due to oesophageal perforation or Boerhaave’s syndrome is a potentially life-threatening disorder because of the inevitable development of mediastinitis. Primary pneumomediastinum rarely requires treatment as the alveolair rupture heals and the leaked air resolves spontaneously. By contrast, secondary pneumomediastinum due to oesophageal perforation often requires surgical intervention [8]. Pneumomediastinum typically manifests as sudden chest pain or dyspnoea and, less commonly, with dysphagia and hoarseness [8]. Our first patient, however, presented with a sore throat, painful swallowing, complaints of pain in the neck and headache. The second patient presented with initial chest pain and hoarseness. The detection of subcutaneous air in the neck and anterior chest wall and Hamman’s sign, a crunching sound synchronous with the heartbeat, are important physical findings. Radiographically, pneumomediastinum manifests on the posteroanterior projection as a thin radiolucent line parallel to the outline of the heart, as the mediastinal pleura is displaced laterally, and on the lateral projection as the accumulation of air retrosternally [8]. When chest pain and dyspnoea occur, panic attacks must be differentiated from pneumomediastinum. Panic disorders are often seen in patients with AN [6]. As a consequence, we believe that pneumomediastinum should be ruled out in anorexia patients presenting with signs of a first panic attack. Spontaneous pneumomediastinum is rarely associated with AN. Only 21 publications related to this association are reported by Pubmed. Purging behaviour, such as self-induced vomiting, is common among AN patients. Vomiting is a known cause of pneumomediastinum, placing anorexia patients, who frequently vomit, at risk for pneumomediastinum. However, in only a few of the cases described in the literature was vomiting was the preceding event of the pneumomediastinum. In most reports, like in our patients, pneumomediastinum in anorectic patients is not preceded by vomiting [1, 2, 4, 5, 7, 8]. In our patients, the diagnostic procedure did not reveal an oesophageal or upper airway perforation. The pathophysiologic mechanism of air entrance into the mediastinum can be explained by the rupture of an alveolar wall, which is mostly the cause of an air leak into the mediastinum [8]. Based on the literature and our observations, patients with AN are at risk for spontaneous pneumomediastinum, even if they are not vomiting. We speculate that the state of malnutrition contributes to the risk of spontaneous pneumomediastinum. Animal studies reveal that calorie restriction causes a loss of alveoli and a fall in gas-exchange tissue [10, 11] and thinner alveolar walls [10]. To survive periods of not eating, organisms sacrifice comparatively non-essential structures for gluconeogenesis to provide glucose for the brain and amino acids to maintain muscle. Because the total organismal and lung oxygen consumption fall during calorie restriction, some lung tissue is expendable [10]. Also in human studies, atrophic changes, such as large alveoli and thin alveolar walls, due to starvation were found in Jewish people in the Warsaw Ghetto [14]. With thinner alveolar walls and the loss of alveoli, malnourished individuals are at risk of alveolar wall rupture. Since known factors of increased intra-alveolar pressure were absent in our first case and since clinical examination and radiographic study did not reveal an esophageal or upper airway perforation, we must assume that subclinical alveolar leaks with subsequent air dissection, pneumomediastinum and diffuse soft tissue emphysema occurred because of weakness of the alveolar wall and thinning of the connective tissue caused by severe malnutrition. Therefore, even with a minimal increase of intra-alveolar pressure, such as that which may occur during usual daily activities, such as a choking incident, can become the cause of air leaks, as in our second case.

Ann_Surg_Oncol-4-1-2234448

Confirmation of Sentinel Lymph Node Identity by Analysis of Fine-Needle Biopsy Samples Using Inductively Coupled Plasma–Mass Spectrometry

Background The sentinel lymph node (SLN) biopsy technique is a reliable means of determining the tumor-harboring status of regional lymph nodes in melanoma patients. When technetium 99 m-labeled antimony trisulfide colloid (99 mTc-Sb2S3) particles are used to perform preoperative lymphoscintigraphy for SLN identification, they are retained in the SLN but are absent or present in only tiny amounts in non-SLNs. The present study investigated the potential for a novel means of assessing the accuracy of surgical identification of SLNs. This involved the use of inductively coupled plasma–mass spectrometry (ICP-MS) to analyze antimony concentrations in fine-needle biopsy (FNB) samples from surgically procured lymph nodes. The sentinel lymph node (SLN) biopsy technique, introduced in the early 1990s, has made it possible to establish the tumor-harboring status of the regional node field in melanoma patients with a minimally invasive procedure.1 Within 3 years of the publication describing the procedure, the accuracy of SLN biopsy was confirmed in two further studies in which the SLN status was found to accurately reflect the status of the entire lymph node field.2,3 The SLN biopsy technique has since become widely accepted as a method of staging the disease of patients with melanoma who have clinically negative lymph nodes. The status of the SLN is used to determine whether further surgical or adjuvant therapy is appropriate, and to assess patient prognosis. The tumor-harboring status of the SLN is the single most important prognostic factor for melanoma patients, surpassing Breslow thickness, ulceration, and mitotic rate.4 To provide accurate prognostic information and to guide appropriate management, it is imperative that all “true” SLNs are accurately identified, excised, and assessed for their tumor-harboring status. In a number of cases, a false-negative result may occur, whereby a patient with disease originally assessed as SLN-negative subsequently develops recurrence in a regional node field. It is disturbing to note that false-negative rates of up to 24.8% have been reported from some major melanoma treatment centers and cooperative groups.5–16 Available evidence suggests that these failures occur not because the SLN biopsy concept is flawed, but because of methodological and technical shortcomings in nuclear medicine, surgery, and histopathology. In an attempt to identify how many of these false-negative cases might have been because of surgical failures—that is, the removal of a lymph node that was not the SLN identified by the preoperative lymphoscintigraphy—we have developed a technique for assaying antimony in tissue sections. This work is based on the concept that technetium 99 m–labeled antimony trisulfide colloid (99 mTc-Sb2S3) particles used for preoperative lymphoscintigraphy in Australia are retained in the SLN but are absent or present in only tiny amounts in non-SLNs. It has been shown that antimony is preferentially retained by the SLN and can be measured in nodal tissue sections by inductively coupled plasma–mass spectrometry (ICP-MS).17 It has also been demonstrated that human tissue contains only negligible amounts of antimony.17 ICP-MS enables validation of the SLN biopsy procedure26 and can identify false-negative results attributable to inaccurate SLN removal.18 Although the morbidity associated with SLN biopsy is low,11 it is an invasive procedure, with a definite risk of complications and morbidity.19,20 In addition, the financial implications of SLN biopsy must be considered. One estimate of the cost of performing a SLN biopsy procedure on an outpatient basis was US$12,19321—a high cost for providing melanoma treatment services to the community. The development of a minimally invasive technique for the identification of SLN metastases may minimize or eliminate the morbidity of the procedure and provide a potential cost benefit. It should be possible to identify SLNs by preoperative lymphoscintigraphy, localize the SLNs with ultrasound, and then perform a percutaneous fine-needle biopsy (FNB) under ultrasound control.22 Preliminary studies of proton magnetic resonance spectroscopy (MRS) analysis of FNB samples from lymph nodes indicate that this technique has the capacity to distinguish nodes containing metastatic melanoma from uninvolved nodes with high sensitivity, specificity, and accuracy.22,23 In addition to accurately determining the tumor-harboring status of the lymph node by this technique, it is also valuable to verify that the sampled lymph node is a “true” SLN, particularly if it has been assessed as being free of metastatic tumor. The objective of this study was to develop a method for confirming that a FNB specimen was obtained from a true SLN by measuring the level of antimony (present in radiocolloid used during preoperative lymphoscintigraphy) in the FNB sample by ICP-MS. Materials and Methods Instrumentation All measurements were made with an Agilent 7500ce Inductively Coupled Plasma Mass Spectrometer equipped with a MicroMist glass concentric nebulizer, a Quartz-Scott spray chamber (Peltier cooled, 2°C), and an Agilent three-channel peristaltic pump. The operating conditions were optimized daily to ensure maximum sensitivity. Typical operating conditions are summarized in Table 1. The mass spectrometer was operated in spectrum mode with an integration time of 300 ms on each of the following isotopes: 121Sb, 123Sb, 103Ru (internal standard/control). TABLE 1.Operating parameters for Agilent 7500ce Inductively Coupled Plasma–Mass SpectrometryParameterConditionRadiofrequency power1500 WSample depth8 mmPlasma gas flow 15.0 L min−1Carrier gas flow1.05 L min−1Peristaltic pump.15 rpsSample read delay60 sRinsing time45 sDwell time100 msReplicates6rps, revolution per second. Reagents and Chemicals All reagents used were of the highest purity available. Seventy percent double-distilled nitric acid, 37% hydrochloric acid (Arastar), and 30% hydrogen peroxide were obtained from Sigma-Aldrich, Australia. Ultra-high-purity water was produced by passing distilled water through a Milli-Q deionizing system (Millipore, Australia). Standards and Certified Reference Materials Certified reference material GBW 07601 Human Hair Powder (Langfang, China) was selected because of its certified antimony levels. Multiple samples of the reference material were also analyzed with each batch of samples and during method development to ensure the accuracy and precision of the analytic technique. SLN Biopsy Procedure Preoperative lymphoscintigraphy was performed to identify the node fields receiving direct lymphatic drainage. This process involved intradermal injections of 99 mTc-Sb2S3 around the primary cutaneous melanoma site, followed by early and delayed imaging with a scintillation camera.24 The location of the SLNs was marked on the overlying skin by the nuclear medicine physician to assist the surgeon in locating SLNs during surgery. The SLN biopsy procedure was performed within 24 hours of the radiocolloid injection, so that residual radioactivity in lymph nodes could be measured intraoperatively with a handheld gamma probe (Navigator GPS, RMD Instruments, Watertown, MA). Fifteen minutes before the operative procedure, multiple intradermal injections of Patent Blue V dye (Guerbet, Roissy, France) were made around the primary cutaneous melanoma site. SLN identification was based on visualization of the nodal blue dye staining and results from the preoperative lymphoscintigraphy, with gamma probe confirmation. The experimental protocols of this study were approved by the University of Sydney Ethics Review Board in accordance with the precepts established by the Declaration of Helsinki. FNB Samples The FNB collection process involved puncturing the fresh SLN specimen within 30 minutes of its surgical removal with a 25-gauge needle attached to a 3-mL plastic syringe. Multiple passes were then made through each quarter of the specimen. A total of 47 FNB samples (from 32 presumptive SLNs and 15 nodes considered to be non-SLNs) were collected. All samples were placed in polypropylene vials containing 300 μL of phosphate-buffered saline (.27 mM of KCl, 13.69 mM of NaCl, 1.52 mM of KH2PO4; pH 7.2) made up in perdeuterated water (phosphate-buffered saline–D2O) and immediately snap-frozen in liquid nitrogen and stored at −70°C.22 Microwave Digestion Procedure Each FNB sample was transferred to a polypropylene tube and prepared by the previously described microwave digestion procedure.17 Briefly, the sample was digested in a solution containing 300 μL of nitric acid, 300 μL of hydrochloric acid, and 500 μL of hydrogen peroxide. Each sample was digested five times in a 500 W microwave oven on the defrost setting for 30 seconds each time. The digest was then quantitatively transferred to a second polypropylene tube, made up to 10 g with a 1% nitric acid solution, and assayed by ICP-MS. Statistical Analysis The unpaired t-test was used to analyze differences between the levels of antimony in the FNB digests from SLNs and non-SLNs. A P value of less than .05 was considered statistically significant. Results The certified concentration of antimony in GBW 07601 is .095 ± .012 μg/g (mean ± SD). Replicate analyses of reference materials during method validation and sample analysis gave a mean value of .088 ± .007 μg/g and a relative standard deviation of <9%, confirming the accuracy and precision of the method. The limit of detection (LOD) was evaluated by the 3σ criterion (the LOD is given by mb + 3σb, where mb is the blank measurement mean and σb is standard deviation of five blank measurements) and found to be .048 parts per billion (ppb). The matrix-matched calibration standards were in the range of 0 to 20 ppb and gave an R2 value of >.9998 during method validation and sample analysis. The relative standard deviation of the slopes of the calibration curves was found to be <10%. The concentration of antimony in the digested SLNs and non-SLNs was measured, and the results are presented in Tables 2 and 3, respectively. The mean and median concentrations of antimony were .898 and .451 ppb, respectively, in the SLNs (range, <LOD−3.248) and .0145 and .0068 ppb in the non-SLNs (range, <LOD−.1159). These results indicate that the levels of antimony in FNBs from SLNs were significantly greater than from non-SLNs (P < .00005) (Fig. 1). TABLE 2.Concentration of antimony (ppb) in fine-needle biopsy sample digests collected from sentinel lymph nodesSample No.Antimony concentration (ppb)1<LOD2<LOD3<LOD4<LOD5.06666.09797.17388.29859.302010.308511.325712.380713.381214.386415.390316.404517.497118.607919.640720.724321.8582221.091231.438241.444251.724261.760271.847282.013292.150302.298312.756323.248LOD, limit of detection (.048 ppb).TABLE 3.Concentration of antimony (ppb) in fine-needle biopsy sample digests collected from nonsentinel lymph nodesSample No.Antimony concentration (ppb)1<LOD2<LOD3<LOD4<LOD5<LOD6<LOD7<LOD8<LOD9<LOD10<LOD11<LOD12<LOD13<LOD14<LOD15.1159LOD, limit of detection (.048 ppb).Fig. 1.Antimony concentration in digested fine-needle biopsy samples from sentinel lymph nodes and nonsentinel lymph nodes. Discussion Current standard clinical management of patients with primary cutaneous melanoma includes wide local excision of the primary tumor and SLN biopsy for patients considered to be at high risk of having regional node field metastases.25 Technical failures of the procedure could be attributable to errors in lymphoscintigraphy, sentinel lymphadenectomy, or histologic evaluation. In such cases, the potential for disease recurrence in a previously mapped lymph node basin exists.6,8–10,12,14 A reliable technique to confirm the identity of a SLN has the potential to reduce false-negative results attributable to surgical errors. In a previous study, we demonstrated that the analysis of antimony in sectioned nodal tissue could be used to distinguish SLNs and non-SLNs removed from the same nodal basin.26 This technique can be used within the bounds of currently used histopathologic protocols and has the advantage of being applicable to archival paraffin-embedded tissue. It has also been established that false-negative results may be caused by the removal of a lymph node incorrectly classified as a sentinel lymph node.18 Because neither blue dye nor radioisotope persist after procedures required for microscopic examination, they cannot be used retrospectively to confirm that the “true” SLN had been removed. The only other reported method to confirm SLN identity involves the injection, along with the blue dye, of carbon particles that are retained by the node.27,28 However, these large, dense carbon particles may hinder optimal histologic examination by obscuring metastatic melanoma cells. As far as we are aware, this method is not in routine clinical use. FNB digests from four nodes presumed to be SLNs (13%) were found to contain extremely low levels of antimony, below the LOD. In a previous study, we also identified SLNs that contained extremely low levels of antimony.26 The most likely reason for the low antimony levels would be inaccurate classification of the node that had been removed as a SLN, when it is in fact a non-SLNs. Other possible explanations to account for the low antimony levels include variable distribution of antimony throughout the node, limited migration of the colloid to the lymph nodes, or tumor deposits preventing colloid uptake. Haigh et al.27 investigated the distribution of carbon dye in SLNs and found a high concentration of carbon particles around the point of entry of afferent lymphatic channels. It is likely that antimony exhibits an analogous distribution pattern. If so, FNB samples that fail to include this region of the SLN may contain falsely low antimony levels. In view of these factors, it would be useful to set a criterion standard for classifying a given node as sentinel or not on the basis of antimony levels assessed by ICP-MS. However, defining such a criterion on the basis of our small sample set is difficult. In a previous study, we assessed antimony levels in archival tissue sections of paired tumor-positive SLNs and tumor-negative non-SLNs removed from the same regional node field during the same operative procedure from individual patients.26 The aims of this study were to determine whether antimony concentrations could be used to confirm whether removed SLNs were “true” SLNs and to differentiate SLNs from non-SLNs. The median concentration of antimony in the SLNs was .526 ppb and in non-SLNs was .043 ppb (P = .004). By using a cutoff point of .18 ppb (the median concentration of antimony in all SLNs and non-SLNs) to differentiate SLNs from non-SLNs, 20 of 24 SLNs and 20 of 24 non-SLNs were correctly identified by the SLN biopsy procedure. Although SLN biopsy is a highly accurate method for staging regional lymph nodes, it is an invasive surgical procedure that is costly and is associated with an inherent risk of complication and morbidity.19,20 Pathologic assessment of SLNs is laborious, time-consuming, and costly, and involves examination of multiple hematoxylin and eosin–stained histopathologic sections as well as sections stained with immunohistochemical techniques. The development of a rapid nonsurgical technique that allows detection of metastatic tumor deposits would be of great benefit. It would eliminate the need for surgical excision of lymph nodes and reduce the complexity of pathologic assessment of SLNs. However, to ensure that potential false-negative results are minimized, it is imperative to ensure the accuracy of SLN identification. Several studies have identified proton MRS as a candidate for the nonsurgical assessment of lymph nodes.22,23 The technique monitors changes in the chemical composition of cells during tumor development and can identify differences that are not morphologically discernible. In these studies, the spectra collected from SLN FNB samples harboring tumor cells contain choline and taurine peaks, which are absent in disease-free samples.23 Another sensitive technique for the detection of melanoma metastases in SLNs is the assessment of tyrosinase mRNA by reverse transcriptase–polymerase chain reaction (RT-PCR).29,30 However, as far as we are aware, RT-PCR assessment of tyrosinase mRNA has not been studied in FNB samples of SLNs. A major potential problem of RT-PCR analysis of FNB samples of SLNs is the issue of false-positive and false-negative results in such specimens and this would need to be assessed before the technique could be used in clinical practice. The results of our preliminary work provide evidence that determining antimony levels in FNB specimens by ICP-MS can confirm SLN identity and can differentiate SLNs from non-SLNs. This technique may be a useful adjunct to other techniques assessing tumor-harboring status of SLNs on FNB specimens such as MRS. Potentially, proton MRS analysis of a FNB from a node confirmed to be a true SLN by ICP-MS could provide a reliable method for determining the tumor-harboring status of SLNs. Although the results are promising and raise the possibility that SLN assessment may be performed on FNB specimens (rather than histologic specimens of excised SLNs) in the future, further validation studies are necessary before such techniques are used in widespread clinical practice. Furthermore, the highly specialized and expensive nature of the equipment used for ICP-MS will likely limit the technique to specialist centers.

J_Gastrointest_Surg-3-1-1852387

Platelet Function in Acute Experimental Pancreatitis

Acute pancreatitis (AP) is characterized by disturbances of pancreatic microcirculation. It remains unclear whether platelets contribute to these perfusion disturbances. The aim of our study was to investigate platelet activation and function in experimental AP. Acute pancreatitis was induced in rats: (1) control (n = 18; Ringer’s solution), (2) mild AP (n = 18; cerulein), and (3) severe AP (n = 18; glycodeoxycholic acid (GDOC) + cerulein). After 12 h, intravital microscopy was performed. Rhodamine-stained platelets were used to investigate velocity and endothelial adhesion in capillaries and venules. In addition, erythrocyte velocity and leukocyte adhesion were evaluated. Serum amylase, thromboxane A2, and histology were evaluated after 24 h in additional animals of each group. Results showed that 24 h after cerulein application, histology exhibited a mild AP, whereas GDOC induced severe necrotizing AP. Intravital microscopy showed significantly more platelet–endothelium interaction, reduced erythrocyte velocity, and increased leukocyte adherence in animals with AP compared to control animals. Thromboxane levels were significantly elevated in all AP animals and correlated with the extent of platelet activation and severity of AP. In conclusion, platelet activation plays an important role in acute, especially necrotizing, pancreatitis. Mainly temporary platelet–endothelium interaction is observed during mild AP, whereas severe AP is characterized by firm adhesion with consecutive coagulatory activation and perfusion failure. Introduction Acute pancreatitis (AP) is characterized by an inflammatory affection of the exocrine pancreatic tissue and disturbances of pancreatic microcirculation.1 Depending on the severity of AP, irreversible perfusion failure with consecutive tissue hypoxia and necrosis can complicate the course of the disease and trigger systemic inflammatory and septic complications.2 The pathophysiology of AP has been investigated with regard to microcirculatory changes in several studies.1–5 Attention was paid especially to erythrocyte flow patterns, leukocyte–endothelium interaction, and rheological approaches to improve perfusion and inhibit irreversible tissue damage.1,3–5 Leukocyte–endothelium interaction as an early event of the inflammatory response has been characterized as a key step in the pathophysiology of AP.6 Besides, activation of the humoral coagulation cascade plays an important role in the development of microcirculatory disorders in AP.7,8 However, the role of platelets as the cellular elements of hemostasis that can functionally link inflammatory cells and humoral coagulation factors has not been investigated. The aim of this study was to investigate platelet activation and function in experimental AP. Materials and Methods Animals The experiments were performed in 54 male Wistar rats weighing 270 to 335 g. Animals were fasted overnight with free access to water before the experiments. Care was provided in accordance with the guidelines published in the “Guide for Care and Use of Laboratory Animals” (National Institutes of Health, publication no. 85-23, 1985). Surgical anesthesia was induced with intraperitoneal injection of pentobarbital (25 mg/kg) and intramuscular injection of ketamine (40 mg/kg) for the procedures of catheter placement and induction of pancreatitis. Anesthesia during intravital microscopy was induced by intravenous injection of pentobarbital (10 mg/kg). Polyethylene catheters (inner diameter 0.5 mm) were placed in the right jugular vein and left carotid artery, tunneled subcutaneously to the suprascapular area, and brought out through a steel tether that allowed the animals’ free movement and access to water during the experiments. Monitoring blood samples Mean arterial pressure and heart rate were monitored during intravital microscopy by an electromechanical pressure transducer (Baxter Uniflow, Baxter Healthcare Cooperation, Deerfield, IL, USA). Arterial blood samples for determination of serum amylase were obtained before (baseline) and 24 h after (end point) pancreatitis was induced. Serum amylase was determined by standard laboratory methods (Hitachi automatic analyzer, Boehringer Mannheim, Germany). Animal models Animals were divided into three groups. In each group, pancreatic microcirculation was evaluated in 12 animals by intravital microscopy, and morphological changes were assessed in six animals by histology. In the control group, animals underwent sham operation and received Ringer’s solution only. Acute pancreatitis of graded severity was induced in the other groups either as mild or severe AP. Mild AP was induced by intraarterial infusion of cerulein (5 μg kg−1 h−1) for over 6 h. Cerulein was reconstituted in saline solution, and infusion volume was 4 ml/kg/h. Severe necrotizing pancreatitis was induced by infusion of bile salt (glycodeoxycholic acid [GDOC] 2.5 mM/l) into the pancreatic duct in combination with intraarterial infusion of cerulein (5 μg kg−1 h−1) for over 6 h as described by Schmidt et al.9 in detail. Bile-salt infusion into the pancreatic duct was performed in a volume- (1.2 ml/kg), time- (5 min), and pressure- (30 mmHg) controlled manner.In each of the models, animals received saline solution during the observation period (0.9%, 4 ml kg−1 h−1). Intravital microscopy was performed 12 h after the induction of pancreatic injury, and histological changes and blood samples were assessed 24 h after the infusions were started. Platelet preparation One milliliter of whole blood was withdrawn before intravital microscopy. Platelets were separated and stained according to the method originally described by Massberg et al.10 Briefly, platelets were stained by rhodamine 6G and separated by 2 cycles of centrifugation under the addition of prostacyclin. After suspending and washing the separated platelets, blood cell count was performed to calculate the number of platelets per microliter and to rule out animal-specific differences in the number of platelets. Platelets were then reinjected, and intravital microscopy was performed. Intravital microscopy The abdomen was reopened, and the pancreas was carefully exteriorized in a horizontal position through the midline incision after the animal was placed on the right side. The duodenal loop with the head of the pancreas was carefully fixed on an anatomically designed stage in a temperature-controlled (37°C) Ringer’s bath. Afterward, intravital microscopy was performed as described below. The animals were killed after the completion of intravital microscopy by a pentobarbital overdose. Erythrocyte and leukocyte assessment A 0.5 ml/kg of erythrocytes (hematocrit 50%) labeled with fluorescein isothiocyanate (FITC) as described before11 was applied intravenously. In addition, 1 ml/kg of rhodamine-6G solution was applied intravenously to label leukocytes in vivo.12 Intravital microscopy was performed after an equilibration period of 15 min using a fluorescent microscope (Leitz, Wetzlar, Germany) with a 20-fold water immersion objective. An epi-illuminescent xenon lamp with an excitation filter of 450–490 nm was used for visualization of FITC-labeled erythrocytes and an excitation filter of 540–630 nm for rhodamine-labeled leukocytes. Platelet assessment After platelet reinjection, intravital microscopy was performed by an epi-illuminescent xenon lamp with an excitation filter of 540–630 nm. Off-line analysis Images were transferred to a monitor and simultaneously recorded on a videotape recorder. In each animal, five capillary fields of the exocrine pancreas and five postcapillary venules (20–40 μm) were recorded for 1 min. Off-line analysis was performed using a specially designed computer program (Capimage, Dr. Zeintl, Heidelberg, Germany). Erythrocyte velocity and platelet velocity were determined for 10 cells in each capillary field and venule. Additionally, temporarily (rolling) and permanently (sticking) adherent leukocytes and platelets were determined in pancreatic venules and capillary fields. Rolling cells were defined as cells with less than 66% of red blood cell velocity, whereas sticking cells are those that were adherent to the vessel wall for the whole observation period.13 Edema A portion of pancreatic tissue was trimmed of fat and weighed. Pancreatic water content was determined by the ratio of the initial weight (wet weight) of the pancreas to its weight after incubation at 60°C for 72 h (dry weight). Histology The pancreas was immediately removed after killing and was fixed in 4% buffered formalin solution. It was then embedded in paraffin, cut, and stained with hematoxylin eosin. Histopathological evaluation was performed in a blinded fashion. For quantification of edema, inflammation, and necrosis, a modification of the scoring system originally described by Schmidt et al.9 was used, ranging from 0 to 3 (no pathological changes to severe injury). Assessment of thromboxane A2 Thromboxane A2 was measured in frozen serum by commercially available enzyme-linked immunosorbent assay (University of Freiburg, Germany). Statistical analysis Results are shown as mean ± SEM. Student’s t test was used when the data had a normal distribution, whereas Kruskal–Wallis and Mann–Whitney tests were utilized when the distribution was not normal. Statistical significance was accepted at the 5% level (p < 0.05). Results Serum amylase Twelve hours after the induction of AP, serum amylase increased significantly compared to control animals. Hyperamylasemia was comparable in both mild and severe AP indicating the presence of pancreatic cell damage. However, amylase was not a marker for the extent of tissue damage or disease severity (Table 1). Table 1Serum Parameters, Wet–Dry Ratio, and Histopathology ControlMild APSevere APSerum parametersAmylase (U/l)586 ± 11627,200 ± 4,012*27,317 ± 3,220*Thromboxane A2 [pg/50 μl]15.3 ± 10.347.8 ± 12.1*61.9 ± 15.8*Wet–dry ratio2.87 ± 0.796.96 ± 0.95*4.77 ± 0.70HistopathologyInflammation0.25 ± 0.421.31 ± 0.08*1.95 ± 0.17*†Necrosis0.08 ± 0.201.10 ± 0.11*1.70 ± 0.23*†*p < 0.05 vs control group†p < 0.05 vs mild acute pancreatitis Serum thromboxane A2 Thromboxane A2, as a marker of platelet activation, showed significantly higher levels in both AP groups compared to control animals after 24 h. Thromboxane liberation correlated with severity of AP, with the highest levels being present in animals with necrotizing AP (Table 1). Intravital microscopy Erythrocyte velocity decreased significantly in mild as well as severe AP in both capillaries and venules compared to control animals. Platelets showed comparable flow features. Flow velocity decreased under both AP conditions, with a highly significant decrease in severe AP in venules and capillaries (Table 2). These changes were paralleled by increased interaction between leukocytes and endothelium (Table 2). Platelet adhesion in capillaries and venules increased significantly in both mild and severe AP (Figs. 1 and 2). Reversible adhesion (rolling platelets) were comparable during both forms of AP, whereas the increase in irreversible adhesion (sticking platelets) depended on the severity of AP and showed peak platelet–endothelium adherence in necrotizing AP (Figs. 1 and 2). Table 2Results of the Intravital MicroscopyIntravital microscopyControlMild APSevere APErythrocyte velocity (capillary) (mm/s)0.65/0.020.42/0.01*0.36/0.01*Erythrocyte velocity (venule) (mm/s)0.93/0.110.77/0.170.58/0.10*†Platelet velocity (capillary) (mm/s)0.54 ± 0.040.35 ± 0.03*0.29 ± 0.03*Platelet velocity (venule) (mm/s)0.67 ± 0.050.63 ± 0.020.53 ± 0.05*Rolling leukocytes (capillary)1.3 ± 0.24.5 ± 1.4*9.0 ± 1.7*†Rolling leukocytes (venule)1.3 ± 0.214.8 ± 1.2*18.9 ± 1.9*Sticking leukocytes (capillary)1.1 ± 0.310.2 ± 1.8*7.2 ± 0.7*Sticking leukocytes (venule)0.7 ± 0.15.6 ± 0.9*13.5 ± 2.0*†*p < 0.05 vs control group†p < 0.05 vs mild acute pancreatitisFigure 1Intravital microscopy, capillary platelet adhesion (one per field). Control group (gray), mild acute pancreatitis (white), and severe acute pancreatitis (striped). Reversible platelet adhesion in mild and severe acute pancreatitis (left columns); irreversible platelet adhesion (right columns). *p < 0.05 vs control group, †p < 0.05 vs mild acute pancreatitis.Figure 2Intravital microscopy, venular platelet adhesion (one per 100 μm). Control group (gray), mild acute pancreatitis (white), and severe acute pancreatitis (striped). Reversible platelet adhesion in mild and severe acute pancreatitis (left columns); irreversible platelet adhesion (right columns). *p < 0.05 vs control group, †p < 0.05 vs mild acute pancreatitis. Tissue edema (wet/dry ratio) Supramaximal cerulein stimulation induced a significant increase in pancreatic water content compared to control animals. In contrast, there was only a slight increase in tissue edema after GDOC treatment (Table 1). Histopathology Control animals showed no histopathological changes after sham operation and 24 h infusion therapy. Histopathology of mild AP was characterized by significant edema formation, inflammatory tissue infiltration, and acinar cell necrosis. In severe AP, the changes regarding inflammation and necrosis were significantly more pronounced (Table 1). Discussion In the present study, we have investigated platelet function in experimental models of AP. We chose two animal models to induce a mild edematous or a severe necrotizing course of AP. Both models are established, well characterized, and have been used in numerous studies.9,14,15 The induction of AP in these models results in a standardized grade of tissue damage, either mild or severe, with very little variance within each group. Therefore, the use of these models allows us to rule out the significant influence of preparatory or other methodological problems on the comparability of the results. Analysis of platelet function by intravital microscopy has been established and standardized for examination of liver and small bowel perfusion by Massberg et al.10 We have modified this method to investigate the pancreas.15 In the present study, we could demonstrate that this method is not only suitable for the examination of healthy pancreas but also for the detailed analysis of pancreatic microcirculation in mild and severe AP. Acute pancreatitis is characterized by an impairment of microcirculation due to an activation of inflammatory cells with a consecutive increase of leukocyte–endothelium interaction. These pathophysiological events mediate an inflammatory tissue infiltration, edema, and hemorrhagic lesions. While the inflammatory response is well investigated, the platelet function and the role of the coagulation cascade have not yet been investigated in detail. It is well known that the inhibition of certain coagulatory steps, e.g., by applying hemodiluting or anticoagulatory substances, improves the outcome of AP.16,17 Coagulation and hemostasis comprise two interacting pathways: humoral coagulatory factors leading to the activation of fibrinogen as the final step of the coagulation cascade and cellular factors, which are represented by activated platelets. Different mechanisms of platelet interaction are responsible for their physiological function, namely, interactions with endothelium, leukocytes, and humoral coagulatory and inflammatory proteins.18,19 In the present study, we could demonstrate that the platelet–endothelium interaction increases during AP and correlates with the degree of its severity. Comparable to leukocyte–endothelium interaction, temporary and permanent adhesions of platelets to the vessel wall were evident in our experiments. This correlates with the activation patterns that have been observed in vivo in ischemia models of the liver and the pancreas,15,20 as well as in vitro.21 Therefore, it seems likely that these activation patterns reflect the severity of the pancreatic affection, leading to reversible adhesion in mild AP and irreversible adhesion in more severe organ affection. Especially, the firm adhesion of platelets contributes to microcirculatory disturbances and may induce perfusion failure and tissue necrosis in the progression to severe AP. The significantly elevated thromboxane levels correlate well with platelet activation and microcirculatory failure observed during intravital microscopy. The increase in serum thromboxane elucidates one mechanism of our results as it executes a direct platelet stimulation and leads to the conversion of “resting” to “activated” platelets with the consecutive adhesive action. Furthermore, thromboxane does not only activate platelets but also acts as a complex pathophysiological mediator with multiple other targets. Its effects include leukocyte activation, upregulation of proinflammatory cytokines, and strong vasoconstricting effects. These are mediated via phosphatidylcholine and phosphatidylcholine-specific phospholipase-C pathway leading to a tonic contraction in smooth muscles and upregulating other vasoactive substances.22,23 Especially, this vasoconstrictor mechanism may additionally contribute to perfusion failure in the course of AP as observed in our study. How far platelet inhibition itself could be an approach to attenuate the course of AP experimentally or clinically is hypothetical but should certainly be addressed to in further studies. Possible aims could be adhesion molecules such as selectins or platelet receptors and also synthesis of thromboxane and prostaglandins. Platelet activation was accompanied by leukocyte activation in the present study. An interaction between these two cell types has been demonstrated by the different authors in the past.24–26 Among others, P-selectin seems to be one of the most important adhesion molecules, which links the inflammatory and procoagulatory cascades and has the potency to activate leukocytes and platelets as the cellular elements of either pathway.18,19,25,26 Besides their adherence to endothelial cells, activated platelets form stable aggregates with leukocytes. This results in a combined inflammatory and coagulatory contribution to thrombus formation and is also mediated by P-selectin and beta-integrins.27–29 Especially, the formation of microthrombotic vessel occlusion with microcirculatory perfusion failure and consequent ischemia, hypoxia, and tissue necrosis was reflected by the intravital microscopic results in the present study. Conclusion The results of the present study show that activation and adhesion of platelets play an important role during AP. Platelet–endothelium and platelet–leukocyte interactions as well as thromboxane liberation show a correlation with the severity of experimental AP and seem to be of distinct importance in the progression from mild to severe necrotizing AP. A possible therapeutic use of these pathophysiological events should be evaluated in further studies.

Photosynth_Res-3-1-1769343

Optical spectroscopic studies of light-harvesting by pigment-reconstituted peridinin-chlorophyll-proteins at cryogenic temperatures

Low temperature, steady-state, optical spectroscopic methods were used to study the spectral features of peridinin-chlorophyll-protein (PCP) complexes in which recombinant apoprotein has been refolded in the presence of peridinin and either chlorophyll a (Chl a), chlorophyll b (Chl b), chlorophyll d (Chl d), 3-acetyl-chlorophyll a (3-acetyl-Chl a) or bacteriochlorophyll a (BChl a). Absorption spectra taken at 10 K provide better resolution of the spectroscopic bands than seen at room temperature and reveal specific pigment–protein interactions responsible for the positions of the Qy bands of the chlorophylls. The study reveals that the functional groups attached to Ring I of the two protein-bound chlorophylls modulate the Qy and Soret transition energies. Fluorescence excitation spectra were used to compute energy transfer efficiencies of the various complexes at room temperature and these were correlated with previously reported ultrafast, time-resolved optical spectroscopic dynamics data. The results illustrate the robust nature and value of the PCP complex, which maintains a high efficiency of antenna function even in the presence of non-native chlorophyll species, as an effective tool for elucidating the molecular details of photosynthetic light-harvesting. Introduction Photosynthetic organisms have bioengineered a wide variety of protein-based structures to carry out photosynthetic light-harvesting (Frigaard et al. 2004). High-resolution molecular structures of many of these systems are providing important clues into the molecular features that control light-harvesting in photosynthesis (Germeroth et al. 1993; Hofmann et al. 1996; Hu et al. 1995; Koepke et al. 1996; Kühlbrandt et al. 1994; Law et al. 2004; Liu et al. 2004; McDermott et al. 1995; Roszak et al. 2003). To understand fully how these naturally occurring pigment–protein complexes harvest light so efficiently, it is necessary to employ spectroscopic tools that reveal directly the dynamics and efficiency of energy transfer and the nature of the excited energy states associated with the bound pigments. The spectroscopic studies are aided by the ability to make systematic alterations in the structures of the light-harvesting complexes, and then to examine how these changes manifest themselves in the spectroscopic observables, dynamics of energy transfer and efficiency of antenna function (Akahane et al. 2004; Cammarata et al. 1990; Chadwick et al. 1987; Crielaard et al. 1994; Croce et al. 1999; Davidson and Cogdell 1981; Frank 1999; Fraser et al. 1999; Morosinotto et al. 2002; Olivera et al. 1994; Olsen et al. 1997; Plumley and Schmidt 1987; Remelli et al. 1999; Struck and Scheer 1991; Struck et al. 1992). The peridinin-chlorophyll-protein (PCP) complex from the dinoflagellae, Amphidinium carterae, is an example where many of these approaches have been applied. Atomic resolution structural analysis by X-ray diffraction to 2.0 Å resolution (Hofmann et al. 1996) has been augmented by pigment reconstitutions and protein refolding studies (Miller et al. 2005; Polívka et al. 2005). Steady-state and time-resolved spectroscopy (Bautista et al. 1999a; Kleima et al. 2000a; Kleima et al. 2000b; Krueger et al. 2001; Linden et al. 2004; Shima et al. 2003; Zigmantas et al. 2002; Zimmermann et al. 2002) has been carried out on the excited singlet and triplet states associated with the bound pigments. The X-ray crystallographic analysis of PCP revealed a trimeric structure of subunits (Hofmann et al. 1996) where eight peridinins and two chlorophyll a (Chl a) molecules (Fig. 1) bind non-covalently in a roughly C2 symmetrical arrangement in each of the three individual subunits. The peridinins are assembled in clusters of four within the subunits, surrounding, and in van der Waals contact with, one chlorophyll a (Chl a) (Fig. 1). The two Chl molecules within a subunit have a center-to-center distance of 17.4 Å. Steady-state and ultrafast optical spectroscopic investigations on peridinin in solution and on the main-form PCP (MFPCP) have explored its photoinduced excited-state processes and energy transfer pathways (Carbonera et al. 1999; Damjanovic et al. 2000; Kleima et al. 2000; Krueger et al. 2001; Ritz et al. 2000; Shima et al. 2003; Zimmermann et al. 2002). In addition, a variant form, the high-salt PCP (HSPCP), has been studied in the same manner (Ilagan et al. 2004; Sharples et al. 1996). Fig. 1Structure of the pigments in a monomeric subunit of the MFPCP complex. The coordinates were taken from Protein Data Bank with 1PPR code. Also shown are the structures of peridinin and the chlorophylls used in this study Adding to the arsenal of PCP complexes that can be used for structural and spectroscopic studies, Miller and coworkers (Miller et al. 2005) prepared and Polívka and coworkers (Polívka et al. 2005) studied the N-domain and full-length PCP apoproteins expressed in Escherichia coli and reconstituted with the total pigment extract from native PCP to produce a fully functional complex with a high peridinin-to-Chl a energy transfer efficiency. Also, they demonstrated that in the N-domain PCP apoprotein the bound Chl a molecules could be replaced by other Chls using the N-domain apoprotein incorporating the requisite stoichiometric amount of peridinin. In this paper, we present a low temperature, steady-state, optical spectroscopic investigation of the spectral features of PCP complexes in which recombinant apoprotein has been refolded in the presence of peridinin and either chlorophyll a (Chl a), chlorophyll b (Chl b), chlorophyll d (Chl d), 3-acetyl-chlorophyll a (3-acetyl-Chl a) or bacteriochlorophyll a (BChl a). We use fluorescence excitation spectroscopy to examine in detail the peridinin-to-Chl energy transfer efficiencies of the various complexes at room temperature and correlate the results with previously reported ultrafast, time-resolved optical spectroscopic dynamics data (Polívka et al. 2005). The low temperature data provide better resolution of the spectroscopic bands than seen at room temperature and reveal specific pigment–protein interactions responsible for the observed Chl transition energies. Materials and methods Sample preparation The PCP complexes were reconstituted from N-domain apoprotein with peridinin and either Chl a, Chl b, Chl d, 3-acetyl-Chl a or BChl a (Fig. 1) as described (Miller et al. 2005). Size exclusion chromatography shows the reconstituted PCP to be approximately 40 kDa in apparent molecular mass; i.e., the same as that of the native PCP, suggesting that the samples used here consist of monomeric protein structures. The purified reconstituted PCP complexes were stored at 4°C until ready for use in the spectroscopic experiments. For the cryogenic experiments, the PCP complexes were dissolved in a buffer containing 50 mM Tricine, 20 mM KCl buffer and 60% (v/v) glycerol at pH 7.6. Chl a and Chl b were obtained from Sigma-Aldrich. Chl d and BChl a were extracted from the cyanobacterium, Acaryochloris marina, and the bacterium, Rhodospirillum rubrum, respectively. Total pigments were extracted from A. marina cells with 2-butanol and the extract dried in a gentle stream of nitrogen and Chl d was purified by HPLC as described below. BChl a was extracted from R. rubrum whole cells with 50:50 (v/v) acetone:methanol. The extracted solution was filtered, dried and redissolved in 50:50 (v/v) petroleum ether:methanol. The methanol phase, in which BChl a was dissolved, was collected and dried with nitrogen gas. The BChl a solution was subjected to an alumina column chromatography with increasing percent of acetone in hexane. 3-Acetyl-Chl a was obtained as an oxidation product of BChl a during its isolation from R. rubrum. All Chls were further purified using a Millipore Waters 600E HPLC system employing a Nova-Pak C18 column. For Chl a and Chl b, isocratic runs of acetone:methanol:water (75:20:5, v/v/v) were used as the mobile phase with a flow rate of 0.5 mL min−1. For Chl d and BChl a, the mobile phase used a combination of solvent A (acetonitrile:methanol—50:50, v/v) and solvent B (solvent A:water—50:50, v/v). The run was programmed as follows: 0–5 min, linear gradient from 80% solvent A and 20% solvent B to 99% solvent A and 1% solvent B; and, 5–30 min, isocratic 99% solvent A and 1% solvent B with a flow rate of 1.0 mL min−1. 3-acetyl-Chl a was separated from BChl a using solvent A (acetonitrile:methanol:water—87:10:3, v/v/v) and solvent B (hexane:propanol—4:1, v/v) as the mobile phase. The mobile phase was programmed as follows: 0–20 min, linear gradient from 100% solvent A to 95% solvent A and 5% solvent B; and, 20–30 min, isocratic 95% solvent A and 5% solvent B with a flow rate of 1.0 mL min−1. The collected samples were dried with a gentle stream of nitrogen gas and then redissolved in 2-methyl tetrahydrofuran (2-MTHF). Spectroscopy Absorption spectra were taken at 10 K using a Cary 50 UV-visible spectrometer and a Janis model STVP100 helium vapor flow cryostat using 1 cm pathlength polymethacrylate cuvettes. The fluorescence and fluorescence excitation spectra of the PCP complexes were taken at 77 K using a Jobin Yvon Horiba Fluorolog-3 spectrofluorometer model FL3-22 on samples immersed in a custom built optical dewar (Kontes) as described (Ilagan et al. 2004). Results Absorption spectroscopy The 10 K absorption spectra of the PCP complexes reconstituted with peridinin and either Chl a, Chl b, Chl d, 3-acetyl-Chl a or BChl a are shown in Fig. 2. All of the complexes exhibit strong absorption in the region 450–575 nm associated with peridinin. For the complexes reconstituted with Chl a and Chl b, a double peak is evident in the Soret region (near 450 nm), but this is not seen in complexes containing Chl d, 3-acetyl-Chl a or BChl a. In the long-wavelength absorption region of the spectra, the Qy band of all the Chls is very sharp and appears at 646 nm for Chl b, 667 nm for Chl a, 682 nm for 3-acetyl-Chl a, 697 nm for Chl d and 782 nm for BChl a. All the Qy bands are blue-shifted compared to the spectra taken at room temperature. Also, except for Chl d that exhibits a weaker shift of 59 ± 6 cm−1, all are shifted by the same amount, 110 ± 15 cm−1. Also, in the PCP complex containing BChl a, a second peak on the long-wavelength side of the primary Qy feature is observed at 799 nm (Fig. 2) and a trace of 3-acetyl-Chl a can be seen in the absorption spectrum of the complex with BChl a. Fig. 2Absorption spectra of the PCP complexes reconstituted with peridinin and either Chl a, Chl b, Chl d, aChl a (3-acetyl-Chl a) or BChl a. The spectra were taken at 10 K and were normalized at 500 nm For comparison and analysis of the Chl spectra of the complexes, absorption spectra of the isolated and purified Chls in 2-MTHF were taken at 10 K. 2-MTHF forms a clear glass at that temperature and has been used previously in this type of analysis (Ilagan et al. 2004). Figure 3 shows the 10 K absorption spectra of the Chl solutions in the Qy region plotted on a wavenumber scale. The spectra were analyzed by Gaussian deconvolution and were well fit using a sum of two components that were allowed to vary in amplitude, position and width built on a broad baseline. The Gaussian deconvolutions of the solution spectra were all very similar (Fig. 3) and revealed a prominent spectral origin and a broad phonon wing at higher energy. The wavelengths of the prominent peaks computed from the wavenumber fits are 670 nm for Chl a, 654 nm for Chl b, 697 nm for Chl d, 689 nm for 3-acetyl-Chl a and 777 nm for BChl a. The results indicate the chlorophyll–protein interaction induces a blue-shift in the case of Chl a, Chl b and 3-acetyl-Chl a, no shift for Chl d and a red-shift for BChl a. Fig. 3Gaussian deconvolutions of the 10K absorption spectra in the Qy region of purified Chls in 2-MTHF solvent. The fit (dashed line) is the sum of two Gaussian components built on a broad background. The data for the Gaussian deconvolution of the Chl a spectrum was taken from a previous study (Ilagan et al. 2004) Figure 4 shows the 10 K absorption spectral profiles of the reconstituted PCP complexes in the Qy region on a wavenumber scale fitted using Gaussian functions that were also allowed to vary in amplitude, position and width. The spectra from the PCP complexes containing Chl a, Chl b and BChl a required four Gaussian components on a broad absorption background (Fig. 4A, B, E) for a good fit. The spectra from the complexes having Chl d and 3-acetyl-Chl a required only two Gaussian functions on a broad absorption background (Fig. 4C, D) for a good fit. Fig. 4Gaussian deconvolutions of the 10K absorption spectra in the Qy region of PCP complexes reconstituted with peridinin and: (A) Chl a; (B) Chl b; (C) Chl d; (D) aChl a (3-acetyl-Chl a); and (E) BChl a Spectral reconstruction Linear combinations of the absorption spectra of peridinin and the various Chls taken in 2-MTHF at 10 K were used to reconstruct the spectral line shapes of the complexes in the region 375–600 nm. According to previously published procedures (Ilagan et al. 2004), individual peridinin spectra were constrained to have identical intensities, but were allowed to vary with respect to wavelength. The spectra of the individual Chls were allowed to have any intensity and wavelength. For all of the complexes, two Chl and eight peridinin spectra were used to fit the lineshapes (Fig. 5A–E). The spectra of all the complexes were best fit using two blue-shifted peridinins and three separated pairs of peridinin spectra (Fig. 5A–E), one of which needed to be significantly red-shifted to accommodate the long-wavelength region of the spectra. In the complexes containing Chl a and Chl b, the Soret bands of the two Chls are split and have noticeably different intensities. This required two separate Chl spectra to reproduce the absorption lineshapes in this region (Fig. 5A, B; Table 1). No splitting of the Soret bands was evident for the complexes with Chl d, 3-acetyl-Chl a and BChl a, so the spectra were able to be fit in this region assuming that the two Chls were identical. A summary of the wavelengths of the spectral origins of the individual peridinins and the Chl bands assigned in this analysis is given in Table 1. Fig. 5Reconstruction of the 10K absorption spectra of the reconstituted PCP complexes in the 375–600nm region. The analysis was carried out by linearly combining individual 10 K absorption spectra of peridinin and Chl taken in 2-MTHF solvent as described in the textTable 1Spectral origins of the eight peridinins and the positions of the Soret peaks and Qy bands of the two Chls derived from the spectral analysis of the 10 K absorption spectra of the reconstituted PCP complexesPCP complexPeridininChlReferenceSoretQy123456781212MFPCP470485522522526526544544434439665667(Ilagan et al. 2004)HSPCPna489533533533533533na436441667676(Ilagan et al. 2004)Chl a472 ± 2486 ± 2510 ± 1510 ± 1529 ± 1529 ± 1548 ± 1548 ± 1434 ± 1440 ± 1667668Chl b470 ± 3485 ± 2508 ± 1508 ± 1527 ± 2527 ± 2546 ± 2546 ± 2458 ± 1461 ± 1644647Chl d474 ± 2490 ± 2513 ± 1513 ± 1530 ± 1530 ± 1546 ± 1546 ± 1456 ± 2456 ± 2697697aChl a467 ± 2482 ± 2511 ± 2511 ± 2530 ± 2530 ± 2543 ± 1543 ± 1446 ± 1446 ± 1683683BChl a472 ± 2486 ± 2507 ± 2507 ± 2531 ± 1531 ± 1545 ± 1545 ± 1365 ± 1365 ± 1782782The numbers are in nm units and the uncertainties represent the range of values that give an acceptable fit to the dataMFPCP is main-form PCP, HSPCP is high-salt PCP, aChl a is 3-acetyl-Chl a, na is not applicable Fluorescence spectroscopy Fluorescence spectra of the reconstituted PCP complexes at 77 K are shown in Fig. 6. The spectra display a prominent narrow band at 671 nm (full width at half maximum, FWHM = 8.5 nm) for Chl a, 647 nm (FWHM = 10 nm) for Chl b, 699 nm (FWHM = 8 nm) for Chl d, 686 nm (FWHM = 9.5 nm) for 3-acetyl-Chl a and 787 nm (FWHM = 14.5 nm) for BChl a. All of the spectra show structured vibronic features to the red of the main band. In the PCP complex with BChl a, a shoulder at 798 nm on the red side of the main peak is clearly evident. Fig. 6Fluorescence spectra of the PCP complexes reconstituted with peridinin and either Chl a, Chl b, Chl d, aChl a (3-acetyl-Chl a) or BChl a taken at 77 K. The spectra were normalized to their λmax values Fluorescence excitation spectroscopy Figure 7 shows the fluorescence excitation spectra of the complexes recorded at room temperature overlaid with their 1-T spectra, where T is transmittance. The Chl fluorescence was monitored at various wavelengths between 675 and 850 nm, and the fluorescence excitation spectra were found to be independent of the monitoring wavelength. The peridinin-to-Chl energy transfer efficiencies obtained from the spectra were averaged over the 450–515 nm wavelength range and were: 94 ± 2% (Chl a), 92 ± 2% (Chl b), 96 ± 3% (Chl d), 99 ± 1% (3-acetyl-Chl a) and ∼100% (BChl a). Fig. 7Overlay of the room temperature fluorescence excitation (ex) and 1-T (where T is transmittance) spectra of reconstituted PCP complexes containing: (A) Chl a; (B) Chl b; (C) Chl d; (D) aChl a (3-acetyl-Chl a); and (E) BChl a. The intensity of each fluorescence excitation spectrum was normalized to correspond to the intensity of its corresponding 1-T spectrum in the Qy region Discussion Low temperature absorption spectra The 10 K absorption spectra of the reconstituted PCP complexes (Fig. 2) show significantly higher resolution than the spectra taken at room temperature (Polívka et al. 2005). This allows a more detailed look at how changes in the structure of the incorporated Chl affect the spectral profiles. The 10 K lineshape of the PCP complex reconstituted with Chl a (Fig. 2) is very similar to that from the native PCP complex (Ilagan et al. 2004), the major characteristics being the splitting of the Soret band near 435 nm and the shoulder on the red side of the peridinin maximum. These features are clearly evident (Figs. 2, 5A) in both reconstituted and native complexes (Ilagan et al. 2004). This supports the notion that the structures of these two complexes are very similar. The 10 K absorption spectra of the reconstituted PCP complexes (Fig. 2) also display blue-shifted Qy bands compared with thse spectra of the complexes taken at room temperature, and except for Chl d that exhibits a shift of 59 ± 6 cm−1, are all shifted by 110 ± 15 cm−1. This is opposite to the low temperature-induced red-shift observed for the 850 nm-absorbing BChl molecules in LH2 complexes from photosynthetic bacteria which was explained in terms of a structural change of the LH2 complex upon glass formation at low temperatures that increased excitonic coupling between the BChls (Wu et al. 1997). This cannot be the case for the PCP complexes here because the two Chl molecules are too far apart (17.4 Å) to be excitonically coupled. However, the blue-shift may result from a slight change in the planarity of the Chl macrocycles. Assuming that glass formation induces a structural change (Wu et al. 1997) in the immediate environment of the Chls in the PCP complexes, one may expect a corresponding change in the conformation of the macrocycles. It has been shown that these conformational changes may induce either red- or blue-shifts of Chl Qy bands (Gudowska-Nowak et al. 1990). The fact that the shift is very similar (110 ± 15 cm−1) for all chlorophyll species except Chl d suggests that the structural change affects the macrocycle in the vicinity of Ring I because all the Chls except Chl d have a large substituent in the position R1 (Fig. 1). Thus, if one assumes a low-temperature-induced protein structural change in the immediate environment of Ring I, this may lead to a distortion of the Chl macrocycle and consequently to the observed blue-shift. For Chl d, the smaller aldehyde substituent at position R1 may be less susceptible to protein structural changes and to inducing conformational distortion of the Chl macrocycle. The Gaussian deconvolutions of the Chl solution spectra in the Qy region (Fig. 3) serve as guides in the analysis of the absorption spectra of the reconstituted PCP complexes. Comparing the spectrum in the Qy region of the Chl a-reconstituted PCP (Fig. 4A) with that of Chl a in solution (Fig. 3A) shows that two distinct pairs of Chl a Gaussian functions built on a broad background are needed to reproduce the spectrum of that complex. Likewise, for the PCP complex containing Chl b, two pairs of Gaussian components having different prominent spectral peaks and zero-phonon lines built on a broad background were needed to reproduce the spectrum of the protein complex (Fig. 4B). However, for the PCP complexes containing Chl d or 3-acetyl-Chl a, only two Gaussian components on a broad background were required to fit the spectra in the Qy band region indicating that Chl d and 3-acetyl-Chl a in these complexes are in highly symmetric environments. This conclusion is supported by the observation that the PCP complexes reconstituted with Chl a and Chl b show a splitting in the Soret band region (Fig. 5A, B), but all the other complexes do not. This indicates a less symmetric environment for the Chls bound in the complexes containing Chl a or b. The observation that the spectra of the PCP complexes containing Chl a and Chl b show splittings in the Soret and Qy regions attributable to environmental asymmetry, but the spectra from the Chls in the other complexes do not, may also be traceable to the different functional groups on Ring I of the Chls. A view of the published structure of the MFPCP complex focusing on Chl 601 in the vicinity of Ring I shows that this part of the Chl macrocycle is in close proximity (within 4 Å) to peridinins 612 and 614 (Fig. 8). The distance between the vinyl group and the peridinins is closer than to any amino acid residue in the protein. Leucine-204 comes closest at ∼5 Å (Fig. 8). The splitting of the Soret and Qy spectral bands in the Chl a and Chl b PCP complexes implies that the vinyl group on Chl a and Chl b may adopt different configurations leading to distinct interactions between these Chls and the peridinins. In contrast, the carbonyl groups on Ring I of Chl d, 3-acetyl-Chl a and BChl a may have only a single configuration and may also sterically hinder the approach of peridinins 612 and 614 toward the Chl in those complexes, thereby minimizing the interaction between pigments, rendering the Chl spectra equivalent. Fig. 8A view of the structure of the MFPCP complex focusing on the Ring I of Chl 601. Per 612, Per 614 and Leu 204 are in close proximity to the Ring I of Chl 601. The coordinates of the structure were taken from Protein Data Bank with 1PPR code The results from the deconvolution of the spectral features for the PCP complex containing BChl a are unusual. In addition to the main Qy absorption band at 782 nm (Figs. 2, 4E), a shoulder at 799 nm is observed which is highly reminiscent of the 4 K absorption spectrum of the antenna BChl a–protein (Fenna–Matthews–Olson, FMO) complex from Prosthecochloris aestuarii (Johnson and Small 1991; Wendling et al. 2002; Whitten et al. 1978). The spectral features in the Qy region are attributed to a combination of site shifts and dipolar interactions among the seven BChl a molecules within one subunit of the FMO protein (Wendling et al. 2002). The crystal structure of this FMO complex has been determined to 2.8 Å resolution (Matthews et al. 1979) and later refined to 1.9 Å resolution (Tronrud et al. 1986). The FMO complex consists of three identical subunits and each subunit contains seven BChl a molecules where the center-to-center distances between BChls vary from 11 to 14 Å (Matthews et al. 1979). This is smaller than the center-to-center distance of 17.4 Å reported from the X-ray crystallographic analysis of the MFPCP complex containing Chl a (Hofmann et al. 1996). Why the BChl-containing complexes exhibit this long-wavelength feature in their spectra but the complexes containing the other Chl molecules do not, is unclear. Nevertheless, the Soret band and the Qy features of the BChl molecules in the PCP complex are reproduced (Figs. 4E, 5E) assuming only a single prominent feature and a phonon line. Thus, similar to the complexes containing Chl d or 3-acetyl-Chl a, this suggests a highly symmetric environment for the BChls bound in this reconstituted PCP complex. The Chl–protein interaction induces a blue-shift of the Qy band in the low temperature spectra of Chl a, Chl b and 3-acetyl-Chl a, no shift for Chl d, and a red-shift for BChl a compared to the spectra taken in 2-MTHF low temperature glasses (Figs. 3, 4). This may be explained in a manner similar to that used above to interpret the blue-shifts of the Qy bands induced upon lowering the temperature. Switching from a 2-MTHF glass to the protein may induce a conformational change of the macrocycle that leads to the observed shifts (Gudowska-Nowak et al. 1990). Structural differences between Chl species occur due to different functional groups attached at positions R1 and R2 (Fig. 1), but again, it appears that the R1 site is key. The small aldehyde group of Chl d does not provide sufficient interaction leading to distortion of the macrocycle either in 2-MTHF glass or in the protein. Thus, essentially no protein-induced shift is observed. In contrast, the vinyl group on Ring I of Chl a and Chl b respond to the protein environment and induce a structural change that is manifested spectroscopically as a blue-shift of the Qy band. The red-shift of BChl a in the protein can be attributed to the acetyl group in the position R1. This functional group has been shown to cause a substantial red-shift of absorption spectra of BChl a in LH2 complexes from photosynthetic bacteria (Cogdell et al. 2002). The rotation of the carbonyl group with respect to the plane of Ring I causes changes in π-electron conjugation and this argument has been used to interpret the difference between the B820 and B850 absorbing BChls in different LH2 complexes (Cogdell et al. 2002). However, if this were the sole effect, a protein-induced red-shift would be expected of the 3-acetyl-Chl a that also has an acetyl group in the R1 site. This is not observed. Thus, we must assume that the saturation of the –C=C– bond in Ring II for BChl also plays a role in determining the direction of the shift. Energy transfer efficiencies Light absorption by peridinin promotes the molecule from its ground state, S0, to an excited state, denoted S2. The S2 state of peridinin then decays in ∼100 fs to a lower-lying S1 state which cannot be directly accessed from S0 due to symmetry selection rules (Akimoto et al. 1996; Krueger et al. 2001; Zigmantas et al. 2003). Since the lifetime of the S1 state of peridinin in solution was found to be highly dependent on the polarity of the solvent (Bautista et al. 1999; Zigmantas et al. 2001; 2003), an intramolecular charge transfer (ICT) state was invoked to account for this behavior (Bautista et al. 1999), and quantum computations have supported this model (Vaswani et al. 2003). Moreover, it has been argued from pump-dump-probe experiments on peridinin in solution that S1 and ICT states are distinct but connected on a picosecond time-scale (Papagiannakis et al. 2004). However, other time-resolved and two-photon spectroscopic methods have suggested that it is the S1 itself that possesses the charge transfer character (Shima et al. 2003; Zigmantas et al. 2003). Since it is not completely resolved whether the S1 state in PCP is distinct from or the same as the ICT state, we shall use the notation S1/ICT to indicate both possibilities. The peridinin-to-Chl energy transfer efficiencies for the complexes can be computed assuming a two-state (S2 and S1/ICT) donor model (Fig. 9). The overall efficiency, ε, is given by Desamero et al. 1998: where kET2 and kET1 are the rate constants for energy transfer from the S2 and S1/ICT states of peridinin, kIC2 is the rate constant for internal conversion from S2 to S1/ICT state and kIC1 is the rate constant for S1/ICT to S0 internal conversion. The rate constants obtained directly from the ultrafast, time-resolved optical spectroscopic experiments carried out (Polívka et al. 2005) are shown in Table 2. The value for the rate constant for S2 to S1/ICT internal conversion, kIC2, is assumed to be 15 ± 1.5 ps−1 (τIC2 = 66 ± 6 fs) from fluorescence up-conversion measurement on the native PCP complex (Linden et al. 2004). The value for kIC1 is assumed to be 0.063 ps−1 (τIC1 = 16 ps) from a measurement of the efficiency and dynamics of energy transfer in the native PCP complex (Zigmantas et al. 2003). The overall energy transfer efficiencies calculated from Eq. 1 are compared with the values obtained here from the steady-state measurements (Fig. 7) and summarized in Table 2. As one can see from the table, the efficiencies determined using both techniques are high in all cases and also in very reasonable agreement with each other. Fig. 9Schematic diagram of energy levels and energy transfer pathways between peridinin and Chl in the PCP complexes. kET1 and kET2 (solid lines) are rate constants for energy transfer from the S2 and S1/ICT states of peridinin to Chl. kIC2 and kIC1 (dashed lines) are rate constants for internal conversion from S2 to S1/ICT and S1/ICT to S0, respectivelyTable 2Rate constants and energy transfer efficiencies obtained from the ultrafast, time-resolved optical spectroscopic experiments on the PCP complexes reconstituted with various ChlsPCP complexkET2 (ps−1)kIC2 (ps−1)kET1 (ps−1)kIC1 (ps−1)ε (from fluorescence excitation), %ε (from time-resolved absorption), %Chl a10 ± 2.515 ± 1.50.34 ± 0.0340.06394 ± 291 ± 7Chl b10 ± 2.515 ± 1.50.17 ± 0.0170.06392 ± 284 ± 14Chl d13 ± 3.015 ± 1.50.59 ± 0.0590.06396 ± 395 ± 2aChl a11 ± 2.815 ± 1.50.45 ± 0.0450.06399 ± 193 ± 4BChl a10 ± 2.515 ± 1.52.22 ± 0.220.063∼10098 ± 2kET2 and kET1 are the rate constants for energy transfer from the S2 and S1/ICT states of peridinin (Polívka et al 2005). kIC2 and kIC1 are the rate constants for internal conversion from S2 to S1/ICT and from S1/ICT to S0, respectively. kIC2 was obtained from fluorescence up-conversion measurement on the native PCP complex (Linden et al 2004). kIC1 was deduced from the efficiency and dynamics of energy transfer in the native MFPCP given in Ref. (Zigmantas et al 2002). kIC2 and kIC1 were assumed to be unchanged in all the PCP complexes. The peridinin-to-Chl energy transfer efficiency, ε, was obtained from steady-state fluorescence excitation experiments and from ultrafast, time-resolved absorption spectroscopic measurements at room temperature. The uncertainties in the values of ε from fluorescence excitation experiments were derived from the standard deviations obtained in the 450–515 nm wavelength range. The uncertainties in the values of ε from time-resolved absorption measurements were obtained by propagating the published errors in the dynamicsaChl a 3-acetyl-Chl a In this work, it was shown that optical spectroscopy carried out at low temperatures provides a more highly resolved view of the spectral features of light-harvesting complexes than seen at room temperature. The steady-state optical spectroscopic methods applied to PCP complexes in which recombinant apoprotein has been refolded in the presence of peridinin and either Chl a, Chl b, Chl d, 3-acetyl-Chl a or BChl a illustrate the robust character of the PCP protein which maintains high efficiency of peridinin-to-chlorophyll energy transfer function in the presence of non-native Chl species. The efficiencies of energy transfer deduced from the steady-state measurements are in good agreement with those from ultrafast, time-resolved experiments. The PCP complex continues to be an effective tool for elucidating the molecular details of photosynthetic light-harvesting via systematic modification of its structural components.

Breast_Cancer_Res_Treat-3-1-2001218

Letrozole as upfront endocrine therapy for postmenopausal women with hormone-sensitive breast cancer: BIG 1-98

The BIG 1-98 trial is a large, randomized, independently conducted clinical trial designed to compare the efficacy of upfront letrozole versus tamoxifen monotherapy and to compare sequential or up-front use of letrozole and/or tamoxifen as an early adjuvant therapy for patients with early breast cancer. We report on the results from the primary core analysis of the BIG 1-98 trial of 8,010 patients, which compares monotherapy with letrozole versus tamoxifen. This pre-planned core analysis allowed the use of patient data from the monotherapy arms of letrozole and tamoxifen and from the sequential arms prior to the drug switch point. Patients randomized to letrozole had a 19% improved disease-free survival (hazard ratio [HR] = 0.81; P = 0.003), due especially to reduced distant metastases (HR = 0.73; P = 0.001). A 14% risk reduction of fatal events in favor of letrozole was also observed (P = NS). The results from the monotherapy arms alone confirmed the findings from the primary core analysis. Based on the results from this trial, the aromatase inhibitor letrozole (Femara®) is currently recommended as a part of standard adjuvant therapy for postmenopausal women with endocrine-responsive breast cancer and has recently been approved in the early adjuvant setting in both Europe and the United States. A subsequent analysis after additional follow-up will address the question of monotherapy versus sequential therapy. Introduction and rationale During the last century the management of primary breast cancer has evolved from gross surgical intervention to a sophisticated approach involving surgical, radiotherapeutic, hormonal, chemotherapeutic, and targeted biologic strategies. Currently, pharmacologic treatment is tailored according to disease and patient characteristics, including tumor size, histologic grade, lymph-node involvement, hormone receptor (HR) status, and human epidermal growth factor receptor-2 (HER2) overexpression. The benefits from adjuvant chemotherapy, adjuvant hormonal therapy, and adjuvant therapy with trastuzumab have been well-established [1, 2]. Patients with early breast cancer presenting with estrogen receptor (ER)-positive and/or progesterone receptor (PgR)-positive tumors typically receive adjuvant hormonal therapy. Adjuvant therapy with tamoxifen, a selective estrogen receptor modulator (SERM), significantly reduces the risk of breast cancer recurrence [1]; the Early Breast Cancer Trialists Collaborative Group reported that 5 years of tamoxifen reduces the annual breast cancer death rate by 31% and is significantly more effective than just 1–2 years of tamoxifen in ER-positive breast cancer [1]. However, intrinsic estrogenic activity [3–5] and increased risk of endometrial cancer [6] and thromboembolism [7, 8] have emerged as disadvantages when using tamoxifen. Based on current published data, extending the course of adjuvant tamoxifen beyond 5 years is not beneficial [9] despite the persistent risk of relapse in patients with HR+ breast cancer [10]. In the first-line treatment of advanced breast cancer, the third-generation aromatase inhibitors (AIs) have shown superior or equivalent efficacy compared with tamoxifen [11–15]. Letrozole demonstrated significant superiority in time to progression and overall response rate [12] and, in addition, an early survival benefit [11]. This indicates that a subset of the breast tumors is inherently less sensitive to tamoxifen [16] and that resistance to tamoxifen is acquired more quickly [17]. This superiority of letrozole over tamoxifen in the advanced setting led to the hypothesis that this AI may also be superior to tamoxifen when administered in the adjuvant setting. The Breast International Group (BIG) 1-98 was designed, coordinated, analyzed, and reported by an independent academic group and currently is the largest ongoing adjuvant trial in breast cancer investigating the role of an AI. This review summarizes the design and results from the primary core analysis of the BIG 1-98 trial, which compared monotherapy with letrozole to monotherapy with tamoxifen and identified letrozole as a better alternative to tamoxifen in this setting. It also summarizes some of the additional published research using the BIG 1-98 database. Trial design and patients BIG 1-98 was a randomized, phase III, double-blind trial with two randomization options: two-arms (A or B) and four-arms (A, B, C or D) (see Fig. 1) for women with operable invasive HR+ (ER+ and/or PgR+) breast cancer. The treatment arms were: (A) initial therapy with tamoxifen for 5 years, (B) initial therapy with letrozole for 5 years, (C) initial therapy with tamoxifen for 2 years followed by letrozole for 3 years, or (D) initial therapy with letrozole for 2 years followed by tamoxifen for 3 years. The doses administered were 2.5 mg/day of letrozole and 20 mg/day of tamoxifen. Fig. 1BIG-98 trial design Between March 1998 and March 2000, 1,835 women were randomly assigned to arms A or B, and between April 1999 and May 2003 a further 6,193 were assigned to arms A, B, C, or D. The primary core analysis compared treatment effects of patients randomized to receive letrozole initially (arms B and D) and those assigned to receive tamoxifen initially (arm A and C). In the sequential treatment groups (arms C and D), only events that occurred up to 30 days after switching treatments were included in the analysis. In total, 8,010 patients were included in the primary core analysis: 4,003 in the initial letrozole group (arms B and D) and 4,007 in the initial tamoxifen group (arms A and C) (see Fig. 2). Patient characteristics were similar in the two treatment groups. Overall, the trial included 41.3% node-positive and 57.3% node-negative women, while 1.4% had unknown nodal status. Positive HR status was an eligibility criterion. The majority of patients (63.1%) had both ER+ and PgR+ tumors. A total of 25.3% of women received adjuvant or neoadjuvant chemotherapy. Fig. 2CONSORT (Consolidated Standards of Reporting Trials) flowchart of the BIG 1-98 trial. The primary core analysis includes all 8,010 assessable patients, but events and follow-up in the sequential treatment groups (L → T and T → L) are truncated at 30 days after switching to the other treatment. L denotes letrozole and T tamoxifen. Reprinted from [18, Supplementary Appendix] End points The primary end point of the trial was disease-free survival (DFS), defined as the time from randomization to first occurrence of: invasive recurrence in ipsilateral breast, chest wall, regional site (internal mammary/axilla), or distant site (including ipsilateral supraclavicular); contralateral breast cancer (invasive); second malignancy (non-breast); or death without prior cancer event. Protocol-specified secondary end points in the BIG 1-98 trial were: overall survival (OS), defined as time from randomization to death from any cause; systemic DFS, defined as time from randomization to distant recurrence; second non-breast malignancy or death from any cause (ignoring local and contralateral-breast events); and safety. Three additional end points were defined in the statistical analysis plan: (1) DFS excluding second, non-breast malignancies; (2) time to recurrence, defined as time from randomization to first breast cancer recurrence (excluding second, non-breast cancers and censoring data on patients who died without a prior cancer event); and (3) time to distant recurrence, defined as the time from randomization to the first breast cancer recurrence at a distant site. Efficacy analyses Primary, secondary, and additional end points At 25.8 months of follow-up, letrozole improved DFS by 19% (P = 0.003). The cumulative incidence of breast cancer relapse was significantly reduced with letrozole compared with tamoxifen (see Fig. 3). The difference was evident from 1 year after randomization and at 5 years was 10.3% in the letrozole group compared with 13.6% in the tamoxifen group (P < 0.001). In addition, letrozole significantly improved systemic DFS compared with tamoxifen (hazard ratio 0.83; 95% CI 0.72–0.97). There was significant improvement with letrozole compared with tamoxifen for the additional end point of DFS excluding second non-breast cancers (hazard ratio 0.79; 95% CI 0.68–0.92). Letrozole was particularly effective in decreasing the risk of distant recurrence by 27% compared with tamoxifen (hazard ratio 0.73; 95% CI 0.60–0.88; P = 0.001) (see Fig. 4). A non-significant 14% improvement in OS was observed in patients receiving letrozole. Thus, 166 deaths (4.1%) were observed in the letrozole group compared with 192 deaths (4.8%) in the tamoxifen group. Fig. 3Cumulative incidence of a breast cancer relapse in the BIG 1-98 trial. Reprinted from [18] with permission from the Massachusetts Medical SocietyFig. 4Cox proportional-hazards model of data from the BIG 1-98 trial. The size of the boxes is inversely proportional to the standard error of the hazard ratio. The dashed vertical shows the hazard-ratio estimate for the overall analysis of the primary study end point (disease-free survival). Reprinted from [18] with permission from the Massachusetts Medical Society Subgroup analyses Prospectively planned subgroup analyses, using the Cox proportional-hazards model, for primary, secondary, and additional end points are summarized in Fig. 4. Subgroup analysis showed that the beneficial effect of letrozole on DFS was seen in ER+ tumors irrespective of PgR receptor status or patient age [18]. The role of ER and PgR in trials comparing AIs with tamoxifen remains an area of continued research [19, 20]. To explore this further, a central assessment of ER, PgR, and HER2 was recently completed for 6,500 patients in BIG 1-98. Results from the first 3,533 patients on the two monotherapy arms with ER+ tumors (by central assessment) and centrally assessed PgR and HER2 confirmed the results reported above from the local assessment of receptor status [18, 21], indicating that PgR status in ER+ tumors does not predict responsiveness to letrozole when compared with tamoxifen. The small group of patients with HER2 overexpression/amplification in the tumor had a higher rate of recurrence with both treatments, but the superiority of letrozole over tamoxifen was similar irrespective of HER2 status [21]. For the primary end point of DFS, the relative risk reductions for letrozole compared with tamoxifen were 29% in patients with node-positive tumors (hazard ratio 0.71, 95% CI 0.59–0.85), 24% in patients with tumors >2 cm (hazard ratio 0.76, 95% CI 0.63–0.92), and 30% in patients with prior chemotherapy (hazard ratio 0.70, 95% CI 0.54–0.92). An additional logistic regression analysis of BIG 1-98 was performed to retrospectively identify clinical and pathological factors predictive of early breast cancer recurrence [22]. The final model, based on 5,980 patients from the four-arm option and 212 events, identified the following significant factors: tumor size (P < 0.001), ER/PgR status (P < 0.001), node positivity (P < 0.001), and tumor grade (P < 0.001). There was a significant interaction between node positivity and treatment (P = 0.003). Patients with the greatest risk of recurrence had ≥4 positive nodes, tumors ≥5 cm, ER+/PgR− tumors, and grade 3 tumors. The increase in risk associated with increased node positivity was greater for patients randomized to tamoxifen than to letrozole [22]. Letrozole-only versus tamoxifen-only arms The superiority of letrozole was confirmed in a protocol-defined supplementary analysis, which was restricted to the 4,922 patients randomized to the monotherapy tamoxifen or letrozole arms. At a median follow-up of 51 months, letrozole provided a significant benefit for the end points DFS (P = 0.007), DFS excluding secondary malignancy (P = 0.01), time to recurrence (P = 0.004), and time to distant recurrence (P = 0.03) (see Table 1) [23]. Table 1Efficacy end points in patients randomized to treatment with letrozole (n = 2,463) or tamoxifen (n = 2,459) for 5 years in the BIG 1-98 trial [23]End pointEvents Hazard ratio95% CIP-valueLetTamDFS (primary protocol definition)3524180.820.71–0.950.007Overall survival1942110.910.75–1.110.35Systemic DFS3313740.870.75–1.010.07DFS (ignoring second non-breast cancer)3073640.830.71–0.960.01Time to recurrence2312910.780.65–0.920.004Time to distant recurrence1932340.810.67–0.980.03DFS disease-free survival, Let letrozole, Tam tamoxifen, CI confidence interval Safety All adverse events were graded according to the Common Toxicity Criteria of the National Cancer Institute (version 2). Predefined adverse events were specifically asked and documented at each study visit. Furthermore, the IBCSG Coordinating Center conducted a medical review (reviewers were blinded to randomization) of all grade 3–5 cardiovascular events and other grade 3–5 adverse events that were considered clinically relevant but whose cause was unclear, and all deaths of women in whom there was no prior cancer-related event. The results of BIG 1-98 showed that letrozole was well-tolerated and had a safety profile different from tamoxifen (see Table 2) [18]. A more detailed analysis of cardiovascular side effects, including baseline risk factors and cholesterol values over time, has recently been presented [24]. Table 2Cardiovascular adverse events and significant other adverse events among patients included in the BIG 1-98 safety analysis [18]Adverse event Incidence of any grade (%)P-valueLetrozole (n = 3,975)Tamoxifen (n = 3,988)Cerebrovascular accident or transient ischemic attack1.01.00.91Thromboembolic event1.53.5<0.001Cardiac event4.13.80.61Other cardiovascular event0.50.20.04Vaginal bleeding3.36.6<0.001Hot flashes33.538.0<0.001Night sweats13.916.20.004Fracture5.74.0<0.001Arthralgia20.312.3<0.001 Tamoxifen compared with letrozole was associated with an increased risk of thromboembolism, vaginal bleeding, and more endometrial biopsies (9.1% vs. 2.3%, respectively; P < 0.001), with a higher incidence of invasive endometrial cancers (0.3% vs. 0.1%, respectively; P = 0.18). In addition, tamoxifen was associated with a higher incidence of hot flushes (38.0% vs. 33.5%, respectively; P < 0.001) and night sweats (16.2% vs. 13.9%; respectively; P = 0.004). There was a higher incidence of arthralgia and skeletal events with letrozole compared with tamoxifen, including a higher rate of fractures (5.7% vs. 4.0%, respectively; P < 0.001) and a shorter time to first fracture. Hypercholesterolemia was among the adverse events listed on the case-report forms and was graded at each study visit during treatment [18]. A total of 43.6% of letrozole-treated and 19.2% of patients in the tamoxifen group had hypercholesterolemia, reported at least once during treatment [18]. Nevertheless, more than 80% of reported hypercholesterolemia was grade 1, and thus of uncertain clinical significance. In addition, serum total cholesterol values remained stable throughout the trial in the letrozole arm but decreased in the tamoxifen arm by approximately 13%, which is consistent with the known lipid-lowering effect of tamoxifen [25]. Thus, median changes in cholesterol values from baseline were 0, 0, and −1.8% at 6, 12, and 24 months in the letrozole group and −12.0, −13.5, and −14.1% in the tamoxifen group. The overall incidence of cardiovascular events was similar for the letrozole and tamoxifen groups. Although there were higher incidences of grade 3–5 cardiac events and cardiac failure in the letrozole arm, the incidences were low in both groups (2.1% vs. 1.1%; P < 0.001 and 0.8% vs. 0.4%, P = 0.01, respectively) in this older patient population at competing risk for cardiovascular events [18]. Discussion Clinical implications of BIG 1-98 The results of the primary core analysis [18], which included all available data from patients randomly assigned to the monotherapy arms and data from the sequential therapy arms censored at the time of the therapy switch, as well as the results from a recently published analysis limited to patients randomly assigned to the continuous therapy arms [23], demonstrate a significant benefit of letrozole over tamoxifen. Letrozole is at least as well-tolerated as tamoxifen, offering patients and physicians a true alternative. However, bone metabolism is differently affected by letrozole and tamoxifen. Patients receiving AIs are at increased risk for bone loss and osteoporosis and should therefore receive appropriate monitoring and medical intervention as part of daily practice. As a result of the findings from BIG 1-98, letrozole was approved both in Europe and the United States as an early adjuvant treatment for postmenopausal women with HR+ breast cancer. The 2007 St. Gallen international consensus guidelines [26] and updated National Comprehensive Cancer Network (NCCN) guidelines [27] recommend letrozole as an option for adjuvant treatment of early breast cancer. In the 2007 St. Gallen Guidelines, use of an AI is considered as an alternative to tamoxifen for postmenopausal women with low-, intermediate-, or high-risk tumors that are classified as endocrine-responsive or endocrine response uncertain [26]. Similarly, in the NCCN guidelines, letrozole is a recommended adjuvant hormonal therapy for all postmenopausal women with hormone-responsive tumors, regardless of HER2 status [27]. The current guidelines do not recommend one AI over another but emphasize that treatment should be selected on the basis of clinical trial evidence in specific settings [27], nor do the guidelines provide recommendations as concerns the optimal use of the AIs as upfront monotherapy or sequenced with tamoxifen. Results from the Arimidex, Tamoxifen, Alone or in Combination (ATAC) trial provided evidence for superior DFS with anastrozole versus tamoxifen used as initial adjuvant hormonal therapy in postmenopausal women with HR+ breast cancer [28, 29]. At a median follow-up of 68 months, no survival advantage has been observed in the ATAC trial, and it remains an open question whether the DFS advantage observed in AI trials will translate into an OS advantage. Both letrozole and anastrozole have demonstrated superiority over tamoxifen as initial adjuvant therapy [18, 28], but a direct comparison of letrozole with anastrozole awaits the results of a randomized head-to-head trial (Femara Anastrozole Clinical Evaluation [FACE]) [30, 31]. The adjuvant FACE trial compares upfront therapy with letrozole 2.5 mg with anastrozole 1 mg daily for up to 5 years in postmenopausal, HR+, node-positive breast cancer patients. In addition, recruitment of a direct comparison of anastrozole and exemestane (MA.27) has been completed recently. Other trials demonstrated better disease control when an AI was given after 2–3 years of adjuvant tamoxifen [32], but so far, no trial has reported on a regimen of an AI given 2–3 years before tamoxifen. Initial treatment with the “gold standard” tamoxifen followed by an AI may be a logical long-term strategy because of the lack of complete cross-resistance between these hormonal strategies. On the other hand, the greater anti-estrogenic potency and higher anti-tumor activity of AIs over tamoxifen, as demonstrated in preclinical models and randomized clinical trials [12, 17, 33, 34], may suggest that it is preferable to use an AI upfront to avoid early relapses that may occur while on tamoxifen therapy. Thus, the key question, “should AIs be given as initial therapy or used sequentially after tamoxifen?” is as yet unanswered. Physicians often extrapolate data from switch trials, e.g., the Intergroup Exemestane Study [35, 36] or the MA.17 trial [33, 37] to sequential trials (e.g., BIG 1-98, Austrian Breast and Colorectal Cancer Study Group [ABCSG] 8 [38]). In these sequential trials, events are included in the analysis from treatment start and not from point of switch after 2–3 years of tamoxifen. Sequential and switch trials investigate obviously the same intervention but are conducted in different patient groups, thus, results are expected to be different and are different indeed. Best use of AIs remains an open question, at least until results of BIG 1-98 from the sequential use of letrozole and tamoxifen, in comparison with continuous monotherapy, as well as from the Tamoxifen and Exemestane Adjuvant Multicenter trial investigating exemestane monotherapy versus tamoxifen followed by exemestane [39] and from an updated analysis of ABCSG-8, become available. Conclusions BIG 1-98 has shown that the AI letrozole results in better disease control than tamoxifen when given as initial endocrine therapy for postmenopausal women with hormone-responsive early breast cancer. Letrozole significantly reduced the risk of recurrence and of distant recurrence and has a reasonable safety profile [18]. The comparison between the monotherapy and the sequential treatment arms within the BIG 1-98 trial are eagerly awaited and are expected to have an important impact on the management of breast cancer.

Biotechnol_Lett-4-1-2248218

Purification and bioactivity of exendin-4, a peptide analogue of GLP-1, expressed in Pichia pastoris

Exendin-4, a peptide analogue of glucagon-like peptide-1 (GLP-1), has been developed for treatment of type 2 diabetes. Herein, the secretive exendin-4 fusion protein, expressed by methanol induction in Pichia pastoris system, was purified to homogeneity by chromatography followed by enterokinase cleavage of the fusion protein and subsequent purification of the recombinant exendin-4. Purity of the recombinant exendin-4 was 95.6%. Bioactivity assay revealed that it had glucose-lowering and insulin-releasing action in vivo. Introduction Exendin-4 is a 39 amino acid peptide (53% structural homology to GLP-1) and was first isolated from the salivary secretions of Gila monster lizard (Heloderma suspectum) (Eng et al. 1992). It shares many of the glucoregulatory actions with GLP-1 (Buse et al. 2004; Kendall et al. 2005) and has aroused great attention for its potential for the treatment of diabetes. Clinical and non-clinical studies have shown that exendin-4 has several beneficial anti-diabetic actions that include glucose-dependent enhancement of insulin secretion, glucose-dependent suppression of inappropriately high glucagon secretion, slowing of gastric emptying, reduction of food intake and body weight, and an increase in β-cell mass (Buse et al. 2004; Edwards et al. 2001; Kendall et al. 2005; Kolterman et al. 2003; Xu and Kaneto 2006). For many years exendin-4 has been synthesized chemically at a high price and is thus unsuitable for mass production. It therefore needs to be produced in quantity by genetic recombinant technology. Exendin-4 has been successfully expressed in E. coli system and it proved having glucose-lowering action in vivo (Yi et al. 2006; Yin et al. 2005). However, no successful expression in other expression systems has been reported until now. For large-scale production, our laboratory recently constructed a recombinant Pichia pastoris that can produce exendin-4 constitutively and secrete it into the broth (Zhuang et al. 2007). In this study, the purification and bioactivity of recombinant exendin-4 expressed in P. pastoris are reported. Materials and methods Microorganism and expression of exendin-4 fusion protein Pichia pastoris GS115 was transformed with plasmid pPIC9 containing the gene encoding exendin-4, which was obtained from GenBank database (Accession No. AAB22006). The details of the vector construction and transformation were described elsewhere (Zhuang et al. 2007). Recipes and fermentation procedures were all followed by the supplier’s protocol and were described briefly as follows. Fermentation of the pPIC9/exendin-4 transformant of was carried out using a 30 l fermentor. After 48 h of growth at 28°C, the transformant P. pastoris was induced by methanol for 72 h. The culture was centrifuged at 5,000g for 15 min to collect the supernatant which was used as the source of recombinant protein. Purification of fusion protein For the purification of exendin-4, the total protein in the supernatant was first concentrated by ultrafiltration using a 10 kDa molecular mass cutoff membrane. The fusion protein was subsequently purified by ion-exchange chromatography, hydrophobic interaction chromatography (HIC) and size-exclusion chromatography. The void volume and eluted fractions from the three chromatographic columns were monitored at 280 nm. Fractions were collected and subjected to SDS-PAGE analysis. The purification steps were shown in Table 1. Table 1Summary of exendin-4 purificationPurification stepTotal protein (mg)Protein of interesta (mg)Recovery (%)Fusion exendin-4Broth supernatant1,640580100DEAE A-251,20039067Phenyl FF76028549G-7541021437Recombinant exendin-4G-25214264.5The concentrated broth supernatant was dialyzed with buffer A (20 mM Tris/HCl, pH 6.0, 1 mM EDTA) and was applied to a DEAE Sephadex A-25 (DEAE A-25) (100 ml bed volume) equilibrated with buffer A beforehand, at 1 ml/min. The column was washed with 300 ml buffer A and then eluted with a linear gradient of 0 to 1 M NaCl in buffer A. The eluted fractions containing exendin-4 were pooled and dialyzed. The desalted protein was mixed with ammonium sulfate at 1 M and applied onto a phenyl-Sepharose Fast Flow (Phenyl FF) column (50 ml bed volume) pre-equilibrated with buffer B [20 mM Tris/HCl, pH 6.0, 1 M (NH4)2SO4, 1 mM EGTA] at 1 ml/min, the column was washed with 150 ml buffer B and eluted with a linear (NH4)2SO4 gradient from 1 to 0 M in buffer B. The peak with exendin-4 was concentrated and dialyzed with buffer C (50 mM PBS, pH 6.0, 10% (v/v) glycerol). The last step of purification was through a Sephadex G-75 column (1 × 50 cm) pre-equilibrated with buffer C at 0.5 ml/ min. For recombinant protein purification, the pool of fractions containing purified fusion protein from the three columns was dialyzed in 20 mM Tris/HCl (pH 6.0)aThe amount of protein of interest was determined by quantifying the amount in each gel lane by densitometry (Totallab V1.11 software) Cleavage of the fusion protein by enterokinase After desalting, the fusion protein was incubated with enterokinase (1 U/1 mg fusion protein) at 37°C for 8 h to obtain the recombinant exendin-4. The reaction mixture was analyzed by Tricine SDS-PAGE. The recombinant exendin-4 was purified by Sepharose G-25 column and its purity was assayed by HPLC. SDS-PAGE, Tricine SDS-PAGE, immunoblotting and protein determination Proteins were analyzed by SDS-PAGE under reducing conditions using 12% (v/v) gels. Tricine SDS-PAGE was carried out with the supplier’s protocol (Amersham Bioscience). The separated proteins were stained with Coomassie Brilliant Blue R-250 or electroblotted to nitrocellulose membrane for western blotting. After blocking with 3% (v/v) non-fat milk in 0.05% Tween/PBS (PBST), the membrane was washed 3 times with PBST for 10 min and incubated further with a 1:2,000 dilution of polyclonal rat anti exendin-4 antibody for 1 h, followed by washing as was described above. The membrane was incubated with a 1:1,000 dilution of HRP-conjugated goat anti-rat IgG as the second antibody, washed with the same procedure as above and revealed with DAB substrate (Sigma). Protein concentration was determined by the method of Bradford using bovine serum albumin as a standard. Biological activity assay The effects of recombinant protein exendin-4 on plasma glucose concentration and insulin levels were examined using 8 week-old male Wistar rats. Animals were housed six per cage (three repetition) at 24°C environmental conditions with free access to food and water. They were allowed one week to adapt to their environment before the experiment. Food was withdrawn for 24 h before intraperitoneal injection of glucose (20 mM/kg body weight) alone or in combination with exendin-4 (10 nmol/kg body wt). Test solutions were administered in a final volume of 1 ml/kg body weight (O’Harte et al. 2000). Blood samples were collected at 15, 30 and 60 min after the injection from the eye socket with capillary into chilled heparin Eppendorf tubes. Blood samples were centrifuged and plasma samples were stored at −20°C before determination. Plasma glucose and insulin level were assayed respectively by a glucose assay kit and insulin enzyme linked immunosorbent assay kit (Dingguo, Beijing). Results were expressed as mean ± SE and were analyzed by one-way analysis of variance (ANOVA) by SPSS 11.0 software. Differences were considered significant at P < 0.05. Results Expression and detection The results of expression and detection exendin-4 fusion protein are given in Fig. 1. Fig. 1Expression and detection of exendin-4 fusion protein. (a) SDS-PAGE analysis of the expression level of fusion exendin-4 at different time in hours. After 56 h, the amount of the fusion protein expression reached the maximum (580 mg/l). Fifteen micro liter of supernatant were applied to a 12% gel and then stained with R-250. Lane 1, control transformant GS155/pPIC9 after 72 h culture; Lanes 2–10, recombinant GS115/pPIC9/exendin-4 after 8, 16, 24, 32, 40, 48, 56, 64 and 72 h of culture; The position of the target proteins induced by methanol is indicated by arrow. (b) Western blotting analysis of expressed protein in P. pastoris. Lanes 1–8 expressed protein of recombinant GS115/pPIC9/exendin-4 after 0, 8, 16, 24, 32, 40, 48 and 56 h, lane 9 expressed protein of control transformant GS115/pPIC9 Purifying the fusion protein and obtaining the recombinant exendin-4 by enterokinase cleave Purification of the fusion protein and production of the recombinant exendin-4 by enterokinase cleavage are given in Table 1 and further details were shown in Fig. 2. Fig. 2(a) Purification of the exendin-4 fusion protein by three chromatography steps. Lane 1, broth supernatant; lane 2, DEAE A-25; lane 3, Phenyl FF; lane 4, G-75; M is protein marker (KDa). (b) Purification of the recombinant protein exendin-4 by a G-25 column. Lane 1, reaction mixture of fusion protein cleaved by enterokinlase. Recombinant exendin-4 (4.3 kDa) is marked by arrow; lane 2, the non-target protein eluted from the G-25 column; lane 3, purified recombinant peptide exendin-4 from G-25 column. (c). The purity analyzed by HPLC was 95.6% which was performed on a 250 mm × 4.6 mm C5 column. The column was eluted with the linear gradient of acetonitrile (90–0%) in 0.1% trifluoroacetic acid for 35 min at 0.5 ml/min. The target peak was at 18.3 min Bioactivity of exendin-4 in vivo Figure 3 showed the results of bioactivity of exendin-4 in vivo. Compared with the control group, plasma glucose concentration in GLP-1 group was significantly reduced (P < 0.05) at 30 min, and the same phenomenon was observed after 60 min in exendin-4 group (Fig. 3a). There was no difference between the glucose-lowering capabilities of treatment group and control group before 15 min. The capabilities of stimulating insulin secretion showed significant response to GLP-1 and exendin-4 supplementation. The maximal increase in insulin in rat treated with GLP-1 was 2 times higher than the control group before 30 min. The plasma insulin concentration of exendin-4 treated group was also significantly raised (P < 0.05) at 30 min compared with the control group and the capabilities of stimulating insulin secretion by exendin-4 could keep to 60 min (Fig. 3b). Fig. 3Biological activity assay of exendin-4 in rats. The capabilities of glucose-lowering (a) and insulin-releasing (b) after intraperitoneal glucose alone (20 mM/kg) (control group), or glucose in combination with either exendin-4 (trial group) or GLP-1 (positive group) (10 nmol/kg). The time of injection is indicated by the arrow (0 min). Values are mean ± SE (n = 6). Asterisk denotes statistically significant differences (P < 0.05) between control and trial (or positive) groups by one-way ANOVA Discussion For many years, the production of exendin-4 was by chemical synthesis which necessitated a high cost. Recently, bioengineering methods of producing exendin-4 have appeared. Yi et al (2006) and Yin et al (2005) cloned the exendin-4 gene and obtained stable expression in E. coli. However, the E. coli-expressed exendin-4 existed initially in inclusion bodies and bioactivity was realized only after renaturation, which made the purification process complicated and led to a low yield. Furthermore, clinical application of bacterially produced products may be affected by the possible presence of endotoxins that sometimes contaminate protein preparations expressed by E. coli. We have several reasons to select P. pastoris as an alternative recombinant expression host for exendin-4 synthesis. P. pastoris has the advantages of large production, genetically stable expression strains, the potential to secrete recombinant proteins into culture medium, and simple inexpensive culture conditions (Patrick et al. 2005). In this study, the highest expression level of fusion protein exendin-4 was estimated at about 580 mg/l culture, which was higher than E. coli system. In addition, Pichia pastoris-synthesized recombinant exendin-4 was effectively purified from the culture medium through four chromatograph steps in this study. The whole purification process was simple and easy. These properties make P. pastoris a superior system to E. coli for preparation of exendin-4. To the best of our knowledge, this is the first study to directly investigate the effects of recombinant exendin-4 expressed in P. pastoris on glucose-lowering and insulin-releasing in vivo. The results suggested that the ability of native peptide GLP-1 to reduce plasma glucose to a certain level was at 30 min, while the activity of exendin-4 appeared at 60 min (Fig. 3a). It indicated that the exendin-4 has a longer duration of action and a greater potency than GLP-1 (Fig. 3a). Insulin is one of the most important hormonal immune responses and is linked to glucose-lowering responses. Although reports on influence of administration of exenatides on insulin activity are not very consistent, this function can be enhanced by injection administration of exendin-4 or its analogue (Gedulin et al. 2007; Kolterman et al. 2003; Nielsen et al. 2004; Xu and Kaneto 2006). The present study observed similar phenomenon. Furthermore, we observed that the capabilities of stimulating insulin secretion by GLP-1 was maintained up to 30 min, while exendin-4 was up to 60 min (Fig. 3b). Thus the half-life of exendin-4 in plasma is longer than GLP-1. There are two main reasons for this. One reason is that the exendin-4 lacks many of the neutral endopeptidase substrate sites present in GLP-1 and has a nine-AA sequence at the C-terminus absent from GLP-1 (Hupe et al. 1995). Another one is that the N-terminal sequence (His:Gly:Glu) of exendin-4 is not recognized by dipeptidylpeptidase (IV) (DPP (IV)), which rapidly cleaves the His:Ala:Glu sequence found on the N-terminus of GLP-1 (O’Harte et al. 2000). In our opinion, we should not rule out the presumption that the exendin-4 (4.3 kDa) has higher molecular weight than GLP-1 (3.1 KDa), which can reduce the glomerular filtration in vivo. It is important to note that in the three independent experiments, the recombinant exendin-4 did not have obvious effects on one or two rats of the six. It was probably because of individual differences. The results of the three independent experiments were comparable. According to the research, the recombinant exendin-4 may also have effects at smaller doses but remains to be proved. In summary, we have achieved high-level secretion of biologically active exendin-4 in P. pastoris. The result of the study indicated that although the heterogeneity-expressed exendin-4 in P. pastoris was slightly different from native GLP-1 in biological value, it still maintained good bioactivity. High purity and stability of the recombinant peptide of exendin-4 after purification makes it possible to be a therapeutical drug in the future. In the next step, we will focus on structure analysis and evaluation of the safety of exendin-4.

J_Abnorm_Child_Psychol-3-1-1915632

Stability in Bullying and Victimization and its Association with Social Adjustment in Childhood and Adolescence

This study examined the concurrent and longitudinal associations between stability in bullying and victimization, and social adjustment in childhood and adolescence. Participants were 189 girls and 328 boys who were studied in primary school and in secondary school. The mean age of the participants was 11.1 years in primary school and 14.1 years in secondary school. The measures consisted of peer reported social and personal characteristics. Children who bullied in childhood and adolescence were less liked and more disliked in childhood, and more aggressive and disruptive both in childhood and adolescence, than children who bullied only in childhood or adolescence. Children who bullied or who were victimized only in childhood did not differ largely in adolescence from the children that were never bullies or victims. Children who were victimized in adolescence closely resembled those who were victimized in childhood and adolescence in terms of being liked or disliked, being nominated as a friend, and shyness. The study stresses the need to distinguish between stable and transient bullies and victims.   Bullying in school classes refers to negative physical or social actions that are repeated over time by one or more other persons towards a person that can not easily defend (Olweus, 1991). Bulling involvement seems relatively stable over time (Boulton & Smith, 1994; Kumpulainen, Räsänen, & Henttonen, 1999) and has been related to various psychosocial adjustment problems in childhood and adolescence. Much research on bullying involvement, especially on victims, has focused on internalizing indicators of adjustment (see for a review Hawker & Boulton, 2000). Little is known about the overt, interpersonal behavioral characteristics of bullies and victims, and even less is known about the association between stability in bullying and victimization and these social characteristics. The aim of the present study was to examine the link between stability in bullying and victimization, and individual differences in social behaviors that are salient to the peer environment. More specifically, we were interested in differences in childhood and adolescent social adjustment of transient and stable bullies and victims. Bullies, victims, and adjustment In primary schools between 20% and 30% of the children are victims of bullying, while between 10% and 20% of the children are bullies (Smith et al., 1999), both in Western and non-Western countries (Eslea et al., 2003). Cross-sectional research on the adjustment of bullies revealed that bullies are more rejected and less popular (Boulton & Smith, 1994), and display more antisocial, aggressive and disruptive behavior than non-involved children (Pellegrini, Bartini, & Brooks, 1999; Rigby & Cox, 1996). Furthermore, longitudinal studies suggest that childhood bullying is associated with social maladjustment in adolescence (Kumpulainen & Räsänen, 2000). Research on the adjustment of victims showed that these children are socially isolated and rejected, and have fewer friends (Hodges, Malone, & Perry, 1997; Hodges & Perry, 1997). In addition, victims tend to be more submissive in their interactions with peers (Schwartz, Dodge, & Coie, 1993) and show overt signs of helplessness and distress (Perry, Williard, & Perry, 1990). Longitudinal research on bullies, victims and adjustment usually linked being a bully or a victim in childhood to adjustment measures in adolescence, without accounting for the chronicity in bullying or victimization. For example, Kumpulainen and Räsänen (2000), reported that children who bullied at age 8 and 12 displayed more externalizing behavior, hyperactivity, and relationship difficulties when they were 15 years old, while children who were victimized at age 12 were more likely to experience relationship problems such as being irritable, isolated, and rejected three years later. Camodeca, Goossens, Meerum Terwogt, and Schuengel (2002) found that stable victims were more reactively aggressive than transient victims. Neither of these studies, however, examined whether the participants were still bullies or victims at the follow-up assessment. Consequently, the longitudinal associations that have been reported may have been confounded by concurrent associations between adolescent bullying involvement and adolescent adjustment. Adequate descriptions of longitudinal associations between childhood bullying involvement and adolescent social adjustment may require distinguishing between children who remain involved in bullying from childhood to adolescence (i.e. stable bullies and victims) and those whose involvement is restricted to childhood (i.e., transient or unstable bullies and victims). Various studies showed that a substantial number of children display a stable pattern of bullying or victimization (e.g., Boulton & Smith, 1994; Camodeca et al., 2002; Salmivalli, Lappalainen, & Lagerspetz, 1998). Boulton and Smith (1994) found correlations of around.60 between bullying at age 8 and one year later, while Kumpulainen, Räsänen, and Henttonen (1999) reported that 25% of 8-year old bullies had turned into stable bullies by the time they were 12, and 15% of the victims had become stable victims. However, very few studies have examined the associations between stability in bullying involvement and interpersonal behavioral characteristics. Stability in being a bully or a victim Stability in being a bully or a victim in school can be caused by two mechanisms: continuity in social environment and continuity in children’s interactional styles. According to Caspi, Elder, and Bem’s cumulative continuity model (1987; Caspi, Bem, & Elder, 1989) and to Scarr’s genotype environment correlations model (Scarr & McCartney, 1983; Scarr, 1985; Scarr, 1992), children’s social maladaptive behaviors may direct them into social environments that perpetuate these behaviors. The idea is that children with specific behaviors select and create environments that promote and maintain their behavior. This may hold for bullies as well as victims. It is well known that antisocial boys affiliate with boys who are similarly aggressive and deviant (Cairns, Cairns, Neckerman, Gest, & Gariepy, 1988), and that victims are more likely to have friends who are less accepted by peers or who themselves are victimized (Hodges et al., 1997; Salmivalli, Huttunen, & Lagerspetz, 1997), thus reinforcing their behavior. Additionally, peers may also reinforce bullying and victimization by conferring reputations that may lock bullies and victims into their specific roles (DeRosier, Cillessen, Coie, & Dodge, 1994). These reputations are difficult to change as long as the school class constellation remains unchanged (Hymel, Wagner, & Butler, 1990). The social environment may reinforce bullies’ behaviors yet in another way. Observational studies (Craig & Pepler, 1997; Pepler & Craig, 1995; Pepler, Craig, & Roberts, 1998) showed that peers often do not intervene in bullying. Bullies may perceive this lack of intervention as a signal that peers condone their bullying behavior. Furthermore, studies on participant roles reveal that some peers may even actively reinforce the bullies by encouraging gestures or by laughing during bullying episodes (Salmivalli, Lagerspetz, Björkqvist, Osterman, & Kaukiainen, 1996). Stability in being a bully or a victim may also be due to continuity in bullies’ or victims’ interactional styles (i.e., interactional continuity, Caspi et al., 1987, 1989) that place them at risk for prolonged bullying or victimization, even after changes in the social environment. Being stably involved in bullying in the formative years of primary school may deprive children from positive social experiences, inhibiting the acquisition of prosocial skills and fostering social skills deficits. Because these children do not learn how to adequately react in social interactions, they may develop dysfunctional interactional styles which may make them prone to social adjustment problems later in life. In this way, a vicious cycle between bully’s or victims’ interactional styles and their bullying or victimization may be established (cf. Kochenderfer-Ladd & Wardrop, 2001). For example, Snyder et al. (2003) showed that for some children victimization was situational, while for other children victimization became a trait like status, possibly because of this dysfunctional interactional style. Stability in being a bully or a victim, and adjustment Although the processes contributing to stability in bullying or victimization are well described, very few studies have examined the behavioral profiles of different bully or victim trajectory types over time. Research has indicated that most bullies desist after some time, but a small group persists (Broidy et al., 2003; Loeber & Hay, 1997; Le Blanc & Loeber, 1993). Because of their aggressive interactional styles and their lack of opportunities to develop social skills, these children are at risk for higher levels and more serious forms of aggression, poor peer relations and social maladjustment later in life (Pope & Bierman, 1999). This behavioral style reflects Olweus’ (1991) notion that bullying is not an isolated phenomenon but rather a component of more general antisocial and rule-breaking behavioral patterns. Regarding victims, only two studies seem to have longitudinally examined the adjustment of stable and transient victim groups. Juvonen, Nishina, and Graham (2000) found that stable victims were lonelier and reported a lower self-worth than transient victims (ie. those who were victimized one year earlier but were no longer victims). Noteworthy, transient victims did not differ from stable non-victims on loneliness, self-worth or depressive symptoms. Covering a two-year period, Smith, Talamelli, Cowie, Naylor, and Chauhan (2004) found that stable victims had fewer friends at school and scored higher on self-reported and teacher-reported emotional problems and peer problems. As was found by Juvonen et al. (2000), in general, the transient victims did not differ largely from the stable non-victims. The latter seems to indicate that psychosocial problems only occur at the time when the victimization takes place. It also supports the cessation hypothesis (Kochenderfer-Ladd & Wardrop, 2001) which states that the psychosocial problems are likely to disappear as soon as the victimization ends. The present study The present study investigated whether stable bullies or victims showed specific behavioral patterns that distinguished them from children who were involved in bullying only in childhood or adolescence. Whereas previous studies on bullying and victimization often focused on psychological adjustment in terms internal characteristics such as self-esteem, loneliness, or emotional stability, our aim was to examine the interpersonal, social adjustment as it was reported by the peers. The bully groups and victim groups were compared on social characteristics that have been identified in previous research as highly relevant correlates for describing bullies or victims in childhood or adolescence. These characteristics included being liked and being disliked (Hodges & Perry, 1999; Pellegrini et al., 1999), friendships (Hodges et al., 1997; Hodges & Perry, 1999), aggressive and disturbing behaviors (Boulton & Smith, 1994; Snyder et al., 2003), help seeking, shyness (Boulton & Smith, 1994) and offering help and cooperation (Boulton & Smith, 1994). We used peer reports to assess these constructs for children may have distorted views of their own social competence (Lemerise & Arsenio, 2000) and self-reports may thus not always be the most reliable means of obtaining information about children’s social functioning in peer groups. The longitudinal group design allowed us to test a number of hypotheses. The hypotheses concerning bullies were in part guided by the work of Olweus (1991) and Loeber (Loeber & Hay, 1997; LeBlanc & Loeber, 1993) and the cumulative continuity hypothesis (Caspi et al., 1987) and genotype-environment correlations model (Scarr, 1983, 1992; Scarr & McCartney, 1983). We hypothesized that due to the accumulative effects of their negative peer interactions stable bullies would exhibit a pattern of more severe, negative and hostile behaviors in childhood and adolescence, and would have poorer adjustment, compared to bullies who bullied for a restricted period of time (i.e., only in childhood or adolescence). Because we assumed that adolescents who first started to bully in adolescence may have done so in order to obtain dominance (Pellegrini et al., 1999) and not because of a specific stable behavioral pattern, we expected that they would show specific maladaptive features in adolescence but not in childhood. Concerning the victims, we hypothesized that stable victims would show higher levels of peer perceived social problems in adolescence compared to childhood or adolescence-only victims. This was based on the idea that stable victims were exposed to victimization for a substantially longer period and thus have had fewer opportunities to acquire positive social skills and experiences. In line with the cessation hypothesis (Kochenderfer-Ladd & Wardrop, 2001) and the findings provided by Juvonen et al. (2000) and Smith et al. (2004) we expected that childhood-only victims would not differ from the stable non-involved children in adolescence. Although we did not have theoretically based hypotheses, we tested whether it would be possible to already predict in childhood which children would become new victims in adolescence (i.e., adolescence-only victims). Finally, we tested whether gender moderated the associations between stability and adjustment. Because we did not have theoretically derived hypotheses about the interaction the interaction effects of gender, testing these effects was entirely exploratory, and no research question specifically addressed this issue. Method Participants Participants were 517 children (189 girls, 328 boys) who took part in the third and fourth wave of a longitudinal study that started in 1985. In the first two waves (1985 and 1986), 231 boys were examined (see for detailed information Cillessen, van Yzendoorn, van Lieshout, & Hartup, 1992). In waves 3 (1991) and 4 (1994), the initial boys were again assessed, but now the assessment also included the boys’ classmates, resulting in total samples of 2521 and 3361 children, for Wave 3 and 4, respectively. The sample of the present study consisted of 517 children who were present both in Wave 3 (childhood data) as well as in Wave 4 (early adolescence data). Because the 231 initial longitudinal participants were boys, some of whom attended schools with a predominantly male population in secondary education (e.g., technical education), there were more boys than girls in the present study. Attrition bias checks using t-tests comparing the present sample with the larger childhood and adolescence samples that were dropped from this study (i.e., the Wave 3 and Wave 4 samples minus the participants who were in the longitudinal sample) revealed that the present study’s participants did not differ on bullying or victimization, nor on any other study variable, from the larger cross-sectional childhood or adolescence sample. They did also not differ on ethnicity or education, nor on socioeconomic status of the parents. In 1991 (Wave 3), participants’ age was 11 years, 1 month (SD = 11 months) and ranged from 10 to 13. The children came from 100 elementary school classes in the Nijmegen-Arnhem area in The Netherlands. In 1994 (Wave 4), all participants were enrolled in secondary education, and were distributed across 131 school classes. Their mean age was 14 years and one month (SD = 11 months), and ranged from 13 to 16. Ninety-five percent of the participants were of Dutch origin, while five percent of the adolescents were ethnic minorities: 1.5% came from Surinam, the Netherlands Antilles, and the Molucca Islands; 2% from Mediterranean countries; and 1.5% from elsewhere. Socio-economic status was based on the classification of the parents’ occupations. The lower socio-economic status contained occupations such as factory workers, while middle socio-economic status consisted of occupations like for example teachers and nurses. The higher socio-economic status included occupations such as lawyers, and physicians. It showed that 22% of all children had a lower, 43% a middle, and 35% a higher socio-economic status. Overall, at the two waves the participants appeared to be representative of the Dutch school population in the geographic area in which the children lived and at the time the assessments were made. Procedure Both the childhood data and adolescence data were obtained by classroom data collections, arranged separately for each school class (for detailed descriptions, see Haselager, Hartup, van Lieshout, & Riksen-Walraven, 1998, and Scholte, van Lieshout, & van Aken, 2001, for childhood and adolescent data assessment, respectively). In order to ensue that school staff and parents obtained as much information as they needed before the study was conducted, letters were sent to schools and children took home these letters for their parents to read. In these letters parents were informed about the study and were asked for passive consent. As in other large scale studies that we conducted in the Netherlands, there were no parents in the present study who did not allow their children to participate. Consent was also obtained from the children and adolescents themselves and from school authorities. Participants were guaranteed confidentiality in the collection and maintenance of the data. In both assessment waves the data collection session started with a brief introduction and class instruction, given by a trained examiner. During the one-hour assessment, the children were asked to fill out a questionnaire booklet. Before answering the bullying involvement questions, the children were provided a definition of “bullying other children” (Olweus, 1989) by the examiner. Measures Childhood measures (peer nominations) We developed a sociometric questionnaire containing 11 items on peer reported social behavior. This measure was similar to the widely used instrument developed by Coie, Dodge, and Coppotelli (1982). Children were asked to nominate a maximum of three classmates who best fitted the descriptions, in a rank ordered way such that the classmate who best fitted a description was nominated first, followed by the second and third best fitting classmate. Same- as well as cross-sex nominations were allowed on each description but self-nominations were not. The children were presented a roster of their classmates to use as a reference in making their nominations on the sociometric questions. Bullying other children was assessed by the question “Which children in class often bully other children, or pick on them.?” Being bullied (victimization) was assessed by the question “Which classmates are often bullied and picked on by other children.?” Being liked and being disliked were based on the questions “Which children in your class do you like most?” and “Which children in your class do you like least,?” respectively. Aggression referred to starting fights in class, while ‘Disruption’ assessed disturbing and disruptive behavior. Cooperation assessed being considerate and cooperative, while ‘Offering help’ indicated offering help to others. Seeking help assessed which children sought help often, while Shyness referred to being shy. Being nominated as a friend, finally, was based on the question “Which three children in you class are your friends.” This variable was used as an indicator of a child’s social integration in the class. Being nominated as a friend was not synonym with being liked because it was possible that children liked classmates without being friends with them.Table 1Intercorrelations among social adjustment measures in childhood123456781. Being liked2. Being disliked−.38***3. Friend nominations.71***−.35***4. Aggression−.20***.65***−.21***5. Disruption−.14**.49***−.15***.71***6. Cooperation.53***−.36***.51***−.30***−.26***7. Shyness−.16**−.06−.16***−.13**−.16***−.048. Offering help.60***−.21***.58***−.06−.04.61***−.15**9. Help seeking−.05.25−.06.18***.18***−.08.09.01**p < .01. ***p < .001.Table 2Correlations between social adjustment measures in childhood and adolescenceAdolescencechildhood1234567Insecurity1. Being liked.25***−.16***.17***.01−.06.17***−.16***−.14**2. Being disliked−.17***.28***−.13**.24***.19***.16***−.05.073. Friend nominations.25***−.14***.18***−.04−.08.21***−.11*−.11*4. Aggression.00.18***.02.38***.28***−.14**−.15***.005. Disruption.01.17***.05.40***.30***−.14**−.21***−.016. Cooperation.18***−.13***.11*−.08−.07.21***−.02−.087. Shyness−.10*−.00−.09*−.16***−.12**−.02.30***.15**8. Offering help.16***−.09*.14**.06.01.15**−.13**−.13**9. Help seeking.10*.04.09*−.03−.01.03−.05−.09*p < .05.**p < .01.***p < .001. Adolescence measures (peer nominations) Adolescents were also presented a roster with the names of their classmates. In grades 1 through 3 of secondary education, adolescents were in the same group all year, and this group served as the nomination reference group. The reference group of the adolescents who were in grade 4 of secondary education consisted of their root class, which consists of those classmates with whom they spent most of the classes, and with whom they share the same mentor. The sociometric questionnaire administered in the adolescent sample contained 25 questions referring to liking and disliking, bullying and victimization, number of friends, and peer group behavior (Scholte et al., 1997). Nine of the 25 items were similar to those asked in childhood (see below), while 16 new items assessed peer-perceived self-confidence, emotional stability, and school achievement. In the present study we only used items in adolescence that were similar to the items in childhood, added with the item on “insecurity’ which in a way reflected the childhood question on seeking help. Participants could nominate up to five classmates on each of the questions, and were asked to nominate first the classmate who fitted an item best, followed by the classmate who fitted an item second best et cetera. In the present study, only the first three nominations on each item were used in the analyses, which was similar to the number of nominations on each item in the childhood data. Same and cross-sex nominations were permitted, self-nominations were not allowed. Assistance was provided if needed. Bullying others and Being bullied were assessed by the questions “Which classmates bully others,” and “Which classmates are being bullied,” respectively. Being liked and being disliked, being nominated as a friend, aggression, disruption, cooperation, and shyness were assessed by similar questions as used in the childhood assessment. Insecurity was assessed by the item ‘Which classmates are insecure and seem to lack self-confidence.’ The intercorrelations among these indicators of social adjustment are given in Table 1. Sociometric nominations were processed as follows. For each participant, scores on each item were determined by summing all received nominations from classmates on that item. These two raw scores were transformed into within-class probability scores (p-scores) assuming a generalized binomial distribution, thus correcting for unequal numbers of nominations made among children and differences in class size (Newcomb & Bukowski, 1983). The p-scores were then z-standardized across all participants. Each peer nomination scale in childhood and adolescence consisted of one item. Because peer nominations involve aggregating across multiple respondents, - in our study on average 23 per class- peer nominations based on one item tend to be reliable (Coie, Dodge, & Kupersmidt, 1990), because, in contrast to self-reports, they are based on the reports of many informants which may decrease the influence of individual bias (Boulton & Smith, 1994). In order to be able to describe the differences between the different groups of victims and bullies more clearly, the childhood and adolescence social adjustment variables were standardized within the total sample and these standardized scores were subsequently analyzed. Results The correlations between the childhood measures and adolescence measures are presented in Table 2. As can be seen, correlations between the childhood and adolescence data were low to moderate, with the highest correlations between childhood aggression and disruption and adolescence aggression and disruption (r’s between .28 and .40). Classification of bullies, victims and non-involved children To determine which children and adolescents were bullies or victims, we used the peer nomination scores on Bullying others and Being bullied. Children and adolescents who scored 1 standard deviation above the mean on Bullying others were considered bullies, and children and adolescents who scored one standard deviation above the mean on Being bullied were regarded as victims. This procedure has been used in other studies (e.g., Pellegrini et al., 1999; Solberg & Olweus, 2003). Of all children in childhood, 9 % were victims (n = 49; 17 girls of the total sample of girls, 32 boys of the total sample of boys), 19% were bullies (n = 100; 6 girls), and 71% were not involved in bullying (n = 368; 163 girls). In adolescence these percentages were 11% (n = 55; 14 girls) for victims, 20% (n = 104; 12 girls) for bullies, and 69% (n = 358; 160 girls) for non-involved children. Chi-square analyses indicated that while boys and girls were equally likely to be victims, boys were more likely to be bullies in childhood (χ2 (2) = 44.57, p < .001) and in adolescence (χ2 (2) = 37.34, p < .001). Because the focus of this study was on bullies or victims, children who scored one standard deviation above the mean on being bullied and bullying either in childhood or adolescence (i.e., bully-victims, n = 9) were not included in the analyses. To examine the longitudinal involvement in bullying, a chi-square analysis was conducted with bully status in childhood and adolescence as factors. The chi-square analysis for the total sample (χ2(4, N = 517) = 111.30, p < .001) and the subsequent test of standardized residuals (Haberman, 1973) showed that victimization and bullying were relatively stable across this three year period. Forty-six percent (n = 46) of the childhood bullies persisted into adolescence (i.e., Stable Bullies), whereas the others had either stopped being involved (i.e., Childhood Bullies, 45%, n = 45) or had turned into victims (9%, n = 9). Forty-three percent (n = 21) of the children who were victims in childhood were still victims in adolescence (i.e. Stable Victims), 51% of the childhood victims (n = 25) were not involved in bullying in adolescence (i.e. Childhood Only Victims), while 6% (n = 3) had turned into bullies. Of all the children not involved in bullying in childhood, 15% (n = 55) started bullying in adolescence (i.e., Adolescence Only Bully), and 7% (n = 25) became victims (i.e., Adolescence Only Victim). The children who were not involved in childhood nor adolescence were considered Non-involved, and served as a reference group. Gender differences in bullying and victimization continuity indicated that male and female childhood victims were equally likely to become a stable victim, but that the continuity of bullying other children was low in girls and high in boys (χ2 (6, N = 517) = 66,13, p < .001). Except for one girl, all stable bullies were boys. To examine the association between the stability in bullying and victimization, and social adjustment, the three groups of bullies as well as the three groups of victims were compared. For reasons of conceptual clarity, children who bullied in childhood and had turned into victim in adolescence and those who were victims in childhood but became bullies in adolescence were dropped from further analyses. Bullying group by gender interactions were examined, as well as gender main effects. The latter were significant in childhood (Wilks Λ = .93, F (9, 343) = 2.94, p < .01) and adolescence (Wilks Λ = .93), F (8, 342) = 2.99, p < .01) and indicated that both in childhood and adolescence, boys were more disliked, more aggressive and disruptive, less cooperative, and less shy than girls. In addition, boys were less often nominated as a friend in childhood than girls did, and were less insecure in adolescence. Because the Stable Bullies included only one girl, gender by bully group interactions could not be examined in analyses that included the Stable Bullies group.Table 3Childhood and adolescent adjustment of childhood only, adolescence only, and stable bullies, and non-involved childrenBulliesChildhood only (n=45)Adolescence only (n=55)Stable (n=46)Non-involved (n=288)F-valuePartial eta squaredChildhoodBeing liked.08b.20b−.37a.14b3.75*.03 Being disliked.44c−.28a1.39b−.36a86.31***.38 Friend nominations.06b.08b−.38a.15b3.79*.03 Aggression1.28b−.28a1.78c−.42a224.32***.61 Disruption1.18c−.12b1.69d−.36a137.25***.49 Offering help.03.11−.10.08.49.00 Cooperation−.30a.09b−.59a.20b10.80***.07 Shyness−.30ab−.27ab−.28a.06b4.40**.03 Help seeking.35b−.21a.02a−.10a3.51*.02Adolescence Being liked.35.11−.18.142.31.02 Being disliked−.37a.50b.87b−.33a43.42***.23 Friend nominations.28.11−.05.12.86.01 Aggression−.26a.84b1.77c−.34a118.42***.45 Disruption−.20a.54b1.26c−.26a46.81***.26 Cooperation.10b−.28a−.41a.19b7.27***.05 Shyness−.29a−.28a−.51a.02b6.85***.05 Insecurity−.16−.14−.21−.011.01.01Note. All scores are z-scores. Means with different superscripts are significantly different from each other.*p < .05. **p < .01. ***p < .001. Social adjustment of bullies in childhood To examine whether the Stable Bullies experienced more social problems in childhood than the other bullies and non-involved children, and whether Adolescence Only Bullies showed signs of maladjustment already in childhood, a MANOVA was conducted with bullying groups (Childhood Only Bullies, Adolescence Only Bullies, Stable Bullies, and Non-involved) and gender as independent variables and the childhood adjustment scores as dependent variables.1 A significant multivariate relation was found, Wilk’s Λ = .32, approx F (27, 1230) = 21.81, p < .001. Except for offering help, all univariate tests were significant (see Table 3). Childhood Only Bullies scored higher on help seeking behavior than the Stable Bullies. However, Stable Bullies were less liked and more disliked by their peers, were less often nominated as a friend, and displayed more aggression and disruption than all other children. Adolescence Only Bullies scored lower on being disliked, aggression, disruption, and help seeking, and higher on cooperation than Childhood Only Bullies.Table 4Childhood and adolescent adjustment of childhood only, adolescence only, and stable victims, and non-involved childrenVictimsChildhood only (n= 25)Adolescence only (n= 25)Stable (n=21)Non-involved (n=288)F-valuePartial eta squaredChildhood Being liked−.52a−.22ab−.88a.14b9.28***.07 Being disliked.63b−.28a1.34c−.36 a38.54***.25 Friend nominations−.60a−.23ab−.58a.15b7.44***.06 Aggression.15b−.32a.18b−.42a14.82***.11 Disruption−.13−.48−.19−.361.53.01 Offering help−.37a−.33ab−.44a.08b3.44*.03 Cooperation−.49a.02b−.52a.20b5.87***.05 Shyness.40.12.42.061.42.01 Help seeking.30b−.06a.87c−.10a8.14***.07Adolescence Being liked−.14b−.92a−.75a.14c10.61***.08 Being disliked−.21a.75b.81b−.33a23.88***.17 Friend nominations−.07b−.93a−.80a.12b9.80***.08 Aggression−.28−.25−.29−.34.09.00 Disruption−.26.05.04−.262.12.02 Cooperation−.11ab−.40a−.58a.19b6.50***.05 Shyness−.19a1.03b1.27b.02a17.10***.13 Insecurity−.02a.37ab.88b−.01a4.37**.04Note. All scores are z-scores. Means with different superscripts are significantly different from each other.*p < .01.**p < .01.***p < .001. Social adjustment of bullies in adolescence Next, we wanted to answer the question whether the Stable Bullies differed from the other bullies and non-involved children in adolescence. To address this question, a MANOVA was conducted in which the three bullying groups and gender served as the independent variables, and the adolescent measures as dependent variables. A significant multivariate effect emerged for bully group (Wilk’s Λ = .49), approx. F (24, 1218) = 14.14, p < .001. The results of the univariate and post hoc analyses (Student Newman-Keuls, p < .05) are presented in Table 3, and indicated no differences between the groups on being liked, number of friends, and insecurity. Childhood Only Bullies scored more positive on the other measures that did the other bullies, and in fact did not differ largely from the Non-involved children, except for their lower scores on shyness, indicating that they turned out to be normally functioning adolescents who did not display problematic social behavior. Stable Bullies turned out to be distinctive from the Adolescence Only Bullies in that they displayed higher levels of aggression and disruption. As the Stable Bullies consisted only of boys while the other groups also contained girls, the analyses were repeated comparing the Stable Bullies with only their male counterparts from the other groups. These analyses resulted in similar findings. Social adjustment of victims in childhood The next research goal was to examine whether the Stable Victims displayed lower social adjustment in childhood than the other victim groups and non-involved children and whether Adolescence Only Victims could already be identified in childhood. To address this goal, a MANOVA with the victim groups (Stable Victims, Childhood Only, and Adolescence Only Victims and non-involved children) and gender as independent, and childhood social adjustment as dependent variable were conducted. Results indicated a significant multivariate effect for group (Wilk’s Λ = .68), approx. F (27, 1002) = 5.33, p < .001. Follow-up univariate analysis of variance were significant for all variables except for disturbance and shyness (Table 4). Compared to the Childhood Only Victims, Stable Victims scored significantly higher on being disliked and help seeking behaviors. In contrast to the other victims, Adolescent Only Victims did not differ from the Non-involved children on any of the behavioral characteristics. There was no significant multivariate gender by group interaction. Social adjustment of victims in adolescence The last question that we addressed was whether Stable Bullies showed more signs of maladjustment in adolescence than the other victim groups and whether those who were victimized only in childhood would show social maladjustment problems in adolescence. A MANOVA was run with the victim and noninvolved groups, and gender as independent variables, and the adolescent measures as dependent variables. Findings revealed a significant multivariate effect for victim group (Wilk’s Λ = .74) approx. F(24, 992) = 4.58, p < .001. The results of the univariate tests and post hoc comparisons are given in Table 4. As can be seen, Childhood Only Victims scored more positively than the Stable and Adolescence Only Victims on being liked and being disliked, received friendship nominations, shyness, and insecurity. In fact, except on being liked, they did not differ from the Non-involved children, which indicated that being victimized only in childhood was not reflected in specific social behavior three years later. As Table 4 further shows, Stable Victims did not significantly differ from the Adolescence Only Victims on any of the peer nominated indicators of social adjustment. A multivariate victim group by gender interaction was found, (Wilk’s Λ = .88), approx. F (24, 992) = 1.86, p < .01. Univariate analyses showed a significant group by gender interaction for shyness (F (3, 349) = 5.07, p < .01 which revealed that in the Stable Victim group boys scored substantially higher on peer reported shyness than girls, whereas in the other three groups, boys scored lower. Discussion The findings of our study give insight into the peer perceived behavioral profiles of children who were bullies or victims for a restricted period of time and those who were involved over a period of three years. They show that stable bullies and victims displayed a behavioral pattern in childhood and adolescence that clearly distinguished them from the children whose bullying or victimization was restricted to childhood. The latter did not show social adjustment problems later in adolescence, while the former did. Children who were victimized only in adolescence showed similar signs of social maladjustment in adolescence as children who were consistently victimized. Our study shows that half of the childhood bullies turned into stable bullies, and half of the childhood victims into stable victims. In contrast to stability in victimization, continuity in bullying was very gender specific: Whereas only one out of 12 girls continued to bully in adolescence, this held true for almost half of the boys. An explanation for this gender specificity in stability may be that bullying behavior in our study referred more to overt aggression, which is more typical for boys, than to relational aggression (cf. Crick & Bigbee, 1998), which is more typical for girls. It is also possible that for girls, in contrast to boys, bullying may be more related to specific social situations rather than to an individual characteristic reflecting an underlying antisocial and aggressive personality pattern (Salmivally et al., 1998). With respect to the behavioral profile of bullies, this investigation revealed that, compared to Childhood Only Bullies, the Stable Bullies were less liked and more disliked, were less often nominated as a friend, were more aggressive and disruptive, and scored lower on help seeking in childhood. This finding supports our hypothesis which states that, because of their socially deviant behavioral profile, children who will continue to bully over time could already be distinguished in childhood from those who will desist after some time (Loeber & Hay, 1997). This negative behavioral pattern may result in accumulation of negative social consequences such as continued peer rejection and fewer opportunities to acquire adequate coping skills. Through this interactive continuity (Caspi et al., 1987) the deviant behavioral pattern is likely to be maintained into adolescence. This was reflected in finding that the Stable Bullies showed signs of peer perceived social maladjustment that distinguished them from Adolescence Only Bullies, which was in line with our hypothesis. Olweus’ (1991) notion that bullying reflects a stable aggressive and antisocial, rule-breaking personality pattern predisposing children to social maladjustment and delinquency in adolescence only matches the behavioral profile of the Stable Bullies. Given that Childhood Only Bullies did not differ much from the Non-involved children in adolescence in their peer perceived social adjustment, an important conclusion of our study is that half of the children who bully in childhood, may not constitute a group at risk for later social problems. The possible role of friends in relation to bullying deserves closer attention. In childhood, Stable Bullies were less often nominated as a friend than the Childhood Only Bullies and Non-involved children. Although we do not know whether these nominations were reciprocated, they might suggest that Stable Bullies have fewer reciprocal friends or at least fewer peers who consider them to be friends. The explanation may be that because of their behavioral profile, these bullies are less attractive as a friend, which is in line with studies showing a negative association between socially deviant behavior and positive peer relations (Rose & Asher, 1999). However, caution must be taken in interpreting these findings, because the differences between the childhood only and stable bullies in the number of received friendship nominations were no longer present after the male only classes were removed. This finding might indicate that in male only classes stable bullies receive fewer friendship nominations than childhood only bullies, while this may not be true for classes where boys and girls are equally present. Because it is not yet clear why this is the case, further study is warranted. In adolescence, although Stable Bullies (and Adolescence Only Bullies) were still more disliked than all other adolescents, they had similar number of peers who thought of them as friends, given that no differences existed between them and other adolescents in number of received friendship nominations. Recently, Cillessen and Mayeux (2004) showed that in adolescence antisocial behaviors including bullying are increasingly linked to social status, suggesting that bullying may become more accepted. Consequently, associating with and becoming friends with bullies may also become more accepted. Since friends may passively or actively encourage bullying (cf Salmivalli et al., 1996) bullies may feel reinforced and continue with their behavior (cumulative continuity, Caspi et al., 1987). The fact that Stable Bullies are as often nominated as a friend as other adolescents may reflect bistrategic orientations as described by Hawley (2003; Hawley, Little, & Pasupathi, 2002). That is, they display antisocial behavior but at the same time seem to be able to somehow convey to certain peers that they are their “friends” even if the bully him or herself does not necessarily consider that person a friend. This suggests that these bullies are more likely to be skilled manipulators rather than to be socially inadequate (Sutton et al., 1999). The environmental influences, combined with the Stable Bullies’ behavioral pattern, may make them relatively resistant to behavior change. Regarding the victims, Stable Victims showed a pattern of peer perceived social maladjustment in childhood that clearly distinguished them from the Childhood Only Victims. They were more disliked by their peers and were more likely to be perceived as seeking help from others. As the study by Boulton and Smith (1994) suggests, a pattern of consistently seeking help may signal that these children lack social self confidence. This, rather than the lack of prosocial behaviors such as cooperation or offering help may predispose children to remain victimized from childhood to adolescence (Boulton & Smith, 1994; Egan & Perry, 1998). This feature may prevent them from successfully interacting with peers in childhood and adolescence (interactive continuity, Caspi et al., 1987), depriving them of positive peer experiences. It may also affect the way they create their own environment (Caspi et al., 1987; Scarr, 1985, 1992; Scarr & McCartney, 1983) in that they may be more inclined to withdraw from social interactions and make them prone to be targeted by bullies in the group throughout their school life. We found that Stable Victims did not have a more problematic social behavioral profile in adolescence than Adolescence Only Victims did. This is in contrast to our expectation that the longer the victims experienced victimization, the more impaired their social behaviors would be (Kochenderfer-Ladd & Wardrop, 2001), and thus that Stable Victims would show the most problematic social adjustment. Because adolescents who are victimized only in adolescence show the same social adjustment problems as the adolescents who have been victimized for a long period of time, this finding might indicate that the duration of victimization is relatively unrelated to the severity or magnitude of peer reported social adjustment problems in adolescence. Our findings showed that in general, victimized or noninvolved boys were less shy than girls, however, those who were stably victimized scored higher on shyness in adolescence. This is consistent with other findings of negative social implications of shyness for boys (Kerr, 2000) and suggests that being shy in adolescence may place boys at particular risk for being victimized for a prolonged period. One of the positive findings of the present study seems to be that Childhood Only Victims may become normally adjusted children in adolescence, at least in terms of their peer reported social adjustment. This suggests that victimization experiences that are restricted to childhood do not necessarily translate into impaired social functioning observed by the peer context. This finding does not support the widely held assumption that being a victim of bullying in childhood is related to social adjustment problems in adolescence (e.g., Kumpulainen & Räsänen, 2000; see also Parker & Asher, 1987). It extends more recent studies because it shows that victimization is not only concurrently associated with psychological adjustment (Juvonen et al., 2000; Smith et al., 2004) but also with social adjustment. Like the findings in these other two studies, our findings may lend support for the cessation hypothesis (Kochenderfer-Ladd & Wardrop, 2001) and might indicate that social problems can disappear once the victimization is over. Nevertheless, it should be remembered that we assessed social adjustment as perceived by the peer group, and that findings might have been differed when more subjective experiences of feelings would have been used. Finally, the Adolescence Only Victims did not differ from the Non-involved children in their peer reported childhood social adjustment. This implies that children who will become victims later in their school career do not necessarily show specific social behavioral patterns that are salient to their peers and on the basis of which they could already be identified in childhood. An alternative explanation might be that becoming a victim in adolescence may actually have little to do with social behavior per se. Rather than because of their specific social behavior, adolescents may become victimized because of how they dress, how they look etc, which becomes more salient at this age. As the present study showed, these victims do have social adjustment problems in adolescence, but these problems may have resulted from being victimized rather than caused it. The present study has a number of positive features. It is among the first to longitudinally examine stability in bullying and victimization in relation to peer perceived social adjustment, during the transition from primary to secondary education. Bullying and victimization have been studied in a cultural context (i.e., the Netherlands) that has not been reported on extensively. This issue may bear some importance given that dimensions of peer relations can have differing connotations by culture (e.g., Schneider, 2000). For example, a recent study revealed that large variations existed between western counties such as England, Spain, Italy, and Ireland not only in the prevalence of bullies and victims but also in how bullying and victimization were related to social relations (Eslea et al., 2003). This indicates that findings from one western country may not be generalized to another. Our findings revealed that in general the behavioral profiles of bullies and victims in the Netherlands were similar to those reported in studies in other western cultures such as the USA (Hodges & Perry, 1999; Pellegrini & Long, 2002) and Great Britain (Boulton & Smith, 1994). Nevertheless, several caveats should be kept in mind. First, we only examined bullies and victims, but not bully-victims. This latter group turned out to be very small in our sample (ie., 2%, n = 9), comparable to other studies (Boulton & Smith, 1994; Olafsen & Viemerö, 2000; Solberg & Olweus, 2003). Additional exploratory analyses that we conducted on the bully-victims showed that the only feature that distinguished them from the other victims was their aggressive and disturbing behavior in class, on which they scored higher than all other victims, both in childhood and adolescence. In fact they were as aggressive as the Stable Bullies in childhood and adolescence. Nevertheless, due to the small sample size these results are only exploratory, and more research seems warranted to further describe the social correlates of bully-victims. Second, because the Stable Bullies contained only one girl, gender interactions could not be examined in analyses comparing this group of bullies with all other groups of bullies and the findings regarding Stable Bullies may thus not generalize to female bullies. Third, although this study was longitudinal in nature, causality in terms of victims’ social behavior triggering bullies bullying behavior or vice versa was not implied. Fourth, this study used peer reports to classify children and adolescents, and to assess the dependent variables. Using one source of information increases the risk of shared-method variance and inflated associations between independent and dependent variables. In addition, several behaviors assessed in our study were based on single items, which may raise some concerns about the validity of these behaviors. Finally, we have focused on the social adjustment of the bullies and victims as it is perceived and reported by their peer environment. Even though the use of peer reported social adjustment measures was informative, we could not identify which of the children who were not victimized in childhood became victims in adolescence. We did not examine children’s subjective experiences such as self-esteem, social insecurity, and loneliness. Exploring these individual subjective experiences may prove more valuable in predicting who is at risk to become victimized in adolescence and may add to our understanding of the consequences of being a bully or victim. Despite these caveats the present investigation shows that it is highly relevant for future scientific research on bullying, victimization, and adjustment to distinguish between children who are only involved in childhood or adolescence, and those who are chronically involved from childhood into adolescence. The positive message of our study is that many of the childhood victims and bullies did not seem to show social adjustment problems in adolescence, as perceived by their peers. The more troublesome message is that between 40 and 50% of the childhood bullies and victims will continue to be involved in bullying in adolescence. These children are the ones who are likely to display peer perceived social behavioral problems in adolescence. However, while this study is among the first to examine the behavioral profiles of stable and transient bullies and victims, this suggestion may be premature, and more research on the stability of bullying and victimization is needed.

Cell_Tissue_Res-4-1-2386754

Effect of tissue-harvesting site on yield of stem cells derived from adipose tissue: implications for cell-based therapies

The stromal vascular fraction (SVF) of adipose tissue contains an abundant population of multipotent adipose-tissue-derived stem cells (ASCs) that possess the capacity to differentiate into cells of the mesodermal lineage in vitro. For cell-based therapies, an advantageous approach would be to harvest these SVF cells and give them back to the patient within a single surgical procedure, thereby avoiding lengthy and costly in vitro culturing steps. However, this requires SVF-isolates to contain sufficient ASCs capable of differentiating into the desired cell lineage. We have investigated whether the yield and function of ASCs are affected by the anatomical sites most frequently used for harvesting adipose tissue: the abdomen and hip/thigh region. The frequency of ASCs in the SVF of adipose tissue from the abdomen and hip/thigh region was determined in limiting dilution and colony-forming unit (CFU) assays. The capacity of these ASCs to differentiate into the chondrogenic and osteogenic pathways was investigated by quantitative real-time polymerase chain reaction and (immuno)histochemistry. A significant difference (P = 0.0009) was seen in ASC frequency but not in the absolute number of nucleated cells between adipose tissue harvested from the abdomen (5.1 ± 1.1%, mean ± SEM) and hip/thigh region (1.2 ± 0.7%). However, within the CFUs derived from both tissues, the frequency of CFUs having osteogenic differentiation potential was the same. When cultured, homogeneous cell populations were obtained with similar growth kinetics and phenotype. No differences were detected in differentiation capacity between ASCs from both tissue-harvesting sites. We conclude that the yield of ASCs, but not the total amount of nucleated cells per volume or the ASC proliferation and differentiation capacities, are dependent on the tissue-harvesting site. The abdomen seems to be preferable to the hip/thigh region for harvesting adipose tissue, in particular when considering SVF cells for stem-cell-based therapies in one-step surgical procedures for skeletal tissue engineering. Introduction Tissue engineering is an emerging field in modern medicine. Therapies involve the combination of cells and scaffold materials that can be loaded with bioactive factors, ideally resulting in the regeneration or replacement of lost or damaged tissues and organs. Multiple cell sources have been investigated for their possible applicability in tissue engineering. Embryonal stem cells are the most potent stem cells; however, their use is controversial and has mayor ethical considerations (Dresser 2001). Mesenchymal stem cells (MSCs) can be obtained from the adult and are widely used because of their differentiation potential. In addition to bone marrow, periosteum (Nakahara et al. 1991), muscle (Asakura et al. 2001), and adipose tissue (Zuk et al. 2002) also appear to be sources of MSCs. Subcutaneous adipose tissue is a particularly attractive reservoir of progenitor cells, because it is easily accessible, abundant, and self-replenishing. It is derived from the mesodermal germ layer and contains a supportive stromal vascular fraction (SVF) that can be readily isolated (Gronthos et al. 2003; Zuk et al. 2001). This SVF from adipose tissue consists of a heterogeneous mixture of cells, including endothelial cells, smooth muscle cells, pericytes, leukocytes, mast cells, and pre-adipocytes (Oedayrajsingh-Varma et al. 2006; Peterson et al. 2005; Prunet-Marcassus et al. 2005). In addition to these cells, the SVF contains an abundant population of multipotent adipose-tissue-derived stem cells (ASCs) that possess the capacity to differentiate into cells of mesodermal origin in vitro, e.g., adipocytes, chondrocytes, osteoblasts, and (cardio)myocytes (Erickson et al. 2002; Guilak et al. 2004; Halvorsen et al. 2001; Hattori et al. 2004; Planat-Benard et al. 2004; Rangappa et al. 2003; Zuk et al. 2001). Because of these favorable characteristics, interest has been growing in the application of ASCs for cell-based therapies such as tissue engineering. For clinical practice, an advantageous approach would be to harvest ASCs and immediately give them back to the patient within the same operation, the so called “one-step surgical procedure” (Helder et al. 2007). This overcomes long-lasting culture expansion of ASCs on the one hand, but necessitates the use of the SVF of adipose tissue on the other hand, since fast selection procedures for stem cells in the SVF are not yet available. Therefore, the success of this procedure requires SVF-isolates to contain sufficient ASCs capable of differentiating into the desired cell lineage. In view of this, we have previously investigated the effect of three different surgical procedures for the harvesting of adipose tissue, i.e., resection, tumescent or conventional liposuction, and ultrasound-assisted liposuction, on the yield and function of the stem cells. We have demonstrated that the SVF isolates from adipose tissue harvested by ultrasound-assisted liposuction contain fewer stem cells, and that the stem cells have a longer population doubling time, leading us to conclude that resection and tumescent liposuction are preferable to ultrasound-assisted liposuction for harvesting adipose tissue, if the cells are to be used for tissue-engineering purposes (Oedayrajsingh-Varma et al. 2006). In the present study, we have investigated whether the yield and functional characteristics of ASCs in the SVF are affected by the most frequently used adipose-tissue-harvesting sites. We have previously demonstrated that the yield of nucleated cells in the SVF of the adipose tissue from these different tissue-harvesting sites is similar (Oedayrajsingh-Varma et al. 2006). In the current study, the frequency of ASCs in the SVF cell isolates has been determined by using limiting dilution and colony-forming unit (CFU) assays. In addition, we have investigated the frequency of CFUs showing an osteogenic differentiation capacity. SVF cells have been subsequently cultured in order to obtain homogeneous cell populations and to acquire sufficient cells to determine their chondrogenic differentiation potential in micromass cultures. Homogeneity has been checked by determining the growth kinetics and phenotypic characteristics of the ASCs. To verify the maintenance of multidifferentiation potential, osteogenic and chondrogenic induction has been assessed in these homogeneous ASC cultures. Materials and methods Donors Samples of human subcutaneous adipose tissue were obtained as waste material after elective tumescent liposuction or resection and donated after informed consent from healthy donors operated on at the Departments of Plastic Surgery of two clinics in Amsterdam, The Netherlands. Adipose tissue was taken from the abdomen (n = 12) and the hip/thigh region (n = 10) during cosmetic surgery; 22 female donors were included in this study. The average age (mean age 40, range 24–62 years) and body mass index (BMI; mean BMI 25.5; range 22.2–29.6 kg/m2) were similar for both groups (Table 1). Table 1Population characteristics (BMI body mass index, THS tissue-harvesting site, ASC adipose-derived stem cell, Res resection, T-LS tumescent liposuction, ND not determined)DonorAge (years)BMI (kg/m2)THSProcedureASC yield (%)102-00016222.6AbdomenRes9.20203-00072624.4AbdomenT-LS8.50303-00184223.9AbdomenRes12.50403-00215729.6AbdomenRes4.40504-000337NDAbdomenT-LS2.20604-00133628.1AbdomenRes2.20704-00153726.9AbdomenRes0.93805-000439NDAbdomenT-LS1.50905-00074228.3AbdomenT-LS8.301006-00035026.5AbdomenT-LS3.61106-00064224.2AbdomenT-LS4.81206-00074630.5AbdomenT-LS9.71303-00102424.4Hip/thighT-LS0.301404-000440NDHip/thighT-LS0.821504-00082726.6Hip/thighT-LS0.161604-000936NDHip/thighT-LS0.161705-00053422.2Hip/thighT-LS0.601805-00062823.2Hip/thighT-LS0.211905-00084324.6Hip/thighT-LS7.202006-00053323.2Hip/thighT-LS0.102106-00105226.2Hip/thighT-LS2.702206-00124223.7Hip/thighT-LS0.16 Cell isolation and storage Isolation of the SVF from adipose tissue was performed as previously described (Oedayrajsingh-Varma et al. 2006). The isolation protocol included a Ficoll density centrifugation step to remove contaminating erythrocytes. After isolation, 4×106 SVF cells were resuspended in a mixture (1:1) of Dulbecco’s modified Eagle’s medium (DMEM) and cryoprotective medium (Freezing Medium, BioWhittaker, Cambrex, Verviers, Belgium), frozen under “controlled rate” conditions in a Kryosave (HCI Cryogenics, Hedel, The Netherlands), and stored in the vapor phase of liquid nitrogen according to standard practice at the Department of Pathology of the VU University Medical Center and following the guidelines of current Good Manufacturing Practice. Limiting dilution assay To assess the frequency of ASCs in the SVF of adipose tissue, SVF cells were seeded in normal culture medium, consisting of DMEM supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine (Invitrogen, Gibco, Calif., USA) in 96-well plates at 15×103 cells/well in the upper row. Two-fold dilution steps of the cells were made in subsequent rows. All cultures were performed in duplicate. Medium was changed twice a week. After 3 weeks, each well was individually scored for the number of cells. A well containing a cluster of at least 10 adhered fibroblast-like cells was considered as being positive. The frequency of ASCs was calculated from the rows of cells for which 25%–75% of the wells were scored as positive. CFU assays CFU assays were performed to check the consistency of the limiting dilution assay method and to determine the frequency of CFU capable of differentiating into the osteogenic lineage from the abdomen (n = 7) and hip/thigh region (n = 6). SVF cells were resuspended in normal culture medium, consisting of DMEM supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine (Invitrogen, Gibco). Two 6-well plates were prepared in which the SVF was diluted ten-fold across both columns, resulting in a upper column containing 104 and a lower column containing 103 nucleated SVF cells. For the CFU-fibroblast (CFU-F) assay, the fixation time was 11–14 days, depending on the amount and growth kinetics of the colonies (merging of colonies was avoided). At the appropriate time point, the medium was removed, and the cells were washed with phosphate-buffered saline (PBS), fixed with 4% formaldehyde for 10 min, and subsequently colored in a 0.2% toluidine blue solution in borax buffer for about 1 min. Excess stain was washed off with distilled water, and colonies were counted. Cells of the duplicate 6-well plate were submitted to a CFU-alkaline phosphatase (CFU-ALP) assay. Cultures were performed in normal medium for 7 days in order to obtain colonies and to remove contaminating cells, after which osteogenic medium was added for 2 weeks. Following this period, cells in the CFU-ALP plate were rinsed with PBS, fixed in 4% formaldehyde, and incubated for 10 min in a 0.2 M TRIS-hydrochloride (pH 10), 0.2 M calcium chloride, 0.1 M magnesium chloride solution, whereafter a solution containing 0.2 M TRIS-hydrochloride (pH 10), 0.2 M calcium chloride, 0.1 M magnesium chloride, and 600 μl nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl-phosphate was added for 30 min. The percentage of the colonies staining positive for ALP was determined. Culturing of SVF cells ASCs from the abdomen and hip/thigh region were cultured up to passage 2 for an adequate and quantitative comparison of stem-cell proliferation and differentiation capacity. Single-cell suspensions of cryopreserved SVF cells were seeded at 5.0×106 nucleated cells/cm2 in normal culture medium. The cultures were maintained in a 5% CO2 incubator at 37°C in a humidified atmosphere. The medium was changed twice a week. When reaching 80%–90% confluency, cells were detached with 0.5 mM EDTA/0.05% trypsin (Invitrogen) for 5 min at 37°C and replated. Cell viability was assessed by using the trypan blue exclusion assay. A homogeneous population of ASCs was thus obtained from abdomen and hip/thigh region and was subsequently checked by determining growth kinetics and by analyzing the surface-marker expression profile of the ASCs. Growth kinetics of ASCs To determine the growth kinetics of cultured ASCs, ten T25 flasks per donor were seeded with 1×105 cultured ASCs (passage 2 or 3). At several time points (between days 2 and 12) after seeding, cells from two duplicate flasks were harvested and counted. ASC numbers were plotted against the number of days cultured, and the exponential growing phase of the cells was determined. The population doubling time was calculated by using the formula: where N1 was the number of cells at the beginning of the exponential growing phase, and N2 was the number of cells at the end of the exponential growing phase. Flow cytometry Single-cell suspensions of cultured ASCs from abdomen and hip/thigh region were phenotypically characterized by using fluorescence-activated cell sorting (FACS; FACSCalibur, Becton Dickinson, USA) as previously described (Varma et al. 2007). All monoclonal antibodies (mAbs) were of the immunoglobulin G1 (IgG1) isotype. Cells were stained with fluorescently labeled antibodies (conjugated to fluorescein isothiocyanate, phycoerythrin, or allophycocyanin) against CD31, CD34, CD45, CD54, CD90, CD106, HLA-DR, and HLA-ABC (BD Biosciences, San José, Calif.), CD166 (RDI Research Diagnostics, Flanders, N.J.), CD105 (Caltag Laboratories, Burlingame, Calif.), CD117 (PharMingen, San Diego, Calif.), and CD146 (Chemicon, Temecula, Calif.). Nonspecific fluorescence was determined by incubating the cells with conjugated mAb anti-human IgG1 (DakoCytomation, Glostrup, Denmark). Chondrogenic and osteogenic differentiation The chondrogenic and osteogenic differentiation capacities of the cultured ASCs from the abdomen (n = 4) and hip/thigh region (n = 4) were studied. Chondrogenic differentiation was induced in cultured ASCs as previously described, with some modifications (Oedayrajsingh-Varma et al. 2006). In short, a 50-μl drop of a concentrated ASC cell suspension (8×106cells/ml, passage 2) was applied to a glass slide and allowed to attach at 37°C for 1 h. Then, 750 μl chondrogenic medium, consisting of DMEM, plus ITS+ Premix (final concentration in medium when diluted 1:100 was 6.25 μg/ml insulin, 6.25 μg/ml transferrin, 6.25 ng/ml selenous acid, 1.25 mg/ml bovine serum albumin (BSA), 5.35 μg/ml linoleic acid; BD, USA), 10 ng/nl transforming growth factor- β1 (TGF-β1; Biovision, ITK-diagnostics), 1% FCS, 25 μM ascorbate-2-phosphate (Sigma, St. Louis, Mo.), 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine, was overlaid gently. Cells were maintained in a 5% CO2/1% oxygen custom-designed hypoxia workstation (T.C.P.S. Rotselaar, Belgium) at 37°C in a humidified atmosphere, as this was shown to enhance chondrogenic differentiation (data not shown). Chondrogenic media were changed every 2–3 days. For osteogenic differentiation, ASCs (passage 2) were seeded at 5000 cells/cm2 and cultured in monolayer in osteogenic medium, consisting of normal culture medium supplemented with 10 mM β-glycerol phosphate, 50 μg/ml ascorbate-2-phosphate, and 100 ng/ml bone morphogenetic protein 2 (BMP-2, Peprotech EC, London, UK). Osteogenic medium was changed twice a week. (Immuno)histochemistry (Immuno)histochemistry was performed as described previously (Oedayrajsingh-Varma et al. 2006). Cell nodules that formed under chondrogenic culture conditions were stained with Alcian blue (Sigma-Aldrich, Zwijndrecht, The Netherlands) at acidic pH for detection of proteoglycans. For the detection of collagen Type II, staining was performed with mouse monoclonal antibody II-II6B3 (1:50; Developmental Studies Hybridoma Bank, Iowa, USA) against human collagen Type II in PBS containing 1% BSA. Osteogenesis was visualized after 21 days of culture in osteogenic medium by Von Kossa staining to establish the formation of a calcified matrix, typical for mature osteoblasts. The protocol used was as described previously (Varma et al. 2007) with only one modification of counterstaining the cytoplasm of the cells using fast green. Calcified extracellular matrix was visualized as black spots. Spectrophotometric ALP activity Early differentiation of MSCs into immature osteoblasts is characterized by ALP enzyme activity, with human MSCs expressing ALP as early as 4 days after induction, and maximum levels being observed at around 14 days after induction (Jaiswal et al. 1997). Therefore, cellular ALP activity was measured after culturing the ASCs in osteogenic medium for 14 days. Cells were lysed with distilled water, and the ALP activity and protein content were determined. To determine ALP activity, p-nitrophenyl phosphate (Merck, Darmstadt, Germany) at pH 10.3 was used as the substrate, as described by Lowry (1955). ALP activity was expressed as micromole per microgram of protein in the cell layer. The amount of protein was determined by using a BCA Protein Assay reagent Kit (Pierce, Rockford, Ill., USA), and the absorbance was read at 540 nm with a microplate reader (Biorad Laboratories, Hercules, Calif., USA). Real-time polymerase chain reaction RNA isolation and reverse transcription were performed as previously described (Oedayrajsingh-Varma et al. 2006). Real-time polymerase chain reactions (PCR) were performed by using the SYBRGreen reaction kit according to the manufacturer’s instructions (Roche Diagnostics) in a LightCycler 480 (Roche Diagnostics). cDNA (approximately 5 ng) was used in a volume of 20 μl PCR mix (LightCycler DNA Master Fast Startplus Kit, Roche Diagnostics) containing a final concentration of 0.5 pmol primers. Relative housekeeping gene expression for 18 S-rRNA (18 S) and relative target gene expression for aggrecan (AGG), collagen Type II (COL2B), and collagen Type X (COL10α1) regarding chondrogenic differentiation, and for collagen Type I (COL1α), osteopontin (OPN), and runt-related transcription factor 2 (RUNX-2) regarding osteogenic differentiation were determined. Primers (Invitrogen) used for real-time PCR are listed in Table 2. They were designed by using Clone Manager Suite software program version 6 (Scientific & Educational Software, Cary, N.C., USA). The amplified PCR fragment extended over at least one exon border, based on homology in conserved domains between human, mouse, rat, dog, and cow, except for the 18 S gene (encoded by one exon only). Amplified Col2B PCR products were electrophoresed on a 2% agarose gel and stained with ethidium bromide. For real-time PCR, the values of relative target gene expression were normalized to relative 18 S housekeeping gene expression. Table 2 PCR primer sets used for reverse transcription/PCR (18 S 18 S subunit, AGG aggrecan, COL1αI collagen Type I, COL2B collagen Type II, COL10a1 collagen Type X, OPN osteopontin, RUNX-2 runt-related transcription factor 2) Gene Primer sets Accession number, product length (bp) 18 S Forward: 5′ GTAACCCGTTGAACCCCATT− 3′ Human, NM_10098, 151 bp Reverse: 5′ CCATCCAATCGGTAGTAGCG 3′ AGG Forward: 5′CAACTACCCGGCCATCC 3′ Human, NM_001135, 160 bp Reverse: 5′GATGGCTCTGTAATGGAACAC 3′ COL1αI Forward: 5′ AAGCCGAATTCCTGGTCT 3′ Human, NM_000088, 195 bp Reverse: 5′ TCCAACGAGATCGAGATCC 3′ COL2B Forward: 5′ AGGGCCAGGATGTCCGGCA 3′ Human, NM_033150, 195 bp Reverse: 5′ GGGTCCCAGGTTCTCCATCT 3′ COL10a1 Forward: 5′ CACTACCCAACACCAAGACA 3′ Human, NM_000493, 225 bp Reverse: 5′ CTGGTTTCCCTACAGCTGAT′ OPN Forward: 5′ TTCCAAGTAAGTCCAACGAAAG 3′ Human, AF_052124, 181 bp Reverse: 5′ GTGACCAGTTCATGAGATTCAT 3′ RUNX-2 Forward: 5′ ATGCTTCATTCGCCTCAC 3′ Human, NM_001024630, 156 bp Reverse: 5′ ACTGCTTGCAGCCTTAAAT 3′ Real-time PCR data analysis With the Light Cycler software (version 4), the crossing points were assessed and plotted versus the serial dilution of known concentrations of the standards derived from each gene by the Fit Points method. PCR efficiency was calculated by Light Cycler software, and the data were used only if the calculated PCR efficiency was between 1.85–2.0. Statistics Kolmogorov-Smirnov tests were used to determine the normalcy of measurements and, if appropriate, their logarithmics. For the evaluation of yield and growth kinetics, means between two groups in one variable were compared by using the independent sample two-tailed t-test. Partial correlation was expressed as the Pearson correlation coefficient, r. For evaluation of gene expression, a repeated measures analysis of variance was used to determine significant differences when increasing time-points in one donor within one variable were compared. If levels of gene expression were below the detection limit (0.05), values were set at 10−2 (or log level at −2). All statistical tests used a significance level of α = 0.05. Results Effects of tissue-harvesting site on frequency of ASCs In a previous study, we demonstrated that the yield of nucleated cells in the SVF of adipose tissue from different tissue-harvesting sites was similar (Oedayrajsingh-Varma et al. 2006). To investigate whether the tissue-harvesting site affected the frequency of ASCs in the SVF in the present study, limiting dilution and CFU-F assays were performed. The outcomes of both types of assays were similar (Fig. 1a, b). When combined, the SVF of adipose tissue harvested from the abdomen contained 5.1 ± 1.1% ASCs (mean ± SEM), whereas the percentage of ASCs in the SVF of adipose tissue harvested from the hip/thigh region was much lower (1.2 ± 0.7%; Fig. 1c). This difference in ASC frequency between adipose tissue from the abdomen and hip/thigh region was significant (P = 0.0009). Fig. 1Effect of adipose-tissue-harvesting site on the frequency of adipose-derived stem cells (ASCs) in the stromal vascular fraction (SVF). After isolation of the SVF from adipose tissue of both tissue-harvesting sites, the frequency of ASCs in the SVF isolates was determined by using: (a) a limiting dilution assay (LD) and (b) a colony-forming unit (CFU) assay (CFU-F), with similar results. When combined (c), a significant difference was detected in ASC frequency between adipose tissue harvested from the abdomen and adipose tissue harvested from the hip/thigh region (P-values: a limiting dilution: P = 0.003, b colony-forming unit: P = 0.05, c combined: P = 0.0009) Effect of tissue-harvesting site on frequency of CFUs having osteogenic differentiation potential Parallel to the CFU-F assay (Fig. 1b), CFU-ALP assays were also performed to determine the percentage of the CFUs capable of osteogenic differentiation. For adipose tissue from the abdomen, 54.9 ± 12.1% (mean ± SEM) of the CFUs stained positive for ALP, whereas the CFUs from adipose tissue of the hip/thigh region displayed an ALP positivity of 72.8 ± 8.1% (Fig. 2). Apparently, no differences in osteogenic potential existed between ASCs from the two tissue-harvesting sites (P = 0.43). Fig. 2Effect of adipose-tissue-harvesting site on the osteogenic diffentiation capacity of CFU from the abdomen or hip/thigh regions. No significant difference is apparent in the CFU-alkaline phosphatase (CFU-ALP) frequency from the abdomen and hip/thigh region when corrected for CFU-fibroblast (CFU-F). Phenotypic characterization and growth kinetics of cultured ASCs Cultured ASCs (passages 2 to 4) from abdomen and hip/thigh regions were phenotypically characterized. ASCs from both tissue-harvesting sites were demonstrated to be homogeneous populations staining positive for stem-cell-associated markers CD34, CD54, CD90, CD105, CD166, and HLA-ABC, and negative for hematopoietic/leukocytic/endothelial markers such as CD31, CD45, CD106, CD146, and HLA-DR (Table 3). Table 3Surface-marker expression of human cultured ASC at passages 3–4. Results are expressed as mean fluorescence (MF), with isotype control <4.25 (− MF<8.5; + 8.5<MF<100; ++ 100<MF<1000; +++ MF>1000).Cell-surface markerCultured ASC (n = 4)CD29++CD31–CD34+CD45–CD54++CD90+++CD105++CD106–CD146–CD166+HLA-ABC++HLA-DR– To determine growth kinetics, the population doubling time of ASCs from passages 2 to 3 was determined. When ASCs numbers were monitored over time, a cell growing curve was obtained showing an exponential growing phase, after which the cells reached confluency (Fig. 3a,b). The mean population doubling time of the ASCs in the exponential growing phase was about 2 days when the adipose tissue was harvested from the abdomen and hip/thigh regions (abdomen: 2.1 ± 0.8; hip/thigh: 2.3 ± 0.3; mean ± SEM; Fig. 3c). Fig. 3Effect of the adipose-tissue-harvesting site on growth kinetics of ASCs in vitro. a Growth kinetics of ASCs of a representative donor when adipose tissue was harvested from the abdomen. b Growth kinetics of ASCs when adipose tissue was harvested from the hip/thigh region. c Population doubling time was calculated from the exponential growing phase of the cells. There was no significant difference in population doubling time of ASCs from the abdomen and hip/thigh region (P = 0.78, independent Student t-test). Effect of tissue-harvesting site on osteogenic differentiation potential of ASCs Differentiation of cultured ASCs into the osteogenic lineage was induced by culturing the cells in monolayer in osteogenic medium containing BMP-2. Specific RNA expression of osteogenically induced ASCs from all donors tested increased over time, RUNX-2 being up-regulated 18-fold (P = 0.002) and COL1α being up-regulated seven-fold (P = 0.024) after 7 days (Fig. 4a,b). OPN gene expression was increased after 7 days; however, this increase was not significant (Fig. 4c). No significant differences were detected in osteogenic gene expression between ASCs derived from the abdomen and ASCs derived from the hip/thigh region, at all three time points tested. Fig. 4Effect of adipose-tissue-harvesting site on the osteogenic differentiation of cultured ASCs in vitro. a–cRUNX-2 (runt-related transcription factor 2; P = 0.002), COL1α1 (collagen type Ia; P = 0.024), and OPN (osteopontin; P = 0.38) gene expression was measured after 0, 4, and 7 days (d) of osteogenic induction, by using quantitative real-time polymerase chain reaction (qRT-PCR). No significant differences were detected in osteogenic gene expression between ASCs derived from the abdomen and ASCs derived from the hip/thigh region, at all three time points tested. d ALP activity was significantly increased after 14 days in the osteogenically stimulated cells (stim) compared with that in unstimulated cells (con; P = 0.047). No statistically significant difference was apparent in ALP activity between ASCs derived from abdominal fat and ASCs from hip/thigh fat. e–g Von Kossa staining of ASCs from abdomen (e) and hip/thigh region (f) after 21 days of culture in osteogenic medium and in control medium (g), showing mineralized matrix visible as black spots. ALP activity in the ASCs was measured after 14 days of osteogenic stimulation. ASCs cultured in control medium served as negative controls. ALP activity in the osteogenically stimulated cells was significantly increased (P = 0.047) compared with that in control cells (Fig. 4d). No statistically significant difference was apparent in ALP activity between ASCs derived from abdominal fat and ASCs from hip/thigh fat. Calcification of the osteogenic matrix was confirmed by using Von Kossa staining; black spots could be observed after 3 weeks of osteogenic stimulation of ASCs from both origins (Fig. 4e: abdomen, Fig. 4f: hip/thigh). No calcification was seen in ASC cultures expanded in control medium (Fig. 4g). Effect of tissue-harvesting site on chondrogenic differentiation potential of ASCs The chondrogenic differentiation potential of ASCs was analyzed after culturing the cells in a micromass in chondrogenic medium containing TGF-β. Within 24 h of culture, most of the cells formed nodules (n = 7). PCR-amplified COL2B mRNA expression was detectable but not quantifiable after 7 days in ASCs from both the abdomen and hip/thigh region in most but not all donors (n = 5). As shown in Fig. 5a, cells from both tissue-harvesting sites displayed COL2B mRNA. Under non-chondrogenic conditions, no COL2B could be detected. AGG and COL10α1 mRNA expression in all donors tested increased over time, AGG being up-regulated 2.4-fold (P = 0.041) at day 7 when compared with day 4, and COL10α1 being up-regulated four-fold (P = 0.024) after 7 days (Fig. 5b,c). No significant differences were detected in chondrogenic gene expression between ASCs derived from the abdomen and ASCs derived from the hip/thigh region, at all three time points tested. Fig. 5Effect of the tissue-harvesting site on the chondrogenic differentiation of cultured ASCs in vitro. a Both abdomen (lane 1) and hip/thigh region (lane 2) display COL2B mRNA. Under nonchondrogenic conditions, no COL2B could be detected (lane 3). b, c Aggrecan (AGG; P = 0.041) and collagen 10A (Col10a; P = 0.024) gene expression, respectively, was up-regulated after 7 days (d), as measured by qRT-PCR. No significant differences were detected in chondrogenic gene expression between ASCs derived from the abdomen and ASCs derived from the hip/thigh region, at all three time points tested. d, f Cartilaginous matrix expression was visualized in both tissue-harvesting sites by staining proteoglycans (Alcian blue) and COL2 (Col2-II6B3 antibody), respectively. e At higher magnification, the ASC nodules resembled cartilage-like tissue, composed of spherical cells surrounded by lacunae and lying in a proteoglycan-rich extracellular matrix. Alcian blue staining demonstrated proteoglycan deposition in the ASC nodules from both the hip/thigh region and abdomen (Fig. 5d). At higher magnification (Fig. 5e), the ASC nodules resembled cartilage-like tissue, composed of round cells, surrounded by lacunae and lying in a proteoglycan-rich extracellular matrix that appeared positive for collagen Type II by immunostaining (Fig. 5f). Discussion In this study, we have investigated whether the yield and functional characteristics of ASCs are affected by the adipose tissue-harvesting site, i.e., abdomen and hip/thigh regions. We have found a difference in the frequency of ASCs between adipose tissue harvested from the abdomen and the hip/thigh regions. SVF isolates derived from abdominal fat contain significantly higher frequencies of ASCs. When cultured, the growth kinetics and surface-marker expression of ASCs from both tissue-harvesting sites are similar. We have detected no differences in osteogenic or chondrogenic differentiation potential between these cultured ASCs from the two tissue-harvesting sites. Adipose tissue is a highly heterogeneous tissue, not only among individuals, but also when comparing different fat depots within one individual. Donor-dependent differences have been demonstrated to exist in (stem) cell yield, proliferation, and differentiation capacity, probably caused by differences in age (Hauner and Entenmann 1991; van Harmelen et al. 2003), BMI (Aust et al. 2004; Hauner et al. 1988; Jaiswal et al. 1997; van de Venter et al. 1994), and diseases such as osteoarthritis and diabetes (Barry 2003; Murphy et al. 2002; Ramsay et al. 1995). Donors used in this study for harvesting adipose tissue were healthy female donors up to 62 years old. Some patients who will benefit from tissue engineering of cartilage will be older or might suffer from disease, and males will also be affected. Therefore, future research should include these donor types to determine whether yields and functional characteristics are influenced by these variables. In our donor population, no significant correlation has been detected between the frequency of ASCs and the age of the donor (P = 0.32, r = 0.27) or between the frequency of ASCs and BMI (P = 0.42, r = −0.22; data not shown). These data are in agreement with other studies that have demonstrated no correlation between BMI or age and numbers of ASCs per gram of adipose tissue (Hauner and Entenmann 1991; van Harmelen et al. 2003). Most importantly, this means that these variables cannot be responsible for the difference that we have found between the frequency of ASCs and the tissue-harvesting site. In addition to donor-dependent heterogeneity, intra-individual differences between fat depots have been demonstrated, e.g., with regard to the metabolic response of adipose tissue to various hormonal and neurological stimuli (Guilak et al. 2004; Lacasa et al. 1997; Masuzaki et al. 1995; Monjo et al. 2003; Rodriguez-Cuenca et al. 2005) and the cellular composition of the adipose tissue (Peptan et al. 2006; Prunet-Marcassus et al. 2005). In this study, we have focused on the intra-individual differences in the yield and function of ASCs from two fat depots: the abdomen and the hip/thigh region. We have demonstrated that adipose tissue derived from the abdomen contains significant higher frequencies of stem cells compared with adipose from the hip/thigh region. Whereas SVF isolates from the abdomen in this study contain about 5.1% of ASCs, the frequency of ASCs in SVF from the hip/thigh region is only 1.2%. Moreover, this 1.2% is the average of a population in which some donors hardly possess any adipose-derived stem cells (see spreading ASC frequency hip/thigh in Fig. 1). Despite this more than four-fold difference, the frequency of ASCs in adipose tissue from the hip/thigh region is still much higher compared with the frequency of MSCs in the bone marrow compartment, which is as low as 0.001%–0.01% (Pittenger et al. 2000), thereby making the hip/thigh region still a much more attractive stem-cell source for tissue engineering therapies. What are the implications of these stem-cell frequencies for clinical practice? We have shown that 0.5–2.0×108 SVF cells can be harvested from 100 g adipose tissue, an amount that can easily be obtained from a patient (Aust et al. 2004; De Ugarte et al. 2003; Oedayrajsingh-Varma et al. 2006; Zuk et al. 2001). With an ASC frequency of 5.1%, the SVF isolates contain between 2.6–10.2×106 stem cells, which is an amount that appears to be sufficient for cell-based therapies as compared with the amount of cells used by others (Erickson et al. 2002; Fan et al. 2006; Williams et al. 2003; Zheng et al. 2006). This implies that time-consuming culturing and expanding steps of the stem cells can be avoided. In comparison, a bone marrow transplant of 100 ml contains approximately 6.0×108 nucleated cells (Zuk et al. 2002), of which only 0.001%–0.01% (0.06–0.006×106 cells) are stem cells (Pittenger et al. 2000). To determine the frequency of CFU capable of differentiating into the osteogenic lineage, a CFU-ALP assay has been performed on cells from the abdomen and hip/thigh region. The frequencies of CFU-ALP ( ± 3%) are almost comparable with those of the CFU-F (Figs. 1b, 2). This is higher than the value that Mitchell et al. (2006) have found in their clonogenic assay of freshly isolated stromal cells (±0.5%). However, they have used a different assay method and shorter incubation period, and their frequency increases to the same level (5%) after progressive passaging (Mitchell et al. 2006). SVF cells have been cultured up to passage 3 to obtain a homogeneous population of ASCs as starting material for differentiation studies toward the chondrogenic lineage. When comparing the growth kinetics of these cultured ASCs from the abdomen and hip/thigh region, the cell doubling time appears to be similar, being approximately 2 days. This is in accordance with the findings of others who have compared the replication rate of adipocyte precursor cells from various tissue-harvesting sites (Hauner et al. 1988; Pettersson et al. 1985; Roncari et al. 1981; Zuk et al. 2001). On the other hand, when comparing SVF cells from omental and subcutaneous fat, van Harmelen et al. (2004) have found a difference in cell proliferation rate; however, this may be explained by differences in methodological approach, since they use stromal cells instead of cultured ASCs. As we have shown that different fat depots contain different numbers of stem cells, these differences in proliferation rate may be caused by differences in initial stem-cell numbers when using SVF cells. This is reflected in the finding that, in our study, SVF cells derived from abdominal fat reach 80%–90% confluency within 5 days, whereas SVF cells derived from adipose tissue of the hip/thigh region take more than 9 days to reach 80%–90% confluency when seeded in the same density (data not shown). In addition to the determination of growth kinetics, we have phenotypically characterized the cultured ASCs. The surface-marker expression profile is in accordance with those found by others (Mitchell et al. 2006; Schaffler and Buchler 2007; Varma et al. 2007), making both tissue sites fully comparable with each other (Tables 1, 3). The homogeneous ASC population has been induced to the osteogenic and chondrogenic lineages. Having determined the osteogenic differentiation capacity of ASCs, we have shown significant up-regulation of osteogenic gene expression, ALP activity, and matrix mineralization. Interestingly, although no significant difference has been detected in ALP activity between ASCs from the abdomen and hip/thigh regions, ASCs from the hip/thigh region tend to show higher values of ALP activity after induction. This might be related to the underlying bone tissue, thereby implying that the ASCs of the hip/thigh region are less multipotent and more committed to the osteogenic lineage. The chondrogenic differentiation capacity of the ASCs has been demonstrated by the up-regulation of AGG and COL10α1 gene expression and the production of matrix proteins. No difference has been detected in the chondrogenic differentiation potential between ASCs from the abdomen and hip/thigh regions. However, in the PCR studies, we have not succeeded in quantifying COL2α and Col2B mRNA expression for any of the donors tested. Others have obviously faced the same difficulty when trying to measure the up-regulation of COL2B genes in MSCs (Huang et al. 2004; Winter et al. 2003; Zuk et al. 2002). Although intending to measure collagen Type X mRNA expression as a hypertrophic and therefore late marker of chondrogenesis, we have noticed that this mRNA is expressed earlier than collagen Type II mRNA. This is surprising, as we would expect stem cells to have to differentiate into chondrocytes before they can become hypertrophic. However, this unexpected hierarchy of chondrogenic gene expression has also been found by Mwale et al. (2006). Moreover, both Mwale et al. (2006) and we have found AGG to be constitutively expressed in MSCs. Because of this constitutive expression of AGG and the early up-regulation of COL10α1, Mwale et al. (2006) warn against using these molecules as markers for chondrogenesis and chondrocytic hypertrophy. We think that these genes can nevertheless be used as markers for differentiation into the chondrogenic lineage, albeit being exclusively shown as the quantitative up-regulation of gene expression, and always in combination with other chondrogenic markers (an awareness of the possible difference in the function of COL10α1 in chondrogenesis in adult stem cells when compared with embryonic stem cells is also necessary). Our lack of detection of any significant differences in the osteogenic and chondrogenic differentiation potential when comparing ASCs from the two tissue-harvesting sites seems to be in contrast with studies of Hauner and Entenmann 1991) who have found differences in the adipogenic differentiation potential between SVF cells from abdominal and femoral adipose tissue. However, since Hauner and Entenmann 1991) have used fresh SVF cells instead of culture-passaged ASCs, the variation in differentiation potential might be attributable to differences in numbers of ASCs in the SVF isolates of the two regions, as we have shown in this study. Other factors responsible for the variation in differentiation potential can be ascribed to other specific histological characteristics of the adipose tissue at the anatomical site, such as vascularity and amount of fibrous tissue (Lennon et al. 2000; Peptan et al. 2006; Pittenger et al. 2000), and to differences in the regulation of gene expression (Djian et al. 1983; Peptan et al. 2006; Pittenger et al. 2000). We therefore conclude that the yield of ASCs is dependent on the tissue-harvesting site. In planning the optimal one-stage procedure for the regeneration of cartilage tissue, factors that can positively influence the outcome of the operation must be taken into account. In view of this, the abdomen seems to be preferable to the hip/thigh region for harvesting ASCs.

Int_Arch_Occup_Environ_Health-3-1-1915645

Workdays, in-between workdays and the weekend: a diary study on effort and recovery

Objectives Effort-recovery theory (Meijman and Mulder in Handbook of work and organizational psychology, Psychology Press/Erlbaum, Hove, pp 5–33, 1998) proposes that effort expenditure may have adverse consequences for health in the absence of sufficient recovery opportunities. Thus, insight in the relationships between effort and recovery is imperative to understand work-related health. This study therefore focused on the relation between work-related effort and recovery (1) during workdays, (2) in-between workdays and (3) in the weekend. For these three time periods, we compared a group of employees reporting relatively low levels of work-related effort (“low-effort group”) and a group of employees reporting relatively high levels of work-related effort (“high-effort group”) with respect to (1) activity patterns, (2) the experience of these activity patterns, and (3) health and well-being indicators. Introduction Much research has shown that high levels of job demands are related to increased levels of physical and psychological health problems across time (e.g., De Lange et al. 2003). Despite this strong focus on the relations between job demands and health, relatively little attention has been paid to the psychological and physiological processes that may explain why health is adversely affected by high job demands. One notable exception is effort-recovery (ER) theory (Meijman and Mulder 1998; Geurts and Sonnentag 2006). ER theory argues that working inevitably requires effort as an appeal is made to workers’ abilities and their willingness to dedicate these abilities to the work task. Expending effort at work (“work-related effort”) produces two kinds of outcomes: the tangible result of work activities, i.e. a product or service, and the psychological and physiological “costs” or load reactions (e.g., fatigue) associated with working. These load reactions are usually short-lived and reversible: they disappear after respite from work. However, under certain circumstances the recovery process may be insufficient or inadequate, and then short-term work-related load reactions may turn into adverse and more chronic health problems, such as prolonged fatigue, chronic tension, and sleep deprivation (Åkerstedt 2006; Härmä 2006; Sluiter et al. 2001; Van Hooff et al. 2005). Recovery opportunities after work may be inadequate in terms of quantity (time) and/or quality. Recovery time may be insufficient in case of prolonged exposure to high demands, for instance, when workers continue to pursue job-related activities during non-work time (e.g., by working overtime) or engage in other demanding (e.g., domestic) activities. Recovery is particularly at stake when during private time an appeal is made upon the same psychophysiological systems that were activated on the job. The quality of recovery may be endangered when individuals’ psychophysiological systems show prolonged activation even if not exposed to any special demands during the recovery period. This may happen when workers have difficulty to relax at home after a stressful working day. For example, Brosschot et al. (2005) showed that when workers worry in their private time about the past or upcoming working day, the psychophysiological systems that were activated on the job remain activated, thus impeding the recovery process (cf. Ursin and Erikson 2004). Due to repeated or prolonged activation of psychophysiological systems, these systems are in danger of chronic overactivity, producing lasting changes in homeostatic mechanisms (i.e., allostatic load, McEwen 1998). Consequently, these originally adaptive systems may start to malfunction by showing either hyperactivity (the systems fail to shut-off) or hypoactivity (the systems are not turned on when needed), constituting a serious health risk. For example, chronic stress may cause the immune system to be not sensitive enough (hypoactivity), allowing infectious agents (viruses and bacteria) to enter the body and cause infectious diseases. Alternatively, the system may become overreactive so that the immune system itself causes ill health (such as autoimmune diseases and allergic diseases, Clow 2001). The present study Effort and recovery are nowadays salient research topics (Zijlstra and Sonnentag 2006). The present study builds on and extends this body of knowledge in at least four regards: Firstly, although the effort-recovery process is assumed to unfold on a daily basis, there is only a limited number of studies examining this process from such a day-to-day perspective (e.g., Cropley et al. 2006; Meijman et al. 1992; Rook and Zijlstra 2006; Sonnentag 2001; Sonnentag and Zijlstra 2006; Totterdell et al. 1995). The majority of research in this area still focuses on either cross-sectional or on global long-term relations between job demands, lack of recovery and health (e.g., Kompier 1988; Sluiter et al. 2001). Thus, in order to obtain more insight in de day-to-day relations between effort and recovery, the present study examines the relation between work-related effort and recovery on a daily basis, both during and after working time. Furthermore, although weekends may offer important opportunities for recovery, they are hardly included in previous studies. Exceptions are Fritz and Sonnentag’s (2005) diary study, which showed that well-being after the weekend was higher when individuals had engaged in social activities during the weekend. Also, Totterdell et al. (1995) reported that sleep, mood and social satisfaction were worse on the first rest day following work shifts in comparison with subsequent rest days. In a study among shift-working nurses, Rook and Zijlstra (2006) found weekends to be important for recovery as well. To increase the understanding of the weekend as potential recovery period, the present study also included the weekend. Thirdly, only limited attention has been given to actual activity patterns during work and non-work time in research on effort and recovery until now (see for exceptions: Fritz and Sonnentag 2005; Sonnentag 2001; Sonnentag and Bayer 2005). This is remarkable, as several work psychological theories (e.g., action theory, Frese and Zapf 1994; Taris and Kompier 2005) assume that job characteristics affect worker well-being through worker behavior: it is what people do that makes them feel tired or enthusiastic. Thus, in order to fully understand effort-recovery patterns during and in-between workdays, we must know how people spend their time on work as well as on home activities. Therefore, the present study provided a detailed assessment of employees’ activity patterns during and in-between working time. Finally, what can be a burden for one individual may constitute a pleasure to the other. Consequently, insight in activity patterns in the work and private domain is insufficient to fully understand workers’ effort-recovery patterns, and preferably workers’ experience of the time spent on (non)work activities must be examined in this context (see also the recommendations by Sonnentag 2001). Until now, the extent to which workers experience their daily work and home activities as effortful and/or pleasant has nonetheless remained largely ignored. Therefore, the present study provided a detailed assessment of how employees experience their activities during and in-between working time in terms of effort and pleasure. We distinguished between workers who reported a relatively high level of work-related effort (i.e., who generally experienced their workdays as effortful) during a standard work week (further referred to as the “high-effort group”) and workers who reported a relatively low level of work-related effort (“low-effort group”). This division of our sample was employed in order to maximize the contrast between the two subgroups in terms of reported effort. The two effort-groups were compared with respect to (1) activity patterns (i.e., the time spent on/frequency of engaging in work activities, domestic activities, active leisure, and passive leisure), (2) experiences of activities (i.e., the specific effort and pleasure experienced while engaging in a specific work or home activity), and (3) health and well-being indicators (i.e., fatigue, sleep quality, sleep time, preoccupation with work, and work motivation). Fatigue is included an indicator of (lack of) recovery. As sleep provides the most “natural” recovery opportunity for humans, sleep quality and sleep time are incorporated as well (Åkerstedt 2006). Preoccupation with work is assessed, because it may prolong physiological activation and therefore interfere with the recovery process (Brosschot et al. 2005). Finally, to avoid focusing exclusively on the “negative” consequences of working, work motivation is added in this study to acknowledge that work may be related to positive aspects of worker behavior as well. These constructs were measured in three time periods: (1) during work time, (2) in-between successive workdays, and (3) during the weekend. In order to minimize the amount of time elapsed between the occurrence and the reports of a certain activity or experience, we utilized a diary design covering five uninterrupted weekdays directly followed by two weekend days. In this vein, the risk of retrospection bias was reduced (Bolger et al. 2003). This study examines three interrelated research questions:How is work-related effort associated with (a) time spent on work activities, (b) experiences of work activities, and (c) health and well-being during the workday?As the distinction between the two groups is based on employees’ reports of work-related effort, we expect that the high-effort group will also report to have expended higher effort on (at least some of) the specific work activities compared to the low-effort group (Hypothesis 1a). Support for this hypothesis is important from the perspective of validation of the effort-measure used to differentiate between the two effort-groups. As the high-effort group should have invested higher levels of effort during the work day than the low-effort group, we expect to observe higher levels of fatigue at the end of the workday (Hypothesis 1b) as well as a (stronger) increase in fatigue during the workday (Hypothesis 1c) in the first group. We do not hold a priori expectations concerning the experiences of pleasure associated with work activities and with respect to the time spent on and the frequency of engaging in each work activity.How is work-related effort associated with (a) time spent on home activities, (b) experiences of home activities, and (c) health and well-being in-between successive workdays?We distinguish among four categories of home activities, i.e. (1) domestic activities (e.g., household chores), (2) overtime work, (3) active leisure activities (e.g., exercising), and (4) passive leisure activities (e.g., reading for pleasure, watching TV, listening to music; see also Sonnentag 2001). The latter category is considered as “passive”, whereas the other three categories of activities demand effort to some extent, and are therefore labeled as “active”. Based on ER theory, it can be argued that our capacity to expend effort is limited, and that the more effort is expended at work, the less remains for home activities. Accordingly, the high-effort group is expected to spend less time on and to engage less often in active home activities, and consequently, will spend more time on and engage more often in passive leisure activities (Hypothesis 2a). Because of the supposed limited amount of energy left in the high-effort group, we further expect that this group will experience engagement in active home activities as more effortful relative to the low-effort group (Hypothesis 2b). As we do not have a priori expectations regarding differences between the groups in the pleasure experienced in home activities, possible differences are examined in an exploratory fashion. Further, we expect that those who have expended high effort on the job (high-effort group) will report higher levels of fatigue and more sleep complaints in-between workdays compared to the low-effort group (Hypothesis 2c). In order to obtain a full picture of the participants’ recovery in-between workdays, sleep time is also examined. Finally, we assume that workers who have expended higher effort during working time, will also be more preoccupied with their job after work (Hypothesis 2d). This expectation is in line with Sonnentag and Bayer’s (2005) finding that those who experienced high workload during the workday found it more difficult to detach from work during evenings than others. We do not formulate a priori expectations regarding possible differences in work motivation between the two effort-groups.How is work-related effort associated with (a) time spent on home activities, (b) experiences of home activities, and (c) health and well-being during the weekend days?The hypotheses formulated for the period in-between workdays (in "Research question 2") can be extended to the weekend. Hence, we expect that those who have spent high effort on the job during week days (the high-effort group), will—during the weekend—spend less time on and engage less often in active and will spend more time on and will engage more often in passive home activities (Hypothesis 3a), experience active home activities as more effortful (Hypothesis 3b), report more fatigue and more sleep complaints during the weekend (Hypothesis 3c), and will be more preoccupied with the upcoming workweek, than the low-effort group (Hypothesis 3d). Method Participants and procedure This study was conducted in two stages among academic staff members of a medium-sized Dutch university. Of the 696 employees who were tenured and worked at least 3 days a week, only those could participate who (1) did not have a job outside this university (to keep the variation in work activities within acceptable limits), and (2) lived with a partner who worked at least 2.5 days a week (to increase the likelihood that the participants fulfilled at least some home obligations). Of the 146 employees who agreed to participate, 133 (19%) completed a general questionnaire (first stage of the study). Data from 13 of these 133 were removed, as they apparently did not meet one or both of the selection criteria. To already reduce the influence of one possible confounder (i.e., working hours) of the associations between work-related effort and the variables of interest, this study was restricted to employees who worked at least 32 contractual hours a week. As a result, our sample comprised 93 employees (69.6% male; 67.7% ≥1 child living in the household; Mage= 45.0 years, SD = 7.6; 49.5% was assistant professor, 16.1% associate professor, 12.9% full professor, 21.5% other jobs, e.g. researcher or lecturer). Due to strict privacy regulations, we did not know how many of the employees that were approached for participation in the study actually met our inclusion criteria (i.e. had no job outside the university and lived together with a partner who worked at least 2.5 days a week). Therefore, we do not have insight in how many employees were in fact eligible for participation in the study, meaning that the overall response rate and the representativeness of our sample are unknown. In the second stage of this study starting about ten days after the completion of the general questionnaire, the daily variables of interest were assessed by means of short questionnaires that were completed three times a day, from Monday to Sunday: (1) a morning questionnaire (to be completed after awaking in the morning, between 7:30 and 8:30 a.m.), (2) an afternoon questionnaire (to be completed around 6 p.m.), and (3) an evening questionnaire (to be completed before bedtime, between 10 and 11 p.m.). Only diaries that were completed within an acceptable time range around the requested time were included in the final database. We thus removed morning questionnaires that were completed more than 2 h after awakening; afternoon questionnaires that were completed before 4:30 p.m., after 8 p.m., or less than 3 h after the morning questionnaires; and evening questionnaires that were filled in less than 2 h after the afternoon questionnaire or after 3 a.m. This procedure resulted in 76.2% valid morning diaries, 73.4% valid afternoon diaries, and 72.5% valid evening diaries. Variables derived from the general questionnaire (general measures) Job types included “assistant professor”, “associate professor”, “full professor”, and “other”, such as researcher and teacher. Age was measured in years; Gender was coded as “0” for “male” and “1” for “female”. Parental status was coded as “0” for having no children living in the household and “1” for having at least one child living in the household. The last three variables are potential confounders in the relationships of interest and are therefore included as covariates in further analyses. General fatigue was assessed with the ten-item fatigue assessment scale (Michielsen et al. 2003). An exemplary item is “I am bothered by fatigue” [1 “(almost) never”, 5 (almost) always”], with higher scores reflecting higher levels of fatigue (α = 0.86). Work engagement was measured with five items adapted from Rothbard (2001). An example is “When I am working, I often lose track of time” (1 “strongly disagree”, 5 “strongly agree”; α = 0.78). Job pressure was measured with five items from the job content questionnaire (Karasek 1985), that were rephrased as questions [e.g., “Do you have to work very fast?” 1 “(almost) never”, 4 “(almost) always”; α = 0.74]. Job control was measured with six items from Van Veldhoven et al. (2002). An exemplary item is: “Can you take a short break if you feel this is necessary?” [1 “(almost) never”, 4 “(almost) always”; α = 0.67]. Social support from colleagues [e.g., “My colleagues show their appreciation for the way I do my job”, 1 “(almost) never”, 4 “(almost) always”; α = 0.86], and Social support from supervisor [e.g., “My supervisor shows her/his appreciation for the way I do my job”, 1 “(almost) never”, 4 “(almost) always”; α = 0.90] were both measured with four items adapted from Geurts et al. (1999). Positive affect and negative affect were measured by means of the positive and negative affect schedule (PANAS; Watson and Clark 1988). Following Rothbard (2001), we distinguished between positive and negative affect regarding work and positive and negative affect regarding family. Sample items for negative affect are “upset” and “distressed”, and examples for positive affect are “enthusiastic” and “proud” (1 very slightly or not at all, 5 extremely), with higher scores indicating higher negative or positive affect (negative affect: Cronbach’s α = 0.83 for work and 0.84 for home; positive affect: Cronbach’s α = 0.87 for work and 0.90 for home). Life events Participants could report for ten events (e.g., birth of a child, financial troubles, change of job) whether or not they had experienced this event during the past year. The number of events experienced was summed. Measures derived from the daily questionnaires (daily measures) To limit the participants’ burden, the questionnaires contained a combination of validated scales as well as single-item report-marks. Work-related effort In the afternoon questionnaire, participants were requested to indicate with a report mark the extent to which they considered the preceding workday as effortful (1 “not at all”, 10 “extremely”). Activities Time spent daily on work activities Participants received a list of 13 major work activities and indicated the time (0 “none”, 1 “<1 h”, 2 “1–2 h”,..., and 7 “>6 h”) they had spent on each activity during regular work time, i.e., until 6 p.m. (afternoon questionnaire), and during nonwork time, i.e., after 6 p.m. (evening questionnaire). We recoded this time range to obtain an estimate of the actual time in hours by assuming that the actual time spent on an activity would lie halfway the two extremes (e.g., the category “<1 h” was recoded as “0.5”). Time spent on research activities by day comprised the time spent on “conducting research”, “data-analysis”, “reading specialist literature”, and “writing papers” (until 6 p.m.). Time spent on teaching activities by day included the time spent on “preparing a lecture”, “giving a lecture”, “reading (Ph.D.) students’ assignments”, and “appointments with (Ph.D) students” (until 6 p.m.). Time spent on administrative activities by day consisted of time spent on “preparing a meeting”, “attending a meeting” and “e-mail/phone”. The category “informal contact with colleagues” was entered in the analyses separately. A 13th activity, “other”, was not incorporated in further analyses, as on average only 0.42 h were devoted daily to these activities. All work activities are potentially relevant to all participants as in the Netherlands lecturers also have some research time, and researchers will usually also teach. Overtime work was computed by summing the time spent on all 13 work activities after 6 p.m. (this university did not offer evening classes) during weekdays, and by summing the total time spent on work activities before and after 6 p.m. on Saturday and on Sunday. Time spent daily on home activities Participants indicated in both the afternoon (until 6 p.m.) and evening questionnaires (from 6 p.m.) the amount of time they spent that day on ten categories of home activities (largely based on those used in Sonnentag’s (2001) diary study). Answer possibilities and recoding procedure were identical to those used for work activities. To ease interpretation of the categories, participants received examples of activities in each category. Time spent on domestic activities was calculated by summing the total time (i.e., before and after 6 p.m.) devoted each day to “household activities”, “doing odd jobs in or around the house”, “doing the groceries”, “care giving activities” and “businesslike activities”. Time spent on active leisure activities comprised the total time spent daily on “physical activities”, “creative activities” and “social activities”. The total time spent on passive leisure activities was computed by summing the time devoted to these activities (e.g., reading for pleasure, watching TV, listening to music) before and after 6 p.m. The tenth category, “other”, was omitted from further analysis as the mean time spent on these activities ranged from only 0.15 h on Saturday to 0.21 h during weekdays. Experiences Participants indicated for each engaged work and home activity, the extent to which they considered it as effortful and as pleasant (1 “not at all”, 10 “extremely”). An estimate of the average daily effort and pleasure for each category of activities was obtained by computing a weighted mean score. Thus, the summed product of hours spent on each activity within a category and the effort (pleasure) experienced while executing the activity was divided by the total hours spent on the activities in the respective category. By employing such a weighted score, the time spent on an activity is controlled for, assuring that the effort (pleasure) score really reflects effort (pleasure). Health and well-being Fatigue at work (weekdays’ afternoon questionnaire) was measured with eight items adapted from Van Veldhoven et al. (2002), for example “I felt tired mentally” (1 “not at all”, 10 “extremely”). Participants rated each item twice: (1) with respect to the first hour of the workday (Cronbach’s α = 0.87) and (2) with respect to the last hour of the workday (Cronbach’s α = 0.86). Fatigue was measured in the morning, afternoon and evening questionnaires. Participants rated their current state of fatigue (“How fatigued do you currently feel?”) with a report mark varying from “1” (“not at all”) to “10” (“extremely”). Sleep complaints (each morning questionnaire) were assessed using a five-item sleep quality scale (Van Veldhoven et al. 2002), slightly adapted to make it suitable for day-to-day measurement. An exemplary item is: “Last night I woke up several times” (1 “yes”, 0 “no”, α = 0.73 across all seven consecutive days). Note that each day’s values for this scale refer to the previous night. Sleep time (each morning questionnaire) was computed by calculating the self-reported number of hours in-between the time they went to sleep last night (“what time did you go to sleep last night?”) and the time they woke up this morning (“what time did you wake up this morning?”). Again, each day’s values for this scale refer to the previous night. Preoccupation with work (each morning questionnaire) was assessed with one self-developed item: “I am already mentally involved with the things I have to do at work today [next week]” (1 “not at all”, 5 “extremely”). Work motivation (each morning questionnaire) regarding the upcoming workday (during weekdays) or the next workweek (during weekend-days) was assessed with one self-developed item: “I feel like starting the next workday [workweek]” (1 “not at all”, 5 “extremely”). A table with correlations between the study variables can be obtained from the first author on request. Figure 1 gives an overview of the measurement structure. Fig. 1Overview of the measurement structure Creation of the two effort-groups The global report mark for work-related effort as assessed in the afternoon questionnaire was used to create the two effort-groups. A workday was labeled as effortful if a report mark of six or higher was given. The number of effortful workdays was summed for each participant to obtain an estimate of how effortful he/she considered the workweek. To increase reliability only participants who gave a report mark during at least three out of the five possible workdays were selected, resulting in a final sample of 72 of the 93 original participants. The low-effort group (n = 27; Meffort = 3.39) consisted of participants who considered none (out of three) or only one workday (out of four or five) as effortful (>6). The high-effort group (n = 24; Meffort = 6.77) included participants who labeled two or three (out of three), three or four (out of four) or four or five (out of five) workdays as effortful. Statistical analyses Data were analyzed by means of (M)ANCOVA, which allows the examination of relationships between a categorical independent variable (the effort-subgroups) and continuous dependent variables (Maxwell and Delaney 2005). Gender, age, number of children in the household and number of contractual work hours (32 or more) were included as covariates in the analyses, because these may affect the relationships between work-related effort and the other variables of interest in this study. One key assumption of MANCOVA is that the criterion variables are multivariately normally distributed (Maxwell and Delaney 2005). To examine whether this assumption could be maintained, the distributions of the criterion variables were inspected for univariate normality, both for the total sample and for the low and high effort subgroups. The skewness of the criterion variables was for 61 out of 64 variables in the study lower than 1.00. As this number is already expected on basis of chance, this finding suggests that the assumption of a multivariate normal distribution of the criterion variables could be maintained for practical purposes. Results Preparatory analyses Table 1 shows the descriptive statistics for the general measures for the total sample as well as for the two effort groups. The total sample can be characterized as middle-aged, rather engaged, and not very tired. The mean level of fatigue in the sample does not significantly differ from that in a heterogeneous sample of 1,123 employees (M = 1.97, SD = 0.57, T(1214) = 1.29, ns; Geurts et al. 2005). Participants report relatively high levels of work pressure and job control. Levels of job control are higher than those in a heterogeneous sample of 1,740 employees (M = 2.54, SD = 0.63, T(1831) = −10.29, P < 0.001; Geurts et al. 2005). Table 1Means and standard deviations for the total sample and for the two effort-groups for the measures derived from the general questionnaireTotal sample (N = 93)Low-effort group (N = 27)High-effort group (N = 24)MSDMSDMSDAge44.957.6346.746.6944.177.80Work engagement3.940.693.920.634.050.67Fatigue1.890.611.870.572.040.60Work pressure2.470.542.230.532.490.60Job control3.220.423.250.463.300.39Social support colleagues2.590.652.620.532.600.66Social support supervisor2.350.872.490.792.190.90Positive affectivity work3.610.533.570.623.510.51Positive affectivity home3.590.633.740.713.470.52Negative affectivity work1.930.571.930.502.040.67Negative affectivity home1.830.561.830.591.950.47Life events1.981.532.041.342.171.81 To investigate possible differences in the composition of the two effort-groups, these groups were compared with respect to the general measures. No significant differences were observed regarding age (T = 1.27, df = 49, ns), gender (χ2 = 0.07, df = 2, ns), parental status (χ2 = 0.14, df = 1, ns), job type (χ2 = 1.43, df = 3, ns), general fatigue (T = −1.00, df = 49, ns) and work engagement (T = −0.72, df = 49, ns). Also, the MANOVA executed with respect to job characteristics (job pressure, job control, social support from colleagues and supervisor) was not significant, F(4, 45) = 0.98, ns. Furthermore, the groups report comparable levels of positive and negative affect (work: T(49) = −0.71, ns; family: T(49) = −0.13, ns) and positive affect (work: T(49) = −0.39, ns; family: T(49) = 1.01, ns). Finally, the two effort-groups did not differ significantly regarding the number of life events experienced (T(49) = −0.29, ns). Thus, in sum, there were no significant differences between the two effort-groups with respect to the general measures. Differences between Saturday and Sunday Preliminary analyses showed that the variables under study did not differ significantly between Saturday and Sunday (F-values ranged from F(1, 27) = 0.00, ns for effort reported for active leisure, to F(1, 8) = 1.95, ns for effort reported for overtime work). Therefore, further analyses are based on mean scores across the two weekend days. Research question 1 Table 2 presents the means, standard deviations and F-statistics for the daily variables for the total sample and for each of the two effort-groups. As to work activities, two analyses were conducted. First, for each participant the percentage of days on which time was spent on each work activity was computed. MANCOVA revealed that these percentages did not differ significantly between the two effort-groups. Secondly, for each participant we computed the mean time they spent daily on each work activity across the five weekdays. Again, MANCOVA did not reveal any significant difference between the two effort-groups. Thus, the two effort-groups did not differ significantly in their work activity patterns during the workday. Table 2Activity patterns, experiences and recovery indicators during the workdayHypothesisTotal sample (N = 93)Low-effort group (N = 27)High-effort group (N = 24)F (df)PMSDMSDMSD–% Days Multivariate 1.29 (4, 42)0.29 Research6133603762300.60 (1, 45)0.81 Teaching7530673779273.33 (1, 45)0.07 Administrative9016901686200.00 (1, 45) 0.99 Informal contacts5533553355300.32 (1, 45)0.57–Time Multivariate0.31 (4, 42)0.87 Research1.51.31.71.51.61.30.29 (1, 45)0.59 Teaching2.01.42.01.72.11.40.36 (1, 45)0.55 Administrative1.91.01.91.21.71.00.02 (1, 45)0.90 Informal contacts0.40.30.40.30.30.20.89 (1, 45)0.35Hypothesis 1aEffort (1–10)  Multivariate9.94 (4, 29)<0.001 Research4.872.203.241.666.441.7826.53 (1, 32)<0.001 Teaching4.471.763.241.495.871.2229.59 (1, 32)<0.001 Administrative4.051.892.741.345.441.8025.64 (1, 32)<0.001 Informal contacts2.601.481.851.183.301.627.76 (1, 32)<0.01–Pleasure (1–10)  Multivariate 0.07 (4, 29)0.99 Research7.181.097.141.377.231.050.00 (1, 32)1.00 Teaching6.691.046.881.346.610.600.01 (1, 32)0.94 Administrative5.761.525.941.465.920.910.02 (1, 32)0.89 Informal contacts7.520.957.311.217.540.880.17 (1, 32)0.68Hypotheses 1b and Hypothesis 1cHealth and well-being Work-related fatigue first hour1.891.031.690.651.960.91Time: 0.00 (1, 45)0.96Group: 4.63(1, 45)<0.05 Work-related fatigue last hour2.561.302.130.863.001.43Time × group: 6.09 (1, 45)<0.05F-statistics, P-values, and means and standard deviations for the total sample and for the two effort-groups To study possible differences in experiences, two MANCOVA’s were conducted, both based on mean scores across the five workdays. The first analysis revealed that the two effort-groups differed significantly in the average amount of effort reported with respect to the four work activities. Univariate tests showed that the high-effort group experienced each activity as more effortful (Hypothesis 1a supported). The second analysis revealed that the two effort groups did not differ significantly with respect to the pleasure they derived from their work activities. Possible differences between the two effort-groups in fatigue at work were examined in a 2 (Time: first hour vs. last hour) × 2 (Group: low vs. high effort) repeated-measures ANCOVA. The development of fatigue during the workday differed significantly between the two effort-groups (significant time × group interaction). Post-hoc analyses showed that there were no significant differences between the two groups in their level of fatigue during the first hour of the workday (T = −1.20, df = 49, ns). However, the high-effort group reported a significantly higher mean level of fatigue during the last hour of the workday (T = −2.66, df = 49, P < 0.05), indicating that the high-effort group reported more fatigue at the end of the workday (Hypothesis 1b supported), and showed a stronger increase in fatigue (Hypothesis 1c supported) during the workday. In sum, the participants in the two effort-groups do not engage in different types of work activities, nor do they experience their work activities differently in terms of pleasantness. However, the high-effort group reports to spend significantly more effort on each of the work activities, experiences significantly higher work-related levels of fatigue at the end of the work day, as well as a stronger increase in fatigue during the workday. Research question 2 Means, standard deviations and F-statistics are presented in Table 3. As to home activity patterns, two analyses were performed. First, for every participant, we computed the percentage of workdays they spent time on each type of home activity (domestic, active leisure, overtime work, and passive leisure). For each of these activities, an ANCOVA was conducted. Results showed that the groups only differed significantly with respect to active leisure activities: whereas the high-effort group spent on less than half of the work days (43%) time on this type of activities, the low-effort group spent on more than half of the work days (62%) time on this type of activities. Secondly, we conducted four ANCOVA’s based on each participant’s mean time spent daily on each of the four activities during the five weekdays, but these revealed no differences between the two effort-groups. These results provide partial support for Hypothesis 2a by showing that participants in the high-effort group engage on average less often in active leisure activities. Table 3Activity patterns, experiences, recovery indicators, work involvement and work motivation in-between workdaysHypothesisTotal sample (N = 93)Low-effort group (N = 27)High-effort group (N = 24)F (df)PMSDMSDMSDHypothesis 2a% Days Domestic8620902283181.34 (1, 45)0.25 Active leisure 5633622643318.12 (1, 45)<0.01 Overtime4923432758344.92 (1, 45)<0.05 Passive leisure7327782870340.63 (1, 45)0.43Hypothesis 2aTime Domestic 2.31.62.21.42.01.51.47 (1, 45)0.23 Active leisure0.90.70.90.50.70.71.90 (1, 45)0.18 Overtime0.90.80.80.71.11.02.42 (1, 45)0.13 Passive leisure1.31.11.30.81.20.80.36 (1, 45)0.55Hypothesis 2bEffort (1–10) Multivariate 10.90 (4, 27)<0.001 Domestic 2.961.572.461.223.321.760.80 (1, 30)0.38 Active leisure3.581.762.911.364.561.997.24 (1, 30)<0.05 Overtime4.341.763.131.485.821.1526.11 (1, 30)<0.001 Passive leisure2.061.241.500.822.811.583.37 (1, 30)0.08–Pleasure (1–10) Multivariate 0.67 (4, 27)0.62 Domestic 5.791.406.081.345.601.210.38 (1, 30)0.54 Active leisure7.311.237.431.207.710.750.03 (1, 30)0.86 Overtime6.341.296.331.496.160.950.07 (1, 30)0.80 Passive leisure6.871.437.191.056.871.330.21 (1, 30)0.65Hypothesis 2c and 2dHealth and well-being Fatigue t13.761.932.901.474.871.72Time: 2.16 (2, 44)0.13 Fatigue t24.921.743.901.565.881.40Group 22.46 (1, 45)<0.001 Fatigue t35.711.984.872.076.521.50Time × group 0.06 (2, 44)0.95 Sleep complaints1.541.121.251.001.950.954.12 (1, 45)<0.05 Sleep time7.090.886.851.057.130.701.85 (1, 45)0.18 Preoccupation work3.321.003.190.903.750.954.41 (1, 45)<0.05 Work motivation3.400.703.490.853.210.503.00 (1, 45)0.09F-statistics, P-values, and means and standard deviations for the total sample and for the two effort-groups In order to investigate possible differences in their experiences of home activities, two MANCOVA’s were computed, both based on mean scores across the five weekdays. The first analysis showed an overall significant difference between the two effort-groups in the extent to which they considered home activities as effortful. Univariate tests demonstrated that the high-effort group considered active leisure activities and overtime work as more effortful (Hypothesis 2b supported). The second analysis revealed that the two effort-groups did not differ significantly as to their pleasure regarding their home activities. Regarding health and well-being, three analyses were conducted, each based on mean scores across the five workdays. For fatigue, a 3 (time: morning vs. afternoon vs. evening) × 2 (group: low vs. high effort) repeated-measures ANCOVA indicated that fatigue did not vary significantly as a function of time. However, the two effort-groups did differ significantly in their average level of fatigue (main effect of Group). Post-hoc analyses demonstrated that the high-effort group reported higher levels of fatigue (M = 5.76) compared to the low-effort group (M = 3.89, T = −4.78, df = 49, P < .001; Hypothesis 2c supported for fatigue). The development of fatigue during the day did not vary significantly as a function of effort-group (time × group interaction, ns). Furthermore, ANCOVA revealed that the high-effort group reported significantly more sleep complaints (Hypothesis 2c supported for sleep complaints). The third analysis (ANCOVA) showed that the two effort-groups did not differ significantly with regard to sleep time. Concerning preoccupation with work, ANCOVA revealed that the high-effort group was significantly more preoccupied (Hypothesis 2d supported). A similar analysis conducted for work motivation did not reveal any significant differences between the two effort-groups. In sum, the high-effort group engaged less often in active leisure activities in-between successive workdays, but did not differ significantly from the low effort-group regarding the experience of pleasure associated with these activities. Further, the high-effort group experienced the home activities as more effortful. In addition, we systematically observed higher levels of fatigue; more sleep complaints, and a higher preoccupation with work in the high-effort group in-between workdays. Research question 3 Table 4 presents the relevant means, standard deviations and F-statistics for the total sample and for the two effort-groups. To map possible differences in activity patterns between the two effort-groups, two analyses were conducted. First, we computed for each participant the percentage of weekend days on which time was spent on each home activity: The four ANCOVA’s (one for each percentage) conducted for these percentage revealed no differences between both effort-groups. Secondly, with respect to the time spent on the four types of home activities, also for each activity an ANCOVA was conducted. Results revealed one important difference in activity patterns: the high-effort group spent significantly more time on overtime work during the weekend compared to the low-effort group (Hypothesis 3a partially supported). Table 4Activity patterns, experiences, recovery indicators, work involvement and work motivation during the weekendTotal sample (N = 93)Low-effort group (N = 27)High-effort group (N = 24)F (df)PMSDMSDMSDHypothesis 3a% Days Domestic 9713981096140.13 (1, 45)0.74 Active leisure7032762963303.16 (1, 45)0.08 Passive leisure8726852779330.27 (1, 45)0.61 Overtime4339374150422.23 (1, 45)0.14Hypothesis 3aTime Domestic 5.02.45.02.05.32.50.14 (1, 43)0.71 Active leisure2.52.02.41.62.42.60.11 (1, 43)0.74 Passive leisure2.61.62.71.52.71.90.38 (1, 43) 0.54 Overtime1.21.40.80.91.41.76.14 (1, 45)<0.05Hypothesis 3bEffort (1–10)  Domestic 3.331.802.651.443.911.975.87 (1, 45)<0.05 Active leisure3.241.922.411.643.872.136.50 (1, 42)<0.05 Overtime4.542.013.421.725.491.809.01 (1, 24)<0.01 Passive leisure2.091.421.731.242.651.595.49 (1, 41)<0.05–Pleasure (1–10)  Domestic 6.241.106.201.316.071.020.15 (1, 45)0.70 Active leisure7.680.887.580.977.610.820.02 (1, 42)0.89 Overtime6.031.576.281.785.961.250.58 (1, 24)0.45 Passive leisure7.470.957.411.107.570.860.14 (1, 41)0.71Hypothesis 3c and 3dHealth and well-being Fatigue (t1)3.432.102.691.694.002.24Time: 3.39 (2, 41)<0.05 Fatigue (t2)4.262.043.561.644.722.21Group: 7.80 (1, 42)<0.05 Fatigue (t3)5.692.094.562.196.381.57Time × group: 0.15 (2, 41)0.87 Sleep complaints1.041.110.941.071.401.311.44 (1, 44)0.24 Sleep time7.881.057.741.107.910.940.69 (1, 43)0.41 Preoccupation work2.411.072.090.942.691.152.28 (1, 44)0.14 Work motivation3.300.983.500.912.890.995.27 (1, 44)<0.05F-statistics, P-values, and means and standard deviations for the total sample and for the two effort-groups Conducting multivariate analyses for “pleasure” and “effort” would result in very restricted sample sizes (n = 13 in both groups). Therefore, only univariate tests were computed, revealing that the high-effort group considered all four activities significantly more effortful (Hypothesis 3b supported). Again, the two effort-groups did not differ significantly with respect to pleasure associated with their activities. Three analyses were conducted to examine possible differences between the two effort-groups regarding health and well-being indicators. A 3 (time: morning vs. afternoon vs. evening) × 2 (group: low vs. high-effort) repeated-measures ANCOVA revealed a main effect of Time. Post-hoc analyses indicated that fatigue increased significantly during the day, that is, was lowest in the morning (M = 3.32), somewhat higher in the afternoon (M = 4.08) and highest in the evening (M = 5.41). Furthermore, overall, the two effort-groups reported different levels of fatigue (significant main effect of group). Post-hoc analyses showed that the high-effort group reported significantly higher levels of fatigue (M = 5.03) than the low-effort group (M = 3.61; Hypothesis 3c supported for fatigue). Finally, fatigue did not vary significantly between the effort-groups as a function of time of the day (non-significant time × group interaction). Two ANCOVAs indicated that sleep complaints and sleep time did not vary significantly between the effort-groups (Hypothesis 3c rejected for sleep complaints). Two additional ANCOVAs indicated that the two effort-groups did not significantly differ with respect to preoccupation with work during the weekend (Hypothesis 3d rejected), but that the high-effort group felt less like starting the next working week (work motivation). In sum, the two effort-groups did not show significantly different activity patterns during the weekend regarding domestic work, active and passive leisure. However, the high-effort group spent significantly more hours on overtime work during the weekend than the low-effort group. Furthermore, the high-effort group experienced all home activities as significantly more effortful, although not as less pleasant, than the low effort-group. We also observed significantly higher levels of fatigue during the weekend and less motivation to start the upcoming workweek in the high-effort group. Discussion The present study was devised to enhance our insight in the associations between work-related effort and recovery from that effort. To this purpose, we compared two groups of employees reporting different levels of work-related effort (high vs. low) with respect to their activities, experiences, and health and well-being in three time-periods: (1) during work time, (2) in-between work days and (3) during the weekend. Activity patterns Our results revealed that the two effort-groups did not differ significantly in terms of their activity patterns at work. However, two significant differences were observed in the home domain. The first manifested itself in-between work days: the high-effort group performed active leisure activities on fewer days than the low-effort group, which is unfortunate, as active leisure activities seem to promote recovery (Sonnentag 2001). A second difference appeared during the weekend. Contrary to our expectations (Hypothesis 3c), employees in the high-effort group spent more time on working overtime in the weekend. This implies that these employees devote part of potential recovery time during the weekend to activities that may interfere with the recovery process (cf. Sonnentag 2001). The amount of time devoted to domestic activities during the weekdays and weekend days did not vary significantly between the two effort-groups. This may be due to the fact that many domestic activities are obligatory in nature (e.g., it is difficult to circumvent doing the household chores). Finally, no significant differences between the groups emerged concerning low-effort activities, both during weekdays and weekend days. Experiences Regarding experiences, we distinguished between effort and pleasure. The high-effort group reported significantly more effort for all work activities. In the home domain, the high-effort group judged overtime and active leisure activities as more effortful during weekdays and judged all activities as more effortful during the weekend. No significant differences between the groups were observed with respect to pleasure, neither during work time nor in-between workdays, nor in the weekend. Hence, work-related effort is independent of the pleasure derived from work and home activities. Health and well-being We observed a stronger increase in work-related fatigue during the workday for the high-effort group than for the low-effort group. Thus, whereas the two groups did not differ significantly in work-related fatigue at the start of the working day, the high-effort group was more fatigued at the end of the working day. This difference persisted in-between workdays. This finding might explain why the high-effort group engaged less often in active leisure in-between workdays than the low-effort group. Also during the weekend, the high-effort group remained significantly more fatigued than the low-effort group. Possibly, this may be due to the fact that the former group spent more time on overtime. A somewhat different pattern of results was observed with respect to sleep complaints: The high-effort group reported more sleep complaints during the week, but not in the weekend. The additional finding that the two groups did not differ significantly with respect to sleep time suggests that work-related effort relates to sleep quality, but not sleep quantity. During the week, the high-effort group was apparently more preoccupied with work than the low-effort group. However, it cannot be excluded that this is partly due to our item wording. Although we asked participants to indicate the extent to which they were already preoccupied with the upcoming workday, it would seem possible that this measure (also) reflects the extent to which participants were still ruminating about their past working day. In the weekend, the two groups did not differ significantly in their preoccupation with the upcoming workweek. This is surprising, as the high-effort group spent more time on work-related activities during these days. The high-effort group nonetheless reported less work motivation than the low-effort group. Limitations and suggestions for future research Six issues with respect to the present study must be discussed. First, as we employed a single item report mark to create the two effort-subgroups, the reliability and validity of this measure can be questioned. However, we believe that there are good arguments in favor of employing this report mark: (1) we did not rely on a single observation of this measure, as each participant completed the item on at least three occasions; (2) employees in the high-effort group considered each of the four categories of work activities as more effortful than the low-effort group, thus suggesting that the report mark correctly reflects the effort experienced during the workday; and (3) there is a correlation of 0.85 (P < 0.001) between our single-item effort-measure and a weighted mean score of the effort experienced during the separate work activities. The latter was computed by first weighting the number of hours spent on each activity by the effort expended to this activity; the sum of these weighted scores was divided by the total number of work hours. Thus, our single-item report mark seems to measure a very similar quantity as a much more refined measure of effort. A second point of concern is the procedure used to create the two effort-subgroups. These groups were created based on the number of days participants considered their work as effortful. To probe the possibility that our findings are biased by this somewhat arbitrary procedure we repeated our analyses using a slightly different effort indicator. For each participant who completed the report mark of global work-related effort on at least three occasions, the mean score on this report mark across the week was computed. Based on these scores, two new subgroups were created: One including participants with scores in the highest tertile, and a second with participants having scores in the lowest tertile. Analyses were repeated for these two groups, yielding results that were highly similar to those found for the original subgroups (results can be obtained from the first author on request). Thus, our findings appear robust across different measures of effort expenditure. Thirdly, our study relied exclusively on self-report measures, and this might have resulted in an overestimation of the associations among the variables due to common method variance. However, this should have inflated all relations studied and not just part of these: the fact that some relationships were found while others were not, argues against the influence of common method variance in our study. Besides, alternative measures such as observational or physiological measures are not free of error variance either, and should therefore not be considered superior to self-report measures (Semmer et al. 2004; see also Kompier 2005). Furthermore, by demonstrating (1) that using self-reports does not guarantee finding significant results, (2) that potential biasing variables (social desirability, negative affectivity and acquiescence) do not generally inflate correlations among study variables and (3) that monomethod correlations are not by definition higher than multimethod correlations, Spector (2006) concludes that “the popular position suggesting common method variance automatically affects variables measured with the same method is a distortion and oversimplification of the true state of affairs” (p 221). Thus, all in all we do not believe that common method bias severely biased our findings, although the use of physiological and performance measures in addition to self-reports could provide interesting insights in future research. A fourth issue is the impact of potential third variables. One might argue that differences between the two effort-groups regarding (experiences of) activities and health and well-being indicators might be due to personality characteristics or other person or work-related constructs, rather than to work-related effort. However, in our study we attempted to exclude the influence of these variables to our best ability: The two effort-groups turned out not to differ regarding the number of life events experienced, general work characteristics (work pressure, job control, social support), fatigue, work engagement, age and positive and negative affect. Of course this does not exclude the possibility that other third variables (e.g., other aspects of personality) may have acted as third variables in this study. Fifth, this study did not offer insight in the intriguing question into the origin of the differences in work-related effort between the two subgroups studied. It may be that these differences are at least partly due to differences in participants’ objective work performance (e.g., number of publications or student evaluations), but such measure was not incorporated in this study. Thus, it is unclear how the differences between the two effort-groups in their work-related effort are related to real output differences, and future studies on this topic should also include objective measures of task performance. Finally, the present research employed a very specific sample, consisting of academic staff members who worked at least 32 h a week and who lived together with a partner who worked at least 2.5 days a week, and who, as is common for tenured academics in the Netherlands, have relatively high job security and are not dependent on fund raising. Although we believe that our main findings on the relations among effort, recovery, health and well-being are not unique to this sample, it is desirable to replicate this study for employees in other professions, in other family situations and/or with other working hours. Thus, future studies should employ samples from other contexts to broaden our understanding of effort and recovery patterns. Assets of this study In spite of these limitations, we believe that the present study extends and enhances previous research on effort and recovery in at least four respects. First, this study is among the very few that examine effort and recovery from a day-to-day perspective, allowing us to demonstrate that work-related effort is related to various aspects of daily work and (potential) recovery time. In this vein, this study shows how effort expenditure at work is actually imbedded in everyday life, and how it relates to recovery during time-off-the job. Secondly, this study emphasized the importance of the weekend as a (potential) opportunity for recovery. Whereas some differences between the two effort-groups persisted throughout the weekend (e.g., higher levels of fatigue and effort-investment for the high-effort group), other differences manifested themselves only during the working week (i.e., less active leisure, more sleep complaints and more preoccupation with work for the high-effort group) or only during the weekend (i.e., more overtime work and less work motivation for the high-effort group). Thus, not all workers employ the recovery opportunities offered by the weekend in a similar fashion: some seem to employ the weekend as a means to catch up with their overdue tasks. These results suggest that it would be worthwhile to study the reasons why workers differ in the way they use their weekend. Thirdly, by paying attention to employees’ activity patterns at work and outside work, we were able to show that effort expenditure at work relates to activity patterns in the home domain. Namely, high levels of effort expenditure at work were associated with less engagement in active leisure and more overtime work. This finding thus suggests that for some workers, high effort expenditure at work is not compensated by a corresponding degree of participation in recovery activities. Given that an imbalance between effort and recovery is associated with adverse health outcomes, this particular group of workers may, in the long run, be a risk group for the development of ill health. Fourthly, we demonstrated that experiences associated with engagement in work and home activities are important: higher effort investment at work is related to experiences of higher effort expenditure outside work, but not to experiences of less pleasure regarding work or home activities. Practical implications Based on our study’s results, three practical suggestions can be formulated. Firstly, adequate control opportunities in the job setting will allow workers to adjust their work behavior to their current need for recovery and, thus, to prevent the development of negative load reactions during working. Secondly, employees should be encouraged to engage in leisure activities that potentially contribute to the recovery process, such as active leisure. Finally, the time spent on overtime work should be kept within acceptable limits, as overtime work impedes the recovery process. Employers should not to demand excessive overtime work from their employees, in order to guarantee sufficient (potential) recovery time (see also Beckers et al. in press). Theoretical implications Our study revealed that workers who invest high effort at work differ in their off-the-job activity patterns from those who invest low effort at work: members of the first group are to a lesser degree engaged in active leisure during evenings in-between work days, and they spend more time on overtime work during the weekend. This different activity pattern may have consequences for the recovery process, as previous research suggests that active leisure promotes recovery, whereas overtime work impedes this process (Sonnentag 2001). That recovery is endangered in the high-effort group is also evidenced by the higher levels of fatigue during non-work time (evenings and weekends) and the lower sleep quality during the week. Apart from a different activity pattern, those expending high effort at work also expend high effort on home activities, which also may endanger the recovery process. Therefore, despite the fact that those investing high effort at work do not experience their activities as less pleasant than those expending low effort, they may be considered at risk for developing health problems in the long run.

J_Med_Internet_Res-4-1-1761926

A Framework for the Evaluation of Internet-based Diabetes Management

Background While still in its infancy, Internet-based diabetes management shows great promise for growth. However, the following aspects must be considered: what are the key metrics for the evaluation of a diabetes management site? how should these sites grow in the future and what services should they offer? Introduction Management of patients with chronic conditions is a long-standing challenge for health care organizations. These conditions include diabetes, chronic heart failure (CHF), chronic obstructive pulmonary disease (COPD), Asthma, HIV/AIDS, and cancer. Patients are required to adopt lifelong exercise, diet, and drug regimens to maintain optimal health and avoid the complications of the disease. These complications can arise suddenly and be life threatening; therefore, patients with chronic diseases must be monitored constantly [1]. In recent years, Internet-based home telemonitoring systems have become available [2]. These sites leverage the Internet to record, measure, monitor, manage, and deliver health care. These information-technology solutions are creating a link between patient and caregiver that enables patients to supply a steady stream of valuable health information to caregivers. For example, diabetics can report their blood glucose readings, thus creating a history of their glucose control, which caregivers can use to evaluate the impact of a therapy (eg, short acting insulin) or the need for a different one [1]. Conversely, caregivers have the ability to provide their patients with crucial information and feedback on the management of their disease. For example, patients can be notified about screening appointments for the complications of diabetes. Therefore, patients benefit from an improved control and understanding of the disease; the ability to self-monitor from home reduces the burden of the disease. These solutions have resulted in dramatic improvements in disease management as measured by hospitalizations [1] and in an overall reduction in costs [3]. Further, patients report higher levels of satisfaction and better control of their conditions [4]. Diabetes is a chronic disease that affects 30 million people worldwide [5] and is the seventh leading cause of death in the United States [6]. The total annual economic cost of diabetes in 1997 was estimated to be US $98 billion. That includes US $44 billion in direct medical and treatment costs and US $54 billion for indirect costs attributed to disability and humanity [7] and a significant intrusion in the life of an individual. In managing diabetes, success is measured by positive change in prognostic indicators and outcomes. Below is a list of measurement criteria used in diabetes management [8,9,10]. Greater patient self-efficacy Greater satisfaction with care, continuity, provider, quality of health outcome Decreased HbA 1cand blood glucose levels Improved diet and body weight control Lowered cholesterol Lowered perception of diabetes intrusiveness Improved quality of life Less depression Decreased incidence of diabetic complications. Primarily, diabetes must be managed by the patient because it requires adherence to stringent dietary, physical, and medical regimes [8]. Internet-based diabetes management systems have the potential of reducing the burden of disease both to the patient and to the health care system. A recent study found that a high proportion of patients are willing to use Internet resources in the management of their disease [9]. The driving forces behind the proliferation of technology for disease management is the patients' demands to get real-time help, get real-time information, and keep in contact with their physician [1]. Not surprisingly, several diabetes-specific sites have recently appeared [10], including myDiabetes, Health Hero Network, LifeChart, LifeMasters, and Medifor. The purpose of this paper is to review the patient's and the health care professional's needs in an Internet-based diabetes management solution and to examine how these needs are addressed in practice. An evaluation framework was constructed by grouping the requirements of an Internet-based diabetes management solution into 5 categories: Monitoring, Information, Personalization, Communication, and Technology. Two of the market leaders (myDiabetes and LifeMasters) were selected and evaluated to illustrate the use of the framework. Methods A literature search was conducted on medical databases (Medline, Pre-Medline, EMBASE, Cochrane, and PubMed) and a nonmedical database (Expanded Academic ASAP). The articles were identified by diabetes, chronic disease, internet, and technology. The searches were based on the following AND combinations of these keywords. diabetes AND internet diabetes AND technology chronic disease AND internet chronic disease AND technology The exact search methodology differed among databases due to differences in their user interfaces. The methodology for each database is summarized in Table 1. The abstracts of the articles retrieved by the searches were screened for relevance by the authors. The relevant articles were reviewed in order to compile a comprehensive list of requirements for an Internet-based diabetes management solution. These requirements were identified on the following basis: No interdependence between requirements Requirements can be assessed as present or not present Equal implementation effort required to satisfy the requirements. The implementation effort was quantified by the number of Use Cases as defined by the Universal Modeling Language (UML) [11,12]. The number of Use Cases ranged from 1 to 3 for each requirement. For example, the requirement defined as User defined parameter-Patient allows patients to define which health parameter they wish to monitor. This functionality requires 3 Use Cases: Identify User, Retrieve Parameters, and Save Parameters. The requirements for Internet-based diabetes management were compiled into the criteria of an evaluation framework. The evaluation criteria were grouped into 5 categories: Monitoring, Information, Personalization, Communication, and Technology. The evaluation framework is presented in Table 2 and the evaluation criteria are discussed in detail in the "Evaluation Criteria" section of the "Results" section. To illustrate the use of the evaluation framework, we have applied it to 2 existing Internet-based diabetes management systems: my Diabetes (www.myDiabetes.com) and LifeMasters (www.lifemasters.com). These 2 sites were selected because they were first movers in the arena of Internet-based diabetes management. MyDiabetes.com was one of the first sites going live in July 1999, shortly followed by LifeMasters.com in October 1999. The sites were evaluated from November 1, 2001 through December 15, 2001. The evaluations were performed by 5 independent evaluators who were not aware of each other's ratings. All evaluators are computer literate and are familiar with the use of the Internet. The evaluators included a physician, 3 diabetic patients, and one author [CM]. All the evaluators registered separately with both sites (registration was free). Each evaluator was given a detailed description of the evaluation criteria, as described in the "Results" section, and Table 2, which describes the framework. The evaluators were also given an evaluation form to fill out (effectively Table 3 without results). For each criterion, the evaluators rated the sites as Yes if the criterion was satisfied or No if it was not satisfied. The evaluations were not supervised. Figure 1 and Figure 2 are screen shots of the entry forms for the daily glucose measurements forms at myDiabetes and LifeMasters respectively. This basic function of diabetes monitoring requires the user to input his or her blood glucose levels and the time of the readings. The data is stored, effectively creating a log of the glucose control of the patient. LifeMasters records glucose levels based on relative times such as Bedtime and asks for symptoms of high and low blood glucose as well as diabetic complications. Mydiabetes records the exact time of the blood glucose measurement but does not screen for any symptoms; this is done in another section of the site. Table 1 Search methodologies for databases Database Search Methodology Medline (1966 to October week 5, 2001) AND diabetes chronic disease internet technology 1 and 31 and 42 and 32 and 4The 4 terms were searched separately by entering the search string, exploding the subject, and selecting all subheadings. The search results were combined using the AND condition. The search history is described below: diabetes chronic disease internet technology 1 and 3 1 and 4 2 and 3 2 and 4 Cochrane and Pre-Medline AND diabetes chronic disease internet technology 1 and 31 and 42 and 32 and 4The 4 terms were searched separately. The search results were combined using the AND condition. The search history is described below: diabetes chronic disease internet technology 1 and 3 1 and 4 2 and 3 2 and 4 EMBASE (via ScienceDirect), Expanded Academic ASAP, PubMed AND All Fields All Years diabetes AND internet diabetes AND technology chronic disease AND internet chronic disease AND technology The terms were searched in combination using the AND condition. The terms were searched in All Fields and for All Years indexed. diabetes AND internet diabetes AND technology chronic disease AND internet chronic disease AND technology Figure 1 The myDiabetes entry form for the daily glucose measurements Figure 2 The LifeMasters entry form for the daily glucose measurements Table 2 Evaluation framework Evaluation Criteria Description Monitoring User defined parameters Health care worker Health care professionals can specify the parameters to monitor Patient Patients can specify the parameters to monitor User defined parameter ranges Health care worker Health care professionals can specify the normal ranges for monitored parameters Patient normal ranges for monitored parameters Automated data collection Vital data can be downloaded directly from the measurement device (eg, Glucometer) Alert algorithms to avoid false alarms Entry validation Validation that patient data is not the result of mistyping (eg, Realistic glucose levels) Screening of symptoms Determine if changes in vital data is associated to symptoms indicative of an emergency Patient involvement in alert Involving the patient in the decision to notify a health care professional Multidisciplinary approach Multiple aspects of disease management monitored The monitoring is based on a multidisciplinary approach to diabetes Physical Monitoring of physical parameters (blood glucose, weight, blood pressure, etc.) Social Monitoring of the social aspects of diabetes (stigma, dieting, etc.) Psychological Monitoring of psychological aspects of diabetes (depression, loss of motivation, etc.) Patient access to multiple specialists Allowing for communication to multiple experts (dietitians, endocrinologists, etc.) Proactive outreach Notification to patients medications, health care appointments, etc Notifications to health care professionals are reminded of screening test and visits Feedback Retrieve and review medical information Patients can retrieve their medical data to monitor their progress (tabular or graphical format) Regular Feedback control of diabetes is administered and stored Information Quality of information site should conform to an accepted level of standards Pull Navigation Navigation should be based on a logical categorization of data Search Search Functionality availability Push Notifications The system should notify its users of newly available information of interest based on their profile (eg. New research) Newsletter subscription Users can subscribe to a specific newsletter that is delivered via e-mail of Web browser Personalization Assessment and feedback diabetes should be assessed using standard evaluation tools Collaborative goal setting management should be clearly specified Identification of barriers and supports Using questionnaires to determine each patient's barrier and the appropriate support measures Follow-up support Re-iteration of support measures Construction of personalized management plan Tailored management plan as a central feature of the site (can be represented as schedules) Modification of management plan The ability for users to modify their plans Language and ethnicity Multilanguage delivery and culture conscious content Communication Health professional- patient Synchronous A channel for one-to-one synchronous communication (eg, videoconferencing) Asynchronous A channel for one-to-one asynchronous communication (eg, secure email) Indirect Technical representation of the health care professional Community creation Chat rooms Synchronous many-to-many communication channels Newsgroups / Forums Asynchronous many-to-many communication channels Expert moderation Communication channels are based on the dialogue with an expert Technology Security Authentication Identification of users (usually username/password) Encryption Data transmission security level (eg, 128-bit) Usability and user-acceptance Evaluation of usability and user-acceptance (achieved with questionnaires, usage monitoring etc.) Reliability and availability Service should be available at all times Open architecture on open standard technologies Statistical Analysis Cohen's multi-rater kappa [13,14] was used to evaluate the agreement between raters for the evaluation framework as a whole. The multi-rater kappa was calculated with SPSS statistical software using the mkappasc procedure. Results Evaluation Criteria In this section, we describe in detail the evaluation criteria presented in Table 2. Monitoring Successful patient monitoring is reliant on efficiently extracting the relevant information from a patient without excessive intrusiveness to both patient and health care professional. Several parameters can be monitored; some examples are blood glucose, weight, blood pressure, diet, foot care, smoking, and nutrition [4,15,16]. Health care professionals should be able to designate which parameters they want to monitor and specify the ranges for each patient. The health care professional should be able to indicate which course of action the system should take if the readings are outside the ranges (eg, notification, triage). Patients should also be able to designate parameters in an effort to improve self-management and goal setting (addressed in the "Personalization" section of "Evaluation Criteria") [17]; these, however, should be in addition to - and clearly differentiated from - the parameters specified by the health care professional. Patient-designated parameters should not be shared with the health care professional unless the patient desires that they be shared. The degree of intrusiveness is a fundamental consideration when designing a diabetes management system. A major problem with many disease-management programs using information technology is that they try to collect too much data too often [1]. The desire to collect as much data as possible must be balanced with the disruption it may cause in a patient's life [4]. Successful strategies to reduce intrusiveness are based on automatic data gathering such as Glucometers that transmit glucose readings via the Internet and the use of simplified questionnaires for triage and screening. Intrusiveness to the health care provider is also an important consideration. If systems were designed to send alerts each time a patient's blood sugar readings are outside the normal parameters, the result would be many false alarms. Therefore, systems must have processes in place designed to not overwhelm health care professionals. These processes include entry validation, screening with the use of questionnaires, and patient involvement in the decision to launch an alert [1]. Effective patient monitoring is not limited to the collection of health data, it also requires a multidisciplinary approach, proactive outreach, and feedback. Multidisciplinary Approach The management of diabetes spans multiple medical specialties as evidenced by the use of multidisciplinary diabetes management teams. For example, an endocrinologist will manage medications and glucose levels, a dietitian will design an appropriate diet, and a psychologist will manage the mental aspect of dealing with diabetes. Internet-based diabetes management programs should be based on a multidisciplinary teamwork. This element consistently appears in successful chronic-disease management systems [18]. Patients should have the ability to interact with multiple specialists to manage each facet of their disease and the Internet can provide a communication channel to enhance this interaction. Successful evaluation tools have been created to effectively measure diabetes management outcomes along multiple dimensions (medical, social, psychological, etc.). Some examples of these tools are the Diabetes Quality of Life Measure (DQOL) developed for use in the Diabetes Control and Complications Trial (DCCT) [19] and the SF-36 [20]. Proactive Outreach Proactive outreach and patient tracking are critical success factors for an Internet-based diabetes management system. Proactive outreach consists of notifications sent to patients to take their medication, visit the health care professional, or simply exercise once a day. The benefit of a proactive approach is well documented in the management of other chronic diseases such as chronic heart failure, where increased compliance and monitoring have resulted in a decrease in the number of hospitalizations for cardiovascular diagnoses and hospital days were reduced from 0.6 to 0.2 (P = .09) per patient per year [21]. Proactive outreach also applies to health care professionals. Reminders to physicians of routine testing for patients can be implemented in an Internet-based diabetes management system. A study determined that the use of a diabetes management system increases the likelihood of physicians ordering lipid-profile testing (19%) and retinal exams for their patients [22]. Feedback The role of the patient has become central in the management of chronic disease; therefore, monitoring must integrate the patient [22]. A crucial aspect of patient integration is feedback. Patients must have the ability to review their medical data at anytime. On-line graphical tools can allow patients to visualize their medical information in much the same way a physician would. Feedback also provides a valuable motivational tool that improves compliance [1] and system usage, both of which are linked to an improved outcome in diabetes management [23]. Information The Internet has always served as a source of health information; 70 million of the 110 million American Internet users have searched the Web for health information in the past year. Currently they can choose from 20,000 health care sites with 1,500 more coming on-line each month [24]. A successful Internet-based diabetes management system should be a source of quality information for the patients who use it. The quality of information on the Internet is a source of great debate. The low barriers to publication on the Internet result in the presence of vast amounts of low-quality and inaccurate information. This misinformation or information that is out of date has the potential of misleading and even harming patients. Consequently, independent agencies such as the Health on the Net Foundation [25] were created to certify the content of medical information on the Internet. Information delivery is based on 2 models: pull and push. Pull Model The pull model relies on the patient retrieving the information he or she seeks. Two pathways are provided to this end. The patient can retrieve documents by navigating through the Web site or can retrieve information with a search engine. Navigation requires a clearly-defined information structure. This is effectively implemented with a hierarchical structure that users can follow to retrieve information of increasing level of detail. Navigation should be facilitated by a clear on-screen indication of the user's location in the information hierarchy. Search engines allow users to search for documents based on keywords. Search engine technology is capable of cataloguing documents based on several criteria. In its simplest form, documents will be catalogued based on their text. Therefore, a search will yield all the documents containing the word that was searched for. However, a successful implementation of a search engine will categorize documents based on several criteria such as topic, author, date, and relevance. Users can then use these criteria to refine their searches. Push Model The push model involves presenting the information to the patient who has opted to receive it. Relevant information could include new research or newly-released drugs for patients who have specified an interest. Interest can be formally expressed by the patient or can be inferred by the system in an effort to personalize the service (see the "Personalization" section of "Evaluation Criteria"). Information delivery in the push model can be implemented in several ways. Patients can be presented with the relevant information upon logging into the system. Alternatively, technologies such as mobile phones and pagers can be used for delivery. A successful Internet-based management system will implement both models of information delivery. Personalization Self-management Plan The management of any chronic disease must be personalized to the individuals, as they are ultimately responsible for its success. Consequently, an Internet-based diabetes management system must allow patients to tailor the intervention to their specific needs. Patients benefit from a proactive approach to their management (in which they are not told what to do) and gain a valuable insight into the management options that may be available to them [17]. Patient involvement and contribution to disease management has demonstrated improved results and compliance [26]. The comprehensive management of diabetes can be based on several models. It is not the purpose of this paper to discuss these management models but rather their successful implementation as Internet-based diabetes management systems. One such model [17] (multilevel social-ecological model for self-management and support for behavior change) was implemented as a physical-activity intervention study [17]. This model is based on the creation of a personal action plan that is the result of both the patient's and health professional's requirements [27]. The creation of a personal action plan can be expressed as these self-management action steps: assessment and feedback, collaborative goal setting, identification of barriers and supports, individualized problem solving, follow-up support, and construction of a personal action plan. Glasgow and Bull have identified the strengths and limitations of interactive technologies such as the Internet for Self-Management Action Steps [17]. Nonetheless, a successful implementation of an Internet-based diabetes management system should provide the patient with the ability to navigate through each action step towards the creation of a personal action plan or the equivalent (depending on the disease-management model used). Language and Ethnicity Piette et al [28] demonstrated that an Automated Telephone Disease Management (ATDM) system produced positive results with an ethnically-diverse diabetic-patient population. Internet-based diabetes systems can reach different ethnicities by offering their services in multiple languages. In some groups where language may be a barrier to medical care, such systems may provide substantial benefits. Inevitably, this opens the discussion of Internet demographics splitting patients between haves and have-nots. This is particularly relevant for type II Diabetes where some minority groups are disproportionately affected and have limited access to the Internet. However, the report from the National Telecommunications and Information Administration indicates a rapid change in Internet demographics that is reflective of the general population of the United States [29]. Communication Communication Between Health Professional and Patient Most efforts in health care technology focus on assisting the doctor in diagnosing and treating a disease. This approach tends to omit a key component of the health care cycle: the patient. The new trend in medicine favors the inclusion of the patient as an integral part of the healing process. A review of 22 studies by Stewart et al [30] indicated a positive effect of communication on actual patient health outcome such as pain, recovery from symptom, anxiety, functional status, and physiologic measures of blood pressure and blood glucose. An Internet-based diabetes management system must be a channel of communication between patients and their health care providers. The communication system can follow 3 models: synchronous, asynchronous, and indirect. Synchronous communication allows the patient and health care provider to communicate directly by using teleconferencing or videoconferencing. Traditionally, these services were in the realm of telemedicine [31] where specific technical equipment was installed to allow the communication to happen. However, the advent of multimedia on the Internet does allow for real-time voice-based and image-based communication. Although at its first steps, synchronous communication can be a valuable part of an Internet-based diabetes management system. Equally, the asynchronous communication model is a crucial part of a management system. Simple solutions such as secure text communication between patient and health care provider can be of great benefit in the management of diabetes. A study at the University of Pittsburgh describes a model of asynchronous communication between doctors and patients that reduced some of the differences in communication in terms of expectations, vocabulary used, and other factors [32]. This study was based on a communication system that allowed patients to familiarize themselves with the relevant domain terms at their own pace. The system also allowed physicians to request more information of patients while providing contextual information. This allowed patients to understand the underlying reasons for the questions. Lastly, the indirect communication model is based on the concept of representation of the health care professional by technology. Such solutions have been implemented using software agents, a form of artificial intelligence that interacts with its environment and reacts to changes. In this case, the agent can interact with the patient and carry out a basic dialogue - and functions as information search and triage [33]. While still experimental, the use of indirect communication in Internet-based diabetes care shows great potential. Community Creation Community creation is based on a many-to-many communication channel compared with the one-to-one communication that occurs between health care professional and patient. Community support is a fundamental aspect of self-management of disease. Diabetes patients benefit from discussing topics that concern management of the disease, anxiety as to what the future holds, and interpersonal and social relationships. The Internet can enable the creation of communities based on the same models of synchronous and asynchronous communication models. One study followed a diabetes chat room for 21 months and found that 79% of all respondents rated participation in the chat as having a positive effect on coping with diabetes [34]. Another study established a chat room for adolescents affected by diabetes and moderated by a diabetologist [35]. The results indicated a decrease in HbA 1cand an improved capacity for self-management. Anonymity undoubtedly favors a greater freedom of expression of individual problems. Community creation and maintenance should be an integral part of any Internet-based management systems. The implementation can be as synchronous chat rooms or as newsgroups where users communicate asynchronously by posting their comments. Further, experts can moderate chat rooms. Technology The complex network of human and machine relations involved in managing diabetes via an Internet-based system has strong implications for the design of such a service. Security One of the main concerns with any medical informatics solution is security and privacy of the data. The success of any Internet-based diabetes management system is reliant on the user's trust that the user's data is secure, private, and confidential. This is possible with the recent availability of strong cryptographic tools used for 2 main purposes: authentication and encryption [23]. Authentication Identification of users is a crucial step in gaining access to the system. Users are granted access to data based on their security profile. For example, only the treating physician can modify a specific patient's blood glucose ranges. Therefore, authentication is both the identification of a user (usually with a combination of username and password) and the enforcement of the security profile. Naturally, user identification is required for more-advanced functions like personalization as mentioned earlier. Encryption All data transmitted between a patient and the system must be secure. Several encryption algorithms exist, with different strengths and speeds. Generally speaking, most Web servers can establish secure communication links using Netscape's Secure Socket Layer (SSL), which is de facto the Internet standard. Recently, 128-bit encryption has been made available worldwide. Any transmission of patient data should be encrypted at the highest level. Usability and User Acceptance Testing usability and user acceptance is a critical part of any computerized system and should be a continuous process during the life of the system. Typically, evaluation instruments have consisted of on-line questionnaires, on-line commenting (e-mail), telephone interviews, video-based testing, and tracking of system usage [36]. Many physicians believe that the key success factor in managing diabetes is simplicity [1]. Consequently, the implementation of an Internet-based diabetes management system should strive towards simplicity for both patient and health care professional. Internet technologies can be a great supplement but if the implementation is not user-friendly, it can become a real barrier [1]. Although the technology has enormous potential, developers should not lose sight of the real purpose of these systems: to collect small amounts of data rapidly and efficiently. Therefore, an Internet-based diabetes management system will only be successful if implemented with a simple user interface used to collect the minimum amount of data from the patient (thus reducing its intrusiveness). Reliability and Availability One of the great advantages of the Internet is that it allows users to access systems anytime and from almost anywhere. This results in a need for systems to always be operational, that is, without downtime. Zero downtime (or close to it) requires fault-tolerant systems. Several technical solutions exist both at the software and hardware level. It is outside the scope of this paper to examine all the solutions; however, it is reasonable to expect an Internet-based diabetes management system to not require downtime for maintenance and to have a fault-tolerant hosting environment. Open Platform Open technologies are based on nonproprietary standards; therefore, a system can be built using technologies from multiple vendors. This is particularly useful for future expansions or medications to accommodate for increased scalability and functionality requirements. An Internet-based diabetes management system should be based on an open platform, particularly for data exchange. Open standards for data representation such as the eXtensible Markup Language (XML) are being adopted by multiple industries. Consequently, a system built using XML will be able to interface with multiple systems and devices. The same system could deliver its services via multiple devices (Internet, mobile phone, handheld computer, etc.) effectively making the Internet open platform the standard. Evaluation of 2 Existing Services To illustrate the use of the evaluation framework, we have applied it to 2 existing Internet-based diabetes management systems: my Diabetes (www.myDiabetes.com) and LifeMasters (www.lifemasters.com). To produce an overall evaluation, a criterion was considered satisfactory if the majority of the raters evaluated it positively (Yes rating). The results of the evaluations were numerically converted by assigning a value of 1 to all positive (Yes) ratings and a value of 0 to all negative (No) ratings. The results of all the evaluations are compiled in Table 3. The agreement level is reported for each individual criterion. This was calculated by dividing the number of ratings consistent with the overall rating (the majority) by the number of raters. For example, if a criterion was rated satisfactory or unsatisfactory by 4 out of the 5 raters, the criterion has an agreement level of 80% (4/5). The technology criteria registered the lowest agreement (60%-80%). The different levels of technical expertise of the evaluators may explain this difference. The Personalization criteria also showed lower levels of agreement between evaluators. This is due to the different interpretations of the criteria between evaluators. Personalization remains a difficult dimension to quantify and evaluate. The quality-of-information agreement levels were also low (60%-80%). Both sites displayed the HON code logo and stated that they subscribed to the HONCode principles. However, neither site was HON registered, although - as of December 14, 2001 - LifeMasters was under review process. The multi-rater kappa for myDiabetes was 0.75 and for LifeMasters was 0.65, indicating a substantial level of agreement as defined by Landis and Koch [37]. There was an important difference between the kappa of MyDiabetes and the kappa of LifeMasters. Further testing is required to clarify the reasons for the difference. Table 3 Evaluation Examples Evaluation Criteria myDiabetes.com (Agreement Level) LifeMasters.com (Agreement Level) Monitoring User defined parameters Health care worker No (100%) Yes (100%) Patient Yes (100%) Yes (100%) User defined parameter ranges Health care worker No (100%) Yes (100%) Patient Yes (100%) Yes (100%) Automated data collection No (100%) No (100%) Alert algorithms to avoid false alarms Entry validation Yes (80%) Yes (100%) Screening of symptoms Yes (100%) Yes (100%) Patient involvement in alert No (100%) No (100%) Multidisciplinary approach Multiple aspects of disease management monitored Physical Yes (100%) Yes (100%) Social Yes (100%). Uses DQOL* Yes (80%). Uses SF-36 Psychological Yes (100%). Uses DQOL Yes (80%). Uses SF-36 Patient access to multiple specialists No (100%) Yes (80%) Proactive outreach Notification to patients Yes (100%) Yes (100%) Notifications to health care professionals No (100%) Yes (100%) Feedback Retrieve and review medical information Yes (100%) Yes (100%) Regular feedback Yes (80%) Yes (80%) Information Quality of information Yes (80%). Uses HON Yes (60%). Uses HON Pull Navigation Yes (80%). Categorized Yes (80%). Categorized Search Yes (100%) Yes (100%) Push Notifications Yes (100%) Yes (100%) Newsletter subscription No (100%) No (100%) Personalization Assessment and feedback Yes (80%) Yes (100%) Collaborative goal setting No (100%) No (80%) Identification of barriers and supports No (100%) Yes (80%) Follow-up support No (100%) Yes (80%) Construction of personalized management plan Yes (80%) Yes (80%) Modification of management plan Yes (100%) Yes (100%) Web site personalization Yes (100%) Yes (100%) Language and ethnicity No (100%) No (100%) Communication Health professional - patient Synchronous No (100%) Yes (80%) Asynchronous No (100%) Yes (80%) Indirect No (100%) No (100%) Community creation Chat rooms Yes (100%) No (100%) Newsgroups / Forums Yes (100%) Yes (100%) Expert moderation Yes (80%) Yes (80%) Technology Security Authentication Yes (100%). User and Password Yes (100%). User and Password Encryption Yes (100%). 128-bit Yes (100%). 128-bit Usability and user acceptance Yes (60%). Tested with forums No (80%). Not actively tested Reliability and availability Netscape compatible Netscape compatible Open architecture No (60%). IIS and ASP No (60%). IIS and ASP Total Positive Results 25 out of 40 32 out of 40 * DQOL = Diabetes Quality of Life Measure Figure 3 Evaluation of myDiabetes.com and LifeMasters.com. The value of each axis is normalized by conversion to a percentage of the maximum score Graphical Representation We believe that a graphical representation of the evaluation results is particularly useful for comparing 2 systems and for determining in which direction the systems should expand their services. To this purpose, a radar graph with the 5 axes representing the 5 dimensions of Monitoring, Information, Personalization, Communication, and Technology is a useful representation. The value of each axis is normalized by conversion to a percentage of the maximum score. The evaluation of myDiabetes.com and LifeMasters.com is represented in Figure 3. The results of the evaluation indicate that LifeMasters is a more-complete solution than myDiabetes in all dimensions - except Information, where both sites were equivalent. This is primarily due to LifeMaster's inclusion of the health care professional in the disease-management cycle. On the other hand, myDiabetes is uniquely interfaced with the patient and is quite good in providing a communication channel for community creation, however, communication with health care professional is lacking, hence the lower score than LifeMasters. Discussion The Internet will undoubtedly change the way we deliver health care services. Chronic disease management, which accounts for 60% of the U.S. medical care costs [38], is a desirable target for the efficiencies of the Internet. Chronic-disease management on the Internet is estimated to have a market potential of US $700 billion [24]. Already we are seeing several Internet-based chronic-disease-management sites arising; however, there is little evidence as to how these solutions are answering the needs of the consumer (the patient). Consumer health informatics research greatly contributes to the health care sector by attempting to systematize and codify consumer's needs, values, and preferences and by trying to build and evaluate information systems that interact directly with consumers and patients [39]. In this paper, we have attempted to catalogue the critical success factors for an Internet-based diabetes management system based on the available literature and the authors' experience. The result is a first step towards a comprehensive evaluation framework. The framework is based on the recognition that the management of diabetes via the Internet is based on several integrated dimensions, namely, Monitoring, Information, Personalization, Communication, and Technology. A successful diabetes management system should efficiently integrate all dimensions. Therefore, the framework provides a model for evaluation and, more importantly, for strategic growth planning for existing sites. For example, a site that is deficient in the communication dimension may enhance its offerings by adding a synchronous chat room. This paper reports an initial evaluation of 2 sites. The results indicate a high-level inter-rater agreement as measured by Cohen's multi-rater kappa. However, this is based on a small sample of evaluations (5). Future research should focus on validation of the framework by consistency between larger samples of raters and on correlation with the success of the multiple sites available today. Key metrics for success include the number of enrolled patients; length of time managed; clinical, economic, and quality-of-life outcomes; and patient-satisfaction measures [24].

Neuroradiology-3-1-2082067

Endovascular occlusion of high-flow intracranial arteriovenous shunts: technical note

Endovascular closure of high-flow arteriovenous (AV) shunts in intracranial AV malformations or pial fistulas is technically challenging. In this paper, we illustrate two simple methods to occlude large high-flow AV shunts in a controlled manner. Introduction Endovascular closure of high-flow arteriovenous (AV) shunts in intracranial AV malformations (AVM) or pial fistulas is technically challenging [1–3]. With the use of liquid embolic agents, the high flow through the shunt may cause migration of the embolic agent to the venous outflow tract with possible devastating consequences. In anatomical configurations with a very short arterial feeder, the shunt can only be occluded at the venous side. In this paper, we illustrate two simple methods to occlude large AV shunts in a controlled manner applied in three patients. Case 1 A 26-year-old man was referred for treatment after haemorrhage of a large callosal AVM (Fig. 1). Angiography demonstrated two high-flow AV shunts from the right anterior cerebral artery draining into enormously enlarged veins. Injection of Onyx 34 at the site of the largest shunt resulted in immediate migration to the venous side. Subsequently, a nondetachable microballoon (Magic B2, Balt, Montmorency, France) was flow-directed into the main feeder just proximal to the shunt and inflated to block the flow. At that time, Onyx injection was continued and resulted in gradual occlusion of the fistula in 7 min. Next, the second high-flow shunt was occluded in a similar way. Control angiography confirmed complete occlusion of both AV shunts. A second embolization procedure was scheduled to be performed 3 months after the first. Fig. 1A 26-year-old man with large callosal AVM. a Frontal and lateral right internal carotid angiogram shows an AVM with two high-flow AV shunts from the anterior cerebral artery (arrows). b Lateral subtracted fluoroscopy image obtained during injection of Onyx with a balloon catheter blocking the flow. The arrow points to the distal tip marker of the microcatheter delivering the Onyx. Onyx is deposited at the transition of artery to vein. c Lateral subtracted fluoroscopy image obtained during occlusion of the second shunt Case 2 A 12-year-old boy was referred for treatment of a left frontal AVM discovered on MRI performed for seizures. Angiography under general anaesthesia demonstrated, besides a plexiform part of the nidus, two large AV shunts from the anterior cerebral artery (Fig. 2). The largest shunt had a very short arterial feeder that precluded the use of Onyx, since reflux would have led to migration into the normal vasculature. A microcatheter was navigated through the shunt into the proximal venous part and detachable coils were inserted until the shunt was completely occluded. Next, the second AV shunt was occluded with simple injection of Onyx 34 and continued injection occluded the major part of the nidus. After embolization, still under general anaesthesia, the patient was transferred to the Gamma Knife unit and the small AVM remnant was irradiated. Fig. 2A 12-year-old boy with a left frontal AVM. a Lateral angiogram of the left internal carotid artery shows a large AV shunt with a very short arterial segment (arrow). b Lateral angiogram obtained after occlusion of the venous and arterial side of the shunt with detachable coils Case 3 A 27-year-old woman was referred for treatment of an intracranial vascular disorder that had been discovered on CT scan performed for chronic headaches. Vertebral angiography revealed a high-flow single-hole side-wall pial fistula from the P3–P4 segment of the right posterior cerebral artery draining into a dilated basal vein of Rosenthal (Fig. 3). In addition, a flow aneurysm on the right superior cerebellar artery was present. A microcatheter was navigated through the shunt into the vein and the proximal part of the vein was occluded with detachable coils with preservation of continuity of the posterior cerebral artery. Subsequently, the flow aneurysm was occluded with coils. A follow-up angiogram obtained 6 months later showed continuing closure of the fistula and the aneurysm. Fig. 3A 27-year-old woman with pial fistula. a Frontal vertebral angiogram shows a high-flow single-hole side-wall pial fistula from the P3–P4 segment of the right posterior cerebral artery draining into a dilated basal vein of Rosenthal. In addition, a flow aneurysm of the superior cerebellar artery is present (arrow). b Follow-up angiogram obtained 6 months after occlusion of the venous side of the fistula with detachable coils shows continuing occlusion of the fistula with preserved patency of the posterior cerebral artery Discussion The most commonly used embolic agent in brain AVMs is the fast polymerizing liquid adhesive n-butylcyanoacrylate (NBCA, Histoacryl; Braun, Melsungen, Germany). Pure NBCA polymerizes immediately after contact with blood. Slower polymerization is achieved when NBCA is mixed with iodinated oil (Lipiodol; Guerbet, Roissy, France). The use of a high concentration of NBCA or pure NBCA in high-flow shunts in brain AVMs may occlude the fistula within seconds, but this technique requires courage, experience and skill, since flow of NBCA is to some extent unpredictable [3]. Inadvertent migration of the glue into the draining veins may result in immediate haemorrhage by blocking venous outflow. Detachable coils may be inserted at the fistula site to decrease the flow and thereby facilitate the injection of NBCA [2]. Recently, the new liquid embolic agent Onyx has become available for embolization of brain AVMs (Onyx Liquid Embolic System; Microtherapeutics, Irvine, Calif.). Onyx is nonadhesive and polymerizes slowly. Onyx is available in several concentrations and the high-concentrated Onyx can be used to slowly occlude large AV shunts or pial AV fistulas in a more controlled way than that achieved with NBCA [4]. However, simple Onyx injection is not always feasible: in some very high-flow shunts Onyx may migrate through the fistula into the distal draining veins. We found the adjuvant use of a microballoon to block the flow to enable gradual occlusion of the large high-flow shunts with Onyx in the AVM of patient 1 a very useful approach. In the other patients, one with a very short arterial feeder of an AV shunt in an AVM and one with a side-wall pial fistula, the use of liquid embolic agents was not possible and controlled delivery of detachable coils to the venous side of the shunt was technically easy with complete occlusion of the fistula. This technique is helpful in shunts with a moderately dilated venous outflow tract in which placement of coils is possible. In conclusion, different types of high-flow AV shunts can safely be occluded with endovascular techniques tailored to the specific anatomical configuration of the shunt.

J_Biol_Inorg_Chem-3-1-2039866

Zinc complexes of the biomimetic N,N,O ligand family of substituted 3,3-bis(1-alkylimidazol-2-yl)propionates: the formation of oxalate from pyruvate

The coordination chemistry of the 2-His-1-carboxylate facial triad mimics 3,3-bis(1-methylimidazol-2-yl)propionate (MIm2Pr) and 3,3-bis(1-ethyl-4-isopropylimidazol-2-yl) propionate (iPrEtIm2Pr) towards ZnCl2 was studied both in solution and in the solid state. Different coordination modes were found depending both on the stoichiometry and on the ligand that was employed. In the 2:1 ligand-to-metal complex [Zn(MIm2Pr)2], the ligand coordinates in a tridentate, tripodal N,N,O fashion similar to the 2-His-1-carboxylate facial triad. However, the 1:1 ligand-to-metal complexes [Zn(MIm2Pr)Cl(H2O)] and [Zn(iPrEtIm2Pr)Cl] were crystallographically characterized and found to be polymeric in nature. A new, bridging coordination mode of the ligands was observed in both structures comprising N,N-bidentate coordination of the ligand to one zinc atom and O-monodentate coordination to a zinc second atom. A rather unique transformation of pyruvate into oxalate was found with [Zn(MIm2Pr)Cl], which resulted in the isolation of the new, oxalato bridged zinc coordination polymer [Zn2(MIm2Pr)2(ox)]·6H2O, the structure of which was established by X-ray crystal structure determination. Introduction The 2-His-1-carboxylate facial triad is a structural motif of increasing prominence among structurally characterized metalloenzymes. Next to the rapidly growing subgroup of non-heme iron enzymes [1, 2], several mononuclear zinc enzymes also feature this triad at their active site [3]. The proteases thermolysin, carboxypeptidase, and neutral protease, for example, catalyze the hydrolysis of peptide bonds. The zinc ion in these enzymes is coordinated by three endogenous residues, i.e., one glutamate and two histidine residues (Fig. 1). The fourth coordination site of the pseudo-tetrahedral zinc center is occupied by a catalytically important water or hydroxide ligand. This 2-His-1-carboxylate facial triad is a variation of the most commonly observed structural motif for the active sites of zinc-containing enzymes. Typically, the tetrahedral zinc centers in these enzymes are bound by a combination of histidine (N), glutamate or aspartate (O) and/or cysteine (S) residues. The resulting NxOySz donor set determines its biological function [3]. Accordingly, it is important to understand how a particular combination of donor groups modulates the reactivity of the Zn(II) metal center. Fig. 1Active site of a zinc-containing enzyme that features the 2-His-1-carboxylate facial triad: neutral protease from Bacillus cereus (Protein Data Bank accession code 1NPC.pdb) [4] The study of small synthetic analogues has focused on establishing a basis for such a structure–function relationship for these mononuclear zinc-containing enzymes [3, 5]. Initially, pioneering studies by the groups of Vahrenkamp, Parkin, Kitajima, and others employed all-N donor ligands such as the trispyrazolylborates [6–8]. In order to obtain close structural models for the 2-His-1-carboxylate facial triad containing zinc enzymes, one would preferentially like to construct small analogues based on a tripodal N,N,O ligand framework that incorporates the biologically relevant donor groups. Of the complexes with a mixed N/O donor set reported to date [9–22], only a few make use of such tridentate, tripodal N,N,Ocarboxylato ligands. These tripodal ligands should be particularly well suited for the structural modeling of tetrahedral metal centers [23]. Dowling and Parkin [24] and Ghosh and Parkin [25] reported the first example of an N,N,O model complex with a carboxylato donor group, which resulted from the insertion of carbon dioxide into a borohydride bond. More recently, the groups of Burzlaff [9, 12] and Carrano [11, 18, 19] have reported studies on the zinc coordination chemistry of the bispyrazolylacetate ligand family. Beck et al. [9], for instance, reported that a 2:1 ligand–zinc complex was obtained with bis(3,5-dimethylpyrazol-2-yl)acetate (bdmpza), a ligand closely related to 3,3-bis(1-methylimidazol-2-yl)propionate (MIm2Pr). These 2:1 complexes were obtained regardless of the ratio of reactants, i.e., even with stoichiometric amounts of ligand and zinc(II). Very recently, Friese et al. [22] constructed mononuclear zinc models via a new approach, i.e., by the use of sterically hindered carboxylato and N-donor ligand building blocks. We have been studying the copper and iron coordination chemistry of the new ligand family of the substituted 3,3-bis(1-alkylimidazol-2-yl)propionates as accurate structural mimics of the 2-His-1-carboxylate facial triad [26–28]. These ligands incorporate the biologically relevant donor groups, i.e., two imidazoles and a carboxylate group, into a monoanionic, tripodal framework. Their general biomimetic potential has been illustrated by the structurally characterized copper [27, 28] and iron [26] complexes of these ligands. Here, our studies are expanded to the structural modeling of the zinc-containing enzymes and describe the zinc coordination chemistry of the ligands MIm2Pr and 3,3-bis(1-ethyl-4-isopropylimidazol-2-yl)propionate (iPrEtIm2Pr) (Fig. 2). Furthermore, the attempted synthesis of zinc complexes with 3,3-bis(1-methylimidazol-2-yl)acetate (MIm2Ac), a direct analogue of the bispyrazolylacetates, is described. During the course of these studies, we found that zinc complexes of ligand MIm2Pr catalyze the unexpected and unprecedented conversion of pyruvate to oxalate in aqueous solution.Fig. 2Ligands 3,3-bis(1-methylimidazol-2-yl)propionate (MIm2Pr), 3,3-bis(1-ethyl-4-isopropylimidazol-2-yl)propionate (iPrEtIm2Pr), and 3,3-bis(1-methylimidazol-2-yl)acetate (MIm2Ac) and the synthesis of the zinc complexes [Zn(MIm2Pr)2], [Zn(MIm2Pr)Cl], and [Zn(iPrEtIm2Pr)Cl] Materials and methods Air-sensitive organic reactions were carried out under an atmosphere of dry, oxygen-free N2 using standard Schlenk techniques. Tetrahydrofuran (THF) and diethyl ether were dried over sodium benzophenone ketyl and distilled under N2 prior to use. Methanol was dried over magnesium methoxide and distilled under N2 prior to use. 1H and 13C NMR spectra were recorded using a Varian AS400, a Varian Inova 300, or a Varian Mercury 200 spectrometer, operating at 25 °C. IR spectra were recorded with a PerkinElmer Spectrum One Fourier transform IR instrument. Elemental microanalyses were carried out by the Microanalytisches Laboratorium Dornis & Kolbe, Mulheim a.d. Ruhr, Germany. ESI–MS spectra were recorded using a Micromass liquid chromatography time-of-flight mass spectrometer at the Biomolecular Mass Spectrometry Group, Utrecht University. Bis(1-methylimidazol-2-yl)methane (MIm2CH2) [29,] 3,3-bis(1-methylimidazol-2-yl)propionic acid (HMIm2Pr), potassium and tetrabutylammonium 3,3-bis(1-methylimidazol-2-yl)propionate (K[MIm2Pr] and [Bu4N][MIm2Pr]) [27], potassium 3,3-bis(1-ethyl-4-isopropylimidazol-2-yl)propionate (K[iPrEtIm2Pr]) [26], and zinc bis(trimethylsilyl)amide [30] were prepared according to published procedures. All other chemicals were obtained commercially and used as received. Benzyl bis(1-methylimidazol-2-yl)acetate To a cooled solution of MIm2CH2 (0.52 g, 3.0 mmol) in dry THF (50 mL) at −78 °C was added dropwise a solution of n-butyl lithium in hexane (2.0 mL, 3.2 mmol). The reaction mixture was stirred for 1 h at −78 °C, after which benzylchloroformate (1.19 mL, 3.5 mmol, 50 wt% solution in toluene) was added dropwise. During the addition, a white precipitate formed. The solution was allowed to warm to room temperature overnight. The reaction mixture was quenched with H2O (20 mL) and all volatiles were removed in vacuo. The aqueous layer was extracted with diethyl ether (4 × 20 mL) and the combined organic layers were dried over MgSO4, filtered and concentrated in vacuo. The crude oil was purified by column chromatography (SiO2, ethyl acetate–methanol 2:1). Benzyl bis(1-methylimidazol-2-yl)acetate (BnMIm2Ac) was isolated as a yellow oil in 62% yield (0.58 g, 1.9 mmol). 1H NMR (300 MHz, CD3CN, 25 °C): δ = 3.44 (s, 6H, NCH3), 5.23 (s, 2H, CH2), 5.62 (s, 1H, CH), 6.85 (s, 2H, Him), 6.96 (s, 2H, Him), 7.35 (s, 5H, HPh) ppm. Anal. for C17H18N4O2 (310.35): calc. C 65.79, H 5.85, N 18.05; found C 65.88, H 5.74, N 17.92. IR (solid): ν = 3,122.9, 2,911.5, 1,734.5, 1,547.4, 1,510.2, 1,454.4, 1,373.6, 1,263.4, 1,220.8, 1,150.4, 978.2, 958.3, 907.1, 847.8, 785.7, 776.5, 740.1, 696.7, 683.5, 665.3 cm−1. ESI–MS: m/z = 409.0 ([M–Cl]+, calc. 409.0). [Zn(MIm2Pr)2] To a hot, colorless solution of HMIm2Pr (282 mg, 1.20 mmol) in dry acetonitrile (100 mL) was added a solution of Zn[N(SiMe3)2]2 (232 mg, 0.60 mmol) in dry diethyl ether (20 mL) via a cannula. The resulting colorless, clear solution was stirred for 60 h at elevated temperature, during which gradually a white precipitate formed. The reaction mixture was concentrated in vacuo to give the crude product as a white powder. Recrystallization from an acetone–water mixture (4:1 v/v) gave [Zn(MIm2Pr)2] as a colorless crystalline solid (230 mg, 72%). 1H NMR (300 MHz, D2O, 25 °C): δ = 2.69 (d, 2H, J = 6.3 Hz, CHCH2), 3.74 (s, 6H, NCH3), 4.93 (t, 1H, J = 6.0 Hz, CHCH2), 6.69 (s, 2H, Him), 7.00 (s, 2H, Him) ppm. 13C{1H} NMR (50 MHz, D2O, 25 °C): δ = 30.2, 33.4, 43.6, 122.7, 125.7, 146.8, 177.2 ppm. Anal. for C22H26ZnN8O4 (531.90): calc. C 49.68, H 4.93, N 21.07; found C 49.76, H 5.06, N 20.97. IR (solid): ν = 3,120.2, 2,951.7, 2,907.1, 2,815.2, 1,580.5, 1,508.1, 1,427.1, 1,391.0, 1,306.4, 1,288.0, 1,229.3, 1,168.8, 1,143.3, 1,044.8, 956.9, 906.2, 769.3, 751.1 cm−1. ESI–MS: m/z = 531.18 ([M + H]+, calc. 531.14), 553.18 ([M+Na]+, calc. 553.13), 569.16 ([M + K]+, calc. 569.10). [Zn(MIm2Pr)Cl] To a hot, colorless solution of K[MIm2Pr] (255 mg, 0.93 mmol) in dry methanol (5 mL) was added a solution of ZnCl2 (127 mg, 0.93 mmol) in methanol (5 mL) via a cannula. Immediately upon addition a white precipitate formed and the suspension was stirred at elevated temperature for 1 h. The white precipitate was separated by centrifugation and washed with methanol (3 × 20 mL). [Zn(MIm2Pr)Cl] was obtained as a white powder in almost quantitative yield (305 mg, 98%). 1H NMR (300 MHz, D2O, 25 °C): δ = 2.80 (d, 3H, J = 3.3 Hz, CHCH2), 3.85 (s, 6H, NCH3), 4.99 (t, 1H, CH2CH), 6.99 (s, 2H, Him), 7.14 (s, 2H, Him) ppm. 13C{1H} NMR (50 MHz, D2O, 25 °C): δ = 29.8, 33.4, 42.6, 122.5, 125.5, 146.1, 177.3 ppm. Anal. for C11H13ClN4O2Zn (334.11): calc. C 39.55, H 3.92, N 16.77; found: C 39.46, H 4.05, N 16.85. IR (solid): ν = 3,138.0, 3,117.2, 1,614.8, 1,508.2, 1389.8, 1,299.8, 1,289.5, 1,151.7, 1,142.5, 1,089.7, 976.2, 949.7, 796.3, 765.3, 736.1 cm−1. ESI–MS: m/z = 296.94 ([M–Cl]+, calc. 297.03), 332.92 ([M + H]+, calc. 333.01), 413.97 ([3M–Zn–3Cl + H]2+, calc. 414.09), 531.02 ([2M–Zn–2Cl + H]+, calc. 531.15), 628.97 ([2M–Cl]+, calc. 629.03), 827.10 ([3M–Zn–3Cl]+, calc. 827.17). [Zn(MIm2Pr)Cl(H2O)] Colorless crystals of [Zn(MIm2Pr)Cl(H2O)] suitable for X-ray diffraction were obtained upon standing of a concentrated solution of [Zn(MIm2Pr)Cl] in H2O for several weeks. The coordination polymer obtained is insoluble in all common organic solvents and water of neutral pH. Anal. for C11H15ClN4O3Zn (352.12): calc. C 37.52, H 4.29, 15.91; found C 37.44, H 4.27, N 15.97. IR (solid): ν = 3,408.6, 3,113.5, 3,134.7, 1,594.5, 1,504.5, 1,403.1, 1,306.7, 1,255.5, 1,155.5, 1,136.3, 1,087.4, 983.7, 959.1, 934.9, 849.7, 764.6, 743.1, 717.6 cm−1. [Zn(iPrEtIm2Pr)Cl] To a colorless solution of ZnCl2 (30 mg, 0.22 mmol) in dry methanol (10 mL) was added a solution of K[iPrEtIm2Pr] (85 mg, 0.22 mmol) in dry methanol (15 mL) via a cannula. The clear solution was stirred overnight at room temperature and evaporated in vacuo. The product was redissolved in dichloromethane and insoluble KCl was separated off by centrifugation. The solution was filtered over Celite and concentrated in vacuo to give [Zn(iPrEtIm2Pr)Cl] as a white powder (97 mg, 99%). Colorless crystals suitable for X-ray diffraction were obtained from a methanolic solution of [Zn(iPrEtIm2Pr)Cl] upon standing. 1H NMR (300 MHz, D2O, 25 °C): δ = 1.22 (d, 6H, J = 7.2 Hz, CH3CHCH3), 1.25 (d, 6H, J = 6.8 Hz, CH3CHCH3), 1.44 (t, 6H, J = 7.2 Hz, CH3CH2), 2.88 (d, 2H, J = 6.0 Hz, CHCH2), 3.25 (m, 2H, CH3CHCH3), 4.21 (ABX3, 2H, J = 7.2, CHHCH3), 4.32 (ABX3, 2H, J = 7.2, CHHCH3), 5.02 (t, 1H, J = 5.7 Hz, CHCH2), 6.99 (s, 2H, Him) ppm. 13C{1H} NMR (100 MHz, CD3OD, 25 °C): δ = 16.6, 22.1, 24.1, 28.0, 31.1, 42.9, 116.0, 146.2, 149.9, 175.7 ppm. Anal. for C19H29ClN4O2Zn (446.32): calc. C 51.13, H 6.55, N 12.55; found: C 51.05, H 6.43, N 12.37. IR (solid): ν = 3,101.4, 2,967.9, 2,873.7, 1,626.6, 1,572.8, 1,494.8, 1,448.9, 1,464.1, 1,381.1, 1,324.8, 1,257.2, 1,175.8, 1,153.0, 1,027.6, 976.1, 801.2, 757.7, 660.5 cm−1. ESI–MS: m/z = 409.13 ([2M–2Cl]2+, calc. 409.16), 445.11 ([M + H]+, calc. 445.13), 483.04 ([M + K]+, calc. 483.09), 631.21 ([3M–2Cl]2+, calc. 631.22), 853.29 ([2M–Cl]+, calc. 853.29), 889.20 ([2M + H]+, calc. 889.26), 1297.02 ([3M + H]+, calc. 1297.42). [Zn(BnMIm2Ac)Cl2] To a hot solution of BnMIm2Ac (259 mg, 0.84 mmol) in dry methanol (5 mL) was added a solution of anhydrous ZnCl2 (115 mg, 0.84 mmol) in dry MeOH (5 mL) via a cannula. The reaction mixture was stirred at elevated temperature for 1 h and gradually the product formed as a white precipitate. The product [Zn(BnMIm2Ac)Cl2] was separated by centrifugation and was obtained as a white powder in quantitative yield. Single crystals suitable for X-ray analysis were obtained by slow evaporation of a solution of [Zn(BnMIm2Ac)Cl2] in acetone. 1H NMR (300 MHz, CD3CN, 25 °C): δ = 3.79 (2, 6H, CH3), 5.12 (s, 2H, CH2), 5.60 (s, 1H, CH), 7.13 (d, 2H, J = 0.75 Hz, Him), 7.22 (d, 2H, J = 0.75 Hz, Him), 7.25 (m, 2H, HPh), 7.33 (m, 3H, HPh) ppm. 13C{1H} NMR (100 MHz, CD3OD, 25 °C): δ = 34.9, 40.6, 69.7, 124.5, 126.5, 129.1, 129.5, 129.7, 136.1, 142.3, 166.2 ppm. Anal. for C17H18Cl2N4O2Zn (446.67): calc. C 45.71, H 4.06, N 12.54; found C 45.64, H 3.95, N 12.61. IR (solid): ν = 3,122.9, 2,911.5, 1,734.5, 1,547.4, 1,510.2, 1,454.4, 1,373.6, 1,263.4, 1,220.8, 1,150.4, 978.2, 958.3, 907.1, 847.8, 785.7, 776.5, 740.1, 696.7, 683.5, 665.3 cm−1. ESI–MS: m/z = 409.0 ([M–Cl]+, calc. 409.0). [Zn2(MIm2Pr)2(ox)] (direct synthesis) To a solution of [Zn(MIm2Pr)Cl] (50 mg, 0.15 mmol) in water (5 mL) was added a solution of potassium oxalate hydrate (14 mg, 0.075 mmol) in water (3 mL). The solution was stirred overnight at room temperature, during which gradually a white precipitate formed. The precipitate was separated by centrifugation and washed three times with water (3 × 20 mL). The product [Zn2(MIm2Pr)2(ox)] was dried in vacuo and obtained as a white powder in 82% yield (42 mg, 0.061 mmol). Anal. for C24H26N8O8Zn2·6H2O (685.33): calc. C 36.33, H 4.83, N 14.12; found: C 36.45, H 4.50, N 14.25. IR (solid): ν = 3,476.8, 3,135.4, 1,671.3, 1,651.6, 1,600.2, 1,544.9, 1,504.8, 1,423.6, 1,382.7, 1,320.1, 1,287.2, 1,158.1, 1,137.9, 985.1, 769.4, 745.1, 710.4 cm−1. [Zn2(MIm2Pr)2(ox)]·6H2O, reaction of [Zn(MIm2Pr)Cl] with sodium pyruvate For NMR investigation of reaction mixtures, stock solutions of [Zn(-MIm2Pr)Cl] (20 mg, 0.06 mmol) in H2O (1 mL) and sodium pyruvate (13.2 mg, 0.06 mmol) in H2O (2 mL) were prepared. NMR tubes were filled with 0.1 mL of each stock solution and kept either at room temperature or heated to 50 °C. The NMR samples were analyzed at different times after mixing the reagents. Before acquisition of the NMR spectrum, 0.4 mL D2O was added to the reaction mixture. Crystals usually deposited within 2 or 3 days in the NMR tube. In the case of the catalytic runs, 0.1 mL of a solution containing 20 equiv of sodium pyruvate was added to 0.1 mL of the [Zn(MIm2Pr)Cl] stock solution and dimethyl sulfoxide (5 equiv, 11 μL) was added as an internal standard. The reaction mixture was kept at 50 °C and the NMR spectrum was recorded with addition of 0.4 mL D2O after 24 h. On a preparative scale, to a solution of ZnCl2 (37 mg, 0.27 mmol) and K[MIm2Pr] (75 mg, 0.27 mmol) in H2O (10 mL) was added sodium pyruvate (30 mg, 0.27 mmol). The solution was heated to 50 °C and gradually an off-white precipitate was formed. After 1 week, the product was separated by centrifugation (29 mg, 31%). The IR spectrum of the product was identical to that of independently synthesized [Zn2(MIm2Pr)2(ox)]. X-ray crystal structure determinations of [Zn(MIm2Pr)Cl(H2O)], [Zn(iPrEtIm2Pr)Cl], ZnCl2(BnMIm2Ac)], [ZnCl2(MIm2CH2)], and [Zn2(MIm2Pr)2(ox)]·6H2O Reflections were measured with a Nonius Kappa CCD diffractometer with a rotating anode (graphite monochromator, λ = 0.71073 Å) at −123 °C. The structures were refined with SHELXL-97 [31] against F2 of all reflections. Non-hydrogen atoms were refined with anisotropic displacement parameters. Geometry calculations and checking for higher symmetry were performed with the PLATON program [32]. Further details are given in Table 1. Table 1Crystallographic data for compounds [Zn(MIm2Pr)Cl(H2O)], [Zn(iPrEtIm2Pr)Cl], ZnCl2(BnMIm2Ac)], [ZnCl2(MIm2CH2)], and [Zn2(MIm2Pr)2(ox)]·6H2OCompound[Zn(MIm2Pr)Cl(H2O)][Zn(iPrEtIm2Pr)Cl] [ZnCl2(BnMIm2Ac)][ZnCl2(MIm2CH2)][Zn2(MIm2Pr)2(ox)]·6H2O FormulaC11H15ClN4O3ZnC19H29ClN4O2ZnC17H18Cl2N4O2ZnC9H12Cl2N4ZnC24H26N8O8Zn2·6H2OFW352.09446.28446.62312.50793.36Crystal size (mm3)0.30 × 0.12 × 0.090.18 × 0.03 × 0.030.60 × 0.03 × 0.030.30 × 0.15 × 0.060.24 × 0.18 × 0.09Crystal colorColorlessColorlessColorlessColorlessColorlessCrystal systemOrthorhombicMonoclinicMonoclinicMonoclinicTriclinicSpace groupP212121 (no. 19)Cc (no. 9)C2/c (no. 15)P21/c (no. 14)a (Å)8.7159(1)10.1766(2)35.5761(4)7.3547(1)9.9634(7)b (Å)10.1124(1)23.8561(6)7.4531(1)14.9538(2)12.0617(11)c (Å)15.1177(2)9.1995(2)14.6106(2)12.3489(2)14.8889(9)α (°)––––113.021(3)β (°)–107.0078(14)101.8858(5)117.5968(6)90.365(3)γ (°)––––102.513(3)V (Å)1332.45(3)2135.72(8)3790.97(8)1203.62(3)1599.4(2)Z44842Dcalc. (g cm−3)1.7551.3881.5651.7241.647μ (mm−1)2.0561.2961.5972.4611.580Absorption correctionMultiscanMultiscanMultiscanMultiscanMultiscanAbsorption correction range0.71–0.840.86–0.960.85–0.960.74–0.860.64–0.87sin(θ/λ)max (Å−1)0.650.650.600.650.65Reflections (measured/unique)21,618/3,06311,758/4,64428,566/3,42427,341/2,75935,709/7,342Parameters/restraints241/0251/2237/0193/0438/0R1/wR2 [I > 2σ(I)]0.0182/0.04470.0312/0.06940.0309/0.07040.0306/0.07930.0389/0.0867R1/wR2 (all reflections)0.0190/0.04520.0384/0.07230.0512/0.07860.0389/0.08520.0540/0.0947S1.0621.0411.0811.0891.083Flack parameter [37]−0.006(7)0.305(9)–––Residual density (e Å−3)−0.36/0.19−0.27/0.41−0.31/0.43−0.54/1.23−0.79/0.51MIm2Pr 3,3-bis(1-methylimidazol-2-yl)propionate, iPrEtIm2Pr 3,3-bis(1-ethyl-4-isopropylimidazol-2-yl)propionate, and BnMIm2Ac benzyl bis(1-methylimidazol-2-yl)acetate, 22, bis(1-methylimidazol-2-yl)methane For Zn(MIm2Pr)Cl(H2O)] the structure was solved with SHELXS-97 [33] using direct methods. All hydrogen atoms were located in the difference Fourier map and refined freely with isotropic displacement parameters. For [Zn(iPrEtIm2Pr)Cl] the structure was solved with DIRDIF-99 [34] using automated Patterson methods. Refinement was performed as an inversion twin. All hydrogen atoms were located in the difference Fourier map and refined with a riding model. For ZnCl2(BnMIm2Ac)] the structure was solved with DIRDIF-99 [34] using automated Patterson methods. All hydrogen atoms were located in the difference Fourier map and refined with a riding model. For [ZnCl2(MIm2CH2)] the structure was solved with SIR-97 [35] using Direct methods. All hydrogen atoms were located in the difference Fourier map and refined freely with isotropic displacement parameters. For [Zn2(MIm2Pr)2(ox)]·6H2O the crystal was nonmerohedrally twinned with a twofold rotation about the crystallographic c-axis as a twin operation. This twin law was taken into account during the intensity integration using the program EvalCCD [36]. The structure was solved on the nonoverlapping reflections with DIRDIF-99 [34] using automated Patterson methods. All hydrogen atoms were located in the difference Fourier map and refined with a riding model. The twin fraction refined to 0.127(3). Crystallographic data (without structure factors) for the structures reported in this paper have been deposited with the Cambridge Crystallographic Data Centre (CCDC) as supplementary publication nos. CCDC-637452 for Zn(MIm2Pr)Cl(H2O)], CCDC-637453 for [Zn(iPrEtIm2Pr)Cl], CCDC-637454 for ZnCl2(BnMIm2Ac)], CCDC-637455 for [ZnCl2(MIm2CH2)], and CCDC-637456 for [Zn2(MIm2Pr)2(ox)]·6H2O. Copies of the data can be obtained free of charge from the CCDC (12 Union Road, Cambridge CB2 1EZ, UK; Tel.: +44-1223-336408; Fax: +44-1223-336003; e-mail: deposit@ccdc.cam.ac.uk; Web site http://www.ccdc.cam.ac.uk). Results As part of our efforts to study the coordination chemistry of the substituted 3,3-bis(1-alkylimidazol-2-yl)propionate ligand family, the complexes of MIm2Pr and iPrEtIm2Pr with ZnCl2 were synthesized and structurally characterized (Fig. 2). MIm2Pr is the parent ligand of the family and iPrEtIm2Pr offers the most steric hindrance of the ligands studied so far. The coordination chemistry of MIm2Pr and iPrEtIm2Pr with zinc was found to be stoichiometry-dependent. Zinc complexes of MIm2Pr The formation of different species upon addition of either 0.5 or 1.0 equiv of ZnCl2 to a solution of the potassium salt of MIm2Pr in D2O is evidenced by the respective 1H NMR spectra in D2O (Fig. 3). Fig. 3400 MHz 1H NMR spectra of the titration of K[MIm2Pr] with ZnCl2 in D2O at room temperature: a K[MIm2Pr]; b K[MIm2Pr] and ZnCl2 2:1; c K[MIm2Pr] and ZnCl2 1:1. Asterisks denote the residual solvent peak All four resonances associated with ligand MIm2Pr (spectrum a) shifted upon addition of 0.5 equiv of ZnCl2 (spectrum b). The observation of two sharp signals of equal intensity for the imidazole protons shows that all four imidazole groups are equivalent. This is indicative of symmetric binding of both ligands to the zinc ion. The resonances of the methyl groups and the methine proton at the bridging carbon shift downfield relative to the free ligand, whereas the propionate CH2 protons shift upfield. Note that the methine proton resonance of the free ligand is hidden under the residual solvent signal in spectrum a. The lack of any other resonances suggests the clean formation of a neutral, 2:1 [Zn(MIm2Pr)2] complex. Upon addition of another 0.5 equiv of ZnCl2 several resonances shift and signal broadening is observed (spectrum c). Especially, one of the two imidazole signals is severely broadened. Again, no free ligand is observed in the spectrum. These observations indicate a dynamic system with an overall 1:1 ligand–zinc stoichiometry. The isolation and purification of the complex formed with 2 equiv of MIm2Pr with respect to ZnCl2 proved troublesome. The separation of the inorganic salts formed as a by-product in the synthesis by means of repetitive washing in all cases resulted in partial loss of coordinated ligand (as observed by NMR), regardless of the A[MIm2Pr] salt (A is K+, [Bu4N]+) originally employed. We therefore attempted the synthesis of this complex by the reaction of 2 equiv of the acid HMIm2Pr with Zn[N(SiMe3)2]2. Zinc bis(trimethylsilyl)amide can be synthesized in sizeable quantities [30] and was found to be a convenient reagent in this case. Indeed, the reaction of Zn[N(SiMe3)2]2 with 2 equiv of HMIm2Pr in acetonitrile resulted in the formation of a white precipitate, which was recrystallized from an acetonitrile–water solution and was identified as the 2:1 complex [Zn(MIm2Pr)2]. Unfortunately, all attempts at obtaining single crystals of [Zn(MIm2Pr)2] failed. On the basis of ESI–MS, IR, and the previously described NMR analysis, we propose that [Zn(MIm2Pr)2] (Fig. 4) is isostructural to the crystallographically characterized analogous complex [CuII(MIm2Pr)2] [27]. The position of the symmetric and asymmetric carbonyl stretching vibrations proved to be indicative of the binding mode of MIm2Pr (Bruijnincx et al., submitted) [27]. The νas(C=O) and νs(C=O) are found at 1,581 and 1,391 cm−1, respectively, indicative of a monodentate binding mode of the carboxylate [Δ(νas − νs) = 190 cm−1, identical to Δionic] [38]. Furthermore, the clean ESI–MS spectrum of [Zn(MIm2Pr)2] confirmed the structure of a mononuclear, neutral 2:1 species with [M + X]+ (X is H, Na, K) as the prominent ions. Fig. 4The 2:1 complex [Zn(MIm2Pr)2] The 1:1 complex [Zn(MIm2Pr)Cl)] was synthesized by the reaction of equimolar amounts of ZnCl2 and K[MIm2Pr] in dry methanol. This led to the precipitation and isolation of [Zn(MIm2Pr)Cl] in almost quantitative yield. The broadened signals in the 1H NMR spectrum of [Zn(MIm2Pr)Cl)] in D2O prompted us to study this complex by variable-temperature NMR (Fig. 5). As complex [Zn(MIm2Pr)Cl)] is insoluble in organic solvents, the variable-temperature NMR studies were limited to aqueous solutions, which restricted the temperature range that could be studied. Spectra were recorded in the range from 2 to 70 °C. Fig. 51H NMR resonances of the imidazole rings of [Zn(MIm2Pr)Cl] at various temperatures in D2O Figure 5 shows that at elevated temperatures two sharp imidazole resonances are observed. In the spectrum recorded at 70 °C, the (small) coupling between the two imidazole protons H1 and H2 can also be discerned. The imidazole signal at higher field broadens upon cooling and ultimately splits into two separate (broad) resonances. This implies that at higher temperature the imidazole H1 protons (Fig. 3) become magnetically equivalent, whereas at lower temperature exchange slows down with the result that decoalesence into two broad resonances occurs. This behavior can be tentatively explained by the possibility of two different spatial orientations of the imidazole rings with respect to the carboxylate group of a second ligand upon coordination to zinc (vide infra), which would result in nonequivalent imidazole H1 protons at lower temperature. At higher temperatures these orientations average on the NMR timescale and only one signal is observed. The asymmetric and symmetric carbonyl stretch vibrations of isolated [Zn(MIm2Pr)Cl] were observed at 1,615 and 1,390 cm−1 [Δ(νas − νs) = 225 cm−1], respectively. The asymmetric stretch thus shifted to higher wavenumbers compared with that observed for [Zn(MIm2Pr)Cl] and a different, but still monodentate coordination mode of the carboxylate must therefore be present in [Zn(MIm2Pr)Cl]. The ESI–MS spectrum of [Zn(MIm2Pr)Cl] showed the presence of several different species, with ions corresponding to, e.g., [MIm2Pr + Zn]+, [2(MIm2Pr) + Zn + H]+, [2(MIm2Pr) + 2Zn + Cl]+, [3(MIm2Pr) +2Zn + H]2+ and [3(MIm2Pr) + 2Zn]+. On the basis of the NMR, IR, and ESI–MS data, the structure of [Zn(MIm2Pr)Cl] can be described as an oligomeric/polymeric species both in the solid state and in solution. This was confirmed by the single-crystal structure determination of [Zn(MIm2Pr)Cl(H2O)]. Crystal structure of [Zn(MIm2Pr)Cl(H2O)] Colorless single crystals of complex [Zn(MIm2Pr)Cl(H2O)] suitable for X-ray diffraction were obtained from a solution of [Zn(MIm2Pr)Cl] in H2O upon standing. The molecular structure of [Zn(MIm2Pr)Cl(H2O)] is depicted in Fig. 6, with selected bond lengths and angles presented in Table 2. Fig. 6Molecular structure of [Zn(MIm2Pr)Cl(H2O)] in the crystal, showing three molecules of the carboxylato-bridged coordination polymer, forming a one-dimensional chain in the crystallographic c-direction. C–H hydrogen atoms have been omitted for clarity. Displacement ellipsoids are drawn at the 50% probability level. Symmetry operation a 3/2 − x, 1 − y, z − 1/2Table 2Selected bond lengths (angstroms) and bond angles (degrees) for [Zn(MIm2Pr)Cl(H2O)]BondBond lengthAngleBond angleAngleBond angleZn1–N22.1203 (15)N4–Zn1–N288.05 (6)N2–Zn1–O1a92.41 (5)Zn1–N42.0517 (14)N2–Zn1–O3173.58 (5)O1a–Zn1–O382.24 (5)Zn1–O1a1.9975 (11)O3–Zn1–N490.38 (5)N4–Zn1–Cl1119.50 (4)Zn1–Cl12.2844 (4)O3–Zn1–Cl190.23 (4)Cl1–Zn1–O1a130.40 (4)Zn1–O32.1860 (13)Cl1–Zn1–N295.97 (4)O1a–Zn1–N4109.55 (5)Symmetry operation a 3/2 − x, 1 − y, z − 1/2 The single-crystal structure determination reveals that the structure of the metal complex is that of a coordination polymer. The zinc atom is coordinated by two 1-methylimidazole groups of one MIm2Pr ligand, while the carboxylato group of this ligand coordinates to a neighboring zinc atom. In this way an infinite one-dimensional chain parallel to the crystallographic c-axis is formed. The chlorido anion and a coordinated water molecule complete the coordination sphere around each zinc center and render it five-coordinate. Thus, a new, bridging coordination mode of MIm2Pr is observed in the structure of [Zn(MIm2Pr)Cl(H2O)] comprising N,N-bidentate coordination to one metal center and O-monodentate coorrdination to a second one. This coordination mode complements the alternative N,N,O-tridentate facial capping mode of the ligand, which is commonly observed for divalent transition metals [26, 27, 39]. The coordination geometry around the five-coordinate Zn atom is distorted trigonal bipyrimidal (τ = 0.72) [40]. The equatorial positions are occupied by N4 and the monoanionic carboxylato O1a and chlorido Cl1 ligands with O1a–Zn1–N4, Cl1–Zn1–O1a, and N4–Zn1–Cl1 angles of 109.55(5), 130.40(4), and 109.50(4)°, respectively (Σangles = 360°). The axial positions are occupied by O3 of the water molecule and imidazole N2 [O3–Zn1–N2 173.58(5)°]. The Zn1–N2 and Zn1–N4 bond lengths are different [2.1203(15) and 2.0517(14) Å], consistent with their respective axial and equatorial positions. The Zn1–O1a distance of 1.9975(11) Å is slightly longer than the Zn–Ocarboxylato bond length observed in [Zn(bdatBu2,Me2)Cl]2 (1.942 Å) [12] [bpatBu2,Me2 is (3,5-di-tert-butylpyrazol-1-yl)(3′,5′-dimethylpyrazol-1-yl)acetate], in which a carboxylato group is similarly coordinated to the metal center. The zinc-aquo distance of 2.1860(13) Å in [Zn(MIm2Pr)Cl(H2O)] is rather long compared with most reported Zn–OH2 distances (mean Zn–OH2 distance for five-coordinate zinc complexes is 2.046 Å; Cambridge Structural Database version 5.27) [41]. In the related coordination polymer [Zn(bpatBu,Me)(H2O)](ClO4) [11] [bpatBu,Me is bis(5-tert-butyl-3-methylpyrazol-2-yl)acetate], where the water molecule occupies a position in the basal plane of the trigonal bipyramid, the zinc–water distance, for example, amounts to 1.961 Å. [Zn(MIm2Pr)Cl(H2O)] crystallizes in the noncentrosymmetric space group P212121 and the determination of the Flack parameter indicates that the crystal is enantiomerically pure. The one-dimensional linear chains in Zn(MIm2Pr)Cl(H2O)] are interconnected via hydrogen-bonding interactions into a two-dimensional network (Fig. 7, Table 3). The water molecule binds via its two hydrogen atoms by hydrogen bonding to a carbonyl O and a chlorido anion as hydrogen acceptors of a [Zn(MIm2Pr)Cl(H2O)] molecule in a neighboring strand. Fig. 7Hydrogen-bonding network in [Zn(MIm2Pr)Cl(H2O)]. Hydrogen bonds connect the infinite one-dimensional chains into a two-dimensional network. C–H hydrogen atoms have been omitted for clarity. Symmetry operations i 1 − x, 1/2 + y, 1/2 − z; ii −1/2 + x, 3/2 − y, −zTable 3Selected hydrogen bond lengths (angstroms) and angles (degrees) for [Zn(MIm2Pr)Cl(H2O)]Donor–H···acceptorD–HH···AD···AD–H···AO3–H1O···O2i0.78 (3)1.95 (3)2.7307 (19)176 (3)O3–H2O···Cl1ii0.84 (3)2.31 (3)3.1472 (15)171 (3)Symmetry operations i 1 − x, 1/2 + y, 1/2 − z; ii −1/2 + x, 3/2 − y, −z Zinc complexes of iPrEtIm2Pr In order to assess the influence of the presence of steric bulk in the ligand framework, we also studied the zinc coordination chemistry of ligand iPrEtIm2Pr. The ethyl and isopropyl substituents of this ligand provide significant steric hindrance and, moreover, increase the solubility of its complexes in organic solvents. The 1:1 complex [Zn(iPrEtIm2Pr)Cl] was synthesized by the reaction of equimolar amounts of ZnCl2 and K[iPrEtIm2Pr] in methanol. The complex was isolated in almost quantitative yield. The IR spectrum of [Zn(iPrEtIm2Pr)Cl] showed the asymmetric and symmetric carbonyl stretch vibrations at 1,627 and 1,381 cm−1 [Δ(νas − νs) = 246 cm−1], respectively. The structure of [Zn(iPrEtIm2Pr)Cl] was determined by X-ray crystal structure determination. Crystal structure of [Zn(iPrEtIm2Pr)Cl] Colorless single crystals of [Zn(iPrEtIm2Pr)Cl] suitable for X-ray diffraction were obtained from a solution of [Zn(iPrEtIm2Pr)Cl] in methanol. The molecular structure of [Zn(iPrEtIm2Pr)Cl] is depicted in Fig. 8, with selected bond lengths and angles presented in Table 4. The crystal structure shows that whereas like [Zn(MIm2Pr)Cl(H2O)] [Zn(iPrEtIm2Pr)Cl] also crystallizes as a coordination polymer, the molecular structure of [Zn(iPrEtIm2Pr)Cl] is noticeably different. Unlike in the structure of [Zn(MIm2Pr)Cl(H2O)], no water molecule is found to coordinate to zinc in [Zn(iPrEtIm2Pr)Cl]. Instead of the five-coordinate structure in [Zn(MIm2Pr)Cl(H2O)], the zinc atom in [Zn(iPrEtIm2Pr)Cl] is four-coordinate. It has a pseudotetrahedral coordination geometry with two imidazole groups of one iPrEtIm2Pr ligand, a chlorido, and a carboxylato group of a neighboring ligand coordinated to it. The bridging mode of iPrEtIm2Pr observed in [Zn(iPrEtIm2Pr)Cl] is the same as found in [Zn(MIm2Pr)Cl(H2O)]. The deviation from ideal tetrahedral geometry is reflected in the observed bond angles [91.90(9)–119.66(6)°]. The smallest angle (N2–Zn1–N4) is caused by the inherent geometrical restrictions imposed by the ligand and the largest angle (Cl1–Zn1–O1a) probably reflects the steric influence of the isopropyl groups. The Zn–N, Zn–O, and Zn–Cl bond lengths in [Zn(iPrEtIm2Pr)Cl] are shorter than the corresponding distances in [Zn(MIm2Pr)Cl(H2O)], which reflects the stronger Lewis acidity of zinc(II) in a four-coordinate versus a five-coordinate geometry. The Zn1–O1a bond length of 1.9628(19) Å is comparable to values reported for other tetrahedral N,N,O zinc complexes with a monodentate carboxylate [9, 11, 12, 18]. The different coordination geometry at zinc results in a more folded zigzag polymer chain in [Zn(iPrEtIm2Pr)Cl], compared with the more stretched zigzag chain observed for [Zn(MIm2Pr)Cl(H2O)]. Fig. 8Molecular structure of [Zn(iPrEtIm2Pr)Cl] in the crystal, showing three molecules of the carboxylato-bridged coordination polymer. C–H hydrogen atoms have been omitted for clarity. Displacement ellipsoids are drawn at the 50% probability level. Symmetry operation a x, 1 − y, z − 1/2Table 4Selected bond lengths (angstroms) and bond angles (degrees) for [Zn(iPrEtIm2Pr)Cl]BondBond lengthAngleBond angleAngleBond angleZn1–N22.038 (2)Cl1–Zn1–N2116.85 (7)N2–Zn1–O1a108.85 (8)Zn1–N42.045 (2)Cl1–Zn1–N4112.12 (7)N4–Zn1–O1a103.15 (9)Zn1–O1a1.9628 (19)Cl1–Zn1–O1a119.66 (6)Zn1–Cl12.2146 (7)N2–Zn1–N491.90 (9)Symmetry operation a x, 1 − y, z − 1/2 The structure of [Zn(iPrEtIm2Pr)Cl] and more generally the coordination chemistry of iPrEtIm2Pr with Zn in methanol solution were studied by 1H NMR spectroscopy and ESI–MS. The 1H NMR spectra of a titration experiment comprising the addition of ZnCl2 to the potassium salt of iPrEtIm2Pr in methanol solution are shown in Fig. 9. Fig. 9400 MHz 1H NMR spectra of the titration of K[iPrEtIm2Pr] with ZnCl2 in CD3OD at room temperature: a K[iPrEtIm2Pr] and ZnCl2 1:0; b K[iPrEtIm2Pr] and ZnCl2 1:1; c K[iPrEtIm2Pr] and ZnCl2 2:1; d K[iPrEtIm2Pr] and ZnCl2 3:1. Asterisks denote the residual solvent peak and H2O Comparison of the 1H NMR spectra of K[iPrEtIm2Pr] (Fig. 9, spectrum a) with that of an equimolar K[iPrEtIm2Pr]–ZnCl2 mixture shows significant shifts pointing to the formation of a [Zn(iPrEtIm2Pr)Cl] complex (spectrum b). In the 1H NMR spectrum of [Zn(iPrEtIm2Pr)Cl] one sharp resonance is observed for the imidazole protons at 6.99 ppm, which is shifted to lower field compared with that of the free ligand. The two imidazole rings are equivalent on the NMR timescale, suggesting that they are symmetrically bound to zinc. Importantly, the CH2 resonance of the ethyl groups is shifted and split into two signals of equal intensity (4.20 and 4.32 ppm). Similarly, the methyl groups of the isopropyl substituent now appear as two doublets of equal intensity. These splittings confirm that the CH2(Et) and CH3(iPr) groupings contain diastereotopic protons and methyl groups, respectively. No splitting or change in multiplicity of the resonances associated with the propionate backbone is observed, which is indicative of mirror symmetry of the complex in solution. Although at first sight the NMR data would allow an assignment of the solution structure for [Zn(iPrEtIm2Pr)Cl] as a mononuclear zinc complex with a facially capped ligand, a bridged oligomeric species (similar as the one discussed before for [Zn(MIm2Pr)Cl]) is preferred. ESI–MS data support the latter description since trimeric, dimeric, and monomeric species are observed. Prominent ions corresponding to [Zn(iPrEtIm2Pr)Cl + H]+, [Zn2(iPrEtIm2Pr)2Cl2 + H]+, and [Zn3(iPrEtIm2Pr)3Cl3 + H]+ are, for instance, observed. The structure of [Zn(iPrEtIm2Pr)Cl] in solution can therefore best be described as a coordination polymer similar to [Zn(MIm2Pr)Cl]. Upon addition of 2 equiv of K[iPrEtIm2Pr] to ZnCl2 in CD3OD (Fig. 9, spectrum c), a second species is formed next to the 1:1 complex [Zn(iPrEtIm2Pr)Cl]. The fact that two species are observed, indicates that an equilibrium is obtained, i.e., the clean formation of a symmetrical 2:1 iPrEtIm2Pr to zinc complex, as previously observed for MIm2Pr, does not occur. Further inspection of the 1H NMR spectrum indicated that all resonances of the 1-ethyl-4-isopropylimidazole groups of the new species are split into two signals of equal intensity, e.g., the imidazole ring protons at 7.37 and 7.19 ppm. This suggests that the two imidazole groups are inequivalent and experience a drastically different magnetic environment. For instance, the two resonances of the CH protons of the two isopropyl groups are found at 2.96 and 1.03 ppm. Furthermore, the methine proton at the bridging carbon is still a single triplet, whereas the CH2 group of the propionate backbone has become diastereotopic. Variable-temperature NMR studies over a temperature range from −40 to 60 °C did not show significant spectral changes. The ESI–MS spectrum of a 2:1 K[iPrEtIm2Pr]–ZnCl2 solution showed prominent ions corresponding to the [Zn(iPrEtIm2Pr)2 + H]+ and [Zn(iPrEtIm2Pr)2 + 2H]2+ cations. On the basis of these data, the structure of the new species is proposed to be a 2:1 ligand-to-zinc complex [Zn(iPrEtIm2Pr)2] in which each ligand is coordinated via its carboxylate group and one imidazole group, resulting in a four-coordinated zinc(II) ion (Fig. 10). The coordination of two of the three donor atoms in a 2:1 ligand-to-metal complex has also been observed with bulky N,N,S(thiolate) ligands in the complexes [Zn(L2S)2] [L2SH is (3-tert-butyl-5-methyl-2-thiophenyl)bis(3,5-dimethyl pyrazolyl) methane] and [Zn(L2)2] [L2 is 1-methoxy-2-methyl-1,1-bis(1-methyl-4,5-diphenyl-1H-imidazol-2-yl)propane-2-thiol] [42, 43]. The addition of a third equivalent of K[iPrEtIm2Pr] (spectrum d) results in the disappearance of the signals associated with [Zn(iPrEtIm2Pr)Cl] and both free ligand K[iPrEtIm2Pr] and [Zn(iPrEtIm2Pr)2] are detected. No attempts to isolate [Zn(iPrEtIm2Pr)2] were undertaken. Fig. 10Proposed structure for [Zn(iPrEtIm2Pr)2]. The structure was calculated by molecular mechanics geometry optimization (Spartan SGI [44] MMFF94) Attempted synthesis of a zinc bis(1-methylimidazol-2-yl)acetate complex Recently, the zinc coordination chemistry of bis(pyrazolyl)acetates (bpa) has been actively explored to mimic the 2-His-1-carboxylate facial triad found in several different zinc enzymes [9, 11, 12, 18, 19]. The backbone of bpa is more rigid than the backbone of the ligands MIm2Pr and iPrEtIm2Pr used in the present study. In particular, the bridging N,N-bidentate-O-monodentate coordination mode reported here is less favorable for bpa, for which a strong predisposition to an N,N,O facial capping mode is expected. The synthesis of zinc complexes of MIm2Ac would therefore be desirable, since it would allow a direct comparison of the respective pyrazole and imidazole donor groups (Fig. 11).Fig. 11Ligands benzyl bis(1-methylimidazol-2-yl)acetate (BnMIm2Ac), MIm2Ac, and bis(pyrazolyl)acetate (bpa) and the synthesis of complexes [ZnCl2(BnMIm2Ac)] and [ZnCl2(MIm2CH2)] We attempted the synthesis of ligand MIm2Ac, but found that the common synthetic procedure, i.e., lithiation of MIm2CH2 followed by reaction with carbon dioxide [45], did not result in product formation in acceptable yields. Instead, it was found that the ligand or an ester precursor of the ligand decarboxylates rather easily, as was also observed by Peters et al. [39]. The benzyl ester of bis(1-methylimidazol-2-yl)acetic acid (BnMIm2Ac), however, was sufficiently stable and could be isolated and purified after reaction of lithio bis(1-methylimidazol-2-yl)methane with benzylchloroformate. Since hydrolysis and hydrogenolysis of the ester moiety resulted in decomposition of the ligand, the 1:1 zinc dichloride complex with BnMIm2Ac was synthesized in the hope that autocatalytic ester hydrolysis by the zinc complex would yield the desired zinc bis(1-methylimidazol-2-yl)acetate complex. A similar approach, involving the autocatalytic saponification of a zinc bis(picolyl)glycine ethyl ester complex, has been reported by Abufarag and Vahrenkamp [46]. The reaction of equimolar amounts of ZnCl2 and BnMIm2Ac gave the complex [ZnCl2(BnMIm2Ac)] as a white powder in quantitative yield. The structure of [ZnCl2(BnMIm2Ac)] was determined by X-ray diffraction studies and showed the BnMIm2Ac ligand bound to the zinc center in an N,N-bidentate way (Fig. 12, top). However, stirring of [ZnCl2(BnMIm2Ac)] in H2O at room temperature in an attempt to obtain [Zn(MIm2Ac)Cl] resulted in decomposition of the ligand and formation of [ZnCl2(MIm2CH2)], which was isolated as one of the decomposition products. [ZnCl2(MIm2CH2)] was also characterized crystallographically (Fig. 12, bottom). On the basis of these findings, it was concluded that the MIm2Ac monoanion is too unstable for practical studies and the pursuit of this ligand was abandoned. Fig. 12Molecular structures of [ZnCl2(BnMIm2Ac)] (top) and [ZnCl2(MIm2CH2)] (bottom) in the crystal. Displacement ellipsoids are drawn at the 50% probability level. All hydrogen atoms have been omitted for clarity. Selected bond lengths (angstroms) as follows: [ZnCl2(BnMIm2Ac)], Zn1–N2 2.011(2), Zn1–N4 1.994(2), Zn1–Cl1 2.2242(7), Zn1–Cl2 2.2554(7); [ZnCl2(MIm2CH2)], Zn1–N2 2.0247(17), Zn1–N4 2.0222(17), Zn1–Cl1 2.2110(6), Zn1–Cl2 2.2504(6) Reaction of [Zn(MIm2Pr)Cl] with sodium pyruvate: oxalate formation In an attempt to prevent the formation of coordination polymers of MIm2Pr and to obtain mononuclear complexes with the desired N,N,O facial capping mode of the ligand, several different monoanionic, bidendate coligands were used. In the course of this study, we also investigated the reaction of [Zn(MIm2Pr)Cl] with sodium pyruvate. The addition of 1 equiv of sodium pyruvate to a solution of [Zn(MIm2Pr)Cl] in D2O resulted in the appearance of a new singlet at 2.38 ppm in the 1H NMR spectrum (Fig. 13, top spectrum). The signal is not shifted with respect to the signal observed for free sodium pyruvate, which indicates that the interaction between [Zn(MIm2Pr)Cl] and the pyruvate anion is weak, at best. The ESI–MS spectrum of an aqueous solution of equimolar amounts of [Zn(MIm2Pr)Cl] and sodium pyruvate, however, did show some major ions corresponding to a mononuclear {[Zn(MIm2Pr)(pyruvate)] + Na}+ and a dinuclear {[Zn2(MIm2Pr)2(pyruvate)]}+ cation, in addition to the ions observed in the ESI–MS spectrum of a solution of [Zn(MIm2Pr)Cl] in water. Fig. 131H NMR spectra of a solution of [Zn(MIm2Pr)Cl] in D2O (asterisk) after addition of 1 equiv of sodium pyruvate and the same solution after 24 h However, when the mixture of [Zn(MIm2Pr)Cl] and pyruvate in D2O was followed in time by NMR spectroscopy, the gradual decrease in intensity of the signal of the pyruvate methyl group was noticed until it ultimately disappeared (Fig. 13, bottom spectrum). After several days colorless crystals suitable for X-ray diffraction studies were formed in the NMR tube. To our surprise, the X-ray crystal structure showed the structure of the zinc complex [Zn2(MIm2Pr)2(ox)]·6H2O. [Zn2(MIm2Pr)2(ox)]·6H2O has the structure of a coordination polymer involving [Zn(MIm2Pr)]+ cations bridged by oxalato dianions. To unequivocally prove the relevance of the structure obtained, zinc coordination polymer [Zn2(MIm2Pr)2(ox)] was also synthesized independently. Indeed, the addition of 0.5 equiv of K2C2O4·H2O to an aqueous solution of [Zn(MIm2Pr)Cl] yielded a white, insoluble powder. This powder was identified as [Zn2(MIm2Pr)2(ox)] by elemental analysis and IR spectroscopy. Typical features of the IR spectrum of [Zn2(MIm2Pr)2(ox)] include absorptions at 1,674, 1,638, 1,591, and 1,399 cm−1, which can be attributed to the oxalato and carboxylato groups in the polymer. Crystallization of [Zn2(MIm2Pr)2(ox)] proved difficult given its general insolubility. Single crystals could, however, be obtained via the gel crystallization method [47]. A gel formed from a solution of tetramethoxysilane and potassium oxalate hydrate in water was layered with an aqueous solution of K[MIm2Pr] and ZnCl2. Single crystals were obtained after several days and could be unequivocally identified as [Zn2(MIm2Pr)2(ox)] by X-ray crystal structure determination. Crystal structure of [Zn2(MIm2Pr)2(ox)]·6H2O The molecular structure of [Zn2(MIm2Pr)2(ox)]·6H2O is shown in Fig. 14, with selected bond lengths and angles presented in Table 5.1 The polymeric structure of [Zn2(MIm2Pr)2(ox)]·6H2O comprises two crystallographically inequivalent zinc atoms. Both zinc centers are five-coordinate and differ only in the slightly different bond lengths and angles. Each zinc atom is coordinated in an N,N-bidentate fashion by the 1-methylimidazole groups of one MIm2Pr ligand, two oxygen atoms of a bis O,O-bidentate bridging oxalato group, and a carboxylato O donor atom of a neighboring MIm2Pr ligand. The combination of the bridging oxalato and MIm2Pr ligands constructs a polymeric structure leading to the formation of two-dimensional sheets. Fig. 14Molecular structure of [Zn2(MIm2Pr)2(ox)]·6H2O in the crystal. C–H hydrogen atoms and all solvent water molecules have been omitted for clarity. Displacement ellipsoids are drawn at the 50% probability levelTable 5Selected bond lengths (angstroms) and angles (degrees) for [Zn2(MIm2Pr)2(ox)]·6H2OBondBond lengthAngleBond angleAngleBond angleZn1–O312.258 (2)O11–Zn1–O3189.48 (8)O21–Zn2–O4190.33 (8)Zn1–O322.018 (2)O11–Zn1–O32124.09 (9)O21–Zn2–O42125.19 (8)Zn1–O111.9694 (19)O11–Zn1–N22b108.91 (9)O21–Zn2–N12109.08 (9)Zn1–N24b2.046 (2)O11–Zn1–N24b107.61 (9)O21–Zn2–N14107.85 (9)Zn1–N22b2.088 (2)O31–Zn1–O3277.16 (9)O41–Zn2–O4277.84 (8)Zn2–O412.233 (2)O31–Zn1–N22b161.43 (9)O41–Zn2–N12160.51 (8)Zn2–O422.0179 (19)O31–Zn1–N24b87.69 (9)O41–Zn2–N1487.24 (9)Zn2–O211.9630 (19)O32–Zn1–N22b89.94 (8)O42–Zn2–N1289.12 (8)Zn2–N122.089 (2)O32–Zn1–N24b125.38 (9)O42–Zn2–N14124.42 (9)Zn2–N142.052 (2)N22b–Zn1–N24b89.06 (9)N12–Zn2–N1488.34 (9)C31–O311.248 (4)C41–O411.250 (3)C31d–O321.252 (4)C41c–O421.254 (3)C31–C31d1.544 (6)C41–C41c1.537 (6)Symmetry operations b 1 + x, y, z; c −x, 1 − y, −z; d 1 − x, 1 − y, 1 − z The coordination geometry around each zinc center is best described as a severely distorted trigonal bipyramid, with τ values of 0.60 and 0.59 for Zn1 and Zn2, respectively [40]. The equatorial positions are occupied by N24b/N14, and the anionic carboxylato O11/O21 and oxalato O32/O42 ligands for Zn1/Zn2, respectively. The axial positions are occupied by O31/O41 of the oxalato group and imidazole N22b/N12 for Zn1/Zn2 [O–Zn–N angles of 161.43(9)°/160.51(8)°]. The planar oxalato moieties are rather asymmetrically bound to the metal centers, with the Zn1–O31 and Zn1–O32 distances amounting to 2.258(2) and 2.018(2) Å, for instance. This asymmetry has also been observed in other bisbidentate bridged zinc–oxalato complexes with five-coordinated metal centers and is consistent with the occupation of one equatorial and one axial site of the trigonal bipyramid [10, 48]. The shortest Zn···Zn distances via the bridging oxalato group are 5.5667(7) Å for Zn1···Zn1d and 5.5332(7) Å for Zn2···Zn2c. The distances of the metal centers to the donor atoms of MIm2Pr follow the same trend as observed in [Zn(MIm2Pr)Cl(H2O)]. The crystal structure of [Zn2(MIm2Pr)2(ox)]·6H2O is further stabilized by 14 different hydrogen-bonding interactions, which involve the cocrystallized water molecules, the oxalato bridges, and the carboxylato groups (Fig. 15). The oxalato groups are, for instance, involved in two chelated, three-centered hydrogen bonds with two water molecules via four O–H···O hydrogen bonds. Fig. 15Hydrogen bonding network and crystal packing of [Zn2(MIm2Pr)2(ox)]·6H2O Discussion The aim of this investigation was to explore the biomimetic modeling potential of the MIm2Pr ligand family with respect to mononuclear zinc enzymes featuring the 2-His-1-carboxylate facial triad. Complex [Zn(MImPr)2] shows that the facial capping mode of MIm2Pr, as earlier observed with copper [27, 28] and iron (Bruijnincx et al., submitted) [26], is also accessible for zinc. The crystal structure of the 1:1 complex [Zn(MIm2Pr)Cl(H2O)], however, shows that MIm2Pr can also adopt other coordination modes. The formation of the oligomeric/polymeric structure of [Zn(MIm2Pr)Cl], both in solution and in the solid state, can be attributed to the flexibility of the ligand due to the CH2 spacer of the propionate backbone. The formation of polymeric [Zn(MIm2Pr)Cl(H2O)] is further aided by the insolubility of the resulting coordination polymer. Two strategies were explored to circumvent polymer formation, i.e., increasing the steric bulk of the ligand and elimination of the CH2 spacer. The latter approach demanded the synthesis of the bis(1-methylimidazol-2-yl)acetate analogue of MIm2Pr, which was found to be rather unstable and its zinc complexes could not be obtained. The effect of more steric bulk was studied with iPrEtIm2Pr, which contains the sterically more demanding isopropyl and ethyl substituents. Zinc complexes of iPrEtIm2Pr are polymeric in nature as well, albeit with a different coordination geometry at the zinc atom as observed in [Zn(MIm2Pr)Cl(H2O)]. The first coordination spheres of [Zn(MIm2Pr)Cl(H2O)] and [Zn(iPrEtIm2Pr)Cl] are compared in Fig. 16. The added steric bulk in iPrEtIm2Pr causes the zinc ion to adopt a pseudotetrahedral geometry in [Zn(iPrEtIm2Pr)Cl], instead of the distorted trigonal bipyramid observed for [Zn(MIm2Pr)Cl(H2O)]. The coordination around zinc in [Zn(iPrEtIm2Pr)Cl] resembles the first coordination sphere of the active sites of the zinc enzymes thermolysin, carboxypeptidase A, and neutral protease. The polymeric rather than monomeric nature of the complex, however, clearly limits its potential as a biological mimic. Fig. 16Comparison of the first coordination spheres of the zinc atoms in the coordination polymers [Zn(MIm2Pr)Cl(H2O)] (top) and [Zn(iPrEtIm2Pr)Cl] (bottom) The isolation of [Zn2(MIm2Pr)2(ox)] from the reaction of [Zn(MIm2Pr)Cl] with sodium pyruvate was rather unexpected and is, to the best of our knowledge, the first report of the nonoxidative formation of oxalate from pyruvate (Scheme 1).2 Interestingly, Takeste and Vahrenkamp [51] very recently reported a trispyrazolylborato–zinc complex with pyruvate, representing the first structurally characterized zinc complex of an α-keto acid. Although in their case the zinc–pyruvate complex obtained was stable, unexpected cleavage reactions with other α-keto acids, such as benzoylformic acid, and α-diketones were observed.Scheme 1The observed transformation of pyruvate to oxalate mediated by [Zn(MIm2Pr)Cl] A control experiment showed that an aqueous sodium pyruvate solution is stable under the reaction conditions. Apparently, [Zn(MIm2Pr)Cl] mediates the conversion of pyruvate to oxalate. The product [Zn2(MIm2Pr)2(ox)] invariably deposited as crystals in experiments done on the NMR scale. On a preparative scale (0.3 mmol), the reaction also takes place, yet the product was isolated as an insoluble white powder. The IR spectrum of this powder was identical to that obtained for independently synthesized [Zn2(MIm2Pr)2(ox)]. Most importantly, further experiments revealed that the reaction is catalytic in zinc. For example, addition of 20 equiv of sodium pyruvate to a solution of [Zn(MIm2Pr)Cl] in water under ambient conditions resulted in ten turnovers after 24 h. The disappearance of the pyruvate methyl signal was also observed in a solution containing ZnCl2 and sodium pyruvate without any further additives. Concomitantly, a precipitate was observed in the NMR tube. However, the composition of this precipitate, i.e., the formation of oxalate, could not be unambiguously established. Since it is known that pyruvate keto–enol tautomerization is catalyzed by divalent metal ions such as Zn(II) [52, 53], the disappearance of the methyl signal could in principle also be attributed to consecutive H/D-exchange steps. This would yield the deuterated pyruvate-d3 anion, which is also invisible in 1H NMR spectroscopy. To be able to exclude this possibility, experiments were performed in H2O and samples were diluted with 2 equiv of D2O just before the measurement. The same observations were made as previously, i.e., gradual disappearance of the methyl signal and slow crystallization of the product in the NMR tube. The presence of appreciable amounts of enolate should also lead to the detection of the dimerization product of pyruvate [52, 53]. This pyruvate dimer, which results from an aldol condensation reaction and has characteristic signals at 1.38 (singlet) and 3.35 (AB pattern) ppm [52–54], was not observed under these conditions. Phenyl pyruvic acid (and other α-ketoacids) is oxidized by molecular oxygen to give benzaldehyde and oxalate or the carbon oxides (CO and/or CO2) [55, 56]. This reaction is catalyzed by several metal ions, such as Mn(II), Fe(II), and Cu(II), but not by Zn(II) [57, 58], and has been attributed to the presence of the enol tautomer, which is capable of reacting with dioxygen [61]. In the absence of dioxygen, i.e., under an Ar atmosphere, pyruvate does react with [Zn(MIm2Pr)Cl] to form oxalate. The anaerobic reactivity cannot be attributed to any residual dioxygen traces, since phenyl pyruvate (a substrate more prone to the reaction with dioxygen) did not show any reactivity under these conditions. The observed conversion of pyruvate to oxalate under anaerobic conditions clearly places it outside previously observed formation of oxalate from oxidative transformations of α-keto acids [55, 56, 59–63]. The observation is important, since pyruvate is a major metabolic junction and it is therefore of prime importance to understand its reactivity. At present no biological counterpart has been reported for the pyruvate–oxalate conversion catalyzed by [Zn(MIm2Pr)Cl]. The chemical transformation as reported here, however, takes place at room temperature and physiological pH and should therefore be considered as a possible alternative pathway in the reactivity of pyruvate. The mechanism of this conversion is currently not clear. No other products could be detected by NMR during the reaction, which precludes the formulation of a mass balance for the reaction and clearly hampers a mechanistic interpretation. In any case, the pyruvate to oxalate conversion requires the scission of a C–C bond. On the other hand, the reactivity of zinc enzymes and their model complexes is often of hydrolytic nature. This can be attributed to the generation of a nucleophilic hydroxide species and/or the activation of a coordinated substrate towards attack of such a nucleophile [3]. The formation of oxalate from pyruvate via a hydrolytic mechanism is, however, difficult to envision. Further investigations have to be performed to address the mechanistic questions concerning this transformation. Conclusions The difference in applied steric bulk by the ligands MIm2Pr and iPrEtIm2Pr is reflected in the coordination chemistry of the two ligands towards ZnCl2. Octahedral (MIm2Pr) versus tetrahedral (iPrEtIm2Pr) and trigonal pyramidal (MIm2Pr) versus tetrahedral (iPrEtIm2Pr) coordination geometries were obtained in the 1:1 and 2:1 ligand-to-metal complexes, respectively. The bridging binding mode of the two crystallographically characterized coordination polymers illustrates the intrinsic flexibility of the 3,3-bis(imidazol-2-yl)propionate ligand framework. The polymeric nature of the complexes could limit their biological relevance with respect to the 2-His-1-carboxylate facial triad in the case of zinc. On the other hand, attempts aimed at obtaining structurally more faithful mononuclear complexes led to the observation of the unprecedented Zn-mediated pyruvate to oxalate conversion. This new, nonoxidative transformation of an important metabolic junction is intriguing and warrants further investigation.

Matern_Child_Health_J-2-2-1592139

The Business Case for Preconception Care: Methods and Issues

Only a limited number of economic evaluations have addressed the costs and benefits of preconception care. In order to persuade health care providers, payers, or purchasers to become actively involved in promoting preconception care, it is important to demonstrate the value of doing so through development of a “business case”. Perceived benefits in terms of organizational reputation and market share can be influential in forming a business case. In addition, it is standard to include an economic analysis of financial costs and benefits from the perspective of the provider practice, payer, or purchaser in a business case. The methods, data needs, and other issues involved with preparing an economic analysis of the likely financial return on investment in preconception care are presented here. This is accompanied by a review or case study of economic evaluations of preconception care for women with recognized diabetes. Although the data are not sufficient to draw firm conclusions, there are indications that such care may yield positive financial benefits to health care organizations through reduction in maternal and infant hospitalizations. More work is needed to establish how costs and economic benefits are distributed among different types of organizations. Also, the optimum methods of delivering preconception care for women with diabetes need to be evaluated. Similar assessments should also be conducted for other forms of preconception care, including comprehensive care. Introduction The opportunity to assist a woman before she becomes pregnant to improve her chances of a healthy pregnancy is compelling. Adverse pregnancy outcomes, including birth defects, preterm birth, and perinatal complications, are the leading causes of infant mortality in the United States [1]. Further, developmental disabilities such as cerebral palsy result in direct and indirect economic costs that can exceed $1 million over a given child’s lifetime [2], in addition to the social and emotional costs incurred by families. A number of additional conditions are associated with serious complications of pregnancy, such as uncontrolled diabetes, hypothyroidism, hypertension, use of alcohol or tobacco, and obesity. The economic burden of most of these conditions in conjunction with pregnancy has been poorly studied but is likely to be enormous. This paper discusses the potential economic benefits of preconception care due to the prevention of adverse pregnancy outcomes with a focus on the data that would be needed to assess costs and benefits. Currently, the needed data and analyses are largely lacking. In the absence of information about the costs and benefits of comprehensive preconception care, which should ideally include nutritional counseling, evaluation of medications, and family history evaluation and genetic counseling, this paper focuses on principles and methods. Three published studies have analyzed the expected financial benefits from counseling women with diabetes prior to pregnancy. These studies are compared and lessons drawn, with discussion of implications of preconception care for women with other conditions. In order to argue that preconception care is a “good buy,” its costs and benefits must be assessed from the perspectives of a variety of stakeholders involved in health care services. This paper will first define a “business case” analysis and outline the methods for determining the “return on investment” from the perspective of various health care organizations. Similarities and differences between such an analysis and a standard cost-effectiveness analyses that calculate returns to the health care system as a whole or to society will be explained. Second, the paper reviews the literature on business case analyses from areas other than preconception care. Third, the paper discusses what data sources are available and needed to apply business case methods in preconception care with particular focus on women with diabetes. Finally, the paper draws conclusions about future steps needed to assess the business case for preconception care. What is a business case analysis? Concepts and definitions A business case is an argument for a financial investment couched in terms of potential economic or market advantages to an organization, most commonly a private corporation [3]. The core of a business case is usually a budget impact analysis, a calculation of the financial return on investment (ROI) made by the firm(s) or organization(s) [4]. In addition to financial returns, such as cost savings or cost avoidance, a business case can be based on perceived strategic advantages such as enhanced organizational reputation, market positioning, or increased employee and customer satisfaction. This paper focuses just on financial business case analyses. According to the Institute for Healthcare Improvement (IHI) and the Commonwealth Fund [3]:A business case for a health care improvement intervention exists if the entity that invests in the intervention realizes a financial return on its investment in a reasonable time frame, using a reasonable rate of discounting. This may be realized as “bankable dollars” (profit), a reduction in losses for a given program or population, or avoided costs. In addition, a business case may exist if the investing entity believes that a positive indirect effect on organizational function and sustainability will accrue within a reasonable time frame. A business case related to health care is based on one or several of the perspectives of specific payers, purchasers, or providers. Because health care providers, payers, and purchasers often have different, and sometimes conflicting, economic incentives, the business case for the same intervention will probably differ among these groups. For example, improved ambulatory care may lead to reduced inpatient admissions, which could result in financial savings for payers, lost revenue for hospitals, and higher costs for ambulatory care providers. The resulting misaligment of incentives in the health care system may, and often does, lead to inappropriate provision or use of certain services and underutilization of others [3]. Although business case analyses by definition focus on businesses, they are relevant to other stakeholders as well. In particular, the consumer perspective is also important to keep in mind. If consumers do not consider a service such as preconception care to provide good value for their investment of time and money, low demand and utilization could render the business case unfavorable. Further, government agencies must meet budgets, and the short term financial impact of policy changes may influence funding decisions. In the long run, though, public health programs are expected to optimize population health subject to resource availability, which requires analysis of both short-term and long-term health outcomes and economic consequences. Analytical techniques and outcome measures Financial analyses that underlie a business case are variably referred to as budget impact and ROI analyses. It is important to situate such analyses in the context of two commonly used methods of economic evaluation, cost-effectiveness analysis (CEA) and cost-benefit analysis (CBA) [5]. A CEA calculates the net monetary cost per unit of health outcome achieved. Certain preventive services are both more effective than alternatives and less costly [6]. A CEA that yields evidence of net cost savings from the payer perspective could be used to frame a business case, but CEAs rarely report payer-specific costs. In a CBA, all outcomes are expressed in dollar values, and the summary measure is expressed as either a net economic benefit or benefit-cost ratio. For example, a CBA of prenatal care for undocumented immigrants in California estimated that public funding cost the state $58 million but saved $211 million in long-term costs of care, for a projected net benefit of $153 million and a benefit-cost ratio of 4.6:1 [7]. A financial analysis of a business case can be viewed as a CBA conducted from the perspective of an individual firm or organization. One important distinction between financial analyses and standard economic evaluations is that the latter are supposed to reflect the opportunity cost or value of resources from a societal perspective whereas a financial analysis considers the direct monetary or budgetary impact to a specific organization. The ROI metric is typically calculated as the ratio of dollars in increased revenue or decreased cost divided by investment cost, although this can also be expressed as net cost difference [8]. Various other metrics can be used in a financial analysis, such as payback period or internal rate of return. For example, a recent study estimated the number of years that could be required for employers to realize a positive net benefit from bariatric surgery for obese employees as a result of reduced medical costs (all of which were assumed to be paid by the employer) and loss of work time [9]. ROI analyses require payer-specific costs, which are rarely reported in standard CBA studies. Time horizon and discount rate Business case analyses generally employ a short time horizon; for example, only changes in use and costs over a period of 2 to 3 years following an intervention are typically considered. This reduces the problem of enrollee or employee turnover. When individuals switch plans or employers, the return on investments in the long-term health of that individual made by the former health plan or employer will be realized by someone else. This can make financial returns to employers or health plans substantially lower than the economic benefits to society [3, 10, 11]. A payback period ranging from 5 years to 10 years [9] is not likely to be attractive to employers. In economic evaluations from the societal perspective, it is conventional to discount costs occurring in future years using a discount rate of 3%, which reflects the social return on capital [5]. For a business case analysis employing a short time horizon, long-term future costs or benefits may not need to be counted or discounted. In the short run, when ROI analysis does employ discounting, the discount rate should reflect the opportunity cost of capital in the private sector or individual firm. Prospective and retrospective evaluations Financial and health economic analyses of an intervention can be performed before or after its implementation. A prospective analysis predicts the financial return for a new initiative or program before it begins. This presumes that there is reliable information about the effectiveness of the intervention and accurate historical cost and revenue data. Analysts construct a simulation model that projects likely costs, health outcomes, and benefits based on a set of assumptions. The benefits include anticipated savings in health care costs and, for analyses from the employer perspective, worker productivity [12, 13]. A limitation of prospective analyses is that key assumptions may turn out to be unrealistic. For example, analyses in the 1980s and early 1990s projected cost savings to Medicaid programs from reductions in the number of low birth weight (LBW) births if prenatal care coverage were expanded (e.g., [14–16]). In conjunction with effective political advocacy and policy changes, these arguments succeeded in more than doubling the number of pregnancies covered by Medicaid prenatal care between 1986 and 1991. Despite this expansion, little or no reduction in LBW rates could be confirmed to have occurred among eligible population groups [17]. The lack of an observed effect may reflect deficiencies in the effectiveness of the routine prenatal care that was offered, as well as barriers to access, and does not rule out a protective effect of an optimal prenatal care protocol. Although prenatal care expansion had other health benefits and can easily be defended, the cost-saving argument appears in retrospect to have been overstated [18]. In the absence of information about intervention effectiveness, a hypothetical calculation can be made regarding the potential magnitude of benefits if an effective intervention were to exist. One way to make such a calculation is to take cost-of-illness estimates of the economic burden of diseases and project the potential cost reduction that might occur as a result of prevention. Only short-term costs, which can be much smaller in magnitude than the long-term costs included in cost-of-illness studies, are likely to be relevant to employers or payers [19]. For example, one group of investigators calculated that poor birth outcomes such as preterm birth cost employer-sponsored health plans $5.6 billion in 1990 and projected that prevention of 10% of preterm births would lead to a cost savings of $560 million [20]. A retrospective or ex post analysis is based on observed costs and benefits. For example, a recent ex post CBA of folic acid fortification calculated that the economic returns were substantially greater, and the costs somewhat lower, than had been projected in previously published analyses conducted prior to fortification [21]. Although actual data can yield more reliable results, it is challenging to collect data on multiple outcomes. Furthermore, interactions among the outcomes or confounding of results from other sources can make it difficult to precisely estimate the impact of the initiative or program alone. Lessons from business case literature A number of business case studies of disease management and health promotion initiatives have been published [8, 22, 23]. However, there is a paucity of well-conducted analyses and studies that contain detailed data on intervention and outcome costs [24]. Readers of business case analyses should beware of publication bias [22] and “wish bias” leading to optimistic estimates of effectiveness and compliance with behavioral changes [10]. One group that developed seven business case studies for health care quality improvement initiatives, such as disease management, concluded that health care providers, payers, and purchasers cannot predictably expect to reap positive financial returns from initiatives to improve health care quality within short time horizons [3]. In the majority of case studies, short-term financial returns to health plan entities were negative. Furthermore, returns to health care providers were almost uniformly negative, as any cost savings were generally recouped by purchasers, and not by providers. In several cases, managers felt that there was a business case for investing in improved quality based on organizational benefits such as “retained market share, increased staff loyalty, and reputation among employers” rather than cost savings [3]. Similarly, the Child Health Business Case Working Group developed a set of business case studies [25]. For example, a case study of postpartum screening for risk factors in new mothers (e.g., depression, smoking, domestic violence) followed by referral to community resources found that such a program would lead to higher costs for the practice without additional reimbursements. A positive business case for providers may require a financial mechanism for tying reimbursements by payers to improved outcomes [3]. Integrated health service providers under capitation payments may have more opportunity to internalize the financial gains resulting from improved care [3, 25]. Several studies reported positive financial returns to managed care organizations (MCOs) from prenatal or perinatal interventions targeted to high-risk pregnant women [26–29], although these reports require validation and replication in well-controlled studies. For example, a high-risk pregnancy management program in upstate New York implemented by a managed care program serving Medicaid and other lower-income beneficiaries reported that cost savings from reduced NICU (neonatal intensive care unit) admissions were more than twice as great as the costs of the intervention [28]. A similar program implemented in another upstate New York MCO during the same time period reported investment costs (including provider reimbursements) of $573,355, cost savings of $789,621 resulting from the estimated prevention of 48 LBW births, and a net benefit of $216,266 [27]. The authors reported an internal rate of return of 37% on the investment, which is equivalent to an ROI ratio of 1.37. Economic analyses of preconception care Finally, we turn to the application of business case methods to preconception care. Preconception care refers to three types of services and target populations:Preventive services and screenings offered to women who expect to become pregnant in the near future;Interconception care for women who have given birth and intend to bear another child at some point in the future;Counseling about the impact of pre-existing health risks or conditions for affected women of childbearing age and their impact on pregnancy outcomes. Few economic evaluations of individual screenings or services in the pre-pregnancy period have been published. One CEA estimated that promoting folic acid supplementation to all women of reproductive age would be cost effective, although not cost saving [30]. If multivitamins containing folic acid were targeted instead to women intending pregnancy, supplementation might be cost saving. On the other hand, because most of the averted costs of caring for children with spina bifida occur years in the future and are shared by multiple payers, this type of analysis cannot be used to calculate the business case for folic acid supplementation strategies. One preconception care service with published analyses of short-term financial costs and benefits is preconception counseling for women with diabetes. Women with poorly controlled diabetes before pregnancy have an elevated risk of having an infant with a birth defect [31], fetal death, or macrosomia. Women with poorly controlled diabetes prior to pregnancy can also experience medical risks during pregnancy, including diabetic retinopathy, nephropathy, and ketoacidosis, as well as pregnancy-induced hypertension and uteroplacental insufficiency [32]. A substantial body of evidence on infant health outcomes associated with preconception counseling for women with diabetes yields consistent findings. According to a meta-analysis of 14 epidemiologic studies, pregnancies preceded by preconception care result in one third as many major congenital malformations as pregnancies not preceded by preconception care [31]. The Maine Diabetes in Pregnancy Program found 4-fold lower rates of both congenital anomalies and fetal and neonatal deaths among pregnancies with preconception care [33]. More recently, the Diabetes Complications Control Trial (DCCT) of 1441 adults with Type 1 diabetes randomly assigned to conventional or intensive therapy, including 180 women who became pregnant, reported 8 congenital malformations in the conventional therapy group and only 1 in the intensive therapy group [34]. Thus, in the case of maternal diabetes, randomized trial data confirm the findings of numerous observational studies that control of diabetes prior to pregnancy results in dramatically fewer birth defects. Three economic evaluations of preconception and pregnancy care programs for women with non-gestational diabetes were identified in the published literature [35–37]. All reported calculations of net short-term economic benefit based on reduced costs of maternal and neonatal hospitalizations and did not take into account long-term averted costs of care associated with the prevention of birth defects. Although the analyses were not developed as business case analyses, the focus on short-term costs and benefits is consistent with that approach. Regrettably, the lack of common metrics of cost and health outcomes makes it difficult to make direct comparisons of assumptions and findings of the three studies. The first study is a prospective analysis of a hypothetical program of comprehensive preconception diabetes care consisting of 20 visits prior to prenatal care [35]. The preconception visits were projected to cost $2,638 per enrollee in 1989 dollars. The investigators calculated that reduced maternal and infant hospitalizations would result in net benefits of $1,720 per enrollee. The majority of the hospitalization cost savings were projected to result from an 8-fold reduction in major congenital malformations, from 3.2 to 0.4%. Using the cost of the intervention as the denominator, the benefit-cost ratio from the health care system perspective can be calculated to be 1.6. This means that, on average, for every $1 spent on the program, reduced hospitalization costs of $1.60 would be expected. Although the projected percentage reduction in major malformations appears high relative to the literature, the projected intervention cost was also rather high, since it is uncertain that 20 visits would be needed for preconception care for women with diabetes. The second study, a retrospective analysis of data for the years 1986–1988 from the California Diabetes and Pregnancy Program (CDAPP), found that matched perinatal hospital charges for mother and child were 41% higher for controls than for program participants [36]. The benefit-cost ratio was 5.19, with reduced maternal and infant hospitalization costs of $5.19 observed for each $1 spent on the program. This finding likely contributed to the ongoing funding of the CDAPP. It is difficult to apply these findings to preconception care, though. Only 24 of 102 participants were enrolled in the program prior to 8 weeks of gestation, the time during which major congenital malformations are likely to occur. Although those participants had average charges that were 44% lower than those who enrolled after 8 weeks, women who enrolled in the program prior to conception incurred higher costs per delivery because of the longer time during which services were received and the fact that not all women who enrolled became pregnant. Therefore, positive net financial returns to the component of preconception care were not demonstrated in that study. The third study is a retrospective evaluation of prospectively gathered data on pregnancy-related costs among women with type 1 diabetes who received both preconception care and prenatal care or prenatal care only at selected Michigan hospitals [37]. Women in the preconception care group received an additional two outpatient visits on average, at an added cost of $132. The rate of congenital malformation was 4.2% in the preconception care group compared with 13.5% in the prenatal care-only group, a reduction of 69%. The rate of maternal and neonatal intensive care unit (NICU) hospitalizations was 50% lower in the preconception care group. Combining maternal and child hospitalizations, cost savings averaged $34,000 per woman enrolled in both preconception and prenatal care compared to those who received standard prenatal care. Although no information was presented on the distribution of costs in this study, an implication of the findings is that an integrated health plan could likely achieve cost savings by paying for a modest number of preconception care visits for women with type 1 diabetes. Discussion The economic and business case literature on preconception care is quite limited. There is some evidence of positive financial returns for preconception counseling for women with diabetes, based on savings in hospitalization costs [35–37]. Although none of the studies involved randomized assignment, the relative differences in frequencies of birth defects are consistent with findings from other studies and increase the confidence that can be placed in the estimates of cost reductions. None of the studies indicate who was expected to bear the costs of counseling or the magnitude of cost savings to specific stakeholders. The studies are all based on relatively old data; newer studies with more complete accounting of costs would be more informative. Furthermore, determination of the optimal number and cost of preconception visits is essential. The American Diabetes Association [38] and the American College of Obstetricians and Gynecologists [39] recommend that all women of childbearing age with diabetes receive counseling about the importance of glucose control before becoming pregnant. Despite these guidelines and positive economic analyses, many women with diabetes still do not receive such advice. A recent survey of women with diabetes ages 18 to 45 years enrolled in managed care plans found that only about half recalled receiving advice about the need for glucose control prior to pregnancy [40]. Research is needed to determine why some women with diabetes seek out preconception care and adopt intensive glucose control measures and others do not [41]. Potential barriers, such as opportunity cost of time, financial cost (e.g., co-pays), invasive testing, and medical risks [32] need to be analyzed from the consumer perspective. For example, it has been estimated that for patients with type 2 diabetes controlled by oral agents, following all self-care practices recommended by the American Diabetes Association could require more than 2 hours daily [42]. Better data are needed to establish the business case for preconception care. Although benefits of preconception care can be projected, data derived from the evaluation of actual preconception care initiatives are needed to make a compelling business case. Further, observational data comparing pregnancy outcomes among women who choose to use preconception care and those who do not are subject to biases due to non-randomized study design. A business case analysis of interventions with both demonstrated efficacy (works if followed) and effectiveness (behaviors are actually changed and lead to improved outcomes) is likely to be more persuasive. In order to develop a compelling business case for preconception care, studies of preconception care interventions accompanied by rigorous evaluations with randomized or matched groups of controls are required. Preparing a business case analysis based on intervention data poses challenges. It is logistically difficult but essential to prospectively collect data on costs and outcomes: “To understand which quality-enhancing interventions are likely to produce positive returns on investments, data collection and analysis must include tracking the investment and operating costs of implementing the intervention as well as the changes in revenues and costs that result from the intervention” [24]. Easily accessible or extant data sets typically do not allow one to calculate costs from the perspective of an individual organization. This requires quantifying actual reimbursements, co-pays, deductibles, etc., rather than using standard charges for a clinic visit. It also requires tracking which patients are covered by different health plans and patterns of enrollment and disenrollment. Conclusion and the way forward Research demonstrating the effectiveness and financial returns to investments in preconception care is necessary to make a financial business case for preconception care. In particular, the only economic evaluations that have been published to date have been of specific components of preconception care, such as counseling women with diabetes or promoting preconception use of folic acid supplements. Evaluations of the costs and benefits of different “bundles” of preconception care services or comprehensive preconception care are needed. Further, such evaluations must include payer-specific cost data to be relevant to framing the business case for preconception care. Although more data are needed, such analyses may not be sufficient to ensure access to or provision of preconception care, as the diabetes preconception care experience suggests. Promotion of preconception care through carefully researched health marketing campaigns is also critical [41]. If, as a result of effective health marketing, consumers were to demand preconception care, providers and payers would be more likely to perceive a business case for providing it. Similarly, a business case might also be supported if health care professionals working with women of childbearing age were to advocate that their own organizations provide such services. Health care providers often lack the incentive to adopt a new service that improves outcomes; therefore, particular attention must be paid to the structure of incentives by both public and private payers and purchasers [3]. Public health programs should fund studies evaluating the effectiveness of preconception care and develop marketing strategies to promote awareness of strategies that are proven effective. Studies that test and evaluate preconception care strategies should collect the financial data needed to conduct business case analyses [24]. Controlled evaluations of preconception care need to be accompanied by financial analyses to help make the business case for providing and paying for such services. Better quality evidence of both the financial and health costs of poor pregnancy outcomes and the benefits from preconception and early prenatal care is needed to provide a compelling business case.

Qual_Life_Res-3-1-2039824

Health status, work limitations, and return-to-work trajectories in injured workers with musculoskeletal disorders

Background The purpose of this study was to describe the health status and work limitations in injured workers with musculoskeletal disorders at 1 month post-injury, stratified by return-to-work status, and to document their return-to-work trajectories 6 months post-injury. Introduction Work-related musculoskeletal (MSK) disorders are common health problems and a major contributor to disability and costs in working populations. In Canada, MSK disorders are responsible for 10% of the short-term disability costs and 39% of the estimated long-term disability costs [1]. The natural course of low back pain and other MSK disorders is characterized by recurrent disabling symptoms and can be described as chronic-episodic [2–6]. Similarly, the trajectory of return-to-work (RTW) following a period of work absence due to MSK disorders is a complex and dynamic process, frequently involving recurrences of work absence [7, 8]. Already a decade ago, it was shown that a first return to work after an injury is frequently followed by one or more recurrences of work absence, making a focus on first return to work a limited and potentially misguiding index of RTW outcomes, and one that does not address the important issue of sustainability of return to work [9, 10]. Recently, Pransky et al. [7] pointed out that “despite an abundance of RTW research, the concept of RTW is often poorly defined, and there is not substantial agreement about what constitutes a successful RTW outcome.” Many studies have been focused on (first) return to work as the primary outcome measure, e.g., return to work is used as an indicator for a reduction in disability—usually with the assumption that workers who return to work are completely recovered from the disabling effects of the injury [10]. However, several studies have demonstrated that workers who return to work are not fully recovered from their initial complaints or injury [10–15]. The traditional outcome measures of return to work and time lost from work do not capture important information about the burden of injury that can be shown by self-reported measures of disability and functional limitations. Hence, to obtain a complete picture of the complex RTW process, capturing the recurrences of work absence, the persistence of disability, and their consequences for work performance, it is important to use multiple outcome measures during follow-up. Although a few studies have addressed health outcomes, such as pain, functional status, and general health, in relation to RTW status [11, 12, 14], little is known about depressive symptoms, which have been suggested to increase the total numbers of days on benefit [16] and about limitations at work in injured workers. Furthermore, it is largely unclear how injured workers “transit” in their RTW status over time. So far, we do know that a substantial proportion of workers with cumulative trauma disorders of the upper extremity [15] and compensated back pain (Côté et al., submitted) experience multiple episodes of work absence. The purpose of this study was to describe the health status, assessed by multiple outcome measures, and work limitations, in injured workers with MSK disorders 1 month post-injury stratified by RTW status, and to document their RTW trajectories over a period of 6 months post-injury. Materials and methods Study design The present study was conducted within the sampling frame of a prospective study of Ontario workers with a back or upper extremity (UE) MSK disorder, who filed a Workplace Safety and Insurance Board (WSIB) lost-time injury claim. Data was collected from two sources: self-reports of participants and WSIB administrative data. The participants were interviewed by phone at baseline (1 month post-injury), and 6 months post-injury. Participants provided information on their RTW experience, workplace, healthcare provider, insurer, and physical and mental health. From the WSIB, administrative information on sociodemographics, workplaces, and claims (e.g., site of injury, claim status, time receiving wage replacement benefits) was obtained. This information was linked to the interview data, when the participants provided written consent for linkage. Ethical approval for the study was granted by the University of Toronto’s Ethics Review Board. Participants were given the option to withdraw from the study at any point and to decline data linkage of their questionnaire responses with their WSIB data. It was made explicit to the respondents that study participation would in no way affect their claim with the insurer. Participant recruitment and final study sample Study eligibility required participants to have a new, accepted or pending, back or UE MSK lost-time claim, be absent from work for at least 5 days within the first 14 calendar days post-injury, and be 15 years of age or older. We excluded claimants with a fracture, amputation, burn, hernia, head injury, concussion, or electrocution, those who were not able to understand or speak English, and those with a security problem, who were incarcerated or received institutional care. From January to July 2005, a computer program run on WSIB weekly files resulted in the identification of 14,555 potential participants. Eligibility assessment and recruitment were conducted in three stages: at the WSIB tracking level, at the WSIB recruiting level, and the university-based research unit level (see Fig. 1). Fig. 1Recruitment procedure and flow of participants A total of 632 participants completed the baseline interview 1 month post-injury. The overall response rate was 61% (632 out of 1,038 eligible and contacted potential participants). Verbal consent for the interview data was obtained from all participants. The mean time between the date of injury and the date of interview was 29.6 days (SD 6.2; median 29 days, range 15–46 days). Approximately 98% of the participants were interviewed within 6 weeks post-injury. For the linkage with WSIB data, written consent was obtained from 479 participants, for which WSIB wage replacement data was available for 431 participants. A consent-to-linkage analysis showed that consenters (n = 479) and non-consenters were similar in terms of sociodemographic, workplace, health status, and work absence variables. However, consenters were more likely to have a higher level of education, and male consenters were more likely to be older than male non-consenters (Franche et al., submitted). Definition of the RTW status At baseline, four mutually exclusive RTW status groups were constructed, based on the workers’ responses to the following yes/no questions: “Have you gone back to work at any point since your injury (includes part-time or modified work)?” and “Are you currently working at any job right now?” The four groups were: (1) sustained first return to work (RTW-S), (2) return to work with recurrence(s) of work absence and working at time of interview (RTW-R working), (3) return to work with recurrence(s) of work absence and not working at time of interview (RTW-R not working), and (4) no return to work (No RTW). In the analyses, we collapsed the two return to work with recurrence(s) groups into one group (RTW-R). RTW status was assessed at each follow-up. Measurements: health outcomes and work limitations Pain intensity We used two items from the Von Korff Pain Scale [17, 18] to measure pain intensity. On a 10-point numerical rating scale (0 = “no pain” to 10 = “pain as bad as could be”), participants were asked to indicate their level of perceived pain from their workplace injury (1) at the present time and (2) on average in the past month. Functional status Functional disability associated with back pain was measured using the Roland–Morris Disability Questionnaire [19], a 24-item questionnaire assessing the presence of activity limitations. Responses to individual items (yes/no) are summed up and range from 0 (no disability) to 24 (severe disability). The score is averaged and—for a better comparison with scores of other instruments—transformed to a standardized score of 0–100 (by multiplying each averaged score by 100), with a higher score indicating greater disability. The Roland–Morris has been shown to have good psychometric properties [20–25]. In the baseline sample, the internal consistency (Cronbachs α) was 0.92. The 11-item QuickDASH was used to assess physical function and symptoms in participants with MSK disorders of the upper limb [26]. The QuickDASH is a shortened version of the DASH Outcome Measure [27]. The items are scored from 1 to 5. Responses to the individual items are summed, averaged, and transformed to a standardized score of 0–100, with a higher score indicating greater disability. Initial testing has shown that the QuickDASH has good psychometric properties [26]. The internal consistency in the present study was 0.90. When participants reported pain in both the back and UE, they completed both the Roland–Morris and the QuickDASH. For these participants, scores from each instrument were converted into a z-score and the highest z-score was used as the index of functional status. For participants completing only one measure of functional status, the z-score of that measure was used as the index of functional status. In addition, for those completing both measures, determination of the main pain site, i.e., back or UE, was based on the highest z-score on the Roland–Morris or the QuickDASH. General health The Short Form-12 (SF-12), a 12-item version of the SF-36, was used to measure physical (Physical Component Summary Scale Score; PCS12) and mental (Mental Component Summary Scale Score; MCS12) health-related quality of life [28, 29]. To calculate the PCS12 and MCS12, test items are scored and transformed in an algorithm to norm-based scores with a mean of 50 and a standard deviation of 10 [29]. PCS12 and MCS12 scores range from 0 to 100, a higher score indicating better health. The psychometric properties of the SF-12 are good: coefficients for test–retest reliability, measured over 2 weeks, are 0.89 (PCS12) and 0.76 (MCS12) [28]. Moreover, Luo et al. [30] reported good internal consistency, validity, and responsiveness in patients with low back pain. In the present study, the internal consistency was 0.89 (PCS12) and 0.86 (MCS12) at baseline. Depressive symptoms The 20-item Center for Epidemiologic Studies Depression (CES-D) [31] scale was used to measure depressive symptoms. The items report the frequency of occurrence of symptoms in the past week on a 4-point rating scale ranging from “rarely or none of the time” (<1 day) to “most of the time” (5–7 days). The score ranges from 0 to 60 with a higher score denoting more depressive symptoms. CES-D scores ≥16 are indicative of individuals at risk for clinical depression [31]. The internal consistency was 0.92, measured in the baseline sample. Work limitations We used the 16-item version of the Work Limitations Questionnaire (WLQ-16) to assess limitations at work due to injury or associated treatment [32–35]. The WLQ-16 covers four domains: output demands (4 items), mental demands (6 items), physical demands (4 items), and time management demands (2 items). Items are scored on a 5-point scale, ranging from “none of the time” to “all of the time.” The scores on the individual items are summed, averaged, and transformed to a standardized score of 0–100, with a higher score indicative of more limitations. The internal consistency Cronbachs α’s were 0.82 (output demands), 0.86 (mental demands), 0.78 (physical demands), and 0.76 (time management demands) at baseline. Sociodemographics, days off work, and comorbidity Participants provided information on age, gender, education, living status, number of children under the age of 18, and personal income. Information on occupational status was obtained from the WSIB database. One self-reported question assessed how many full days of work a participant had missed due to the injury. In addition, data on time receiving wage replacement benefits was obtained from the WSIB database. The Saskatchewan Comorbidity Scale was used to measure comorbidity (Jaroszynski et al., unpublished work). The 16/14-item (women/men) self-report scale assesses the presence and severity of health problems. Participants are instructed to indicate whether they currently have a particular health problem/disease and, if so how much it has affected their health in the last 6 months. The response options range from 1 = “not at all” to 4 = “severe.” In the present study, two additional items pertaining to gynecological problems and pregnancy status were added for women. Responses were combined and categorized as: no comorbidity, comorbidity with no/mild effect on health, and comorbidity with moderate/severe effect on health. Statistical analyses Univariate statistics (means, standard deviations, frequency counts) were used to describe participants, for the total cohort and by RTW status, in terms of their baseline sociodemographics, health outcomes, and work limitations. Differences in baseline characteristics between the three RTW status groups (RTW-S, RTW-R, and No RTW) were tested using a χ2 test or analysis of variance. Multiple comparisons, with RTW-S as reference group, were performed with a Tukey correction. Group differences in health outcomes and work limitations, adjusted for identified covariates were tested with GLM analyses, and multiple comparisons were performed with a Bonferroni correction. To describe the RTW trajectories, we used the self-reported RTW status at baseline and 6-month follow-up. All statistical analyses were performed with SPSS 13.0 [36]. Results Baseline characteristics and selection bias analysis A total of 632 participants, 350 (55%) men and 282 (45%) women, completed the baseline interview 1 month post-injury. Table 1 shows the sociodemographic factors for the total sample and by RTW status at baseline. The mean age of the total cohort was 42.2 years (SD 10.8) and approximately 69% lived with a partner. The mean duration of time receiving wage replacement benefits, based on WSIB data, was 19.1 days (SD 8.9; median 20 days). Sixty-six percent of the participants were primarily experiencing back pain and 34% UE pain. Table 1Baseline (1 month post-injury) sociodemographic characteristics, days off work, pain site, and comorbidity for the total study sample (n = 632) and by return-to-work status groupaTotalRTW-SRTW-RNo RTWN = 632N = 293 (46.9%)N = 88 (14.1%)N = 244 (39.0%)GenderN(%)N(%)N(%)N(%)    Female282(44.6)123(42.0)48(54.5)110 (45.1)    Male350(55.4)170(58.0)40(45.5)134(54.9)Age categories    15–29 years93(14.7)45(15.4)9(10.2)35(14.4)    30–39 years137(21.7)58(19.8)29(33.0)48(19.8)    40–49 years228(36.1)109(37.2)31(35.2)87(35.8)    ≥50 years173(27.4)81(27.6)19(21.6)73(30.0)Living with/without partner    Living with partner433(68.5)221(75.4)54(61.4)156(63.9)    Not living with partner199(31.5)72(24.6)34(38.6)88(36.1)Children under age 18    No children 341(54.0)158(53.9)48(54.5)131(53.7)    1 child118(18.7)53(18.1)17(19.3)47(19.3)    2 children118(18.7)57(19.5)15(17.0)44(18.0)    ≥3 children 55(8.7)25(8.5)8(9.1)22(9.0)Education    Some high school112(17.7)46(15.7)16(18.2)49(20.1)    High school completed177(28.0)75(25.6)30(34.1)69(28.3)    Some university or college130(20.6)70(23.9)14(15.9)45(18.4)    University/college completed213(33.7)102(34.8)28(31.8)81(33.2)Occupational status (N = 479)b    White collar89(18.6)47 (20.9)13(17.6)29(16.4)    Pink collar156(32.6)71(31.6)26(35.1)58(32.8)    Blue collar-indoor99(20.7)54(24.0)15(20.3)30(16.9)    Blue collar-outdoor68(14.2)28(12.4)10(13.5)28(15.8)    Missing67(14.0)25(11.1)10(13.5)32(18.1)Personal income    <$20,00095(15.0)31(10.6)12(13.6)47(19.3)    $20,000–39,999240(38.0)111(37.9)32(36.4)96(39.3)    $40,000–59,999180(28.5)88(30.0)28(31.8)64(26.2)    >$60,00081(12.8)46(15.7)9(10.2)25(10.2)    Missing36(5.7)17(5.8)7(8.0)12(4.9)Number of working hours/week at the time of the injury; n (%)    ≤37.5179(28.3)84(28.7)24(27.3)69(28.3)    >37.5–40.0281(44.5)128(43.7)40(45.5)110(45.1)    >40.0172(27.2)81(27.6)24(27.3)65(26.6)Duration of receiving full wage replacement benefits within first 30 days post-injurycn = 205n = 67n = 156    Days mean (SD)19.1(8.9)14.2(7.1)18.7(8.4)25.9(6.5)    Days; median20.013.019.029.0Self-reported full days off work due to injury    Days mean (SD)14.5(7.1)10.5(5.1)14.1(6.9)19.7(6.1)    Days; median14.010.013.519Pain site; n (%)    Back418(66.1)214(73.0)64(72.7)134(54.9)    Upper extremities214(33.9)79(27.0)24(27.3)110(45.1)Comorbidity; n (%)    No513(81.2)238(81.2)72(81.8)197(80.7)    Yes, with no/mild effect on health33(5.2)16(5.5)5(5.7)12(4.9)    Yes, with moderate/severe effect on health86(13.6)39(13.3)11(12.5)35(14.4)a N = 625 classified with regard to RTW status, N = 7 were working at time of the baseline interview, but were not asked item regarding recurrencesb WSIB information available for participants who provided informed consent for linkage of WSIB data with questionnaire datac N = 431 for WSIB temporary total compensation within 30 days of accident; Missing data due to (1) no informed consent for linkage of WSIB data with questionnaire data (N = 153) or (2) absence of data when claim was not a lost-time accepted claim (N = 48) To examine a possible selection bias, we compared the cohort participants (n = 632) to a group of algorithm-selected potential participants (n = 3,712) on characteristics extracted from the WSIB database, where the algorithm mimicked the inclusion criteria of our study.1 The time frame during which their injury occurred was the same as for our study sample. This analysis showed that participants were comparable to potential participants with regards to firm size, industrial sector, and income level (Table 2). However, participants were more likely to be older and female. Women aged 40–49 were more likely to participate than women in the other age categories, and older men were more likely to participate than younger men. With respect to claim status, we compared only participants with accepted claims and available wage replacement data (n = 559) with potential participants, since this data is not available for participants with pending, denied or abandoned claims. Participants were more likely to have a longer duration on wage replacement benefits at 1 and 6 months post-injury and a higher rate of wage replacement re-instatement at 6 months post-injury than potential participants, suggestive of more severe work disability in our cohort. Table 2Comparison of baseline study participants with algorithm-selected potential study participantsVariableFull baseline cohort (n = 632)Algorithm-selected potential participants (n = 3,712)Gender, n (%)    Female282 (44.6)1,365 (36.8)    Male350 (55.4)2,347 (63.2)Age at baseline interview, mean (SD) (median)42.2 (10.8) (43)40.3 (11.3) (40.7)Gender × age at baseline interview, n (%)    Females: 15–29 years40 (6.3)262 (7.1)    Females: 30–39 years59 (9.3)348 (9.4)    Females: 40–49 years109 (17.2)428 (11.5)    Females: ≥50 years74 (11.7)327 (8.8)    Males: 15–29 years53 (8.4)524 (14.1)    Males: 30–39 years78 (12.3)630 (17.0)    Males: 40–49 years119 (18.8)714 (19.2)    Males: ≥50 years99 (15.7)479 (12.9)Firm size, n (%)    <20 employees58 (9.2)335 (9.0)    20–99 employees100 (15.8)680 (18.3)    100–999 employees190(30.1)1,166 (31.4)    ≥1,000 employees150 (23.7)769 (20.7)    Schedule 2134 (21.2)761 (20.5)Industrial sector, n (%)    Automotive, manufacturing, steel109 (17.2)875 (23.6)    Service127 (20.1)812 (21.9)    Education, municipal, Schedule 2a146 (23.1)817 (22.0)    Healthcare85 (13.4)425 (11.4)    Transportation55 (8.7)333 (9.0)    Chemical/processing, electrical, food31 (4.9)212 (5.7)    Construction14 (2.2)154 (4.1)    Agriculture, forest, pulp and paper, mining15 (2.4)83 (2.2)    Unknown50 (7.9)1 (0.0)Occupational status, n (%)b    White collar109 (18.7)624 (16.8)    Pink collar209 (35.9)1,112 (30.0)    Blue collar-indoor132 (22.7)1,026 (27.6)    Blue collar-outdoor86 (14.8)631 (17.0)    Missing45 (7.7)319 (8.6)Weekly earnings in tertiles, n (%)    ≤$447.68116 (18.4)819 (22.1)    $447.68–≤$880.00281 (44.5)1,890 (50.9)    ≥$880.00157 (24.8)978 (26.3)    Missing78 (12.3)25 (0.7)Duration of time receiving wage replacement benefits (30 days post-injury), mean (SD) (median)c20.6 (9.0) (25)15.9 (9.4) (13)Duration of time receiving wage replacement benefits (180 days post-injury), mean (SD) (median)d58.7 (57.2) (33)37.6 (50.2) (14)Occurrence of re-instatement of wage replacement benefits (30 days post-injury), n (%)c    No507 (90.7)3,481 (93.8)    Yes52 (9.3)231 (6.2)Occurrence of re-instatement of wage replacement benefits (180 days post-injury), n (%)d    No463 (82.5)3,286 (88.5)    Yes98 (17.5)426 (11.5)Claim status (30 days post-injury), n (%)    Lost-time-accepted431 (68.2)2,893 (77.9)    Lost-time-pending171 (27.1)765 (20.6)    Lost-time-denied19 (3.0)43 (1.2)    Lost-time-amalgamated5 (0.8)1 (0.0)    No lost-time-accepted6 (0.9)6 (0.2)a Schedule 2 firms do not operate under the collective liability insurance principle, and, as such, are individually responsible for the full cost of the injury/illness claims filed by their workers. Schedule 2 employers include federal, provincial and municipal governments, railways, airlines, shipping, and telephone companiesb Data on the full baseline cohort is restricted to participants with accepted claims (n = 582)c Data on the full baseline cohort is restricted to participants with accepted claims and available wage replacement data 30 days post-injury (n = 559)d Data on the full baseline cohort is restricted to participants with accepted claims and available wage replacement data 180 days post-injury (n = 561)Note: Italics–The main differences between the full baseline cohort and the algorithm-selected potential participants Group differences in baseline characteristics A total of 625 participants were categorized into one of the four RTW status groups. The remaining seven participants were working when interviewed at baseline, but not asked about recurrence(s) due to an error in a skip pattern of the questionnaire, which was subsequently corrected. At baseline (approximately 1 month post-injury), 47% of the participants reported a sustained first return to work, 5% a return to work with recurrence(s) of work absence and working at time of interview, almost 9% a return to work with recurrence(s) of work absence and not working at time of interview, and 39% no return to work. With regards to gender, age, children under age 18, education, occupational status, and working hours per week at the time of the injury no statistically significant differences were observed across the three RTW groups. However, participants who had a sustained first return to work reported more often that they lived with a partner (χ2 = 11.0, p = .004) and reported a higher personal income (χ2 = 11.7, p = .069) than those with a recurrence or no return to work. These variables were used as covariates in subsequent analyses. With respect to the mean duration of time receiving wage replacement benefits, based on WSIB data, significant differences were observed across all three RTW groups (F = 122.6, p = .000). Significant differences were also seen with self-reported full days off work due to the injury (F = 169.7, p = .0000). With respect to pain site, a statistically significant difference was found across the RTW groups: participants with a sustained first return to work and those who experienced a recurrence reported low back pain more often, and participants who did not return to work reported pain in the UE more often (χ2 = 21.6, p = <.0001). A total of 81% of the participants reported no comorbidity, whereas 5% reported no/mild effects on health, and 14% reported moderate/severe effects on health, with no statistically significant group differences. Group differences in health outcomes and work limitations Table 3 presents the adjusted (for age, gender, living status, and income level) estimated means for baseline health outcomes and work limitations by RTW status, with multiple comparison results. Participants with a sustained first return to work reported significantly less pain compared to those with a recurrence and no return to work. Moreover, they also reported significantly less pain in the past month compared to those with no return to work. In participants with back pain, those with a sustained first return to work reported significantly less functional disability compared to those with a recurrence and no return to work. In participants with UE pain, we observed that those with a sustained first return to work reported significantly less functional disability compared to those who did not return to work, but not compared to those who experienced a recurrence. With regards to physical and mental health as well as depressive symptoms, participants with a sustained first return to work reported significantly better health and fewer depressive symptoms than those with a recurrence and no return to work. It is interesting to note that high levels of depressive symptomatology, indicative of being at risk for clinical depression, were found in all participants, especially in those with a recurrence and those who did not return to work. For all outcomes, there were no significant differences between participants who experienced a recurrence and those who did not return to work. With regards to limitations at work, those with a sustained first return to work and those with a recurrence, reported limitations in all domains, but mainly for physical demands and time management demands. As expected, participants with a sustained first return to work reported significantly fewer limitations than those with a recurrence. Table 3Estimated means (95% confidence intervals, CI) of baseline (1 month post-injury) health outcomes and limitations at work by return-to-work status group (N = 625), adjusted for age, gender, living status, and personal incomeRTW-SaRTW-RNo RTWN = 293 (46.9%)N = 88 (14.1%)N = 244 (39.0%)Estimated mean (95% CI)Estimated mean (95% CI)pEstimated mean (95% CI)pPerceived painRange 0–10    Pain at time of interview3.84 (3.54–4.13)5.19 (4.64–5.73).0005.94 (5.62–6.27).000    Pain in the past month8.45 (8.25–9.13)8.78 (8.40–9.16).4218.90 (8.68–9.13).014Roland Morris    Range 0–100; N = 41852.73 (49.29–56.17)68.14 (61.75–74.54).00075.66 (71.30–80.03).000Quick DASH    Range 0–100; N = 21447.58 (42.61–52.55)54.94 (46.01–63.87).46963.15 (58.98–67.33).000Physical SF-12    Range 0–10037.03 (35.98–38.09)32.75 (30.80–34.69) .00031.63 (30.47–32.78).000Mental SF-12    Range 0–10048.55 (47.18–49.91)44.40 (41.88–46.92).01445.58 (44.08–47.07).013Depressive symptoms    CES-D; Range 0–6013.16 (11.74–14.59)18.15 (15.53–20.78).00318.70 (17.14–20.26).000Work limitationsRange 0–100    Physical demands46.00 (42.79–49.22)62.81 (56.80–68.82) .000n.a.    Mental demands17.12 (14.70–19.55)29.41 (24.88–33.95).000n.a.    Output demands18.65 (15.95–21.35)35.59 (30.56–40.61).000n.a.    Time management41.83 (38.43–45.23)61.90 (55.61–68.20).000n.a.a The RTW-S group is the reference group for the multiple comparisons Attrition analysis Six months after injury, the 632 participants who had completed the baseline interview were approached again to complete the follow-up interview. Overall, 446 participants, 238 (53%) men and 208 (47%) women, completed the follow-up interview (retention rate of 70.6%). Reasons for non-response in the follow-up interview were “unable to contact” (n = 92), “avoided contact” (n = 49), and “refused to participate” (n = 45). An attrition analysis, comparing respondents (n = 446) of the 6-month interview with non-respondents (n = 186), revealed that non-respondents were more likely to be younger, to work longer hours at the time of injury, and to specify “back” as their primary pain site. Moreover, male non-respondents tended to be younger than male respondents, whereas in females differences in age were not statistically significant. Otherwise, non-respondents did not differ significantly with respect to other sociodemographic, workplace, health status, and work absence variables tested, including time receiving wage replacement benefits, re-instatement of wage replacement benefits 6 month post-injury, self-reported work absence duration 1 month post-injury, and claim status. Full details of the attrition bias analysis have been reported elsewhere (Franche et al., submitted). RTW trajectories from baseline to 6 month follow-up Figure 2 shows the RTW trajectories for 439 participants, based on self-reported RTW status at baseline and 6 month follow-up. The majority (73%) of workers with a sustained first return to work at baseline were still at work 6 months later. However, 27% had experienced at least one recurrence during that time period. All participants who experienced a recurrence remained, by definition, in this group. Of those participants who had not returned to work at baseline, 59% had a sustained first return to work 6 months later, 17% had made a RTW attempt with a recurrence, and 24% were still off work 6 months post-injury. Six months post-injury, the rate of recurrences of work absence in workers who had made at least one RTW attempt was 38% [n = 153 recurrences/(n = 439 minus n = 40 with no RTW attempt)]. Fig. 2Return-to-work trajectories based on self-reported return-to-work status at baseline (1 month post-injury) and 6-month follow-up (n = 439). aRTW-S, sustained first return to work; bRTW-R working, return to work with recurrence(s) of work absence and working at time of interview; cRTW-R not working, return to work with recurrence(s) of work absence and not working at time of interview; dNo RTW, no return to work Discussion The findings of this cohort study suggest the presence of a pattern in baseline health states and work limitations specific to RTW status, 1 month post-injury. Workers with a sustained first return to work reported less pain, less functional disability, better physical and mental health, fewer depressive symptoms, and fewer work limitations compared to those who experienced a recurrence of work absence or who never returned to work. The study adds to the literature by demonstrating that depressive symptoms and limitations at work are prevalent in workers 1 month post-injury, including in those with a sustained first return to work. A substantial rate of recurrences of work absence over 6 months was found (38%), even in workers who had initially made a sustained first return to work at baseline (27%). Moreover, of those workers who did not return to work at baseline, 17% attempted to return and experienced a recurrence within 6 months of the injury, and 24% were still off work at 6 months post-injury. Our findings are consistent with previous research suggesting that a return to work does not translate into a complete recovery from a MSK disorder [11, 12, 14, 15]. For example, in a study among 205 workers with MSK injury, Evanoff et al. [14] found that 88% of the workers had returned to work (with 83% working at full duty) at 6 month follow-up, and of these 24% reported continuing disability due to the injury (including 20% of those working at full duty). In our study, participants with a sustained first return to work at 1 month post-injury had SF-12 physical and mental health subscales scores below the healthy population average of 50 [29], and high levels of work limitations, specifically for physical demands and time management demands. Thus, the results suggest that many workers, while back at work, still have health problems and experience difficulties in meeting their work demands. Furthermore, the importance of measuring multiple health outcomes, in particular depressive symptoms, was shown by the high levels of depressive symptoms, indicative of being at risk for clinical depression, in injured workers suffering from MSK disorders, particularly in those who experienced a recurrence and those who did not return to work. Our finding of high levels of depressive symptoms in injured workers is in line with earlier studies showing that depressive symptoms are prevalent in MSK-injured workers [37, 38] and highlights the need to address and examine the mental health of workers suffering a workplace injury. Moreover, 6 months post-injury we found a high rate of recurrences of work absence (38%) in workers who had made at least one RTW attempt, which is consistent with studies suggesting that there is considerable movement in and out of work after the first return to work [15 Côté et al., submitted]. In a study among 1,321 US workers who filed a workers compensation claim for back pain, Côté et al. (submitted) observed that 23% and 30% of the workers, 6 and 12 months post-onset of back pain, respectively, experienced multiple work absences. When interpreting the results, the following methodological issues must be considered. Though reasonable for a study among claimants, the overall participation rate of 61% raises the question of selective participation, which may have biased the results. However, the cohort was shown to be representative of the most comparable claimant group with regards to basic demographic and workplace variables, but not with regards to duration of time receiving wage replacement benefits and rates of wage replacement re-instatement, suggesting the presence of more severe disability in the cohort. Hence, the generalizability of the results remains limited with respect to workers with less severe work disability. More importantly, the rates of self-reported recurrence of work absence may be inflated in our cohort. A related issue concerns the loss-to-follow-up of 29%. The attrition analysis demonstrated that non-respondents and respondents were similar with regards to time receiving wage replacement benefits, the occurrence of re-instatement of wage replacement benefits, and self-reported work absence duration. Non-respondents were younger males, worked longer hours, and were more likely to specify “back” as their primary pain site compared to respondents. Future research should further explore the relationship between recurrence(s) of work absence, health outcomes, and work limitations over an extended period of time. We found that workers who experienced a recurrence after a first return to work clearly report more health problems and work limitations than those with a sustained first return to work, and their health status is often comparable to workers who do not return to work. Our trajectory analyses were based on a 6 months time window and it was not yet possible to examine multiple recurrences and their effects on health outcomes and work limitations over a longer period of time. However, 12 month follow-up data will make such analyses possible in the future. Moreover, we have to examine important and meaningful changes in health outcomes and work limitations between baseline and follow-up across the possible RTW trajectories [39], and also study them in relation to a broad range of factors (e.g., RTW interventions, disability management strategies) that might have influenced the RTW process and the outcomes considered. Finally, future research should identify early prognostic factors of the trajectories, particularly focusing on the “problematic” trajectories (recurrences and persistent work absence), so that guidance for an optimal reintegration or for recurrence prevention can be provided. To conclude, the results of this prospective study suggest that workers who had a sustained first return to work report a better health status and fewer work limitations than those who experienced a recurrence after a first return to work or who did not return to work. However, it is also demonstrated that a return to work is not equivalent to a complete recovery from MSK disorders. Workers who had a sustained first return to work still reported meaningful health limitations, as compared to the general population and significant work limitations. Given the considerable impact of disability on worker health and costs for the workers, employers, and society, the findings highlight the importance of considering multiple health outcomes, including depressive symptoms and work limitations, when studying the complex process of return to work and when developing RTW interventions and disability management strategies.

Eur_Spine_J-2-2-1602204

Dorsal dumb-bell melanotic schwannoma operated on by posterior and anterior approach: case report and a review of the literature

The schwannomas are benign tumors originating from Schwann cells which constitute the nerve sheath. The dorsal pigmented type of schwannomas is relatively rare (Goldstein in Am J Med Genet 86:62–65, 2004; Kurtkaya-Yapicier in Histol Histopathol 18(3):925–934, 2003). There are two distinct types of melanotic schwannoma: the sporadic melanotic schwannomas and the psammomatous melanotic schwannomas of Carney complex. We report a case of a patient harboring a right dorsal dumb-bell melanotic schwannoma and left adrenal mass. The patient underwent a surgical procedure for en bloc total removal of the mass by a posterior and anterior approach. Histopathological examination revealed the diagnosis of melanotic schwannoma. At present, we have no reliable marker of histopathological malignancy of melanotic schwannoma so the follow-up period ought to continue for a period of more than 5 years. As the total removal of the melanotic neurinomas is mandatory to prevent possible malignant transformation of the tumor recurrence or regrowth, we believe that the posterior and trans-thoracic approach are the most suitable one for the dorsal dumb-bell schwannoma. Introduction Dorsal dumb-bell melanotic schwannomas are very rare [13]. They are rather frequently associated to Carney’s complex [8–10]. They present problems on different aspects, such as pre-operative diagnosis, surgical approach and post-operative course. We describe the case of a 53-year-old man who presented with left adrenal hyperplasia and right paravertebral mass suggesting a dumb-bell schwannoma. Case report Presentation and examination In January 2003, a 53-year-old salesman, heavy smoker with an elevate level of arterial hypertension, began to complain about pain in his right chest and upper limb. Chest X-ray and thoracic CT scan in the following February revealed a D9–D10 dumb-bell mass extending into the intrathoracic right space for about 2.5 cm. The patient was admitted to our clinic. A spinal dorsal MRI confirmed the presence of the voluminous D9–D10 right paravertebral mass (Figs. 1, 2). Abdomen MRI showed an almost 4 cm swelling of the left suprarenal gland, suggesting a pheochromocytoma (Fig. 3). The neurologic examination and brain MRI were normal. The renal arterial echodoppler, the dosing of vanilmandelic acid, alpha-fetoprotein, carcinoembryonic antigen, urinary catecholamine and cortisol were normal. The renin activity and the plasmatic level of aldosteron resulted normal in orthostasis and increased in recumbency. The medullar adrenal gland scintigrams (123 IIs MIBG-148 MBqs e.v.) resulted negative for pheochromocytoma. Therefore, we excluded a pheochromocytoma and the patient underwent a surgical procedure for total removal of the dorsal mass by posterior and anterior approach in collaboration with the thoracic surgeons. Fig. 1MR images demonstrating a right dumb-bell intra-extradural mass at T9–T10 level extending into the intrathoracic right space for about 2.5 cm. Coronal and axial T1-weighted images with Gd-DTPA enhancementFig. 2Sagittal T2-weighted MRI showing T9–T10 right intervertebral foramen occupied by the tumorFig. 3Coronal abdominal spectral fat saturation inversion recovery (SPIR) MR images showing left suprarenal swelling Operation A D9–D10 laminectomy after radioscopic control was performed. A right dumb-bell intra-extradural mass of blackish aspect was seen. En bloc total removal of a markedly melanotic intradural lesion was achieved using operating microscope. Then the patient was placed in the left lateral decubitus and by trans-thoracic approach the paravertebral mass was radically removed en bloc, also freeing the right D9–D10 intervertebral foramen from the black and well-encapsulated neoformation. Accurate plastic of the paravertebral pleural wall was obtained. Histological examination revealed melanotic schwannoma with mitotic index not superior to one for ten fields on high magnification indicating low risk for malignancy. Postoperative course Post-operative course was uneventful, with marked improvement of the painful symptomatology and good healing of the surgical wounds. The patient was discharged from our ward after 10 days, with a recommendation for ichnographic controls of the left adrenal gland hyperplasia. During a 2-year follow-up period, no recurrence was evident on surveillance MRI. Discussion The schwannomas are benign tumors originating from Schwann cells which constitute the nerve sheath. The first case was described by Bjornboe [13, 15] in 1934 in a patient with neurofibromatosis type I. There are two distinct types of melanotic schwannoma: the sporadic and the psammomatous melanotic schwannomas of Carney complex [8, 14, 24, 27, 29, 36]. In 1985, Carney described a syndrome characterized by mixomas (heart, skin and breast), mucus-cutaneous spotty pigmentations and endocrine overactivity (due to thyroid dysfunction or hypophysial adenoma or Cushing syndrome secondary to adrenocortical pigmented primary nodular hyperplasia) [8, 9]. A year later, a Mendelian dominant heredity was defined responsible for the symptomatologic complex [43]. About 50% of patients with melanotic psammomatous schwannomas are affected by the syndrome of Carney and one-fifth of them are carriers of multiple lesions [21, 42]. Generally, melanotic schwannoma affects the spinal nerves, the central nervous system and the autonomous nerve system, but extra-nervous locations can also exist. The dorsal location represents 30.5% of the spinal melanotic schwannoma [41]. Until today we were able to find 47 spinal cases of extramedullary melanotic neurinoma in literature (Table 1) [1, 2, 3, 5, 6, 7, 11, 12, 13, 16, 18, 20, 21, 26, 28, 30, 31, 33, 34, 35, 38, 39, 44, 45]. The clinical presentation is not specific but similar to that of other extramedullary spinal tumors (radicular pain, dysesthesias, progressive sensorial-motor deficits). Preoperative diagnosis is based on magnetic resonance imaging (MRI): melanotic schwannomas appear hyperintense on T1-weighted sequences and hypointense on T2-weighted sequences compared to the nonmelanotic type of schwannomas that are hypo-isointense on T1 and hyperintense on T2 [1, 13, 17, 25, 37]. Table 1Review of 47 cases of spinal extramedullary melanotic schwannoma reported in literatureReferencesAgeGenderLevelAprile et al. [1]70FL3Bagchi et al. [2]40MT6–T7Belak et al. [3]44–DorsalBosman et al. [4]43ML4–L5Bouziani et al. [5]46MLumbarBuhl et al. [6]28ML5–S1Bunc et al. [7]––T12–L2Cornejo et al. [10]36ML3–L4Di Gregorio et al. [11]––SacralErlandson et al. [12]36ML5–S1Goasguen et al. [13]66FC2–C3Graziani et al. [15]40FT3Gregorios et al. [16]45FT2Iizuka et al. [18]58FT10Killeen et al. [19]26FS1Krichen et al. [20]27MC6–C7Kuchelmeister et al. [21]53FC5–C6Le Cam et al. [23]35FL4Leger et al. [24]36MC4Lowman et al. [26]17FT12–L1Lowman et al. [26]26FC6Ludvikova et al. [27]56FSpineLudvikova et al. [27]58MSpineMandybur et al. [28]59MT7Martin-Reay et al. [29]32MSacrumMcGavran et al. [30]12FT2McGavran et al. [30]49FC8Mennemeyer et al. [31]23FL1Mennemeyer et al. [31]25MT7Mennemeyer et al. [31]36MS1Parent et al. [33]63FS1Paris et al. [34]49FC8Parker et al. [35]18FDorsalPrieto-Rodriguez et al. [36]38MT5Schmitz et al. [38]42FCervicalTawk et al. [39]61MT7Vallat-Decouvelaere et al. [41]5 casesF and MC–T–L–SZhang et al. [44]5 casesF and MSpinal nerve rootZonenshayn et al. [45]27FL2 and L4–L5 The histological differential diagnosis with the other spinal pigmented tumors is with the meningeal melanocitoma, the pigmented neurofibroma, the paraganglioma, the ganglioneuroma, the fibrous form of meningioma and above all with the primitive melanoma or relative metastasis [13, 15]. Even if the association of the left adrenal swelling and arterial hypertension with the dumb-bell thoracic melanotic schwannoma can be correlated to the Carney complex, in our opinion our patient did not harbor such a syndrome. In fact only the adrenocortical pigmented primary nodular hyperplasia is a characteristic stigmata of Carney complex. Moreover, our patient did not show cutaneous signs of neurofibromatosis and his family history was negative for Carney complex stigmata. The total body CT scan showed the left adrenal mass and excluded any other typical lesion of the syndrome. The increased plasmatic level of renin and aldosteron in recumbency in our patient, that was on heavy medical therapy for high level of arterial blood pressure, may be correlated to left adrenal hyperplasia or solitary adrenal adenoma [40] and not to adrenocortical pigmented nodular hyperplasia, which is the cause of ACTH-independent adrenal Cushing’s syndrome [8] and normally do not trigger elevation of plasmatic aldosteron. Moreover, in our case the histological examination of the neurinoma did not show the most common, psammomatous aspect of Carney complex [8, 13]. In summary, we did not perform a left adrenal biopsy as we ruled out the Carney syndrome and malignancy of the adrenal mass by clinic and radiological investigations. Hypertension, adrenal mass and dumb-bell melanotic schwannoma do not imply Carney complex. In our patient the hypertension was not correlated to the Carney complex, but to the hyperaldosteronism caused by left adrenal hyperplasia or solitary adrenal adenoma. Nevertheless, neurosurgeons ought to keep in mind the connection between the melanotic schwannoma and the Carney complex, since more than 50% of the cases described in literature have been related to such syndrome. A complete endocrinological evaluation and a total body CT should always be performed together with a genetic study in the suspicion of Carney complex. The risk of possible malignant transformation of the neurinoma should also be taken into consideration as cases of recurrences or metastases have been described even 13 years after surgery [4, 6, 19, 21, 23, 32, 41]. It is not possible to have any reliable histopathological indication of possible malignant transformation after the first surgical operation [42]. The follow-up ought to continue for a period of more than 5 years. We have both to check the primitive tumor regrowth or recurrence and the possible remote locations, particularly in the lungs. From the surgical point of view the total removal of the melanotic neurinomas is mandatory to prevent possible malignant transformation of the tumor remnants. The therapeutic potential of radiotherapy is not proven [42]. The prognosis depends, therefore, almost entirely on the extent of surgical removal of the tumor [3, 44]. So we believe that the posterior and anterior approach is the most suitable one to remove the dorsal dumb-bell schwannoma completely, both in the thoracic cavity and in the intervertebral foramen, and to avoid possible liquoral leakage.

Eur_J_Health_Econ-_-_-1388082

Determining the “Health Benefit Basket” of the Statutory Health Insurance scheme in Germany

The issue of defining health benefit catalogues has recently gained new importance in Germany as a result of the creation of the new Institute for Quality and Efficiency. The Institute was designed to support the Federal Joint Committee conducting effectiveness studies for benefit coverage decisions. The Committee and the contractual partners (sickness funds and providers) define the benefit catalogues for the Statutory Health Insurance in the framework of Social Code Book V, Germany’s most relevant health care scheme. Unlike other countries, the German federal government limits its regulatory role to defining procedures that determine the scope of Statutory Health Insurance services. The explicitness of the benefit catalogues varies greatly between different sectors. While benefits in outpatient care are rather explicitly defined, benefit definitions for inpatient care are vague. It is argued that the establishment of the new Institute and the development of the DRG system are initial steps towards a more effective and explicit benefit catalogue. Organizational structure and actors involved in the definition of benefit catalogues A fundamental aspect of the German health care system is the sharing of decision-making powers between the federal government, the individual states, and designated self-governmental institutions. Responsibilities are traditionally delegated to membership-based, self-regulated institutions of payers and providers that are involved in financing and delivering health care. In the largest scheme (which covers 88% of the population), the Statutory Health Insurance (SHI), sickness funds, their associations and associations of SHI-affiliated physicians and dentists are recognized as quasipublic corporations. These corporatist bodies constitute the self-regulated structures that operate the financing and delivery of benefits covered by the SHI scheme within the legal framework of the Social Code Book (SGB) V [1]. In joint committees of payers (associations of sickness funds) and providers (associations of physicians and/or dentists and/or the Hospital Federation) legitimized actors define benefits, prices, and standards (federal level) and negotiate horizontal contracts to control and sanction their members (regional level). The vertical implementation of decisions taken at senior levels is combined with strong horizontal decision making and contracting among the legitimized actors involved in the various care sectors [2]. Physicians treating SHI-insured patients are organized into 17 regional physicians’ associations. The Federal Association of SHI Physicians is responsible for cooperation on the federal level. SHI-accredited dentists are organized the same way as physicians through 17 dentists’ associations and the Federal Association of SHI Dentists. The German Hospital Federation is also involved in the decision-making process. The payers’ side is made up of autonomous sickness funds organized on a regional and/or federal basis. They are obliged to raise contributions from their members and to determine the contribution rate necessary to cover expenditures. Their responsibilities include contracting, negotiating prices, quantity and quality assurance measures. Services covered by such contracts are usually accessible to all fund members without any prior approval by the fund, except for preventive spa treatments, rehabilitative services and short-term home nursing care. If there is any doubt, the sickness funds must obtain an expert opinion on the medical necessity for treatment from the Medical Review Board, which serves as a joint institution of the sickness funds. The most important body in the benefit negotiations between sickness funds and physicians concerning the scope of benefits is the Federal Joint Committee. Based on the legislative framework the Committee issues directives relating to all sectors of care. The main body of the Committee consists of nine representatives of the federal associations of sickness funds, nine representatives from provider groups, two neutral members with one proposed by each side, and a neutral chairperson-accepted by both sides. In addition, nine nonvoting representatives of formally accredited patient organizations have the right to participate in consultations, and to propose issues to be assessed and decided upon. The directives of the Committee are legally binding for all actors in the SHI scheme. These directives primarily concern the coverage of benefits and assure that SHI services are adequate, appropriate, and efficient. The actual criteria defining benefits vary widely between sectors and types of catalogues. The most important benefit catalogues in the German SHI scheme and its underlying criteria are displayed in Tables 1 and 2. This article concentrates on HC1 (services of curative care) of the International Classification for Health Accounts (ICHA) taxonomy [3].Table 1 Catalogues and criteria used for defining benefitsLegal statusDecision maker(Original) purposePositive/negative definition of benefitsDegree of explicitnessaIf itemized: goods, procedures only; linked to indicationsUpdatingCriteria used for defining benefitsNCECEBOtherSHI GBR LawParliamentEntitlementsP/N1/2–When necessary+–––––SHI FJC general directivesDirectiveFJCEntitlementsP2–When necessary+++(+)b––SHI FJC special directives: positiveDirectiveFJCEntitlementsP2/3Goods, procedures, indicationWhen necessary+++(+)b––SHI FJC appendices to directives: negativeDirectiveFJCEntitlementsN3Goods, procedures, indicationWhen necessary+++(+)b––SHI DRGContractCommittee on Hospital PaymentReimbursementP3ProceduresEvery year+++–––SHI EBMContractValuation Committee (Physicians)ReimbursementP3ProceduresWhen necessary+++–––SHI BEMAContractValuation Committee (Dentists)ReimbursementP3ProceduresWhen necessary+++–––SHI BEL-IIContractValuation Committee (Dentists)ReimbursementP3GoodsWhen necessary+++–––Statutory long-term care insurance: GBRLawParliamentEntitlementsP/N1–––+––––FJC Federal Joint Committee, SHI Statutory Health Insurance, GBR General Benefit Regulation, DRG diagnosis-related group, EBM Uniform Value Scale, BEMA Uniform Value Scale–dentists, BEL-II Uniform Value Scale–dental technicians, N need, C costs, E effectiveness, CE cost-effectiveness, B budgeta 1, “all necessary”; 2, areas of care; 3, itemsb Although explicitly mentioned, it is provided principally for medical devicesTable 2 Benefit-defining laws/ decrees and cataloguesStatutory Health Insurance (SHI) - general benefit regulationSHI - general directives of the Federal Joint CommiteeSHI - special directives of the Federal Joint Commitee (positive)SHI - appendices to directives of the Federal Joint Commitee (negative)SHI - DRGSHI - EBMSHI - BEMASHI - BEL-IIStatutory long term care insurance - general benefit regulationCatalogue: type of document, actors and contentsHC.1.xxx (X)xHC.1.2xxx (X)xHC.1.3.1xxx (XI)xHC.1.3.2xxx (I)x (XII)xxHC.1.3.3xxx (II)x (XI)xHC.1.3.9xxx (III)x (XI)HC.2.1/2.2xxHC.2.3xHC.3.1xxHC.3.2xxHC.3.3xx (IV)xHC.4.1xxxHC.4.2xxxxHC.4.3xx (V)HC.5.1.1xxHC.5.1.2xxx (VI)HC.5.2xxx (VII)HC.6.1xxx (VIII)HC.6.3xHC.6.4xxx (IX)xxHC.6.5xFJC Federal Joint Committee, SHI Statutory Health Insurance, GBR General Benefit Regulation, DRG diagnosis-related group, EBM Uniform Value Scale, BEMA Uniform Value Scale–dentists, BEL-II Uniform Value Scale–dental technicians, I Directives on the Provision of Prosthetic Services II Directives on Psychotherapy III Directives on Nonphysician Care IV Directives on Home Nursing Care V Directives on Patient Transport VI Directives on OTC VII Directives on Medical Aids VIII Directives on Maternity Care IX Directives on A. Early Detection of Cancer, B. Dental Prophylaxis & C. Medical Examinations for the Early Detection of Diseases X Appendix to Directive according to SGB V, Sect. 137c (to evaluate hospital procedures) XI Appendix to Directive on Medical Procedures, XII Appendix to Directive on New Dental Procedures In-patient curative care If curative care (i.e., to detect, cure, prevent the worsening, or relieve the discomforts of accompanying diseases) cannot be achieved by ambulatory treatment (SGB V, Sect. 39), the insured party is entitled to inpatient treatment in accordance with SGB V, Sect. 27. This health care entitlement is linked to a copayment of €10 per calendar day to a maximum of 28 calendar days per year [SGB V, Sect. 39(4)1]. Hospital services are granted in accordance with the care ability of each hospital and with the level of care assigned to each hospital. In each individual case the provision of services needs to be suitable and adequate for the insured. This includes medical treatment, nursing care, the provision of pharmaceuticals, cures and therapeutic appliances, as well as board and accommodation [4]. Hospital care may be only provided in hospitals included in the hospital plan of the respective federal state, in university hospitals, or in hospitals that have concluded a service provision contract with the sickness funds (SGB V, Sect. 108). While the spectrum of services provided by the respective hospitals is indirectly determined by the hospital plan (which also determines governmental subsidies for investments), the reimbursement for the provided services is decided in negotiations between each hospital and the association of sickness funds. The Federal Joint Committee presides over matters of exclusion of health care services, and/or the evaluation and examination of treatment methods; the Committee handles these matters in response to requests from the federal associations of sickness funds and the German Hospital Federation. The method under examination will be scrutinized as to its suitability to provide adequate, expedient, and economical care for the insured persons, with general state-of-the-art medical knowledge taken into consideration. Should the examination reveal that the method does not meet the aforementioned, it may no longer be provided at the expense of the SHI system. In such instances, the Federal Joint Committee issues a corresponding directive according to SGB V, Sect. 137c (1) (see Fig. 1). Health care services in the framework of clinical studies are not subject to the directive. This means that all health care services that are not excluded by a directive of the Federal Joint Committee may be provided at SHI’s expense.Fig. 1 DRG Case Fees Catalogue for inpatient care The SHI Reform Act 2000 required the selection and implementation of a case fee system for reimbursement effective as of 1 January 2003. On 27 June 2000 the federal associations of sickness funds, the Association of Private Health Insurance, and the German Hospital Federation adopted the Australian system of diagnosis-related groups (DRGs) as the basis for developing a German DRG system. On 10 May 2001 they founded the Institute for the Payment System in Hospitals (InEK) which is intended to support the introduction and the further development of the DRG system. The InEK is controlled and supervised by the Committee on Hospital Payment, an institution consisting of representatives from the contracting partners (see Fig. 1). The matters addressed by the DRG Institute consist of defining the DRG case groups, the maintenance of the DRG system, and its severity classification system, the development of a coding directive and proposals for adapting German modifications of the International Classification of Diseases ICD-10 and the Operating Procedures System (OPS) into the DRG system. The Institute is also responsible for the calculation of DRG cost weights and individual adjustments within the DRG system. As the basis for the new pricing system a uniform case fee catalogue with fixed payments for services and benefits, valid throughout Germany, was developed. The catalogue lists all procedures (services) performed in hospitals in accordance with respective clinical diagnoses. At the same time the DRG system constitutes the catalogue of services and benefits covered by the SHI scheme for inpatient care. The inclusion of new health care services in the DRG system is reflected at the beginning of each year when a new version of the OPS and the ICD-10 is made available and is linked to a DRG [5]. The Case Fees Catalogue of 2005 consists of 876 DRGs, of which 33 are not remunerable with a case fee, and an additional list of 71 negotiated extra remunerations. The German DRG system is subdivided into 23 major diagnosis categories (MDCs) which refer in principle to a body system or cause of a disease. The MDC category also defines the first of the four digits of a DRG. The second and third digits of a DRG indicate the respective partition. The partition differentiates between surgical procedures (01–39), other procedures (40–59), and medical (conservative) procedures (60–99) carried out during a hospital stay, thus linking a DRG to benefits provided in a hospital. The fourth digit further subdivides a DRG according to a patient’s clinical complexity level, which is comprised of such factors as complexity of secondary diagnoses, cause of discharge and patient gender [6]. For inpatient services not covered by the DRG system (e.g., new methods of treatment), agreements are made with the hospitals concerned. The local contractual partners inform the contract partners at the federal level of such agreements, who may then decide to initiate an evaluation process in order to exclude these new services from the benefit package [SGB V, Sect. 137c; V, Sect. 6 (2), Hospital Payment Act]. In principle however, as noted above, all health care services that are not explicitly excluded by a directive of the Federal Joint Committee can be provided at the expense of the SHI. Outpatient care The provision of medical and dental care must be regulated and secured by agreements between the respective regional physicians’ association/regional dentists’ association and the regional associations of the sickness funds (SGB V, Sect. 72). Whereas, in accordance with SGB V, Sect. 137c, medical care in hospitals shall be, “adequate, expedient and cost-effective”, for ambulatory care, in accordance with Sect. 135, the criteria to be applied are “diagnostic and therapeutic expedience, medical necessity and cost-effectiveness.” Thus the inclusion and/or exclusion of health care services from the benefit catalogues differ in the two sectors. In the outpatient sector a service provided must be confirmed to fulfill the criteria “expedience, necessity and cost-effectiveness” in order to be included into the catalogue of services and benefits. In contrast to that, health care services in the inpatient sector is excluded from the benefit catalogue of the sickness funds only if the criteria are proven to be unfulfilled. For this reason it is possible that the health care services provided in the inpatient sector are not included in the benefit catalogue of the outpatient sector [7]. Basic medical and diagnostic care Insured persons are entitled to preventive care, detection, and treatment of diseases [SGB V, Sect. 28 (1)]. This entitlement also embraces complementary services by nonphysicians and practitioners, provided that they are prescribed by a physician. The legislative authority, however, does not define in detail the entitlements of the insured persons, but regulates the procedures with which the institutions of self-governance and the contractual partners determine the scope of SHI services [8]. In accordance with SGB V, Sect. 92 (1), the Federal Joint Committee issues directives in respect of adequate, expedient and cost-effective medical care for the insured persons. The directives consist of a general part that explains their aim, their users and mentions the corresponding paragraph in the SGB V. After the initial section the directives become more detailed. For example, the Directive on Medical Procedures that regulates the in- and exclusion of benefits in the outpatient sector initially defines the term of a new service and the conditions an evaluation is depended upon. Thereafter it is stated that the regional physicians’ associations, the Federal Association of SHI Physicians and the federal associations of sickness funds have the right to propose services for their inclusion. Then the criteria for the inclusion of services, the classification of evidence and the decision-making process are described in detail. The services included or excluded through the evaluation process are listed in the annexes [9]. While the Federal Joint Committee decides on the in- and exclusion of services into the benefit package, the Valuation Committee, which consists of seven representatives of the Federal Association of SHI Physicians and representatives of the federal associations of sickness funds, defines the actual benefit catalogue for the insured, the Uniform Value Scale (EBM). The EBM defines, as an integral component of the Federal Framework Contract–Physicians (BMV-Ä), the scope of medical care to be provided under the SHI throughout Germany. If the Valuation Committee fails to reach a consensus, at least two of its members or the Federal Ministry for Health and Social Security may demand that the extended Valuation Committee in accordance with SGB V, Sect. 87 (4), be brought in to resolve a split decision. Resolutions are to be submitted to the Ministry of Health, which, in the event of unresolved objection, may define alternative executions. The BMV-Ä is concluded between the Federal Association of SHI Physicians and the federal associations of sickness funds (SGB V, Sect. 82). In addition to the scope of health care provided under the SHI, the BMV-Ä regulates participation in ambulatory care, the pertinent aspects of quality assurance, and entitlement to benefits. Thus the EBM and the directives of the Federal Joint Committee are both integral parts of this contract. In Sect. 2 of the BMV-Ä, the description of a service in the EBM is stipulated as a condition for the provision of the respective service. As a result, the EBM constitutes the catalogue of services and benefits covered by the SHI (see Fig. 2).Fig. 2 Uniform Value Scale for physicians for outpatient care The broad structure and the contents of the EBM are stipulated in SGB V, Sect. 87 [10]: (a) The EBM displays the health care services covered by the SHI scheme and their monetary value in relation to one another in the form of a points system. (b) A basic remuneration for general practitioners is defined. (c) Health care services are grouped into packages of similar services. (d) Differentiation is made between the health care services to be provided exclusively by general practitioners and those to be provided exclusively by specialists. (e) The respective health care services are assigned exclusively to the groups of specialists that are allowed to provide them. The EBM catalogue is structured into six main chapters and various sections. Chapter I describes general regulations regarding the provision and reimbursement of health care services. Chapters II–IV contain health care services related to different physician groups and/or special criteria. Chapter V lists the general health care services provided by most physicians reimbursed with case fees. Chapter VI contains appendices (e.g., a list of services which are already contained in other services and are therefore not reimbursed additionally) [11]. As an appendix to the BMV-Ä there is an agreement that applies to care provided by general practitioners under SGB V, Sect. 73. It defines the provision of medical treatments and the early detection of diseases. The definition of individual services to be provided is included in the EBM. In addition to these central agreements, which are uniform for all sickness funds, there are numerous “small” contracts determining the scope of the health care services covered by the German SHI scheme. Outpatient dental care While benefits for ambulatory physician services are legally defined in generic terms only, legislation regulating dental care is much more detailed in the SGB V. One reason for this is that the respective committee of the joint institutions until 2003 failed to provide more explicit definitions [12]. The basic entitlements of the insured to dental care are defined in SGB V, Sect. 28 (2): The insured are entitled to prevention, early detection, and treatment of diseases of the teeth, the mouth, and the jaw. Consequently only prophylactic treatment, basic dental care, and dental prosthetic services are covered by the sickness funds [13]. Similar to the definition of benefits for basic medical care, the directives of the Federal Joint Committee broadly define when patients are entitled to a benefit. However, they do not define specific items that must be included. Therefore the Dental Valuation Committee, which consists of representatives of the federal associations of the sickness funds and the Federal Association of SHI Dentists defines the Uniform Value Scale for Dentists (BEMA; see Fig. 3). The BEMA lists services that are reimbursed by the sickness funds, thereby explicitly defining the SHI benefit catalogue. The services of dental technicians producing the material needed for orthodontic or prosthetic services are listed in a similar framework, the Uniform Value Scale for Dental Technicians (BEL-II) which is negotiated by the same Committee.Fig. 3 Uniform Value Scales for dentists and dental technicians Orthodontic treatments, except those for the treatment of abnormalities, are to begin during childhood and are excluded for insured parties over the age of 18 years (SGB V, Sect. 29). To prevent overprovision of services dentists must prepare a cost schedule that is reviewed by the sickness funds. Prosthetic services are only partially covered by the sickness funds and are therefore defined more explicitly. The insured receive a so-called “subsidy” as a percentage of a “standard” treatment, defined by the Federal Joint Committee in a directive according to SGB V, Sect. 56. The directive currently in force defines a standard treatment for 52 findings. For each standard treatment all reimbursable services of the dentists and the dental technicians are listed separately according to the BEMA and the BEL-II [14]. Sickness funds usually cover 50% of the standard treatment costs. This proportion can increase to 70% or 80% if a patient can prove yearly preventive dental checkups over the past 5 or 10 years, and the patient’s efforts for dental hygiene are observable. Higher payment levels, up to full coverage of the costs of the standard treatment, are provided only for persons of very low income. Patients are free to choose nonstandard treatments [SGB V, Sect. 55 (5)] or include additional services [SGB V, Sect. 55 (4)]; however, the amount of sickness funds’ payments remains unchanged. Outpatient care performed by nonphysicians The term “cures” subsumes health care services in Germany that are provided by nonmedical practitioners, which include professional, recognized therapists, such as physiotherapists and occupational therapists [15]. The entitlement of the insured to cures is found in SGB V, Sect. 32. It is limited by copayments for insured parties over the age of 18 years under SGB V, Sect. 61 (3). A further limitation on entitlements is imposed under SGB V, Sect. 34 (4), “Excluded Pharmaceuticals, Cures and Medical Aids.” The Ministry of Health is entitled to exclude cures from the catalogue of services and benefits covered by the SHI through decrees, with the approval of the Federal Council (upper chamber of the federal Parliament). However, a corresponding legal decree does not exist at present. The scope of services covered by the SHI scheme is explicitly described and regulated by the Directive on Non-physician Care issued by the Federal Joint Committee under SGB V, Sect. 92 [16]. The prescription of more cost-effective measures with equal efficacy, for example, drugs and other therapeutic appliances that achieve the same therapeutic objective, is to be given precedence. The benefits are listed in the directive in connection with an indication. New benefits and/or an extension of the indications for a given benefit may only be prescribed after the Federal Joint Committee has recognized their therapeutic value and included them into its directive (SGB V, Sect. 138). The federal associations of sickness funds and representatives of non-physicians compile a Catalogue of Non-physician Care. The catalogue facilitates the implementation of the directive on Non-physician Care issued by the Federal Joint Committee (see Fig. 4; SGB V, Sect. 125), which regulates: (a) the content, scope and frequency of cures, (b) further training measures and quality assurance (c) the content and scope of collaboration between non-physicians and the prescribing SHI physician, (d) measures to meet the aim of cost-effectiveness, and (e) specifications for remuneration structures.Fig. 4 Catalogue of Non-physician Care Conclusions Despite the existence of various catalogues and directives for the SHI scheme, the benefit package is not defined in detail because the obligation of the catalogues and their explicitness varies largely. Inpatient services not listed in the DRG catalogue can still be covered by the SHI scheme as long as they are not explicitly excluded by the Directive according to SGB V, Sect. 237c. However in the ambulatory sector only those procedures listed in the SHI EBM or in the SHI BEMA are covered as benefits in the outpatient sector. With the exception of the Catalogue of Non-physician Care the benefits described in the DRG, EBM, BEMA, and BEL-II are the aggregate results of decisions taken at various levels, and they are not linked to specific indications. The reason for this is that they were originally defined for reimbursement and were not meant to define the SHI benefit basket in full detail. For example, as DRGs aggregate multiple procedures and diagnoses, benefits (medical procedures) provided under one DRG will vary from case to case. Additionally, the patient clinical complexity level of a DRG is determined by diagnoses including comorbidities, gender and cause of discharge and not on the basis of the actual services provided. Therefore the scope of a DRG is very broad. Conversely, the development of a DRG catalogue can also be seen as a starting point towards a more explicitly defined benefit catalogue, and subsequently lead to benefit catalogues where all approved interventions are listed and grouped around the relevant diagnoses [2]. In recent years strong efforts have been made by the German government to move towards a more explicitly defined benefit basket. The creation of the Federal Joint Committee out of four smaller committees for the different sectors of care can be considered an improvement. The number of issued directives since the inception of the committee supports the assumption that it is more productive than its predecessors. This development suggests that the German health care system is moving towards a more explicitly defined benefit catalogue [17]. Until now the use of cost-effectiveness studies as part of the decision criteria for the inclusion of new benefits is widely lacking. The criteria of cost-effectiveness was only taken into consideration for benefit decisions on medical devices. However, it is likely that it will be considered for other benefits in the future as well. The creation of a supporting institute to the Federal Joint Committee, the Institute for Quality and Efficiency, in 2004, which increasingly commissions effectiveness studies, was one major step in that direction [2]. Although this will increase the information base for decisions [18, 19], the future impact of the cost-effectiveness criteria on the decision-making process and therefore on the structure of the health basket still remains unclear.

Clin_Rev_Allergy_Immunol-3-1-2071970

The Future of Biologic Agents in the Treatment of Sjögren’s Syndrome

The gain in knowledge regarding the cellular mechanisms of T and B lymphocyte activity in the pathogenesis of Sjögren’s syndrome (SS) and the current availability of various biological agents (anti-TNF-α, IFN- α, anti-CD20, and anti-CD22) have resulted in new strategies for therapeutic intervention. In SS, various phase I and II studies have been performed to evaluate these new strategies. Currently, B cell-directed therapies seem to be more promising than T cell-related therapies. However, large, randomized, placebo-controlled clinical trials are needed to confirm the promising results of these early studies. When performing these trials, special attention has to be paid to prevent the occasional occurrence of the severe side effects. Introduction Sjögren’s syndrome (SS) is a chronic lymphoproliferative autoimmune disease with disturbances of T lymphocytes, B lymphocytes, and exocrine glandular cells [1]. SS can be primary (pSS) or secondary SS (sSS), the latter being associated with another autoimmune disease [e.g., rheumatoid arthritis, systemic lupus erythematosus (SLE)]. Lymphocytic infiltrates are a characteristic histopathological finding in SS. These infiltrates consist of T and B cells. The expression of different cytokines, such as tumor necrosis factor-α (TNF-α) and interferon-α (IFN-α), during the formation and proliferation of these infiltrates has been investigated. There is an overexpression of TNF-α, which is secreted by CD4+ T lymphocytes, mononuclear cells, and epithelial cells [2]. The intraglandular synthesis of TNF-α causes destruction of acini by up-regulation of Fas at the surface of the glandular epithelial cells, stimulation of secretion of type 2 and 9 matrix metalloproteases by epithelial cells, and overexpression of different chemokines [3–5]. IFN-α is produced by activated plasmacytoid dendritic cells in primary SS (pSS), and numerous IFN-α-producing cells have been detected in labial salivary glands [6]. IFN-α promotes the autoimmune process by increasing autoantibody production and through the formation of endogenous IFN-á inducers. IFNs have potent immunomodulating properties and are thought to trigger a systemic biological response [7]. Besides the presence of proinflammatory cytokines, described in the previous paragraph, recent studies have shown an important role for B cells in the pathogenesis of SS. Presence of autoantibodies and hypergammaglobulinemia are both considered to reflect B cell hyperactivity. Systemic complications of SS are associated with this B cell hyperactivity [8]. Moreover, about 5% of SS patients develop malignant B cell lymphoma [9]. B cell activating factor (BAFF), also known as B lymphocyte stimulator (BLyS), is an important factor in local and systemic autoimmunity [1]. Dysregulated BAFF expression is implicated in disease progression and perpetuation of humoral autoimmunity. Overproduction of BAFF in transgenic mice has been shown to result in B cell proliferation and antibody production resulting in inflammation and destruction of the salivary glands, as well as kidney failure similar to observations seen in SLE [10]. In humans, circulating BAFF levels are increased in patients with pSS and correlate with disease activity [11]. Recent insights in the cellular mechanisms of T and B lymphocyte activity in the pathogenesis of SS and the current availability of various biological agents have resulted in new strategies for therapeutic intervention. The use of these biological agents in the treatment of SS will be discussed in this review. Biological Agents Currently, biological agents have been introduced in various systemic autoimmune diseases, as rheumatoid arthritis and SLE. Biological agents most frequently applied in autoimmune diseases are monoclonal antibodies, soluble receptors, and molecular imitators [12]. These biological agents enhance or replace conventional immunosuppressive therapy. In contrast to rheumatoid arthritis and SLE, no biological agent has been approved yet for the treatment of SS, but several phase II and III studies have been or are currently conducted. The biological agents used in SS trials are IFN-α and agents targeting TNF-α and B cells (anti-CD20, anti-CD22). Although no trials have been performed yet with BAFF antagonists, this might be a promising therapy [13] and will be discussed in this review, as well. Anti-TNF-α Monoclonal Antibodies There are three main biological agents targeting TNF-α: the chimeric monoclonal IgG1 antibody infliximab, the receptor fusion protein etanercept, and the fully humanized monoclonal antibody adalimumab. In an open-label study, short-term treatment with infliximab was reported to be very effective in active pSS over a 3-month period [14]. Sixteen patients received three infusions (3 mg/kg) at weeks 0, 2, and 6, which led to significant improvement in all clinical and functional parameters, including global assessments, erythrocyte sedimentation rate, whole salivary flow rate, tear secretion (Schirmer test), tender joint count, fatigue score, and sensation of dry eyes and dry mouth. Three patients, all with short disease duration (<3 years), were considered to be in complete remission up till 1 year. In 10 out of the 16 patients, SS symptoms, particularly mouth dryness, relapsed after a median of 9 weeks. In a follow-up study, a maintenance regimen of one infusion every 12 weeks was evaluated in these 10 patients. Retreatment induced an improvement of signs related to SS that was comparable with the effects from the three loading infusions [15]. To confirm these promising results from an uncontrolled study, the Trial of Remicade In Primary Sjögren’s Syndrome study was designed. In this multicenter, double-blinded, placebo-controlled, randomized clinical trial, 103 patients with active pSS were included and treated with infliximab infusions (5 mg/kg) or placebo at weeks 0, 2, and 6. Follow-up was 22 weeks. Primary endpoint was an improvement of >30% of two of three VAS scores measuring joint pain, fatigue, and dry eyes. There were several secondary endpoints of which one was the basal salivary flow rate. In contrast to the previously mentioned uncontrolled studies, no evidence of efficacy of infliximab treatment on all clinical and functional parameters could be demonstrated in this randomized controlled clinical trial [2]. A trial on 15 pSS patients (mean disease duration 3.6 years) with 25-mg etanercept, subcutaneously twice a week for 12 weeks, did not reveal a reduction in sicca symptoms and signs, neither did the repeated treatment for up to 26 weeks. Only in the subset of four patients with severe fatigue a decrease in fatigue was observed [16]. Another trial evaluating subcutaneous administration of etanercept vs placebo for 12 weeks (28 patients) also showed no clinical efficacy [17]. No trials of adalimumab treatment in pSS have been reported in the literature yet. In conclusion, TNF-targeting treatment could not be proven to be of benefit in reducing the complaints of pSS patients. IFN-α IFNs are proteins with antiviral activity and potent immunomodulating properties. SS patients have an activated type I IFN system [6]. Such a role for IFN-α appears to contradict the reports described below, that low doses of IFN-α administered via the oromucosal route increase the unstimulated salivary output. However, it is hypothesized that oral IFN-α treatment may act by increasing saliva secretion by up-regulation of aquaporin 5 transcription without significantly influencing the underlying autoimmune process [6, 7]. In a phase II study, treatment of pSS patients with IFN-α administered via the oromucosal route (by dissolving lozenges) was demonstrated to be effective (improvement of salivary output, decreased complaints of xerostomia) and safe [18]. Based on these promising results, a randomized, parallel group, double-blinded, placebo-controlled clinical trial (497 pSS patients) was designed. Patients were randomized into two groups and received a 24-week daily treatment with either 450 IU IFN-α (150 IU three times per day) or placebo in a ratio 3:2, administered by the oromucosal route. This randomized, controlled clinical trial failed to demonstrate a significant effect on the primary endpoints (VAS score for oral dryness and stimulated whole salivary flow) in the IFN-α group relative to the placebo group. There was a significant increase in unstimulated whole saliva in the patients treated with IFN-α, which correlated positively and significantly with improvement in seven of eight symptoms associated with oral and ocular dryness. No adverse events were observed [7]. In conclusion, no clinical evidence for the efficacy of IFN-á treatment in pSS patients has been shown yet; however, an improvement of unstimulated whole saliva was observed. Further research is needed to objectify the effect of IFN-á on salivary gland tissue. Anti-CD20 Monoclonal Antibodies Anti-CD20 (rituximab) is a chimeric humanized monoclonal antibody specific for the B cell surface molecule CD20, which is expressed on the surface of normal and malignant pre-B and mature B lymphocytes. CD20 mediates B cell proliferation and differentiation. This antibody has been demonstrated to prevent B cells from proliferating and to induce lysis of B cells by complement-dependent and antibody-dependent cytotoxicity mechanisms as well as by direct induction of apoptosis [19]. Rituximab is currently used for the treatment of low-grade B cell lymphomas [20]. In controlled studies, it was shown to be safe and effective in the treatment of rheumatoid arthritis [21–23]. Moreover, open-label studies in SLE patients are promising [24]. In an open-label phase II study, 15 patients with pSS were treated with 4 infusions of rituximab (375 mg/m2 once weekly) and followed up for a 3-month period. Eight of the 15 patients were early pSS patients (mean disease duration 28 months, all had residual salivary gland function at baseline), and 7 patients had a concomitant mucosa-associated lymphoid tissue (MALT) lymphoma (mean disease duration 79 months). In the early pSS patients, rituximab treatment resulted in significant improvement of subjective symptoms and an increase in salivary gland function. All patients showed a rapid depletion of peripheral B cells within a few weeks, accompanied by a decrease in IgM-RF levels [8]. Repeated parotid gland biopsies in five of the early patients after treatment showed redifferentation of the lymphoepithelial duct lesions into normal striated ducts, possibly indicating regeneration of salivary gland tissue (unpublished data). Five of the eight pSS patients without a MALT lymphoma received a second course of rituximab (after 9–11 months) due to recurrence of symptoms. Retreatment resulted in the same significant improvement of the salivary flow rate and subjective symptoms compared to the results of the first treatment, together with a decrease in B cells and IgM-RF levels. Six of the seven MALT/pSS patients were initially effectively treated with rituximab. The remaining MALT/pSS patient had progressive MALT disease and severe extraglandular SS disease within 3 months after the start of rituximab treatment. Cyclophosphamide was added, which led to stable disease of both MALT and SS. One of the six patients initially responding had a recurrence of MALT lymphoma after 9 months and was successfully retreated with rituximab. The other patients are still in remission (unpublished data). In another open-label study, 16 pSS patients received only two weekly rituximab infusions (375 mg/m2), with a follow-up of 36 weeks. Again, treatment resulted in rapid complete depletion of peripheral B cells. At week 12, a significant improvement of VAS scores for fatigue and dryness was recorded, and at week 36, a significant improvement for VAS scores for global disease, fatigue, dry mouth, dry eyes, and dry vagina, but also in the number of tender joint and tender joint counts was seen [25]. Both in the study of Pijpe et al. [8] and the study of Devauchelle-Pensec et al. [25], patients with a short disease duration showed more improvements than patients with longer disease duration. Two trials retrospectively evaluated the effect of rituximab (four infusions of 375 mg/m2) in 18 pSS patients (mean disease duration 10 years) with systemic features. Self-reported dryness improved in six patients (VAS scores not known for three patients, no improvement in the other nine patients). Both studies reported good efficacy of the treatment on systemic features [26, 27]. In conclusion, in phase II trials, it has been shown that rituximab seems to be effective for at least 6–9 months in pSS patients with active disease, improving both subjective and objective complaints. Retreatment with rituximab resulted in a similar good clinical response. In pSS patients with longer disease duration, without residual salivary gland function, rituximab treatment seems to be effective for systemic features. To confirm these promising results, randomized placebo-controlled clinical trials are needed. Anti-CD22 Monoclonal Antibodies Epratuzumab is a fully humanized monoclonal antibody specific for the B cell surface molecule CD22. CD22 is expressed on the surface of normal mature and malignant B lymphocytes. CD22 appears to be involved in the regulation of B cell activation through B cell receptor signaling and cell adhesion [28]. In an open-label phase I/II study, safety and efficacy of epratuzumab were investigated in 16 pSS patients. Follow-up was 6 months. These pSS patients received four doses of 360 mg/m2 epratuzumab intravenously. Mean disease duration before therapy was 2.9 years, and none of the patients had received prior B cell-targeted therapy. Most improvements occurred in the Schirmer test, unstimulated whole salivary flow and the VAS score for fatigue. The new developed disease activity score consisted of the four domains: dryness of the eyes, dryness of the mouth, fatigue, and laboratory parameters. Based on this score, 53% achieved at least 20% improvement in at least two domains at 6 weeks. Corresponding rates for 10, 18, and 32 weeks are 53, 47, and 67%. Remarkably, the number of responders was higher 6 months after the treatment administration than earlier. Peripheral B cells decreased with a median decrease of 54 and 39% at 6 and 18 weeks, respectively. In conclusion, epratuzumab seems to be an effective treatment. Randomized, placebo-controlled clinical trials are needed before epratuzumab can be advised for general treatment in pSS patients [29]. Anti-BAFF BAFF is a B cell-activating factor that acts as a positive regulator of B cell function and expansion. BAFF levels were found elevated in serum and saliva in SS patients, but no correlation could be shown between serum and saliva levels [30]. However, circulating levels of BAFF in pSS patients were shown to be a marker for disease activity [11]. To the best of our knowledge, no trials have been performed with anti-BAFF treatment in SS yet, but such an approach might be considered for future trials. Currently, two human BAFF antagonists have been developed, a human antibody (anti-BLyS) that binds to soluble BAFF and a fusion protein of one of the BAFF receptors [31, 32]. Especially, SS patients with elevated BAFF levels, hypergammaglobulinemia, elevated levels of autoantibodies, and associated B cell lymphoma might be candidates for anti-BAFF treatment [33]. Safety and Tolerability of Biological Agents The most important side effects of treatment with biological agents are direct mild infusion reactions. Several patients developed a serum sickness-like disease a few days after the second infusion that might be related to the formation of antibodies against the biological agent [human anti-chimeric antibodies (HACAs) or human anti-human antibodies]. A few patients developed infections during treatment with a biological agent; however, some patients concomitantly used other immunosuppressive therapies. Therefore, the direct relation between the biological agent and the infection is unsure. All adverse events reported in the trials described in this review are reported in Table 1. According to this table, the most frequent side effects of treatment with biological agents are mild infusion reactions. The most severe side effect of the various treatments used in SS patients was the development of a serum sickness-like disease. This adverse effect of treatment occurred in 16% (8 of 49) of the patients treated with rituximab. HACA formation was observed in patients developing a serum sickness-like disease and occurred only in patients receiving low-dose corticosteroids and no other immunosuppressive drugs. It is assumed that higher doses of corticosteroids during treatment might prevent the occurrence of serum sickness. Table 1Adverse events after treatment with biological agents in SS Agent/doseNumber of patients in trial (number treated with the agent)Premedication/concomitant immunosuppressive therapyInfusion reactionInfectionsSerum sicknessHACA/HAHA formationOtherAnti-TNF-α monoclonal antibodiesSteinfeld [14]Infliximab intravenous, 3 mg/kg16 (16)n.r./no1 (6%)2 (13%) (respiratory tract)–n.r.–Steinfeld [15]Infliximab intravenous, 3 mg/kg10 (10)n.r./no4 (40%)2 (20%) (enteritis, tonsillitis)–n.r.–Marriette [2]Infliximab intravenous, 5 mg/kg103 (54)n.r./continuation of hydroxychloroquine and corticosteroids (≤15 mg/day)2 (4%)2 (4%) (1 cutaneous, 1 respiratory tract)–n.r.2 (breast cancer, auto-immune hepatitis)aZandbelt [16]Etanercept subcutaneously, 25 mg15 (15)n.r./pilocarpine at a constant dose–1 (7%) (parotitis)–n.r.–Sankar [17]Etanercept subcutaneously, 25 mg28 (14)n.r./allowed to use long-term medication1 (7%)1 (7%) (skin lesion)b–n.r.–IFN-αShip [18]IFN-α oromucosal, 150 IU, 450 IU109 (87)n.r./non.a.––n.r.–cCummins [7]IFN-α oromucosal, 450 IU497 (300)n.r./non.a.–––23 (7.7%)d (34% gastrointestinal, 25% musculoskeletal)Anti-CD20Pijpe [8]Rituximab intravenous, 375 mg/m215 (15)25 mg prednisolon intravenously/patients with severe extraglandular manifestations (n = 3) received immunosuppressive therapy2 (13%)1 (7%) (zoster)4 (27%)e4 (27%)–Devauchelle-Pensec [25]Rituximab intravenous, 375 mg/m216 (16)n.r./no––1 (6%)n.r.–Gottenberg [26]Rituximab intravenous, 375 mg/m26 (6)n.r./hydroxychloroquine (n = 1), methylprednisolone (n = 3)1 (17%)–1 (17%)n.r.–Seror [27]Rituximab intravenous, 375 mg/m212 (12)n.r./cyclophosphamide (n = 1), hydroxychloroquine (n = 1), leflunomide (n = 1)1 (8%)–2 (17%)n.r.–Anti-CD22Steinfeld [29]Epratuzumab intravenous, 360 mg/m216 (16)0.5–1 g acetominophen, 25–50 mg antihistamine./no2 (13%)2 (13%) (sinusitis, dental abscess)–3 (19%)6 (38%) (TIA, osteoporotic fracture, diarrhea, dyspepsia, palpitations, paresthesia)n.a. Not applicable, n.r. not reported, HACA human anti-chimeric antibodies, HAHA human anti-human antibodiesaOne patient in the placebo group developed benign lymph node enlargementbOne patient in the placebo group developed a prolonged upper respiratory tract infectioncIn this study, there were mild adverse events; however, there were no significant differences between the groups. Adverse events were not specified.dEight patients (4.1%) in the placebo group developed adverse eventseOne of these 4 patients developed serum sickness after retreatment [8] Future Perspectives Biological agents are promising therapies for SS. Randomized studies failed to show a clinical effect of anti-TNF-α and IFN-α in the treatment of SS. Notwithstanding the unfortunate results of anti-TNF-α and IFN-α, B cell depletion (both anti-CD20 and anti-CD22) seems very promising. Again, this promising effect, as was previously also assumed for anti-TNF-α and IFN-α, must be confirmed in larger randomized controlled clinical trials. HACAs have been reported to occur at a higher rate in patients with an autoimmune disease. It seems that monoclonal antibodies are more immunogenic in active autoimmune disease, independent of the type of disease. Additional use of immunosuppressive therapy in these patients might be mandatory to prevent serious side effects. These unwanted side effects might also be prevented by the use of fully humanized antibodies. The currently available humanized antibodies are promising, but need further study. Moreover, there is still a need for improved assessment parameters to monitor treatment effects, both subjectively and objectively. For studies on intervention of SS, evaluation of the parotid gland might be of use because function, composition of saliva, and histology can be evaluated on the same gland at different time points. Activity scores are currently under development by Bowman and Vitali [34, 35]. Finally, as soon as effective intervention treatments have been established, the cost-effectiveness of these currently very expensive antibodies needs to be analyzed to select those patients that might benefit the most from this kind of treatment.

Oecologia-4-1-2311384

The effect of temperature on growth and competition between Sphagnum species

Peat bogs play a large role in the global sequestration of C, and are often dominated by different Sphagnum species. Therefore, it is crucial to understand how Sphagnum vegetation in peat bogs will respond to global warming. We performed a greenhouse experiment to study the effect of four temperature treatments (11.2, 14.7, 18.0 and 21.4°C) on the growth of four Sphagnum species: S. fuscum and S. balticum from a site in northern Sweden and S. magellanicum and S. cuspidatum from a site in southern Sweden. In addition, three combinations of these species were made to study the effect of temperature on competition. We found that all species increased their height increment and biomass production with an increase in temperature, while bulk densities were lower at higher temperatures. The hollow species S. cuspidatum was the least responsive species, whereas the hummock species S. fuscum increased biomass production 13-fold from the lowest to the highest temperature treatment in monocultures. Nutrient concentrations were higher at higher temperatures, especially N concentrations of S. fuscum and S. balticum increased compared to field values. Competition between S. cuspidatum and S. magellanicum was not influenced by temperature. The mixtures of S. balticum with S. fuscum and S. balticum with S. magellanicum showed that S. balticum was the stronger competitor, but it lost competitive advantage in the highest temperature treatment. These findings suggest that species abundances will shift in response to global warming, particularly at northern sites where hollow species will lose competitive strength relative to hummock species and southern species. Introduction Peat bogs play a large role in the global sequestration of C. Although northern peatlands cover only 2% of the total land surface, they store about one-third of the world soil C in the form of peat (Gorham 1991). It is therefore important to know how ombrotrophic bog ecosystems, which form a large part of northern peatlands, will respond to predicted climate changes, especially since the rise in temperature is expected to be above the global average at high latitudes (Christensen et al. 2007) where the majority of peat bogs occur (Kivinen and Pakarinen 1980; Gunnarsson 2005). Vegetation in ombrotrophic bogs is often dominated by different Sphagnum species. The Sphagnum species account for the bulk of the C sequestration in peat because of their recalcitrant litter (Coulson and Butterfield 1978; Clymo and Hayward 1982; Limpens et al. 2003). Therefore, it is crucial to understand how Sphagnum vegetation in bogs will respond to global warming in order to predict the role of bogs as C sinks in the future. Several studies have revealed differences in production (Lindholm and Vasander 1990; Gerdol 1995; Asada et al. 2003; Gunnarsson 2005) and in decomposition rate (Rochefort et al. 1990; Johnson and Damman 1993; Belyea 1996; Limpens and Berendse 2003) between different Sphagnum species. These differences between species are often related to differences in microhabitat preference (Gunnarsson 2005). Within a bog, different Sphagnum species occur at different heights above the water table and at different positions along pH and nutrient gradients (Andrus 1986; Sjörs and Gunnarsson 2002; Limpens et al. 2003). The most obvious division in microhabitat preference is between hollow species, which grow in pools and at shallow water levels, and hummock species, which grow at deeper water levels. Hummock species can also grow at higher water levels, but they are then usually outgrown by hollow species (Rydin 1986, 1993, 1997). However, the competitive ability of species may differ between years, seasons and locations, and as such, competitive replacement occurs very slowly, if at all. This results in a relatively stable competitive balance between species (Rydin 1997). Several studies found a positive relation between Sphagnum productivity and temperature (Moore 1989; Sonesson et al. 2002; Gunnarsson 2005). The positive effect that increased temperature might have on C sequestration in bogs is, however, often diminished by the positive effect of temperature on decomposition rates (Hobbie 1996). Increased decomposition rates also lead to increased rates of nutrient release from the peat layer, enhancing production rates even further. A change in temperature influences not only the production and decomposition of individual Sphagnum species but also the competitive balance that exists between species. An important challenge facing ecologists is to predict how climate change will alter species distributions in ecosystems (Mooney 1991). Robroek et al. (2007b) already found different responses in biomass production among species when temperature was increased. It can be imagined that when a species with high production and/or a low decomposition rate increases its relative abundance in a bog, this will increase the C storage capacity of the system. Not only are there different competing species within a bog, but there are also differences in dominant species between bogs when different climatic regions are compared. In this study, we used species from two different sites. At the site in northern Sweden, Sphagnum balticum and Sphagnum fuscum are the dominant species while at the site in southern Sweden these species also occur, but Sphagnum magellanicum and Sphagnum cuspidatum are the most abundant species. This corresponds with the general distribution of these species in Europe since both S. magellanicum and S. cuspidatum occur further south than S. fuscum and S. balticum (Daniels and Eddy 1985). To examine the effect of temperature on the competition between species, we performed a greenhouse experiment in which we studied the effect of four temperature treatments on the growth of the four species: S. fuscum and S. balticum from a northern Swedish site and S. magellanicum and S. cuspidatum from a southern Swedish site. Three combinations of species were made to study the effects of temperature on interspecific competition. In our experiment we tried to answer the following questions:What is the effect of increased temperature on the growth of different Sphagnum species? We expect all species to show an increase in both height increment and biomass production as a direct result of increased temperature and indirectly through increased nutrient availability.What is the effect of temperature on competition between species? Since S. fuscum and S. magellanicum grow in drier and therefore also warmer microhabitats than S. balticum and S. cuspidatum respectively, we expect these species to be better adapted to higher temperatures. S. magellanicum grows at more southern sites than S. balticum, so we expect this species to be better adapted to higher temperatures. Consequently, we hypothesize that increased temperature will have a positive effect on competitive abilities of the hummock and southern species, leading to a relatively larger increase in height increment and biomass production with temperature than for hollow and northern species. Materials and methods Plant material In August 2004, Sphagnum cores (diameter 16 cm, height 18–22 cm) were collected at two different sites in Sweden. From the northern site Lappmyran (64°09′N, 19°35′E), 30 Sphagnum fuscum (Schimp) H. Klinggr cores and 40 Sphagnum balticum (Russ.) C. Jens. cores were collected. This site is a string flark mire or mixed mire with ridges of hummocks and hollows where S. fuscum is dominant on the hummocks and S. balticum in the dryer parts of the hollows. When identifying the species from this site in the lab, we found specimens of both S. balticum and Sphagnum angustifolium (Russ.) C. Jens. These species are difficult to distinguish, which both Russow (Smith 1978) and Klinggraff (Daniels and Eddy 1985) recognized when they identified both species as varietas of Sphagnum recurvum. In our samples, we were unable to quantify the exact percentages of S. balticum and S. angustifolium. As result, whenever we mention S. balticum, we are referring to a mixture of S balticum and S. angustifolium. From the southern site Saxnäs Mosse (56°51′N, 13°27′ E), 40 Sphagnum magellanicum (Brid.) cores and 30 Sphagnum cuspidatum (Hoffm.) cores were collected. At this site S. magellanicum occurs on the lawns and low hummocks and S. cuspidatum in the hollows and pools. Cores were taken from monospecific stands of each Sphagnum species (>95%) with sparse vascular plant cover (<5%). The cores were placed in plastic containers (diameter 16 cm, height 22 cm). Vascular plants were clipped flush with the Sphagnum, and other Sphagnum species were removed with tweezers. The containers were brought to Wageningen and stored outside for 8 weeks before the greenhouse was available. As a result, containers from both sites could acclimate to the same climate to some extent. Experimental design At the start of the experiment, the containers were brought into the greenhouse and randomly divided over the treatments and five replicate blocks, with seven species combinations (four monocultures and three mixtures) and four temperature treatments per block. All four species were kept in monoculture and the following three species combinations were made to study interspecific competition: two northern species S. fuscum with S. balticum, two southern species S. magellanicum with S. cuspidatum and a northern with a southern species S. balticum with S. magellanicum. To study competition between a northern and a southern species we chose the combination of S. balticum and S. magellanicum because they occur at similar water levels. To make the combinations, the cores were cut into four equal quarters and two quarters of both species were placed alternately in an empty container. In potting the species combinations, we made certain that the surface of the mixture was uniform. We did not cut monocultures in four quarters, but another experiment showed no difference in water content between cut and uncut monocultures (Robroek et al. 2007a). The experiment was conducted in four adjacent climate controlled greenhouse compartments from November 2004 till April 2005 for a total of 154 days. Each compartment was assigned one of four temperature treatments. Treatments and containers were switched between compartments every 2 weeks to minimize any effect of the different compartments. The position of the blocks and the position of containers within the blocks were also switched every 2 weeks. The average day temperatures in the four temperature treatments were 11.2, 14.7, 18.0 and 21.4°C, respectively (Table 1). During the dark period of 8 h, the day temperature was lowered by approximately 3–9.3, 11.6, 15.5 and 18.9°C, resulting in mean temperatures of 10.6, 13.7, 17.2 and 20.6°C in temperature treatments 1, 2, 3 and 4, respectively. In the region of the northern site, the mean temperature in July is 14.7°C (Alexandersson et al. 1991) and in the southern site the mean temperature in July is 17.0°C (Malmer et al. 2003). A light period of 16 h was applied. If light intensity was low during this period, SON-T AGRO 400 (Philips Powertone 400) lamps were used. Table 1Day and night mean values of temperature (°C), relative humidity (%) and vapour pressure deficit (VPD) (kPa) of the four treatments, ±SE, n = 154TreatmentTemperatureRelative humidityVPDDayNightDayNightDayNightT111.2 ± 0.0493 ± 0.0581.9 ± 0.486.8 ± 0.40.24 ± 0.010.16 ± 0.01T214.7 ± 0.03116 ± 0.0576.9 ± 0.381.8 ± 0.50.39 ± 0.010.25 ± 0.01T318.0 ± 0.02155 ± 0.0375.2 ± 0.476.1 ± 0.40.51 ± 0.010.42 ± 0.01T421.4 ± 0.03189 ± 0.0472.7 ± 0.473.0 ± 0.50.70 ± 0.010.59 ± 0.01 Relative humidity during the day was set at 75%. In the greenhouse it was not possible to keep the relative humidity exactly the same with all temperatures. The rise in temperature between treatments corresponded with a decrease in relative humidity, which caused an additional increase in vapour pressure deficit (VPD) with temperature (Table 1). The difference in relative humidity only explained 33% of the increase in VPD with the highest temperature. If relative humidity would have been equal in all compartments, VPD would still have been twice as high at temperature 4 as at temperature 1. Our highest VPD of 0.7 kPa with a temperature of 21.4°C is actually still quite low compared to field conditions. Although not many data on VPD are published for similar ecosystems, Hobbie and Chapin (1998) did find the following values for VPD in Toolik Lake, Alaska, in June and July: in open field sites, 0.02–1.71 kPa with temperatures ranging from 5.9 to 22.5°C; and under small plastic greenhouses, 0.08–3.83 kPa with temperatures ranging from 6.4 to 31.1°C. Dorrepaal et al. (2003) measured a VPD of 1.54 under normal conditions and 1.41 in open-top chambers with an average temperature of 15°C in June and July in Abisko, Sweden. The water level was set to 1 cm below capitula at the start of the experiment. All species were subjected to the same water level, so that temperature was the only changing environmental variable. This relatively high water level was used for all species because hollow species cannot grow at low water levels; while hummock species can survive the environmental conditions of hollows quite well; nonetheless, they are absent from these areas because of biotic factors (Rydin and McDonald 1985). Grosvernier et al. (1997) found that growth in height and dry weight is equal for S. fuscum grown at water levels of 1 cm and 40 cm below moss surface while, for S. magellanicum and especially for Sphagnum fallax, growth in height and dry weight is much greater with the high water level. During the experiment an artificial rainwater solution, an 8,000-fold dilution of a sea water solution (Garrels and Christ 1965), was added twice a week to bring the water level back to 1 cm below capitula. The amount of water added was used as a measure of evaporation. The drop in the water table was highest in the highest temperature treatment, but water level never dropped more than 6 cm below moss surface between two water additions. In a number of containers, Sphagnum grew higher than 1 cm above the container. To keep the water level at 1 cm below the moss surface, a plastic ridge was glued onto the containers and the crack was filled with silicone kit. Water content was measured using a theta probe (Delta-T Devices, Cambridge, UK) before the final harvest of the experiment. This was done 3–4 days after watering the containers for the last time, so the water content would reflect possible differences between treatments. Measurements Height increment of the Sphagnum carpet was measured non-destructively every month using a variation of the cranked wire method (Clymo 1970). We used plastic rods that were inserted to a depth of approximately 8 cm and anchored by plastic broom bristles, this method kept the cranked wires firm at the same place so they did not move with Sphagnum growth. Two plastic rods were inserted in the monocultures and one plastic rod was inserted in each quarter of the mixtures. The rods had a diameter of 1.5 mm and did not seem to interfere with the growth of the surrounding Sphagna. At the end of the experiment, columns with a diameter of 5 cm were cut around each cranked wire and cut off at 5 cm length. Each column was put in a plastic Ziplock bag and fresh weight was determined. All bags were stored at 1°C till further measurements could be taken. Capitula were defined as the top 1 cm of each individual plant and stem as the 1–4 cm part. Capitula and stems were separated per column and oven dried at 70°C for at least 48 h and then weighed. The weight of the total sample was used to calculate bulk density and the biomass production per squared centimetre to account for changes in capitulum density. For nutrient analyses, samples of capitula were pooled per species for each container. Total N, P and K concentrations were determined by digesting 300 mg of homogeneous, milled material with H2SO4, salicylic acid, Se and H2O2. All samples were analysed for total N and P spectrophotometrically using an auto-analyzer (Skalar). K concentrations were measured with an atomic absorption spectrophotometer (Varian AAS). To compare capitulum bulk density and nutrient concentrations with field values, we collected five samples (d = 5 cm) from monocultures of the four Sphagnum species in the two Swedish sites in August 2006. Measurements on capitulum bulk density and nutrient concentrations were executed as described above. To measure the change in cover of the species in the mixtures, digital pictures were made at the start and at the end of the experiment. In these pictures we measured the total surface cover per species in each pot with Image J (Abramoff et al. 2004). The biomass production per unit area (g m−2) was calculated as follows: Data analysis Data were tested for normality and equality of variance. When necessary, data were square-root transformed to achieve homogeneous variances. Block effect was tested as random factor. When no block effect was detected, which was usually the case, block was omitted from the analysis to gain extra df. All analyses were conducted using the SPSS statistical package for Windows (12.0). One container with a monoculture of S. balticum under temperature 4 was heavily affected by a fungal infection, probably Lyophyllum palustre. After 4 months, 90% of the plant material had died. This container was further omitted from the analyses. Height increment in monocultures was tested per species with one-way ANOVA with temperature as independent factor. Height increment, biomass production, cover change, bulk density, water content and nutrient concentrations were tested per species. Two-way ANOVAs were performed with temperature and species combination as independent factors. The effect of temperature, species and competition on the concentrations of N, P and K was tested using a three-way ANOVA. Differences between the treatments were analysed using a Tukey post hoc test. To test how the environmental variables temperature, N concentration and water content influenced biomass production we performed a stepwise regression analysis per species. Results Height increment In the monocultures, the height increment increased with temperature in each species (Fig. 1, Table 2). For S. cuspidatum this effect was least distinct with the lowest height growth at temperature 2 and no difference in growth between temperatures 1, 3 and 4. Fig. 1Height increment of species in monocultures (cm). Data are mean values ± SE, n = 5 except for Sphagnum balticum at T4, n = 4. Different letters indicate significant differences (P < 0.05) between temperature treatments within each species. For temperature treatments, see Table 1Table 2F-values and P-values and direction of main effects influencing height increment and biomass production per speciesa, corrected for cover, using a two-way ANOVATemperatureCombinationT × combinationFPEffectFPEffectFPHeight incrementSphagnum fuscum (+Sphagnum balticum)49.31<0.001+1.170.288+0.920.445S. balticum (+S. fuscum)51.19<0.001+9.130.005−0.610.616S. balticum (+Sphagnum magellanicum)19.58<0.001+10.320.003−3.240.035S. magellanicum (+S. balticum)28.71<0.001+8.870.006−1.880.154S. magellanicum (+Sphagnum cuspidatum)18.53<0.001+4.860.035−1.140.350S. cuspidatum (+ S. magellanicum)10.80<0.001+18.83<0.001−0.860.470Biomass productionS. fuscum (+S. balticum)20.99<0.001+18.57<0.001−2.310.095S. balticum (+S. fuscum)19.41<0.001+5.480.026−2.040.128S. balticum (+S. magellanicum)9.68<0.001+12.380.001−3.470.028S. magellanicum (+S. balticum)17.50<0.001+44.99<0.001−1.670.193S. magellanicum (+S. cuspidatum)11.75<0.001+23.36<0.001−2.000.133S. cuspidatum (+ S. magellanicum)5.150.005+2.200.148−0.180.913aNumber of observations n = 40, except for S. balticumn = 39 Only S. fuscum showed no difference in height increment between monoculture and mixture. All other species showed a growth reduction in mixtures (Table 2). There were no significant interactions between the effect of temperature and competition on height increment, except for S. balticum in combination with S. magellanicum (Table 2). Temperature increased height increment of S. balticum in monoculture, while there was no effect of temperature in mixture with S. magellanicum (Figs. 1b, 2b). When we compared height increment between the two species in each mixture, the “wet” species S. balticum and S. cuspidatum had higher values than the “dry” species S. fuscum and S. magellanicum at temperature treatments 1 and 2 (Fig. 2). When temperature increased, the differences in height increment between the two species disappeared. Fig. 2Height increment of the individual species at each temperature treatment (see Table 1) within the three mixtures of species: aSphagnum fuscum + S. balticum, bSphagnum magellanicum + S. balticum and cS. magellanicum + Sphagnum cuspidatum. Data are mean values ± SE, n = 5. Different letters indicate significant differences between treatments (P < 0.05) Cover The hollow species S. balticum and S. cuspidatum increased in area in 53 out of 60 containers when growing in the mixtures. In the other seven containers (three with S. magellanicum + S. balticum, two with S. fuscum + S. balticum, two with S. magellanicum + S. cuspidatum), cover changed less than 5%. For S. balticum there were significant effects of temperature (F = 3.67, P = 0.023) and neighbouring species (F = 4.95, P = 0.033) on expansion (Fig. 3a, b). Maximum change in cover was for S. balticum in combination with S. fuscum with temperature 2 and 3. There was no effect of temperature on expansion of S. cuspidatum (Fig. 3c, F = 0.88, P = 0.470). Fig. 3Cover change for hollow species: aS. balticum in combination with S. fuscum, bS. balticum in combination with S. magellanicum and cS. cuspidatum in combination with S. magellanicum. Data are mean values ± SE, n = 5. There were no significant differences between treatments (P < 0.05). For temperature treatments, see Table 1 Biomass production The response of biomass production to the temperature treatments was similar to the response of height increment (Table 2). In all species biomass production increased with increased temperature. S. fuscum showed the strongest response; this species increased its biomass production 13-fold from the lowest to the highest temperature treatment in monocultures (Fig. 4a). Only for S. balticum in combination with S. magellanicum was there a significant interaction between temperature and species combination in biomass production (Table 2). Temperature increased biomass production of S. balticum in monoculture and in mixture with S. fuscum, while there was no effect of temperature in mixture with S. magellanicum. Fig. 4Biomass production corrected for change in cover for species in monocultures (mono) and in mixtures (mix) for each temperature treatment (see Table 1) and species combination: aS. fuscum (fus) + S. balticum (bal), bS. magellanicum (mag) + S. balticum and cS. magellanicum + S. cuspidatum (cus). Data are mean values ± SE, n = 4–5. Different letters indicate significant differences between treatments (P < 0.05) When differences in biomass production between monocultures and mixtures are compared per temperature treatment, it shows that S. fuscum suffered from competition with S. balticum at temperature 3 because production was lower in mixture than in monoculture, but it no longer suffered at temperature 4 (Fig. 4a). At temperature 4 S. balticum does suffer from competition with S. fuscum, but not at lower temperatures. In the mixtures of S. magellanicum and S. balticum, S. magellanicum suffers from competition at temperature 1 and 2, whereas S. balticum suffers from competition at temperature 3 and 4 (Fig. 4b). Biomass production of S. cuspidatum did not show any effect of competition with S. magellanicum, while biomass production of S. magellanicum did suffer from competition at temperatures 1, 2 and 3 (Fig. 4c). In the stepwise regression for biomass production of S. balticum, water content was selected as the most explanatory variable (R2 = 0.36, P < 0.001). For biomass production of S. fuscum (R2 = 0.52, P < 0.001), S. magellanicum (R2 = 0.36, P < 0.001) and S. cuspidatum (R2 = 0.12, P = 0.031), temperature was selected as the most explanatory variable. Bulk density Bulk densities were lower at higher temperatures for all species, except for bulk density of the capitula and stems of S. cuspidatum (Table 3). Compared to field values, capitulum bulk density of S. fuscum and S. balticum decreased with temperature, while capitulum bulk density of S. magellanicum and S. cuspidatum seemed to increase at low temperatures. Table 3Bulk density of capitula section, stem section and total sample. Data are mean values ± SE, pooled within temperature treatment (see Table 1) and within competition treatment. Different letters indicate significant differences between treatments within a species (P < 0.05). There were no significant interactions between temperature and competition treatments. fus S. fuscum, bal S. balticum, mag S. magellanicum, cus S. cuspidatumField valueTemperatureCombinationT1T2T3T4Monofus–balmag–balmag–cusBulk density capitula (mg cm−3)S. fuscum38 ± 235 ± 2b37 ± 2b21 ± 1a17 ± 1a30 ± 2b25 ± 2a––S. balticum32 ± 126 ± 1b16 ± 1a14 ± 1a14 ± 1a20 ± 1b18 ± 2b15 ± 1a–S. magellanicum19 ± 121 ± 1b18 ± 1b13 ± 1a13 ± 1a17 ± 1–15 ± 117 ± 1S. cuspidatum19 ± 123 ± 122 ± 219 ± 119 ± 123 ± 1b––19 ± 1aBulk density stems (mg cm−3) S. fuscum18 ± 1b16 ± 1b15 ± 2ab11 ± 2a17 ± 1b14 ± 1a––S. balticum15 ± 1b11 ± 1a8 ± 1a9 ± 1a13 ± 1b9 ± 1a11 ± 1ab–S. magellanicum11 ± 1c9 ± 0bc8 ± 1ab7 ± 0a9 ± 1–9 ± 18 ± 0S. cuspidatum10 ± 110 ± 17 ± 17 ± 18 ± 1––9 ± 1Bulk density total sample (mg cm−3) S. fuscum22 ± 1c21 ± 1bc17 ± 1ab13 ± 1a20 ± 1b16 ± 1a––S. balticum17 ± 1b12 ± 1a10 ± 1a10 ± 1a14 ± 1b11 ± 1a12 ± 1ab–S. magellanicum13 ± 1c11 ± 0b9 ± 1a8 ± 0a11 ± 1–10 ± 110 ± 0S. cuspidatum12 ± 1c12 ± 0bc10 ± 1ab9 ± 0a11 ± 1––11 ± 0Values for material from the sites of origin (Field values, n = 5) are given for comparison. Number of observations for each species for each temperature treatment: S. fuscum and S. cuspidatum, n = 10; S. balticum, n = 14–15; S. magellanicum, n = 15. Number of observations for each species for each combination, n = 19–20 Bulk densities of both capitula and stems were lower in mixtures than in monocultures of S. fuscum and S. balticum. Similarly, in S. cuspidatum the bulk density of the capitula was lower in mixtures than in monocultures. There were no significant interactions between temperature and species combination. Nutrient concentrations The concentrations of all nutrients (N, P and K) differed between species and rose with increasing temperature. N concentration was also affected by the interactions of temperature by species (F = 2.66, P = 0.007) and species by combination (F = 3.73, P = 0.006). At temperatures 1 and 2, N concentrations were lower for S. fuscum and S. balticum than for S. magellanicum and S. cuspidatum, as was also the case in field values. At temperature 3, only the N concentration of S. balticum was lower than that of S. magellanicum. At temperature 4, N concentrations of S. balticum were lower than those of S. magellanicum and S. fuscum (Table 4). Compared to field values, all species showed an increased N concentration with temperature. Table 4N, P and K concentration (mg g−1) of the capitula. Data are mean values ± SE, pooled within temperature treatment (see Table 1) and within competition treatment. Different letters indicate significant differences between treatments within a species (P < 0.05). There were no significant interactions between temperature and competition treatments. For abbreviations, see Table 3Field valueTemperatureCombinationT1T2T3T4Monofus–balmag–balmag–cusN (mg g−1)S. fuscum6.5 ± 0.28.4 ± 0.2a8.6 ± 0.4a11.3 ± 0.6b12.0 ± 0.7b9.8 ± 0.510.4 ± 0.5S. balticum5.3 ± 0.18.4 ± 0.38.7 ± 0.49.7 ± 0.59.6 ± 0.39.5 ± 0.48.7 ± 0.49.0 ± 0.3S. magellanicum11.1 ± 0.511.7 ± 0.512.6 ± 0.312.8 ± 0.312.9 ± 0.312.4 ± 0.312.3 ± 0.312.7 ± 0.3S. cuspidatum10.0 ± 0.39.8 ± 0.611.3 ± 0.610.5 ± 0.510.8 ± 0.811.5 ± 0.4b9.7 ± 0.4aP (mg g−1) S. fuscum0.39 ± 0.030.19 ± 0.07a0.30 ± 0.02ab0.40 ± 0.03b0.41 ± 0.04b0.31 ± 0.040.34 ± 0.04S. balticum0.36 ± 0.030.13 ± 0.090.26 ± 0.040.29 ± 0.030.29 ± 0.020.32 ± 0.040.22 ± 0.050.19 ± 0.05S. magellanicum0.29 ± 0.010.10 ± 0.06a0.18 ± 0.04ab0.27 ± 0.04ab0.37 ± 0.08b0.26 ± 0.070.21 ± 0.050.21 ± 0.03S. cuspidatum0.31 ± 0.030.10 ± 0.080.18 ± 0.040.23 ± 0.020.26 ± 0.040.24 ± 0.040.15 ± 0.04K (mg g−1)S. fuscum54 ± 0.24.9 ± 0.2a5.5 ± 0.3a7.1 ± 0.3b7.1 ± 0.2b6.4 ± 0.35.9 ± 0.3S. balticum4.7 ± 0.44.7 ± 0.3a5.2 ± 0.2a6.4 ± 0.3b6.5 ± 0.3b5.6 ± 0.35.9 ± 0.35.5 ± 0.3S. magellanicum2.2 ± 0.23.8 ± 0.1a4.7 ± 0.2b6.6 ± 0.2c6.9 ± 0.2c5.6 ± 0.35.3 ± 0.45.4 ± 0.3S. cuspidatum2.2 ± 0.34.5 ± 0.3a5.0 ± 0.2a6.6 ± 0.3b6.9 ± 0.4b6.1 ± 0.3b5.5 ± 0.3aField values (n = 5) are given for comparison. Number of observations for each species for each temperature treatment: S. fuscum and S. cuspidatum, n = 10; S. balticum: n = 14–15; S. magellanicum, n = 15. Number of observations for each species for each combination, n = 19–20 P and K concentrations were affected by temperature (P concentration, F = 9.32, P < 0.001; K concentration, F = 82.77, P < 0.001) and species (P concentration, F = 3.05, P = 0.031; K concentration, F = 3.86, P = 0.011). P and K concentrations were higher at high temperature for all species (Table 4). P-values decreased, compared to field values, while K concentration increased compared to field values. Evaporation and water content Water content decreased with temperature treatment (Table 5). Water content was lowest in S. balticum in mixture with S. fuscum. Evaporation increased with temperature and therefore with VPD for all species (Table 5). Evaporation in S. fuscum in both monoculture and mixture was lower than in other species. Table 5Water content (% volume) at 5 cm depth and evaporation (mm day−1) per temperature treatmenta (see Table 1) and per species combinationb. Data are mean values ± SE. Different letters indicate significant differences between treatments (P < 0.05). There were no significant interactions between temperature and competition treatmentsWater contentEvaporationTemperature 188 ± 2c11.2 ± 0.02a 279 ± 3b21.5 ± 0.03b 360 ± 3a31.7 ± 0.04c 460 ± 3a42.0 ± 0.05dMonoculturesS. fuscum83 ± 5cS. fuscum1.3 ± 0.05aS. balticum66 ± 6bS. balticum1.7 ± 0.08bcS. magellanicum74 ± 4bcS. magellanicum1.8 ± 0.08cS. cuspidatum78 ± 4bcS. cuspidatum1.8 ± 0.08cMixturesS. fuscum74 ± 5bcS. fuscum + S. balticum1.5 ± 0.07aS. balticum49 ± 5aS. magellanicum69 ± 5bS. magellanicum + S. balticum1.6 ± 0.07bS. balticum68 ± 5bS. magellanicum82 ± 5cS. magellanicum + S. cuspidatum1.7 ± 0.08bcS. cuspidatum73 ± 5bcNumber of observations for each temperature treatment: for water content, n = 50; for evaporation, n = 35. Number of observations for each competition treatment: for water content and for evaporation, n = 20 Discussion Effect of temperature on growth As expected, all species in monoculture increased height increment with temperature (Fig. 1). All species also showed a looser growth form with higher temperatures, which resulted in lower densities (Table 3). Despite the lower bulk density at higher temperatures, biomass production still increased with temperature for all species (Table 2, Fig. 4). The response of height increment to temperature was about a factor 2 larger than the response of biomass production. Growth of Sphagnum was probably increased by a higher rate of photosynthesis and a higher N availability. Photosynthesis in Sphagnum has been shown to increase with temperature up to an optimum around 20–25°C (Skre and Oechel 1981; Harley et al. 1989). When temperature increased, the nutrient concentrations in Sphagnum were higher (Table 4). Moreover, N and K concentration increased compared to field values (Table 4). As biomass production also increased, N and K uptake must have increased considerably with increasing temperature. No nutrients were added with the rainwater solution, so the higher availability of nutrients must have come from Sphagnum itself and the peat below. Probably the lower Sphagnum parts in the containers decomposed faster when the temperature was higher, making more N and K available for growth. This temperature-induced stimulation of nutrient mineralization is probably larger than it would be in field conditions as the peat soil in the containers is surrounded by warm air on every side, resulting in relatively high soil temperatures in our experiment. However, it can be expected that increased temperature as a result of climate change will also enhance decomposition rates in field situations (Hobbie 1996), thereby increasing nutrient availability for both Sphagnum and vascular plants. The lower N concentrations of S. fuscum and S. balticum compared with S. magellanicum and S. cuspidatum at temperatures 1 and 2 are probably caused by the different sites of origin. In the northern site, the N deposition is much lower (approximately 0.2 g N m−2 year−1) than in the southern site (approximately 0.7 g N m−2 year−1) where S. magellanicum and S. cuspidatum were collected. In this experiment we provided near optimal growing conditions for Sphagnum with high water level and high humidity, which explains why the Sphagnum plants could grow with a lower bulk density and still keep their capitula moist. However, Sphagnum water content still decreased with an increase in temperature. In field conditions it would probably not have been possible for the Sphagnum plants to change their density so much since it would have led to problems in maintaining the water content of the capitula. The effect of temperature on height increment might therefore also have been less. In a few field studies, no positive or even negative effects of temperature on production were found. However, in these experiments there were other factors besides temperature that could have had a negative effect on biomass production, like drought stress (Hobbie et al. 1999; Weltzin et al. 2001; Gunnarsson et al. 2004) and higher vascular plant cover, which could have resulted in etiolation of the Sphagnum plants and reduced biomass production because of increased shading (Gunnarsson et al. 2004). In other studies also positive effects of temperature on height growth and production were found for different species, in both greenhouse (Robroek et al. 2007b) and field experiments (Moore 1989; Gerdol 1995; Sonesson et al. 2002; Dorrepaal et al. 2003). In accordance with our hypothesis, Asada et al. (2003) suggested that Sphagnum species respond positively to increased temperature and that hollow and lawn species are more sensitive to temperature than hummock species. This, however, does not correspond with the low response of S. cuspidatum to temperature in our experiment. Dorrepaal et al. (2003) also found a decrease in bulk density with increased temperature. Since the response of height increment and bulk density to temperature were in opposite directions, the dry matter production of S. fuscum did not show a significant increase with temperature. Our relatively high height increment and low density of S. fuscum compared to the values of Dorrepaal et al. (2003) can probably be explained by the much higher water levels in our experiment. Dorrepaal et al. (2003) used water levels of −20 to −40 cm below moss surface during their measurements. However, the Sphagnum densities we measured (Table 3) agreed quite well with densities from other studies (Lindholm and Vasander 1990; Moore et al. 1998; Asada et al. 2003). From our results, we conclude that the height increment and biomass production of all Sphagnum species in bogs can increase with an increase in temperature. However, the literature shows us that, particularly under field conditions, this potential response may not be realized in instances of competition from vascular plants, drought stress or extreme temperature increases. Effect of temperature on competition We hypothesized that increased temperature will have a positive effect on competitive abilities of S. fuscum and S. magellanicum. This hypothesis was confirmed in the competition with S. balticum, since the height increment and biomass production of S. balticum at the highest temperature treatment were no longer larger than those of S. fuscum and S. magellanicum (Figs. 2, 4). However, S. balticum still increased its cover at the expense of S. fuscum and S. magellanicum in all mixtures (Fig. 3a, b). This could be expected as the water level we used was relatively high and in the range of the natural habitat of these species. These results correspond to the findings from a field experiment of Rydin (1986), who found an expansion of S. balticum at the expense of S. fuscum at a water level of −1 cm. At water levels of −5 and −10 cm, he found a small expansion of S. balticum, but at a water level of −15 cm S. fuscum showed large expansion. In agreement with the hypothesis, the difference in height increment and biomass production between S. fuscum and S. balticum disappeared as temperature increased (Figs. 2a, 4a). S. fuscum kept increasing height increment with temperature, but S. balticum already reached maximum height growth at temperature 2, which is comparable to the summer temperature in the site of origin of these species. Also, the expansion in cover became lower for S. balticum at temperature 4 (Fig. 3a). As a result, the difference in biomass production between S. fuscum and S. balticum slowly decreased with an increase in temperature until there was no significant difference at the highest temperature. Apparently, the hummock species S. fuscum is better able to profit from a high increase in temperature than S. balticum, even at relatively high water levels. This is confirmed by the stepwise regression analysis which shows that biomass production of S. fuscum is correlated most to temperature, and biomass production of S. balticum to water content. As could be expected, S. fuscum as a hummock species has a higher water content than S. balticum, also at higher temperatures (Table 5), and therefore can profit more from increased temperature, because water does not become limiting for growth. S. fuscum also seems to be the better competitor for nutrients. The N and P concentrations in S. fuscum tended to be higher in the mixtures than in the monocultures, whereas in S. balticum the opposite was the case (Table 4). Indeed, S. fuscum is known for its high N uptake rate (Jauhiainen et al. 1998). Our results show that S. balticum loses competitive strength with increasing temperature. With a slight increase in temperature in northern bogs, S. balticum may remain the stronger competitor at high water levels, but at higher temperature increases, S. fuscum will gain a competitive advantage over S. balticum, also because increased temperature may lead to lower water tables in bogs. Recent climate change simulation models predict an increase of 2.5°C in summer temperature in northern Europe, with increasing precipitation in northern Scandinavia (Christensen et al. 2007). This limited rise in temperature, almost similar to the increase from temperature 2 to 3, would not affect competition between S. balticum and S. fuscum very strongly. However, the effect of a predicted increase in winter temperature of 5.0°C (Christensen et al. 2007) could shift the competitive balance between S. balticum and S. fuscum further, because increased snow cover and spring temperature have been shown to increase S. fuscum production (Dorrepaal et al. 2003). For S. magellanicum and S. cuspidatum, temperature 3 is comparable to the summer temperature in their site of origin. There was no effect of a temperature increase from temperature 3 to 4 on biomass production and relative performance of the species. This implies that with the predicted increase in temperature of 2.5°C (Christensen et al. 2007), the competitive balance between S. magellanicum and S. cuspidatum will remain unchanged. The mixtures of S. balticum and S. magellanicum show the largest difference in height increment and biomass production at temperature 2, which is comparable to the summer temperature in the site of origin of S. balticum. Also the expansion in cover of S. balticum was highest at this temperature. When temperature increases, the difference in height growth and biomass production became smaller and then disappeared in mixtures, whereas in monocultures the difference in biomass production between species remained. This implies that S. balticum as a northern species can enhance its production with an increase in temperature as long as it grows in monocultures. If S. magellanicum expanded its distribution further to the north as a result of global warming, it would be a strong competitor for S. balticum, thereby reducing S. balticum production. Gunnarsson et al. (2004) also showed that S. balticum performs less well under competition with increased temperatures. They performed a competition experiment in a poor fen, close to the site of origin of the S. balticum material. They found that when Sphagnum papillosum was transplanted to an area with S. balticum, S. papillosum decreased in area by 30%. However, when temperature was increased by 3.6°C, S. papillosum increased by 42%. This treatment is comparable to the increase from temperature 2 to temperature 3 in our experiment but with lower water tables, ranging from 5 to 17 cm under the moss surface. Generally, we conclude from our experiment that an increase in temperature can favour hummock and lawn species compared to hollow species. This corresponds to the findings of Mauquoy et al. (2002), who showed from peat core analyses that in periods with lower temperature, due to decreased solar activity, there was a shift in representation from lawn and hummock species to hollow species in ombrotrophic mires in Denmark and the UK. In conclusion, our findings suggest that production rates can increase and that Sphagnum species’ abundances can shift in response to global warming, particularly at northern sites. There, hollow species such as S. balticum will lose competitive strength relative to hummock species such as S. fuscum and southern species such as S. magellanicum. What the consequences for the C balance of bog ecosystems will be depends not only on the production rates, but also on decomposition. In general, decomposition rates also increase with temperature and increased N availability (Hobbie 1996; Limpens and Berendse 2003) and hummock species decompose slower than hollow species (Rochefort et al. 1990; Johnson and Damman 1993; Limpens and Berendse 2003), but virtually nothing is known about the temperature sensitivity of decomposition of bog plant species.

Pediatr_Radiol-3-1-2077923

Fractional anisotropy in white matter tracts of very-low-birth-weight infants

Background Advances in neonatal intensive care have not yet reduced the high incidence of neurodevelopmental disability among very-low-birth-weight (VLBW) infants. As neurological deficits are related to white-matter injury, early detection is important. Diffusion tensor imaging (DTI) could be an excellent tool for assessment of white-matter injury. Introduction The last decade has seen improved survival for very-low-birth-weight (VLBW) preterm infants [1, 2]. The survivors, however, still show a high incidence of neurodevelopmental disability, despite advances in fetal and neonatal intensive care [3–5]. These neurological deficits are usually related to white-matter injury [6, 7]. Thus early detection of abnormal white-matter maturation is important in the design of preventive, protective, and rehabilitative strategies for the management of the critically ill newborns [8, 9]. The clinical evaluation of these infants may not provide adequate diagnostic or prognostic information on white-matter injury. Various neuroimaging techniques have become available, however, that may be of help. Cranial US is such a technique used in premature infants, but it is not as sensitive as MR imaging, which shows better soft-tissue contrast [10]. Yet, conventional T1-weighted (T1-W) and T2-weighted (T2-W) MR imaging sequences also have limited value in the evaluation of the preterm brain, as they do not allow visualization of specific white-matter tracts before the onset of myelination. Most white-matter tracts of the premature cerebrum are unmyelinated except for a few pathways, e.g. the pyramidal tract that matures early as the infant approaches term-equivalent age [11, 12]. Diffusion tensor imaging (DTI) is unique in its ability to visualize and quantify white-matter tracts in the human brain. It is superior to T1-W and T2-W imaging in detecting unmyelinated or premyelinated fibre tracts [13] and, therefore, likewise in assessing the microstructural organization of the developing white matter. DTI anisotropy measurements offer great potential in the study of white-matter damage in VLBW infants. So far, however, the clinical use of this technique is restricted by the lack of normal reference values for VLBW infants. Although there are reports on fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values in preterm infants, only four groups have reported values for infants of gestational age less than 32 weeks without brain injury (Table 1). Each of the populations studied included no more than four infants without white-matter damage on conventional imaging, scanned within days after birth and gestational age less than 32 weeks [14–17]. Because the sizes and shapes of the regions of interest (ROIs) in these studies differ, comparison between the studies is difficult. We therefore set up a retrospective study aimed at providing reference values for the clinical interpretation of DTI images in VLBW infants. Table 1Overview of relevant literatureReferenceGestational age (weeks)Scan weekNo. of patientsNo. with gestational age <32 weeksMR field strength (T)ROI14<3728–4327<51.5FTa1524–3628–3917<51.5Manual1624–3628–3914<51.5Manual1726–3028–406<51.5ManualaROI placement with fibre tracking. Materials and methods Subjects The Erasmus MC Ethical Review Board approved the study and written informed parental consent was obtained for each subject. The images used in this study had been obtained within the framework of a study in which premature infants of different gestational ages underwent serial conventional and DTI acquisitions to evaluate white-matter development. The inclusion criteria for our study were birth at gestational age 25–32 weeks, no evidence of white-matter injury on conventional MRI and scanned within 4 days of birth. Developmental outcome assessed at age 1–3 years needed to be normal. Gestational age was calculated from the mother’s last menstrual period or estimated from early US (<18 weeks of pregnancy). Exclusion criteria were intraventricular haemorrhage, ventriculomegaly, congenital infection, brain malformation or a multiple congenital anomaly syndrome. Furthermore, preterm infants whose images showed severe motion degradation were excluded since evaluation of these scans was not possible. During the 25 months of the study period (March 2004 to April 2006), 41 infants were scanned and 32 met our study criteria. Four MR examinations were excluded due to severe motion artefacts, so 28 patients were included. Neurodevelopment monitoring The infants’ hospital charts were reviewed for neurodevelopmental outcome. They had all been neurologically assessed at 3, 6, 12 and 24 months, and Denver scores were used to monitor neurodevelopment [18]. At 24 months the Bayley scales of infant development, second edition (BSID-II) had also been applied [19]. All patients had shown normal neurodevelopment according to the most recent examination. Conventional MR imaging Images were acquired using standard scanning protocols. All imaging was performed on a 1.5-T GE EchoSpeed scanner (GE Medical Systems, Milwaukee, Wis.). The imaging protocol included T1-W (spin echo, TR/TE 500/11 ms) and T2-W sequences (spin echo, TR/TE 3,000/120 ms) with a slice thickness of 4 mm and 0.4 mm gap. For DTI an echoplanar sequence with diffusion gradients (b = 1,000 s/mm2) applied in 25 non-collinear directions was used with a slice thickness of 3 mm and no gap. An average of 20 slices was recorded within 4 min using TR/TE 9,150/98–91 ms. The FOV was 20 cm, the scan matrix 128×128 and the reconstruction matrix 256×256. All patients were scanned using an MR-compatible incubator with a specialized high-sensitivity neonatal head coil (Lammers Medical Technology, Lubeck, Germany) that allowed DTI imaging at high spatial resolution and high signal-to-noise ratio (SNR). The incubator provided controlled temperature and humidity as well as MR-compatible pulse oximetry and ventilation. This set-up allowed imaging in this most vulnerable patient population in a stable and safe microenvironment [20–22]. We used mouldable earplugs and neonatal earmuffs to reduce the noise. No sedation was given. Pads around the infant’s head kept movement to a minimum. Diffusion tensor postprocessing Diffusion tensor images were transferred to a GE Advantage Windows workstation (General Electric Medical Systems, Milwaukee, Wis.) for postprocessing using Functool 2000 software (General Electric). DTI measures the diffusion of water in each voxel and the extent to which water diffuses in particular directions as a result of the microstructural characteristics of the tissue imaged. High anisotropy indicates that the magnitude of diffusion is very unequal in different directions. The complex nature of anisotropic diffusion in the brain has been described by a diffusion tensor, which contains information about the magnitude of diffusion in different directions. Each tensor contains a set of three eigenvalues, which are related to the major, intermediate, and minor axes of a diffusion ellipsoid. The principal eigenvector (λ1) specifies the direction in which water diffusion is greatest. FA measures the fraction of the magnitude of the diffusion tensor that can be ascribed to anisotropic diffusion [23–25]. For isotropic diffusion (λ1 = λ2 = λ3), FA is zero, and in the case where there is a strongly preferred direction of diffusion (λ1>>λ2 = λ3), FA approaches a value of one. The colour maps were based on major eigenvector orientation with red representing right–left, green representing anteroposterior, and blue representing superoinferior anatomical directions. The tracts selected for quantization in the study included commissural tracts (corpus callosum: splenium and genu), projection tracts including those of the posterior limb of the internal capsule (PLIC), the anterior limb of the internal capsule (ALIC) and the optic radiation (OR), and association tracts (external capsule, EC). Regions of interest DTI measurements were taken from multiple ROIs positioned bilaterally within individual white-matter tracts. For the placement we used standard-size, round-shaped 16-pixel ROIs. To optimize our ROI placement we first detected the patient scan slice that represented slice “O” of the atlas of Wakana et al. [26]. We then placed the ROI, combining the FA, ADC and colour map to verify the right location. When the two researchers both agreed on the position, the ROI was used as a seed point to perform fibre tracking (minimum length 10 mm, minimum anisotropy 0.10). We compared the fibres that were visualized with the fibre bundles described in the atlas of Wakana et al. for verification (Figs. 1, 2 and 3). We then looked for the maximum pixel value within the ROI (Figs. 4 and 5). Two researchers experienced in neonatal DTI achieved consensus on ROI placement and measurements. Fig. 1Example of an FA map (gestational age 28 weeks) on which the ROIs are placedFig. 2Confirmation of ROI placement on the colour map (gestational age 28 weeks)Fig. 3Fibre tracking was performed in each of the ROIs to confirm the correct location. This is an example showing the corticospinal tracts (gestational age 28 weeks)Fig. 4Delineation of the corpus callosum using free-hand ROI placement on an FA map (gestational age 28 weeks)Fig. 5Automated calculation of the maximum FA pixel value within the ROI on an FA map (gestational age 28 weeks) Statistical analysis The relationship between FA of the white matter tracts and gestational age was analysed by correlation analysis (Pearson product moment correlation, SPSS 13.0.1). Tract comparison was done using one-way analysis of variance (ANOVA). Results Patient characteristics Gestational ages at birth ranged from 26 to 32 weeks (mean 28 weeks 5 days). Mean weight at birth was 1,148 g and mean head circumference was 26.4 cm. All patients had shown normal neurodevelopment as defined by the Denver or Bayley scoring system according to the most recent examination. FA and ADC of ROIs of standard pixel size We plotted the average FA and ADC values of 16-pixel ROIs of the different white-matter tracts against gestational age. No statistically significant differences based on a two-tailed paired Student’s t-test (at a significance level of P<0.05) were found comparing FA and ADC values between the left and right hemispheres in each subject. Therefore, the values obtained from the left and right were averaged to obtain the mean FA and ADC of the different ROIs. We found a significant correlation between gestational age and FA of the PLIC (r = 0.495, P<0.01; Fig. 6). No significant correlations were found for the other tracts (Table 2). We found no correlation between average ADC values and gestational age in the studied tracts (Table 2). Fig. 6Average FA values of 16-pixel ROIs of the PLICTable 2Tract statistics (FA and ADC mean values and standard deviations) and significant differences in FA and ADC between tractsStructureFAADCMean (SD)Tracts showing significant differences (P<0.05)Mean (SD)Tracts showing significant differences (P<0.05)Posterior limb of internal capsule0.349 (0.028)a, b, c1.09 (0.05)l, m, n, o, pAnterior limb of internal capsule0.242 (0.033)a, d, e, f1.27 (0.091)lExternal capsule0.175 (0.188)b, g, h, i1.29 (0.097)mOptic radiation0.270 (0.042)g, j, k1.30 (0.126)nCorpus callosum (genu)0.42 (0.048)d, e, h, j1.24 (0.111)oCorpus callosum (splenium)0.442 (0.056)c, f, i, k1.27 (0.131)p Maximum FA values of the PLIC The maximum FA values of the ROI were plotted against gestational age and also showed a significant correlation between gestational age and FA of the PLIC (r = 0.569, P<0.01; Fig. 7). We saw no other significant correlations between FA and gestational age. Compared with standard-size ROI mean FA values of the PLIC, the maximum values were a mean of 0.05 higher. Fig. 7Maximum FA ROI values of the PLIC Comparison between tracts DTI values varied between white-matter structures (Table 2). FA was highest in the commissural tracts of the corpus callosum (splenium>genu) and deep projection tracts (internal capsule). FA was lowest in association tracts (external capsule). Discussion Our aim was to provide clinicians with reference values for the evaluation of diffusion tensor images of white-matter tracts in VLBW infants. A literature search identified four similar studies. Comparison with these studies seems irrelevant as each included no more than four infants evaluated below a gestational age of 32 weeks within the first days of life. We were able to study scans of 28 VLBW infants without white-matter damage within the first 4 days of life. We found a statistically significant correlation between gestational age and FA of the PLIC. Differences were observed between white-matter tracts: FA was highest in commissural, followed by deep projection and association tracts. Others have found the same hierarchy in the perinatal brain [16] and in fully myelinated adult tracts [27]. Different anisotropy indexes can be used to study the anisotropy of white-matter tracts: FA, relative anisotropy (RA), axial anisotropy etc. We focused on FA, the most frequently used anisotropy index. FA has a high sensitivity for studying anisotropy in white-matter tracts [16]; it has a higher SNR than RA for anisotropic regions [28, 29]. Previous studies have shown that FA is the most sensitive and significant discriminating DTI parameter in pair-wise comparisons between different tracts in premature infants [16]. In most tracts FA increases significantly by the week, whereas weekly decreases in ADC values are smaller and less discriminative in most regions [16]. There are a growing number of reports of FA values of injured tracts [30–32]. Nevertheless, we should realize that FA is only one of the functions that describe the diffusion tensor. Although aberrant FA values seem to be a sensitive detector of abnormal tissue one could also use the relative magnitudes of the three principal eigenvectors to describe diffusion. What determines anisotropy in white-matter tracts? The value of the anisotropic index is determined by size, shape and composition of physical obstructions, as well as the space between them. Myelin, fast axonal transport, the axonal cytoskeleton (neurofilaments and microtubules), and local susceptibility gradients do not seem to have any major impact on anisotropic water diffusion [26, 33]. Indeed, intact membranes are thought to be the main contributing factor to anisotropy. The increasing anisotropy of white matter takes place before the histological appearance of myelin [15]. This increase has been attributed to the ‘premyelinating state’ [13, 34]. The corpus callosum, with a coherent parallel organization, has the highest FA values. Callosal fibre tracts, however, do not show a significant increase in FA with increasing gestational age, as the PLIC does. This might be due to the fact that myelination progresses more slowly in the callosal fibres than in the projection fibres at this age. During development a decrease in ADC values of the white-matter tracts is expected. We found no correlation between average ADC values and gestational age in the studied tracts. When looking at the available data reported previously we only saw the ADC values decrease significantly with gestational age when the ADC values of the infants older then 35 weeks were included. The fact that we could not find any significant correlation could be due to the limited number of infants scanned. The alternative is that at an earlier gestational age there is no significant decrease in ADC in the studied tracts. Studying DTI parameters of the white-matter tracts of VLBW infants is challenging for many reasons. One of these challenges is to determine a standard for the size and shape of the ROIs. A ROI is a controlled identification of a given area of an image for numerical analysis and the area of anatomy being scanned that is of particular importance in the image. Different authors have used different ways to set their ROIs. The reason we compared two techniques was to achieve better reproducibility. Maximum values (the maximum pixel value within a ROI) might be an alternative in which the value is given by the software. However, theoretically there is a bigger change of artefacts. Fibre tracking, colour maps and ADC maps are established but time-consuming techniques for the verification of tracts. We trust that automatic verification will become common practice in the future. A possible solution for ROI comparisons between researchers is the use of a neonatal brain atlas coordinate system. Individual brain images could then be transformed into a common coordinate space and the ROIs could be placed at specific topographic coordinates. Our research group is currently looking into this option. Another serious challenge is the SNR and spatial resolution constraints due to the very low anisotropy of premyelinating white matter and the tiny size of white-matter tracts in premature newborns. Using a custom-made MR-compatible incubator with a high-sensitivity neonatal head coil that improved image quality, spatial resolution and patient comfort, we were able to overcome this challenge. In this retrospective study a b value of 1,000 s/mm2 was used to allow comparisons between early and later scans at different ages. We realize a lower b value would be optimal because the ADC value in neonatal white matter is higher than that of more mature brains. The optimal b value can be calculated as indicated by Jones et al. [35]. Like all VLBW infants discharged from our NICU, our study patients were routinely seen by trained paediatricians and by paediatric physical therapists at the outpatient department for neurodevelopmental follow-up. Neurodevelopmental outcome was defined according to the most recent neurological examination and Denver and Bayley scores. The children’s ages at the most recent assessments varied, making the results difficult to compare; also Denver and Bayley scores are limited in their prognostic value below the age of 2 years. Better understanding of normal preterm white matter development is essential to encourage the use of DTI for evaluation and treatment of white-matter injury. Early diagnosis of white-matter abnormalities means that early intervention might be possible. We are exploring the feasibility of perinatal brain repair, and new MR imaging techniques such as DTI will enable us to improve our understanding of how the developing brain responds to our interventions. Conclusion Our study gives anisotropy values for VLBW infants with normal outcome that can be used as reference values. This work adds to our understanding of normal preterm white-matter development.

Ann_Hematol-4-1-2226003

Iron and thrombosis

Although essential for cell physiology, an increase or depletion of body iron has harmful effects on health. Apart from iron deficiency anemia and iron overload-related organ tissue damage, there are increasing evidences that body iron status is implicated in atherosclerotic cardiovascular diseases. The hypothesis formulated in 1981 that iron depletion may protect against cardiovascular events is intriguing and has generated a significant debate in the last two decades. Indeed, to study this phenomenon, several investigators have tried to design appropriate experimental and clinical studies and to identify useful biochemical and genetic markers of iron status. The results of the literature on the effect of iron deficiency and overload on vascular health are critically reviewed in this study from a pathogenic and clinical point of view. Introduction Iron is an essential nutrient for living cells because of its role as a cofactor for enzymes in the mitochondrial respiration chain, in the DNA synthesis, and being the central molecule for binding and transport of oxygen by hemoglobin and myoglobin. While the lack of iron leads to growth arrest and anemia, an increased accumulation of this metal is associated with toxic radical formation and progressive tissue damage. It is interesting to note that both iron deficiency and excess have been associated with an increased risk of developing thromboembolic events [1–7]. This review will analyze, from a clinical and pathogenic point of view, the existing literature data on the relationship between iron and arterial and venous thrombosis. Iron deficiency and thrombosis There are several reports in the literature on thrombotic complications in iron-deficient children and adults [8–26]. Secondary thrombocytosis has been implicated in many cases. Indeed, iron deficiency is a cause of reactive thrombocytosis, usually mild [27]. For instance, within a study group of children with iron deficiency, reactive thrombocytosis was found in up to one-third of them [28]. Nagai et al. [29] reported a case of severe iron deficiency with marked thrombocytosis (1,020 × 109/l) that was complicated by central retinal vein occlusion. By contrast, Kinoshita et al. [26] described two cases of cerebral venous sinus thrombosis associated with iron deficiency and normal platelet count. Hartfield et al. [13] reported six children with iron deficiency who developed an ischemic stroke or venous thrombosis. Four of them had a concomitant thrombocytosis. In a prospective case–control study, Stolz et al. [30] found that severe anemia, along with thrombophilia and hypercholesterolemia, were independent risk factors for cerebral venous thrombosis. The mechanisms causing reactive thrombocytosis in iron deficiency anemia are not completely understood. Iron is an important regulator of thrombopoiesis [31, 32]. Whereas normal iron levels are required to prevent thrombocytosis by inhibiting thrombopoiesis, a minimum amount of iron is required to maintain platelet production. Thus, while thrombocytosis is usually associated with a mild iron deficiency and is the result of a lack of inhibition of thrombopoiesis, a severe defect of this metal may be accompanied by thrombocytopenia. However, studies on thrombopoietic cytokines failed to show any effect on reactive thrombocytosis in iron deficiency [27]. For instance, Akan et al. [33] assayed the serum levels of thrombopoietin, erythropoietin, leukemia inhibitor factor, IL-6, and IL-11 in patients with iron-deficient anemia with or without elevated platelet count. Only erythropoietin level was elevated, correlated with thrombocytosis, and decreased with iron replacement. The other cytokines remained unchanged after therapy, suggesting that they probably do not play any significant role in iron deficiency-associated reactive thrombocytosis. Recently, Bilic and Bilic reported that the amino acid sequence homology of thrombopoietin and erythropoietin may explain the thrombocytosis in children with iron deficiency anemia [34]. By contrast, two other reports suggested that the relationship between iron deficiency and reactive thrombocytosis is more complex than a mere consequence of a crossreactivity between erythropoietin and thrombopoietin [35, 36]. In addition to the increased thrombotic risk associated with high platelet count, other authors have suggested that the decrease in antioxidant defense in iron deficiency anemia may cause increased oxidant stress, which in turn may result in a tendency toward platelet aggregation [37]. Thus, the abnormal platelet count and function observed in iron deficiency anemia could act synergistically to promote thrombus formation, especially in the setting of an underlying atherosclerotic disease [23]. However, as not all cases of iron-related thrombotic events occur in patients with concomitant high platelet count [13, 26], other pathogenic mechanisms have been proposed in these last years. Thus, iron deficiency may contribute to a hypercoagulable state by affecting blood flow patterns within the vessels because of reduced deformability and increased viscosity of microcytic red blood cells [13]. Furthermore, anemic hypoxia secondary to iron deficiency could precipitate situations of increased metabolic stress (i.e., infections) in particularly vulnerable areas of the brain supplied by end arteries, such as the basal ganglia, thalamus, and hypothalamus [38]. This phenomenon could explain the association between iron-deficient anemia and reversible focal deficits and stroke found by some authors [1, 13, 39, 40]. Iron overload and thrombosis Accumulation of iron in excess of physiologic requirements has been implicated in the development of several chronic illnesses, including cardiovascular diseases [7]. As previously reported, iron is a prooxidant cofactor associated with an increased production of hydroxyl radical in cardiovascular tissues and increased progression of atherosclerosis in experimental models [41–44]. Indeed, in an experimental mouse thrombosis model, a moderate iron overload markedly accelerated thrombus formation, impaired vasoreactivity, and enhanced the production of reactive oxygen species and systemic markers of oxidative stress [41]. It is interesting to note that the administration of dl-cysteine, a reactive oxygen species scavenger, completely abrogated the iron load-induced thrombus formation thus corroborating the hypothesis that iron accelerates thrombosis through a prooxidant mechanism. Similarly, laboratory investigations have demonstrated iron-dependent generation of reactive oxygen species in endothelial cell cultures and increased aortic atherosclerosis in the apolipoprotein E-deficient mice and cholesterol-fed rabbits with increased iron intake [42–44]. The research in this field has focused on the evaluation of the impact of iron depletion and iron overload on cardiovascular outcomes. Sullivan first postulated in 1981 the “iron hypothesis” by which the chronic iron depletion has a protective effect against ischemic heart disease and may account for the reduced risk of cardiovascular events in menstruating women [45–47]. Basing on these observations, some investigators have studied the effect of serial blood donations on the coronary heart disease risk, but their results were inconsistent [48–53]. However, an important support to the hypothesis of a potential link between blood donation and reduced cardiovascular risk came from a recent study conducted by Zheng et al. [54] who found that high-frequency blood donors had decreased serum ferritin levels, a marker of body iron stores; decreased serum 3-nitrotyrosine levels, a marker of oxidative stress; and greater flow-mediated dilation in the brachial artery, a marker of vascular function. It is interesting to note that in a clinical study, the iron chelation with deferoxamine improved endothelial function in patients with coronary artery disease [55]. By contrast, in a multicenter, randomized controlled trial (the iron [Fe] and Atherosclerosis Study [FeAST]) on 1,277 patients with symptomatic peripheral arterial disease, the reduction of body iron stores by phlebotomy did not significantly decrease all-cause mortality or death plus nonfatal myocardial infarction and stroke [56]. Similarly, in a prospective analysis of the second National Health and Nutrition Examination Study (NHANES II), Sempos et al. [57] observed either no association (in Caucasian men) or a possible nonsignificant increased risk (in Caucasian women) of cardiovascular or coronary heart disease death among individuals with low ferritin concentrations. On the other hand, the association between biochemical markers of body iron load and the risk of developing cardiovascular disease have been investigated by several studies [7, 58–72]. In the Kuopio Ischemic Heart Disease Risk Factor Study (KIHD), serum ferritin levels were found to be one of the strongest risk factors for acute myocardial infarction among Finnish men [61]. Similar findings were reported from a Canadian study [65] that observed an increased risk of myocardial infarction among subjects in the highest serum iron category and from the Bruneck study [64] in which a positive association between serum ferritin levels and ultrasound measures of progression of carotid atherosclerosis over a 5-year follow-up period was found. A relationship between serum ferritin levels and carotid atherosclerosis was also identified by a recent study conducted by Wolff et al. [71]. Haidari et al. [60] observed a significant correlation between serum ferritin levels and risk of coronary heart disease in male Iranian patients. However, a number of epidemiological studies did not find an association between iron status and coronary artery disease [61, 69, 70, 72]. For instance, Bozzini et al. [61] found that the mean serum ferritin concentrations were slightly higher in coronary heart disease patients than in controls, but this difference disappeared after adjusting for sex and C-reactive protein. The 1996 discovery of HFE gene mutations responsible for most cases of hereditary hemochromatosis has led to the use of genetic markers of iron overload, which are not influenced by external factors such as inflammation, in epidemiologic studies. Thus, several authors have investigated in recent years the relationship between C282Y and H63D mutations in the HFE gene and the risk of cardiovascular diseases [73–88]. Three prospective population-based studies have reported an association between heterozygotes and vascular events [75–77]. The first study was from a subgroup of the original Finnish KIHD cohort [75]. Of 68 individuals, 8 (11.8%) were diagnosed with acute myocardial infarction, and of 1,150 noncoronary heart disease participants, 77 (6.7%) were carriers of C282Y. The crude relative risk of myocardial infarction was 2.0 (95%CI = 0.9–4.1) and the adjusted relative risk was 2.3 (95%CI = 1.1–4.8). In a cohort of 12,239 Dutch postmenopausal women, the C282Y carrier status was assessed among 531 women who died of cardiovascular disease and 555 randomly selected women who did not die of cardiovascular disease [76]. This study reported a relative risk of 1.6 (95%CI = 1.1–2.4) for total cardiovascular death. Finally, in the United States Atherosclerosis Risk in Communities (ARIC) study [77], a C282Y carrier frequency of 9.9% among 243 coronary heart disease cases and 6.1% among 535 controls was reported. The crude relative risk of coronary heart disease associated with C282Y carrier status was 1.6 (95%CI = 0.9–3.0) and was 2.7 (95%CI = 1.2–6.0) after being controlled for other risk factors. To support of these findings, Gaenzer et al. [78] found an association between increased iron stores and impaired endothelial function (measured as endothelium-dependent dilation and intima-media thickness) in patients homozygous for C282Y mutation. Iron-depletion therapy normalized the endothelial function in such patients thus reducing the increased risk of cardiovascular events. However, the majority of the studies disagreed with these results [61, 79–89]. Indeed, Bozzini et al. [61] found a similar rate of carriers of C282Y mutation among patients with coronary atherosclerotic disease and controls. In a case–control study on 1,098 subjects, Rossi et al. [80] found that C282Y mutation was not a predictor of asymptomatic carotid atherosclerosis. Franco et al. [81] reported that the HFE genes were not associated with coronary or peripheral atherosclerosis in patients aged less than 50 years. Similarly, in the West of Scotland Coronary Prevention Study (WOSCOPS), Gunn et al. [86] found that the presence of a C282Y mutation in the HFE gene did not predict the occurrence of coronary events over a mean follow-up of 4.9 years. It is interesting to note that Yunker et al. [89] analyzed the relationship between biochemical and genetic markers of iron overload and carotid intima-media thickness and brachial flow-mediated vasodilation by high-resolution ultrasound in 907 males, but neither ferritin nor hemochromatosis genotype were related to brachial endothelial function and carotid atherosclerosis. In addition, a recent large study from Denmark found no increased risk of coronary heart disease among carriers of the C282Y mutation or individuals who had compound heterozygosity for the C282Y and H63D mutations [79]. Other investigation have focused on the association between genetic markers of iron overload and idiopathic dilated cardiomyopathy and stroke, but their results were conflicting [90–94]. Finally, some authors have suggested that HFE C282Y could interact with other predisposing factors for venous thromboembolism, such as factor V Leiden, thus exacerbating their prothrombotic effect [95]. Table 1 summarizes the most important studies on the association between genetic markers of iron overload and cardiovascular diseases. Table 1Summary of the most important studies on the association between HFE gene mutations (C282Y and H63D) and the risk of cardiovascular diseasesAuthors [reference]Study designPopulationResultsTuomainen et al. [75]Prospective1,150 individualsC282Y heterozygosity is associated with a 2.3 RR for AMI compared with noncarriersRoest et al. [76]Prospective12,239 postmenopausal womenC282Y heterozygosity is associated with a 1.6 RR for TCD compared with noncarriersRasmussen et al. [77]Prospective243 CHD cases and 535 controlsC282Y heterozygosity is associated with a 2.7 RR for CHT compared with noncarriersGaenzer et al. [78]Case–control41 C282Y/C282Y cases and 51 controlsC282Y homozygosity is associated with impaired endothelial functionBozzini et al. [61]Case–control546 CHD cases and 303 controlsC282Y mutation is not associated with CHDRossi et al. [80]Case–control1,098 subjectsC282Y mutation is not a risk factor for asymptomatic carotid atherosclerosisFranco et al. [81]Case–control256 CHD cases and 272 controlsC282Y and H63D mutations are not associated with CHDEllervik et al. [79]Prospective9,178 individualsC282Y and H63D mutations are not associated with CHDCase–control2,441 CHD and 1,113 AMI cases vs 8,080 controlsC282Y and H63D mutations are not associated with CHDGunn et al. [86]Case–control482 CHD cases and 1,104 controlsC282Y mutation is not associated with CHDCampbell et al. [88]Case–control924 AMI cases and 1,029 controlsC282Y mutation is not associated with CHDYunker et al. [89]Case–control907 individualsHFE genotype is not related to brachial endothelial function and carotid atherosclerosisAMI: acute myocardial infarction, CHD: coronary heart disease, RR: relative risk, TCD: total cardiovascular death Conclusions It is interesting to note that although with different pathogenic mechanisms, both iron deficiency and overload have been associated with an increased thrombotic risk in experimental and clinical studies. However, several aspects need to be still elucidated in this field. In particular, large prospective controlled trials are needed to elucidate the role of genetic markers of iron stores and the impact of long-term iron depletion on morbidity and mortality from cardiovascular events.

Eur_Radiol-4-1-2292493

Breast tumor characteristics of BRCA1 and BRCA2 gene mutation carriers on MRI

The appearance of malignant lesions in BRCA1 and BRCA2 mutation carriers (BRCA-MCs) on mammography and magnetic resonance imaging (MRI) was evaluated. Thus, 29 BRCA-MCs with breast cancer were retrospectively evaluated and the results compared with an age, tumor size and tumor type matched control group of 29 sporadic breast cancer cases. Detection rates on both modalities were evaluated. Tumors were analyzed on morphology, density (mammography), enhancement pattern and kinetics (MRI). Overall detection was significantly better with MRI than with mammography (55/58 vs 44/57, P = 0.021). On mammography, lesions in the BRCA-MC group were significantly more described as rounded (12//19 vs 3/13, P = 0.036) and with sharp margins (9/19 vs 1/13, P = 0.024). On MRI lesions in the BRCA-MC group were significantly more described as rounded (16/27 vs 7/28, P = 0.010), with sharp margins (20/27 vs 7/28, P < 0.001) and with rim enhancement (7/27 vs 1/28, P = 0.025). No significant difference was found for enhancement kinetics (P = 0.667). Malignant lesions in BRCA-MC frequently have morphological characteristics commonly seen in benign lesions, like a rounded shape or sharp margins. This applies for both mammography and MRI. However the possibility of MRI to evaluate the enhancement pattern and kinetics enables the detection of characteristics suggestive for a malignancy. Introduction BRCA1 and BRCA2 are the most well known gene mutations responsible for an increased risk for developing breast cancer. A BRCA1 or BRCA2 mutation carrier (BRCA-MC) has approximately a 3% risk of getting breast cancer before the age of 30. This risk increases to almost 50% when she reaches the age of 50 and becomes 50–80% at the age of 70 [1, 2]. To reduce this risk, these women can choose between bilateral prophylactic mastectomy [3], oophorectomy [4] or chemoprevention [5]. In breast cancer, close surveillance contributes to a more favorable stage of disease at detection and may reduce the rate of death from breast cancer [6, 7]. In the surveillance or general screening for breast cancer, mammography still plays a prominent role. However, due to the young age and thus in most cases dense breast tissue, the sensitivity for mammography is moderate. False-negative rates of up to 62% have been reported for mammography in screening gene mutation carriers [8, 9]. A malignant lesion in the breast is mammographically best detected if it presents itself as an ill-defined or spiculated mass, a group of microcalcifications or as an architectural distortion. A smoothly outlined well-defined mass detected on mammography has a chance of less than 1% of being malignant [10, 11]. Tilanus et al. [12] and Kaas et al. [13, 14] have evaluated the mammographic appearance of breast cancer in BRCA-MC. Tilanus and coworkers found the mammographic appearance suspicious for a malignancy in only 38% of the gene carriers in comparison with 71% in a control group. “Prominent pushing margins” caused by a continuous front of tumor cells not separated by connective tissue were described in the BRCA-MC group as the main reason for a false-negative evaluation of mammograms [12]. Kaas and coworkers concluded in their study of 31 breast cancer cases in BRCA-MCs that all mammographically detected lesions should be further evaluated by ultrasound and biopsy regardless of their appearance [13]. Well-defined mammographic tumors correlated in 83% with histologic circumscribed tumor margins in BRCA1-MCs [14]. For women with an increased risk for developing breast cancer magnetic resonance imaging (MRI) should be included for close surveillance [15, 16]. The superior sensitivity of MRI (81%) for the detection of breast cancer in these women compared with mammography (40%) has been proven in literature [17]. However the classification of a lesion detected on MR as benign or malignant still remains a challenge. Morphological and dynamic features are important in breast MRI interpretation. Focal masses with smooth borders are associated with a high negative predictive value for malignancy [17]. An irregular lesion contour, inhomogeneous enhancement pattern and rim enhancement have been reported as features indicating malignancy [18]. The dynamic evaluation is often based on the enhancement characteristics 2–7 min after the injection of a paramagnetic contrast agent. In this approach, the decrease of signal intensity, often referred to as a type 3 curve or washout, is highly predictive for breast cancer, with a likelihood of malignancy of 87% [19]. Until now the appearance of breast malignancies in BRCA-MCs has only been investigated for mammography. In this study we analyzed the MRI characteristics of BRCA-MC-associated tumors compared with sporadic cases of breast cancer. Materials and methods All available (35) BRCA-MCs with a biopsy-proven malignancy, imaged with MRI for screening [9] or pre-operative evaluation in the period from July 2000 until November 2006, were included in the study: 23 BRCA1 carriers and 12 BRCA2 carriers. In order to compare tumor characteristics with sporadic cases of breast cancer an age, tumor type and tumor size matched control group was composed from 206 consecutive sporadic breast cancer cases imaged with MRI in the period from November 2001 until January 2007. All BRCA-MC cases were age matched within 5 years with sporadic breast cancer cases. Cases were also matched for tumor type (IDC, ILC or DCIS) and pathological tumor size. For size matching, the BRCA-MC cases were matched to the closest tumor size in the sporadic cases available, with a limit for the maximum size difference of 0.5 cm for tumors smaller than 1 cm, 1-cm difference for tumors up to 5 cm and 1.5-cm difference for tumors larger than 5 cm. BRCA-MC cases that could not be matched following these criteria were excluded. Mammograms were obtained in the mediolateral oblique and craniocaudal direction on a digital mammographic unit (Senograph 2000 D or a Senograph DS, GE Healthcare, Wis., USA). Detection, density of the lesion compared with breast tissue, lesion morphology, and size were scored. In the morphologic assessment, lesion type was classified as either a mass, a calcifications or as an architectural distortion. Lesion shape was described as rounded, lobulated or irregular and lesion margins as sharp, vague or spiculated. The size of the tumor was measured by determining the longest axis through the displayed lesion. Spiculae surrounding a solid lesion were interpreted as desmoplastic reaction and not included in the measurement. MRI investigations were performed on a 1.5-Tesla system with a double breast coil (Magnetom Vision, Sonata or Symphony, Siemens, Erlangen, Germany). In the scanning, we used a coronally orientated three-dimensional fast low-angle shot (FLASH 3D) with the following parameters: TE 4 ms, TR 8.1 ms, FA 20°, FOV 360 mm, TA 96 s, image resolution 1.5 mm × 1.5 mm × 1.5 mm for all patients scanned prior to June 2004 and TE 4 ms, TR 7.5 ms, FA 8°, FOV 320 mm, TA 87 s, image resolution 1.3 mm × 1.3 mm × 1.3 mm for all patients scanned after June 2004. Prior to the MR examination, an intravenous catheter was inserted. All patients were placed in the prone position, with the breasts in the double breast coil and positioned at the isocenter of the magnet. After localizer images were obtained in three directions and a precontrast FLASH 3D series was recorded, 0.1 mmol/kg bodyweight gadolinium chelate (Magnevist, Schering, Germany or Dotarem, Guerbet, The Netherlands) was administered using a power injector (Spectris, Medrad, USA) at 2.5 ml/s followed by a 15-ml saline flush at the same injection rate. Thereafter, five post contrast FLASH 3D series were recorded. All MRI examinations were retrospectively evaluated on a dedicated breast MRI workstation (Dynacad, Invivo, USA) scoring lesion detection, size, morphology and enhancement kinetics. Maximum intensity projections, coronal images and axial reconstructions of both the T1 weighted and subtracted images and time-intensity curves were displayed. The morphologic assessment included lesion shape, margin appearance and enhancement pattern. Lesion shape was classified as being rounded, lobulated or irregular. Margins were described as sharp, vague or spiculated. The enhancement pattern of a lesion was classified as homogeneous, heterogeneous or rim enhanced. Lesion enhancement kinetics were evaluated according to the criteria described by Kuhl et al. [19]. Type 1 shows persistent enhancement and is highly suggestive for a benign lesion. Type 2 shows a plateau after initial increased enhancement, where the maximum signal intensity is reached approximately 2–3 min after contrast injection and remains constant. This type of curve is seen in both benign and malignant lesions. In a type 3 curve, the peak enhancement is reached in the early postcontrast phase, and this is followed by a decrease of signal intensity (wash-out). The latter curve is strongly suggestive for a malignant lesion. The dynamic curves were evaluated based on a single voxel or by selecting a region of interest within the lesion, the workstation allowed the readers to use both methods. Because of the possible bias in this retrospective study, a BI-RADS classification [20] could not be scored objectively and was therefore not included in the evaluation. All studies were evaluated retrospectively by two radiologists in conference and consensus. BRCA-MCs and controls were mixed during the evaluation. Except from the knowledge of a malignancy being present, the radiologists were blinded to any other clinical information. Mammography and MRI images were evaluated in separate sessions. From the histopathology reports, the tumor type, size and mitotic activity index (MAI) were recorded. The study was approved by the institutional review board; since the study was performed retrospectively, informed consent was not required according to the review board. In the statistical evaluation, differences in patient and tumor characteristics between the BRCA-MC and control group were analyzed using an independent sample t-test if variables were continuous and normally distributed. For categorical variables, the Pearson chi-square test was used and we used Fisher’s exact test when any of the expected values was less than five. Pearson’s correlation coefficients were calculated for both mammographic size and MRI size versus pathologic size. All statistical analyses were performed using SPSS statistical software (version 12.0.1). P values <0.05 were considered to indicate statistical significance. Results Six BRCA-MC cases could not be matched according to the criteria defined; these cases were excluded from the study. Four BRCA-MCs were excluded because no match could be found based on patient’s age; the other two were excluded because no match could be found based on tumor size. The mean age and tumor size of the excluded cases were, respectively, 33 years (range 27–36, median 35, SD 3.4 years) and 1.4 cm (range 0.6–2.8 cm, median 1.1 cm, SD 0.8 cm). A total of 29 BRCA-MC cases were included for this study. In the BRCA-MC group, five women were symptomatic (17%); 21 women were symptomatic in the control group (83%). Mean age in the BRCA-MC group was 42 years (range 32–68 years, median 40 years, SD 8.0 years); this was 44 (range 37–64 years, median 43 years, SD 5.6 years) for the control group. The mean pathological tumor size was 2.0 cm (range 0.4–7.0 cm, median 1.4 cm, SD 1.5 cm) in the BRCA-MC group and 2.3 cm (range 0.6–7.0 cm, median 1.9 cm, SD 1.7 cm) in the control group. No significant difference was found for patient age (P = 0.289) or maximal pathological tumor size (P = 0.371). The mean tumor size on mammography was 2.1 cm (range 0.5–7.0 cm, median 1.5 cm, SD 1.49 cm). The mean tumor size on MRI was 2.4 cm (range 0.6–7.1 cm, median 1.8 cm, SD 1.75 cm). There was a significant correlation between imaging measurements and pathological measurements; 0.664 (P < 0.001) for mammographic measurements and 0.808 (P < 0.001) for MRI measurements. In both the BRCA-MC and control groups, 23 cases were based on invasive ductal carcinoma (IDC), two cases on invasive lobular carcinoma (ILC), one case on ductal carcinoma in situ grade 1 (DCIS1) and three cases on ductal carcinoma in situ grade 2 (DCIS2). In the BRCA-MC group, a mean MAI of 33.0 (range 6–100, median 27, SD 27.1) was found, compared with 17.5 (range 1–60, median 14, SD 15.9) in the control group. The difference in MAI between the BRCA-MC and control groups was found to be significant (P = 0.044). In the control group, one patient refused to undergo mammography because of implants. Overall mammography detected 44 of 57 lesions and MRI detected 55 of 58 lesions. Therefore, the overall detection is significantly better with MRI than with mammography (P = 0.021). Mammography detected 22 (76%) lesions in the BRCA-MC group and 22 (79%) in the control group. No significant difference was found (P = 0.807). All lesions missed on mammography in the BRCA-MC were IDC. In the control group, five cases of IDC and one case of ILC were missed. Mammographic lesion characteristics are presented in Table 1. Lesions in the BRCA-MC group were significantly more described as rounded (12//19 vs 3/13, P = 0.036) and were more often described to have sharp margins (9/19 vs 1/13, P = 0.024). Lesions in the control group were significantly more described as irregular (10/13 vs 6/19, P = 0.029). From the six BRCA-MCs that were excluded, only three lesions were detected on mammography. In two cases a mass was detected and in one case calcifications. Both these masses were described as rounded with sharp margins. Table 1Mammographic lesion characteristics for both groups BRCA-MC groupControl groupP valueLesions detectedn = 22n = 22Lesion typeMass19130.042Arch. distortion140.345Calcification250.216Lesion densityaHyperdense810.050Isodense11120.050Lesion morphologyaRounded1230.036Lobulated1–1.000Irregular6100.029Lesion marginsaSharp910.024Vague880.473Spiculated240.194aLesion density, morphology and margins of mass-like lesions only On MRI, 27 lesions (93%) were detected in the BRCA-MC group, 28 (97%) in the control group. No significant difference was found (P = 0.553). The lesions missed in the BRCA-MC group were both cases of DCIS, one case of DCIS1, seen on mammography as an architectural distortion, and one case on DCIS2, seen on mammography as a mass. The lesion missed in the control group was based on DCIS2, seen on mammography as calcifications. Their was no significant difference found for the detection of breast cancer between mammography and MRI within the BRCA-MC (P = 0.18) or the control group (P = 0.13). Morphological and dynamic MR characteristics are presented in Table 2. Lesions in the BRCA-MC group were significantly more often described as rounded (16/27 vs 7/28, P = 0.010), with sharp margins (20/27 vs 7/28, P < 0.001) and to show rim enhancement (7/27 vs 1/28, P = 0.025). Lesions in the control group were significantly more often described as irregular (18/28 vs 8/27, P = 0.010), with vague margins (15/28 vs 6/27, P = 0.017) and with a heterogeneous enhancement pattern (22/28 vs 12/27, P = 0.009). No significant difference between the two groups was found for enhancement kinetics (P = 0.667). From the six BRCA-MCs that were excluded, five were detected on MRI. Four of these lesions were described as rounded, one as irregular. The delineation was described as sharp in four and as vague in one of these cases. The enhancement pattern was described as homogeneous in two, heterogeneous in one and as rim-enhancement in two of these cases. All five cases showed a type 3 curve. Table 2MRI lesion characteristics for both groups BRCA-MC groupControl groupP valueLesions detectedn = 27n = 28Lesion morphologyRounded1670.010Lobulated331.000Irregular8180.010Lesion marginsSharp207<0.001Vague6150.017Spiculated160.049Enhancement patternHomogeneous850.304Heterogeneous12220.009Rim710.025Enhancement kineticsType 1121.000Type 2460.525Type 322200.380 Discussion In this study, the overall false-negative rate for mammography was significantly higher compared with MRI. Although it is expected that the sensitivity for MRI in both the control group and the BRCA-MC is higher compared with mammography, no significant difference was found in this study within the two groups. This is probably due to the relatively small numbers of cases in this study. Kaas et al. [13] described in their mammographic study on BRCA1 and BRCA2 mutation carriers a sensitivity of 64% for the detection of a tumor in the original reports. In this study, 76% of the lesions were visible on mammography. Since this was a retrospective study and the radiologist was aware of the fact that a tumor was present at the time of the evaluation, no conclusion can be drawn from the difference in detection between both studies. On mammography, the mass like lesions detected in the BRCA-MC group were significantly more often described as rounded. Also, lesions were found to differ in margin appearance; tumor margins in the BRCA-MC group were significantly more often described as sharp. A smooth, nonspiculated mass has previously been described by Tilanus et al. [12] as a reason for a false-negative mammographic evaluation in BRCA-MC. Thus, although Sickles et al. [21] described that nonpalpable, circumscribed, noncalcified breast masses (probably benign) should be managed with periodic mammographic surveillance regardless of lesion size and patient age, and Sardanelli et al. [22] published that both the well-defined margins and the rounded shape are more often associated with benign lesions, these findings are not applicable to the BRCA-MCs studied in this or other studies on this subject. Tilanus et al. [12] stated that any mammographic mass in BRCA-MCs must be regarded with suspicion. A similar conclusion was published by Kaas et al. [13]. An additional evaluation using ultrasound and biopsy of all lesions detected in BRCA-MC is mandatory, regardless of their morphological appearance. In their study of 28 BRCA-MCs, Hamilton et al. [23] also described the appearance of breast tumors on ultrasound. On ultrasound, 53% of the tumors were classified as either benign or indeterminate, making a biopsy of any detected mass inevitable. We are also of the opinion that any solid lesion detected in BRCA-MCs should be evaluated by a biopsy. In the screening of women with an increased risk for developing breast cancer, more tumors are detected by MRI compared with mammography [8, 9]. In this study, all lesions missed on MRI were cases of DCIS. MRI is known to have a lower sensitivity in detecting DCIS compared with invasive carcinomas [24], especially in low-grade DCIS. The low or intermediate contrast uptake that is often observed in pure DCIS and the absence of a type 3 curve can result in a false-negative evaluation [24]. Kriege et al. [9] found in the screening of 1,909 women with an increased risk for developing breast cancer, including 358 carriers of germ-line mutations, that MRI missed five cases of DCIS that were detected on mammography, with six noninvasive tumors detected in total in the study. However, Kuhl et al. [25] reported MRI to be more sensitive for DCIS compared with mammography in a prospective study of women with an increased risk for developing breast cancer. Although MRI proved to be more sensitive for the detection of DCIS compared with mammography, not all cases of DCIS were detected by either modality. Therefore, at this point both modalities are still needed in the screening of women with an increased risk for developing breast cancer. Similar to mammography, on MRI a significantly higher number of lesions were described as rounded and with sharp margins in the BRCA-MC group. Furthermore, the number of lesions with ‘rim-enhancement’ was found to be significantly higher in the BRCA-MC group. The presence of this enhancement pattern has been associated with malignant lesions [26, 27]. Because MRI enables the radiologist to evaluate the enhancement pattern of the lesion, where mammography does not, these lesions will become more suspectedly malignant, even though other morphologic features are more often seen in benign lesions. The association of rim-enhancement (Fig. 1) with central necrosis or insufficient microvessel growth can be an indicator for the growth rate of tumors. Jimenez and coworkers have described centrally necrotizing carcinomas to have an accelerated clinical course and early systemic metastasis [28]. An accelerated growth rate can be associated with a high MAI. In this study, the MAI was found to be significantly different between both groups. This is in agreement with Tilanus et al. [12], who also found the mitotic count to be significantly higher in tumors found in gene mutation carriers. The higher rate of sharp tumor margins and rim-enhancement may thus be explained by the more aggressive nature of tumors in BRCA-MCs. Several authors have implied that, due to the rapid growth rate of tumors in gene mutation carriers, the screening frequency should be adjusted [13, 29]. Komenaka et al. [29] suggest a higher screening frequency in carriers because of the high number of interval cancers found in their group. Half of these interval malignancies were already positive for lymph-node involvement. The 13 carriers in their study were screened with mammography. Mammography in carriers is not sensitive, particularly because the women are young and thus more often have dense breasts [30, 31]. The use of MRI in screening has already been a step forward since MRI can detect smaller tumors, often occult for mammography, that are less likely to have progressed into lymph-node involvement [9]. Fig. 1a An MLO mammogram from the right breast of a 42-year-old BRCA1-mutation carrier. b A coronal subtraction MR image of the same breast. An 11-mm sharply delineated rounded lesion is present on the mammogram projecting over the upper quadrants (arrow). On MRI, the same lesion was detected (arrow) with rim-enhancement. The rim-enhancement makes this lesion morphologically suspect malignant. Ultrasound guided core biopsy proved this lesion to be an invasive duct carcinoma Rounded, homogeneous enhancing lesions found on MRI are in general not considered suspicious. A homogeneous enhancement pattern, found in nine BRCA-MCs, does not contribute to the malignant nature of these lesions. Therefore, enhancement kinetics are of value. In this study, the dynamic analysis showed in both the BRCA-MC and the control groups a type 3 curve in, respectively, 82% and 71% of the cases. As described by Kuhl et al. [19], a type 3 enhancement curve is highly indicative for malignancy. Using this characteristic, even rounded, sharply delineated, homogeneous enhancing lesions become suspect malignant (Fig. 2). Fig. 2a A coronal subtraction MR image from the right breast of a 49-year-old BRCA1-mutation carrier. Lateral located in the right breast, a sharply delineated rounded, homogeneous enhancing lesion is visible with a longest diameter of 9 mm (arrow). b The relative enhancement versus time curve. The type 3 curve seen for this lesion was the only characteristic indicating a possible malignancy. Pathology proved this lesion to be an invasive duct carcinoma Despite all the findings discussed in this study, it remains questionable if the characterization of a lesion detected on either mammography or MRI in BRCA-MCs is even necessary. As the chance for these women to develop breast cancer is significantly increased, almost any detected lesion will in practice be classified as suspect malignant until proven otherwise. Sardanelli et al. [17] found a positive predictive value for MR in women with an increased risk of only 53% due to a high number of false positives. As stated previously for mammographically detected lesions, additional evaluation by core biopsy is the only definitive classification for lesions detected in this group of women. In the case of MRI screening in high risk women, short-term follow-up, target ultrasound or MRI-guided biopsies are therefore often indicated [32]. What the best strategy in this group of women will be, also in terms of cost effectiveness, needs to be further studied. We conclude that in BRCA-MC malignant lesions frequently have morphological characteristics that are commonly seen in benign lesions, like a rounded morphology or a sharp delineation. This applies for both mammography and MRI. However, the possibility of MRI to evaluate the enhancement pattern and enhancement kinetics of lesions enables the radiologist to detect characteristics suggestive for a malignancy.

Mol_Cell_Biochem-4-1-2226060

Modulation of calcification of vascular smooth muscle cells in culture by calcium antagonists, statins, and their combination

Background Vascular calcification is an organized process in which vascular smooth muscle cells (VSMCs) are implicated primarily. The purpose of the present study was to assess the effects of calcium antagonists and statins on VSMC calcification in vitro. Methods VSMC calcification was stimulated by incubation in growth medium supplemented with 10 mmol/l β-glycerophosphate, 8 mmol/l CaCl2, 10 mmol/l sodium pyruvate, 1 μmol/l insulin, 50 μg/ml ascorbic acid, and 100 nmol/l dexamethasone (calcification medium). Calcification, proliferation, and apoptosis of VSMCs were quantified. Results Calcium deposition was stimulated dose-dependently by β-glycerophosphate, CaCl2, and ascorbic acid (all P < 0.01). Addition of amlodipine (0.01–1 μmol/l) to the calcification medium did not affect VSMC calcification. However, atorvastatin (2–50 μmol/l) stimulated calcium deposition dose-dependently. Combining treatments stimulated calcification to a degree similar to that observed with atorvastatin alone. Both atorvastatin and amlodipine inhibited VSMC proliferation at the highest concentration used. Only atorvastatin (50 μmol/l) induced considerable apoptosis of VSMCs. Conclusion In vitro calcification of VSMCs is not affected by amlodipine, but is stimulated by atorvastatin at concentrations ≥10 μmol/l, which could contribute to the plaque-stabilizing effect reported for statins. Introduction Atherosclerosis is the principal cause of coronary artery disease, stroke, and peripheral artery disease, and is the major cause of mortality in the Western hemisphere. Vascular calcification is a prominent feature of atherosclerosis, and it is associated with an increased risk of myocardial infarction [1]. Vascular calcification refers to the deposition of calcium phosphate mineral, most often in the form of hydroxyapatite, in the vessel wall. Calcification of the vessel wall and heart valves is associated with ageing, diabetes, and uremia [2–4]. Vascular calcification is now considered to be an organized, regulated process comparable to bone mineralization. The presence of various components associated with bone mineralization, such as bone morphogenetic protein, osteocalcin, osteopontin, osteoblast-like cells, and matrix vesicles in atherosclerotic lesions supports this concept [5–7]. Vascular cells such as vascular smooth muscle cells (VSMCs) and pericyte-like cells play an important role in vascular calcification [8]. The lipophilic calcium antagonist (CA) amlodipine has been shown to restore cholesterol-induced membrane bilayer abnormalities in VSMCs derived from the atherosclerotic rabbit aorta [9, 10], thereby restoring normal calcium homeostasis. Other mechanisms, through which CAs may affect atherosclerosis development include inhibition of proliferation and migration of VSMCs [11, 12], and inhibition of lipoprotein oxidation [13]. In addition, CAs modify binding of monocytes to the endothelium, and activate synthesis of matrix components [14]. The effects of CAs on atherosclerotic calcification have not been widely studied, however. Lipids may contribute to atherosclerotic calcification [15, 16]. Recently statins have been shown to decrease the progression of coronary artery calcification and aortic valve calcification [17, 18]. Therapy with a combination of a CA and a statin might be more atheroprotective than either treatment alone [19, 20]. Combination therapy with CA and statin was shown to improve endothelial function and arterial compliance, to diminish LDL atherogenicity, and to slow the progression of atherosclerosis. So far, the effects of the combination therapy with CA and statin on vascular calcification have not been studied extensively. We used an in vitro model of vascular calcification by neonatal rat VSMCs, isolated by outgrowth from aortic explants. The purpose of this investigation was to identify the factors, and their concentrations, that are mandatory to induce calcification of aortic VSMCs in vitro. Since calcium is such a major component of advanced atherosclerotic lesions, we furthermore studied whether the CA amlodipine and the statin atorvastatin, alone and in combination, affected calcification of aortic VSMCs in vitro. Materials and methods Cell culture Vascular smooth muscle cells (VSMCs) were obtained from segments of aortas explanted from 2-day old Wistar rats. Aortic segments were obtained aseptically and cut open longitudinally. The endothelium was removed by gently rubbing the luminal side of the aortas over the surface of a tissue culture dish (Falcon). Subsequently, the aortas were placed, lumen side down, on the bottom of a tissue culture flask (Greiner), and allowed to adhere for approximately 3 h. Then tissues were immersed in growth medium consisting of Dulbecco’s modified Eagle’s medium (DMEM) (Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Life Technologies), penicillin (100 U/ml), and streptomycin (100 μg/ml) (both supplied by BioWhittaker Europe). Seven days later the aortic tissues were removed, and the VSMCs that had grown out were detached by trypsinization. The detached cells were resuspended in growth medium and seeded in tissue culture flasks (Greiner), 6 or 12-well plates or on glass cover slips. Immunocytochemistry Subconfluent cultures at early passage were examined for the presence of α-smooth muscle actin, SM myosin heavy chain and calponin using specific antibodies (anti-SM actin clone 1A4 (1:400), Sigma; anti-SM myosin heavy chain (1:100), SanverTech; anti-calponin (1:10,000), Sigma). After fixation in 1% formalin in PBS on ice for 30 min, cells were permeabilized with 0.1% Triton X-100 for 30 min, and subsequently incubated with the primary antibody for several hours at 4°C. Secondary antibodies were conjugated with fluorescein isothiocyanate (FITC) or Cy3 (Sigma). To identify cell nuclei, the cells were also stained with Hoechst 33342 (10 μg/ml; Molecular Probes). Immunofluorescent images were obtained using a fluorescence microscope (Nikon Eclipse) equipped with 20×, 40×, and 100× objectives and a digital camera (Nikon DXM1200). In vitro calcification of VSMCs Calcification of VSMC cultures was induced by the method of Shioi et al. [21] with minor modifications. When confluent, the incubation medium of VSMCs was switched from growth medium to calcification medium. Calcification medium consisted of DMEM (high glucose, 4.5 g/l) supplemented with 15% FBS, penicillin (100 U/ml), streptomycin (100 μg/ml), 8 mmol/l CaCl2, 10 mmol/l sodium pyruvate, 1 μmol/l insulin, 50 μg/ml ascorbic acid, 10 mmol/l β-glycerophosphate, and 100 nmol/l dexamethasone. The medium was replaced with fresh medium every 2–3 days. In several experiments, the concentrations of CaCl2, ascorbic acid, β-glycerophosphate, and dexamethasone were varied. In another series of experiments, calcification medium was supplemented with amlodipine (0.01–1 μmol/l), atorvastatin (2–50 μmol/l), or a combination of these drugs. The CA amlodipine was dissolved in absolute ethanol at a concentration of 1 mmol/l. This stock solution was to be diluted at least 1:1,000 in culture medium. Atorvastatin, an inhibitor of HMG-CoA reductase, was dissolved in absolute ethanol at a concentration of 10 mmol/l. This stock solution was to be diluted at least 1:200 in culture medium. Cells treated with calcification medium without added drugs were used as controls. Appropriate amounts of solvent (ethanol) were added to these controls. Addition of ethanol at these levels did not affect cell growth. After 3 weeks of incubation, calcification was quantified. Assessment of calcium deposition VSMCs were decalcified with 0.6 N HCl for 24 h. The calcium contents of the supernatants were determined by spectrophotometer using the o-cresolphthalein method (Roche Diagnostics). After decalcification, the cells were washed with PBS and scraped from the culture plate. The protein content was measured using the BCA protein assay (Pierce). The calcium content of the cell layer was normalized to protein content. Proliferation assay VSMCs were seeded at a density of 5 × 103–2 × 104 cells/well in a 96-well plate and allowed to attach overnight. Subsequently, the standard culture medium (DMEM, 10% FBS, antibiotics) was replaced by calcification medium supplemented with amlodipine, atorvastatin, or a combination of both. Cell proliferation was assessed at day 4 and day 9 using the Cell Proliferation Kit II (XTT, Roche), which is a colorimetric assay for the non-radioactive quantification of cell proliferation and viability. Detection of apoptosis To identify apoptotic cells by assessment of condensation of nuclear chromatin, VSMC cultures were stained with Hoechst 33342 (10 μg/ml; Molecular Probes) for 10 min in the dark. Immunofluorescent images were obtained using a fluorescence microscope (Nikon Eclipse) equipped with 20x, 40x, and 100x objectives and a digital camera (Nikon DXM1200). In 6 frames per treatment, an average of 250 nuclei per frame were assessed for the presence of apoptosis. Statistical analysis Results are expressed as mean ± SEM. For statistical analysis, SPSS 10.0 for Windows was used. Since the data were not normally distributed, non-parametric tests were used for comparisons between groups. Overall comparisons between groups were performed with the Kruskall–Wallis test. If only two groups were compared, Mann–Whitney rank sum tests were used. P values less than 0.05 were regarded as significant. Results VSMCs that have been cultured from rat aortic explants for approximately 7 days stained positive for α-smooth muscle-actin, smooth muscle myosin heavy chain, and calponin, confirming the smooth muscle cell nature of these cells (Fig. 1, online supplement). VSMCs of passage 3–6 were subsequently used in calcification experiments. Fig. 1(A) Rat aortic VSMC isolated by outgrowth, phase contrast. Immunofluorescent staining of neonatal rat VSMC incubated with (B) (and insert) anti-smooth muscle actin Ab, (C) anti-smooth muscle myosin Ab, and (D) anti-calponin Ab. Original magnification B and C ×100, D and insert ×200. VSMC = vascular smooth muscle cell Calcium deposition and dependence of extracellular Ca2+ concentration Cells were incubated with calcification medium supplemented with various Ca2+ concentrations for 21 days. At Ca2+ concentrations ≤3 mmol/l, hardly any calcium deposition was observed, but at Ca2+ concentrations >3 mmol/l, a dose-dependent increase in calcium deposition was observed (Fig. 2A). The amount of calcium deposition was positively correlated to calcium concentration in the culture medium (P < 0.01). We chose to continue with 8 mmol/l of Ca2+-ions added to the calcification medium. Fig. 2(A) Dose-dependent effects of CaCl2 on calcification of neonatal rat VSMCs. (B) Dose-dependent effects of β-glycerophosphate on calcification of neonatal rat VSMCs. VSMCs were treated for 21 days with calcification medium containing varying concentrations of Ca2+ ions or β-glycerophosphate. Control cultures (=con) were incubated with DMEM, 10% FBS, and antibiotics. Calcium deposition was quantified by o-cresolphthalein method. The data are presented as mean ± SEM (n = 3–9). *P < 0.05 vs. control cultures. ◆P < 0.05 vs. 2 and 3 mM calcium in culture medium. #P < 0.05 vs. all other treatments Calcium deposition and dependence of extracellular phosphate concentration Shioi et al. have demonstrated that β-glycerophosphate accelerates in vitro calcification of VSMCs and induces extensive calcium deposition in a manner analogous to in vitro mineralization by osteoblasts [21]. In our model of vascular calcification, the amount of calcium deposition was positively correlated to the β-glycerophosphate concentration in the calcification medium (P < 0.01; Fig. 2B). Calcium deposition and dependence of extracellular dexamethasone concentration To determine the contribution of added dexamethasone, we incubated the VSMCs with various concentrations of dexamethasone (10–1,000 nmol/l). No significant differences in calcium deposition were observed between the cells incubated with 0, 10, 100, and 1,000 nmol/l dexamethasone (Fig. 3A). Fig. 3(A) Dose-dependent effects of dexamethasone on calcification of neonatal rat VSMCs. (B) Dose-dependent effects of ascorbic acid on calcification of neonatal rat VSMCs. VSMCs were treated for 21 days with calcification medium containing varying amounts of dexamethasone or ascorbic acid. Calcium deposition was quantified by o-cresolphthalein method. The data are presented as mean ± SEM (n = 6–12). #P < 0.05 vs. all other treatments. ◆P < 0.05 vs. 0 and 5 μg/ml ascorbic acid Calcium deposition and dependence of extracellular ascorbic acid concentration To determine the contribution of ascorbic acid to calcium deposition, we incubated VSMCs with various concentrations of ascorbic acid (0, 5, 50, and 500 μg/ml). At the highest concentration (500 μg/ml) ascorbic acid was associated with significantly more calcium deposition than at all other concentrations (Fig. 3B). However, 500 μg/ml ascorbic acid caused a considerable decrease in pH of the culture medium. Since calcium deposition was positively correlated with ascorbic acid concentration (P < 0.01), we decided to use 50 μg/ml ascorbic acid in future calcification studies, the highest concentration of ascorbic acid that did not cause acidification of the medium. Effects of amlodipine on in vitro VSMC calcification To study the effect of the CA amlodipine on VSMC calcification, VSMCs were incubated for 2–3 weeks with calcification medium supplemented with various concentrations of amlodipine (0.01–1 μmol/l). Incubation of neonatal rat VSMCs with amlodipine had no effect on VSMC calcification, at none of the concentrations tested (Fig. 4A). Fig. 4 (A) Dose-dependent effects of amlodipine on calcification of neonatal rat VSMCs. (B) Dose-dependent effects of atorvastatin on calcification of neonatal rat VSMCs. (C) Effects of amlodipine, atorvastatin and a combination of both treatments on neonatal rat VSMC calcification. VSMCs were treated for 21 days with calcification medium containing varying concentrations of atorvastatin or amlodipine, a combination of both, or none of them (control). Calcium deposition was quantified by o-cresolphthalein method. The data are presented as mean ± SEM (n = 15). *P < 0.05 when compared to untreated control cultures. #P < 0.05 when compared to all other treatments Effects of atorvastatin on in vitro VSMC calcification To study the effect of the statin atorvastatin on VSMC calcification, VSMCs were incubated for 2–3 weeks with calcification medium supplemented with various concentrations of atorvastatin (2–50 μmol/l). Atorvastatin increased VSMC calcification dose-dependently (Fig. 4B). At a concentration of 2 μmol/l atorvastatin, calcium deposition was increased by 30% (P = 0.04) when compared to VSMCs incubated with atorvastatin-free calcification medium. At concentrations of 10 and 50 μmol/l atorvastatin, calcium deposition was increased 2.3-fold and 6.1-fold, respectively (P < 0.001) when compared to VSMCs incubated with atorvastatin-free calcification medium. Effect of a combination of CA and statin on in vitro VSMC calcification To study the effect of combining amlodipine and atorvastatin treatment on VSMC calcification, VSMCs were incubated for 2–3 weeks with calcification medium supplemented with 0.1 μmol/l amlodipine, 10 μmol/l atorvastatin, or a combination of these drugs in the same concentrations. Incubation of VSMCs with 0.1 μmol/l amlodipine had no effect on VSMC calcification. Incubation of VSMCs with 10 μmol/l atorvastatin resulted in a 2.2-fold increased calcium deposition when compared to control cultures treated with calcification medium only (Fig. 4C) (P < 0.001). Treatment with a combination of amlodipine and atorvastatin also resulted in a 2.2-fold increased calcium deposition when compared to control cultures treated with calcification medium only (P = 0.026). The combination therapy resulted in significantly more calcium deposition than treatment with amlodipine alone (P = 0.003), and as much calcium deposition compared to treatment with atorvastatin alone (n.s.). Effects of amlodipine, atorvastatin, and their combination on VSMC proliferation VSMC proliferation was assessed at day 4 and day 9. After 4 days of incubation in calcification medium supplemented with amlodipine (1 μmol/l), VSMC proliferation was decreased by 21% (P < 0.001) as compared to proliferation of VSMCs cultured in amlodipine-free calcification medium (Fig. 6A, online supplement). After 9 days of incubation, VSMC proliferation had increased by 14% (P < 0.05) when treated with 0.01 μmol/l amlodipine, but at an amlodipine concentration of 1 μmol/l, VSMC proliferation had decreased by 26% (P < 0.01) as compared to proliferation of VSMCs cultured in amlodipine-free calcification medium (Fig. 5A). Atorvastatin, at a concentration of 50 μmol/l, decreased VSMC proliferation by 50% (P < 0.001; Fig. 5B). Treatment of VSMCs with the combination of 0.1 μmol/l amlodipine and 10 μmol/l atorvastatin did not result in significant changes in cell proliferation compared to VSMCs incubated in the absence of these drugs. Fig. 5Effect of amlodipine and atorvastatin on VSMC proliferation. VSMC proliferation was assessed at day 4 and day 9 using the Cell Proliferation Kit II (XTT)(Roche). VSMC proliferation is depicted as percentage relative to untreated cells (=100%) (A) Effect of amlodipine. (B) Effect of atorvastatin. #P < 0.05 vs. control cultures. *P < 0.05 vs. control cultures Effects of amlodipine, atorvastatin, and their combination on VSMC apoptosis VSMCs incubated in calcification medium only and VSMCs incubated with amlodipine (0.01–1 μmol/l) or atorvastatin (2 and 10 μmol/l) in calcification medium for 72 h hardly contained any apoptotic nuclei. Amlodipine at the highest concentration tested (1 μmol/l), caused apoptosis of on average 0.05% of the cells (n.s. compared to control). At the highest concentration of atorvastatin (50 μmol/l), however, 5.6% of the nuclei were apoptotic (P < 0.05 vs. control, 2 and 10 μmol/l atorvastatin; Fig. 6). Fig. 6VSMCs were treated with 1 μmol/l amlodipine (A, B) or 50 μmol/l atorvastatin (C, D) for 72 h. Nuclei were visualized with Hoechst 33342. Apoptotic nuclei are circled in panel C and indicated with arrows in panel D. Microscopy: magnification A and C ×100, B and D ×400 VSMCs treated with the combination of 0.1 μmol/l amlodipine and 10 μmol/l atorvastatin for 72 h had a number of apoptotic nuclei that did not significantly differ from those observed in 10 μmol/l atorvastatin only, 0.1 μmol/l amlodipine only, and VSMCs incubated in calcification medium in the absence of these drugs. Discussion In an in vitro model of vascular calcification of rat aortic VSMCs, calcium deposition was dependent upon the extracellular concentration of organic phosphate (P < 0.01), calcium ions (P < 0.01), and ascorbic acid (P < 0.01). While studying the effect of the CA amlodipine and the statin atorvastatin, alone and in combination, on calcification, we found that, at none of the concentrations tested (0.01–1 μmol/l) did amlodipine have any effect on VSMC calcification in this model, nor on development of apoptosis. In contrast, atorvastatin stimulated VSMC calcification at a concentration of ≥10 μmol/l (P < 0.05), and 50 μmol/l inhibited VSMC proliferation and induced apoptosis (P < 0.05). VSMCs retain remarkable plasticity, even in adult animals. VSMCs can undergo rapid and reversible changes in its phenotype in response to changes in local environmental conditions [22]. Due to production of a variety of cytokines by subendothelial macrophages, proliferation, and migration of VSMCs to the intimal layer can occur, thereby contributing to the process of intimal thickening and atherosclerosis. Neonatal VSMCs resemble VSMCs in atherosclerotic plaque, since they retain the proliferative phenotype. Neonatal aortic explants exhibit rapid outgrowth of VSMCs, and we chose to use these cells to develop a model of vascular calcification. VSMCs were shown to migrate from the aortic explant within several days, and were identified as VSMCs by immunohistochemistry [8]. Dialysis patients have accelerated atherosclerosis, with extensive calcification of both the intima and the media [23]. Hyperphosphatemia has been implicated in this process [24]. Inorganic phosphate (Pi) levels have previously been shown to regulate human aortic smooth muscle cell (HSMC) culture mineralization in vitro [25, 26]. In the present study, the amount of calcium deposition in neonatal rat VSMCs correlated positively to the β-glycerophosphate concentration in the calcification medium. In the present study, the amount of calcium deposition in rat aortic VSMCs correlated to the calcium concentration in the calcification medium. This observation was not a result of passive calcium deposition, since incubation of VSMCs with normal growth medium supplemented with 8 mmol/l CaCl2 (final concentration) did not result in calcium deposition by these cells. Shioi et al. [21] have previously demonstrated that culturing bovine VSMCs in the presence of β-glycerophosphate, ascorbic acid, and insulin can induce diffuse calcification in a manner analogous to in vitro mineralization of osteoblasts. The present study also showed that β-glycerophosphate is crucial for the induction of calcification in rat aortic VSMCs. Ascorbic acid augments this process. Dexamethasone was demonstrated to be less vital in this model of aortic VSMC calcification. Statins possess potent lipid-lowering effects. Besides, statins exert pleiotropic effects on vascular wall cells, which include improvement of endothelial function, stabilization of the atherosclerotic plaque, and suppression of inflammation [27, 28]. Although several studies have failed to demonstrate an effect of statins on the progression of calcific aortic stenosis [29, 30], statins have also been shown to decrease the progression of coronary artery calcification and aortic valve calcification [17, 18, 31, 32]. Accordingly, Kizu et al. have demonstrated that statins inhibit calcification in an in vitro model of inflammatory vascular calcification [33]. Using interferon-γ, 1α, 25-dihydroxyvitamin D3, tumor necrosis factor-α, and oncostatin M to induce calcification in human VSMCs, it was demonstrated that cerivastatin and atorvastatin inhibited calcification dose-dependently. However, one should realize that this model differs from the model we have used in the present study. Whereas Kizu et al. used inflammatory mediators to induce calcification of the VSMCs, we used increased levels of CaCl2 and organic phosphate in the culture medium to promote calcification. Statins are known to have anti-inflammatory effects, and therefore, their inhibitory effect on calcification induced by inflammatory mediators is to be expected. In addition, low doses of statins have also been shown to protect human aortic VSMCs from inorganic phosphate-induced calcification [34]. In contrast to the aforementioned studies, the present study showed a dose-dependent stimulatory effect of atorvastatin (10 and 50 μmol/l) on calcification of VSMCs incubated in calcification medium. However, the dose of atorvastatin used in the present study was much higher than in the study described by Son et al. (500x higher at the highest dose) [34], and most likely explains the discrepancy between the results obtained by Son et al. and the results of the present study. Reynolds et al. [35] have demonstrated that human VSMCs undergo vesicle-mediated calcification in response to changes in calcium and phosphate concentrations in the culture medium. Elevated calcium and phosphate concentrations resulted in increased release of vesicles and stimulation of apoptosis. Calcification was initiated by release of membrane-bound matrix vesicles from living cells and also by apoptotic bodies from dying cells. Vesicles released by VSMCs after prolonged exposure to calcium and phosphate contained preformed basic calcium phosphate, and calcified extensively. The present study confirms that statins stimulate apoptosis of VSMCs. Apoptotic bodies can calcify extensively [36], so a likely mechanism of induction of calcification by atorvastatin is through induction of apoptosis. Indeed, atorvastatin stimulated both apoptosis and calcification in our model. Myocardial infarction and unstable angina usually arise from the disruption of mildly stenosed atherosclerotic lesions. Such vulnerable plaques typically contain a large amount of lipid and have a preponderance of inflammatory cells at the shoulders of the plaque and a thin fibrous cap. Calcium is found infrequently in the culprit lesion of ruptured plaque. Thus, plaque vulnerability is not associated with size, but with composition. Beckman et al. [37] presented the concept “that calcium is not a critical substrate for plaque disruption and is, in fact, associated with more stable plaques.” Therefore, statin-induced alterations in the plaque’s composition from lipid to fibrous tissue without or with calcification would decrease the plaque’s vulnerability to rupture, thereby decreasing the risk of acute coronary syndromes. Prolonged intensive lipid-lowering has been demonstrated to increase calcium content of plaques as assessed by MRI in vivo, and lowered plaque lipid content at about the same extent, without a pronounced effect on plaque size [38]. Since atherosclerotic plaque composition plays an important role in plaque stability, with lipids destabilizing the plaque and calcification stabilizing the plaque, the effect of statins on plaque composition may be favorable. In addition to lipid-lowering and calcification-inducing effects, we observed that statins have an anti-proliferative effect on VSMCs. Less proliferation of VSMCs may destabilize the plaque. However, this only occurred at high doses of atorvastatin, whereas calcification induction occurred at lower dose of atorvastatin. Although several in vitro and in vivo studies have demonstrated that combining CA and statin therapy might be more atheroprotective than either treatment alone [19, 20, 39], the present study found no difference between the combination of amlodipine and atorvastatin therapy, and atorvastatin therapy alone on VSMC calcification. In conclusion, extracellular β-glycerophosphate and CaCl2 concentrations are important determinants of in vitro calcification of rat aortic VSMCs. Ascorbic acid stimulates this process. Dexamethasone was demonstrated to be less vital in this model of VSMC calcification. In vitro calcification of rat aortic VSMCs is not affected by amlodipine treatment, but is stimulated by atorvastatin treatment. As high concentrations of atorvastatin have a pro-apoptotic effect, the pro-calcification effect of atorvastatin may be explained by the production of apoptotic bodies that act as foci for calcium deposition. The latter finding may explain the plaque-stabilizing effect reported for statins.

Pediatr_Nephrol-3-1-2100434

Growth hormone axis in chronic kidney disease

Chronic kidney disease (CKD) in children is associated with dramatic changes in the growth hormone (GH) and insulin-like growth factor (IGF-1) axis, resulting in growth retardation. Moderate-to-severe growth retardation in CKD is associated with increased morbidity and mortality. Renal failure is a state of GH resistance and not GH deficiency. Some mechanisms of GH resistance are: reduced density of GH receptors in target organs, impaired GH-activated post-receptor Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling, and reduced levels of free IGF-1 due to increased inhibitory IGF-binding proteins (IGFBPs). Treatment with recombinant human growth hormone (rhGH) has been proven to be safe and efficacious in children with CKD. Even though rhGH has been shown to improve catch-up growth and to allow the child to achieve normal adult height, the final adult height is still significantly below the genetic target. Growth retardation may persist after renal transplantation due to multiple factors, such as steroid use, decreased renal function and an abnormal GH–IGF1 axis. Those below age 6 years are the ones to benefit most from transplantation in demonstrating acceleration in linear growth. Newer treatment modalities targeting the GH resistance with recombinant human IGF-1 (rhIGF-1), recombinant human IGFBP3 (rhIGFBP3) and IGFBP displacers are under investigation and may prove to be more effective in treating growth failure in CKD. Learning objectives: To review the normal growth pattern in childhood and its alteration in chronic kidney disease. To understand the mechanisms believed to be responsible for growth failure in chronic kidney disease. To understand the new evidence supporting therapy for growth failure in chronic kidney disease. To identify targets for future therapies for treatment of growth retardation in chronic kidney disease. Introduction Growth retardation is a common result of chronic kidney disease (CKD) in childhood. Children with CKD fail to achieve the final adult height consistent with their genetic potential [1]. Data from the North American Pediatric Renal Trials and Collaborative Studies (NAPRTCS) 2005 database revealed that 36.9% of children with CKD had statutory growth impairment. Even though the growth failure was correlated with the degree of renal impairment, those with mild reduction of glomerular filtration rate (GFR) also exhibited short stature. Children were short at initiation of dialysis, with younger age associated with more severe growth failure and little improvement in height standard deviation score (SDS). The mean height deficit for all renal failure patients was −1.85 (less than the third percentile of their healthy peers) at transplantation. This deficit was greater for male patients and for younger subjects. Those below age 6 years benefited most from transplantation in demonstrating acceleration in linear growth. For subjects aged 6–12 years, the linear growth remained stable, and those older than 12 years of age had no increase in their height deficit score [2–4]. Despite advances in medical care, growth failure in CKD is associated with increased morbidity and mortality. Furthet al. demonstrated from the United States Renal Data System (USRDS) database that patients with severe-to-moderate growth failure had increased hospitalization rates and increased risk of death [5]. Growth retardation is assessed by the SDS or height deficit score, which measures the patient’s height in comparison with that of unaffected children of similar age. Wong et al. demonstrated that, among pediatric patients on dialysis or after transplantation, each SDS decrease in height was associated with an increase in death by 14% [6]. Alteration of normal growth patterns in CKD Linear growth is unique to childhood. Complex biological processes are responsible for maintaining normal growth. Infancy is the fastest growing period of childhood; one-third of the total growth occurs in the child’s first 2 years of life, and this is mainly nutrient dependant. This growth is markedly decreased in congenital CKD, with the greatest height deficit occurring in the child’s first year of life, especially during the first 4 months [7]. Thus, the earlier the onset of kidney disease, the more severe the growth disturbance. Additional factors affecting growth are inadequate protein and calorie intake, water and electrolyte losses in polyuric and salt wasting conditions, anemia, metabolic acidosis, renal osteodystrophy, and resistance to hormones mediating growth. Early nutritional intervention and prevention of metabolic deficits of renal failure can preclude the development of growth failure in this period. Conservative treatment (no use of growth hormone) in pre-dialysis patients from birth to 3 years, revealed a favorable growth rate, with height velocity at 22.2 cm/year, 10.9 cm/year and 7.6 cm/year for each year, respectively, all of which were higher than the lower two standard deviation scores [5]. Further discussion of these factors is beyond the scope of this article. During the mid-childhood period, a constant growth rate of 5–7 cm/year is demonstrated, mainly under the control of growth hormone (GH) and thyroid hormone. Children with congenital CKD grow at the percentile achieved at the end of 2 years of life. Even though one of the most important predictors of growth impairment is the degree of renal insufficiency, significant short stature has been seen at all levels of renal function [3, 7]. At the onset of puberty, the growth hormone/insulin-like growth factor (GH/IGF-1) axis is activated by small increases in estrogen and testosterone in girls and boys, respectively. The onset of puberty is delayed in adolescents with CKD; peak height velocity is delayed by approximately 2.5 years. The pubertal growth spurt is delayed, shortened and associated with a reduced growth velocity. The pubertal height gain is about 65% of that seen in healthy children without CKD and is likely due to the shortened growth spurt [1]. These effects of delayed puberty in CKD are mediated by a loss of the normal pulsatile hypothalamic release of the gonadotropin-releasing hormone (GnRH). GH and IGF axis Growth hormone is the key endocrine regulator of postnatal growth (Fig. 1) [8]. The anterior pituitary secretes growth hormone in a pulsatile manner stimulated by the growth hormone releasing hormone (GHRH) and inhibited by somatostatin. GH mediates its somatotropic actions directly and mainly through IGF-1. Ghrelin, a growth hormone-releasing peptide expressed in the stomach and hypothalamus, is involved in hormonal and nutritional regulation of GH release [9]. Fig. 1GH/IGF-1 axis in CKD: deranged somatotropic axis in chronic renal failure. The GH/IGF-I axis in CRF is changed markedly, compared with the normal axis shown here. In CRF the total concentrations of the hormones in the GH/IGF-I axis are not reduced, but there is reduced effectiveness of endogenous GH and IGF-I, which probably plays a major role in reducing linear bone growth. The reduced effectiveness of endogenous IGF-I likely is due to decreased levels of free, bioactive IGF-I as levels of circulating inhibitory IGF-binding proteins (IGFBPs) are increased. In addition, less IGF-I is circulating in the complex with acid labile subunit (ALS) and IGFBP-3 as a result of increased proteolysis of IGFBP-3. Together, these lead to decreased IGF-I receptor activation and a decreased feedback to the hypothalamus and pituitary. Low free IGF-I and high IGFBP-1 and IGFBP-2 levels probably contribute to reduced renal function and lead to reduced stature. The direct effects of GH on bone, which are poorly understood, also are blunted. Reprinted from [8] with permission Growth hormone and the IGF-1 axis play a major role in growth failure in CKD. Random fasting serum levels of GH are normal or increased in children and adults with CKD, depending on the extent of renal failure. The half-life of GH is prolonged, due to decreased metabolic clearance secondary to decreased functional renal mass in proportion to the degree of renal dysfunction. A high normal calculated GH secretion rate and amplified number of GH secretory bursts have been reported in pre-pubertal children with end-stage renal disease, likely due to attenuated IGF-1 feedback [8]. This has led to the concept of GH insensitivity or resistance in uremia. Pubertal patients with advanced CKD have a reduced GH secretion rate, indicating altered sensitivity of somatotropic hormones to the stimulatory effects of sex steroids [10]. GH resistance GH receptors One mechanism for GH resistance is a reduced density of GH receptors in target organs. Determination of the concentration of serum growth hormone binding protein (GHBP), which is a cleaved product of the GH receptor, may be used to assess GH receptor density in tissues, particularly liver, since GHBP is derived mainly, but not exclusively, from the liver. GHBP is low in children and adults with CKD and proportionate to the degree of renal dysfunction. Serum GHBP correlates with both spontaneous growth rate and response to GH therapy, and it is an indirect indicator of sensitivity to both exogenous and endogenous GH [11, 12]. However, there is controversy as to the reliability of serum GHBP level as a marker of GH receptor levels in specific tissues [13]. Janus kinase/signal transducer and activator of transcription signaling Another mechanism for GH resistance in uremia is a defect in post-receptor GH-activated Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling (Fig. 2) [14]. GH action is mediated by the binding of GH to the growth hormone receptor (GHR), resulting in its dimerization and the auto-phosphorylation of the tyrosine kinases, Janus kinase 2 (JAK2), which, in turn, stimulates phosphorylation of signaling proteins, STAT proteins STAT1, STAT3 and STAT5. Upon activation, these STAT proteins translocate to the nucleus and activate GH-regulated genes. An intact JAK2-STAT5b signaling pathway is essential for GH stimulation of IGF-1 gene expression. In uremia, a defect in the post-receptor GH-activated JAK2 signal transducer and STAT transduction is described as one of the mechanisms of GH resistance [14]. Fig. 2Growth hormone-mediated JAK/STAT signal transduction. GH activates several signaling pathways via JAK2, including the JAK/STAT pathway [22, 23]. Binding of GH to its receptor (GHR) activates JAK2, which then self-phosphorylates. This is followed by phosphorylation of the GHR and, subsequently, STAT 1a, STAT 3, STAT 5a, and STAT 5b, members of a larger family of cytoplasmic transcription factors. These phosphorylated STATs form dimers that enter the nucleus, where they bind to specific DNA sequences and activate their target genes, IGF-1 and some suppressors of cytokine signaling (SOCS). Deletion of STAT5 expression leads to retarded body growth, and STAT5b is required for GH-mediated IGF-1 gene expression. In renal failure phosphorylation of JAK2 and the downstream signaling molecules STAT5, STAT3, and STAT1 is impaired, as are the nuclear levels of phosphorylated STAT proteins. This important cause of uremic GH resistance may result, in part, from up-regulation of SOCS2 and SOCS3 expression with suppressed GH signaling and also from increased protein tyrosine phosphatase activity, with enhanced dephosphorylation and deactivation of the signaling proteins. Reprinted from [14] with permissionThe JAK2/STAT pathway is regulated, among other factors, by suppressor of cytokine signaling (SOCS) proteins, which is induced by GH. These proteins bind to JAK2 and inhibit STAT phosphorylation [15]. Up-regulation of SOCS has been described in inflammatory states and may play a similar role in CKD [16]. IGF and IGFBP Daughaday et al. proposed the somatomedin hypothesis: GH leads to an increase in IGF-1 (formerly called somatomedin) c) from the liver, which, in turn, reaches the growth plate to mediate its action [17]. However, it was later demonstrated that all the effects of GH are not mediated through IGF-1. Also, the liver is not the only source of IGF-1, and it is produced in other tissues. Circulating IGF-1 mainly derived from the liver acts as an endocrine hormone. IGF-1 produced locally acts as a paracrine/autocrine hormone. The majority of circulating IGF in the serum exists as a 150 kDa complex, including the IGF molecule, IGF binding protein 3 (IGFBP3) and the acid labile subunit (ALS). IGF-1 gene knockout in mice results in miniature mice that die soon after birth. Knock-out of IGF-1 in the liver alone results in relatively normal sized mice with a 75% reduction in circulating IGF-1 levels, suggesting that local IGF-1 may be more important for body growth than liver IGF-1 [18]. ALS knock-out mice also exhibit relatively normal growth and development, despite a 65% reduction in circulating IGF-1 levels. Disruption of both liver IGF-1 and ALS genes results in 85–90% reduction in circulating IGF-1 levels and significant reduction in linear growth, suggesting that a threshold concentration of IGF-1 is necessary for normal growth [19].IGF-1 is the predominant IGF during the rapid growth period of puberty. The levels of IGFs (IGF-1 and IGF-2) are normal in pre-terminal renal failure. In end-stage renal disease IGF-1 level is normal or decreased and levels of IGF binding proteins (IGFBP) are increased, thus resulting in a net decrease in IGF bioactivity. Free IGF-1 is decreased in proportion to renal failure [20, 21].IGF-1 is an anabolic hormone; it binds to the IGF-1 receptor (IGF-1R) at the α subunit, where it activates tyrosine kinase intrinsic to the IGF-1Rβ subunit. In experiments with rats with CKD, skeletal muscles show decreased serum levels of IGF-1 and IGF-1 mRNA, while IGF-1R mRNA and IGF-1R numbers are increased, with normal binding affinity. However, there is resistance to post-receptor signal transduction; both auto-phosphorylation of the IGF-1R tyrosine kinase and activity of the IGF-1R tyrosine kinase to the exogenous insulin receptor substrate-1 (IRS-1), a natural substrate for IGF-1 receptor tyrosine kinase, are diminished in skeletal muscle of CKD rats. These abnormalities may account for the resistance of IGF-1 to protein turnover in skeletal muscles in CKD [22].IGFs are transported in plasma bound to IGF binding proteins (IGFBPs), which are responsible for extending their half-life and controlling their bioavailability. There are six different IGFBPs, based on their respective amino acid sequence. Of the IGFs in the serum, 75% circulate as a ternary complex bound to IGFBP3 and another subunit, acid labile subunit (ALS), to form major 150 kDa complexes. This acts as a reservoir and cannot cross the endothelial barrier. Some 20–25% of IGFs are present as smaller binary complexes with other IGFBPs forming the minor complex (35 kDa) [23]. Approximately 97% of IGF is bound; 1% of IGF-1 is in a free bioactive form. In healthy children there is a 25% excess of IGFBPs over IGF, while, in pre-terminal renal failure, IGFBPs are 150% in excess of IGFs, and in end-stage renal disease they are 200% in excess, thus reducing the bioavailability of IGF [21].IGFBP3 is the most abundant of all the binding proteins. IGFBP3 and IGFBP5 are similar in many aspects; structurally, they are both closely related, both potentiate IGF action and are up-regulated by GH. The serum level of intact IGFBP3 is normal in CKD, while the fragmented fraction (29 kDa) is increased but has a reduced affinity for peptides [24]. IGFBP5 level is not altered, but it exists mostly in the fragmented form.Intact IGFBP1, IGFBP2, IGFBP4 and IGFBP6 are elevated in CKD in relation to the degree of renal dysfunction [25]. IGFBP1, IGFBP2 and IGFBP6 are growth inhibitors of IGF-dependant proliferation of chondrocytes. Increased levels of IGFBP1 and IGFBP2 have been shown to correlate negatively with height [26]. Ulinski et al. analyzed serum levels of IGFBP4 and IGFBP5 in 89 CKD patients. There was a fourfold increase in IGFBP4, while IGFBP5 was not significantly increased. IGFBP4 level was inversely correlated with glomerular filtration rate and standard height. IGFBP5 level was positively correlated with standard height [25]. Ghrelin Ghrelin is a newly discovered, 28-amino acid peptide, with growth hormone releasing properties [growth hormone secretagogue (GHS)], expressed most abundantly in the stomach and also in smaller quantities by pancreatic islet cells and hypothalamus. Ghrelin levels are increased by fasting, cholinergic stimulation, estrogen and recombinant hIGF-1, and are decreased by food intake [27]. Ghrelin has been shown to increase the release of GH which is amplified by co-administration of GHRH, and acts at the level of the hypothalamus and pituitary. The role of ghrelin in the growth abnormalities in CKD has yet to be defined. Treatment with recombinant human growth hormone Studies in the past have shown that therapy with recombinant human growth hormone (rhGH) in CKD is safe and efficacious, and that it increases growth rate and improves standardized height [28, 29]. Haffner et al. studied the effect of growth hormone treatment on final adult height of children with CKD (Fig. 3) [1]. In contrast to the controls that had persistent growth failure, children treated with rhGH demonstrated sustained catch-up growth. In treated children the standardized height increased from baseline to a mean final adult height of 1.6 ± 1.2 SD below normal, while, in untreated control children, it decreased from baseline to a mean final adult height of 2.1 ± 1.2 SD below normal. Of the children treated with rhGH, 65% reached a final adult height within the normal range, although it was significantly below the genetic target [1]. Fig. 3Growth hormone treatment in chronic kidney disease: change from initially predicted adult height at baseline in 38 children (32 boys and six girls) with chronic renal failure who received growth hormone treatment compared with 50 control children with chronic renal failure who did not receive growth hormone, according to gender. Values are means ± SD. Asterisks indicate significant differences from the previous period (P < 0.001) and daggers indicate significant differences from the children who were not treated with growth hormone (P < 0.001). Reprinted from [1] with permission Poor growth outcomes after renal transplantation are associated with corticosteroid use, persistent CKD and abnormalities in the GH-IGF-1 axis. The use of rhGH after transplantation leads to catch-up growth associated with increase in IGF-1 levels [26]. Using the NAPRTCS database of 513 renal transplant recipients receiving rhGH, Fine and Stablein demonstrated that final adult height was superior in rhGH treated patients compared to controls. Allograft function and graft failure rates were similar in both groups, with no increase in adverse events [30]. RhGH treatment in renal failure is associated with an elevation in serum free IGF-1 [31]. RhGH increases IGFBP3, IGFBP4, and IGFBP5, decreases IGFBP1 and has no effect on IGFBP2 and IGFBP6. RhGH also increases IGFBP4, but does not correlate with change in height SDS, since most of the IGFBP4 in serum is fragmented [25, 31]. RhGH-mediated increase in IGFBP3, IGFBP5 and ALS results in increased ternary complexes and correlates positively with the increment in height SDS [31]. Among the growth-promoting effects of treatment with rhGH in CKD are anabolic effects, demonstrated by increase in body weight and mid-arm muscle circumference. The recommended dose approved for treatment of growth failure in CKD is 0.35 mg/kg per week (28 IU/m2 per week) [3]. Targets for future therapy: Treatment with recombinant IGF-1 The GH resistance associated with CKD may be amenable to treatment with recombinant human IGF-1 (rhIGF-1). In children with GH-receptor deficiency or GH-inactivating antibodies, rhIGF-1 treatment resulted in a modest increase in growth velocity and height SDS, although less than that expected with rhGH [32]. In addition, short-term administration of rhIGF-1 has been shown to increase glomerular filtration rate and renal plasma flow in patients with end-stage renal disease and in healthy subjects. Vijayan et al. demonstrated sustained improvement in renal function in a 31-day randomized, double-blinded, placebo-controlled trial of 15 patients with advanced CKD, who had received intermittent doses of rhIGF-1 [33]. One reason for the use of IGF-1 treatment in CKD is that while patients are GH sufficient, they are GH resistant. Therefore, rhIGF-1 may be more beneficial than GH as therapy in CKD and merits further investigation as an agent for treatment of growth failure and improvement in renal function in CKD [34]. Combined use of rhGH and rhIGF-1 Animal studies have shown that the combined use of rhGH and rhIGF-1 has an independent and additive effect. While whole-body growth in uremic rats was similar with treatment with either agent, an additive effect on longitudinal growth and anabolism was demonstrated when both agents were administered together [35]. Combined use of rhIGF-1 and rhIGFBP3 One of the concerns with the use of rhIGF-1 in children with normal GH production is that rhIGF-1 may suppress endogenous GH, IGFBP3 and ALS production, which may have a detrimental effect on growth. The other adverse effect of the use of rhIGF-1 observed in GH receptor-deficient patients is hypoglycemia, due to low levels of IGFBP3 and increased level of free IGF-1 available for binding to insulin receptors [32]. Both these effects can be reduced by the combined use of rhIGF-1 and rhIGFBP3 [36]. IGFBP displacers At least two different IGF-1 analogs have been made, which have an affinity for IGFBPs but have no effect on IGF receptors, thus “displacing” IGF-1 from IGFBPs and elevating free IGF-1 levels. Animal studies in hypophysectomized rats and dwarf dw/dw rats treated with IGFBP displacers demonstrated increased kidney size, improved renal function, and the stimulation of weight gain and bone growth. This effect was enhanced when treatment with IGFBP displacers was combined with rhIGF-1 [37]. Conclusion This review of the literature lends support to the concept that CKD is associated with GH “resistance”. Despite adequate treatment with rhGH and improvement in catch-up growth, children with CKD display a final adult height that is often below the genetic target. The potential for newer therapies with rhIGF-1, combined use of rhGH and rhIGF-1, combined use of rhIGF-1 and rhIGFBP3 or IGFBP displacers to improve both the short and long term outcomes in the treatment of the disturbances in the GH/IGF-1 axis in CKD awaits future investigations.

Eur_Spine_J-2-2-1602184

Penetrating spinal injury with wooden fragments causing cauda equina syndrome: case report and literature review

Study design: Case report Objective: To report an unusual case of cauda equina syndrome following penetrating injury to the lumbar spine by wooden fragments and to stress the importance of early magnetic resonance imaging (MRI) in similar cases. Summary of background data: A 22-year-old girl accidentally landed on wooden bannister and sustained a laceration to her back. She complained of back pain but had fully intact neurological function. The laceration in her back was explored and four large wooden pieces were removed. However 72 h later, she developed cauda equina syndrome. MRI demonstrated the presence of a foreign body between second and third lumbar spinal levels following which she underwent emergency decompressive laminectomy and the removal of the multiple wooden fragments that had penetrated the dura. Results: Post-operatively motor function in her lower limbs returned to normal but she continued to require a catheter for incontinence. At review 6 months later, she was mobilising independently but the incontinence remained unchanged. Conclusion: There are no reported cases in the literature of wooden fragments penetrating the dura from the back with or without the progression to cauda equina syndrome. The need for a high degree of suspicion and an early MRI scan to localise any embedded wooden fragments that may be separate from the site of laceration is emphasized even if initial neurology is intact. Introduction Penetrating injuries to the spine, although less common than blunt trauma from motor vehicle accidents, are important causes of injury to the spinal cord [1, 2, 8, 12, 13, 18]. They are essentially of two varieties—gunshot or stab wounds. Gunshot injuries to the spine are more commonly described and are associated with a higher incidence of neurological damage. On the contrary, the prognosis is better in stab wounds where surgery plays a greater role [9, 12, 14, 16, 18]. Very few case reports have been published on the onset of cauda equina syndrome (CES) following stab wounds. Injuries with sharp knifelike objects and rarely glass have been known to cause stab wounds to the spine [1, 10, 13, 14, 17]. However, there are no reports in the literature of penetrating injury by wood to the cord or cauda equina. We report a unique case of CES that developed in a patient almost 72 h following a penetrating injury to her back by a large wooden fragment. Case history A 22-year-old hairdresser accidentally slipped from the top of the stairs on to the wooden bannister which broke and as she landed at the bottom of the stairs, a piece of the wood caused a laceration to her back at the upper lumbar paraspinal region. She was taken to the nearby hospital where an open laceration to the area was found. She was complaining of severe back pain but had intact neurology in her legs. The wound was explored and debrided and four large pieces of wood were removed following which she was commenced on antibiotics. Her back pain persisted and 72 h later, she started complaining of incontinence with bilateral paraesthesia and pain down her legs. On examination, she had mildly reduced power (4/5) distally in both legs with reduced sensation in her perineum. She was transferred to the Neurosurgical unit and a MRI scan of the thoracolumbar region (Fig. 1) was taken. This revealed an area of low signal intensity between the soft tissues extending towards the spinal canal into the conus at the level of the second and third lumbar spines (L 2 and L 3) with a high signal in the cord above. No obvious spinal haematoma was present. She underwent an emergency decompressive laminectomy of L 2 and L 3 with removal of several fragments of wood. At surgery, the spinous process and right hemilamina of L 2 were noted to be fractured and multiple wooden fragments were found to have entered the spinal canal (Fig. 2) penetrating the dura in the midline causing contusion and division of some of the nerve roots of the conus. These wooden pieces were carefully removed using the microscope, the dura repaired and a closed drainage system left in. Swabs for microbiological examination were obtained from the surgical site and post-operatively prophylactic intravenous cefuroxime 750 mg, thrice a day for the next 7 days were commenced. She was advised strict bed rest in supine position for 5 days while the drain was left in situ. Over this period of time, the pain and weakness in her legs resolved completely but she continued to have numbness in the perineal region and continued to require the urinary catheter. A repeat post-operative MRI scan demonstrated the high signal in the cord as before with no residual foreign body in the canal. At discharge the wound appeared satisfactory and she was mobilising independently. When reviewed 6 months later, she was otherwise asymptomatic apart from persistent altered sensation in her perineal area. Fig. 1T 2 weighted Sagital MRI of the lumbar spine (left) demonstrates the wooden piece as an area of low signal intensity (arrow) within the soft tissues, extending into the spinal canal with an area of high signal in the cord above. T 2 weighted axial MRI (right) shows the wooden piece to have penetrated the theca on the left sideFig. 2Per-operative view (left) demonstrates the wooden piece within the spinal canal (arrows) with size of the multiple wooden fragments demonstrated (right) Discussion The most widely reported cause of penetrating injury to spine is gunshot injury [1, 2, 8, 12, 13, 15, 18] with relatively few case reports describing injuries by glass or knifelike objects [1, 10, 13, 17]. Wooden foreign bodies have been reported to penetrate the cranium, orbit, face and limbs [3, 6]. However to our knowledge, there are no reports in the literature of PI to the spine by wooden fragments with or without development of CES. The only remote resemblance to our case is that by Lunawat et al. [9] who reported the presence of tiny pieces of wood in the spinal canal of an 18-year-old man who presented insidiously with weakness in his legs and who had suffered a PI to his abdomen 6 years ago. Our case is unique due to several reasons. It is the first report of a penetrating injury by wooden fragments into the lumar spinal canal producing CES. The mechanism of injury was the direct force of the sharp wooden fragment penetrating the upper lumbar paraspinal region. We believe that after piercing the skin and subcutaneous tissue, the piece of wood probably fragmented. While some of the fragments remained superficially in the deep subcutaneous tissue overlying the thoracolumbar fascia, others were pushed towards the spinous process and right hemilamina of L 2 which were fractured and subsequently the wooden fragments pierced the dura contusing the nerve roots of the conus. What is interesting is that she developed CES almost 72 h following the injury. Whether the fragments actually moved into the canal later as a result of her movement in bed following the initial wound debridement or whether she developed oedema at the site of nerve root injury is unknown. We found MRI scan very useful to demonstrate and localise the foreign body and also to exclude any intra- or extradural haematoma or contusion in the cord or cauda equina. In our case some contusion was demonstrated in the cord (Fig. 1). Previous reports of penetrated wooden pieces into the neck or extraspinal areas of the body have commented on the risk of misinterpretation of CT appearance of wood which appears as an area of high attenuation [3, 6]. Wood is highly echogenic and shows clear acoustic shadowing on sonography [10]. MRI is a useful adjunct to both CT and sonography for the detection of non-metallic foreign bodies such as wood [15]. Surgery in PI is indicated for progressive neurological deficits, persistent cerebrospinal fluid (CSF) leak or for incomplete neurologic deficits with radiological evidence of compression [5]. However, surgery at the cauda equina, in such cases, may not be easy due to the dural tear associated with impingement of the foreign body with the contused nerve roots. In our patient urgent surgery was necessary as she developed progressive neurological deficit leading to CES with radiological evidence of a large foreign body causing thecal compression. Post-operatively, we took the extra precaution to prevent CSF leak from the wound by advising the patient a strict bed rest for 5 days. In our institute we advise 3–5 days of strict bed rest for patients with unintentional dural tears during Lumbar spinal surgery. This has been the practise by others too [19]. We do not routinely prescribe antibiotics for such cases. However, given the nature of this particular case, we decided to use prophylactic antibiotics in spite of negative bacteriological results. Incidentally it has been claimed that extraspinal sepsis is much more common than spinal (CSF) infection following PI to the spine [5]. We did not use methylprednisolone as there is no convincing evidence to suggest any advantage, and moreover there have been reports of myopathy following the administration of methylprednisolone in patients with acute spinal injury [7, 11]. Finally, neurological recovery from injury to the cauda equina is unpredictable and may be influenced by several factors [4] such as patient age, energy transfer to the neurovascular structures and timing of neural decompression, although the latter is debated [12, 14, 15]. Late motor recovery has been reported to occur after PI to the cauda equina up to 16 years after injury in one study [8]. Our patient had excellent symptomatic relief of her back and bilateral leg pain following surgery with resolution of the motor weakness in her legs. However, the perineal numbness has persisted and we would have to wait and see if this improves in future. Conclusion Penetrating wound to the upper lumbar spine is common with gunshot injuries but has never been described with wooden fragments. MRI scan provides excellent visualisation of the foreign body and also helps to exclude any underlying spinal haematoma or contusion. For all similar cases where a laceration containing wood is found, the authors recommend a high index of suspicion and a MRI scan must be performed early to look for separate fragments of wood deeper down. Neural decompression and removal of the foreign bodies should be performed to prevent neurological deterioration, infection and possible CSF leak.

Eur_J_Nucl_Med_Mol_Imaging-3-1-1914225

Cardiovascular molecular imaging of apoptosis

Introduction Molecular imaging strives to visualise processes at the molecular and cellular level in vivo. Understanding these processes supports diagnosis and evaluation of therapeutic efficacy on an individual basis and thereby makes personalised medicine possible. Apoptosis The term ‘apoptosis’ is derived from the Greek (apo = from and ptosis = falling, commonly pronounced ap-a-tow’-sis) and denotes a regulated process of cell suicide resulting in a cell corpse that is distinct from the necrotic cell [1, 2]. Whereas necrosis is manifested by cellular swelling, plasma membrane rupture and rapid release of intracellular constituents into the environment, causing an inflammatory response [2], apoptosis is characterised by cell shrinkage, long-lasting maintenance of plasma membrane integrity and lack of inflammatory responses in the vicinity of the dying cell [2]. It has long been thought that apoptosis and necrosis are the only two modes of cell death counterbalancing mitosis [3]. However, recent insights have taught us that cells can follow more roads to death than only apoptosis and necrosis. In 2001, Leist and Jaättelä suggested a model in which various forms of cell death could be positioned on a gliding scale between two extremes, apoptosis and necrosis [4]. The mode of execution depends on the cell type and the cell death-inducing trigger. Most knowledge about various modes of cell death has been generated in the field of oncology, where cell death plays an important role in the progression and treatment of cancer [5]. Several intermediate forms of cell death have been identified and characterised [6, 7], including mitotic catastrophe and autophagic cell death. Mitotic catastrophe can be the result of unsuccessful chromosome segregation, an event that requires malfunctioning of multiple cell cycle checkpoints. Catastrophic cells are large and non-viable cells with a compromised nuclear blueprint leading eventually to cell death [8]. Mitotic catastrophe appears to be a rare event in cardiovascular diseases. Autophagic cell death is characterised by the total destruction of the cell through autophagy. It is a process in which cytoplasmic constituents are degraded through the lysosomal machinery [9]. Autophagic cell death occurs both in oncology [10, 11] and in cardiovascular diseases [12]. The various described modes of cell death differ in morphology and biochemistry [13]. The caspase cascade is the most remarkable biochemical distinction [14]. It is activated during apoptosis but hardly, or not at all, during other modes of cell death. Biochemistry of apoptotic cell death Two major biochemical routes dominate apoptotic cell death. They involve the activation of the caspase cascade through either death-receptor mediated signal transduction or stress-induced release of cytochrome C from mitochondria [15]. Caspases involved in apoptosis can be grouped into initiator (caspases 8 and 9) and effector caspases (caspases 3 and 7). The principal event of the caspase cascade is the activation of an inactive procaspase into a proteolytically active caspase. Activation of initiator caspases occurs through auto-activation, catalysed by the interaction with ligated death receptors (procaspase 8) or cytochrome C-containing apoptosome (procaspase 9). Effector caspases are activated through intra-chain cleavage by active initiator caspases [16]. In both cases, activation results in the exposure of active sites that selectively cleave peptide bonds after an aspartate residue in caspase substrates. Several checks and balances exist that control burst of the caspase cascade if triggered by minor undesired initiating events. Inhibitors of apoptotic proteins (IAPs) suppress proteolytic activity by binding to activated caspases. Inhibition of IAPs is in turn required to allow execution of apoptosis. Smac/DIABLO residing in the mitochondria and co-released with cytochrome C inhibits IAPs and permits propagation and amplification of the proteolytic signal through the caspase cascade. Anti-apoptotic members of the Bcl-2 family such as Bcl-2 and Bcl-X prevent the release of cytochrome C and Smac/DIABLO from mitochondria. Pro-apoptotic members such as Bax and tBid, which is the result of Bid cleavage by activated caspase 8, neutralise the protective effects of the anti-apoptotic members and provoke mitochondria to release their pro-apoptotic cargo. The intricate scheme of the apoptotic machinery offers several potential targets for molecular imaging. For example, knowledge about the mechanism of caspase activation has triggered the design of several strategies to measure apoptosis [17–19]. The hurdle these methods have to take is entrance of the reporter of caspase activity into the cell. Consequently, in vivo applications of such strategies face unfavourable biodistribution profiles and high background levels. Targets embedded in the plasma membrane encompass a more promising set of molecules for molecular imaging. Phosphatidylserine An essential part of the apoptotic program consists of the appearance of ‘eat me’ flags at the cell surface. Phagocytes recognise these flags and respond by engulfing the dying cell before it leaks pro-inflammatory components into the surrounding tissue. The ‘eat me’ flags are, alone or in combination, specific for the dying cell, allowing phagocytes to make the right choice in an environment filled with living cells. In 1992, Valerie Fadok and co-workers reported that phosphatidylserine (PS) becomes exposed on the surface of apoptotic lymphocytes, where it functions as an ‘eat me’ signal for phagocytes [20]. PS, a negatively charged aminophospholipid, is predominantly found in the cytofacial membrane leaflets of living cells. The exofacial plasma membrane leaflet contains predominantly phosphatidylcholine and sphingomyelin while PS is almost completely lacking. The PS asymmetry results from the ATP-dependent action of the aminophospholipid translocase that transports PS from the outer to the inner leaflet [21, 22]. During apoptosis the aminophospholipid translocase is inhibited and concomitantly a scramblase is activated. The combined action results in the surface expression of PS whilst the plasma membrane integrity remains intact. Surface-expressed PS provides an attractive target for molecular imaging of apoptosis. Annexin A5 In the past decade, a molecular imaging protocol was established to detect and measure apoptosis in vitro and in vivo in animal models and patients [23–26]. This molecular imaging protocol is based upon the protein annexin A5 (Anx A5), which binds with high selectivity and with great affinity (Ka = 7 nM) [27, 28] to the PS abnormally expressed on the cell membrane of apoptotic cells [21, 22, 29]. Anx A5 is a member of a multiprotein family of more than 160 proteins that share the property of Ca2+-dependent binding to negatively charged phospholipid surfaces [30]. The biophysical, biochemical and biological properties of Anx A5 and other members of the annexin family have been extensively described in a number of excellent reviews [30–34]. Detecting apoptosis with Anx A5 The development of annexin-based detection of apoptosis started in 1992 when Fadok et al. revealed that PS is expressed on the cell surface of apoptotic cells [35]. This revelation led Koopman et al. to the design of an apoptosis detection assay on the basis of fluorescence (fluorescein isothiocyanate)-labelled Anx A5 [29]. This Anx A5 affinity assay was further developed by labelling Anx A5 with biotin or with several radionuclides to facilitate various protocols for measuring apoptosis both in vitro [36] and in vivo [23–25, 37–41] in animal models. The availability of 99mTc-labelled Anx A5 produced under GMP regulations led to the first studies of non-invasive detection of apoptosis in patients [26, 42–46]. Detection of apoptosis using alternative methods In addition to Anx A5, several other proteins may have a high specificity and tight binding to PS. If labelled with a fluorescent or nuclear probe, these proteins could possibly be used for in vivo detection of apoptotic cells in the field of cardiovascular disease and in other disciplines. In addition, other steps in the apoptotic signalling cascade, such as the activation of caspase-3, can be utilised to visualise activation of the apoptotic machinery (Fig. 1). Fig. 1Targets for apoptosis detection. During apoptosis, initiator caspases are activated, via either cell death receptor-mediated or mitochondrial signalling. These initiator caspases, in turn, trigger the activation of effector caspases, such as caspase-3. The activation of caspase-3 results in the typical characteristics of apoptosis, such as DNA fragmentation, substrate cleavage of cytoplasmic proteins and cell membrane alterations. These apoptotic characteristics and the activation of caspase-3 offer targets for molecular imaging of apoptosis Synaptotagmin I One such PS binding protein is the C2 domain of synaptotagmin I, which binds to anionic phospholipids in cell membranes [47]. By conjugation of the C2 domain, synaptotagmin can be used for in vivo molecular imaging of apoptosis. Zhao and co-workers conjugated the C2 domain with superparamagnetic iron oxide (SPIO) particles, a very effective T2 relaxing MRI contrast agent, and injected the product in tumour-bearing mice. They revealed that a murine lymphoma (EL4) tumour model treated with cyclophosphamide and etoposide showed an increase in apoptotic cells from a basal value of 4% to 32% during chemotherapy. This was readily detected with C2-labelled SPIO particles and T2W MRI. Tumour regions with the greatest MRI changes correlated with regions having the highest proportion of apoptotic and necrotic cells [47]. More recently, the same group labelled the C2A domain of synaptotagmin I radioactively (99mTc) or fluorescently (FITC) and used the labelled synaptotagmin in a reperfused acute myocardial infarction model (AMI) rat model. They found both ex vivo and in vivo accumulation of the radiotracer in the area at risk. Although some of the uptake was caused by passive leakage due to elevated vascular permeability in the area at risk, the majority was caused by specific binding to PS [48]. Both Anx A5 and synaptotagmin I detect apoptotic cells by selective binding to externalised PS. An alternative method to detect apoptosis may be to use a target more upstream in the apoptotic cascade, the effector caspases. 5-Pyrrolidinylsulphonyl isatins In one study the peptide-based irreversible pan-caspase inhibitor Z-VAD-fmk was radio-iodinated and suggested as a caspase imaging agent. However, due to poor cellular permeability, intracellular targeting of activated caspases was limited, preventing in vivo application [49]. In more recent work, non-peptidyl caspase inhibitors of the 5-pyrrolidinylsulphonyl isatin-type were proposed instead of a peptide-based caspase, such as Z-VAD-fmk. These caspase binding radioligands (CBRs) may be capable of directly targeting apoptosis in vivo. CBRs are expected to form intracellular enzyme inhibitor complexes by means of binding to the activated caspases. Kopka and co-workers succeeded in labelling several 5-pyrrolidinylsulphonyl isatins with iodine-125 [50]. However, none of these compounds have been used for non-invasive imaging in patients yet. ApoSense The ApoSense family is a group of low molecular weight amphipathic apoptosis markers targeting the cell membrane of apoptotic cells. ApoSense was developed by the Israel-based company NST. ApoSense has been used in vitro and in vivo in several disease models associated with cell death, such as radiation-induced lymphoma, renal ischaemia/reperfusion and cerebral stroke [51, 52]. However, it is still not known to which cell membrane target ApoSense exactly binds. Therefore, it remains a challenge to link ApoSense binding to the signalling cascades in apoptosis signalling. Role and detection of apoptosis in CVD Apoptosis of cardiomyocytes after myocardial infarction Several forms of cell death have been observed in the infarcted myocardium. The central ischaemic zone is dominated by necrotic cell death, whereas in the periphery of the area at risk mainly apoptotic cell death is observed. Apoptotic cell death in the infarct periphery is not the exclusive domain of the cardiomyocyte. Also non-myocytes, including endothelial cells, macrophages and blood cells, contribute to a comparable degree to apoptotic cell death [53]. Since the process of cell death can be pharmaceutically manipulated, thereby limiting the extent of myocardial cell loss, detection and quantification of apoptotic cells are of importance for determination of the full extent of reversible myocardial damage. Once the apoptotic program is activated, several targets are present both for imaging apoptosis and for therapeutic interventions. As described earlier, the C2A domain of synaptotagmin I binds to the exposed PS on apoptotic cells. Zhao and co-workers developed a radiotracer for PS expression by labelling synaptotagmin via 2-iminothiolane thiolation with 99mTc, forming a technetium C2A glutathione S-transferase complex (99mTc-C2A-GST) [48]. In vivo planar imaging of AMI in rats was performed on a gamma camera using a parallel-hole collimator. The uptake of 99mTc-C2A-GST within the area at risk was quantified by direct gamma counting. The foremost finding was that in seven of seven rats the infarct was clearly identifiable as focal uptake in planar images. This and other findings led the authors to conclude that the C2A domain of synaptotagmin I labelled with a radioisotope binds to both apoptotic and necrotic cells. Ex vivo and in vivo data indicate that, because of elevated vascular permeability, both specific binding and passive leakage contribute to the accumulation of the radiotracer in the area at risk. However, the latter component alone is insufficient to achieve detectable target-to-background ratios using in vivo planar imaging. Annexin A5 has been more widely used to target PS exposure. Hiller and co-workers [54] recently reported on a new Anx A5-based T1 contrast agent. They linked Anx A5 to gadolinium diethylenetriamine penta-acetate (Gd-DTPA)-coated liposomes. In their model of perfused isolated rat heart, ligation of the left coronary artery for 30 min was followed by reperfusion. T1 and T2* images were acquired using an 11.7-T magnet before and after intracoronary injection of the contrast agent. A significant increase in signal intensity, visible in those regions containing cardiomyocytes in the early stage of apoptosis, was found. The group of Weissleder also reported on Anx A5-based MRI contrast agents [55], demonstrating the construction of a magneto-optical nanoparticle AnxCLIO-Cy5.5, which was tested in a murine model of transient coronary artery (LAD) occlusion. The synthesis of the probe was achieved in three steps. First, the amino-CLIO nanoparticle was labelled with Cy5.5 and activated with SPDP [N-succinimidyl 3-(2-pyridyldithio)propionate], to yield a compound termed 2PySS-CLIO-Cy5.5. Second, Anx A5 was reacted with SATA (N-succinimidyl S-acetylthioacetate). In the final step of the process, SATAylated Anx A5 was linked with 2PySS-CLIO-Cy5.5 to yield the multimodal AnxCLIO-Cy5.5 nanoparticle. It was demonstrated that it is feasible to obtain high-resolution MR images of cardiomyocyte apoptosis in vivo with the use of the nanoparticle AnxCLIO-Cy5.5. In early work preformed by our group a fluorescent invasive Anx A5-based imaging protocol for myocardial apoptosis was developed in mice [25]. The promising findings of this study encouraged use of modified human recombinant Anx A5 to construct (Anx A5)-n-1-imono-4-mercaptobutyl, labelled with pertechnetate for clinical studies. One milligram of 99mTc-Anx A5 (584 MBq) was injected intravenously 2 h after reperfusion in seven AMI patients. In six of the seven patients, increased uptake of 99mTc-Anx A5 was seen in the infarcted region of the heart on single-photon emission computed tomography (SPECT) images. No uptake was seen in the heart outside the infarcted area. In all individuals with increased uptake, a matching perfusion defect (on MIBI SPECT) was observed (Figs. 2, 3). These results demonstrated the feasibility of non-invasive monitoring of apoptosis in patients. These techniques may allow measurement not only of the extent of reversible damage but also of the efficacy of novel therapies targeting apoptosis. Fig. 2Transverse tomographic images in a patient with acute anteroseptal infarction. a Arrow shows increased 99mTc-labelled Anx A5 uptake in the anteroseptal region 22 h after reperfusion. b Perfusion scintigraphy with sestamibi 6–8 weeks after discharge shows an irreversible perfusion defect which coincides with the area of increased 99mTc-labelled Anx A5 uptake (arrow). L liver. Reprinted with permission from Elsevier (The Lancet, 2000, 356, 211)Fig. 3Transverse tomographic images in a patient with acute anterior wall myocardial infarction. a Arrow shows increased uptake of 99mTc-labelled Anx A5 in the infarct area 17.5 h after reperfusion. L liver. b Perfusion defect on sestamibi perfusion scintigraphy 6–8 weeks after discharge matches the uptake of Anx A5 (arrow). Reprinted with permission from Elsevier (The Lancet, 2000, 356, 211) Atherosclerotic lesions and plaque instability Extensive histopathological studies and investigations on the initiation of atherosclerosis in transgenic mice have generated a quite detailed picture of the development of atherosclerosis in the various stages of disease [56–63]. The onset of plaque formation is characterised by circulating levels of oxidised low-density lipoproteins, which are capable of penetrating the endothelial border while inflicting oxidative damage on the endothelial cells in the process. Subsequently, the reactionary expression of inflammatory markers on the endothelial cell surface attracts circulating monocytes to the site of injury, which, upon arrival at the scene, differentiate into macrophages and start ingesting the available lipid and other oxidised particles. Uninhibited intracellular lipid accumulation compromises the function of macrophages and converts them into foam cells [64]. The accompanying secretion of pro-inflammatory cytokines and matrix-metalloproteinases (MMP) likely contributes to the hostility of the plaque environment and weakening of collagen matrix structures, both indirectly through loss of smooth muscle cells and directly through MMP activity [65]. At a certain point, excess damage to macrophages and smooth muscle cells induces apoptosis, causing remnants of lipid-laden dead cells to form small extracellular lipid droplets. A luxating moment in atherogenesis is thought to occur when the remaining population of functional phagocytes becomes unable to engulf all apoptotic remnants of dying cells in the environment [65]. Once the lesion contains a large, necrotic, lipid pool that is covered by a fibrous cap, primarily consisting of SMCs and extracellular matrix, it is named an atheroma. A critical finding for vascular risk prediction is that although plaque rupture generally requires progression of lesion size, acute events resulting from rupture may just as well originate from only mildly stenotic lesions [66]. Coronary catheterisation studies in patients admitted with acute coronary syndromes have shown that most of the culprit lesions are below the 50% stenosis threshold [67–69]. In reality, the extent of the lesion may be underestimated, since conventional coronary catheterisation does not take the extent of outward remodelling of the coronary artery at the site of atherosclerotic lesions into account. The emerging cardiac multislice CT technology may be better equipped to take outward remodelling into account and to assess the total coronary plaque burden in patients. Nevertheless, these data imply that though plaque dimensions will in general be sufficient for imaging-target visualisation, focussing on mere plaque size entails unsatisfactory prognostic information for clinical use. From a biological perspective, several inflammatory parameters are associated with vulnerable lesions, such as elevated extracellular levels of MMP and pro-inflammatory cytokines, including a variety of interleukins and chemotactic proteins, as has been shown in both experimental and clinical studies [70–73]. Moreover, high-risk lesions are notorious for displaying high levels of inflammation and apoptotic cells; in particular, large amounts of apoptotic macrophages have been found to reside in fragile and ruptured fibrous caps [59, 65, 74]. In 2000, Kolodgie et al. [59] examined 40 culprit lesions from cases of sudden coronary death and reported apoptotic macrophages to constitute up to 50% of the total macrophage population, preferentially localising at the site of rupture. Also, once atheromas have formed, plaques have been found to be far more sensitive to SMC death than regular arterial walls, as recently demonstrated by the group of Bennett [61]. Since all these features share their relation to inflammation and more specifically apoptosis, detection of apoptosis in vivo seems a sensible approach to vulnerable plaque identification. This approach builds on the pivotal role that apoptosis plays in promoting plaque progression and disruption; however, the underlying mechanism is a complex interplay in which apoptosis affects endothelial cells, smooth muscle cells and macrophages in a different manner and with different consequences. Given the firmly established direct causal relationship between apoptosis and lesion instability [59, 75], in vivo visualisation of the apoptotic plaque status could represent a significant leap forward in the field of therapeutic management for individual patients (Fig. 4). Fig. 4Feasibility of non-invasive imaging of apoptosis by radiolabelled Anx A5. Left lateral oblique gamma images of experimental atherosclerotic (a–c) and control (d–f) rabbits injected with 99mTc-Anx A5; L and K indicate liver and kidney activities, respectively. Images at the time of injection (a, d) and at 2 h after injection (b, e) are shown. Although blood pool activity is seen at the time of injection (a) in the atherosclerotic animal, tracer uptake is clearly visible in the abdominal aorta (with lesions) at 2 h (b). c Ex vivo image of b shows intense 99mTc-Anx A5 uptake in the arch and abdominal region. Annexin-positive areas were confirmed to contain macrophage- and apoptosis-rich regions in the atherosclerotic plaque by histology. d–f show the corresponding images in the control animal. Note that the aorta is indistinguishable from background at 2 h after injection (e). The blood pool at the time of injection in the control animal (d) is comparable to that in the atherosclerotic animal. f Ex vivo aortic image of the control animal demonstrates the absence of 99mTc-labelled Anx A5 uptake. From Kolodgie et al. [75] The expanding knowledge about plaque biology has led to the identification of a variety of potentially suitable molecular markers for vulnerable plaque imaging [76]. Studies have been reported on the targeting of several inflammation-related molecular targets, including vascular adhesion molecules (ICAM1, VCAM1, E-selectin), neoangiogenesis molecules (αvβ3 integrin), scavenger receptors on macrophages, MMP proteolytic activity, macrophage metabolic activity and apoptosis (PS) [60, 62, 75–83]. Although some of these approaches have yielded promising experimental data, there is still a paucity of clinical studies confirming the feasibility of these targeted approaches in humans. Of the clinical results available, the most impressive have been achieved using 99mTc-Anx A5 SPECT imaging and [18F]fluorodeoxyglucose (18FDG) positron emission tomography (PET) imaging. The latter has been used for many years to detect metabolically hyperactive processes in nuclear oncology. Since the glucose analogue 18FDG is a marker for foci of increased metabolic activity [84] and it is known that macrophages involved in atherosclerotic plaques have a high metabolic rate, there is sound logic behind this strategy. In addition, 18FDG imaging capitalises on the merits of oncological medicine since 18FDG is a clinically approved tracer that can be detected with high sensitivity and accuracy by PET technology, whereas PET conjugates for most alternative plaque imaging targets are still in the process of development. 18FDG imaging of arterial inflammation has been performed with rather promising results in retrospective studies of oncological 18FDG scans, in experimental settings and in in vivo clinical studies by Rudd and colleagues [76, 82, 85, 86]. In an eight-patient pilot study, they were able to correlate 18FDG uptake in carotid lesions of patients scheduled for endarterectomy with histologically assessed macrophage density (p < 0.005). However, a practical difficulty of performing 18FDG imaging in the search for coronary lesion instability lies in the physical activity of the cardiac muscle, which results in high glucose requirements and thus severe background signal. In addition to detection of the vulnerable plaque with 18FDG, imaging of the vulnerable plaque with 99mTc-Anx A5 has yielded some exciting results as well. The first demonstration of the feasibility of in vivo imaging of plaque instability with 99mTc-Anx A5, in animals, was given in 2003 by Kolodgie et al. It was demonstrated in a rabbit model of aortic de-endothelialisation followed by a high-fat diet that 99mTc-Anx A5 showed a ninefold higher binding to the atherosclerotic vessels than to the control vessels [75]. Further validation of the Anx A5 imaging concept in the vulnerable plaque was done by Johnson and colleagues in 2005 [87]. In a porcine model of atherosclerosis the feasibility of SPECT imaging with 99mTc-Anx A5 to assess atheroma progression in coronary arteries was demonstrated. This was ultimately challenging owing to the low mass of the imaging target and cardiopulmonary motion. As swine and human hearts have similar cardiac dimensions, these results indicate that non-invasive functional imaging of coronary plaque instability could be realistic in the near future. The first clinical demonstration of the feasibility of nuclear vulnerable plaque imaging with 99mTc-Anx A5 was given by Kietselaer et al. in 2004 [43]. In a pilot study, four patients were included who were scheduled for carotid endarterectomy and had a history of recent (n = 2) or remote (n = 2) transient ischaemic attack (TIA). Before surgery 99mTc-Anx A5 was administered and SPECT imaging was performed. In the two patients who had suffered a TIA more than 3 months before imaging and had since been treated with statins and antiplatelet agents, no tracer uptake of Anx A5 was observed (Fig. 5). Histological analysis confirmed stable lesions with negligible macrophage infiltration and intra-plaque haemorrhage. In contrast, the two patients who had experienced transient ischaemia shortly before imaging (3 and 4 days, respectively) demonstrated marked uptake of Anx A5 in the culprit carotid vessel, while the contralateral arteries remained clear of tracer binding. Anx A5 binding was histologically localised to apoptotic macrophage membranes. An interesting finding during this study was that echo-Doppler examination in one of the patients with a recent TIA identified a severe stenotic lesion in the contralateral carotid artery prior to imaging. On the 99mTc-Anx A5 scan, however, the lesion was negative, though it might have been suspected as the culprit lesion by conventional techniques. This finding underlines the potential added value of functional versus anatomical imaging in predicting the likelihood of future thrombotic events. Fig. 5Images of unstable atherosclerotic carotid artery lesions obtained with radiolabelled Anx A5. a Transverse and coronal views obtained by SPECT in patient 1, who had a left-sided TIA 3 days before imaging. Although this patient had clinically significant stenosis of both carotid arteries, uptake of radiolabelled Anx A5 is evident only in the culprit lesion (arrows). b Histopathological analysis of an endarterectomy specimen from patient 1 (polyclonal rabbit anti-Anx A5 antibody, ×400) shows substantial infiltration of macrophages into the neointima, with extensive binding of Anx A5 (brown). c In contrast, SPECT images of patient 3, who had had a right-sided TIA 3 months before imaging, do not show evidence of Anx A5 uptake in the carotid artery region on either side. Doppler ultrasonography revealed a clinically significant obstructive lesion on the affected side. d Histopathological analysis of an endarterectomy specimen from patient 3 (polyclonal rabbit anti-Anx A5 antibody, ×400) shows a lesion rich in smooth muscle cells, with negligible binding of Anx A5. ANT anterior, L left. Copyright © 2004 Massachusetts Medical Society. All rights reserved. [43] Among the discussed imaging modalities for vulnerable plaque detection, nuclear imaging (PET in particular) is far superior to MRI, being capable of detecting probes at the picomolar range [88]. PET further outperforms SPECT imaging regarding spatial resolution, with 4- to 6-mm versus 10- to 15-mm resolution in the centre of the field. The major difficulty in evaluating PET or SPECT usually originates from orientation and localisation of the measured signal in the body. To overcome this, nuclear imaging modalities can be co-registered and overlaid with conventional CT or MRI imaging, hence providing anatomical landmarks in relation to molecular information. Improvements in the availability and costs of PET/CT or PET/MRI imaging devices will therefore imply great advantages. Given the high costs of imaging, diagnostic algorithms in the future most likely will combine relatively cheap serum biomarkers that indicate plaque instability with the subsequent use of expensive imaging technology in selected cases. This indicates that the specific niche for both serum and imaging biomarkers needs to be defined in large prospective studies. In addition, much is to be gained from refinements of the radioligand with regard to blood clearance and biodistribution, since molecular imaging typically requires fast binding of the probe to the target, rapid subsequent clearance from the blood and not too short radioactive half-lives. Heart failure Heart failure (HF) is growing to epidemic proportions, partially due to the better treatment of AMI. The pathogenesis of heart failure is still poorly understood. There may be an important role for cell death in the progression of congestive HF. A number of papers have reported on the differences in the amount of cell death in healthy and failing hearts. The baseline rate of apoptosis in healthy human hearts is around one to ten cardiac myocytes per 105 nuclei. On the other hand, 80–250 heart muscle cells per 105 cardiac nuclei undergo apoptosis at any given time in late-stage dilated cardiomyopathy. However, it remains unclear whether this cell death is a coincidental finding, a protective process or a causal component in the pathogenesis. In work done by Wencker and co-workers, the uncertainty of the role of cell death was addressed. In a transgenic mouse model with a conditionally active caspase expressed exclusively in the myocardium, they demonstrated that even very low levels of cardiomyocyte apoptosis (23 per 105 nuclei, compared with 1.5 per 105 nuclei in controls) caused a lethal dilated cardiomyopathy. Inhibition of this active caspase largely prevented the development of cardiac dilation and contractile dysfunction. These and other data suggest that cardiomyocyte apoptosis and necrosis may be a causal mechanism of heart failure, contributing to adverse left ventricle (LV) remodelling. Inhibition of this cell death process may constitute the basis for novel therapies. Consequently, anti-apoptotic agents have been developed and tested in animal models of HF with promising results [89]. Although, as discussed earlier, it is possible to detect cell death in MI leading to HF, the small amount of cell death present in heart failure (80–250 cardiomyocytes per 105 cardiac nuclei) makes detection with the current techniques very challenging. However, the feasibility of detecting apoptosis in patients with progressive idiopathic dilated cardiomyopathy was demonstrated in a proof of principle study done by Kietselaer [90]. Nine patients with idiopathic dilated cardiomyopathy (IDCM) were evaluated, as well as two family members with hypertrophic cardiomyopathy in different stages of disease. Patients were imaged with 99mTc-labelled Anx A5. 201Tl images were used for orientation. Other causes of IDCM were excluded using laboratory testing, echocardiography and coronary angiography. It was found that five patients with IDCM showed focal uptake of radiolabelled Anx A5 and that one patient showed global uptake in the left ventricle (Fig. 6, left panel). No enhanced uptake could be visualised in the other six patients (Fig. 6) and controls. All cases that showed Anx A5 uptake in the heart displayed recent onset or recent worsening of left ventricular function and functional class. In cases that did not show Anx A5 uptake, LV function and clinical status remained stable. Kietselaer concluded that non-invasive imaging using 99mTc-Anx A5 detects apoptosis in patients suffering from IDCM, providing further evidence for the role of apoptosis in the pathogenesis of IDCM. In addition, Anx A5 imaging identifies high-risk patients who might benefit from cell death blocking strategies. Fig. 6Dual-isotope imaging using 201Tl for left ventricular contour detection and, simultaneously, radiolabeled annexin A5 in patients with dilated cardiomyopathy. a Dilated cardiomyopathy patient with rapid deterioration of left ventricular function. Note focal uptake in apex and lateral wall, and slight septal uptake. b Dilated cardiomyopathy patient in acute heart failure. Note global uptake of radiolabeled annexin A5. c Dilated cardiomyopathy patient in stable clinical condition. Uptake is absent even when image is enhanced to the extent that background radioactivity can be observed. d Family member of patient in panel b. No clinical evidence is seen of dilated cardiomyopathy. Note absence of uptake of radiolabeled annexin A5. ANT, anterior; INF, inferior; LAT, lateral; SEPT, septal. Reprinted by permission of the Society of Nuclear Medicine from [90] In addition to imaging of apoptosis, the detection of the molecular substrates of early-stage alterations leading to HF, such as vascular remodelling in infarcted tissue, is an emerging technique. Vascular remodelling is associated with non-contractile scar tissue formation that may contribute to adverse LV remodelling. However, none of these molecular techniques has reached the clinical stage yet. Cardiac allograft rejection Until 1997, almost 46,300 heart transplantations had been carried out worldwide [48]. Despite this encouraging number, only 50% of cardiac transplant recipients will survive 10 years and almost all will die within 20 years unless re-transplantation is preformed [19]. One of the possible complications for patients who have received a donor heart is transplant rejection. Under the current medical guidelines, endomyocardial biopsies (EMBs) are recommended 15–20 times in the first year after transplantation to monitor potential allograft rejection. EMB is an invasive diagnostic procedure and not without risk of complications. Allograft rejection is pathologically characterised by perivascular and interstitial mononuclear inflammatory cell infiltration associated with myocyte necrosis and apoptosis [91, 92]. Because the process of apoptosis can be pharmaceutically manipulated, identifying apoptosis non-invasively is of critical importance [93]. In a study done by Narula and co-workers [42], 18 patients who had undergone heart transplantation within the past year were subjected to non-invasive imaging with Anx A5. Patients received an injected dose of 555–1,110 MBq (15–30 mCi) of 99mTc-Anx A5 and underwent SPECT imaging 1 h after injection (Fig. 7). In 13 out of these 18 patients, no myocardial uptake of 99mTc-Anx A5 was observed. In the remaining five patients, uptake of 99mTc-Anx A5 in the myocardial region was observed, which was associated with at least moderate transplant rejection (Fig. 8), of International Society of Heart and Lung Transplantation (ISHLT) grade 2/4. Histological analysis revealed active caspase-3 in EMB specimens. This study demonstrated the clinical feasibility of detecting cardiac allograft rejection induced apoptosis, which, after confirming these findings in larger clinical studies, might lead to a reduction for the need of EMBs. Fig. 7SPECT images of 99mTc-labelled Anx A5 in three patients (10, 3 and 4) are demonstrated. a In patient 10, myocardial uptake (as outlined by solid circles) is clearly seen in all tomographic orientations and can be differentiated from the left ventricular cavity, especially in the transaxial slice. b In patient 3, by contrast, no Anx A5 uptake is observed, and only background activity is seen. c Patient 4 demonstrates Anx A5 activity in the blood pool originating from the great vessels (GV great cardiac vein, PA pulmonary artery, Ao aorta) and ventricular contours, but no activity is observed in the myocardium. The scan of patient 10, with myocardial uptake, was further processed (Fig. 8). RV right ventricle, LV left ventricle. Adapted by permission from Macmillan Publishers Ltd: Nature Medicine [42], copyright 2001Fig. 8Diffuse myocardial uptake of 99mTc-labelled Anx A5 in cardiac allograft. Patient 10 underwent Anx A5 imaging 9 months after orthotopic heart transplantation. SPECT imaging 3 h after intravenous injection of radiolabelled Anx A5 demonstrated multifocal myocardial uptake of radiotracer. Smoothing of the images and cardiac SPECT processing in the short axis, vertical long axis and horizontal long axis revealed myocardial perfusion scan-like images with diffuse uptake in the whole myocardium, suggesting apoptosis in the myocardium. Adapted by permission from Macmillan Publishers Ltd: Nature Medicine [42], copyright 2001 Future perspectives Of all the apoptotic imaging agents, Anx A5 has made the most successful transition from the test tube to the clinical arena. Novel imaging modalities like PET/CT and SPECT/CT are making a significant contribution to the growing role of molecular imaging. The combination of anatomical imaging with CT and biological imaging using SPECT or PET yields a new synergistic imaging modality that provides detailed information on the molecular (patho)physiological processes in relation to exact anatomical orientation. The next challenge is to assess the incremental clinical value of molecular imaging and its ability to change patient management decisions. For this purpose, prospective cohort studies need to be designed to evaluate the clinical meaning of molecular imaging scan results in relation to the progression of the underlying disease. For instance, data on the association between Anx A5 uptake in plaques in the carotid arteries and clinical event rate are still lacking. The availability of such large clinical datasets may allow for better stratification of patients and therefore more optimal treatment decisions. Finally, we and others have found that not only apoptotic cells but also viable cells are detectable with Anx A5. Our work on the vulnerable plaque showed that several processes leading to plaque instability are associated with PS expression, such as activated macrophages (inflammation) and aging red blood cells (intra-plaque haemorrhage). This ability of Anx A5 to visualise exposed PS in different biological conditions opens novel opportunities for imaging.

Cardiovasc_Eng-3-1-2137944

Determination of myocardial energetic output for cardiac rhythm pacing

This research is aimed to the determination of the changes in the cardiac energetic output for three different modes of cardiac rhythm pacing. The clinical investigation of thirteen patients with the permanent dual-chamber pacemaker implantation was carried out. The patients were taken to echocardiography examination conducted by way of three pacing modes (AAI, VVI and DDD). The myocardial energetic parameters—the stroke work index (SWI) and the myocardial oxygen consumption (MVO2) are not directly measurable, however, their values can be determined using the numerical model of the human cardiovascular system. The 24-segment hemodynamical model (pulsating type) of the human cardiovascular system was used for the numerical simulation of the changes of myocardial workload for cardiac rhythm pacing. The model was fitted by well-measurable parameters for each patient. The calculated parameters were compared using the two-tailed Student’s test. The differences of SWI and MVO2 between the modes AAI and VVI and the modes DDD and VVI are statistically significant (P < 0.05). On the other hand, the hemodynamic effects for the stimulation modes DDD and AAI are almost identical, i.e. the differences are statistically insignificant (P > 0.05). Introduction The medical management of cardiovascular diseases, especially of arrhythmia, has become increasingly reliant upon the use of implanted electronic devices for the treatment of bradycardia or tachycardia (Winters et al. 2001; Gregoratos et al. 2004). The simple pacemakers of the 1970 gave way to the more advanced devices capable of real-time diagnosis of arrhythmia-related disorders and their treatment. The pacemaker technology and clinical practice permits the use of either single-chamber ventricular pacemakers or dual-chamber pacemakers for patients who require cardiac pacing in some indications. The single-chamber ventricular pacemakers are less expensive and their implantation is simpler than the implantation of the dual-chamber pacemakers. Several randomized studies comparing the dual-chamber “physiological” pacing to the ventricular pacing (Mattioli et al. 1998; Connolly et al. 2000; Lamas et al. 2002) have shown that the physiological pacing was more beneficial in some subgroup of patients despite the fact that both variants caused mortality, stroke and heart failures were no significant differences. Clinical studies focusing on the quality of life with respect to the type of the pacing system implantation were also conducted. The results of the studies confirmed that the implantation of a permanent pacemaker improves health and quality of life. The quality-of-life benefits associated with dual-chamber pacing as compared with the ventricular pacing were observed only in the subgroup of patients with the sinus-node dysfunction (Lamas et al. 1998). The DAVID trial (Dual Chamber and VVI Implantable Defibrillator) demonstrated increase heart failure hospitalisation and mortality in patients programmed to the DDDR mode compared to ventricular backup pacing. The Multicenter Automatic Defibrillator Implantation Trial II (MADIT II) evoke the question of minimising the unnecessary right ventricular pacing (Wilkoff et al. 2002; Moss et al. 2002; Galley et al. 2004). The myocardial energy demands during the cardiac pacing were studied experimentally (Montgomery et al. 1991; Portman and Ning 1995). The clinical data demonstrating the energetic changes during the pacing on the patients with dilated cardiomyopathy were shown by Baller et al. (2004). Kerckhoffs et al. (2005) carried out mathematical modelling of the right ventricular pacing and comparison with spontaneous depolarization. In the present study we are using the numerical model of the human cardiovascular system developed in the Institute of Thermomechanics, Czech Academy of Sciences (Převorovská and Maršík 2004) with the purpose to find the parameters of the myocardial energetics during the cardiac rhythm pacing. Methods and materials Patients For our research purposes the patients were designated on the basis of selective criteria. The inclusion criteria were established sick sinus syndrome and pacemaker implantation lasting more than 6 months. The exclusion criteria were left ventricle dysfunction, clinical significant valvular disease and other pathological structural heart changes. The patients with other than sinus rhythm, AV block under 110 beat/min during stimulation or bad image quality from echocardiography examination were also excluded. Our research finally involved thirteen patients with the sick sinus syndrome (eight males and five females with the average age 73 years, ranging 59–82 years). These patients with permanent dual-chamber pacemaker implantation were taken to echocardiography examination. Study protocol The numerical model of cardiovascular system was used to calculate the myocardial energetic parameters such as the stroke work index and the myocardial oxygen consumption for three types of pacing. The stroke work index SWI (gm/m2) is used as a measure of myocardial contractility or work performed by the heart at every contraction. The work performed by the heart can be simply by the fluid dynamic formula The stroke work index of the left ventricle LVSWI is then determined by the stroke volume SV (ml), by the difference of the mean arterial pressure MAP (mmHg) and the pulmonary artery occlusion pressure PAOP (mmHg) (considered to be the indirect measure of the left atrial pressure) by the formula (http://www.lidco-ir.co.uk/html/clinical/nhp.asp)1: the constant 0.0136 converts the units mmHg l/m2 to gm/m2. BSA (m2) is the body surface area given by the DuBois formula (Wang et al. 1992): where BW (kg) is the body weight and BH (cm) is the body height. The myocardial oxygen consumption MVO2 (ml O2/beat) is calculated by the relationship proposed by Suga 1990, which is based on the time-varying elastance model of the ventricle. This relationship represents linear correlation between the total pressure-volume area PVA (J/beat) and MVO2 (1 ml O2 = 20 J) where a is the slope coefficient and b is unloaded MVO2. PVA consists of the external mechanical work EW (J/beat) within the contraction cycle and the potential energy PE (J/beat) The values EW (representing the work done by the ventricle to eject blood) and PE (the work against the viscoelastic properties of the myocardium) can be estimated using the pressure-volume loops (see Fig. 1) Fig. 1Ventricular pressure–volume loopsThe following symbols are used in Eq. 5:ppeakS (Pa) maximal value of the systolic ventricular pressureppED (Pa) end-diastolic ventricular pressureVED (m3) end-diastolic ventricular volumeVES (m3) end-systolic ventricular volumeVo (m3) ventricular volume at zero pressure (point of intersection on the volume axis, see Fig. 1). The input data for the numerical model, such as the anthropometrical data of the patients (weight, height) or the heart frequency, were obtained from their physical examination. The various model parameters, such as the calcium, sodium, potassium channel’s conductance, were fitted to obtain the stroke volume, blood pressure and pulmonary artery pressure established by the echocardiography examination which was performed during the spontaneous heart rhythm and the AAI, VVI, DDD pacing modes. Each mode was examined for stimulation frequencies 70, 90 and 110 beat/min. Each examination was performed 5 min after the mode change to ensure steady state. Standard views of the parasternal long-axis view and the Doppler outflow tract across the aortic valve were taken using the Agilent Sonos 5500 ultrasound device. The stroke volume was determined from the left ventricular outflow (aortic valve) diameter measured on the parasternal long-axis view and from the outflow velocity on the aortic valve measured from the apical five-chamber view (Huntsman et al. 1983; Lewis et al. 1984) by the Doppler probe. The blood pressure was measured by the sphyngomanometer for each stimulation mode. Then, the pulmonary pressure was measured by the Doppler echocardiography on tricuspid valve. Medical Ethical Committees of the University Hospital Královské Vinohrady, Prague, approved the study protocol and the informed consent was obtained from all patients. Mathematical model For the modelling purposes, the real cardiovascular system was compartmentalised (see Fig. 2). The 24-segment hemodynamical model (pulsating type) of the human cardiovascular system imitating the electrochemical and mechanical activity of the heart muscle was used for the numerical simulation of the myocardial workload changes during the cardiac rhythm pacing. Fig. 2Scheme of the human cardiovascular system applied to heart workload calculations The system is approximated by twenty-four compartments, which are characterized by elasticity, resistivity and transfer of the relevant chemical components. The pulsating heart consists of the following compartments: RA—Right Atrium, RV—Right Ventricle, LA—Left Atrium, LV—Left Ventricle. The pulmonary circuit includes compartments representing the pulmonary artery (PA), arteries (Paa), capillaries (Pc), postcapillary venules (Pvv), and pulmonary veins (PV). Systemic circuit connects the aorta (Ao), systemic arteries (Saa), capillaries (Sc), head arteries (HA), capillaries (Hc), and head veins (HV). In addition, the systemic circuit includes the coronary circulation with coronary arteries (CA), capillaries (Cc), veins (CV), shunt (CS), and also the bronchial circulation with bronchial arteries (BA), left bronchial veins (BVL), capillaries (BRc), and right bronchial veins (BVR). The compartments of the pulmonary and systemic blood circulation are connected in series with the four pulsating heart compartments. The suggested model describes the one-dimensional flow of incompressible blood through the network of elastic blood vessels. The heart compartments are considered to be made of anisotropic and viscoelastic incompressible material. The mathematical formulation describing the blood flow in the cardiovascular system is based on the mass, energy and momentum balance. The behaviour of the cardiovascular system is described by the heart rate and by its hemodynamical variables, i.e., blood pressure, volume and flow, and by the cardiovascular parameters, such as the compliances and resistances in the corresponding compartments. The blood inertia and physicochemical variables, such as the cardiac action potential and the calcium, sodium and potassium ion concentrations, are included to support more accurate performance of the cardiovascular system. The Beeler-Reuter model (Beeler and Reuter 1977) describes the behaviour of excitable cardiac cells. Discussion and conclusions All experimental results obtained with the investigated patients and simulated values resulting from the applied numerical model were elaborated by the following statistical analysis. We use the null hypothesis to demonstrate the agreement or discrepancy between the values of the myocardial energetic parameters (stroke work index, myocardial oxygen consumption) obtained by the numerical simulation for the three pacing modes. All comparisons of the data were made by the two-tailed Student’s test. The statistical analysis was performed with 5% significance level and the degree of freedom df = 24. The data for the stroke work index are expressed as the mean value ± standard deviation. The calculated values of the stroke work index at the three pacing modes are shown in the Fig. 3. It was shown that the differences in the compared pacing modes AAI and VVI and the modes DDD and VVI are statistically significant (P < 0.05). The comparison of the stimulation modes DDD and AAI shows that the differences are statistically insignificant (P > 0.05). Fig. 3Stroke work index calculated by the numerical model of cardiovascular and respiratory system The calculated values of the myocardial oxygen consumption for three pacing modes are demonstrated in the Fig. 4. The myocardial efficiency, which is an important factor for heart performance, changes according to the stimulation mode. The changes resulting from the hemodynamical model fitting by the well-measurable data, i.e., heart rate, stroke volume, systolic pulmonary artery pressure, systolic and diastolic aortic pressure. The conclusion is that the difference between the modes AAI and VVI and the modes DDD and VVI is statistically significant (P < 0.05). The statistical analysis of the AAI and DDD modes indicates that the compared data are not significantly different (P > 0.05). Fig. 4Myocardial oxygen consumption calculated by the numerical model of cardiovascular system in various pacemaker modes From the comparison of the obtained numerical values for the stroke work index between the AAI and VVI pacing modes and between DDD and VVI modes it follows that the differences are statistically significant (P < 0.05). The results are consistent with the published data (Leclercq et al. 1995; Nielsen et al. 1998). The results of the statistical analysis for the AAI and DDD stimulation modes show that the differences between the compared groups are statistically insignificant (P > 0.05). These results are consistent with the clinical observation where can achieve a similar clinical outcome in AAI and DDD modes in this group of patients (Masumoto et al. 2004). The increase of the stroke work index in AAI or DDD modes was caused by saving the AV synchrony in the patient with the normal left ventricle systolic function. During the AAI pacing mode the implantation of the pacemaker is easier. This fact supports the clinical superiority of the AAI mode against the DDD mode in the sick-sinus group patients with preserved left ventricular function. The results of the comparison of the obtained numerical values for the myocardial oxygen consumption between the AAI and VVI and between DDD and VVI modes show that the difference between the modes is statistically significant (P < 0.05). The increase of the myocardial oxygen consumption corresponds to the increase of the stroke work index. From the comparison of AAI and DDD modes we can see that the difference is statistically insignificant (P > 0.05). The insufficient number of the designated subjects gives the limitation of our research study because very strict exclusion criteria exist for the concomitant disease. Many pacemaker patients were not eligible for echocardiography examination because they had hypertension, left ventricle dysfunction or valvular disease. We conclude that mathematical modelling of the cardiovascular system is a recommended method for the estimation of the myocardial energy demand in patients with heart diseases.

Behav_Genet-3-1-1914301

Variance Decomposition Using an IRT Measurement Model

Large scale research projects in behaviour genetics and genetic epidemiology are often based on questionnaire or interview data. Typically, a number of items is presented to a number of subjects, the subjects’ sum scores on the items are computed, and the variance of sum scores is decomposed into a number of variance components. This paper discusses several disadvantages of the approach of analysing sum scores, such as the attenuation of correlations amongst sum scores due to their unreliability. It is shown that the framework of Item Response Theory (IRT) offers a solution to most of these problems. We argue that an IRT approach in combination with Markov chain Monte Carlo (MCMC) estimation provides a flexible and efficient framework for modelling behavioural phenotypes. Next, we use data simulation to illustrate the potentially huge bias in estimating variance components on the basis of sum scores. We then apply the IRT approach with an analysis of attention problems in young adult twins where the variance decomposition model is extended with an IRT measurement model. We show that when estimating an IRT measurement model and a variance decomposition model simultaneously, the estimate for the heritability of attention problems increases from 40% (based on sum scores) to 73%. Introduction In quantitative genetics, one is interested in the extent to which variation in certain characteristics is heritable. Heritability is expressed in terms of the proportion of the variance of a trait in a population that can be attributed to genetic differences. This genetic variance component can be estimated in, for example, the classical twin design (Boomsma et al. 2002a) in which the covariance structures of monozygotic and dizygotic twins are compared. However, it is not always straightforward to estimate variance components. A variance component is only meaningful when measures are expressed on a scale of at least interval level. Moreover, many statistical methods require the phenotype to be normally distributed. Many phenotypes are not expressed in clearly defined units and are at best ordinal in character (e.g., conservatism, extraversion). Some traits have even only a nominal character (e.g., psychiatric disorders). There are several ways of dealing with such nominal data. One possibility is to focus on concordance rates and compute recurrence risk ratios (Risch 1990, 2001). Alternatively, one might assume a latent continuous trait with a threshold above which individuals are affected and estimate the heritability on that latent trait (Lynch and Walsh 1998; Falconer 1965; Crittenden 1961). This method can also be used with ordinal data. For some traits, it is convenient to have multiple indicators (items). For example one might have for a particular disease 10 symptoms that each can be scored as absent (0) or present (1). For each individual one can then compute a sum score that indicates to what extent the individual is affected by the disease. Such sum scores usually show a normal distribution or do so after an appropriate transformation. It is typically assumed that the normally distributed scores or transformations thereof reflect a continuous interval scale and the variance of the sum scores is subsequently decomposed. This approach follows classical test theory (CTT) where it is assumed that the observed score (the sum score) is the aggregate of a true score and a random component, usually referred to as measurement error. When decomposing the variance of sum scores, the measurement error variance (the unreliability) ends up as part of the non-shared environmental variance. As a result, when the reliability of a scale is low (i.e., the measurement error is large) and the analysis is based on sum scores, the heritability of the actual trait is significantly underestimated. Modelling sum scores is appropriate if the sum scores are highly reliable (for instance because they are based on a large number of correlated items) and well validated. Furthermore, there should be enough variation and the distribution should be more or less normal. Finally, there should be no data missing. If these requirements do not hold, item response theory (IRT) provides a well-established alternative to classical test theory. This paper introduces the basics of the IRT framework, after which its advantages over a sum score approach are discussed. Next, it is argued that IRT models should be estimated simultaneously with the variance decomposition model, which can be done using a Bayesian approach with Markov-chain Monte Carlo estimation. Lastly, a simulation study shows the potential bias when estimating variance components on the basis of sum scores and the Bayesian method is illustrated with an empirical data set on attention problems. Item response theory models In IRT models—as opposed to CTT—the influence of the items and the respondents are explicitly modelled by distinct sets of parameters. In these models, an assumed continuous latent variable θ reflects the trait and every item is identified by thresholds β where a response in one category becomes more likely than a response in an adjacent category. It is usually assumed that the latent variables θj are drawn from a normal distribution, that is, θj are independently and identically distributed N(μ, σ2), though this assumption is not always necessary to identify the model parameters. The probability of the presence of the symptom i in individual j, p(Yij = 1), is a function of the difference between the individual’s trait score θj and the parameter βi, with βi indicating the location on the scale where the presence of a symptom becomes more probable than its absence. In the case of multiple symptoms, we have with denoting the cumulative standard normal distribution function. That is, the probability of the presence of symptom i in person j is a function of both a person’s liability score θj and a symptom (or item) parameter βi. In the IRT framework, this model is referred to as the one-parameter normal ogive model, or 1PNO (Lawley 1943; Lord 1952, 1953). This model is identified with a location restriction, for example, μ = 0. The variance of the latent trait, σ2, can be estimated and can be interpreted as the covariance of the items: the larger the variance, the higher the reliability of the scale. An alternative parameterisation replaces the normal ogive by a logistic curve, that is, where This version of the model is known as the one-parameter logistic model (1PLM), or Rasch model (Rasch 1960). To illustrate the model, consider an individual with a score θj of 1 on the latent trait, and a particular item with parameter β = 1. Then the probability of a positive response from this individual on this item equals exp(1 − 1)/(1 + exp(1 − 1)) = exp(0)/(1 + exp(0)) = 1/2 = 50%. An individual with a score higher than 1 has a higher probability of showing a positive response, whereas an individual scoring lower than 1 has a lower probability. Individuals with a latent score of −1 have a probability of exp(−2)/(1 + exp(−2)) = 12%. With a simple multiplicative transformation of the scale, the logistic and normal ogive curves are very similar and indistinguishable for all practical work (see, for instance, Lord 1980). In the Rasch model, as well as in the 1PNO model, all items have the same correlation (‘‘factor loading’’) with the underlying latent trait. Analogous to factor models, it is possible to estimate factor loadings that differ across items. In the IRT framework these factor loadings are referred to as discrimination parameters αi. These parameters indicate the extent to which an item i loads onto the latent trait, and the model becomes An alternative form in the literature replaces αθ − β with α(θ − β). This leads to a somewhat different interpretation of the β-parameters (they are scaled differently) but it only involves a reparameterisation. Essentially, a one-parameter model can be described by a two-parameter model where all α parameters are equal. In order to identify the model and estimate α, however, the variance of the latent trait should be fixed. Thus, a one-parameter model with a large variance of the latent trait is equivalent to a two-parameter model with large discrimination parameter α that is equal for all items together with a fixed variance of the latent trait. The two-parameter model must be identified by both a location and a scale restriction. The former can be the same restriction as above, that is, μ = 0. The latter can be the additional restriction that the variance of the latent distribution is equal to one, that is, the model is identified by assuming a standard normal distribution, N(0,1), for the latent ability parameters θj. Alternatively one fixes one of the discrimination parameters to unity. Generally, however, this identification solution is not advisable, because the standard errors of the parameters blow up if the discrimination parameter chosen for the identification is poorly identified. IRT models for polytomous data Often, measurement is based on items or symptoms with more than two categories. For example, answers can be coded as 0 (not at all), 1 (somewhat, sometimes) and 2 (a lot, often). Typically in CTT approaches in behaviour genetics the sum of these item scores is regarded to represent a person’s score on the trait of interest and is used for the statistical inference. There are several IRT models for ordered categories (e.g., Samejima 1969; Masters 1982). These have different rationales and are not reparameterisations of each other, but the practical implications for preferring one over the other are often negligible. Here we describe a continuation-ratio model (Tutz 1990; Verhelst et al. 1997). This model allows the transformation of a polytomous item into a set of dichotomous items, which facilitates model estimation. The response to a polytomous item is viewed as a set of responses to an ordered sequence of virtual dichotomous items: it is assumed that the respondent is administered virtual items until an incorrect or negative response is given. So, in this approach, an item with M categories labelled m = 0,..., M − 1, the response is dummy-coded into M − 1 dichotomous quasi-items. As an example, for an item with m = 3 categories we make two new virtual items. A score of 2 would be coded as correct responses to both virtual items. A score of 1 on the original item would be coded as a correct response to the first virtual item and an incorrect response to the second virtual item. A score of 0 would be coded as an incorrect response to the first virtual item and the second virtual item would be coded as not administered (missing). Now the responses to all virtual items can be modelled by an IRT model for dichotomous items, such as the models given by Eqs. 1, 2 or 3 and can be estimated by any IRT software package that can handle dichotomous items in combination with missing data. There are also IRT packages that estimate models for polytomous items directly (e.g., Multilog; Thissen et al. 2003). Advantages of using an IRT framework compared to analysing sum scores We will discuss four advantages of using IRT: (1) it supports construct validity and the scoring rule (e.g., a scoring rule might consist of taking the unweighted sum of symptoms as an estimate of a person’s liability), (2) it supports the use of incomplete item administration designs and handling of missing data, (3) it supports accounting for measurement error, and (4) it can handle floor and ceiling effects. An IRT framework allows one to explicitly model the relationship between item scores and the phenotype of interest. Any combination of items can of course be summed (weighted or unweighted), but this does not guarantee that the sum score reflects a meaningful construct. The meaningfulness of the measurement can be directly assessed in an IRT framework. Fit to an IRT model is empirical evidence that the observed responses can be explained by an underlying structure. The latent variable of the IRT model should, of course, be an appropriate representation of the construct to be measured. The IRT model that fits the data determines the score rule of the measurement instrument. If, for instance, a one-parameter model does not fit the data, but a two-parameter model does, the sum score where the items scores are weighted with their respective discrimination parameters is a sufficient statistic for θj (Lord and Novick 1968). So some items can be more important or sensitive indicators of a trait than others. Modelling the item data in a variance decomposition analysis allows the separate evaluation of model fit regarding the measurement model and the variance decomposition model. In addition, group differences can be modelled, through differences in means, variances and variance components, and through differences in the way symptoms relate to the latent trait. For instance, one or more symptoms may show a higher incidence rate in one group (indicated by a difference in β-parameters across groups, e.g., females and males), or be a more sensitive indicator for the trait in a particular group (indicated by a difference in α-parameters across groups). Such violations of measurement invariance are usually referred to as differential item functioning (DIF). A practical advantage of the analysis of data using an IRT framework is the use of incomplete item administration designs and handling of missing data. In some situations, intentionally incomplete item administration designs can greatly improve the efficiency of data collection. With an IRT approach one can also effectively deal with problems specific to longitudinal research where items differ across waves. When using IRT models in a maximum likelihood or a Bayesian framework, it is easy to include individuals that have missing data on one or more items if the data are missing at random (Little and Rubin 1987). When data are not missing at random, the non-randomness can be modelled within an IRT framework by expanding the model with an IRT model that describes the pattern of the missing data (see, for instance, Moustaki and Knott 2000; Moustaki and O’Muircheartaigh 2000; Holman and Glas 2005). The encompassing framework for handling missing data using IRT offers an important advantage over classical test theory. In classical test theory sum scores are only meaningful if the items are the same in all individuals and at all measurement waves. The third advantage of the analysis of data using an IRT framework is that it accounts for measurement error. Unreliability suppresses the correlation between measurements (attenuation). Particularly when using a scale with only a few items, the correlations amongst sum score variables may be grossly attenuated. Clearly, this has important implications for the estimation of variance components in genetic research. In an IRT framework, the problem can be solved by, instead of focussing on sum scores, considering the correlations between latent variables (see, for instance, Béguin and Glas 2001; Fox and Glas 2003). These so-called latent correlations can be seen as estimates of correlations corrected for attenuation. A simulation study and an application of IRT to real data in a later section will show the possible extent of such attenuation effects on the estimation of heritability. The fourth advantage of IRT has to do with floor and ceiling effects. A problem of analysing sum scores that represent indices of psychopathology is that these scores show a skewed distribution in the general population (Van den Oord et al. 2003; Derks et al. 2004). These skewed distributions result from the fact that many behavioural phenotypes are assessed using questions that relate to symptoms that are relatively rare in the population. These distributional violations may have important implications for the inference regarding relative variance components when analysing sum scores (Derks et al. 2004). In an IRT framework one is essentially free to specify the distribution of the latent trait (in some cases, it can even be estimated). In most cases, with polygenic traits, a normal distribution seems the most reasonable alternative (a mixture approach may be more suitable for traits with only a few large QTL effects). When in turn the variance of the normally distributed latent trait is decomposed into genetic and non-genetic variance, the inference is unbiased if the assumptions of the model are correct. Variance decomposition: the one-step and the two-step approach In IRT models, the latent scores θj are typically assumed to be random draws from a normal distribution. When we are interested in the extent to which individual differences on the latent trait are heritable, we only need to decompose the variance of the θjs using, for example, the classical twin design. There are two approaches. The first approach is to first estimate the parameters of the IRT model using standard IRT software (such as, Bilog, Multilog, Parscale, Testfact, ConQuest, OPLM), and then to have the same software estimate each individual score on the latent trait. Next, one uses these estimates of the θjs as observed values in a standard variance decomposition analysis. This we call the two-step approach. There are several disadvantages to this two-step approach. First of all, in the IRT model fitting phase, the usual IRT estimation software cannot handle the dependency in the data inherent in twin and family designs. In some cases, with simple designs such as with sibling pairs only, weighting of the data would come a long way in solving this problem, but with more complex family designs, weighting is not a satisfactory solution. Second, when estimating latent scores for each individual, the estimates of the θjs, just like sum scores in the CTT tradition, are not simply observations but estimates with error variance. When computing the confidence intervals for the heritability estimates in the second phase, this uncertainty on the latent scores is not taken into account and the heritability confidence intervals are consequently too narrow and the estimates biased downwards. Moreover, in an IRT framework, the confidence intervals for estimates of individual latent scores are dependent on their location on the scale (actually, the number of items with β-parameters that are similar in magnitude to the person score θ and the items’ discriminatory power, α), whereas in the variance decomposition, it is assumed that measurement error (as included in the non-shared environmental variance component) is independent of location (cf. CTT). For example, many psychopathology scales have only items that refer to relatively rare symptoms. As a consequence, many individuals in the general population score 0, which does not necessarily imply that all actually have the trait to the exact same degree. In other words, the scale provides very little information on the trait on the low end. In contrast, the upper end of the scale usually shows more variation, which may imply that the measures are more reliable (more items that discriminate between individuals). Thus, a priori it seems likely that psychopathological scales have more discriminatory power at the upper end of the scale than at the lower end. Of course, this is not a bad thing, since these scales were designed to discriminate between the healthy and the sick. Therefore it seems reasonable to forego the assumption of equal reliability across the scale and take differing reliabilities into account. Actually, using the two-step approach the heritability coefficient estimate will be about the same as when the analysis is carried out on sum scores. This is because IRT estimates and sum scores correlate highly, well over 0.90 in the case of two-parameter models. When applying a one-parameter model, the correlation will be practically one, because a basic assumption of the Rasch model is that a sum score is a sufficient statistic for the score on the latent trait. Therefore, all persons with the same sum score will get the same estimate on the latent trait. Thus, a third problem of the two-step approach is that it neither solves the attenuation problem, nor the non-normality, nor the ceiling effects. In order to take full advantage of the IRT approach, it is critical to estimate both the measurement model and the variance decomposition model simultaneously, using a one-step approach. However, computationally this is rather challenging. Below, it is shown how this can be done using software for Bayesian estimation procedures. In an application in a later section, we demonstrate the one-step approach for the estimation of heritability with both simulated and empirical data. Bayesian estimation using a Markov chain Monte Carlo algorithm In twin studies, a widespread method of estimating variance components is through structural equation modelling (SEM). For continuous traits with normal distributions, this is a flexible approach in that it is able to accommodate all linear models and allows for testing of equality of means, variances, covariances and variance components across subpopulations. However, with more elaborate models with discrete or categorical observed variables, SEM maximum likelihood (ML) estimation or ML procedures for estimating generalised linear mixed models such as GLAMM (Rabe-Hesketh and Skrondal 2005) soon reach computational boundaries. An alternative method is Bayesian statistical modelling with Markov chain Monte Carlo (MCMC) estimation algorithms (see also Eaves et al. 2005). In the Bayesian approach, inference is based on the posterior density of the model parameters, P(η|Y), where η represents the vector of model parameters and Y the observed data. By Bayes’ rule, the density P(η|Y) is proportional to the product of the likelihood of the data given the model parameters P(Y|η) and the marginal density for η, P(η), that is, The marginal distribution of η is termed the prior distribution (prior in the sense of before the data have been taken into account), and must be specified by the user. The model provides us with the likelihood function P(Y|η), and hence the posterior distribution of η is determined (posterior in the sense of after the data have been taken into account). The posterior distribution is a description of the probabilities of possible values for η given the observed data and forms the basis for statistical inference. We may, for example, take the mean or the median of this distribution as our point estimate for η. Further, the interval between the 2.5th and the 97.5th percentile of the posterior distribution provides the so-called central 95% credibility region, which is analogous to a 95% confidence interval in the ML framework. For more on Bayesian statistics, the reader is referred to the introductions by Box and Tiao (1973) and Gelman et al. (2004). Sometimes it is easy to compute the posterior distribution analytically, but very often this is not possible. One can then use computer simulation to draw a sample of η-values from the posterior distribution. The mean or median of the posterior distribution can then be approximated by the mean or median of the sampled η-values, and approximate credibility regions can be determined in a similar way. In practice, the joint posterior distribution of all model parameters is usually quite complicated. Therefore, the complete set of parameters is split up into a number of subsets in such a way that the conditional posterior distribution of each subset given all other parameters has a tractable form and can be easily sampled from. This approach is known as Gibbs sampling (Geman and Geman 1984; Gelfand et al. 1990; Gelman et al. 2004), which is a special case of an MCMC algorithm. When however the conditional posterior distribution of a subset of the parameters is not easy or even impossible to sample from directly, other MCMC algorithms can be used, where one samples from a similar proposal distribution and uses a decision rule to either accept or reject a sample so that the accepted values can be regarded drawings from the target distribution. In each iteration of an MCMC algorithm, a sample is taken from each conditional posterior distribution for each subset of the parameter space, given the current values of the other parameters. After a number of so-called ‘‘burn-in’’ iterations, necessary for a chain to achieve stationarity (i.e., approaching the target distribution: the joint posterior distribution) sufficiently closely, the subsequent draws can be regarded as sampled from the joint posterior distribution. The application of the Bayesian approach with MCMC sampling to IRT models is mainly motivated by the fact that IRT models with complex dependency structures require the evaluation of multiple integrals to solve the estimation equations in a likelihood-based framework. This problem is avoided in an MCMC framework. In recent years, the fully Bayesian approach has been adopted to the estimation of IRT models with multiple raters, multiple item types, missing data (Patz and Junker 1999a, b), testlet structures (Bradlow et al. 1999, Wainer et al. 2000), latent classes (Hoijtink and Molenaar 1997), models with a multi-level structure on the ability parameters (Fox and Glas 2001, 2003) and the item parameters (Janssen et al. 2000), and multidimensional IRT models (Béguin and Glas 2001). In behaviour genetics, the approach has been taken up by Eaves and his co-workers (Eaves et al, 2005; Eaves et al. 2004). In IRT research, the Gibbs sampler is used in two versions: a version with a normal ogive representation such as in Eq. 1, introduced by Albert (1992), and a version with a logistic representation introduced by Patz and Junker (1999a). Below, a logistic version will be used for simulated and real data, implemented in the freely obtainable MCMC software package WinBUGS (http://www.mrc-bsu.cam.ac.uk/bugs/). Genetic models may be specified in WinBUGS as follows. Under the assumption that an ACE variance decomposition model (additive genetic, shared environmental and non-shared environmental effects) is appropriate for a latent trait θ, the model can be parameterised as a linear random effects model (see also Van den Berg et al. 2006a): where ck denotes the environmental effect for being a member of family k, and ejk denotes the environmental effect of being individual j in family k. The genetic component is split into a1 and a2 to model the different genetic correlations amongst monozygotic (MZ) and dizygotic (DZ) twins (cf. Jinks and Fulker 1970). The genetic correlation in MZ twins is usually assumed 1.0 and in DZ twins 0.5, in other words, the genetic covariance in MZ twins is twice as large as in DZ twins. Therefore, if we let the random effect a1 be constant within all families and we let a2 vary within families only for DZ twins (but be constant for MZ twins), and then fix the variances of a1 and a2 to be equal, the genetic covariance in MZ twins will be twice as large as in DZ twins. The variance of a1 and a2 together, VAR(a1) + VAR(a2) = 2 * VAR(a1) can then be interpreted as the variance due to additive genetic effects. We assume that a1 ∼ N(0, ½ σ2a), a2 ∼ N(0, ½ σ2a), c ∼ N(0, σ2c), and e ∼ N(0, σ2e). The case for the ADE model can be derived similarly (Van den Berg et al. 2006a). For some estimation problems, it might be computationally more convenient to model sum and differences scores, instead of the latent scores for the twins separately (Van den Berg et al. 2006a; Robert and Casella 2004, p. 396; cf. Boomsma and Molenaar 1986). Simulation To illustrate the effect of attenuation on heritability estimates, 101 datasets were generated consisting of 400 MZ twin pairs and 600 DZ twin pairs. A standard normally distributed latent trait was simulated with an additive genetic component of 72% and a non-shared environmental component of 28%. The 1PL IRT model was used to simulate responses to 14 dichotomous items, where the β parameter values ranged from 0.5 to 3.5, with increments of 0.25. This corresponds to questionnaire items that are rarely endorsed by people. The simulated item data were fitted using a model with additive genetic and non-shared environmental effects (AE model) on a latent trait and a 1PL measurement model. Next, sum scores were computed and these were analysed with an AE model. Since the distribution of the sum scores is positively skewed, the AE analysis was also performed after a logarithmic transformation of the sum scores. The simulations were carried out using the software package . For each replicated data set, we computed the twin correlations for the latent scores, the twin correlations of the sum scores and the twin correlations for the log-transformed sum scores. The three types of analyses were carried out in WinBUGS. After a burn-in phase of 1000 iterations, the characterisation of the posterior distribution for the model parameters was based on 1000 iterations from 2 independent Markov chains. From each of the 3 (analyses) * 101 (replicated data sets) marginal posterior distributions for the heritability we took the mean and the median as point estimates. Further simulations were carried out to illustrate the attenuation effect and the bias in variance components. For simple genetic models, the twin correlations are sufficient statistics for the variance decomposition. Therefore it is enough to show how correlations based on sum scores behave as a function of number of items and beta parameters. Data were simulated using bivariate normally distributed latent values, with correlations 0.9, 0.7, 0.5, 0.3 and 0.1. These latent values were used to simulate corresponding sum scores using a one-parameter logistic IRT measurement model under a variety of conditions. First of all, we used different degrees of discrimination of the items (i.e., the variance of the latent trait: 0.676, 1 and 100). Second, we varied the way in which the items are distributed across the scale, either evenly scattered so that sum score distributions are symmetrical, or only scattered on the upper half part of the scale, that is, using only items that less than 50% of the population endorses, which results in positively skewed sum score distributions (cf. Derks et al. 2004; van den Oord et al. 2003). Third, we varied the number of items (5, 10, 20, 50, 100) to investigate attenuation. Simulation results Taking the median parameter values from the 101 data sets, the simulated latent data correlated 0.72 in MZ twins and 0.36 in DZ twins, just as would be expected. The sum scores correlated 0.45 in MZ twins and 0.21 in DZ twins (medians of the 101 data sets) and the log-transformed sum scores correlated 0.41 and 0.20, respectively. Thus, twin correlations are severely attenuated when analysing sum scores, even with 14 items. Analysing the simulated item data with a 1PL IRT model, using the one-step approach, we recovered the true 72% value for the heritability coefficient closely (see Table 1). When analysing the raw sum scores using a normal AE model, either with or without transformation, the heritability point estimate dropped considerably, to about 42%. Thus, when the true model is an IRT model and the number of items is limited, an analysis of raw or transformed sum scores can lead to extensive underestimation of heritability. Table 1Simulation results. Reported heritability values are the medians of the 101 posterior means and medians, standard deviations between parenthesesMethod of analysisHeritability coefficient point estimatesPosterior meanPosterior median1PL IRT model0.7232 (0.0585)0.7245 (0.0589)Sum scores continuous model0.4364 (0.0393)0.4369 (0.0395)Log-transformed sum scores0.4046 (0.0403)0.4047 (0.0406) For each condition of latent correlation, number of items, and variance of the latent variable, we simulated 100,000 twin pairs and correlated their sum scores. Figure 1A shows the result for the condition where the variance was 1 and the items were nicely scattered across the distribution of the latent values, between −2½ and 2½ times the standard deviation (1). The attenuation effect is clearly dependent on the number of items: with 100 items, the correlation on the basis of the sum scores is very close to the true correlations. An analysis treating the sum scores as bivariately normal and applying a variance decomposition will approximate the true proportions. Moreover, the degree of the attenuation is proportional to the true correlation: with 5 items, a true correlation of 0.9 will be attenuated to a correlation of 0.55 (61%) and a true correlation of 0.1 will be attenuated to a correlation of 0.06 (60%). Therefore, when the analysis on 5 items is based on the sum score, and the true MZ correlation equals twice the DZ correlation, this ratio is maintained when analysing sum scores. Thus, when applying an AE model, heritability will be underestimated, but no artifactual shared environmental effects or dominance genetic effects will appear as a result of analysing sum scores. Fig 1Correlations of simulated sum scores as a function of true correlation at the latent level, variance of the latent trait (quality of the scale), and scatter of the item β parameters (entire scale or only top half, i.e., all > 0) Figure 1B shows the result for a scale with slightly worse discrimination: the variance of the trait is now only 0.767. The items are again nicely scattered, between −2½ SD (−2.05) and 2½ SD (2.05). Thus, we retain the spread of the β values in terms of the SD, so that the expected proportion of individuals scoring a particular number of items remains equal across simulation situation; the resultant distribution of the sum scores will be equal. But now, due to the decreased sensitivity of the scale, the number of items has a more pronounced effect on the attenuation. The sum score correlations are now lower than under the model with variance = 1. However, the attenuation effect is still proportional to the true correlations. Figure 1C shows an extreme situation where the items have high discriminatory power. The variance is now 100, and the items are evenly scattered between −25 and 25. Note that again, we retain the scatter of the beta values in terms of the SD, and again the sum score distribution will not be different from the earlier simulations. However, with such a sensitive scale, practically everybody that scores less than 1 SD below the mean will show a sum score of 16% of the total number of items. Everybody with a latent score higher than 1 SD below the mean will show a sum score of 84% of the number of items. Moreover, the data will show a scalogram pattern, for example with 3 items with increasing difficulty, the only observed patterns will be 111, 110, 100 and 000. Such a pattern will not be observed when the variance is 1, and even less so with a variance of 0.767: more individuals will then show patterns like 101 and 011, etc. Again, attenuation occurs when the number of items is limited, but the effect is much less pronounced, and again the attenuation is proportional across the different correlations. In this situation, an analysis of sum scores will yield reasonable estimates for the variance components given a sufficient number of items. Actually, when the raw item data follow the scalogram pattern, the true correlations and the corresponding variance components will be recovered when applying a threshold model (Lynch and Walsh 1998). This is also true when the data follow a scalogram pattern but the items are not evenly scattered across the scale and the sum score distribution is skewed: applying a threshold model will recover the true correlations (cf. Derks et al. 2004). However, when applying an ordinary variance component analysis, ignoring its non-normality will yield biased estimates, underestimating the effects of shared environment and overestimating the effects of dominance (cf. Derks et al. 2004). This because when the items are not evenly scattered and the sum score distribution is skewed, the attenuation effect is no longer proportional to the true correlations (Fig. 1D): small correlations are more severely attenuated than large correlations. In case the true DZ correlation equals half the true MZ correlation, DZ:MZ = 1:2, the correlations of the sum scores will show a smaller ratio, DZ:MZ < 1:2, usually an indication of dominance genetic effects or epistasis. This is hard to see from the Fig. 1D, but with 5 items, the simulated sum score correlation is 0.83667 when the true correlation is 0.9 (92.96%), 0.42429 when the true correlation is 0.5 (84.86%), and 0.076575 when the true correlation is 0.1 (76.58%). Suppose we could analyse the true correlations, 0.9 and 0.5. One would then conclude that additive genetic variance accounts for 80% of the variance, non-shared environmental effects 10% and the shared environmental effects for the remaining 10%. Now if we would base our analysis on the observed sum score correlations 0.84 and 0.42, we would conclude that there are no shared environmental effects. One can imagine that when the true correlations are 0.90 and 0.45 one would conclude dominance effects to be absent, whereas if one would analyse observed sum scores correlations, one would find evidence for dominance genetic variance, the extent of which is dependent on the number of items. Now, scalogram pattern data that fit a Guttman scale model are extremely rare. More often, item data follow a pattern that can be explained by the more lenient IRT model. Figure 1E shows the attenuation effect when the true model is a one-parameter IRT model with variance 1, where all items are endorsed by fewer than half the participants (i.e., all β parameters larger than the average latent score). Again we see that under the usual IRT model, the attenuation effect depends on the number of items and again we see that due to the skewness of the sum score distribution, the attenuation is not proportional to the true correlation. For example, with five items the simulated sum score correlation equals 0.591256 for true correlation 0.9 (66%), 0.316222 for true correlation 0.5 (63%), and 0.056567 for true correlation 0.1 (57%). When true correlations are again 0.9 and 0.5, the most likely model would be, when based on an analysis of the sum scores with 5 items, 5% dominance genetic variance, 61% additive genetic variance and 34% non-shared environmental variance. Thus, also under the IRT model, analysing sum scores leads to an underestimation of shared environmental effects and an overestimation of dominance genetic effects when the sum score distributions are skewed. An application We illustrate the decomposition of variance using an IRT measurement model with data from the Netherlands Twin Registry (NTR; Boomsma et al. 2002b). Attention problems were measured with the Young Adult Self-Report (YASR; Achenbach 1997). We used data collected in the year 2000 from 460 males and 966 females from MZ twin pairs, 288 males from DZ same-sex twin pairs, 561 females from DZ same-sex twin pairs, and 305 males and 441 females from opposite sex twin pairs. All twins were between 18 and 30 years (inclusive). All available data were used, including data from incomplete pairs and individuals with several items missing. It was assumed that data were missing at random (cf. Van den Berg et al. 2006c). The attention problems (AP) subscale of the YASR consists of seven items (see Table 2) with three ordered response categories (0 = Not true, 1 = Somewhat or Sometimes True, 2 = Very True or Often True). In children, sum scores typically show a high heritability with a significant non-additive genetic component (Rietveld et al. 2004). In young adults, AP sum scores also showed heritability (40%), but no non-additive genetic component (Van den Berg et al. 2006c). Table 2Items of the attention problems subscale of the young adult self-report (YASR; Achenbach 1997)ItemDescription1I act too young for my age2I have trouble concentrating or paying attention3I daydream a lot 4My school work or job performance is poor5I am too dependent on others 6I fail to finish things I should do7My behaviour is irresponsible Here, we estimate A and E variance components using a 1PL measurement model. A main effect of sex, δ, was modelled on the latent trais. The seven original items with three response categories were transformed into 14 dichotomous dummy items for each individual as described above. A separate β-parameter was estimated for each dummy item, so that for each original item there are two β-parameters. For the variance components, locally non-informative (‘‘flat’’) inverse gamma priors were used, and for the β and δ parameters we used locally non-informative normal priors. The parameterisation modelled the variances of sum and differences scores for the latent trait (Van den Berg et al. 2006a). The appendix gives the WinBUGS script. Three independent MCMC chains were used with randomised starting values. The chains converged rapidly to the stationary distribution with relatively low autocorrelations. The first 1000 iterations were discarded as burn-in samples, and a further 1000 iterations were used for inference. Results Table 3 gives the descriptives of the marginal posterior distribution of the parameter values. The estimate for heritability based on the mean of the posterior distribution is 73%. The main effect of sex on the latent trait, with females scoring higher than males, is just significant, as zero is not included in the central 95% credibility region. Values of the β-parameters are all around zero or positive, indicating that the AP scale is most sensitive for individuals with considerable attention problems but has a hard time discriminating individuals with relatively few problems with attention. This results in the severely skewed distributions of sum scores. Table 3Descriptives of marginal posterior distributions for the AE variance decomposition model using the 1PL IRT model for polytomous items with a main effect for sexParameterMeanSD2½th percentileMedian97½th percentileσ2a0.840.070.710.840.99σ2e0.320.060.200.320.44δ−0.130.05−0.24−0.13−0.02β110.250.050.150.250.34β122.760.102.562.762.96β21−0.760.05−0.86−0.76−0.66β222.440.082.302.452.60β31−0.430.05−0.53−0.43−0.33β321.840.081.711.841.98β411.900.061.781.902.02β423.960.203.583.964.36β510.220.050.130.220.32β523.030.102.833.023.23β610.620.050.530.620.73β623.900.153.633.904.19β712.570.072.442.572.71β724.610.314.054.605.27h20.730.050.630.720.82Note: First index of the betas refers to the item (see Table 1) and the second to the threshold The estimate for the heritability (73%) is much larger than the one reported earlier based on sum scores (40%, Van den Berg et al. 2006c). In the current sample, twin correlations for sum scores are very much like those reported earlier (MZ:0.45, DZ:0.17). By applying an IRT measurement model the twin correlation estimates for the latent trait are much higher, 0.76 for MZ twins and 0.30 for DZ twins. For comparison, when using a two-step approach, first estimating IRT model parameters in Multilog and then estimating latent scores for each individual (correlation between sum score and IRT estimate: 0.98), the results showed twin correlations nearly identical to those based on sum scores. The 1PL IRT measurement model could easily be extended to include discrimination parameters (‘‘factor loadings’’). It is most convenient to constrain these to be positive through the specification of lognormal priors where for instance α = exp(γ) and γ ∼ N(0, 100). In this case, the heritability estimate was not affected by this extension of the model (results not shown). Discussion We have compared an IRT model with a sum score approach with indirectly measured phenotypes. Under a range of conditions, the IRT framework is to be preferred over using sum scores. For example, in longitudinal studies with data missing by design or changing measurement instruments, when some items in a questionnaire change across birth cohorts or across different ages or when item data are missing, a sum score approach may no longer be appropriate, but in many cases the analysis can still be meaningfully carried out in an IRT framework using parameter expansion (see, for instance, Glas 1998). When a simple IRT model does not fit the data, one could consider deleting or changing bad fitting items, and/or deleting bad fitting persons. Alternatively, one could consider using more general IRT models that offer many possibilities of obtaining model fit. General frameworks for multi-level and multi-dimensional IRT models are outlined in Skrondal and Rabe-Hesketh (2004) and De Boeck and Wilson (2004). In the specific context of genetic modelling, it might also occur that a particular subset of items show relatively high genetic correlations compared to the remaining items. In that case a more appropriate model would be an independent pathway model for categorical or ordinal traits (see for instance Van den Berg et al. 2006b). Good fit to a one-dimensional IRT model is empirical evidence that the observed item responses can be explained by one continuous underlying trait. When it further can be concluded that the scale is meaningful (based on item analysis and association with external measures to assess its validity), and the assumption of measurement invariance across different subpopulations is tenable (Lubke et al. 2004), the approach effectively deals with non-normal distributions of sum scores in for instance psychopathology (Van den Oord et al. 2003). Moreover, when the measurement model and the variance decomposition model are estimated simultaneously, the variance decomposition deals appropriately with the dependency in the data when estimating IRT model parameters and testing the model’s assumptions, and the IRT measurement model deals appropriately with the estimation of the heritability coefficient (correcting for attenuation to obtain an unbiased point estimate) and the reporting of the confidence intervals (correcting for location-dependent uncertainty of person scores on the latent trait). Our simulations showed the dramatic extent of the attenuation effect and the bias in estimating variance components due to imperfect measurement. Particularly when sum score distributions are skewed, underestimation of shared environmental effects and overestimation of dominance genetic effects may occur. The bias in variance components was also illustrated with an empirical data set: instead of finding a heritability estimate of 40% for attention problems with a sum score (Van den Berg et al. 2006c), a heritability estimate of 73% was obtained when including a measurement model and estimating it simultaneously with the variance decomposition model. This example provides an additional illustration of the bias in variance components due to the analysis of sum scores. However, it should be noted that model fit was not assessed, nor was the assumption of measurement invariance tested. This requires further study. The crucial element of the one-step approach that leads to unbiased point estimates is the inclusion of the appropriate probabilistic measurement model so that the estimation takes into account the unreliability of the measurement. The probabilistic modelling allows for the fact that twins with identical response patterns may have different scores on the latent trait, and also, that twins with non-identical response patterns may have exactly the same score on the latent trait. Discriminatory power of the items and the number of items are both crucial to the heritability estimated based on sum scores: the fewer the items and the worse the discrimination of the items (i.e., the smaller the variance of the latent trait in the one-parameter model; the smaller the factor loadings in the two-parameter model), the more biased the estimation will be when the analysis is performed on sum scores. High quality scales with a large number of items (say, more than 50) with high discriminatory power that are scattered across the entire scale can indeed be analysed with sum scores, but any other scale should be analysed using the IRT framework if one is interested in an unbiased heritability estimate with trustworthy confidence intervals. Future work should focus on the assessment of model fit in the context of genetic models. It is only sensible to apply a one-step IRT approach when the data actually conform to an IRT measurement model. If data do not fit an IRT model, for instance when there is differential item functioning across subpopulations, the approach will still lead to biased estimates. A crucial first step therefore is assessing model fit and checking measurement invariance.

J_Comput_Aided_Mol_Des-4-1-2311385

Recommendations for evaluation of computational methods

The field of computational chemistry, particularly as applied to drug design, has become increasingly important in terms of the practical application of predictive modeling to pharmaceutical research and development. Tools for exploiting protein structures or sets of ligands known to bind particular targets can be used for binding-mode prediction, virtual screening, and prediction of activity. A serious weakness within the field is a lack of standards with respect to quantitative evaluation of methods, data set preparation, and data set sharing. Our goal should be to report new methods or comparative evaluations of methods in a manner that supports decision making for practical applications. Here we propose a modest beginning, with recommendations for requirements on statistical reporting, requirements for data sharing, and best practices for benchmark preparation and usage. Introduction The field of computational chemistry, particularly as applied to drug design, has become increasingly important in terms of the practical application of predictive modeling to pharmaceutical research and development. Tools for exploiting protein structures or sets of ligands known to bind particular targets can be used for binding-mode prediction, virtual screening, and quantitative prediction of activity. A serious weakness within the field is a lack of standards with respect to statistical evaluation of methods, data set preparation, and data set sharing. Our goal should be to report new methods or comparative evaluations of methods in a manner that supports decision making for practical applications. In this editorial, we propose a modest beginning, with recommendations for requirements on statistical reporting, requirements for data sharing, and best practices for benchmark preparation and usage. There are two fundamental premises in making such a proposal. First, we must believe that the goal of reporting new methods or evaluations of existing methods is to communicate the likely real-world performance of the methods in practical application to the problems they are intended to solve. Ideally, the specific relationship between methodological advances and performance benefits will be clear in such reports. Second, we must understand that the utility of the methods of broad utility in pharmaceutical research application are predicting things that are not known at the time that the methods are applied. While this seems elementary, a substantial proportion of recent reports within the field run afoul of this observation in both subtle and unsubtle ways. Rejection of the first premise can reduce scientific reports to advertisements. Rejection (or just misunderstanding) the second premise can distort any conclusions as to practical utility. This special issue of the Journal of Computer-Aided Molecular Design includes eleven papers, each of which makes a detailed study of at least one aspect of methodological evaluation [1–11]. The papers collected within this issue make the detailed case for the recommendations that follow; the recommendations are intended to provide guidance to editorial boards and reviewers of work submitted for publication in our field. In surveying the eleven papers, we feel there are three main areas of concern: data sharing, preparation of datasets, and reporting of results. Concerns within each area relate to three main subfields of molecule modeling, i.e. virtual screening, pose prediction, and affinity estimation, and to whether protein structural information is used or not. We describe the issues in each area and then present recommendations drawn from the papers herein. Data sharing The issues Reports of new methods or evaluations of existing methods must include a commitment by the authors to make data publicly available except in cases where proprietary considerations prevent sharing. While the details are different across the spectrum of methods, the principle is the same: that sharing data promotes advancement of the field by ensuring study reproducibility and enhancing investigators’ ability to directly compare methods. However, the details of this matter a great deal, both for docking methods and for ligand-based methods. Docking will be used to briefly illustrate the problem. Many reports make claims of sharing data by, for example, providing a list of PDB codes for a set of protein–ligand complexes used in evaluating docking accuracy. In a very narrow sense, this might accommodate a notion of sharing. However, this is inadequate for four reasons:PDB structures do not contain all proton positions for proteins or ligands. Many docking approaches require all atoms, and nearly all require at least the positions of the polar protons. Without the precise protein structures used, in a widely used file format, it is not possible to reproduce the results of a report or make comparisons of other methods to those reported [7, 9, 11].Ligands within PDB structures do not contain bond order information and often do not even contain atom connectivity at all. Lacking this information, it is not possible to know what protonation state or tautomeric state was used to produce a particular result [4, 7–9].Docking methods have different sensitivities to input ligand geometries, both with respect to absolute pose and with respect to other aspects such as conformational strain and ring conformations. Since docking methods do not search ligand pose space exhaustively, absence of precise input ligand structures produces the same issue of reproduction and comparison as in (1) [4, 7–9].Different methods of protein structure preparation can yield subtle biases to different types of docking and scoring approaches. Very small changes in heavy atom or proton positions, as come with various relaxation strategies, can yield large changes in the positions of extrema for scoring functions. Provision of coordinates for all atoms allows other investigators to understand and differentiate the effects of methodology from the effects of protein structure preparation [4, 7–9]. Recommendations on data sharing Authors of reports on methodological advances or methods comparisons must provide usable primary data so that their results may be properly replicated and assessed by independent groups. By usable we mean in routinely parsable formats that include all atomic coordinates for proteins and ligands used as input to the methods subject to study. The commitment to share data should be made at the time of manuscript submission. Exceptions to this should only be made in cases where proprietary data sets are involved for a valid scientific purpose. The defense of such an exception should take the form of a parallel analysis of publicly available data in the report in order to show that the proprietary data were required to make the salient points [8]. Preparation of datasets The issues As stated earlier, the ultimate goal is predictions of things that we do not alreadyknow. For retrospective studies to be of value, the central issue is the relationship between the information available to a method (the input) to the information to be predicted (the output). If knowledge of the input creeps into the output either actively or passively, nominal test results may overestimate performance. Also, if the relationship between input and output in a test data set does not accurately reflect, in character or difficulty, the operational application of the method to be tested, the nominal reported performance might be unrelated to real world performance. Here, we will briefly frame the issue by discussing the differences between the operational use of methods and the construction of tests to measure and document their effectiveness for both protein structure, e.g. docking, and ligand-based methods in their areas of application. Docking Pose prediction. Here the goal is to prove that a method can predict how a ligand may bind, but not whether it can bind. In the operational case, we typically have a protein structure in complex with a ligand (or several such examples). We desire accurate prediction of poses for novel ligands that are potentially quite different from those whose bound structures are known. For method evaluation, the construction of prediction tests varies, but there are two basic forms: Cognate docking. The most common test of pose prediction involves a set of protein structures, each bound to a ligand, and with that ligand being the one to be tested. This represents the easiest form of the problem, since the conformation of the protein contains information pertinent to recovering the correct pose of the ligand. Most commonly, the protein coordinates are used as provided experimentally, with some variation in addition of protons, with the ligand in a randomized starting pose. Examples of information ‘leak’ include using of the cognate ligand pose as input [7], adding protons to the protein to favor the cognate pose [7, 9], choosing tautomer or charge states based on knowledge of the bound structure [8], and inappropriate use of bridging water molecules [9]. An extreme example would be optimizing the protein–ligand complex under the same scoring function used for docking, and then using this new, non-crystallographic information as the “test” data [7].Cross docking. The less common (but more relevant) formulation employs a protein structure with a bound ligand, but where the ligands to be predicted are different. The issue of similarity between the known ligand and the ligand being tested should be raised, but this is certainly more realistic, since the potential protein rearrangement from the apo form has been partially embedded in the structure but not optimized for each test ligand [7, 8]. Virtual screening. Predicting whether a ligand will bind, but not its affinity or its pose. In an operational application, we typically have a protein structure (or several), and we may have a few ligands known to bind a site of interest. The goal is to find novel ligands from some library of compounds. Operationally, we do not have the bound structures of the ligands we are trying to find, nor do we generally have a specific protein structure in which we are guaranteed a hospitable geometry. Many of the same mistakes that can be made with pose prediction can also be made to prefer known ligands over decoys, but there are additional hazards: The decoys do not form an adequate background [5–8, 10]. One of the frustrations in evaluating a study is how to judge the background against which a method is framed. It is very easy to generate a set of decoys that any method can tell apart from actives, and much more difficult to construct an informative collection.All the actives are chemically similar [2, 4, 5, 8, 10]. This is more relevant to ligand-based methods, but also applicable to docking because operationally finding chemically similar molecules as being potentially active is of little value in that these will likely be found by other methods. Scoring. Prediction of affinity is the hardest problem for molecular modeling and is as yet unsolved. In the operational case, we typically have multiple protein structures with ligands and may also have a wealth of structure-activity data for multiple ligand series. Frequently the problem here is accurately predicting the activity of what may be considered an obvious analog in virtual screening. We do not know the precise bound geometry of the specific ligand whose activity we are predicting. Affinity prediction tests can be done absent any affinity data on related analogs. However, to date, successful predictions without prior affinity information have been so anecdotal and untransferable that the field seems willing to accept any input of prior structural information. Hence, inclusion of information as to the protein’s disposition upon binding that is not available in an operational setting is considered acceptable.More typically, structural information and the activities of one or more closely related analogs are available. Here there are fairly regular reports of success, if given complete structural information. Chemical similarity is assumed, thus placing this technique in the domain of lead optimization, not lead discovery. As illustrated in at least one of the reports here [3], such methods are not currently successful when properly considered with control computations that include, for example, correlations of affinity with molecular weight. Ligand-based modeling Pose prediction. This is rarer than the use of ligand information in virtual screening but not operationally uncommon. The goal is to find the alignment of ligands to a protein using one or more known protein–ligand complexes. If the known and predicted ligands are one and the same, then issues from cognate ligand apply, for instance using torsions from the crystal structure, rather than deriving such information. If the known and test ligands are different, then the caveats from cross-docking apply, for instance are the test ligands diverse enough to make this experiment meaningful. Virtual screening. We have some number of ligands known to bind a particular site competitively, or, minimally, a single compound that exhibits a desired activity. The goal is to find novel ligands from some library of compounds. The incremental value of obvious analogs of known ligands is small as such would typically be found from SAR expansion from the known active (and is relevant in the narrow case of expanding hits after, for example, an HTS screen). Quite frequently, test cases are constructed where both the input ligands and testing ligands are all trivial analogs of a common central structure [2, 8, 11]. This stems, in part, from the simple fact that the ligands available for constructing tests are most frequently synthesized as part of a design process in which creating analogs is a useful exercise. However, such test cases do not reflect a key feature of the practical application in lead discovery: ligands that are obvious analogs of existing lead compounds will not exist in libraries to be screened for new leads.The relevant test cases are those in which the ligands to be retrieved are not analogs of the input ligands. This is, to a degree, a subjective issue. However, construction of such cases can be done, for example, by choosing input ligands that were discovered long before the test ligands or by choosing input ligands that have substantially different overall biological properties (e.g. side effects) than the test ligands [2]. Scoring. Predicting affinities of ligands from the affinity of one or more ligands, whether relative or absolute. In practice, we generally have structure-activity data for multiple ligand series. Frequently the problem here is accurately predicting the activity of what would be considered an obvious analog in virtual screening. We do not generally know the bound geometry of the specific ligand whose activity is to be predicted. This methodological area of QSAR has its own set of relatively well-understood foibles and is not addressed in detail in this issue. The descriptions of test case construction above involve different degrees of challenge in proportion to the amount of information provided to a method. The problems often encountered in reviewing or reading papers is that methods claim a lower level of information concerning the answers than is actually true. This is seldom intentional, no matter the provocation to believe otherwise, but a reflection of the difficulty in preparing a ‘clean’ test. Recommendations on dataset preparation Protein structure selection and preparation. Protein structure selection should take into account more than just the nominal resolution [4, 5, 9]. There are other measures such as coordinate precision that are more appropriate but require the use of structures where an R and Rfree are reported. In addition, checking to see if density actually exists for the poses being predicted is suggested, although this requires structure factors to have been deposited along with protein coordinates.Protein structure optimization must not be done by making use of the known geometry of the ligand that is the subject of a prediction [5, 7]. At most, selection of sensible protonation states, tautomers, and rotamers of ambiguous or underspecified groups should be done one time for each protein structure. Much fuller disclosure of preparation procedures is required than is typically seen.The most relevant tests of methods will employ proteins whose structure was determined with a ligand other than the one being predicted or a close analog thereof [8].The number and diversity of protein targets needs to be sufficient to enable to draw statistically robust conclusions [4, 6, 10, 11]. Some typical targets (e.g. HIV protease) are quite atypical [4] and in small datasets may dominate results [10, 11]. Decoy set construction. There is clearly a consensus that decoy sets can have a significant impact on results [4–8, 10, 11]. The contributed papers here provide no clear consensus as to what constitutes an acceptable set of decoys, although there are lessons as to what not to do, for instance using molecules that might actually be actives, or have unusual properties compared to known actives. At present, the best suggestions seem to be to make decoys relatively ‘drug-like’, so as to mimic real, i.e. operational screens. We also recommend the practice of employing multiple decoy sets and including those developed by other investigators to facilitate study comparison and collation. Active ligand set construction. There is consensus that the degree of “obvious similarity” among actives has important effects, particularly in evaluating ligand-based methods [1, 2, 4, 7], but there is less agreement on how to either measure this or to control for it. Our recommendation is that such effects should be quantified in reports, where possible, by, for example, using 2D similarity methods to provide a baseline for the difficulty of a retrieval task or to provide a numerical characterization of the diversity of active ligand sets [2]. In addition, suggestions are made in this issue to either use only single representatives of a chemical class or to weight each according to its order of discovery [1, 4, 6]. Both ideas seem eminently worth further evaluation. Ligand preparation. All ligands (whether active or decoys) must be prepared using automated procedures that are unbiased and which will not yield systematic differences between populations of molecules that will generate a systematic performance bias [7–9]. For instance, assign protonation states of ligands and decoys by the same protocol, and generate conformations from just connectivity records of both ligands and decoys. Parameter tuning. Many papers in this issue show how the choice of parameters influences the apparent quality of results [3, 4, 9]. There is a dichotomy of opinion on whether “tuned” performance figures are relevant to future application of a method when the correct answer is unknown. Our recommendation is that if one chooses to report tuned performance, one must also report performance using standard parameters. Even within the constraints outlined above, data set preparation and parameter selection can yield a wide range of results. This is acceptable to illuminate which choices are of most benefit to users of the different methods. However, without strong requirements for data sharing (the subject of the previous section), this benefit will be diluted. Further, without baseline requirements for statistical reporting (the subject of the next section), this diversity will lead to an unacceptable degree of incomparability between different reports. Reporting results The issues The issues surrounding what to report are substantially in dispute, and this has led to an alarming inability to compare multiple studies, except in the case where all primary data are available and where one is willing to make an independent analysis. Here there seem to be two schools of thought. The first is that molecular modeling is a special enterprise, distinct and different from other efforts at prediction. As such it is seen as a part of the process to select or invent measures that illustrate a particular point. The second school holds that molecular modeling is in fact similar to many other areas of science and commerce and that by ignoring standard practices in other, more established, fields, we do a disservice to modeling.Pose prediction. The almost universal measure for pose prediction is RMS, i.e. the root-mean-square difference between heavy atom positions seen in crystallographic refinement and predicted by a method, generally corrected to allow for internal symmetries within the ligand in question [8]. What is at issue is the manner in which RMS is reported. The desire, as with enrichment metrics, is for a single value to capture the performance of a method over a collection of test cases. The most commonly reported is the proportion of successes at some particular threshold of RMS (for instance, an arbitrary 2.0 Å RMS), but a number of investigators report average RMS instead. Neither is ideal, but mean RMS is less useful for two reasons. First, it can be skewed by small numbers of poor poses (each with very large RMS) [5]. Second, its magnitude can be directly manipulated by clever choice of poses against which to measure success [5, 7].Virtual screening. In many senses, this is the most disputatious area. The standard measure has been “enrichment” defined to be the ratio of the observed fraction of active compounds in the top few percent of a virtual screen to that expected by random selection. The reason enrichment is so prevalent is that it is synonymous with the purpose of virtual screening: to enable the selection of a subset of compounds with improved chances of drug discovery. However, by nearly all other considerations it is a poor measure. Most regrettable is its dependence on the ratio of actives to inactives, which makes enrichment a property of a method and an experimental set-up rather than an intrinsic property of the method [10]. A number of metrics have been proposed, many of which share this clearly undesirable quality [1, 7, 8, 10].Affinity estimation. Ideally the analysis of a prediction of affinity ought to be the simplest of tasks. Given a set of experimental values and a predicted set, merely calculate the average difference. If modeling could actually predict affinities, this might be a reasonable approach. As it is, the best generally hoped for is a rough correlation between activity and score and even in these cases there are obvious, and not so obvious, pitfalls. However, a rough correlation between activity and score is frequently obtained simply by equating activity with, for example, a monotonic function of molecular weight [3].General. There are other more subtle issues. One is the presentation of results where the answers have fed back to the input (training/test set contamination). This is generally easy to spot and usually means a method is without merit. More subtle errors tend to be where forms for cross-validation are followed (proper separation into training and test systems), but where the true independence of these two sets is never called into question [10]. If the test set is not sufficiently different to the training set then there is no assurance against over-parameterized approaches. Finally, reports that profess to predict affinities seldom provide some reliable estimate of experimental affinity. The practice of combining results from multiple experiments is only acceptable if experimental conditions are similar. Anecdotal stories abound of different labs within the same company failing to be able to reproduce each other’s binding affinities, often with difference of an order of magnitude or more. It seems sheer folly to think a test set from truly heterogonous sources can be called reliable. Recommendations for reporting results Pose prediction. Success rates using multiple RMS thresholds should be reported. At a minimum we recommend 3.0, 2.0, 1.0, and 0.5 Å. In fact, we encourage investigators to report full cumulative histograms of RMS performance for both top scoring and best-predicted poses. This will generally take very little additional space in a report, and it provides much more information to the reader. For example, if there are a large proportion of reported RMS values that appear to have greater precision than the experiment, this is detectable by inspection of the histogram [5, 7]. Statistically it is not impossible in a fair prediction for a measurement to be within, say, 0.1 Å of an experimental measurement that is only accurate to 0.5 Å, but it is unlikely. We also suggest that if an estimate of the precision of the experimental coordinates is available that it must be reported. This, then, provides an excellent bulwark against over-fitting to the known results. Virtual screening. Based on multiple reports in this issue, we recommend reporting the area under the curve for ROC plots (AUC) [1, 3, 7–10]. These have for a long time been a standard metric for other fields and for good reasons. The argument against using AUC values to judge methods is that they are global measures, i.e. reflect the performance throughout a ranked list. Thus, the notion of “early enrichment” may not be well characterized by just AUC, particularly when virtual screening methods yield AUC values short of the 0.8–1.0 range. Therefore we make two suggestions. First, enrichment percentages should be reported at the following four values: 0.5%, 1%, 2%, and 5%. Second, that a formulation of enrichment is used that reports the ratio of true positive rates (the Y axis in an ROC plot) to the false positive rates of 0.5%, 1%, 2%, and 5% (found on the X axis in an ROC plot). Thus “enrichment at 1%” becomes the fraction of actives seen along with the top 1% of known decoys (multiplied by 100). This removes the dependence on the ratio of actives and inactives and directly quantifies early enrichment. It also makes standard statistical analysis of error bars much simpler [10]. Affinity estimation. First, standard correlation measures must be reported. We recommend Pearson’s correlation (due to its intuitive appeal and ubiquity) as well as Kendall’s Tau (due to its robustness in cases where Pearson’s correlation can yield spurious values). Both are easy to calculate, and errors for both are simple to compute. Second, we recommend that papers claiming a correlation with affinity ought to also present the correlations achieved with simpler measures, to include molecular weight, cLogP, and hydrogen bond donor/acceptor counts [3, 8]. Thirdly, authors must be held responsible for realistic estimates of the accuracy of experimental affinities, in particular when such results are from heterogeneous sources. General. First, if data and dataset preparation are completely disclosed, then the issue of the precise manner of reporting in a paper becomes less vital. Authors may choose to emphasize whatever measures they wish but interested readers should be able to construct alternate measures. Secondly, the most lamentable aspect of reporting in our field is the lack of error bars on reported metrics and of the quantification of statistical significance more generally. This is the single simplest, most effective, and most needed reform that an editor can insist upon and that a reviewer should look for. Multiple papers here suggest approaches that should be applied [1, 5, 7, 10]. There can be no excuse for a paper on a modeling method to be published claiming one method is superior to another without proper statistical validation. Finally, we hold to the aforementioned second school of thought i.e. that molecular modeling should be held to the same standards as other fields. As such, our most general recommendation is to report standard metrics as a requirement and alternates as desired by authors. Conclusions Molecular modeling is a relatively young field. As such, its growing pains include the slow development of standards. Our hope for this special issue of JCAMD is that with the help of the arguments made in the contributed papers, the modest recommendations made here will form the kernel of standards that will help us as a community to both improve the methods we develop and to reduce the disparity between reported performance and operational performance.

Dig_Dis_Sci-3-1-1914222

Alterations in Epithelial and Mesenchymal Intestinal Gene Expression During Doxorubicin-Induced Mucositis in Mice

In the current study we aimed to gain insight into epithelial-mesenchymal cross-talk and progenitor compartment modulation during doxorubicin (DOX)-induced mucositis in mice. Intestinal segments were collected on various days after DOX treatment. DOX-induced damage at day 1–2 was characterized by increased epithelial proliferation and apoptosis and a decrease in the expression of epithelial differentiation markers. Concurrently, T-cell factor-4 (TCF4) levels increased and the epithelial differentiation enhancing factor, bone morphogenic protein-4 (BMP4), decreased. During severe damage (day 3), BMP4 levels were significantly increased, which inversely correlated with epithelial proliferation. At the same time, the expression of the epithelial differentiation markers was increasing again. At day 7, BMP4 levels were down-regulated, while the levels of the epithelial differentiation markers and TCF4 were normalized again. These data suggest that in response to DOX-induced damage, BMP4 and TCF4 are modulated in such a way that homeostasis of the progenitor compartment is partly preserved. Introduction The highly proliferating small intestine is very susceptible to chemotherapy-induced damage [31, 47]. This side effect, often referred to as mucositis, is very painful and can be life threatening due to the enhanced risk of bacterial translocation caused by loss of epithelial barrier function. Each year, around 500,000 patients worldwide suffer from this side effect [7] for which there is still no definitive prophylaxis or treatment. Drug-induced damage to the highly proliferating stem cells and progenitors located in the crypts of Lieberkühn showed that different classes of cytostatic drugs affect epithelial crypt cells of different topographical and hierarchical status [12, 13, 24]. Methotrexate (MTX), 5-fluorouracil and vincristine for example, induce damage high in the proliferative compartment at crypt position 9–11. Damage induced by MTX is well characterized and shows severe morphological changes characterized by epithelial flattening, villus-atrophy, and specific de-differentiation of enterocytes. The last one is indicated by a decreased expression of the enterocyte-specific enzymes but a maintained expression of goblet and Paneth cell-specific genes [29, 38, 41, 45, 46]. Doxorubicin (DOX) is a cytostatic drug that also frequently causes severe mucositis. However, DOX-induced damage is not well characterized [28]. DOX preferentially attacks cells at crypt position 4–6, which is just above or at the same level as the intestinal stem cells [12, 32]. DOX is therefore thought to induce more severe damage, especially to the proliferating compartment at the bottom of the crypts, as for example MTX. In clinical practice, DOX is often used in treatment of solid tumors, leukemia, and lymphomas in both adult and childhood cancer patients [16, 18, 21, 28]. Recent studies revealed a dynamic cascade of events leading to chemotherapy-induced mucositis [35]. Not only are the specialized epithelial cells affected by cytostatic-drug treatment, but also the underlying submucosal connective tissue. Under normal physiological circumstances, the epithelium maintains a cross-talk with the mesenchyme. Epithelial-mesenchymal interactions are critical for the normal morphogenesis and maintenance of the crypt-villus axis [15, 23]. Bone morphogenic protein (BMP)- and Wnt-signaling pathways are two of the main pathways involved in embryonic and adult intestinal development [5, 33]. BMPs are morphogenes that constitute a large group of structurally and functionally related proteins of the TGF-beta superfamily. BMP signaling pathways have a key role in organogenesis [2, 37], gastrointestinal development and intestinal homeostasis in adults [8–11, 14, 27]. After intestinal development, BMP4 is exclusively expressed in the intravillus and intracrypt mesenchymal cells, including those adjacent to the intestinal stem cells [8, 10]. Paracrine BMP signaling occurs specifically in the villus from the mesenchyme to the adjacent epithelium [8], suggesting involvement in epithelial-mesenchymal interactions as a mesenchymal signaling molecule. The Wnt-signaling pathway also has been implicated in the regulation of the intestinal epithelial proliferation/differentiation balance in vitro [43]. In the intestine of mice deficient for transcription factor TCF4, the main Wnt pathway transcription factor in the intestinal epithelium, loss of proliferative compartments and epithelial cell differentiation are found [17]. Interestingly, the BMP- and Wnt pathways appear to be linked, as was shown by the fact that BMP signaling suppresses Wnt signaling to ensure a balanced control of stem cell proliferation and subsequent epithelial differentiation [8, 10]. The objective of this study was to develop an experimental mucositis mouse model to characterize DOX-induced intestinal damage and subsequent repair. In addition, we aimed to correlate the alterations in morphology, epithelial homeostasis, and gene expression with changes in BMP4 and TCF4 expression. This, in order to gain insight into possible modulation of the epithelial-mesenchymal cross-talk and progenitor compartment during chemotherapy-induced intestinal damage and regeneration. Materials and methods Animals Animal experiments were performed with permission of the Animal Ethics Committee of the Erasmus MC-Sophia. Upon arrival at our institute, 10-week-old male BALB/c mice (Harlan, Horst, The Netherlands) were housed individually during the whole experiment in micro-isolator cages under specific pathogen-free conditions with free access to a standard palletized diet (Hope farms, Woerden, The Netherlands) and water. After 1 week of adjustment to the new environment, the mice were divided into three groups and injected intravenously with doxorubicin (DOX) (Doxorubicin, Pharma Chemie, Haarlem, The Netherlands) on two subsequent days. At day −1 and 0, the first group of mice was injected with a low dose of DOX of 6 and 4 mg/kg (low dose) respectively, a second group was injected with a medium dose of 8 and 5 mg/kg (medium dose) and a third group was injected with a high dose of 10 and 6 mg/kg (high dose). Controls were given equivalent volumes of 0.9% NaCl. Mice in the low- and high-dose group were sacrificed at day 1, 2, 3 and 7 after the final DOX injection; mice in the medium dose group were only sacrificed at day 3 and 7. One hour before sacrifice, the mice were injected with 120 μl 10 mg/ml 5- Bromo-2’deoxyUridine (BrdU) (Sigma-Aldrich, Zwijndrecht, The Netherlands), an uridine analog, to locate the proliferating cells. Per time point 4–6 DOX-treated animals and 2–4 control animals were sacrificed. Segments of mid-jejunum were collected and either processed immediately for histological analyses or snap-frozen in liquid nitrogen for storage at −80°C and subsequent protein isolation. Histochemistry Five-millimeter segments of mid-jejunum were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), dehydrated and embedded in Paraplast Plus (Sherwood Medical, Den Bosch, The Netherlands) as previously described [45]. Four μm sections were routinely stained with hematoxylin (Vector Laboratories, Burlingame, CA) and eosin (Sigma-Aldrich) to study morphological alterations of the crypts and villi. Immunohistochemistry was performed as described previously [45] with some minor modifications. The sections for BrdU staining required an extra adjustment to this protocol of HCL incubation, washing with borate buffer, and pepsin treatment as described before [34]. In short, the sections were blocked as described and incubated overnight with the following antibodies diluted in PBS: to visualize BrdU incorporation, mouse monoclonal anti-BrdU (1:250, Roche Applied Sciences, Indianapolis, IN) was used, as an enterocyte marker rabbit polyclonal anti-rat Sucrase-Isomaltase (SI) (1:9000 in PBS, kindly provided by Dr. K.Y. Yeh [49]) was used and as a goblet cell-specific marker rabbit polyclonal anti-rat trefoil factor family (TFF3: 1:3000, kindly provided by Prof. Dr. D.K. Podolsky) was used. Furthermore, BMP expression was visualized with anti-BMP4 (1:100, R&D Systems, Abingdon, UK). Immunoreactions were detected using Vectastain ABC Elite Kit (Vector Laboratories, Burlingame, CA) and 3,3′-diaminobenzidine tetrahydrochloride (Sigma-Aldrich, Zwijndrecht, The Netherlands). Crypt and -villus length Longitudinal sections of crypts and their corresponding villi were selected so that the base (marked by Paneth cells), middle and top of the crypt were all in the plane of section and thus well orientated. The depth of ten crypts and the length of ten villi were measured on three slides per animal, four animals per time point, with the use of a Nikon Eclipse E800 microscope and IM 500 software. Protein dot blotting The expression of enterocyte markers was detected and quantified as described previously [46]. Briefly, 5-mm segments of the mid-jejunum were homogenized and protein concentrations were measured using the BCA Protein Assay Reagent (Pierce, Rockford, IL) and 50 μg protein of each homogenate was dot-blotted on nitrocellulose (Protran BA83, 0.2 μm; Schleicher & Schuell, Dassel, Germany). Hereafter, blots were blocked for 1 h with blocking buffer containing 50 mM Tris, pH 7.8, 5% (wt/vol) nonfat dry milk powder (Campina Melkunie, Eindhoven, The Netherlands), 2 mM CaCl2, 0.05% (vol/vol) Nonidet P40 (BDH, Brunschwig Chemie, Amsterdam, The Netherlands) and 0.01% Antifoam B (Sigma-Aldrich). Blots were incubated overnight at 4°C with rabbit polyclonal anti-rat SI (1:1000 [49]), and rabbit polyclonal anti-rat trefoil factor family (TFF3: 1:1500) diluted in blocking buffer. After washing with blocking, the buffer blots were incubated with 125I-labeled protein A (specific activity 30 mCi/mg, Amersham Biosciences, Roosendaal, The Netherlands) for 2 h at room temperature. Specific binding of 125I-labeled protein A to the enterocyte marker antibodies was measured using PhosphorImager detection. The elicited signal was quantified by ImageQuant software (Molecular Dynamics, B&L systems, Zoetermeer, The Netherlands) and the expression of TFF3 and SI was expressed per 50 μg protein of tissue. Average expression levels of TFF3 and SI in the mid-jejunum were calculated per mouse, followed by calculation of the mean expression of TFF3 and SI per time point studied. Subsequently, the average expression of TFF3 and SI of control mice was set at 100%. The specificity of the above-described antibodies was previously confirmed by Western-blot analysis [46]. Western-blot analysis The same protein homogenate was used as described for protein dot blot analysis. Twenty μg of protein was loaded per lane and run on a 12.5% SDS-PAGE. The separated proteins were transferred to nitrocellulose membranes (Protran BA83, 0.2 μm) and blocked for 1 h at room temperature in blocking buffer as described above. The blots were incubated overnight at 4°C with primary antibodies diluted in blocking buffer: mouse monoclonal anti-human PCNA, clone PC10 (1:250, Novo Castra Laboratories, Newcastle upon Tyne, UK) and mouse monoclonal anti-human BMP4, clone 3H2 (1:100), (Novocastra Laboratories, Newcastle upon Tyne, UK), rabbit polyclonal anti-human cleaved Caspase-3 antibody (1:1000, Cell Signaling, Beverly, MA), and mouse monoclonal anti-human TCF-4, clone 6H5-3 (1:250, Upstate, Waltham, MA). After washing with PBS, 0.2% Tween-20 blots bound antibodies were revealed using HRP conjugated goat anti-mouse or rabbit anti-goat secondary antibodies (1:1000) and SuperSignal West Femto Luminol Enhancer kit (Pierce, Rockford, IL). The signal was detected and quantified by the ChemiGenius gel documentation system (Syngene, Cambridge, UK) and the expression of the specific proteins analyzed was expressed per 20 μg protein of tissue. Average expression levels of PCNA, BMP4, TCF4, and caspase-3 in the mid-jejunum were calculated per mouse, followed by calculation of the mean expression of these specific proteins per time point studied. Subsequently, the average expression of PCNA, BMP4, TCF4, and caspase-3 in control mice was set at 100%. Statistical analysis Changes in protein expression levels during damage and regeneration were statistically analyzed using the Kruskal-Wallis H-test and the Mann-Whitney U-test. A p < 0.05 was considered statistically significant. Data are presented as the mean ± standard error of the mean (SEM). Results Dose-response analysis of DOX-induced mucositis in an experimental mouse model To optimize the dose of DOX necessary to induce severe intestinal damage (i.e., villus atrophy, crypt loss and flattening of the epithelial cells) a dose-response curve was performed. Thereto, mice were divided in three treatment groups: low dose, medium dose, and high dose (see Materials and methods, section Animals for details). In the high-dose treatment group, two of the six mice died at day 4. Necropsy showed an excess of fluid in the abdominal cavity of unknown source. Because of the elapsed time after death, morphological evaluation could not be performed. Morphological analysis of the low-dose treatment group at days 1 and 2 (data not shown), revealed only a slight increase in crypt length, which became more pronounced at day 3 (Fig. 1B). An increase in crypt length was also seen in the medium dose group at day 3. (Note the medium-dose treatment group was not studied at days 1 and 2). No other morphological changes were seen in the low dose- (days 1–3) and medium-dose group (day 3). In contrast, in the high-dose treatment group we did see severe morphological damage (Fig. 1D–F). Specifically, severe villus atrophy, crypt lengthening, crypt loss, perturbation of crypt arrangement, flattening of crypt and villus epithelium and inflammation were observed at day 3 (Fig. 1F). Moreover, the latter morphological damage was already seen, although in a milder degree, at days 1 and 2 after the high-dose DOX treatment (Fig. 1D, day 1 and Fig. 1E, day 2). At day 7, the intestinal morphology was completely restored in the low-dose and medium-dose group (data not shown), but was still affected in the high-dose group. As in the latter group, the crypt epithelium still showed signs of regeneration, like new crypt formation and less epithelial flattening (Fig. 1G). Quantitative analysis of crypt and villus length To quantify the amount of morphological damage induced by the different doses of DOX, the length of the crypts and villi were measured (Fig. 1H–I). In the high-dose group, but not in the low-dose or medium-dose group, a significant increase in crypt length was already observed at day 2 (data not shown). At day 3 the crypt length was significantly increased in each treatment group compared to the control group (Fig. 1H). By day 7, the length of the crypts had returned to control levels in the low-dose and medium-dose group, but were still increased in the high-dose group. This increase in crypt length, however, showed a trend towards normal levels because the crypt length at day 3 was 41% above control level but at day 7 this was still only 13% above control level. The length of the villi did not change in the low-dose or the medium-dose treatment group on each day investigated. However, in the high-dose treatment group the length of the villi were significantly decreased on days 1–3 and 7 (Fig. 1H, day 3 and 1I, day 7). Because severe intestinal mucositis (day 3) and morphological regeneration (day 7) were seen only in the high-dose treatment group, we continued our studies with the high-dose DOX treatment model. Effects of high-dose DOX treatment on enterocyte-specific gene expression To gain insight into the functional capacity of the intestinal epithelium after DOX treatment, a sucrase-isomaltase (SI) immunohistochemical staining was performed. SI is an enterocyte-specific disaccharidase responsible for sucrose degradation. SI is expressed in the brush border of differentiated villus enterocytes directly after weaning [42] and is considered an intestinal epithelial differentiation marker. At day 1 and 2 after DOX treatment SI was expressed in the brush border of enterocytes along the entire villi, comparable to the control epithelium (Fig. 2A, control; B, day 1; C, day 2). At day 3, SI staining showed a patchy pattern with weak staining on less affected parts of the epithelium and even absence of staining on severely damaged epithelial parts of the sections (Fig. 2D). At day 7, during the regenerative phase, the SI staining pattern was comparable to control again (Fig. 2E). To quantify SI protein expression, protein dot-blot analyses were performed using SI-specific antibodies (Fig. 2F). The SI protein expression level at day 1 was similar to the control situation. At day 2, SI expression levels were significantly decreased and remained significantly decreased at day 3. At day 7, the SI expression levels regained to control levels. Effects of DOX treatment on goblet cell-specific gene expression Goblet cell-specific gene expression after high-dose DOX treatment was analyzed by the expression of trefoil factor family 3 (TFF3). TFF3 is a bioactive peptide, produced and secreted by goblet cells, that is involved in epithelial protection and repair [6, 22]. Immunohistochemical staining of TFF3 showed TFF3 expression by goblet cells in crypts and villi of the jejunum (Fig. 3). At day 1, the TFF3 staining was not visibly altered compared to the control situation (Fig. 3B). At day 2, the immunohistochemical detection of TFF3 decreased visibly, especially in the goblet cells localized in the crypts and lower part of the villi where TFF3 protein could hardly be detected/was absent (Fig. 3C). At day 3, during most severe morphological damage, TFF3 staining by goblet cells in the crypt was reconstituted (Fig. 3D). Day 7 showed complete regeneration of TFF3 (Fig. 3E). Subsequently, a protein dot-blot analysis was performed (Fig. 3F) using TFF3-specific antibodies to quantify TFF3 levels. At day 1, the TFF3 expression was already significantly decreased to 50% of control level and maintained at this low level at day 2. At day 3, concomitant with the most severe morphological damage, TFF3 expression returned to control level. TFF3 protein expression was again comparable to control situation at day 7. Effect of high-dose DOX treatment on epithelial proliferation Localization of proliferation was studied by immunohistochemical detection of incorporated BrdU (Fig. 4). In controls (Fig. 4A) BrdU was mainly localized in cells from the bottom of the crypt up to three quarter of the crypt length. At day 1 and 2 the proliferative zone broadened and progressively moved upwards to the crypt-villus junction. At day 3, during severe morphological damage, BrdU-positive cells formed a scattered pattern throughout the remaining crypt structures of the small intestine. In the regenerative phase, at day 7 BrdU-positive cells were located in the lower part of the crypts again.Fig. 1Morphology of the murine small intestine after low, medium- or high-dose DOX treatment. Morphology of the jejunum of a control mouse (A), at day 3 after low-dose DOX (B) and at day 3 after medium-dose DOX (C), both were mildly affected by DOX treatment. Morphology of the jejunum at day 1 (D) and day 2 (E) after high-dose DOX treatment, which appeared progressively affected and severely deteriorated at day 3 (E). At day 7, the regenerative phase, the intestinal morphology of the mice in the low- and medium-dose group was completely restored to control situation (data not shown). The morphology at day 7 of the mice treated in the high-dose group was regenerating, showing new crypt formation but still villus atrophy. These photographs are representative examples of a group of four animals per time-point. To quantify the effect of these different doses of DOX treatment, crypt and villus lengths were measured. Crypt and villus lengths of control mice were set at 100%. At day 3 (A), the length of the crypts in all three DOX treatment groups was significantly increased in comparison to the control group. Villus length in the low- and medium-dose group remained at control level. The length of the villi of the high-dose group decreased significantly, indicating significant villus atrophy after high-dose DOX treatment. At day 7 (B) crypt and villus lengths of the low- and medium-dose DOX treatment groups showed complete restoration. The crypts of the high-dose treatment group were still significantly longer, but this increase in crypt length was declining compared to day 3 (day 3 41%, day 7 13%). The length of the villi was still significantly decreased at day 7, showing no signs of regeneration. *P < 0.05 versus control. The bars are expressed in mean + SEMFig. 2Effect of high-dose DOX treatment on the enterocyte-specific SI expression in the small intestinal epithelium. SI expression of the jejunum of a control mouse (A), at day 1 (B) and at day 2 (C). The brush border staining at day 1 and 2 is comparable to control situation. At day 3 (D), SI staining showed a patchy pattern; some parts of the brush border showed SI expression, on some other parts of the brush border SI expression was absent (→). SI expression at day 7 (E) showed complete regeneration, and was equal to the control situation. SI protein expression levels were analyzed by dot-blot technique and the SI expression of control mice was set at 100% (F). The expression of SI remained stable at day 1, but decreased significantly at day 2 and 3. During the regenerative phase, at day 7, the SI expression reached control level again. *P < 0.05 day 2 and 3 versus control. The bars are expressed in mean + SEMFig. 3Effect of high-dose DOX treatment on the goblet-specific TFF3 expression in the small intestine. Immunohistochemical staining of the goblet cell-specific protein TFF3 (A–E). TFF3 expression of the jejunum of a control mouse (A), at day 1 (B), at day 2 (C), at day 3 (D) and at day 7 (E). In the control situation, TFF3 was expressed in all goblet cells in the crypts and villi, at day 1 (B), the amount of goblet cells expressing TFF3 maintained at control level. At day 2 (C), TFF3 expression decreased visibly, the goblet cells localized in the crypts showed no staining. At day 3 (D), during severe morphological damage, TFF3 expression of goblet cells in the crypt was reconstituted. TFF3 expression at day 7 (E) was regenerated and comparable to control situation. TFF3 protein expression levels were analyzed by dot-blot technique, the TFF3 expression of control mice was set at 100% (F). TFF3 expression decreased significantly at day 1 and day 2, the expression was decreased to 50% of control level. During severe morphological damage at day 3, the TFF3 expression increased almost to control level. Day 7 showed TFF3 expression even above control level. *P < 0.05 day 1 and 2 versus control. The bars are expressed in mean + SEMFig. 4Effect of high-dose DOX treatment on epithelial proliferation. The localization of epithelial proliferation was visualized by detection of BrdU incorporation. BrdU incorporation of the jejunum of a control mouse (A) and mice at day 1 (B), at day 2 (C), at day 3 (D) and at day 7 (E). At day 1, the BrdU-positive cells were slightly migrated upwards in the crypt, at day 2 the BrdU-positive cells were seen along the entire length of the crypts. At day 3 the BrdU expression decreased and showed a scattered pattern on some crypts and villi. The location of BrdU-positive cells at day 7 was confined again to the lower part of the crypts. PCNA, Caspase-3, TCF4 and BMP4 protein expression levels were analyzed by Western-blot technique (F), the PCNA expression of control mice was set at 100% (G). The expression of PCNA protein showed a non-significant trend of increased expression at day 1, and a significant increase at day 2. PCNA expression at day 3 and 7 showed a decreasing trend to around 20% of control level. *P < 0.05 day 1 and 2 versus control. The bars are expressed in mean + SEM The changes in proliferation induced by DOX were quantified by Western-blot analysis of PCNA protein expression [4, 40] using a specific monoclonal PCNA antibody. PCNA protein expression after MTX treatment (Fig. 4F, G) showed a trend of increased expression at days 1. At day 2, this increased expression was significant compared to control levels, indicating that high DOX treatment did not induce an arrest in proliferation at day 1 and 2. At day 3, PCNA expression was decreased to around 20% of the normal level and was still 50% of control level at day 7. The latter data indicate that the inhibition of epithelial proliferation sustained during the regenerative phase. Overall, the PCNA levels correlated with the immunohistochemical data (Figs. 4A–E vs. 4G) Effect of high-dose DOX treatment on enterocyte apoptosis Besides proliferation, the influence of DOX treatment on apoptosis was quantified with a polyclonal antibody against cleaved Caspase-3 (Fig. 4F, 5A). Caspase-3 is one of the key executioners of apoptosis [26]. The expression of cleaved Caspase-3 protein was significantly increased at day 1 and 2, at the beginning of DOX-induced morphological damage. At day 3, during severe morphological damage, and day 7, during regeneration, expression of cleaved caspase-3 was comparable to control level. Increased apoptosis at day 1 and 2 was primarily located along the crypt axis (Fig. 5B showing day 1 after MTX treatment))Fig. 5Effect of high-dose DOX treatment on epithelial apoptosis. Caspase-3 protein expression of the control mice, analyzed by Western-blot technique (Fig  4F), was set at 100% (A). At day 1 and 2, the Caspase-3 levels increased significantly in comparison to the control levels. At day 3, during severe morphological damage the caspase-3 expression maintained at control level, the same was seen at day 7, during the regenerative phase. The increase in apoptosis at day 1 is primarily located in the crypt region (B). *P < 0.05 day 1 and 2 versus control. The bars are expressed in mean + SEM. → indicates apoptotic cells Effect of high-dose DOX treatment on TCF-4 protein expression To gain insight into the effects of DOX on the epithelial stem cell compartment, the expression of TCF-4 was quantified. TCF-4 is a transcription factor of the Wnt-signaling pathway and is expressed in the gut epithelium in a gradient fashion that is highest at the base of the crypts [1, 19]. TCF-4 −/− mice lose the intestinal epithelial progenitor and stem cell population before crypt formation can be established [17]. Therefore, TCF-4 has a role in intestinal epithelial stem cell maintenance. The TCF-4 expression (Figs. 4F and 6) remained very stable despite DOX treatment. Only at day 1 the TCF-4 expression increased significantly, although little in absolute sense. But at day 2, day 3 and day 7, the TCF4 protein expression remained at control level.Fig. 6Effect of high-dose DOX treatment on TCF-4 protein expression. TCF-4 protein expression of control mice, analyzed by Western-blot technique (Fig. 4F), was set at 100%. At day 1, TCF-4 protein expression increased significantly. At day 2, day 3, and day 7, the TCF-4 protein expression remained stable at control level. *P < 0.05 day 1 versus control. The bars are expressed in mean + SEM Effect of high-dose DOX treatment on expression of Bone morphogene 4 (BMP4) To examine the role of BMP4 signaling in intestinal homeostasis the expression of BMP4 was quantified. BMP4 is crucial for epithelial-mesenchymal cross-talk and intestinal homeostasis. BMP4 is expressed in the intravillus and intracrypt mesenchymal cells [8, 10], inhibition of BMP signaling causes intestinal architectural abnormalities [8] and decreased differentiation as seen in tumorigenesis [27]. BMP4 protein expression showed a mild decrease in expression level at day 1 after DOX treatment (Figs. 4F, 7A). At day 2, this decrease in BMP4 expression level progressed and was almost significant (P=0.05) in comparison to the control level. Thereafter, BMP4 day 2 but not to control. Day 7 showed a non-significant decrease in expression compared to control mice. BMP4 was predominantly expressed in the intravillus mesenchym as shown by immunohistochemistry (Fig. 7B, C), both in control as in MTX-treated mice.Fig. 7Effect of high-dose DOX treatment on expression of the morphogen BMP4. The BMP4 protein expression of the control mice, analyzed by Western-blot technique (Fig. 4F), was set at 100% (A). Expression of the morphogen BMP4 protein showed a decreasing trend at day 1, which was at the brink of significance (P=0.05) at day 2 in comparison to control expression. At day 3, the BMP4 expression showed an increasing trend. This increase in expression was significant in comparison to day 2. Day 7 showed a decrease in expression again beneath the level of control expression. The localization of BMP4 expression did not change at day three (C) in comparison to control situation (B). *P < 0.05 day 2 versus control, **P < 0.05 day 2 versus day 3. The bars are expressed in mean + SEM Discussion This study revealed that DOX, in a dose of 10 and 6 mg/kg induced severe morphological damage to the small intestine of mice within 3 days, which was almost completely regenerated by day 7. Moreover, it revealed that the intestine was virtually not or much less affected by lower doses of DOX. Mucositis induced by the chosen dose of DOX was characterized by an increasing degree of intestinal morphological damage at day 1 and 2, which correlated with a significant increase in both apoptosis and proliferation. During this phase of epithelial hyper-proliferation, the epithelial cells lost their highly differentiated status as measured by a significant down-regulation of epithelial-specific SI at days 2–3. The decreased expression of TFF3 at days 1–2 could be caused by a decrease in goblet cell differentiation, but on the other hand, could also be the result of increased TFF3 secretion. At day 3, the time-point when intestinal damage was most severe, the morphology was characterized by severe villus atrophy, a significant rise in crypt length, epithelial flattening, crypt loss, inhibition of proliferation and impaired epithelial differentiation. During morphological regeneration, at day 7, proliferation started to return to control level, and SI and TFF3 expression levels were normalized again. In order to be able to prevent or treat chemotherapy-induced mucositis, it is essential to know if different cytostatic drugs induce the same or different kinds of intestinal damage. Potten et al. demonstrated that there is a general tendency for antibiotics, like DOX, and radiation to damage the lower cell positions in the crypt near or at the position of the stem cells (position 4–6) [12]. Alkylating agents on the other hand mainly damage cells at position 6–8. Anti-metabolites like MTX and a microtubule dissociating agents act on higher cell positions [9–11]. It is, however, unclear if cytostatic drugs attacking at the lowest positions in the crypts cause a different kind of damage than drugs damaging cells at higher positions. If we compare the DOX-induced mucositis as studied presently with the well-characterized MTX-induced mucositis [34, 38, 44–46, 48], then it is clear that there are many similarities and few discrepancies between the two. Although the two drugs affected cells of different hierarchical height [12], they both caused apoptosis, villus atrophy, epithelial flattening, crypt loss and a temporary loss of SI expression and TFF3 expression [45, 46, 48]. Since SI is involved in carbohydrate metabolism and TFF3 is involved in mucosal repair [6, 22, 30, 42], these data suggest impaired absorption and mucosal repair after DOX as well as after MTX treatment. In contrast, both MTX [45] and DOX hardly affect the expression of lysozyme by Paneth cells in the crypts (data not shown). Decreased levels of TFF3 after both DOX and MTX was at the same time as epithelial hyper-proliferation [45, 48], but changes in proliferation, induced by MTX treatment followed a different time-line compared to DOX treatment. MTX causes proliferation inhibition within 1 day, followed by a period of hyper-proliferation during severe intestinal damage [34, 45]. DOX treatment leads to immediate hyper-proliferation (day 1 and 2) with subsequent inhibition of proliferation during severe morphological damage (day 3). Moreover, the cell-fate specific affect of MTX on goblet cells causing goblet cells to accumulate in the crypt and at the top of the villus [45, 46] was not seen after DOX treatment. The reason for these discrepancies remains to be further investigated, but might be directly related to the difference in topographical height (and thus status) of the cells vulnerable to the two different drugs. Overall, however, the similarities in intestinal responses after DOX or MTX treatment are striking. This suggests that there may be common pathways involved in intestinal damage and repair. Historically, chemotherapy- or radiation-induced mucositis was believed to be solely due to damage to dividing epithelial cells at the bottom of the crypts [20]. However, recently it has become clear that other parts of the intestinal mucosa and submucosa might also be involved [35, 36]. Here we provide evidence for a mesenchymal contribution to the damage by showing that BMP4, a very important lamina propria derived-morphogen in the small intestine [8, 10], is affected by DOX treatment. BMP4 was modulated by DOX during the onset of damage at days 1 and 2. BMP4 expression decreased almost significant at day 2, which correlated well with an increasing degree of morphological damage, increased proliferation, and loss of epithelial differentiation as measured by the decreased SI and likely TFF3 expression. Very recently, a link between the BMP and the Wnt pathways has been demonstrated. It was shown that BMP signaling suppresses Wnt signaling to ensure a balanced control of stem cell proliferation and subsequent epithelial differentiation [8, 10]. Here we show that there indeed might be a close relationship between BMP and Wnt pathways, because at day 1 when BMP4 expression is decreased, expression of TCF4, a Wnt effector, increased significantly, which correlated well with increased proliferation and inhibited epithelial differentiation. At day 2, BMP4 remained low, whereas TCF4 returned to a normal level and remained at control level during the following days. At day 3, when damage was most severe, BMP4 was increased, which inversely correlated with proliferation, and correlated with epithelial differentiation, as suggested by the recovery of TFF3 expression level. BMP is involved in epithelial-mesenchymal signaling [8] and therefore we conclude that the data presented in this study indicate that epithelial-mesenchymal cross-talk is modulated during onset of DOX-induced damage in order to stimulate proliferation instead of differentiation and during severe intestinal damage to induce differentiation and inhibit proliferation. During the regenerative phase, at day 7, BMP4 expression level was down-regulated again, which could be a response to the shortage in number of crypts, since a blockage of BMP4 has been shown to cause stem cells to divide, leading to newly formed crypts [8, 10]. Therefore, our findings are in line with the roles of the BMP and Wnt/TCF pathway in epithelial homeostasis/morphogenesis. Furthermore, the decrease in BMP4 might also indirectly cause the observed decrease in SI expression, because inhibition of BMP4 stimulates Wnt signaling, and Wnt signaling itself induces SOX9, a negative regulator of Cdx2, which is a SI transcriptional activator [3, 10, 39]. In addition, BMP4 has been shown to be directly involved in HNF-1α expression, also a well-known activator of SI transcription [25]. Thus, the decrease in BMP4 might result in a decrease in Cdx2 and HNF-1α expression, two of the most important activators of SI transcription [39]. Currently, it is not known whether the Wnt or BMP4 signaling pathways regulate TFF3 gene expression. In conclusion, high-dose DOX induces severe damage to the epithelium, which closely resembles damage induced by MTX, indicating that general mechanisms of damage and repair are involved. We show that signaling pathways involving BMP4 and TCF4 and thus epithelial-mesenchymal cross-talk are modulated by DOX-induced damage in such a way that homeostasis of the progenitor compartment is restored by initially inducing cell proliferation and inhibiting differentiation and subsequently inducing differentiation, inhibiting proliferation and promoting crypt fission. Understanding these mechanisms is essential to develop clinical strategies to prevent chemotherapy-induced mucositis.

J_Biol_Inorg_Chem-4-1-2359827

Spectroscopic evidence for an all-ferrous [4Fe–4S]0 cluster in the superreduced activator of 2-hydroxyglutaryl-CoA dehydratase from Acidaminococcus fermentans

The key enzyme of the fermentation of glutamate by Acidaminococcus fermentans, 2-hydroxyglutarylcoenzyme A dehydratase, catalyzes the reversible syn-elimination of water from (R)-2-hydroxyglutaryl-coenzyme A, resulting in (E)-glutaconylcoenzyme A. The dehydratase system consists of two oxygen-sensitive protein components, the activator (HgdC) and the actual dehydratase (HgdAB). Previous biochemical and spectroscopic studies revealed that the reduced [4Fe–4S]+ cluster containing activator transfers one electron to the dehydratase driven by ATP hydrolysis, which activates the enzyme. With a tenfold excess of titanium(III) citrate at pH 8.0 the activator can be further reduced, yielding about 50% of a superreduced [4Fe–4S]0 cluster in the all-ferrous state. This is inferred from the appearance of a new Mössbauer spectrum with parameters δ = 0.65 mm/s and ΔEQ = 1.51–2.19 mm/s at 140 K, which are typical of Fe(II)S4 sites. Parallel-mode electron paramagnetic resonance (EPR) spectroscopy performed at temperatures between 3 and 20 K showed two sharp signals at g = 16 and 12, indicating an integer-spin system. The X-band EPR spectra and magnetic Mössbauer spectra could be consistently simulated by adopting a total spin St = 4 for the all-ferrous cluster with weak zero-field splitting parameters D = −0.66 cm−1 and E/D = 0.17. The superreduced cluster has apparent spectroscopic similarities with the corresponding [4Fe–4S]0 cluster described for the nitrogenase Fe-protein, but in detail their properties differ. While the all-ferrous Fe-protein is capable of transferring electrons to the MoFe-protein for dinitrogen reduction, a similar physiological role is elusive for the superreduced activator. This finding supports our model that only one-electron transfer steps are involved in dehydratase catalysis. Nevertheless we discuss a common basic mechanism of the two diverse systems, which are so far the only described examples of the all-ferrous [4Fe–4S]0 cluster found in biology. Introduction The anaerobic bacterium Acidaminococcus fermentans ferments glutamate via the 2-hydroxyglutarate pathway to ammonium, carbon dioxide, acetate, butyrate, and molecular hydrogen [1]. A key enzyme in this fermentation is 2-hydroxyglutaryl-CoA dehydratase (CoA is coenzyme A), which catalyzes the reversible syn-elimination of water from (R)-2-hydroxyglutaryl-CoA, resulting in (E)-glutaconyl-CoA. This reaction is of considerable interest since a nonactivated β-C–H bond has to be cleaved (pK ≈ 40); for reviews see [2–4]. The enzyme system from A. fermentans is composed of two protein components, the homodimeric activator or archerase (HgdC) and the heterodimeric dehydratase (HgdAB), whose genes have been cloned and sequenced. After overexpression of hgdC in Escherichia coli, the activator could be purified to homogeneity aided by a C-terminal Strep-tag. The protein [γ2, 2 × (27 ± 1) kDa] contains 4.0 ± 0.1 iron and 3.8 ± 0.1 acid-labile sulfur, which form one [4Fe–4S]+/2+ cluster between the two subunits [5]. The activator can be readily reduced with dithionite or titanium(III) citrate. Most likely in vivo a reduced clostridial [4Fe–4S]-type ferredoxin [6] or the hydroquinone form of a flavodoxin [7], both isolated from A. fermentans, can act as reducing agents for the activator. The activator is extremely oxygen-sensitive; its half life under air is only about 10 s. The oxygen sensitivity results from destruction of the [4Fe–4S] cluster, which is exposed to the solvent [8]. Furthermore, the activator also precipitates in the presence of the dyes that are commonly used for the determination of redox potentials; herefore, the redox potential of the [4Fe–4S]+/2+ cluster (E0′ ≈ −320 mV) can only be roughly estimated from its almost complete reduction by ferredoxin or flavodoxin (E0′ = −420 mV) [7]. The native 2-hydroxyglutaryl-CoA dehydratase HgdAB (αβ, 54 + 42 kDa) contains 4.5 ± 0.3 iron, 3.5 ± 0.4 acid-labile sulfur and 1.0 FMNH2 [5]. In order to obtain catalysis, incubation of the dehydratase with reduced activator in the [4Fe–4S]+ state, ATP, and MgCl2 is necessary. Spectroscopic studies revealed that thereby one electron is transferred from the activator to the [4Fe–4S] cluster of the dehydratase [5, 7, 9], which cannot be reduced by any other means, except by deazaflavin and light in the presence of EDTA as shown for the related 2-hydroxyisocarproyl-CoA dehydratase (J. Kim and W. Buckel, unpublished results). A mechanism has been proposed in which the dehydration is initiated by injection of this electron into the thiol ester carbonyl. The resulting substrate-derived ketyl radical eliminates the adjacent hydroxyl group. The enoxy radical formed can now be deprotonated to the product-related allylic ketyl radical, because radical formation has lowered the pK of the β-hydrogen by about 26 U [10]. In the last step the electron is returned from the allylic ketyl radical to the [4Fe–4S] cluster of the dehydratase and a new cycle can start. With the related 2-hydroxyisocaproyl-CoA dehydratase from Clostridium difficile this radical has been identified by electron paramagnetic resonance (EPR) spectroscopy [11] and it has been shown that such a cycle can endure for about 10,000 turnovers until a new ‘shot’ by the activator is required [12]. Homologous activators with around 60% sequence identity have been characterized from 2-hydroxyglutaryl-CoA dehydratase in Fusobacterium nucleatum [13], lactyl-CoA dehydratase in C. propionicum (T. Selmer and W. Buckel, unpublished results), phenyllactyl-CoA dehydratase in C. sporogens [14, 15], and 2-hydroxyisocaproyl-CoA dehydratase in C. difficile [12]. Furthermore, each genome of almost all anaerobically thriving bacteria and archaea, even that of E. coli (yjiL gene), encodes up to four different homologues of the A. fermentans activator with 30–40% sequence identity. Only the function of the two homologues in Thauera aromatica and other anaerobic benzoate-degrading bacteria could be assessed as the α- and δ-subunits of the four subunits comprising benzoyl-CoA reductase. The catalysis of this enzyme, which in addition requires ATP and reduced ferredoxin as substrates, has been proposed to proceed also via a ketyl radical [16]. Hence such activators, which probably all contain a single [4Fe–4S] cluster between two subunits with a helix–cluster–helix architecture [8], are widespread in the anaerobic world. Proteins with almost the same architecture but with unrelated sequences are the nitrogenase reductases, which appear to have the same function as the activators [17]. The proteins of both families increase the reductive power of an electron by ATP hydrolysis. Nitrogenase reductase from Azotobacter vinelandii, also called Fe-protein or Av2, has been shown to become superreduced to the all-ferrous state by titanium(III) citrate [18–26]. The energy efficiency of nitrogenase would be twofold enhanced if the catalysis proceeded via superreduced Av2, owing to the transfer of two electrons from the corresponding [4Fe–4S]0 cluster per hydrolysis of two ATP molecules [27, 28]. However, the low redox potential Em = −790 mV reported for the (+1/0) couple of the iron–sulfur cluster virtually excludes any physiological relevance [measured with Cr(II)EDTA as a reductant] [29]. Interestingly, a much higher value, Em = −460 mV, was initially obtained by using methyl viologen [26]. Moreover, a recent report indicates that nitrogenase catalysis actually can be supported by superreduced Av2 and may proceed via the (+2/0) couple of the iron–sulfur cluster, if the in vivo reductant flavodoxin hydroquinone (FldHQ) is used [30]. In this new state of the Fe protein, the all-ferrous [4Fe–4S]0 cluster is difficult to discriminate from the +1 and +2 redox states using the absorption spectra, and it appears to be diamagnetic. In contrast, the Av2 [4Fe–4S]0 cluster made with titanium(III) citrate has a distinct absorption spectrum [18], and an unusually high spin state S = 4, as was inferred from elaborate integer-spin EPR and magnetic Mössbauer studies [18, 19, 31, 32] and theoretically explored by density functional theory studies [33, 34]. Here we report an all-ferrous state for the [4Fe–4S] cluster in the activator from A. fermentans after treatment with a tenfold excess of titanium(III) citrate. To our knowledge this is the second biological system which stabilizes a [4Fe–4S]0 cluster. Remarkable spectroscopic and biochemical similarities between activator and the A. vinelandii Fe-protein give reason to favor a common functional principle of these exciting proteins. Materials and methods Metallic 57Fe was purchased from Chemotrade (Düsseldorf, Germany). The metal (20 mg) was dissolved in 0.5 mL 8 M HCl at 80 °C for 12 h. This stock solution was added directly to the medium. Standard I nutrient broth was from Merck (Darmstadt, Germany). E. coli strain XL1-blue was purchased from Stratagene (Heidelberg, Germany). Streptavidin–Sepharose and the plasmid pASK-IBA3 were obtained from IBA (Göttingen, Germany). All other chemicals and substances were from Sigma or Fluka (Deisenhofen, Germany). Anaerobic manipulations were performed in an anaerobic chamber (Coy Laboratories, Ann Arbor, MI, USA) with an atmosphere of 5% H2 in N2. X-band EPR measurements were performed using a Bruker E500 ELEXSYS continuous-wave spectrometer with the Bruker dual-mode cavity (ER4116DM) or a standard cavity (ER4102ST) and an Oxford Instruments helium-flow cryostat (ESR 910). Microwave frequencies were measured with a Hewlett–Packard frequency counter (HP5352B), and the field control was calibrated with a Bruker NMR field probe (ER035M). The spectra were simulated on the basis of a spin-Hamiltonian description of the electronic ground state: where St is the spin of the coupled system and D and E/D are the axial and rhombic zero-field parameters. The simulations were performed with a program which was developed from the S = 5/2 routines of Gaffney and Silverstone [35] and which specifically makes use of the Newton–Raphson iterative search method for transition fields as described there. Frequency-dependent Gaussian line shapes were used with full-width-at-half-maximum σν given in millitesla at g = 1. Alternatively also the Bruker XSOPHE suite [36] was used for the calculation of energy-level schemes and transition roadmaps. Mössbauer spectra were recorded using two spectrometers with alternating constant acceleration. The minimum experimental line width was Γfwhm = 0.24 mm/s (full width at half maximum). The sample temperature was maintained constant either in an Oxford Variox or an Oxford Mössbauer-Spectromag cryostat. The latter is a split-pair superconducting magnet system for fields up to 8 T in which the sample temperature can be kept constant in the range 1.5–250 K. The field is perpendicular to the γ-beam. The 57Co/Rh source was positioned at room temperature inside the gap of the magnet in a reentrant bore tube at a distance of about 85 mm from the sample. The field is zero at this position. All isomer shifts are quoted relative to iron metal at 300 K. Magnetic Mössbauer spectra were simulated with a computer program based on the same spin-Hamiltonian for St = 4 as for the EPR simulations (Eq. 1). The hyperfine interactions for 57Fe were calculated by using the usual nuclear Hamiltonian [37]. Purification of recombinant activator from A. fermentans Standard I medium (8–10 L, Merck, Germany) was inoculated with 25 mL overnight culture of recombinant E. coli (XL1-blue/hgdC in pASK-IBA3) and grown in tightly closed 2-L bottles at 30 °C. When an optical density at 600 nm of about 0.3 was reached, cells were induced with anhydrotetracycline (30 μg/L). After anaerobic growth for a further 3 h, cells were harvested by centrifugation and suspended in 20 mM 3-(N-morpholino)propanesulfonic acid, pH 7.2, 100 mM NaCl, 10 mM MgCl2, 1 mM ATP, 2 mM dithiothreitol or 4 mM dithionite (buffer A). The bacteria were lysed anaerobically using a French press system. Cell debris and membranes were removed by ultracentrifugation for 60 min at 100,000g. The filtered supernatant was loaded with gravity flow onto an 8-mL streptavidin–Sepharose column, which was equilibrated with buffer A. After complete binding, the column was washed with at least ten column volumes of buffer A. The activator was eluted with 2.5 mM d-desthiobiotin in buffer A, then concentrated with an Amicon stirred cell, and immediately used for further experiments. Results Preparation and characterization of superreduced activator The activator as isolated was concentrated to 60–80 mg protein/mL (1.2–1.6 mM) in 50 mM tris(hydroxymethyl)aminomethane hydrochloride pH 8.0. A tenfold excess of titanium(III) citrate was added, which yielded a color change from brown to red. Correspondingly, a new shoulder appeared in the UV–vis spectrum at about 500 nm (Fig. 1). The orange-red color apparently is a typical feature of superreduced [4Fe–4S]0 clusters, and does not originate from titanium(III) citrate directly, because this was removed by gel filtration on a Sephadex G-25 desalting column. It is noteworthy that the same color change was also observed upon full reduction of the nitrogenase Fe-protein from A. vinelandii [18]. Fig. 1UV–vis absorption spectrum of activator reduced with dithionite, and superreduced with titanium(III) citrate As described previously, the activator had to be used right after purification for biochemical studies because of its tendency to precipitate within 4 days [5]. To our surprise, the superreduced activator was nearly as stable as the singly reduced or oxidized enzyme; the activity decayed only with a half life of 4–5 days or more. Remarkably, dehydratase activity could not be observed with 2-hydroxyglutaryl-CoA when titanium(III) citrate reduced activator was used in the activity assay [38]. EPR spectroscopy In a previous EPR and Mössbauer study we demonstrated that dithionite-reduced activator1 possesses a [4Fe–4S]+ cluster with spin S = 3/2 [5]. The unusual magnetic ground state was inferred from X-band EPR spectra as shown in the inset in Fig. 2. The EPR resonances extend over a wide field range of at least 800 mT with a dominating strong absorption-like peak at g ≈ 4.5 and a broad trough at g ≈ 1.3–2. The large anisotropy and particularly the presence of high effective g values indicate the high spin multiplicity with large zero-field splitting (D > gμBB). Although the broad derivative spectrum does not have the classical appearance known for a quartet species from a transition metal complex [39], the pattern could be nicely simulated by using the usual spin-Hamiltonian for S = 3/2 with reasonable zero-field splitting parameters D = 1.3 cm−1 and E/D = 0.213. Just a simple Gaussian distribution of the rhombicity parameter with half width σ(E/D) = 0.11 was sufficient to account for the observed line broadening caused by g strain [5]. Moreover, a minor fraction of the [4Fe–4S]+ clusters (approximately 3%) was found to have the more common spin state S = 1/2, which gives rise to the narrower resonances around g = 2. The corresponding simulations for both spin states of [4Fe–4S]+, S = 1/2 and 3/2, are shown as gray and blue lines in the inset in Fig. 2. Fig. 2Normal-mode (B1⊥B0) X-band electron paramagnetic resonance (EPR) spectrum of titanium(III) citrate reduced activator (low-field part, microwave frequency 9.6337 GHz; power 1.6 mW; modulation 2 mT at 100 kHz). The simulation for St = 4 (green line) is obtained with D = −0.66 cm−1, E/D = 0.165, g = (2.1, 1.99, 2.03) (Footnote 2) and a frequency-dependent Gaussian line width of σν = 70 mT. The inset shows the full range X-band spectrum of dithionite-reduced activator at 10 K with simulations for the [4Fe–4S]+ clusters as published previously [5]. The arrows in the main panel and in the inset indicate the field position of the strongest integer-spin signal from [4Fe–4S]0 clusters with St = 4 The strong absorption-like band at g ≈ 4.5 from S = 3/2 persists also in the EPR spectrum of titanium(III) citrate treated activator (Fig. 2, bold trace). We assign it to [4Fe–4S]+ clusters remaining from incomplete conversion of the reduced activator. The relative abundance of this contribution is estimated by corresponding the Mössbauer spectra to account for about 50% of the total iron (see below). Formation of superreduced clusters, however, is also discernible in the EPR spectrum from the appearance of two new lines in the low-field range at g ≈ 11 and g ≈ 16. Narrow absorption-like signals at such high g values without related strong derivative lines at higher field are completely unexpected for half-integer-spin systems, but rather suggest integer-spin transitions. The interpretation is corroborated by parallel-mode measurements (B1||B0) for which the new low-field resonances are the strongest signals, whereas the spectrum from the reduced activator, particularly the absorption-like band at g ≈ 4.5, is completely attenuated, as shown in Fig. 3. The striking difference of the two detection modes results from the different selection rules Δm = ±1 for B1⊥B0 and Δm = 0 for B1||B0, which essentially discriminate half-integer-spin and integer-spin transitions [40, 41]. Therefore, the low-field transitions at g ≈ 11 and g ≈ 16 must be assigned to an all-ferrous [4Fe–4S]0, cluster for which coupling of the four Fe(II) sites with local spin SFe = 2 can yield only an integer total spin. This clearly indicates superreduction of the activator by titanium(III) citrate. The ground state for such a tetrameric all-ferrous cluster can have, in principle, any total spin ranging from St = 0 to St = 8, depending on the six spin coupling constants between the four ferrous ions. But sophisticated quantum chemical studies for the titanium(III) citrate reduced [4Fe–4S]0 cluster from the Fe protein suggest only St = 0, 4, or 8 as possible solutions for that particular system [33, 34]. Below we show that the all-ferrous [4Fe–4S]0 cluster in superreduced activator actually has total spin St = 4, with properties that closely resemble those of the superreduced state of the Fe protein from A. vinelandii [18, 31]. Fig. 3Parallel-mode (B1||B0) X-band EPR spectra of titanium(III) citrate reduced activator recorded at 3, 10, and 20 K (microwave frequency 9.3404 GHz; power 1.6 mW; modulation 0.57 mT at 100 kHz). The green lines represent simulations for St = 4 with D = −0.66 cm−1, E/D = 0.165, g = (2.1, 2.0, 2.1) (Footnote 2), and a frequency-dependent Gaussian line width of σν = 30 mT The integer-spin signals are remarkably sharp in the parallel-mode EPR spectra with full width at half maximum of only Γfwhm ≈ 4–5 mT for the main derivative peaks, but with very broad and asymmetric underlying shoulders. Such a superposition of broad and narrow powder signals may be caused by complex distributions of spin-Hamiltonian parameters, which owe their origin to microheterogeneity in the surrounding of the [4Fe–4S] clusters. In normal mode, however, the corresponding lines appear to be almost Gaussians of moderate 9–12 mT line width (Fig. 2). The difference may be related to the different orientational selections that contribute to the respective powder spectra [41]. Other ms components of the wave functions determine the transition probabilities between the zero-field split sublevels of the integer-spin system for parallel-mode detection, and thus may respond differently on the structural distortions. In the simulations, however, we could not reproduce the complex experimental line shape of the parallel-mode spectra of the superreduced activator by a simple distribution model for any of the spin-Hamiltonian parameters. Since we wanted to refrain from further sophistications other than Gaussian distributions, we just focused on the main features of the EPR spectra, i.e., the well-resolved sharp lines in the parallel-mode spectra and the Gaussian lines of the normal-mode spectra recorded at base temperature. The EPR spectra of superreduced activator recorded in normal and parallel mode can be consistently simulated with total spin St = 4, as shown in Figs. 2 and 3. The zero-field parameters obtained from global optimization are D = −0.66 (±0.1) cm−1 and E/D = 0.165 (±0.01), with g = (2.2, 1.99, 2.03).2 Particularly the unusual temperature dependence of the two narrow parallel-mode signals in Fig. 3 is basically well reproduced with those values, i.e., at 3 K the simulated line at g ≈ 11 is stronger than the line at g ≈ 16, whereas the intensity ratio is inverted at 20 K. According to the energy splitting scheme shown on the left in Fig. 4, the derivative signals arise from transitions within the low-lying “ms = ±4” and the excited “ms = ±3” non-Kramers doublets, respectively. At the actual conditions of D and E/D, both of these ms doublets are virtually unsplit at zero magnetic field. For fields applied in the molecular z-direction, their X-band resonances are expected at apparent g values of 16 and 12, respectively. However, the transition probability is zero for the “ms = ±4” transition at exact B0||z and is very low for the “ms = ±3” transition, but the Zeeman effect is very anisotropic and causes mixing of the ms wave functions for both doublets, so EPR transitions become allowed for B0 oriented between the principal axes. The major intensity in a parallel-mode powder spectrum actually arises from orientations close to the z-axis. The variation of the resonance field as a function of the field orientation is visualized by the transition roadmaps shown on the right in Fig. 4. Powder derivative signals are expected from these diagrams only at the turning points of the respective traces. The corresponding g values for the “ms = ±3” and “ms = ±4” derivative spectra are indicated by the vertical dotted lines and g labels in diagram B in Fig. 4. Interestingly, there is a third resonance traced in the diagrams, coming from the highest pair of ms levels for the field close to the y-axis (marked with an asterisk). That also has a turning point at g ≈ 16, so the contribution in the EPR spectrum entirely overlaps the “ms = ±4” transitions. Fig. 4A Energy level scheme for St = 4 with D = −0.66 cm−1, E/D = 0.165, g = 2.03, and B0 applied in the z-direction. The ms values given in quotation marks are labels for the magnetic sublevels but do not reflect the true composition of the wave functions. The vertical bold bars indicate parallel-mode transitions for ν = 9.6 GHz. B–D Transition roadmap diagrams for the same system, which are plots of the resonance fields for hypothetical single-crystal rotation experiments. The resonances of the “ms = ±4” and “ms = ±3” non-Kramers doublets are shown in red and blue, whereas transitions within the highest two ms levels are marked with an asterisk: B rotation of B0 in the z → x-plane; C rotation in the y → x-plane; and D rotation in the y → z-plane The roadmap diagram also explains why the g ≈ 11 and the g ≈ 16 derivative signals have different intensities in the parallel-mode spectra at 20 K, although the Boltzmann law arrogates almost equal population of the corresponding non-Kramers doublets at this condition of temperature and field strength. Apparently, the ground-state “ms = ±4” doublet has a higher magnetic anisotropy, so the resonances become spread over a wider field range than for the “ms = ±3” transitions and the signal amplitudes are attenuated. Additionally, also lower intrinsic transition probabilities can be found at g ≈ 16, because higher ms functions become less mixed as a result of the competition between Zeeman- and zero-field interaction. At 3 K, in contrast, the g ≈ 11 signal fades out because the excited states become mostly depopulated. The effect yields substantial information about the magnitude of D and E/D, although the constraint on the parameters obtained from the intensity variations in the parallel-mode spectra in Fig. 3 may be somewhat weakened by the odd overall experimental line shapes of the two signals. However, the actual D and E/D values are nicely supported by the successful simulation of the resonance lines in the normal-mode spectrum in Fig. 2. The parallel-mode spectra of the superreduced activator (Fig. 3) resemble remarkably well the EPR features of the all-ferrous cluster in the Fe-protein [21, 24] with a ground-state spin St = 4 [18, 31]. This cluster also shows a very sharp line in the parallel-mode X-band spectra around g ≈ 16 (actually at g = 16.4), but at g ≈ 11 only a weak, broad feature is observed at elevated temperature, whereas for the activator the line at g ≈ 11 is sharp and appears already at 3 K as a strong sharp signal. The inherent similarity of the spectra illustrates the same spin state for the all-ferrous clusters of the activator and the Fe protein and reveals similar physical properties. The differences in the EPR spectra arise from only small differences in the zero-field interaction, which slightly modulate the properties of the resonating non-Kramers doublets as shown in Fig. 4 (D = −0.75 cm−1, E/D = 0.30 for the Fe protein [31] compared with −0.66 cm−1 and 0.165 for the activator). In view of the fact that the zero-field splitting of the cluster spin manifold is inherited from the single-ion properties of four ferrous sites [42], and that there are 15 different ways to obtain St = 4 from such a spin tetrade [31], the close spectral similarity suggests very similar electronic properties of the all-ferrous clusters from activator and the Fe protein. Mössbauer spectroscopy We used 57Fe Mössbauer spectroscopy to quantify superreduction of the activator samples and to further substantiate the comparison with the Fe protein. Figure 5 shows the zero-field Mössbauer spectrum of titanium(III) citrate reduced activator recorded at 140 K. The asymmetric pattern could be fitted with three quadrupole doublets (I, IIa, and IIb), of which the prominent subspectrum I can be clearly assigned to [4Fe–4S]+ clusters from remaining starting material, because its isomer shift, δ = 0.51 mm/s, and quadrupole splitting, ΔEQ = 0.94 mm/s, are virtually identical to those observed for reduced activator at 80 K (δ = 0.53 mm/s, ΔEQ = 0.95 mm/s) [5]. The relative intensity of subspectrum I accounts for 54% of the total iron. Fig. 5Zero-field Mössbauer spectrum of titanium(III) citrate reduced activator at 140 K. Subspectrum I and the superposition of subspectra IIa and IIb represent the contributions from reduced [4Fe–4S]+ and superreduced [4Fe–4S]0 clusters, respectively. δ(I) = 0.51 mm/s, ΔEQ(I) = 0.94 mm/s, 54% relative intensity; δ(IIa, IIb) = 0.65 mm/s, ΔEQ(IIa) = 1.51 mm/s, ΔEQ(IIb) = 2.59 mm/s; intensity ratio of IIa and IIb fixed at 3:1 Superreduction of the activator and formation of all-ferrous [4Fe–4S]0 clusters is revealed by the presence of the wide split component with higher isomer shift shown as the shaded area in Fig. 5. The corresponding all-ferrous [4Fe–4S]0 clusters from Av2 showed a similar pattern, but one that could be fitted with four subspectra according to four different iron sites. However, three subspectra were strongly overlapping and had similar quadrupole splittings in the range ΔEQ = 1.2–1.7 mm/s, but one deviated, showing ΔEQ = 3.08 mm/s [18, 31]. Such distinct differences cannot be resolved from the spectra of the all-ferrous cluster of the activator because of broad lines and limited experimental resolution, but the basic distribution pattern of three similar sites and one different site seems to persist also for the activator. We could fit our data with two Lorentzian doublets, IIa and IIb, having identical isomer shifts and a fixed intensity ratio of 3:1, so subspectrum IIa accounts for three similar iron sites of the all-ferrous cluster with unresolved signals, and subspectrum IIb represents a fourth, discernibly different site. Details on the significance of the fit model are given in the electronic supplementary material.3 The Mössbauer parameters obtained are δ = 0.65 mm/s for IIa and IIb and ΔEQ = 1.51 mm/s for IIa and ΔEQ = 2.59 mm/s for IIb, respectively. The high isomer shift and large quadrupole splittings are typical for tetrahedral Fe(II)S4 sites with localized valences. Particularly the isomer shift, which is the key parameter for the assignment of valence states, compares very well with that of the monomeric Fe(II) centers in reduced rubredoxin (δ = 0.70 mm/s at 4.2 K) [32, 43] or with that of the diferrous [2Fe–2S] cluster in a plant-type ferredoxin (δ = 0.71 mm/s at 4.2 K) [44]. Finally it matches exactly the value found for the all-ferrous cubane clusters in the superreduced Fe-protein (δ = 0.65 mm/s at 130 K). We assume that the rather strong line broadening found for the activator, Γfwhm(IIa) = 0.7 mm/s, Γfwhm(IIb) = 0.47 mm/s, is caused by further heterogeneity of the cluster sites in the frozen protein sample. We can exclude broadening by paramagnetic relaxation at 140 K since basically the same pattern was observed at 80 K. It was not possible to achieve higher yields of superreduced [4Fe–4S]0 clusters than the approximately 50% mixture with remaining [4Fe–4S]+ clusters observed in the Mössbauer spectra shown here. Prolonged incubation with titanium(III) citrate or a higher excess of reducing agent did not improve the result. All attempts failed to fully convert a sample by the use of any chemical means, whereas photolytic reduction of the activator with deazaflavin led to cluster deterioration and precipitation of the protein. Because of this restriction, the midpoint of the [4Fe–4S]+ → [4Fe–4S]0 transition could not be exactly determined, but we estimate that the redox potential is lower than that of titanium(III) citrate, for which the value E0′1/2 = −720 mV against the standard hydrogen electrode is reported [45] (E0′ = −420 mV; or E0 = 0 mV). This estimation is based on the fact that most likely two protons per electron are released during the oxidation of TiIII to TiIV, which forms a titanyl ion: Hence at pH 8.0, at which the all-ferrous cluster was prepared, the redox potential of titanium(III) citrate should be—according to the Nernst equation—2 × 60 mV more negative, i.e., E0′ = −840 mV. Magnetic Mössbauer measurements of titanium(III)-treated activator were performed at 4.2 K with applied fields of 0.2, 4, and 7 T. They yielded a complex magnetic hyperfine pattern with a large overall splitting of about 10 mm/s as shown in Fig. 6. On the basis of the model applied for the fit of the zero-field data, the spectra could be decomposed into three components. The inner part in the range from −2.5 to 3 mm/s is mostly determined by the contribution from the reduced activator having [4Fe–4S]+ clusters with spin S = 3/2 [5]. The component could be reasonably well simulated (subspectrum I, blue lines) by using isomer shift, quadrupole splitting, zero-field splitting, and magnetic hyperfine parameters as published previously for the S = 3/2 species4 [5]. The relative intensity of subspectrum I was fixed to 54% as taken from the zero-field spectrum. Fig. 6Applied-field Mössbauer spectra of titanium(III) citrate reduced activator recorded at 4.2 K with B = 0.2, 4 and 7 T. The blue lines are spin-Hamiltonian simulations for subspectrum I from reduced [4Fe–4S]+ clusters with spin S = 3/2 [5] (Footnote 3). The thick green lines and the shaded areas mark the superposition of the two subspectra IIa and IIb from superreduced [4Fe–4S]0 clusters with the parameters as given in Table 1. The individual contribution of subspectrum IIb from the unique Fe(II) site with antiparallel aligned local spin is additionally shown as a thin green line above the other spectra The outer hyperfine lines in the magnetic Mössbauer spectra found at about −4 and +5.5 mm/s are the best indicators for the paramagnetic contribution from the all-ferrous [4Fe–4S]0 clusters, which must have large internal fields at the iron nuclei because of the large spin. The magnetic splitting of the corresponding spectra is clearly outside the typical range of hyperfine splitting known for [4Fe–4S]+ clusters with spin S = 3/2 or 1/2. We could readily simulate that contribution from the superreduced activator (Fig. 6, green lines with shaded areas) by using the spin-Hamiltonian for spin St = 4 (Eq. 1), amended with the 57Fe hyperfine interaction with four sites grouped in the two subspectra IIa and IIb introduced above. Particularly the field dependence of the Mössbauer spectra, which depends critically on the zero-field splitting, is well reproduced with the corresponding zero-field splitting parameters D = −0.66 cm−1 and E/D = 0.165 obtained from the EPR analysis. The complete set of Mössbauer parameters used in the simulation of the magnetic spectra are summarized in Table 1. We have to mention, however, that particularly the results obtained for subspectrum IIb are not unique, because of the relatively poor signal-to-noise ratio of the broad spectra and the strong overlap with the remaining starting material, which perturb particularly the analysis of the otherwise informative weak-field spectra. Therefore, the solution given for the anisotropy of the A tensors and the components of the electric field gradient tensors (η and the Euler angles α, β, and γ) have to be regarded as generic. The main objective of presenting the analysis of the magnetic Mössbauer spectra is to show that the data are nicely consistent with the EPR analysis. Table 1Mössbauer parameters for the all-ferrous [4Fe–4S]0 cluster of superreduced activator with spin St = 4 at 4.2 KSubspectrum IIa (three sites, 75% intensity)Subspectrum IIb (one site, 25% intensity)δ (mm/s)0.71 (3)0.71 (6)ΔEQ (mm/s)+1.80 (5)+2.6 (1)η0.8 (4)0.7 (4)α (°)90a90aβ (°)54 (10)125 (40)Ax/gNβN (T)−10.2 (1)+5.0 (5)Ay/gNβN (T)−6.4 (1)+8.0 (5)Az/gNβN (T)−7.23 (2)+7.8 (1)A0/gNβN (T)b−7.95 (7)+6.6 (4)The zero-field parameters used for the simulation were D = −0.66 cm−1 and E/D = 0.165. The Euler angles α, β, and γ rotate the electric field gradient tensor with respect to the principal axes system of the zero-field splitting. The numbers given in parentheses are rough estimates of the error margins obtained by single-parameter variation and visual inspectionaThe angles α and γ were fixed at 90° and 0° to reduce the number of parametersbIsotropic value is 1/3 tr(A) The three iron sites of subspectrum IIa that are indistinguishable at zero field also could not be further discriminated with applied fields, neither by sign or size of the quadrupole splitting nor by the magnetic hyperfine interaction. The isomer shift and quadrupole splitting appear to be uniform, with values close to those at 140 K with appropriate corrections for the lower temperature, and the hyperfine splitting can be described by only a single A tensor. Any possible heterogeneity in subspectrum IIa is masked by the anisotropy of A and the rather poor experimental resolution. The isotropic part A0 of the hyperfine tensor is negative for subspectrum IIa, which is best seen from the behavior of the resolved outer magnetic lines. They both have a strong component that is moving “inward” with increasing strength of the applied field from B = 0.2–7 T, owing to opposing internal and applied fields. Since the corresponding negative sign of A0 is like the sign intrinsically expected for iron(II) with respect to the local spin SFe = 2, the three iron sites forming subspectrum IIa must be basically coupled “parallel” to the total spin St = 4 of the [4Fe–4S]0 cluster. The isotropic value A0 for subspectrum IIb, in contrast, is taken to be positive. This causes the increasing magnetic splitting of a part of the outer hyperfine lines moving outward with increasing applied field. For demonstration of this effect the corresponding subspectrum IIb is also shown separately as a thin green line above the superposition of subspectra IIa and IIb and the experimental traces in Fig. 6. Thus, the local spin of the corresponding unique Fe(II) site is predominantly opposite to the total spin of the [4Fe–4S]0 cluster. The specific scheme of negative and positive A0 values was adopted from the study of the [4Fe–4S]0 cluster in superreduced Fe-protein [31]. There it was also demonstrated that the occurrence of one positive and three negative A values is a necessary common property for any ground multiplet with total spin St = 4 that one can obtain from a tetrade of four spins SFe,i = 2 with isotropic exchange interaction [31]. The effective A-tensor components found for the [4Fe–4S]0 cluster of the superreduced activator are relatively weak compared with the intrinsic a0 value for high-spin Fe(II) sites like that of rubredoxins (isotropic part A0/gNβN = −7.94 and 6.6 T for the activator cluster compared with a0/gNβN = −16.2 T for rubredoxin) [32, 43], but they are in the range of those of the [4Fe–4S]0 centers from the Fe-protein [A0 = −11.2 (±0.2) and 5 MHz [31], which corresponds to A0/gNβN = −8.1 (±0.14) and +3.6 T, respectively]. The apparent reduction of the effective hyperfine coupling in the cubane clusters is caused by competing antiferromagnetic spin coupling. In the corresponding coupling scheme invoked for the all-ferrous cluster of the Fe-protein [31], the spin-projection factors Ki that describe the effective hyperfine interaction with respect to the total spin of the individual sites, A0(i) = Kiao(i), have been obtained as K1 = K2 = K3 = 7/15 (subsumed in subspectrum IIa here) and K4 = −2/5 (subspectrum IIb). With the a0 value for rubredoxin, a0/gNβN = −16.2 T, one obtains A0(IIa)/gNβN = −7.6 T and A0(IIb)/gNβN = 6.5 T. Both values are in reasonable agreement with the experimental result found for the superreduced activator, similar to what was found for the Fe-protein. In summary the spectroscopic properties found for the all-ferrous [4Fe–4S]0 cluster from the superreduced activator resemble quite closely those of the superreduced Fe-protein from A. vinelandii nitrogenase. Discussion In the past, cubane [4Fe–4S]n+ clusters had only been known to undergo one-electron redox reactions using either the [4Fe–4S]+/2+ couple or the [4Fe–4S]2+/3+ couple, but not both [46]. According to which couple the clusters are supporting, the iron–sulfur proteins were graded as ferredoxin-type proteins or high-potential iron proteins (HIPIPs) [47]. The latter are most abundant in phototrophic organisms. Except for the case of the [4Fe–4S]n+ cluster in the HIPIP from Chromatium vinosum, which can be artificially cycled from the +3 to the +1 state [48], the classification appeared to be unique and generally valid, until Watt and Reddy [26] reported that the [4Fe–4S]+ cluster of the nitrogenase Fe-protein from A. vinelandii could be reversibly reduced by methyl viologen to an all-ferrous [4Fe–4S]0 state where all irons have oxidation state +2. The midpoint potential for this reaction was initially given as −0.46 V versus the standard hydrogen electrode (SHE), but later studies with other reductants reported −0.79 V versus SHE [29], in contrast to −0.31 V for the first reduction (+2 ↔ +1). Titanium(III) citrate reduced nitrogenase Fe-protein was the subject of elaborate EPR and Mössbauer spectroscopic investigations, which yielded detailed insight in the electronic structure of the all-ferrous cluster and revealed spin S = 4 for the ground state [18, 31]. More recently, the formation of another superreduced state of Av2 with a diamagnetic [4Fe–4S]0 cluster and a different electronic absorption spectrum was observed, if the in vivo reductant FldHQ was used [30]. Notably, that system has a sufficiently high redox potential to support nitrogenase catalysis with two-electron transfer by using the (+2/0) couple of the iron–sulfur cluster. Hitherto, nitrogenase Fe-protein was the only known enzyme which supported the all-ferrous [4Fe–4S]0 cluster. Also the first example of a synthetic all-ferrous [4Fe–4S]0 complex with ligands other than phosphane was reported only recently [49, 50]. It represents a very valuable analogue of the biological all-ferrous cluster, since it has been structurally and spectroscopically characterized. The Mössbauer spectra of the compound closely match those of titanium(III) citrate reduced Av2. In this paper we have demonstrated that superreduction in biological systems is not a unique phenomenon of the Fe-protein, but also the activator from 2-hydroxyglutaryl-CoA-dehydratase is able to support the [4Fe–4S]2+/+ as well as the [4Fe–4S]+/0 redox cycles of a single cubane cluster. The electronic structures of the superreduced [4Fe–4S]0 cluster from the Fe-protein and from the activator appear to have a series of remarkable spectroscopic similarities. First, the UV–vis spectra of both systems exhibit a shoulder-type signal around 500–530 nm which causes distinct color changes from brown to orange-red upon reduction. We note that reduced iron(II) rubredoxin does not exhibit a similar absorption in the visible region [51]. Moreover, both [4Fe–4S]0 clusters exhibit the same total spin St = 4 in their ground state, and even the zero-field-splitting parameters compare remarkably well (activator, D = −0.66 cm−1, E/D = 0.165; Fe-protein, D = −0.75 cm−1, E/D = 0.33 [31]). The moderate difference of the rhombicity parameters for the superreduced activator, deviating from the almost fully rhombic limit encountered with superreduced Fe-protein, gives rise to the two distinct lines in the parallel-mode EPR spectra of the activator instead of the strong single line found for the Fe protein. The Mössbauer isomer shifts for both centers are very similar and uniform for the respective cluster sites, owing to the same ferrous character of the metal ions. Our data are consistent with the specific distribution pattern of three iron(II) sites with similar quadrupole splitting and similar magnetic hyperfine coupling, and one different iron(II) site with larger quadrupole splitting and positive A tensor found for the Fe protein. In both enzymes the clusters are suspended at the interface of a protein dimer, but the spectra lack any indication of twofold symmetry, as one might expect from the respective protein structures [8, 52]. Since the electric field gradient, which gives rise to the quadrupole splitting, senses only local charge asymmetries, the more uniform appearance of the activator might reflect a smaller asymmetry of the actual cluster coordination in the protein and solvent moiety. Apart from that, the Mössbauer lines of the superreduced activator are unusually broad. Therefore we suppose that the molecular structure is probably not very well defined in the frozen solutions and exhibits microheterogeneity at the cluster site which leads to a distribution of quadrupole splittings for each of the iron(II) sites. Microheterogeneity seems also to be the reason for the unusual shape of the EPR signals measured from the superreduced activator. The resonances show a superposition of sharp lines and very broad shoulders centered at similar resonance fields. The corresponding distribution of spin-Hamiltonian parameters, like the distribution of electric field gradients, may be related to the location of the cluster at a sensitive position in the protein structure between two protein monomers [8], where variations in the conformational substates of the protein may have a particularly strong impact on the geometry and electronic properties of the cluster. Such a distribution of spin-Hamiltonian parameters has been observed previously also for the dithionite-reduced activator with [4Fe–4S]+ clusters having the unusual spin state S = 3/2 [5] (the unusually broad EPR spectrum is redepicted in the inset in Fig. 2). Here, it is interesting to mention that the [4Fe–4S]+ cluster of the Fe-protein has spin S = 1/2 under physiologically reduction conditions, but it also changes to S = 3/2 if it is reduced in the presence of urea [53]. In contrast to the situation for the Fe protein from nitrogenase, it was never possible to obtain a complete superreduction of the activator; at best a ratio close to 1:1 could be achieved for [4Fe–4S]+ and [4Fe–4S]0 clusters. One may speculate whether this is related to the conformational distributions invoked above for interpretation of the spectroscopic heterogeneities. Probably, some fraction within the ensemble of variations may impose such unfavorable conditions on the molecular structure of the cluster and the resulting electronic properties such that further reduction is practically impossible. This interpretation seems to be supported by the observation that hard photoreduction by using deazaflavin as a potent electron source leads to inactivation and precipitation of the protein. However, for the time being also the trivial explanation cannot be excluded, namely, the midpoint redox potential for the [4Fe–4S]+/0 couple could be lower than that of titanium(III) citrate at pH 8.0 (see above), and thus even with a tenfold excess of this agent the cluster only becomes half reduced. We also cannot readily explain why titanium(III) citrate reduced activator is found to be enzymatically inactive, whereas the EPR and Mössbauer spectra reveal the presence of about 50% [4Fe–4S]+ clusters remaining, which should be catalytically competent. We presume that in the enzymatic assay the activator is fully reduced to the [4Fe–4S]0 state because of the very low protein concentration in micromolar range, in contrast to the situation in the highly concentrated spectroscopic samples where the protein is in millimolar abundance. Since the ability of a [4Fe–4S] cluster to undergo two one-electron redox reactions using both the [4Fe–4S]+/2+ and the [4Fe–4S]+/0 couples should be basically related to its geometry, i.e., the metric details of molecular structure, as well as the (dielectric) properties of the protein site harboring the metal center. In this sense it may not be accidental that both proteins, the nitrogenase Fe-protein and the activator of 2-hydroxyglutaryl-CoA dehydratase, have very similar quaternary structures: homodimeric proteins in which a [4Fe–4S] cluster bridges the subunits. In both proteins an α-helix from each subunit points with its N-terminus towards the cluster and thus may stabilize with its dipole a negative charge, which should increase the redox potential. In the Fe-protein of nitrogenase this helix–cluster–helix architecture forms an angle of 150°, which upon binding of ATP and interaction with the iron–molybdenum protein opens to 180° [17]. Whether this conformational change further influences the redox potential of the cluster remains to be established. The helix–cluster–helix architecture of the activator of 2-hydroxyglutaryl-CoA dehydratase forms an angle of 105°, which is 45° closer than that of the Fe-protein of nitrogenase [8]. Upon ATP binding, the activator angle is thought to open in a similar manner [54], but a crystal structure of the complex between the activator and dehydratase stabilized by AlF4−-ADP (J. Kim and W. Buckel, unpublished results) is not available yet. Most likely the common electronic properties of the [4Fe–4S] clusters as well as the common structural features favor a related mechanism of both enzyme systems. Redox potentials are lowered by hydrolysis of ATP in order to achieve chemically difficult reductions of the substrates molecular nitrogen and 2-hydroxyglutaryl-CoA, respectively. However, owing to the observed inability of the superreduced activator to catalyze the enzyme reaction, the [4Fe–4S]2+ → [4Fe–4S]+ cluster transition and the subsequent electron transfer to the dehydratase component (HgdAB) appears to be the only biologically relevant charge transfer. The situation seems to be more complex for the nitrogenase mechanism. It was postulated that the all-ferrous nitrogenase Fe-protein indeed is an active redox state for nitrogen reduction. First, Watt and Reddy [26] reported that the ATP consumption for nitrogen fixation using all-ferrous Fe-protein is only half of that reported for the [4Fe–4S]+ cluster containing Fe-protein (one ATP per two transferred electrons compared with one ATP per one electron) [26]. In the meantime, the principal physiological relevance of the [4Fe–4S]0 cluster appears to have been corroborated for the nitrogenase system [30]. Maybe the enzyme can switch the redox pathway to choose between different electron donors for nitrogen reduction available in the cell. A similar role for the all-ferrous cluster of the 2-hydroxyglutaryl-CoA dehydratase system is still elusive. Electronic supplementary material Below is the link to the electronic supplementary material. Comparison of the fit model for the zero-field Mössbauer spectra of Ti(III) citrate-treated activator with one and two Lorentzian doublets for the all-ferrous cluster. (PDF 374 kb)

Eur_J_Nucl_Med_Mol_Imaging-4-1-2275773

99mTc Hynic-rh-Annexin V scintigraphy for in vivo imaging of apoptosis in patients with head and neck cancer treated with chemoradiotherapy

Purpose The purpose of this study was to determine the value of 99mTc Hynic-rh-Annexin-V-Scintigraphy (TAVS), a non-invasive in vivo technique to demonstrate apoptosis in patients with head and neck squamous cell carcinoma. Introduction Apoptosis is an important mechanism of cell death in response to treatment with radiation and many chemotherapeutic agents [1]. To what extent apoptosis contributes to the overall cytotoxic effect of an anti-cancer treatment modality has been the topic of intense research during the last decade [2–4]. The relative contribution of apoptosis to the occurrence of cell death varies greatly both between different tumour types and normal tissues [5]. Recently, in vivo imaging of apoptosis has proven to be feasible by using radiolabelled Annexin V [6–10]. This endogenous human protein has a high affinity for membrane-bound lipid phosphatidylserine (PS), which becomes exposed at the outer leaflet of the cell membrane bilayer at an early stage of the apoptotic process [11, 12]. PS then serves as a recognition site for macrophages that digest and remove apoptotic cells. The reproducibility of the 99mTc-Hynic-Annexin-V scintigraphy has been demonstrated in head and neck squamous cell carcinoma (HNSCC) patients by serial imaging in untreated patients [13] in which the mean difference in uptake was 6%. We have recently demonstrated that 99mTc-Hynic-rh-Annexin V scintigraphy (TAVS) correlates with radiation-induced cytologically confirmed apoptosis in non-Hodgkin lymphoma [14] and can be used to identify patients that have a favourable prognosis [15]. The current standard of practice for patients with advanced stage HNSCC is treatment with concurrent cisplatin-based chemoradiation [16, 17]. Although this is an effective treatment, it also is accompanied by more severe toxicity than radiation alone. We reasoned that TAVS early during treatment might be used to monitor apoptosis induction in tumour and normal tissue and to give an indication of the radiosensitivity of these structures. In future, this would also offer the possibility to adapt the treatment strategy at an early stage on a patient-by-patient basis. In the present study, we therefore aimed at assessing the feasibility of TAVS as a non-invasive technique to demonstrate treatment-induced apoptosis in vivo, in patients with HNSCC, early during treatment with concurrent chemoradiation. The purpose was to determine the degree of uptake on TAVS in normal tissue, primary tumour and lymph node metastases and to evaluate the treatment-induced Annexin uptake in relation to radiation dose. Furthermore, we questioned whether the differences in uptake would correlate with treatment response. Materials and methods Patients included in this study were recruited between January 2004 and March 2005 from a randomised phase III trial in advanced stage HNSCC, investigating the optimal route of cisplatin delivery during cisplatin-based chemoradiation (RADPLAT). Randomisation was between two treatment arms: Arm 1, standard intravenous (IV) administration of cisplatin 100 mg/m2 (1 h before radiotherapy at days 1, 22 and 43) or arm 2, high-dose selective intra-arterial (IA) delivery of cisplatin at a dose of 150 mg/m2 (on days 2, 9, 16 and 23, within 1 h after radiotherapy delivery). Main eligibility criteria included: inoperable squamous cell carcinoma of the oral cavity, oropharynx or hypopharynx, TNM stage T3-4 status of primary oral cavity or oropharyngeal tumours and T2-3-4 for hypopharyngeal tumours, with any N-status, (functionally) inoperable disease, no distant metastases, age at least 18 years, ability to give informed consent and no prior cerebro-vascular accident. Both the randomised trial and the TAVS study were approved by the medical ethics committee of the hospital. Patients were informed about the nature of the study protocol and signed informed consent before enrolment separately for both the randomised trial and the TAVS protocol. Radiotherapy was given with 4–6 MV photon linear accelerators. Target volume included the primary tumour and the bilateral neck for a dose of 46 Gy in 23 fractions. A boost was given to the known macroscopic tumour extensions at the primary tumour site and lymph node metastases to a dose of 24 Gy in 12 fractions. The total dose delivered was 70 Gy in 35 fractions, five fractions per week, with an overall treatment time of 7 weeks. The radiation technique was either a conventional three-field beam setup (using conventional simulation or virtual simulation with CT scan) or an intensity-modulated radiotherapy (IMRT) plan depending on resources and tumour extent. The IA cisplatin delivery was accomplished by a selective catheterisation procedure using the femoral artery according to the earlier described RADPLAT protocol [18, 19]. Concurrently, with IA cisplatin sodium-thiosulphate was administered to neutralise systemic cisplatin. In both arms, prehydration and posthydration were given. 99mTc-Hynic-rh-Annexin V scintigraphy, diagnostic imaging, radiation treatment planning and image fusion Baseline TAVS was performed within 2 weeks before the start of concurrent chemoradiation (average interval 5 days, range 1–8). Post-treatment TAVS was done within 48 h after the start of cisplatin chemotherapy. At the time of post-treatment TAVS, patients had received 6 Gy of radiation in case of IV cisplatin, 8 Gy in case of IA cisplatin. Each patient received an average of 847 MBq (range 714–1032 MBq) of 99mTc-Hynic-rh-Annexin V (Theseus Imaging Corporation, Boston, USA) by slow intravenous injection 4 h before the planar imaging and single-photon emission tomography (SPECT) imaging. Planar images were used to assess the biodistribution. SPECT of the head and neck region was acquired by the step-and-shoot mode, one step per 3°, 30 s per frame, matrix size 128 × 128, using a dual-head gamma camera (Genesis, Philips, Best, The Netherlands) equipped with low-energy, high-resolution collimators. For SPECT reconstructions, an iterative algorithm was used and the images were postfiltered using a Butterworth filter (cutoff frequency 0.35, order 5). The use of absolute quantitative analysis of SPECT data was validated by comparison of iterative and FBP reconstruction methods. Transaxial, coronal and sagittal slices were visually examined to evaluate tracer uptake at the tumour sites and in normal tissues. The intensity of the obtained images was corrected by normalisation for the injected radioactive dose and body weight. Baseline diagnostic imaging with MRI (1.5-T system; Somatom; Siemens Medical Systems, Erlangen, Germany) or spiral CT (Tomoscan AVE1, Philips, Best, The Netherlands or HiSpeed CT, GE Medical Systems, USA) was performed within 3 weeks before the start of treatment and repeated 6–8 weeks after the end of the treatment for the evaluation of treatment response. Radiation treatment planning was done with our clinical treatment planning system (TPS; U-MPlan, University of Michigan, Ann Arbor, USA). Six patients were treated by IMRT, and the treatment plans were recalculated for the dose at the time of first post-treatment TAVS. The regions of interest (ROI) were delineated manually for each patient on the CT scan. These included the gross tumor volume (GTV) of the primary tumour and/or lymph nodes, the parotid glands and the submandibular glands. In the other seven patients, a standard three-field technique was used by virtual simulation with a CT scan. In these latter patients, the CT scan was imported into the TPS in which the delineations were done. The treatment fields were reconstructed, and the clinically applied dose distribution at the time of TAVS was recalculated in the TPS. The primary tumour and lymph node volumes were calculated by 3-D reconstruction of the delineated GTV. SPECT, MRI and CT were performed separately. For the SPECT scan, we obtained reproducibility of the positioning of the head and neck as during RT by the use of immobilisation mask fixed to the table by adhesive tape. All images and the radiation dose data were transferred for image fusion to our in-house developed workstation for co-registration (Worldmatch Workstation) in DICOM format [20]. Keeping the limitations of the mask fixation in mind, it was decided that for more accuracy, matching was performed on different anatomical regions for different ROIs, e.g. for the analysis of parotid glands or oropharyngeal tumours, the co-registration was done in such a way that it ensured adequate matching for that area only. For co-registration of ROIs in the neck (e.g. lymph nodes or hypopharyngeal tumours), only the neck and vertebral column was used in the matching procedure. Accurate matching of body contours and bone structures was visually verified. SPECT, composite SPECT/SPECT, CT and SPECT/CT images, obtained using a colourwash technique, were simultaneously examined using linked cursor to evaluate Annexin V uptake in tumour and normal tissues. The tumour and normal tissue uptake at baseline and post-treatment were calculated as follows. ROIs were delineated in the planning CT. Each ROI was projected using the registration transformation onto the baseline and post-treatment SPECT scans. The area was next sampled by 10.000 random points, and for each point, the Annexin uptake was determined in the SPECT scans by trilinear interpolation. Of these 10.000 samples, the maximum and mean values were computed. In this way, the uptake in the ROI was accurately sampled, although the pixel size in the SPECT scans is relatively large (about 5 mm). For the quantification of Annexin uptake, no attenuation correction was performed because the images were obtained on a conventional gamma camera without a hybrid system. Subsequently, the difference (ΔU) between the post-treatment and baseline uptake was determined by subtraction of the baseline scan from the post-treatment scan. The subtraction was performed on a point by point basis for all 10.000 points. Then, the mean or maximum value of the difference of both scans (subtraction scan) was computed and expressed as ΔU. No correction for background activity was made, as this was automatically eliminated by subtraction of the baseline activity from the post-treatment activity, assuming that the background activity is equal in the baseline and post-treatment scan. Radiation dose parameters (mean and maximum dose in cGy) within each ROI were calculated and correlated with corresponding ΔU parameters. It was observed that the maximum ΔU and the mean ΔU were closely related: a linear relationship between both two parameters was found for each ROI (tumour: r = 0.93, p < 0.0001; parotid gland: r = 0.94, p < 0.0001; submandibular gland: r = 0.90, p < 0.0001; lymph node r = 0.94, p < 0.0001), indicating that both values would have been representative for the whole ROI. Evaluation of treatment Six to eight weeks after the end of treatment, the results of therapy were evaluated by means of radiological investigations (by MRI or CT scan and/or ultrasound) and examination under general anaesthesia, with biopsies taken in case of suspicious findings. For residual disease in the neck at the time of evaluation, salvage neck dissection was performed if the patient was judged operable. Follow-up visits were planned every 3 months in the first year after therapy, every 4 months in the second year and less frequent thereafter. A follow-up chest X-ray was performed annually. Statistical analysis For quantitative comparison of continuous data Student’s t test was applied. Chi-squared and Fisher’s exact test were used for analysis of categorical data. The Pearson and Spearman rank correlation coefficient r were used to calculate correlations between Annexin uptake and treatment parameters. Locoregional control and survival data were calculated from the start of treatment using the Kaplan–Meier method and log rank testing. Two-sided P values of <0.05 were considered statistically significant. Results Sixteen patients gave their consent for participating in the TAVS study. However, one patient refused post-treatment TAVS scans; in one patient, no CT scan in radiation treatment position was available, and in one patient, co-registration with SPECT was unsuccessful due to misalignment because no immobilisation mask and no tongue depressor could be used during SPECT due to pain (Fig. 1). The number of patients included in this analysis on 99mTc-Hynic-rh-Annexin V scintigraphy in advanced stage HNSCC is therefore 13. The mean age of patients was 53 years (range 26–66 years). See Table 1 for the patient, tumour and treatment characteristics. Fig. 1Sagittal projection of TAVS imaging co-registered with planning CT scan. The solid line represents the contour of the patient during RT-planning CT scan, including tongue depressor. The projected image is the TAVS. a Baseline TAVS; b post-treatment TAVS. Note the mismatch of contours of the head and neck mostly due to different position of the chin in b. The misalignment of the lower neck in a is due to the flexion of the neck. For the analysis, the matching procedure was focused on the mandibular region. This patient was excluded from analysis because of misalignment in bTable 1Patient, tumour and treatment characteristicsPatient numberGenderAge (years)Tumor siteTNM stageTumor volume (cm3)Max ΔU in primary tumourInterval cisplatin infusion and Annexin scan (h)RT techniqueMode of cisplatin administration1M64Oral cavityIII216252.2Conventional 3 fieldIntra-arterially2M58OropharynxIII4720350.3IMRTIntra-arterially3F65OropharynxIV2511253.1IMRTIntra-arterially4M55OropharynxIII263653.5IMRTIntra-arterially5M49HypopharynxII134353.3IMRTIntra-arterially6F66Oral cavityIV3213553.6Conventional 3 fieldIntra-arterially7M60OropharynxIV14719154.3Conventional 3 fieldIntra-arterially8M46OropharynxIV2817052.7Conventional 3 fieldIntra-arterially9M55OropharynxIV414249.9IMRTIntravenously10M57OropharynxIV255151.7IMRTIntravenously11M48OropharynxIV9111751.5Conventional 3 fieldIntravenously12M50HypopharynxIV10013250.3Conventional 3 fieldIntravenously13M26OropharynxIV1055250.1Conventional 3 fieldIntravenouslyM Male, F female, IMRT intensity modulated radiotherapy Annexin uptake in normal tissue and tumour At baseline scintigraphy, reproducible physiologic Annexin uptake was detected on the planar images in liver, kidneys, spleen, gall bladder, bone marrow, colon and urinary bladder, as described earlier [7, 14, 21]. In the parotid glands, weak baseline uptake was present in all cases. The post-treatment TAVS showed moderate to strong increase in uptake in 24 of 26 parotid glands (Fig. 2a). The average of the mean number of counts increased from 55 to 88 (p < 0.001). Visual analysis of the increase in uptake showed that these changes were related to the radiation portals and the dose given to the parotid glands (Fig. 3). The difference in Annexin uptake between the post-treatment and baseline TAVS (ΔU) in each parotid gland was correlated with the radiation dose at the time of post-treatment scintigraphy (n = 13 patients, n = 26 parotids). The mean ΔU showed a positive correlation with the mean radiation dose (Pearson coefficient r = 0.59, p = 0.002): parotid glands that received a higher dose of radiation showed a higher Annexin uptake (Fig. 4a). The increase in Annexin uptake in parotid glands that were treated with parotid-sparing IMRT was less than in parotid glands that were treated with a conventional three-field technique (IMRT 29 counts, three-field technique 59 counts, p = 0.02). The given radiation dose at the time of post-treatment TAVS (6 or 8 Gy) and the mode of cisplatin administration (IA or IV) did not affect the uptake in parotid glands. As xerostomia scoring was not documented sufficiently detailed in the files, we interviewed patients alive at last follow-up for xerostomia grading for the purpose of this study. In nine patients alive at last follow-up, the treatment-induced Annexin uptake in parotid glands was related to the subjective xerostomia rating using the EORTC QLQ HN-35 questionnaire [22]. No relation could be established in this small set of patients (data not shown). Fig. 2Baseline (diamonds) and post-treatment (squares) Annexin activity in parotid glands (a), submandibular glands (b) and tumour (c) for all 13 patients. The data for 26 parotid and submandibular glands represent the right and left gland within each patientFig. 3TAVS imaging co-registered with planning CT scan in frontal plane and axial plane, a and c at baseline, b and d after treatment, from patient number 2. The Annexin uptake is represented by colourwash. In b and d, the isodose lines show the dose distribution in relation to the parotid gland and primary tumour (isodose lines shown, 40, 60, 80 and 95% from outer to inner side of patient). Note the increased treatment-induced Annexin uptake in the right parotid gland, in correspondence to the higher radiation dose distribution, when compared to the left parotid gland. Also note the weak increase in primary tumour Annexin uptake in the right oropharynx after treatment and in the anterior floor of mouth (d)Fig. 4Correlation plot of mean radiation dose (cGy) and the post-treatment increase in Annexin uptake, mean ΔU, in parotid glands (a), normalised for tracer dose and body weight. Circles represent parotid glands treated with IMRT, triangles parotids treated with conventional technique. In b, the correlation between mean radiation dose (cGy) and the post-treatment increase in Annexin uptake, mean ΔU, in submandibular glands is given For the submandibular glands, a similar pattern as with the parotids was observed: absent/weak baseline uptake and moderate to strong increase after the start of chemoradiation (Fig. 2b). The average of the mean number of counts increased from 85 to 132 (p < 0.001). No correlation between the ΔU and radiation dose was noted (Fig. 4b), probably as all submandibular glands were located within the high-dose region. The baseline TAVS showed moderate to strong Annexin uptake in the primary tumour in all patients, indicative of spontaneous apoptosis or necrosis. On the post-treatment TAVS, the uptake in tumour clearly increased in nine cases; in the other four, little or no changes occurred (Fig. 2c). The average of the maximum number of counts increased from 127 to 163 (p = 0.007). The difference between post-treatment and baseline TAVS (ΔU) in primary tumour was not correlated with primary tumour volume (Pearson coefficient r = 0.39, p = 0.21) nor with the mode of chemotherapy administration (p = 0.21). In nine patients, 17 lymph node metastases were present in the neck (mean lymph node volume 5 cm3, range 1–19). The treatment-induced Annexin uptake (maximum ΔU) in primary tumour was positively correlated with the uptake in corresponding lymph node metastases (r = 0.73, p = 0.004) as shown in Fig. 5. To verify that the observed effects of Annexin uptake in different structures could not be attributed solely to the administered radioactive dose per patient but was tissue-specific, we calculated the correlation between the ΔU in tumour and parotid glands. No evidence of such an effect was found (r = 0.28, p = 0.18). Fig. 5Correlation plot of the treatment-induced increase in Annexin uptake, maximum ΔU, in primary tumour and lymph node metastases. Each colour represents an individual patient. Some patients had multiple lymph node metastases originating from the same primary tumour. In these cases, multiple values of treatment-induced Annexin uptake in lymph nodes (on the y-axis) correspond to a single value of the primary tumor (on the x-axis) Response to treatment, locoregional control and survival The median follow-up of all patients alive at last follow-up was 30 months (range 24–38). The complete response (CR) rate at the primary tumour site was 85%. CR rate in the neck in the case of nodal metastases was 100%. The overall CR rate was 85%. The estimated locoregional (LR) tumour control rates were 68% at 2 and 3 years. The rates for disease-free survival (DFS) were 54% at 2 and 3 years. Overall survival (OS) rates at 2 and 3 years were 62 and 40%, respectively. In this small subset of 13 patients from a larger randomised phase III trial, there were no differences in response rates, LR control, DFS or OS between the two treatment arms (IV vs IA). No correlation could be established between baseline or treatment-induced Annexin uptake (ΔU) and any outcome parameter (response rate, recurrences and/or survival). Discussion This study represents our first clinical experience on in vivo imaging of apoptosis with TAVS in HNSCC patients in which we applied co-registration of multiple imaging modalities. By this, we were able to analyse treatment-induced changes in normal and tumour tissue, early during therapy, after one cycle of cisplatin chemotherapy and 6–8 Gy of radiotherapy. Treatment-induced Annexin uptake in the parotid glands could be visualised, indicative of early treatment related apoptosis at mean radiation doses as low as 3–8 Gy (see Fig. 4a) and one course of cisplatin. The Annexin uptake in the parotids showed a radiation dose-response relationship: Glands that had received higher doses of radiation demonstrated increased Annexin uptake. Loss of parotid gland function, leading to xerostomia, is an important long-term side effect of radiotherapy, affecting quality of life of patients. It has well been shown by scintigraphy studies [23] and by salivary flow rate studies [24] that reduction of radiation dose to the parotid glands can maintain parotid function, decrease xerostomia and improve quality of life. Most studies on salivary gland function are performed after a full course of RT (typically 60–70 Gy in 6–7 weeks) [23–26]. Our study design with early in vivo imaging indicated that already after low dose of RT (6–8 Gy), parotid glands may be affected and that the physiologic process leading to loss of parotid gland function already starts early. This is in agreement with observations in experimental rodent studies in which apoptosis was induced early, after doses of up to 5 Gy [27]. Similar results were obtained in a monkey model, with early radiation induced apoptosis of serous acinar cells of salivary glands [28]. It was suggested that these early effects are mainly responsible for the acute sialoadenitis, which, at higher RT doses, might progress in chronic xerostomia by damage to and depletion of ductal stem cells within the gland. In the nine patients assessed for xerostomia, no relation could be established with subjective salivary gland function after radiotherapy and parotid gland Annexin uptake. However, the significance of this analysis is limited, as we assessed both parotid glands separately by TAVS, and xerostomia questionnaires give an overall impression of function. We did not include follow-up salivary flow studies. In short, our results indicate indirect evidence of early radiation-induced salivary gland damage and, in future studies, this needs to be correlated with functional outcome parameters. Due to the large inter-patient variability in the baseline TAVS (Fig. 2c), the subtraction method for assessing treatment-induced changes was chosen over a relative method (e.g. percentage change). In this figure, it is important to notice that the maximum uptake value of the baseline and post-treatment scan might be located in different points within the tumour ROI. Therefore, these data cannot be simply subtracted, but for the calculation of the maximum ΔU in primary tumour and lymph nodes, the baseline scan was first subtracted from the post-treatment scan, and, subsequently, these parameters were calculated from the subtraction scan. The treatment-induced Annexin uptake in primary tumour and pathological lymph nodes showed a positive correlation: Patients with primary tumours with a high Annexin uptake also appeared to have lymph node metastases with high Annexin uptakes. The ΔU in primary tumour and lymph node metastases showed large inter-patient differences (Fig. 5). This variation could not be attributed to differences in the route of cisplatin administration (intravenous or intra-arterial chemoradiation), nor to the radiation doses given (6 or 8 Gy), nor to the primary tumour volume. This positive correlation for ΔU in primary tumour and lymph nodes as well as the large inter-patient differences might represent a tumour-specific apoptotic response. In our series, repositioning of the patient during SPECT scan in RT position was standard procedure, but due to the limitations in fixation, not optimal. The repositioning facilitated the co-registration of SPECT and CT scan. In one patient, misalignment occurred because the positioning was not accurate due to the fact that the RT position could not be reproduced (Fig. 1). So, for co-registration of (functional) imaging to RT-planning CT scan data in HNSCC, we recommend improvements from our protocol to obtain a reproducible positioning, like the use of an immobilisation mask fixed to the SPECT table, the use of the same mattress under the patient and a laser alignment system. In ongoing studies, we have now incorporated these. In the current series, no attenuation correction was applied because the images were obtained on a conventional gamma camera without a hybrid system. For the purpose of this study, the quantification of Annexin uptake remains valid, as we used the relative increase from baseline to post-treatment uptake, and the degree of attenuation will be equal on both scans. In this small series of HNSCC, the treatment-induced Annexin uptake (ΔU) in primary tumour did not predict outcome: No correlation between Annexin uptake and early response was observed probably partly due to the high initial response rates (85%). In addition, TAVS also did not predict the locoregional control rates within the first 2 years of follow-up. In our previous experience on TAVS in other tumour sites (mainly lymphoma), a strong correlation has been found between early response (within 3 months) and Annexin uptake [15]. One explanation for the lack of correlation might be that in the current series, these advanced stages HNSCC with large tumours harbour more necrosis, which is also detected by TAVS because of accessibility of PS at the necrotic cell membrane [29]. Van de Wiele et al. [8] showed that in HNSCC, Annexin scintigraphy correlated with histopathological apoptosis (using TUNEL assay) only in the absence of necrosis. Indeed, in this series, baseline Annexin uptake in the HNSCC patients was clearly demonstrated, whereas it was not present in our previous study in lymphoma patients [15]. This may suggest the presence of necrosis in large tumours from solid origin. Other factors that may affect Annexin uptake are the intra- and peri-tumoural lymphocyte infiltration [30]. For the future, we plan to increase the number of HNSCC patients to undergo the TAVS imaging to prove whether or not it can be used for outcome prediction as it was shown in patients with malignant lymphoma and non-small cell lung cancer [14, 15]. In conclusion, co-registration of Annexin V scintigraphy with radiotherapy-planning CT scan showed a radiation-dose-dependent uptake in parotid glands, indicative of early apoptosis during treatment. The inter-individual spread in Annexin uptake in primary tumours could not be related to differences in treatment schedule or tumour volume, but the Annexin uptake in tumour and lymph nodes were closely correlated. This effect might represent a tumour-specific apoptotic response.

Med_Biol_Eng_Comput-3-1-2039845

A newly developed tool for intra-tracheal temperature and humidity assessment in laryngectomized individuals: the Airway Climate Explorer (ACE)

The aim of this study is to develop a postlaryngectomy airway climate explorer (ACE) for assessment of intratracheal temperature and humidity and of influence of heat and moisture exchangers (HMEs). Engineering goals were within-device condensation prevention and fast response time characteristics. The ACE consists of a small diameter, heated air-sampling catheter connected to a heated sensor house, containing a humidity sensor. Air is sucked through the catheter by a controlled-flow pump. Validation was performed in a climate chamber using a calibrated reference sensor and in a two-flow system. Additionally, the analyser was tested in vivo. Over the clinically relevant range of humidity values (5–42 mg H2O/l air) the sensor output highly correlates with the reference sensor readings (R2 > 0.99). The 1–1/e response times are all <0.5 s. A first in vivo pilot measurement was successful. The newly developed, verified, fast-responding ACE is suitable for postlaryngectomy airway climate assessment. Introduction Total laryngectomy results in a permanent disconnection of the upper and lower airways, and leads to an increase in chronic pulmonary complaints like frequent involuntary coughing, sputum production, and repeated daily forced expectoration in order to clear the airways [5]. Heat and moisture exchangers (HMEs) were developed in an attempt to compensate for the lost functions of the upper airway and have been found to reduce these symptoms and improve the quality of life [1, 4, 7]. In vitro, the moisture exchanging capacity of the HME can be specified with the ISO 9360 standard. However, this standard does not incorporate a standardized method for determining the heat exchange of the HME, and it is not possible to translate the results obtained with the lung model of the ISO 9360 standard to the true in vivo situation. Therefore several questions remain unanswered. Fundamentally: what is the in vivo heat and moisture exchanging capacity of the HME? And clinically, how does the in vivo heat and moisture exchanging capacity relate to that of the upper airway, i.e. where does the HME stand in the range from open stoma to nose breathing? Is a further improvement of the HMEs heat and moisture exchanging capacity possible and desirable, and does this result in a further decrease of the pulmonary complaints due to the use of this rehabilitation device? Measuring intra airway temperature and humidity is a technically complex issue [15]. Assessing the temperature and humidity variations during normal breathing requires fast sensor response characteristics. No appropriate commercial systems for this purpose are available, and only a few self-constructed validated systems have been published [6, 8, 10, 11, 13, 14]. Two groups evaluated the influence of an HME on intra airway temperature and humidity in laryngectomees. In one study [10] the sensor response time was up to 9 s and in the other study [8, 9] a response time of 2 s is reported which has been shown to be sufficiently fast for normal breathing volunteers, but which might be insufficient for patients. Our goal was to develop a verified intra-tracheal airway climate analyser with response characteristics fast enough for assessment of end tidal intra-airway temperature and humidity variations during normal breathing in laryngectomized individuals, and for evaluation of the influence of HMEs on these parameters. Design considerations Placement of a humidity sensor in the airway itself was rejected. Humidity sensors must be kept clean, and a construction to protect a sensor would be so large that the aerodynamics within the airway would be disturbed. A high airflow around the sensor is required to minimize the response time of the sensor. A constant airflow is necessary for a reliable interpretation of the time dependent signal, and in addition facilitates the use of deconvolution to optimize system response characteristics. Therefore, a sensor environment had to be created in which airflow remains constant and high during the whole breathing cycle, i.e. in a “chamber”. In order to disturb airway aerodynamics as little as possible, a small diameter sample catheter for transportation of breathing air outside the airway to a humidity sensor is required. We decided that the lumen surface of the sample catheter should not exceed 10% of the lumen surface of the trachea. A decrease in lumen area of maximal 10% will result, by the law of mass conservation, in an increase in local flow velocity of 10%. This on its turn will lead to a transbronchial pressure drop of maximal 20% which is supposed to be small enough to avoid substantial deformation of the trachea wall. This means that based on a diameter of 2 cm, which is representative for the trachea of an adult person, the maximal diameter of the sample catheter should not be more than 6 mm. In order to prevent condensation of water, the temperature of the measurement system must not be below the temperature of the breathing air. In addition, high humidity levels should be avoided to prevent hysteresis of the humidity sensor. Based on the maximal, i.e. end expiratory, temperature of about 36°C, with a relative humidity (RH) of about 97% [16], the breathing air must be heated to at least 38°C in order not to exceed 90% RH (http://www.thunderscientific.com/web_humicalc/index.php). On the other hand, a heated sample catheter should be thermally isolated in order to prevent patient discomfort and artificial heating of inspired air. We decided to limit the sample rate to less than 10% of the respiratory minute volume. Based on a breathing frequency of at least 12 cycles/minute and a tidal volume of 0.5 l for an adult human at rest, the sample rate should be 0.6 l/min at most. Finally the catheter should be easy to manipulate and it should be possible to insert it into the trachea even if an HME filter is in situ. The primary measuring site for the sensor should be about 1 cm behind the HME, but preferentially, measurements deeper in the trachea should be possible as well. Components Airway Climate Explorer Breathing air is sampled through a 0.8 mm diameter central canal of a multilumen tube (see Fig. 1a, b.). Four of the six surrounding canals are threaded with two resistance wire loops in order to facilitate catheter heating. Resistance wire was chosen to minimize current load in case of unexpected patient contact. Two surrounding canals were threaded with a copper wire loop in order to measure temperature for the system heating control unit. A thermocouple (MLT1402 T-type Ultra Fast Thermocouple Probe (IT-23), ADInstruments Ltd, Oxfordshire, UK) was placed just inside the tip of the central canal, in order to establish a constant airflow around this sensor. The thermal response of the thermocouple is specified as 5 ms, accuracy ±0.1°C. The distal 2 cm of the sample catheter is not actively heated in order to prevent retrograde heating of the thermocouple. The multilumen tube is surrounded by a silicone isolation tube with an outer diameter of 5 mm. In between both tubes, a small layer of air increases thermal isolation. The length of the sample catheter is 30 cm. Fig. 1Drawings of the sample catheter; a 3D overview with enlargement of the tip. b cross section (total diameter 5 mm) The proximal end of the catheter is connected to an aluminium sensor house, incorporating the humidity sensor (see Fig. 2). The humidity sensor most suitable for our purposes appeared to be the capacitive humidity sensor of a radiosonde (RS92K, Vaisala Oyj, Helsinki, Finland). The technical specifications report a response time of <0.5 s at 20°C and <20 s at −40°C, both at an airflow of 6 m/s. The humidity sensor is placed in an aluminium sensor house that is wrapped with copper wire for heating. The humidity sensor is mounted on a sensor beam and consists of a small chip with the sensor polymer and an aluminium wire that in the original radiosonde served as a heating wire, but which we use to measure the temperature of the humidity sensor in situ (see Fig. 3). Apart from the humidity sensor chip, the sensor beam also contains a temperature sensor, but this was not used, as it is located relatively far from the humidity sensor chip. Fig. 2Cross section of the sensor house; I connection to catheter, II humidity sensor, III copper wire for heating of the sensor house, IV droplet interceptor with electrode for fluid detection, V valve for disconnection of pressure gradient in case of fluid detection in the droplet interceptor, VI thermocouple wireFig. 3Details of the humidity sensor (dimensions 4 × 1.5 mm2) The sensor house is heated to avoid prolonged saturation of the sensor [12, 15]. The temperature is kept stable within 0.2°C. Therefore the output of the capacitive sensor which is sensitive for the relative humidity can be directly converted to absolute humidity. Because a radiosonde is not designed for long-term use, the sensor house was constructed such that it could be easily opened to replace the sensor beam. The diameter of the air canal that leads to the humidity sensor is 1.4 mm. The air inflow passes through a droplet interceptor in the sensor house in order to prevent sucked up sputum from being projected onto the sensor chip. In the bottom of the droplet interceptor an electrode is placed in order to detect fluid accumulation, in which case a mechanic valve instantly disconnects the pressure gradient through catheter and sensor house, which is established by a vacuum pump (DIVAC 0.6, Leybold Vacuum BV, Utrecht, the Netherlands). The flow rate is controlled and kept constant at 0.6 l/min, by a mass flow meter/controller (Smart Mass Flow Meter, type 5850S and Smart Control V1.4 Software, Brooks Instruments, Veenendaal, the Netherlands), which is placed between the sensor house and the pump. Figure 4 summarizes the airway climate explorer in a block diagram. In this figure also the custom-made electronic module is shown. The main functions of the module are: processing of the humidity sensor signal and its associated temperature signal, i.e. heating of both the sample catheter via the resistance wire, and the sensor house via the copper wire, each to their own adjustable set points. The electronic principles of the design of the module are described in the appendix. Fig. 4Block diagram of the airway climate explorer For patient safety, the total assembly is connected to a safety-isolating transformer. Data acquisition and signal processing All signals of the total assembly, i.e. voltage output of the sensor polymer and the aluminium wire on the humidity sensor of the ACE, temperature signal of the thermocouple, sample rate of the mass flow controller, and all ancillary equipment (see below), are read out simultaneously via a multichannel data acquisition system with additional software (Powerlab and Chart 5.4.1 software, ADInstruments Ltd, Oxfordshire, UK) on a PC. For deconvolution of the raw humidity output of the ACE, the software application LabVIEW (National Instruments Netherlands BV, Woerden, The Netherlands) is used. Cleaning and disinfection protocol Prior to every in vivo measurement, the sample catheter is cleaned and disinfected with the following procedure. First the catheter is rinsed with tap water. Subsequently the catheter is cleaned with disinfection hand soap (Baktolin Basic, Bode Chemie, Hamburg, Germany) and rinsed again with tap water. Then the catheter is placed in a solution of Biotex (Sara Lee, Utrecht, The Netherlands) and tap water for 10 min, after which it is placed in a 70% alcohol–water solution for 5 min. During the disinfection procedure, chemicals are prevented from entering the central canal of the sample catheter by establishing retrograde airflow, in order to protect the humidity sensor polymer. The central canal is not sterilized because during use there is continuous suction of air towards the vacuum pump. Finally the catheter is dried in room air. Ancillary equipment for frequent in hospital calibration For frequent, in-house, temperature calibrations a can of water is used in which the thermocouple is placed, simultaneously with a calibrated thermometer (Thermalite, Electronic Temperature Instruments Ltd, Worthing, UK), accuracy: ±0.4°C, acting as a secondary reference. By using water with melting ice or warm tap water, the full temperature range can be quickly calibrated. For frequent humidity calibrations, a small climate room was constructed. This climate room consists of a plexiglas box of 26 × 42 × 16 cm3, in which moisturized air can be brought via the distal end of the tube of a neonatal respiratory humidifier (MR730 Respiratory humidifier, Fisher & Paykel Healthcare, Papendrecht, the Netherlands). Air is mixed via an electromotor-driven propeller (25 cm diameter). The box has entries for the tip of the sample catheter and for a calibrated heated humidity and temperature sensor accuracy ±0.6°C and ±2.5% relative humidity (RH), (Testo B.V., Almere, the Netherlands), which is used as a secondary humidity reference. The system can be operated in ambient air (30–60% RH), can be connected to the hospital oxygen supply, and can be set to deliver dry or 100% humidified gas at room temperature. Assessment of step-response characteristics was performed in a two-stream system, which consisted of two tubes that were mounted side by side, connected to—in absence of compressed room air facilities—the hospital’s oxygen supply. Through one tube, dry oxygen flowed, and through the other the oxygen was heated and 100% humidified with the respiratory humidifier. Subsequently, the sample catheter tip was manually switched between both tube lumina. Leakage testing is performed by checking the time required to completely empty a flexible sack that is filled with a well-defined volume of air (1 l). This volume is pushed into the sack with a calibration syringe (Jaeger GmbH/VIASYS Healthcare GmbH, Hoechberg, Germany). Environmental temperature and humidity are monitored with the heated humidity and the temperature sensor. Calibration and verification Full calibration of the system was performed with one single sensor beam. However, due to the disposable design of a radiosonde and the possibility of damaging the sensor during the measurements, most likely several sensor beams will be used over time. Therefore, response time behaviour was checked also for a second sensor beam, and the calibration procedure was designed such that these characteristics can be easily determined for a consecutive beam. Operating temperature Raising the temperatures of the system, and therefore of the humidity sensor, above room temperature, we observed an initial decrease in the response time, but above 40°C the response time increased. Therefore 40°C was chosen as the operating temperature. At this temperature, the sampled air in the catheter reaches 40°C well before arrival in the sensor house, and therefore the influence of breathing temperature variations on sensor temperature is negligible. For measurements at 1 cm deep in the trachea the heated part of the tube does not enter the trachea so that artificial heating of the inspired air will not occur. Static calibration The temperature of the sensor is kept stable within 0.2°C. Therefore the output of the capacitive sensor, which is sensitive for the relative humidity, can be directly converted to absolute humidity. In a climate cabinet (Heraeus Votsch HC7020), the humidity output of the ACE was plotted against a clinically relevant range of absolute humidity values, as measured with the secondary reference sensor, see Fig. 5. Since this relation appeared to be linear, a 2-point calibration is considered appropriate for further frequent humidity calibrations in the hospital. Given the accuracy of the reference sensor and the somewhat limited range between the calibration points to avoid excessive condensation or temperature disequilibrium in the calibration device, the inaccuracy of the system for measuring absolute humidity values after the 2-point calibration is less than 5%. Fig. 5Output of the humidity sensor plotted against the secondary reference The thermocouple measuring the temperature at the tip of the sample catheter was 2-point calibrated against the values of the reference thermometer in ice water and water at body temperature. This procedure can be performed in the hospital, prior to every in vivo measurement. Given the accuracy of the reference thermometer, the accuracy of the system for measuring temperature values after a 2-point calibration is 0.3°C. Assessing the role of condensation in the measuring system In order to assess the necessity of heating the system in order to prevent condense formation within the device, step response measurements were performed with an unheated system and subsequently with the system heated to 40°C. With an unheated system, a delay of up to several seconds was observed in the humidity trace relative to the thermocouple trace after a downward step. The longer the sensor is kept in a high humidity environment, this delay increases; and it is indicative of the formation and subsequent evaporation of the condensate within the system. Even when simulating clinically relevant breathing frequencies, a noticeable delay of 0.15 s was observed. When measuring high humidity at operating temperature for a long time, after the downward step, a very fast response with a delay of less than 0.05 s was observed. This delay is caused by the transport of sampled air from the tip of the catheter towards the humidity sensor and is not relevant for the measurement of end tidal values. In the downward trace this fast response was followed by a shoulder at about 50% of the step, of which the length again increased with the duration at high humidity. This shoulder probably represents the evaporation of the condensate in the unheated tip of the sensor. During in vivo measurements, however, the system will never be exposed to high humidities for an extended period of time. Therefore we simulated a range of breathing frequencies with the two-stream system resulting in a series of block functions of approximately 9, 18, and 35 cpm (cycles per minute). At 9 cpm a short shoulder is still visible, but the condensate evaporates sufficiently fast that the “end-inspiratory” value is reached. At 18 and 35 cpm no shoulder was observed. Raw response time characteristics Using the two-stream system the step response of the ACE was determined. Both for the temperature and the humidity signal we determined the 1–1/e response time. For the downward humidity step we could not use a plain step function because condensation in the tip would occur. Therefore we used a block function with a frequency of 18 cpm (see Fig. 6a). At this frequency the response time is still relatively short compared with the period of the block function and easy to determine from the humidity trace. Fig. 6Response to a block function; solid black line raw humidity trace, solid grey line deconvolved humidity trace, dashed line thermocouple trace. Vertical axis: °C/mg H2O/l. Horizontal axis: seconds. a Response to a block function of approximately 18 cycles per minute. b Overview of the response to a block function with varying frequency (9 to 35 cpm) For the thermocouple, the response to increasing and decreasing temperature steps was <0.2 s. For the first humidity sensor, the response to increasing humidity and decreasing humidity was <0.5 and <0.8 s, respectively, at a frequency of 18 cpm. We verified that the response characteristics at other frequencies were comparable. Both rise and fall times of the second humidity sensor were approximately 60% shorter. The observed raw response times are slightly longer than we had expected based on the specifications of the Vaisala Humidity Sensor, probably because the airflow over the sensor is slower than specified. Although every attempt was made to minimize the volume of the sensor chamber we estimate that the airspeed over the sensor is between 2 and 3 m/s instead of 6 m/s. Deconvolution of the humidity signal Since the raw step response times of tested humidity sensors were considered too slow for patients with high breathing frequencies, it was decided to further process the raw humidity signal. Depending on the tested sensor, the raw step response can be closely approximated by a single or a double exponential function. Therefore it is possible to reconstruct the original signal by deconvolution. Although we observed a different response time for increasing and decreasing humidity steps, we nevertheless chose a single function because of its better numerical stability: For the removal of excessive noise a low pass filter of <300 cpm was applied. Because the raw response characteristics varied in between the humidity sensors and over time, τ was adjusted for each individual measurement. The principle of adjustment of these parameters was to acquire a maximal value for τ, such that no overshoot of the deconvolved humidity trace was visible. We used the thermocouple signal as reference for the true step function. For the first sensor, the deconvolved signal was stable and optimal for increasing humidity with τ = 0.4 s. For the second humidity sensor a τ of 0.1 s appeared to be appropriate. We verified that the raw step response characteristics were reproducible, and that the deconvolution results were not influenced by the step size. After deconvolution of the humidity signal, the response characteristics improved, see Fig. 6a. The 1–1/e step response time to a block function for the first humidity sensor was <0.2 s for increasing humidity and <0.4 s for decreasing humidity at a frequency of 18 cpm. For the second sensor the deconvolved step response characteristics were similar after deconvolution. Figure 6b shows the raw and deconvolved humidity response to a block function over a range of breathing frequencies. It is shown that the frequency dependence of the peak-to-peak values is reduced after deconvolution. At higher frequencies, we nevertheless see that the deconvolved signal increasingly deviates from the true peak-to-peak values. This deviation is more marked in the lower peak, due to the relatively long humidity fall time of the raw trace, compared to the rise time. The magnitude of the deviation in the lower peak ranges from about 1.5 mg H2O/l at a breathing frequency of 9 cpm to 5 mg H2O/l at 35 cpm. Since the step response characteristics of the deconvolved signal were comparable for both tested sensors, the deviations are of comparable magnitude too. At 35 cpm even the amplitude of the fast thermocouple signal does not reach the true peak-to-peak values. The deviation is less than 0.5°C so that we do not consider a correction necessary for the thermocouple signal. The in vivo measurement The total assembly and the influence of a heat and moisture exchanger were tested in one laryngectomized volunteer. The tip of the sample catheter was pushed 1 cm through a small hole that was punched in an HME plaster (Fig. 7). The diameter of the puncture was slightly smaller than the diameter of the sample catheter in order to avoid air leakage. Figure 8 shows a segment of the raw and deconvolved humidity trace and the thermocouple signal. In this particular volunteer τ required adjustment to 0.35 in order to eliminate overshoot. During breathing in rest, intra airway temperature and humidity ranged from 28/34°C and 15/37 mg H2O/l (typical end inspiratory/end expiratory values), respectively. During the total measurement session, breathing frequency ranged from 12 to 44 cpm (mean 17.8 cpm, SD 4.4 cpm). In addition, inspiration time may be relatively short compared to expiration time. See the beginning of the trace in Fig. 8. Fig. 7Sample catheter in situ, pushed through a punched hole in the HME plaster, without HME filter in situFig. 8In vivo humidity trace before and after deconvolution; black trace raw humidity signal, grey trace deconvolved humidity signal. Vertical axis: mg H2O/l. Horizontal axis: seconds Discussion Our results show that we have succeeded in developing an intra-tracheal airway climate analyser with response characteristics fast enough for assessment of end tidal intra airway temperature and humidity variations during normal breathing in laryngectomized individuals. In the literature, only a few systems have been described that simultaneously measure temperature and humidity in the human airway [6, 8, 10, 11, 13, 14]. Ingelstedt [6] reported on intra airway climate in healthy individuals using a well- validated system based on psychrometry, with signal stability 0.3–0.4 s after sudden exposure to increased humidity. However, this method had a few disadvantages. The wet thermo element remained sufficiently moistened for only 5–8 respiratory cycles, after which it had to be retracted in order to be watered by a moistened wick again. Furthermore, even the slightest contact with tracheal secretion increased the thermal inertia of the psychrometer, after which it has to be cleaned in a 4-h lasting procedure. Although contact of the humidity sensor of the ACE by tracheal secretion is less likely, the central canal of the sample catheter is susceptible to obstruction by tracheal secretion. When this would occur, the sample catheter can be disconnected from the sensor house and subsequently the obstructive plug can be removed by establishing reversed airflow. The electrodes in the bottom of the droplet interceptor appeared to work effectively. If the humidity sensor would be damaged, it can easily be replaced. Due to the “in hospital” calibration procedures, the measurements can be continued within a relatively short time span. The system of Primiano et al. [13] consisted of a sample catheter that was connected to a mass spectrometer. In order to prevent condensation, the tip of their sample catheter was constricted, establishing a decrease in air pressure. Extensive validation of the total assembly revealed impressive 10–90% response time characteristics of less that 0.25 s for humidity readings and less than 0.12 s for temperature readings, respectively. Since mass flow meters are expensive and require solid maintenance, the use of such a device was not considered feasible for our project. Although their measurement system has been used for assessment of anaesthesiology heat and moisture exchanger effectiveness, we are not aware that this system has been used for airway climate assessment in laryngectomees. McRae et al. [10, 11] and Keck et al. [8, 9] have assessed tracheal climate and the influence of HMEs in laryngectomees. The set up of Keck et al. consisted of a sample catheter, connected to a sensor house that incorporated a capacitive humidity sensor. As described in [9] the sensor house was heated with a heat mat to 34–35°C, which is slightly lower than the maximum end expiratory temperature observed (35.2°C) so that condensation still might be a possibility. However, the sample catheter was not heated and the verification of the total assembly has not been included in the report. In the present device, evidence of condensation—probably precipitated in the unheated distal 2 cm of the catheter tip (which is in vivo somewhat heated by the body) is observed when offering block functions at a frequency of 9 cpm. Despite this, at 9 cpm the end values were not affected. It is therefore unlikely that also during in vivo measurements condensation will affect the amplitude of the peak-to-peak values. Moreover, the temperature and humidity steps will occur more gradually in the in vivo situation, theoretically decreasing the amount of condense formation in the sensor tip, compared to the offered block function. In the studies by McRae et al. [10, 11], the response time of the humidity sensor used in the setup of McRae was up to 9 s. The frequency of human rest breathing is far too high compared to these response characteristics to obtain representative values for intra-tracheal humidity. However, as the two-stream block function experiment has shown, also in the present measuring device, the amplitude of the deconvolved humidity trace depends on the breathing frequency. The clinical implication of this finding is that, at higher breathing frequencies, the true peak-to-peak values are slightly, but progressively underestimated by the ACE, although probably less than in vitro due to more gradual humidity and temperature variation in the latter situation. The variation in end-inspiratory humidity values and temperature signals, as indicated by the deconvolved humidity signal and the thermocouple in the in-vivo experiment may therefore be partially due to a frequency dependency and not due to a real variation in humidity and temperature values. Quality control The first sensor functioned correctly for 1 year and the second for half a year. Both sensors had to be replaced because of accidental damage during cleaning. The two humidity sensors appeared to have different response characteristics. Sensor age, variation in the production process of the humidity sensor etc. may result in variation between the characteristics of different humidity sensors, but also between the characteristics of one sensor measured over time. Linearity between reference humidity values and sensor output in all consecutive sensors is felt to be a prerequisite for serial production of radiosondes. The small climate room and the 2-point calibration is therefore considered appropriate for daily quality control. A potential source of error is leakage of air, causing dilution of sampled tracheal air with room air. Therefore the simple leakage test should be part of the regular quality control. Although both tested humidity sensors revealed the different step response times of the raw humidity signal, the peak-to-peak amplitude of deconvolved signal was comparable between both sensors. The two-stream step response set up facilitates regular control of the sensor response characteristics and, if necessary, adjustment of the deconvolution function can be easily performed. Limitations and further improvements During measurements deeper in the trachea (e.g. 4 cm), contact of the catheter tip with the airway wall may lead to irresistible cough. Preferably the distal end of the sample catheter should be better pliable in the shape of the trachea, in order to facilitate better control of its tip and reducing airway wall contact. Also visual control of the catheter tip may contribute to prevention of wall contact. The response of the sensor when decreasing the humidity is slower than the response to increasing humidity. However, the same simple deconvolution function is used for increasing and decreasing humidity. A more advanced signal processing might further improve response characteristics and decrease the remaining frequency dependence of the peak tot peak amplitude. In particular for the end-inspiratory values this is important because of the relatively short time span of the inspiration time compared to the expiration time, combined with the relatively slow response characteristics to decreased humidity compared to increased humidity. In conclusion, the Airway Climate Explorer is an easy to use, relatively inexpensive tool for intra airway temperature and humidity measurements. When looking at the end-inspiratory and end-expiratory humidity values as measured by our system, the accuracy is adequate for assessment of intratracheal climate.

Eur_J_Appl_Physiol-4-1-2374881

Health-related physical fitness of adolescents and young adults with myelomeningocele

To assess components of health-related physical fitness in adolescents and young adults with myelomeningocele (MMC), and to study relations between aerobic capacity and other health-related physical fitness components. This cross-sectional study included 50 adolescents and young adults with MMC, aged 16–30 years (25 males). Aerobic capacity was quantified by measuring peak oxygen uptake (peakVO2) during a maximal exercise test on a cycle or arm ergometer depending on the main mode of ambulation. Muscle strength of upper and lower extremity muscles was assessed using a hand-held dynamometer. Regarding flexibility, we assessed mobility of hip, knee and ankle joints. Body composition was assessed by measuring thickness of four skin-folds. Relations were studied using linear regression analyses. Average peakVO2 was 1.48 ± 0.52 l/min, 61% of the participants had subnormal muscle strength, 61% had mobility restrictions in at least one joint and average sum of four skin-folds was 74.8 ± 38.8 mm. PeakVO2 was significantly related to gender, ambulatory status and muscle strength, explaining 55% of its variance. Adolescents and young adults with MMC have poor health-related physical fitness. Gender and ambulatory status are important determinants of peakVO2. In addition, we found a small, but significant relationship between peakVO2 and muscle strength. Introduction During the last decades, life expectancy of persons with myelomeningocele (MMC) has increased and many will nowadays survive into adulthood (Bowman et al. 2001). As a consequence, lifestyle-related diseases, such as cardiovascular disease and diabetes mellitus, will be of increasing concern in this patient group. Therefore, more attention towards a healthy lifestyle is warranted. Similar to the general population, persons with MMC develop their own lifestyle during the transition from adolescence to adulthood. At this age, special attention should be paid to optimize the lifestyle in order to improve health throughout life. Physical fitness is recognized as an important component of health (Lamb et al. 1988; Twisk et al. 2002) and it may be important for the performance of functional activities and quality of life (Noreau and Shephard 1995; Stewart et al. 1994). Low physical fitness may result in high physical strain during the performance of activities (Bruinings et al. 2007). As a consequence, activity levels may decrease due to fatigue and discomfort, exacerbating low physical fitness. Caspersen and co-workers defined several health-related components of physical fitness, i.e. aerobic capacity, muscle strength and endurance, flexibility and body composition (Caspersen et al. 1985). Only a few studies are available on health-related physical fitness in persons with MMC. In a previous study in adolescents and young adults with MMC, we found that average peak oxygen uptake (peakVO2) was 42% lower than normative values of healthy peers, with lower values in non-ambulatory than in ambulatory persons (Buffart et al. in press). Several other studies have also reported low aerobic capacity (Agre et al. 1987; Sherman et al. 1997; van den Berg-Emons et al. 2003). Furthermore, children with MMC, and particularly non-ambulatory children, were found to have reduced strength of lower extremity muscles (Agre et al. 1987; McDonald et al. 1991; Schoenmakers et al. 2004). Hip and knee contractures have been reported in respectively 18% and 36% of adolescents and young adults with MMC and hydrocephalus (Verhoef et al. 2004). Finally, high levels of body fat have been found in persons with MMC (Bandini et al. 1991; Mita et al. 1993; Shepherd et al. 1991); previously we reported that 35% of adolescents and young adults with MMC were obese (Buffart et al. in press). Because these previous studies lack simultaneous assessment of several health-related physical fitness components, the relation between aerobic capacity and other components remains unclear. Insight in these inter-relations may show for example, whether focusing on peripheral factors such as muscle strength would have additional value for improving aerobic capacity. Due to the scarcity of studies in persons with MMC, the first aim of this study was to describe health-related physical fitness i.e. aerobic capacity, muscle strength, joint flexibility and body composition in a relatively large group of adolescents and young adults with MMC, allowing descriptions of subgroups regarding gender and ambulatory status. Secondly, we wanted to study the relation between aerobic capacity and other components of health-related physical fitness, controlled for relevant personal and disease-related characteristics. Studies in persons with other chronic conditions such as spinal cord injury have shown relations between aerobic capacity and muscle strength (Haisma et al. 2006; Janssen et al. 1993; Zoeller et al. 2005). Therefore, also in persons with MMC, we expected aerobic capacity to be related to muscle strength. Method Participants Recruitment Adolescents and young adults with MMC, aged between 16 and 30 years, were recruited from the university hospitals in Rotterdam, Leiden, Utrecht and Amsterdam and all rehabilitation centers in the Southwest of the Netherlands. Exclusion criteria were complete dependence on an electric wheelchair, presence of disorders other than MMC that affect daily physical activity (e.g. rheumatoid arthritis), and presence of disorders that contra-indicate a maximal exercise test (e.g. exercise-induced ischemia or arrhythmias, uncontrolled hypertension and exercise limitation due to chronic obstructive pulmonary disease). We invited 171 persons of whom 50 participated in the study (29%). Main reasons for non-participation were no interest, lack of time or duration of the measurements. No differences were found between participants and non-participants regarding age, gender, level of lesion and presence of hydrocephalus, as measured with an independent t-test or Chi square test (Buffart et al. in press). All participants and parents of adolescents aged less than 18 years gave written informed consent before participating in the study. The Medical Ethics Committee of the Erasmus MC Rotterdam and of all participating institutes approved the study. Characteristics In total, 25 males and 25 females (mean age 21.2 ± 4.5 years) participated in the study. Table 1 presents personal and disease-related characteristics of participants. Neurological level of lesion and the presence of hydrocephalus were obtained from the medical records. Five categories of neurological level were distinguished: thoracic, thoracolumbar, lumbar, lumbosacral and sacral. We considered hydrocephalus to be present when a shunt was placed. Ambulatory status was determined according to the classification of Hoffer and co-workers (Hoffer et al. 1973): (1) community ambulator, walking indoors and outdoors, (2) household ambulator, walking only indoors, and (3) non(functional) ambulator. Non-functional ambulators walk only during therapy sessions and non-ambulators are completely wheelchair dependent. Since main mode of ambulation in daily life is similar between non-functional ambulators and non-ambulators, we combined these two groups. Educational level was categorized as low (pre-vocational practical education or lower level), medium (pre-vocational theoretical education and secondary education) or high (higher education and university) (Donkervoort et al. 2007). Table 1Personal and disease-related characteristics of participants (n = 50) Gender (% male)50Age (mean ± SD in years)21.2 ± 4.5Height (mean ± SD in m)1.57 ± 0.12Body mass (mean ± SD in kg)67.6 ± 15.6Level of lesion (%) Thoracic2 Thoracolumbar14 Lumbar28 Lumbosacral42 Sacral14 Hydrocephalus (%)82Ambulatory status (%)a Community ambulator30 Household ambulator14 Non (functional) ambulator56Educational level (%) Low37 Medium39 High24aNo gender differences as tested with a Chi square test Aerobic capacity Aerobic capacity was measured in a progressive maximal exercise test, based on the McMaster All-Out Progressive Continuous Cycling and Arm test (Bar-Or 1983), on an electronically braked arm or cycle ergometer (Jaeger ER800SH and ER800 respectively; Jaeger Toennies, Breda, The Netherlands). Studying patients with cerebral palsy who were partly wheelchair-dependent, Bhambani and co-workers concluded that maximal exercise testing during the main mode of ambulation elicits the highest oxygen uptake (Bhambhani et al. 1992). Therefore, depending on their main mode of ambulation, participants performed an arm crank test (n = 33) while sitting in their own immobilized wheelchair with cranks at shoulder height, or a leg cycle exercise test (n = 17). The test was preceded by a 3-minute warm-up (5 W for arm ergometry and 20 W for cycle ergometry), followed by a resting period of 5 min. During the test, resistance was increased every 2 min with a variable load, ensuring that total individual exercise duration ranged from 8 to 12 min. The pedal/crank rate was 60 rpm and strong verbal encouragement was given throughout the test. The test was terminated when the subject stopped due to exhaustion. Gas exchange was determined continuously using a breath-by-breath portable measurement system (K4b2, COSMED, Rome, Italy). Calibration was performed before each test with reference gases. Heart rate was measured continuously using a heart rate (HR) monitor which was attached to the system, and participants were fitted with a transmitter belt around the chest (Polar Electro, Finland). Aerobic capacity was defined as the mean oxygen uptake during the last 30 s of exercise (peakVO2, in l/min). In addition, for those measured during cycle ergometry, values of aerobic capacity were expressed as percentage of reference values of Dutch able-bodied sedentary males and females of similar ages, as estimated from a submaximal exercise test on a cycle ergometer using the nomogram of Åstrand (Vos 2001). Peak work load (peakW) was defined as the highest work load maintained for at least 1 min. In addition, the ventilatory anaerobic threshold (VAT) was estimated by the ventilatory equivalent method, when the ventilatory equivalent for O2 (Ve/VO2) and the end-tidal O2 partial pressure (petO2) increased while ventilatory equivalent for CO2 (Ve/VCO2) and end-tidal CO2 partial pressure (petCO2) remained stable (Reinhard et al. 1979; Wassermann et al. 1999). VAT was also expressed relative to the measured peakVO2 (VAT%). HR and respiratory exchange ratio (RER) were used as objective criteria for maximal exercise. Subjective strain was measured at the end of the final stage using the modified Borg scale of rating perceived exertion (RPE), which is a vertical scale labelled from 0 (no effort at all) to 10 (maximal effort) (Borg 1982; Mahler et al. 1987). Muscle strength We measured strength of two large muscle groups of the lower and upper extremity with a hand-held dynamometer (MicroFet, Hoggan Health Industries) using the “break” testing method. We measured strength of hip flexors and knee extensors in persons whose main mode of ambulation was walking. In persons whose main mode of ambulation was wheelchair-driving, we measured strength of shoulder abductors and elbow extensors. The positions and the performance of the measurements were according to van der Ploeg and co-workers (van der Ploeg et al. 1991). The applicator of the dynamometer was held against the distal part of the limb segment, and participants were asked to build up their force to a maximum against it. The examiner applied sufficient resistance just to overcome the force exerted by the participant, and the applicator was then immediately moved away from the limb segment and the measured force was recorded. Each trial lasted approximately 4 s, and three repetitions were performed with 1 min rest in between. We used the average value of three repetitions of the dominant side for further analyses because we found no differences between the dominant and non-dominant side (tested with a paired samples t-test). To assess whether muscle strength was subnormal and to be able to compare strength of upper and lower extremity muscles, values were normalized to Z-scores using reference values of healthy males and females. For hip flexors, shoulder abductors and elbow extensors we used reference values of Phillips and co-workers for males and females aged 20–29 years (Phillips et al. 2000) and for knee extensors, we used reference values of Bohannon and co-workers (Bohannon 1997). In case the examiner could not resist the muscle strength, Z-score was set at 2. We used the lowest Z-score of the upper or lower extremity as indicator of muscle strength, and we considered muscle weakness to be present when Z-score ≤−2. Flexibility As indicator of flexibility of lower extremity we assessed passive mobility of hip and knee joint while participants were lying supine and of the ankle joint while they were sitting. We considered mobility to be restricted when extension of hip and knee joint and ankle dorsal flexion did not reach neutral position. Since no differences were found in mobility restrictions between dominant and non-dominant side (tested with the Wilcoxon signed rank test) we used the results of the dominant side to calculate a sum score of joint mobility ranging from 0 (no mobility restrictions in any joint) to 3 (all three joints have mobility restrictions). Body composition Height was measured with a flexible tape while lying on a bed. In case of joint contractures, measurements were performed from joint to joint. Body mass of ambulatory persons was obtained while standing on a Seca scale and of non-ambulatory persons while sitting on an electronic scale (Cormier, France). Thickness of four skin-folds (biceps, triceps, subscapular, and suprailiac) was measured twice on the right side of the body with a Harpenden caliper (Burgess Hill, UK) and mean values were used for further analyses. In addition, sum of four skin-folds were expressed as percentage of reference values of Dutch sedentary males and females of similar ages (Vos 2001). Statistical analysis Results of health-related fitness components are presented as mean ± standard deviation (SD) for the total group and for subgroups regarding gender and ambulatory status. Simple linear regression analyses were used to study relations between personal and disease-related characteristics and peakVO2 (in l/min), in order to detect relevant determinants to include in the multiple regression models. Multiple linear regression analyses for peakVO2 were carried out in two steps. First, we built a model including relevant personal and disease-related characteristics. In the second step, we studied whether other health-related physical fitness components, i.e. muscle strength (Z-score), sum of four skin-folds (mm), and joint mobility (number of restricted joint), were significantly related to peakVO2, controlling for the personal and disease-related characteristics from step 1. We presented the standardized regression coefficients (β) and explained variance (R2) of the linear regression models. Statistical analyses were performed using SPSS 12.0 for Windows. P ≤ 0.05 were considered significant. In the analyses, we pooled the data of peakVO2 measured during arm and cycle ergometry, and corrected for differences in exercise mode. Due to high collinearity between type of ergometer and ambulatory status (correlation coefficient (r) = 0.92, P < 0.001), we adjusted the analyses for ambulatory status as proxy for type of ergometer, because clinically, ambulatory status would be more meaningful. Furthermore, due to overlap between lesion level and ambulatory status (r = 0.58; P < 80.001), we chose to only correct for ambulatory status in the multiple regression analyses. Ambulatory status was determined during the study, whereas level of lesion was obtained from medical records. The national study on adolescents with spina bifida in the Netherlands (ASPINE) reported that the level of lesion as mentioned in medical records may be unreliable because lesions are determined at different ages, sometimes using the motor level and sometimes using the sensory level, and often lacking descriptions of methods (Verhoef et al. 2004). Results Descriptive results of aerobic capacity, muscle strength, joint mobility and body composition for the total group and for subgroups regarding gender and ambulatory status are presented in Table 2. Average peakVO2 was 1.48 ± 0.52 l/min. For persons measured during cycle ergometry (n = 17), peakVO2 corresponded to 67 ± 15% of reference values. Sixty-one percent of the participants had subnormal muscle strength as indicated by Z-scores and 61% had mobility restrictions in one or more joints. Average sum of four skin-folds was 74.8 ± 38.8 mm, corresponding to 159 ± 77% of normative values (Table 2). According to objective and subjective criteria most participants reached their peak exercise performance. Average peakHR was 174 ± 19 beats per minute, which was 90.4 ± 9.6% of the age predicted maximum (220—age for cycle ergometry; 210—age for arm ergometry) and average peakRER was 1.17 ± 0.22. Average RPE was 6.2 ± 2.2, indicating that participants experienced the exercise as heavy to very heavy. Table 2Descriptive results of health-related physical fitness components for the total group and for subgroups regarding gender and ambulatory statusAll (n = 50) GenderAmbulatory statusMale (n = 25) Female (n = 25) Community (n = 15) Household (n = 7) Non(functional) (n = 28) Aerobic capacity (mean ± SD) PeakVO2 (l/min) 1.48 ± 0.521.78 ± 0.511.18 ± 0.301.85 ± 0.571.44 ± 0.451.29 ± 0.40% of reference valuesa67 ± 1571 ± 1361 ± 1868 ± 1654 ± 2–PeakVO2 (ml/kg min) 22.6 ± 8.228.1 ±  7.017.0 ± 4.729.0 ± 7.722.3 ± 6.619.2 ± 6.8Peak oxygen pulse (ml/bpm)8.7 ± 3.010.1 ±  2.87.3 ± 2.410.7 ± 2.87.8 ± 2.47.8 ± 2.8PeakRER1.17 ± 0.221.17 ± 2.281.18 ± 0.201.15 ± 0.221.27 ± 0.161.16 ± 0.24PeakW (W)91 ± 42113 ± 4369 ± 28123 ± 4297 ± 3573 ± 34PeakHR (bpm)174 ± 19179 ± 16169 ± 20173 ± 21183 ± 14172 ± 18PeakHR % of predicted maximum90 ± 1092 ± 889 ± 1087 ± 1095 ± 891 ± 10VAT (l/min)1.20 ± 0.431.39 ± 0.441.01 ± 0.321.55 ± 0.451.07 ± 0.291.05 ± 0.34VAT%82 ± 1580 ± 1486 ± 1684 ± 1077 ± 2283 ± 16Muscle strength (mean ± SD) Z-score −2.1 ± 1.8−2.3 ± 2.1−1.9 ± 1.5−2.7 ± 2.2−2.0 ± 1.6−1.8 ± 1.7Weak strength: Z-score ≤ −2 (%)615864795754Joint mobility (median [range]) Number of restricted joints 1 [0–3]1 [0–3] 1 [0–3] 0 [0–2] 1 [0–2] 1.5 [0–3] Impaired mobility in any joint (%)615467295782Body composition (mean ± SD)Sum of four skin-folds (mm) 74.8 ± 38.851.2 ± 24.6100.4 ± 35.159.1 ± 29.266.5 ± 34.786.0 ± 42.0% of reference values159 ± 77146 ± 79173 ± 73121 ± 52160 ± 101181 ± 75aOnly for those measured on the cycle ergometer, n = 17 (ten males, seven females) Several personal and disease-related characteristics were related to peakVO2 (Table 3). PeakVO2 was higher in males than in females (β = −0.61; P < 0.001), higher in community ambulatory persons than in household and non-ambulatory persons (β = −0.48; P < 0.001) and higher in persons with a lower level of lesion (β = −0.43; P = 0.002). Age, presence of hydrocephalus and educational level were not related to peakVO2. Table 3Regression models for aerobic capacityIndependent variablesAerobic capacity (peakVO2, in l/min)β P-valueR2Simple regression analysis Personal characteristics Gender−0.61<0.0010.36 Age−0.07 0.65– Lesion level−0.430.0020.17 Hydrocephalus−0.18 0.22– Ambulatory status−0.48<0.0010.22 Educational level0.130.39–Multiple regression analysis Step 1a  Personal characteristics0.50  Gender−0.55<0.001  Ambulatory status−0.40<0.001 Step 2b  Inter-relations  Muscle strength (Z-score)0.220.040.55  Body composition (sum of four skin-folds in mm)0.250.080.53  Joint mobility (number of restricted joints)0.000.98–Significant betas are presented in boldβ standardized regression coefficient, R2 explained variance, Gender male (0), female (1), Lesion level sacral (1), lumbosacral (2), lumbar (3), thoracolumbar (4), thoracic (5), Hydrocephalus no (0), yes (1), Ambulatory status community ambulator (1), household ambulator (2), non(functional) ambulator (3), Educational level: low (0), medium (1), high (2)aDue to large overlap between ambulatory status and lesion level, we only included ambulatory statusbRelations between aerobic capacity and other fitness components controlled for gender and ambulatory status Fifty percent of the variance in peakVO2 was explained by gender and ambulatory status (Table 3). In addition, when controlling for both variables, we found that participants with higher muscle strength had higher values of peakVO2 (β = 0.22; P = 0.04) explaining an additional 5% of the variance in peakVO2. Furthermore, we found that participants with higher sum of four skin-folds tended to have higher values of peakVO2 (β = 0.25; P = 0.08). Discussion Components of health-related physical fitness In the present study, several health-related components of physical fitness were studied simultaneously in a relatively large group of adolescents and young adults with MMC. In general, most participants had poor health-related physical fitness. Compared to the general population and compared to other patient groups, persons with MMC had low aerobic capacity. Regarding community ambulatory persons with MMC, values of peakVO2 were 32% lower than the reference values for able-bodied people and 23% lower than the average peakVO2 of males with spastic diplegia measured during cycle ergometry (Lundberg 1978). However, peakVO2 is influenced by the amount of active muscle mass (Davies and Sargeant 1974,1975; Lewis et al. 1983), which may possibly be reduced due to paresis of lower extremity muscles. PeakVO2 in non-ambulatory persons with MMC was lower than in ambulatory persons which may be caused by the lower amount of active muscle mass during arm ergometry compared to cycling. In able-bodied people, arm exercise induces a peakVO2 ranging from 53 to 73% of that achieved with lower extremity exercise (Bar-Or and Zwiren 1975). If that ratio is also applicable to non-ambulatory persons who may be accustomed to using their arm and shoulder muscles, adapted values of average peakVO2 would range between 1.76 and 2.43 l/min, which is of comparable range to the peakVO2 in ambulatory persons with MMC. Furthermore, compared to males with spinal cord injuries with lesions below T10 measured during arm ergometry (Janssen et al. 2002), the peakVO2 in non(functional)-ambulatory males with MMC was 22% lower. We therefore assumed that the poor aerobic capacity we found in persons with MMC may be influenced by reduced active muscle mass, but also deconditioning is likely to be present. This is supported by previous studies showing that adolescents and young adults with MMC were inactive, and that inactivity was associated with lower aerobic capacity (van den Berg-Emons et al. 2001), particularly in non-ambulatory persons with MMC (Buffart et al. in press). VAT is another indicator of aerobic capacity, which was strongly correlated to peakVO2 in the present study sample (r = 0.84; P < 0.001). In the general population, VAT roughly corresponds to 50–60% of VO2max during leg exercise (Davis et al. 1997), and values of 40–60% have been found in the able-bodied population during arm exercise (Flandrois et al. 1986; Lin et al. 1993; Schneider et al. 1999). Paradoxically, we found VAT to be at 82% of peakVO2 suggesting high aerobic capacity. Comparably, Coutts and McKenzie found that persons with tetraplegia had lower VAT than persons with paraplegia caused by smaller muscle mass, but they had higher VAT%, and it was suggested that values of peakVO2 were low in relation to VAT (Coutts and McKenzie 1995). Fifty-four percent of the non-ambulators and 79% of the community ambulators had subnormal strength in at least one of the major muscle groups. The relatively better performance of non-ambulators compared to ambulators in this respect might be explained by their relatively higher muscle strength in the upper extremities due to the habituation of using their arms for wheelchair propulsion. Furthermore, in contrast to lower extremity muscles, the neurological level of lesion will not disturb innervations of upper extremity muscles, which suggests that the weak upper extremity strength we found in non-ambulators is due to atrophy of muscles as a consequence of disuse. Weak strength of lower extremity muscles in ambulatory persons with MMC corresponds to previous findings in children with MMC (Agre et al. 1987; McDonald et al. 1991; Schoenmakers et al. 2004) and might be related to disturbed innervations as well as to disuse of muscles. Results of the present study support the findings of Verhoef and co-workers that a large proportion of persons with MMC have reduced joint mobility (Verhoef et al. 2004). Similar to the results of Agre and co-workers restrictions in joint mobility in the lower extremities were mainly present in non-ambulatory persons with MMC (Agre et al. 1987). We found high sum of four skin-folds, particularly in females and non-ambulatory, which is in agreement with high levels of body fat reported in previous studies (Bandini et al. 1991; Mita et al. 1993; Shepherd et al. 1991). Relations between aerobic capacity and other components of health-related physical fitness The results of the regression analyses indicated muscle strength was associated to peakVO2 when controlling for gender and ambulatory status. This finding is in accordance with literature on persons with spinal cord injury (Haisma et al. 2006; Janssen et al. 1993; Zoeller et al. 2005). In contrast to gender and ambulatory status, muscle strength is modifiable. Strength training may increase muscle mass and thus metabolizing mass contributing to higher peakVO2 (Janssen et al. 1993). In persons with spinal cord injury, strength training resulted in increased peakVO2 (Cooney and Walker 1986). Because causality cannot be established with the cross-sectional design of the present study, future studies should confirm whether strength training results in increased aerobic capacity in persons with MMC. However, considering the small, but significant, contribution of muscle strength to the explained variance, we assume this specific effect to be small. In this respect, it seems that mainly aerobic training is needed in order to improve aerobic capacity; however, including strength training may have additional value. Body composition tended to be positively related to aerobic capacity, indicating that persons with more body fat had higher values of peakVO2. This relation may be caused by the greater fat-free mass of overweight persons concomitant with a greater body size (Unnithan et al. 2006). Joint mobility was not related to aerobic capacity. However, in non-ambulators we might have underestimated the relation because we did not measure mobility of upper extremity joints. Nevertheless, good flexibility of lower extremity is suggested to be important to prevent problems later in life, such as problems with personal hygiene and transfer capabilities (Agre et al. 1987). Limitations of the study The methodology of the study had some limitations. First, because it is suggested that the primary mode of ambulation elicits the highest values of peakVO2 (Bhambhani et al. 1992), we used different exercise modes to assess aerobic capacity for ambulatory and non-ambulatory persons. However, due to large differences in active muscle mass, arm and leg exercise have different physiological responses. Therefore, we corrected for ambulatory status as proxy measure for type of ergometer when analysing relations between aerobic capacity and the other health-related fitness components. Furthermore, because the main mode of ambulation of ambulatory persons is walking rather than cycling, we may have underestimated peakVO2 in some ambulatory persons with MMC who rarely cycle. However, in clinical practice, the cycle ergometer seems to be more practical because physically disabled people may experience severe balance problems on the treadmill, and on the cycle it is possible to strap the feet to the pedals (Lundberg 1978). Peripheral local fatigue may have caused exercise cessation before reaching maximum oxygen uptake, however, based on the objective (peakHR and peakRER) and subjective criteria of maximal exercise (RPE), it may be concluded that values of peakVO2 are reasonable. Furthermore, muscle strength was measured using hand-held dynamometry. This method is known to be cheap, quickly applicable (van der Ploeg et al. 1984) and reliable (Bohannon 1997). However, for an average examiner, values above 250 N are considered too high to apply sufficient resistance (van der Ploeg et al. 1991), which may lead to less accurate results in strong muscle groups. Using absolute values of strength measured with an isokinetic device, instead of expressing muscle strength as Z-score, might provide more detailed insight into the relation between strength peakVO2. However, the current measurement method and use of Z-scores are considered adequate to determine weakness of major muscle groups. Finally, the response rate was low, which hampers generalization of results. However, personal and disease-related characteristics did not differ between participants and non-participants. Moreover, the prevalence of middle-level (lumbosacral) and high-level (lumbar or thoracolumbar) lesions of the present study sample was comparable to the persons who participated in the national ASPINE study (Verhoef et al. 2004). Despite, a selection bias may have occurred since it could be that the more active and more fit people had higher interest in participating than the less active and less fit ones, which may have led to an overestimation of health-related physical fitness components. In conclusion, the results of the present study show that both ambulatory and non-ambulatory adolescents and young adults with MMC have poor health-related physical fitness. A large part of the variance in aerobic capacity is explained by gender and ambulatory status. Results showed a small but significant relationship between peakVO2 and muscle strength, suggesting that adding strength training to aerobic training may have additional value in increasing peakVO2.

Soc_Sci_Med-2-1-2430159

Most of our social scientists are not institution based… they are there for hire—Research consultancies and social science capacity for health research in East Africa

There is a serious shortage of senior African social scientists to lead health-related research in Africa. This is despite the existence of many African social science graduates, and decades of Northern funded research programmes intended to develop local capacity. To investigate the barriers to developing health social science research capacity in East Africa, 29 in-depth interviews, informal conversations and a group discussion were conducted with professionals in this field. Introduction There is a serious shortage of senior African social scientists to lead or manage health-related research in Africa (World Bank, 2000). This is despite the graduation of many African social scientists, and decades of Northern funded research programmes intended to develop local capacity (Nchinda, 2002; Simon, 2000). Whilst weak research capacity probably affects all areas of health, the HIV epidemic ‘offers a supreme test of how effectively African universities can respond to emerging challenges …’ (Zeleza, 2003, p. 84). Yet the shortage of senior social scientists is particularly apparent in sexual health. For instance, large-scale HIV/AIDS research programmes in both Tanzania and Uganda have been unable to recruit local social scientists to senior posts on international salaries, despite having trained local junior social scientists for over 10 years. Furthermore, international debates on social dimensions of sexual health, such as the hypothesis of permissive African sexuality (Caldwell, J, Caldwell, P, & Quiggin, 1989), are dominated by Northern academics, despite their sensitivity. Health-related social science research capacity is the ability to investigate and define the social dimensions of health problems, set objectives, identify solutions and build sustainable institutions (cf. Sitthi-amorn & Somrongthong, 2000). Limited capacity is problematic at several levels. Most obviously, it requires non-local researchers generally unfamiliar with local life, reliant on interpreters, and prone to cultural misunderstandings with local fieldworkers. Service providers and policy makers have to base decisions on more superficial analyses, but ex-patriate-initiated research is less likely to have such practical application anyway (Costello & Zumla, 2000). At the broadest level, limited social science capacity restricts intellectual sovereignty (Zeleza, 2003), undermining political autonomy (RAWOO, 2002; Sitthi-amorn & Somrongthong, 2000). The main explanations for limited research capacity have been identified in the literature as: inadequate resources for education at every level (Nchinda, 2002; Sall, 2003; Sitthi-amorn & Somrongthong, 2000); the drain of expertise to the North (Pang, Lansang, & Haines, 2002; Ramsay, 2002; Sall, 2003; Zeleza, 2003); dependence on Northern research funding (Jentsch & Pilley, 2003; Lansang & Dennis, 2004); inequitable access to the literature (Lansang & Dennis, 2004); unbalanced North–South research collaborations (Costello & Zumla, 2000; Jentsch & Pilley, 2003) and poor support from government (Nchinda, 2002; Sall, 2003; Sitthi-amorn & Somrongthong, 2000). Some see the perpetuation of inadequate research capacity as replicating the imbalance in global trade relationships (Zeleza, 2003) and essentially semi-colonial (e.g., Costello & Zumla, 2000). Others assume the good intentions of funders and research partners, but identify the perverse consequences of North–South collaborations (Edejer, 1999), such as poaching senior researchers from local institutions. Either way, limited research capacity in Africa should be an ethical issue for Northern researchers working there. Whilst the main causes of weak research capacity are clearly macro-economic, the processes of conducting research might contribute to the problem and may be more readily modified (Green, 2003). Some reports have suggested that research consultancies, whilst augmenting meagre incomes, might marginalise teaching and research and undermine social science scholarship (Kwesiga, Mbago, & Chimanikire, 2000; Menken, Blanc, & Lloyd, 2002; Mkandawire, 1998; Sall, 2003). Being highly prescribed, consultancies have also been said to exacerbate the way African social science research is narrowly policy-bound (Allen, 1986; Sall, 2003), and increase Northern dominance of the research agenda (Mkandawire, 1998). ‘…consultants do not frequently choose to contradict the donor's agenda, … this would decrease their chances of … consultancies in the future.’ (Rossi, 2004, p. 27). Consultancies have been described as particularly problematic in East Africa (Sall, 2003), the University of Nairobi being called a ‘consultancy university’ (Allen, 1986, p. 25). There have been many attempts to strengthen research capacity in developing countries, leading sponsors being the WHO and other UN agencies (e.g., the Special Programme for Research and Training in Tropical Diseases (TDR); UNDP/World Bank/WHO, 2003), national development agencies (e.g., Swiss, Canadian and Japanese), foundations such as Rockefeller and NGOs such as the Population Council. However, there is little evidence about the most effective approach (Simon, 2000), and debates continue over, for instance, investing in individuals or institutions (Costello & Zumla, 2000; Nchinda, 2002), whether post-graduate training in the North exacerbates the brain drain (Nchinda, 2002), and Southern control of research budgets (Lansang & Dennis, 2004; Nchinda, 2002). The role of consultancies, however, has received scant attention and almost exclusively in the grey literature. In order to investigate the poor capacity for health-related social science research in East Africa, the processes perpetuating it and possible ways to improve it, a small-scale exploratory study was conducted in Tanzania, Kenya and Uganda. The general findings have been reported elsewhere (Wight, 2005). This short report focuses on the individualised nature of research activity and the role of individual research consultancies in shaping research capacity. Methods In 2003 and 2004, I conducted in-depth interviews with 29 leading professionals conducting, commissioning or supporting health-related social science research in East Africa (Table 1). The findings are biased towards Uganda, 18 interviewees being Ugandan, four Kenyan, three British, two North American, one Tanzanian and one Nigerian. This was primarily a snowball sample including seven senior social scientists from Makerere, the oldest university in East Africa with by far the largest research function in Uganda, and others from the London School of Hygiene and Tropical Medicine (three), the Universities of Nairobi (two) and Dar es Salaam (one) and from the leading independent research centres and research-supporting NGOs in Uganda and Kenya. None of the new universities were represented. To protect respondents’ anonymity small institutions are not named. The interview schedule covered leadership of local social science research, training and career paths, ways of strengthening research capacity and barriers to this. The full schedule is available (as Appendix A). Informal conversations were held with five senior and one junior researcher from the University of Dar es Salaam and National Institute for Medical Research, Tanzania, and nine junior researchers from Makerere and the MRC Programme on AIDS in Uganda. A group discussion was held with four of the Ugandans (three men, one woman), following the same schedule used for the in-depth interviews. The interviews were summarised according to analytical themes. There was considerable concordance in accounts, which were not patterned by gender. Divergent opinions are considered in the Discussion. Interviewees were circulated the main report for comments and to confirm that their views were presented accurately. Four provided comments. Findings Severity of the problem Nearly all those interviewed thought that there is a serious shortage of social science research capacity in East Africa, the few really good social scientists being overworked and overwhelmed with requests for collaboration. Most academic health-related social science research in Uganda was said to be run by Northerners, yet Uganda was thought to have stronger capacity than Kenya, with Tanzania coming third. Particular limitations identified were in qualitative research, analysis and writing skills, and health-related specialisms. Interviewees stated that the vast bulk of social science research in East Africa is commissioned by NGOs or government departments, mostly funded from the North; consequently, it is highly applied and determined by external priorities. The few opportunities for academic research were said to come primarily from Northern researchers who win the funding, resulting in unbalanced collaborations. Research processes: inter-collegiate support Poor social science capacity was primarily related to under-development and global economic inequalities: very poor schooling, talented students choosing high status vocational courses, poor university facilities and teaching, research funded through Northern institutions, and the drain of senior researchers abroad (Wight, 2005). The problem has been exacerbated by the death of many junior and mid-level Researchers from AIDS (Pfau & Barton, 2004, RAWOO, 2002, Zeleza, 2003). However, the individualised nature of departments and lack of collegiate support were also identified as unhelpful. This was primarily attributed to lack of resources and staff's reliance on individual research consultancies, resulting in no writing skills training and limited publishing experience to share. A head of department observed that senior staff rarely co-author papers with junior colleagues, due to ‘the culture of individualism.’ In her department there was no formal system to support junior researchers, though she was planning a mentoring system. Staff rarely seem to comment on their colleagues’ draft papers; one senior respondent estimated that at Makerere only 1% of colleagues would have time. Research processes: consultancies Most of the research work conducted by social scientists in East Africa is in the form of consultancies. The proportion of academics’ time spent on them is unclear (Kwesiga et al., 2000), partly perhaps to disguise this from supervisors, but most estimates were around 50% of working time. Teaching takes up much of the rest, with very little for academic research.… in Makerere you can spend your entire time just working on very well paid, short-term consultancy studies for NGOs, … who want something done in three weeks, and will pay you very well … (Senior researcher, previously Uganda) Extremely low university salaries create a powerful incentive for consultancies. A research associate's salary might be $250/month, while consultancies can pay $100–$250/day. In one research institute consultancies augment salaries from around $400/month to about $5000. A head of department explained: ‘… to rely on your salary would never make ends meet at all.’ Furthermore, in contrast to regular salaries, most researchers can avoid declaring consultancy fees for tax (30%). Research commissioners, predominantly government departments or NGOs, usually seek a contract with individuals, or sometimes consultancy firms, but rarely with university departments. Private consultancy firms, often constituted for a particular brief, usually employ university staff to help with the bid and subsequent research. Commissioning bodies are reportedly unwilling to pay overheads to institutions, and when they do, they are generally very low, e.g., 5–20% in Makerere departments, 20% at the University of Dar es Salaam, and a maximum of 15% at a Kampala independent research centre. The senior management at Makerere were said to encourage departments to become consulting firms and demand 30% overheads, but this leads university staff to work independently, undercutting university departments and earning more. The predominance of research consultancies is critical for the development of research capacity. Financial insecurity leads researchers to take on any work available, and consequently:There are no research traditions being developed …. We are social scientists but very few are specialists … (Faculty dean) Consultancy work also inevitably restricts academics’ time for teaching and supervision. At Dar es Salaam and Makerere consultancies should not interfere with normal academic work, but this is difficult to enforce:He will leave you. And who loses? This is the person you have trained up to PhD level, and now he is leaving you, and you have no one to teach … (Head of department) Writing consultancy reports provides little incentive to develop analytical skills. Reports generally involve very tight timetables with little opportunity for peers’ critical input, are descriptive and have limited dissemination (sometimes for internal use only). Several researchers said they do not publish from consultancies because they need the funder's permission, but none knew of it being refused. More plausibly, there is rarely time for such writing. Consequently, the CVs of highly experienced researchers often list numerous consultancy reports but very few journal publications, jeopardising their applications for senior jobs. The conflict between consultancies and academic publications reportedly generates a professional culture in which: ‘the point is to try and chase the quick money, and not take advantage of the chance of academic growth … people don’t value it very much.’ A faculty dean commented:Consultancies is not building the capacity of the person who is doing it. …. [Some] have even refused scholarships to do PhDs because they were busy doing consultancies. Only two interviewees questioned the inevitability that consultancies detract from publications or teaching: ‘… consultancies can … be a source of writing … [and] training.’ (Director large research programme). The high remuneration from consultancies, and tensions with teaching responsibilities, might encourage researchers to become full-time consultants. However, few do this because commissioners seek the ‘recognition and visibility’ of ‘high powered people’ in established university posts. Furthermore, it would be too insecure: ‘You would earn much more in the short-term, but then you would be unemployed in the long-term.’ Strengthening research capacity and likely barriers Respondents proposed many ideas to strengthen health-related social science research capacity (see Wight, 2005). Here I focus on those to modify the individualism of research practice and consultancies. Five senior interviewees identified the need to develop writing skills, for instance ‘to guide you through … the very complicated processes … and requirements’ to publish in international journals. Suggestions included experienced and in-experienced staff co-authoring, mentoring systems and support networks. The director of a research-facilitating NGO advocated posts dedicated to writing support, but with salaries adequate to prevent appointees taking on consultancies. The potential advantages of institutional research consultancies were explored, and in particular establishing a norm of significant overheads, e.g., 30%. Everyone approved in principle. Overheads could be used for: libraries, computing and internet access; department-initiated research; disseminating reports; training staff and developing writing skills. Institutional consultancies might facilitate a more collective approach to research and assist management by departmental heads. Furthermore, paying overheads might benefit commissioners since they could require reports to be published, at least in an on-line journal. However, several objections to institutional consultancies were also raised.The culture of institutionalising things is not there. Many think the institution is a barrier to them. And … the bureaucracy, you know, many people would prefer to have the money in their own accounts …. (Head of department) Researchers anticipated the frustrations of inefficient institutional administrations, with long delays in finalising contracts or being paid. It was feared that, since some universities do not allow departmental bank accounts, the central administration might appropriate funds raised through departmental consultancies. Furthermore, fees would not only have to be shared with the institution, but would have to be declared for tax. Consequently, the director of a large programme thought: ‘… people will just try to get around it. They will get consultancies privately.’ Several interviewees said that commissioning agencies would not ‘… want to pay the institutional fee.’All the American universities [have] institutional overheads, but tell DfID that [they] have to be factored in [in Kampala] …: “Oh, no!” …. How am I supposed to run the project without institutional overheads? … they have the mentality that they can do it on the cheap. Africa is poor, but it is not cheap! (Director of large programme) It was also argued that some agencies want to commission specific individual researchers, and that individual consultancies incentivise good work produced on time. Some large donors practice their policies of strengthening institutional capacity by only contracting research through institutions, e.g., the Carnegie and Rockefeller Foundations, the World Bank, and the Swedish and Norwegian development agencies. However, some interviewees thought a concordat with all commissioning agencies to pay minimum overheads would be unrealistic, since East Africa is too dependent on donors. Discussion Although most respondents were Ugandan, the data from Kenya and Tanzania suggest that these findings apply across East Africa, while the broader literature (e.g., Carlsson & Wohlgemuth, 1996; Sall, 2003; Zeleza, 2003) and contacts with researchers elsewhere suggest they are relevant to much of sub-Saharan Africa. There is a serious shortage of health-related social science research capacity in this region, as evidenced by the Northern intellectual leadership of most academic research. This perpetuates ‘the international intellectual division of labour whereby African … social scientists … import appropriate … theory and, at best, export empirical data.’ (Zeleza, 2003, p. 111) African countries’ limited ability to define for themselves their problems and the solutions may have very practical consequences. For instance, Cleland and Watkins (2006, p. 2) argue that Africans’ frustratingly slow response to the HIV epidemic is because ‘the problem, and the remedies, were socially constructed in the West ….’ Like most previous studies, this one points to global economic inequalities as the primary cause of limited research capacity. However, unremarked in nearly all the published literature, these findings also suggest that the problem is perpetuated by the highly individualised character of research in East Africa, fuelled by the dominance of individually-contracted consultancies. ‘Most of our social scientists are not institution based, whether NGO or private. They are there for hire.’ (Faculty dean) Such consultancies seem to stunt research capacity: reports are generally not disseminated, thus not contributing to collective understandings, university departments are denied overheads, and staff are diverted from teaching, supporting colleagues, or publishing. Furthermore, consultancies exacerbate the narrow policy orientation of African social science research (Allen, 1986; Rossi, 2004; Sall, 2003). Given their prominence, this study explored the potential for research consultancies to be used to strengthen research capacity, which to date has been largely ignored in the wider literature. The principle that consultancies should be contracted with institutions, rather than individuals, with overheads of around 30%, was widely accepted. This could fund many initiatives to strengthen research capacity, facilitate a more collective approach to research, and in the longer term might mean commissioning agencies get better value. Some African universities and research centres already regulate the division of consultancy fees between researchers and their institution, e.g., the University of KwaZulu-Natal and the REACH Trust, Malawi (Theobald & Nhlema, in press). However, established individual consultants would probably oppose institutionalisation, given very low university salaries and lack of confidence in departmental administration, due to experience of patronage, mismanagement and corruption (Zeleza, 2003). Furthermore, commissioning agencies were said to prefer individual consultancies as cheaper and more straightforward. By and large, initiatives to strengthen research capacity do not address the issue of research consultancies, although in practice they are in competition for researchers’ commitment. This was clear in an academic research centre sponsored by an international NGO where researchers are prohibited from consultancy work. However, as noted above, some large donors further the development of institutional capacity by only contracting research through institutions, not individuals. Unsurprisingly, respondents’ accounts were shaped by their professional and institutional positions, for instance leading them to defend their staff or externalise problems (see Wight, 2005). African interviewees gave more emphasis to economic factors, research commissioners’ restrictions and exclusion from Northern-dominated academic networks, while Northern interviewees were more likely to contrast East African with Northern professional cultures. However, these cultural differences were usually attributed to underlying structural/economic factors. Only one interviewee, an African, explicitly attributed inadequate research capacity to a global economy of academic research, in which Northern institutions actively maintain their dominance. Notably, the most critical reports of East African research came through informal conversations, rather than recorded interviews. This has only been an exploratory study. Further research needs to clarify: the scale of individual consultancies across East Africa; whether revising commissioning practices would seriously contribute to research capacity; and, hardly represented here, the views of agencies commissioning consultancies. While the underlying causes of poor research capacity require global economic reform, this study also points to the importance of individually contracted research consultancies in perpetuating the problem. Although they greatly augment meagre university salaries, they also seem to divert university staff from academic research and training the next generation of researchers, stunt the institutional capacity of university departments, restrict the sharing of research findings and perpetuate donors’ control of the research agenda. Commissioning bodies committed to strengthening research capacity should consider devising research contracts, and means to improve university administration, that ameliorate rather than exacerbate the problem.

Diabetologia-4-1-2292422

A Kir6.2 mutation causing severe functional effects in vitro produces neonatal diabetes without the expected neurological complications

Aims/hypothesis Heterozygous activating mutations in the pancreatic ATP-sensitive K+ channel cause permanent neonatal diabetes mellitus (PNDM). This results from a decrease in the ability of ATP to close the channel, which thereby suppresses insulin secretion. PNDM mutations that cause a severe reduction in ATP inhibition may produce additional symptoms such as developmental delay and epilepsy. We identified a heterozygous mutation (L164P) in the pore-forming (Kir6.2) subunit of the channel in three unrelated patients and examined its functional effects. Introduction ATP-sensitive potassium (KATP) channels link cellular metabolism to membrane electrical activity by regulating K+ fluxes across the plasma membrane [1]. They are found in multiple tissues but are of particular importance in regulating insulin secretion from pancreatic beta cells [1, 2]. At substimulatory glucose concentrations, KATP channels are open so that the membrane potential is hyperpolarised and Ca2+ influx and insulin secretion are prevented [3]. Glucose metabolism enhances ATP production, resulting in closure of KATP channels, stimulation of electrical activity, opening of voltage-gated Ca2+ channels and exocytosis of insulin granules. KATP channels are hetero-octamers of Kir6.x and sulfonylurea receptor (SUR) subunits [4–6]. Four inwardly rectifying subunits (Kir6.2 in pancreatic beta cells) form the pore of the channel and four auxiliary SUR subunits (SUR1 in pancreatic beta cells) associate with the tetrameric pore and regulate its gating [7–9]. Binding and/or hydrolysis of Mg-nucleotides by the intracellular nucleotide-binding domains (NBDs) of SUR produces channel opening [10–13]. It is believed that reciprocal changes in the intracellular concentrations of ATP and MgADP are involved in the metabolic regulation of KATP channels. Over the last 4 years, many different missense mutations in the genes encoding Kir6.2 (KCNJ11) and SUR1 (ABCC8) have been shown to cause permanent neonatal diabetes mellitus (PNDM). This is a rare disorder characterised by high blood glucose levels that manifests within the first 6 months of life. A subgroup of mutations were associated with a more severe clinical profile characterised by Delayed development of motor, intellectual and social skills, muscle weakness, Epilepsy, facial dysmorphism and Neonatal Diabetes (DEND syndrome) [14–16]. Mutations in Kir6.2 have also been found to cause a remitting relapsing form of neonatal diabetes that resembles transient neonatal diabetes mellitus [17, 18]. All Kir6.2 mutations analysed to date were heterozygous and most were de novo mutations [14, 19–21]. In most cases, the diabetes they caused could be successfully treated with sulfonylureas [14, 20, 22], which directly close KATP channels by binding to the SUR1 subunit of the channel [23]. All PNDM mutations result in a reduced KATP channel sensitivity to inhibition by MgATP in vitro [15, 24]. This is expected to cause an increased KATP current amplitude and reduced insulin secretion. Studies to date suggest that the severity of the clinical phenotype reflects the extent of the reduction of the channel ATP sensitivity. Thus, mutations that produce a small increase in KATP current in the presence of physiological concentrations of MgATP (1–5 mmol/l) lead to PNDM, whereas mutations that cause a larger increase in KATP current give rise to DEND syndrome [15]. In this paper, we identify a KCNJ11 mutation (L164P) that causes neonatal diabetes without obvious neurological complications. In functional studies, we show that the L164P mutation produces a large increase in the resting whole-cell current and a marked reduction in KATP channel sensitivity to inhibition by ATP. These effects are a secondary consequence of an increase in the channel open probability (Po) produced by the mutation. Surprisingly, other mutations that produce a similar increase in Po cause DEND syndrome. The L164P mutant channel was also far less blocked by the sulfonylurea tolbutamide, which explains why the patients were unable to transfer to glibenclamide therapy. Methods Participants Informed consent was obtained from all individuals investigated (or from their parents if they were children). Molecular genetic analysis Genomic DNA was extracted from peripheral lymphocytes using standard procedures. The KCNJ11 gene was amplified and sequenced as described [25]. Other family members were also tested for the novel mutation. Family relationships were confirmed using a combination of six microsatellites on chromosome 11: D11S902, D11S419, D11S1397, D11S1901, D11S921 and D11S1888. Oocyte preparation Female Xenopus laevis were anaesthetised with ethyl 3-aminobenzoate methanesulfonate salt (MS222; 2 g/l added to the water). One ovary was removed via a mini-laparotomy, the incision sutured and the animal allowed to recover. Subsequently, animals were operated on for a second time, but under terminal anaesthesia. Immature stage V–VI oocytes were incubated for 60 min with 1 mg/ml collagenase (Type V; Sigma, Poole, UK) and manually defolliculated. All procedures were carried out in accordance with UK Home Office Legislations and the University of Oxford ethical guidelines. Oocytes were coinjected with ∼0.8 ng wild-type or mutant Kir6.2 mRNA and ∼4 ng mRNA encoding SUR. The final injection volume was 50 nl per oocyte. Isolated oocytes were maintained in Barth’s solution and studied 1–4 days after injection. Electrophysiology Wild-type or mutant Kir6.2 (GenBank D50581) were coexpressed with SUR1 in Xenopus oocytes (GenBank L40624) as described [26]. Because all the patients were heterozygous for the L164P mutation, their pancreatic beta cells will contain a mixture of wild-type and mutant Kir6.2 subunits. To simulate this heterozygosity, we coinjected Xenopus oocytes with a 1:1 mixture of mutant and wild-type Kir6.2 together with SUR1 mRNA. This is expected to give rise to a mixed population of channels composed of homomeric wild-type channels, homomeric mutant channels and heteromeric channels containing between one and three mutant subunits [21]. We refer to this global channel population as heterozygous channels.Whole-cell currents were recorded using a two-electrode voltage clamp in response to voltage steps of ±20 mV from a holding potential of −10 mV, filtered at 1 kHz and digitised at 4 kHz. Oocytes were perfused with a solution containing (in mmol/l): 90 KCl, 1 MgCl2, 1.8 CaCl2 and 5 HEPES (pH 7.4 with KOH). Metabolic inhibition was produced by 3 mmol/l sodium azide.Macroscopic currents were recorded from giant inside-out patches using an EPC10 amplifier (List Medical Electronics, Darmstadt, Germany) controlled with Pulse v8.74 software (Heka Electronik, Lambrecht, Germany). Macroscopic currents were elicited by 3 s voltage ramps from −110 to +100 mV (holding potential 0 mV), or recorded at a constant potential of −60 mV. They were filtered at 0.5 kHz and digitised at 1 kHz. The pipette solution contained (mmol/l): 140 KCl, 1.2 MgCl2, 2.6 CaCl2, 10 HEPES (pH 7.4 with KOH). The internal (bath) solution contained (mmol/l): 107 KCl, 1 K2SO4, 2 MgCl2, 10 EGTA, 10 HEPES (pH 7.2 with KOH) and Mg-nucleotides as indicated. Experiments were conducted at 20–22°C. Solutions were changed using a local perfusion system consisting of tubes of 200 μm diameter into which the tip of the patch pipette was inserted.Nucleotide concentration–inhibition curves were fit with the Hill equation: where G is the KATP conductance in the presence of ATP, Gc is the KATP conductance in the absence of the nucleotide, [ATP] is the ATP concentration, IC50 is the nucleotide concentration at which inhibition is half maximal and h is the slope factor (Hill coefficient).Single-channel currents were measured at −60 mV, filtered at 5 kHz and digitised at 20 kHz. Unitary amplitude and Po were measured from the Gaussian fit to all-points amplitude histograms of tracts of current of 30–90 s duration.Data were analysed with in-house routines developed in the IgorPro platform (Wavematrics, Portland, OR, USA). Data are given as means ± SEM in the text and in the figures. Statistical significance was evaluated using a two-tailed Student t test and p < 0.05 taken to indicate a significant difference. Results Patient characteristics and genetics Three unrelated probands with permanent neonatal diabetes were shown to be heterozygous for the KCNJ11 gene mutation L164P (c.491T>C, p.Leu164Pro). All were female and had unaffected parents of different ethnic origin. Two patients have been reported previously [27, 28]. Mutation testing and microsatellite analysis of DNA from both parents and the child for two of the cases confirmed that the mutation had arisen de novo (the parents of the third case were not available for testing). None of the patients had any neurological complications or obvious developmental delay.The first patient is currently 8 years old. She was born in Singapore from Sri Lankan parents, at 38 weeks of gestation with a birthweight of 2.6 kg. She developed diabetes at 30 weeks of age. When she came to Australia at the age of 6.5 years, her HbA1c level was elevated (7.4%) and her diabetes required insulin treatment (0.45 U kg−1 day−1). Glibenclamide treatment (1.0 mg kg−1 day−1) was trialled for 2 months (at 7 years of age) but did not produce a decrease in her insulin requirement. She continues to require insulin (up to 0.7 U kg−1 day−1) with an HbA1c ranging from 7.2 to 8.2%. She has normal developmental milestones, her neurological examination is normal and she has no evidence of diabetic complications.The second patient, born in Australia of Afghan origin, is currently 2.7 years old. She was born at term following an uneventful pregnancy, with a birthweight of 2.7 kg. She presented at 8 weeks in diabetic ketoacidosis and was subsequently treated (at 2.5 months for 4 weeks, then at 5 months for 3 months) with glibenclamide (up to 1.1 mg kg−1 day−1, regimens of two times per day or three times per day), but had no reduction in insulin requirement (0.4 U kg−1 day−1) [28]. Her mean HbA1c since diagnosis has been 7.6% (range 7.2–8.7%). Assessment at 5 months by a neurologist was entirely normal, with a normal EEG. Subsequently, she has had normal developmental milestones and her neurological examinations have been normal at all follow up visits every 3 months. Diabetic complications have not been assessed due to the young age of the patient.The third patient is from Slovakia [27]. She was born at term with a birthweight of 2.6 kg, and was diagnosed with diabetes at 5 weeks of age. She is currently 20 years old and is treated with insulin (1.24 U kg−1 day−1). Glibenclamide transfer was not attempted because the patient also suffers from hepatitis C. She has poor glycaemic control, with an HbA1c level of 15.2%, and various diabetic complications including retinopathy and nephropathy. Psychomotoric development in childhood was normal. A neurological examination suggested the presence of diabetic motoric polyneuropathy in the lower extremities but there were no further neurological findings. Although her IQ was not tested, she successfully completed a specialised business school training suggesting normal mental development. Functional analysis: effects on whole-cell KATP currents We examined the functional effects of the L164P mutation by heterologous expression in Xenopus oocytes. When wild-type Kir6.2/SUR1 channels are expressed in oocytes they are normally closed because they are inhibited by resting intracellular ATP concentrations ([ATP]i). They can be opened, however, by azide (3 mmol/l; Fig. 1), a metabolic inhibitor that lowers [ATP]i [29]. Heterozygous L164P (hetL164P) currents were about 18-fold larger at rest than wild-type currents, but increased twofold on metabolic inhibition indicating that the channel is only partially closed at resting ATP levels (Fig. 1). In contrast, homomeric L164P (homL164P) channels displayed a much greater resting current and were little affected by metabolic inhibition (Fig. 1). Fig. 1Whole-cell KATP current. Mean steady state whole-cell currents evoked by voltage steps from −10 to −30 mV before (control, white bars) and after application of 3 mmol/l azide (grey bars) and in the presence of 3 mmol/l azide plus 0.5 mmol/l tolbutamide (black bars). The number of oocytes was five to seven in each case. G, KATP conductance; Gc, KATP conductance expressed relative to the conductance in the absence of the nucleotide. WT, wild-type; hetL164P and homL164P channels as indicated The sulfonylurea tolbutamide (0.5 mmol/l) blocked whole-cell KATP currents by 98% but had no effect on homL164P currents (4% block). HetL164P channels were blocked by only 54% (Fig. 1). Functional analysis: effects on KATP channel ATP sensitivity The increase in resting whole-cell KATP currents suggests that the L164P mutation may reduce the channel ATP sensitivity, as found for other PNDM mutations [21]. We first measured the ATP sensitivity of wild-type and mutant KATP currents in the absence of Mg2+, to isolate the effects of ATP on Kir6.2 (in the absence of Mg2+, ATP does not interact with SUR1 [11]). Both homL164P and hetL164P channels had severely impaired ATP sensitivity. HomL164P channels were not blocked at all by ATP, even at concentrations as high as 10 mmol/l. The concentration–inhibition curve for hetL164P showed a striking shift to higher ATP concentrations, with an IC50 of about 100 μmol/l, and a marked pedestal of unblocked current at very high ATP concentrations. The data were best fitted by assuming that in the heterozygous state about 20% of channels are never closed by ATP (Fig. 2a,b and Table 1). Fig. 2ATP-inhibition of L164P channels is less that that of wild-type channels. a, c Currents recorded in inside-out patches excised from Xenopus oocytes expressing hetKir6.2-L164P/SUR1 (hetL164P) or homKir6.2-L164P/SUR1 (homL164P) channels, as indicated, in response to 3 s voltage ramps from −110 to +100 mV. ATP (10 mmol/l) was applied as indicated by the horizontal bars in the absence (a) or presence (c) of 2 mmol/l Mg2+. b, d, mean relationship between [ATP] and KATP conductance (G), expressed relative to the conductance in the absence of the nucleotide (Gc), for wild-type (white circles, n = 9), hetL164P (white/black circles, n = 6) or homL164P (black circles, n = 4) channels. Experiments were carried out in the absence (b) or presence (d) of 2 mmol/l Mg2+. The continuous lines through the black circles were drawn by eye. The smooth curves are the best fit to the Hill equation with IC50 of 11 μmol/l (wild-type) and 100 μmol/l (hetL164P) (b, 0 mmol/l Mg2+) or IC50 of 16 μmol/l (wild-type) and 122 μmol/l (hetL164P) (d, 2 mmol/l Mg2+)Table 1ATP sensitivity of wild-type and mutant channels IC50 (Mg-free)h (Mg-free)IC50 (2 mmol/l Mg2+)h (2 mmol/l Mg2+)A (2 mmol/l Mg2+)%Imax (3 mmol/l MgATP)Wild-type9.6 ± 1.61.30 ± 0.0815.8 ± 3.00.99 ± 0.05n.a.0.01 ± 0.01hetL164P100 ± 8*1.10 ± 0.05118 ± 29*1.16 ± 0.340.29 ± 0.07*33.6 ± 3.4*homL164Pn.a.n.a.n.a.n.a.n.a.98 ± 1*Values are means ± SEM. The number of patches was four to nine in each case.A The fraction of unblocked current used to fit the ATP concentration–inhibition relationship; h Hill coefficient; IC50, ATP concentration (μmol/l) producing half-maximal inhibition; %Imax the per cent unblocked current in the presence of 3 mmol/l MgATP; n.a. not applicable (as no block).*p < 0.05 vs wild-type Molecular mechanism of the reduced ATP sensitivity Mutations that reduce the ATP sensitivity of the KATP channel can act in several ways. They may prevent ATP binding directly. They may impair the mechanism by which nucleotide binding is coupled to channel gating. They may also stabilise the intrinsic open state of the channel (i.e. that in the absence of ATP), which shifts the gating equilibrium in the presence of ATP towards channel opening and thus indirectly reduces the channel ATP sensitivity (e.g. [15, 30, 31]).In a structural model of Kir6.2 [32], L164P lies within the channel pore, at a considerable distance from the ATP-binding site (Fig. 3). It is not predicted to interact directly with ATP. We therefore examined whether the L164P mutation alters intrinsic gating. Experiments were carried out in the absence of ATP, where intrinsic gating can be assessed. The L164P mutation had no effect on single-channel current amplitude (Table 2). However, the intrinsic Po was markedly increased, being 0.86 (n = 6) for homL164P compared with 0.4 (n = 4) for wild-type channels (Fig. 4 and Table 2). This suggests that, at least in part, the L164P mutation alters channel ATP sensitivity indirectly, via an increase in Po. Fig. 3Homology model of Kir6.2 [32]. For clarity, only two subunits are shown. ATP (yellow) is shown in its binding site. Residue L164 is shown in redFig. 4The L164P mutation enhances single-channel activity. Representative single KATP channel currents recorded at −60 mV in inside-out patches from oocytes expressing wild-type or homL164P channels, as indicated. Currents were recorded in the absence of Mg2+ and nucleotidesTable 2Single-channel parameters for wild-type and homL164P channels Poi (pA)Wild-type0.39 ± 0.054.0 ± 0.1homL164P0.86 ± 0.01*4.1 ± 0.1Mean ± SEM values of intrinsic open probability (Po) and single-channel current (i) measured at −60 mVThe number of patches was five to six in each case* p < 0.05 vs wild-type Effects on KATP channel ATP sensitivity in the presence of Mg2+ Previous studies have shown that Kir6.2 mutations associated with neonatal diabetes may not only decrease the sensitivity of Kir6.2 to ATP, but can also enhance channel activation by Mg-nucleotides [24]. We therefore explored the effect of the L164P mutation on the ATP sensitivity in the presence of 2 mmol/l Mg2+.Surprisingly, there was no difference in the effect of ATP in the presence and absence of Mg2+ on homL164P channels (Fig. 2c,d). This contrasts with what has been observed for mutations in the ATP-binding site that completely abolished ATP inhibition in the absence of Mg2+, such as R50P and G334D: channels containing these mutations were activated by MgATP [33, 34]. The IC50 for ATP inhibition of hetL164P channels was also not significantly affected by Mg2+, although the pedestal of unblocked current at high ATP concentrations was increased by about 50% (Table 1). At 3 mmol/l MgATP, a concentration within the physiological range, the unblocked current was 34%, substantially greater than that found for wild-type channels (<1%; Table 1). Functional analysis: effects on MgADP sensitivity The lack of MgATP activation of hetL164P channels could be due to reduced functional coupling between SUR1 and Kir6.2-L164P, or to reduced MgATP binding/hydrolysis at the NBDs of SUR1. To explore the former possibility, we measured the ability of MgADP to activate hetL164P channels preblocked by 100 μmol/l MgATP. It was necessary to preblock the channels as the mutant channel has an open probability close to maximal and further activation is therefore not possible. Figure 5 shows that 30 μmol/l MgADP activates wild-type currents by 3.7 ± 1.8 fold (n = 3) and that 100 μmol/l MgADP did not produce any further increase (3.7 ± 1.2-fold, n = 3). In contrast, hetL164P channels were activated less: 1.4 ± 0.1 fold (n = 3) by 30 μmol/l MgADP and 1.5 ± 0.1 fold (n = 4) by 100 μmol/l MgADP. Fig. 5Sensitivity to MgADP of wild-type and hetL164P channels. a Representative currents (I) recorded at −60 mV from inside-out excised membrane patches from Xenopus oocytes expressing wild-type or hetL164P channels, as indicated. Patches were exposed to 100 μmol/l ADP in the continuous presence of 100 μmol/l ATP: 2 mmol/l Mg2+ was present throughout. b Mean current in the presence of 100 μmol/l MgATP or 30 or 100 μmol/l MgADP plus 100 μmol/l MgATP, normalised to the current in the absence of nucleotides for wild-type (grey bars) and hetL164P (black bars) channels. Bars indicate means ± SEM. The number of patches was three to four in each case Despite the fact that 100 μmol/l MgATP blocked mutant channels less, the lower extent of activation of hetL164P channels by 30 μmol/l MgADP is not due to the fact that channel activity is already maximal, as the amplitude is still significantly less than that in control solution (Fig. 5). Discussion We describe the clinical and functional effects of a KCNJ11 mutation, L164P, associated with permanent neonatal diabetes. This mutation causes a marked reduction in KATP channel inhibition by ATP primarily by stabilising the open state of the channel. This leads to an increase in the whole-cell KATP current, and in beta cells is expected to result in a reduction in insulin secretion. Surprisingly, unlike other mutations that reduced the channel ATP sensitivity by a similar amount, no motor or mental developmental delay was associated with the L164P mutation. Structural considerations and molecular basis for reduced ATP sensitivity In a homology model of Kir6.2 [32], L164 lies partway along the permeation pathway, 35 Å away from the ATP-binding site (Fig. 3). It is therefore unlikely that it acts by reducing ATP binding directly. The side-chains of L164 point into the pore, forming a hydrophobic girdle that is narrow enough to prevent the passage of water and hydrated K+ ions [35], which suggests that L164 may form a hydrophobic gate within the pore. Functional studies support this idea. Following mutation of L164 to cysteine, cadmium ions were able to block the KATP channel with high affinity [36, 37], suggesting that the four cysteines (one on each subunit) come together close enough to form a high-affinity binding site for Cd2+. These data indicate that the side-chain of L164 must face into the pore, and that the pore is very narrow at this position. Substitution of a proline for L164 is expected to produce a kink in the α-helix [38] and disrupt the hydrophobic gate. It is therefore not surprising that the L164P mutation caused a dramatic effect on the Po. Mutation of L164 to cysteine, alanine, valine, threonine or glycine also produces a very large increase in Po [30, 36, 37, 39], consistent with the importance of this residue in channel gating. The fact that the L164P mutation enhanced Po can explain, in part, the reduced ATP sensitivity of the channel. It is also possible that the mutation may have effects additional to stabilisation of the channel open state. Effect of Mg2+ on Kir6.2-L164P/SUR1 channel ATP sensitivity In contrast to all PNDM mutations studied to date [24], MgATP was unable to enhance the activity of either homL164P or hetL164P channels. The lack of MgATP activation of homL164P channels may be due to the fact that Po is already very high and thus there is no scope for further activation. However, little MgATP activation was also observed for hetL164P channels. This is in marked contrast to other mutations that caused enhanced Po (I296L, [16], V59G [15]) where MgATP activated both homomeric and heterozygous channels. It is therefore possible that the L164P mutation reduces the efficacy of coupling between SUR1 and Kir6.2, and thereby decreases the ability of MgATP to stimulate channel activity. In support of this idea, MgADP activation of hetL164P channels was also reduced. This may explain the inability of Mg2+ to reduce ATP inhibition of hetL164P channels, since it is well established that MgATP must be hydrolysed to MgADP to stimulate channel activity.The reason for the lack of Mg-nucleotide activation of hetL164P channels is unclear. Because L164 lies within the pore, far away from the NBDs of SUR1, the effect must be mediated allosterically. It is possible that this is mediated by an interaction between the backbone of L164 itself, or transmembrane domain 2 (within which L164 lies), and the transmembrane domains of SUR1. In the absence of an atomic resolution structure of the KATP channel complex, however, this cannot be definitely determined.HomL164P channels were not blocked by tolbutamide, as expected because of their high Po [40]. The lower efficacy of tolbutamide on hetL164P channels (about 50% block) may reflect the enhanced Po of channels within the heterozygous population containing mutant subunits. It could also reflect impaired coupling between SUR1 and Kir6.2, such as that found for MgADP. Clinical implications Previous studies of KCNJ11 mutations have suggested that there is a good correlation between the percentage of current that remains unblocked in the presence of 3 mmol/l MgATP and the clinical phenotype. Namely, currents that are >30% of maximal are associated with DEND syndrome, and those that lie between 5–10% of maximal with neonatal diabetes alone, compared with a current of <1% for wild-type channels (Fig. 6). The L164P mutation does not conform to this simple relationship between the functional effects of the mutation and the clinical phenotype. The magnitude of the hetL164P current in 3 mmol/l MgATP was 36%, yet none of the patients had extra-pancreatic symptoms. Fig. 6Mean KATP current (expressed as a % of maximum) measured in the presence of 3 mmol/l MgATP from inside-out patches expressing heterozygous mutant channels as indicated. The dashed line indicates the maximal current amplitude normally associated with PNDM. iDEND, intermediate DEND syndrome (i.e. neonatal diabetes with developmental delay [21]). Data for wild-type (WT) and Kir6.2-R201H channels are from [14], for Y330C and F333I from [26], for Q52R, V59G and R201C from [15], for V59M from [24], for I296L from [16] and for G334D from [34]The reason for this anomaly is not clear, particularly as the resting whole-cell current was also very large. Previous studies have also indicated that such large whole-cell currents are invariably associated with more severe clinical symptoms. The lack of responsiveness of the patient to sulfonylureas is consistent with the very large increase in Po and the resting whole-cell current. It is therefore likely that hetL164P channels expressed in Xenopus oocytes are a reasonable model for the pancreatic beta cell KATP channels of the patients. We are therefore forced to conclude that compensatory mechanisms must ameliorate the extra-pancreatic effects of this mutation. The fact that the patients had very different ethnic origins suggests the compensatory mechanism is unlikely to reflect a shared genetic background: however, it does appear to be specific to the L164P mutation, which is unique (to date) in having a marked functional effect in vitro despite causing neonatal diabetes without neurological complications.Importantly, a maximally effective concentration of tolbutamide only blocked hetL164P channels by 50%. All patients to date whose channels are blocked by <65% have not been able to transfer from insulin treatment to sulfonylurea therapy. This suggested that our patients would be unable to transfer to sulfonylurea therapy, as indeed was found to be the case. This is likely to be due to the fact that L164P destabilises the long closed state of the channel, to which sulfonylureas preferentially bind, and which is rarely entered in channels with enhanced Po. Our results further suggest that not all patients with permanent neonatal diabetes will necessarily respond to sulfonylurea treatment, and indicates that knowledge of the functional effect of the mutation is helpful for predicting the drug response in patients.

Behav_Brain_Res-2-1-2424131

Effect of disconnecting the orbital prefrontal cortex from the nucleus accumbens core on inter-temporal choice behaviour: A quantitative analysis

Previous experiments showed that destruction of the orbital prefrontal cortex (OPFC) or the nucleus accumbens core (AcbC) in rats altered choice between two delayed food reinforcers. Application of a quantitative model of inter-temporal choice suggested that lesions of either structure increased the delay-dependent degradation of reinforcer value (delay discounting); destruction of the OPFC (but not the AcbC) also increased the relative value of the larger reinforcer. This experiment examined the effect of disconnecting the OPFC from the AcbC on inter-temporal choice. Rats received excitotoxin-induced contralateral lesions of the OPFC and AcbC (disconnection), severing of the anterior corpus callosum (callosotomy), a combined lesion (disconnection + callosotomy) or sham lesions. They were trained in a discrete-trials progressive delay schedule to press levers A and B for a sucrose solution. Responses on A delivered 50 μl of the solution after a delay dA; responses on B delivered 100 μl after a delay dB. dB increased across blocks of trials; dA was manipulated across phases of the experiment. Indifference delay, dB(50) (value of dB corresponding to 50% choice of B), was estimated for each rat in each phase, and linear indifference functions (dB(50)vs. dA) were derived. The disconnection + callosotomy group showed a lower intercept of the indifference function (implying a higher rate of delay discounting) than the sham-lesioned group; the disconnection group showed a similar but less robust effect, whereas the callosotomy group did not differ significantly from the sham-lesioned group. The results suggest that OPFC–AcbC connections are involved in delay discounting of food reinforcers, but provide no evidence for an involvement of OPFC–AcbC connections in regulating sensitivity to reinforcer size. 1 Introduction It is widely accepted that the capacity of a reinforcer to exert control over operant behaviour is a direct function of its size and an inverse function of the delay that precedes its delivery. These two principles are placed in mutual opposition in inter-temporal choice schedules, in which the subject is required to choose between two reinforcers that differ along both dimensions. For example, the subject may be confronted with a choice between a smaller reinforcer, A, of size qA, delivered after a short delay, dA, and a larger reinforcer, B, of size qB, delivered after a longer delay, dB. Inter-temporal choice schedules have provided valuable insights into the behavioural and neurobiological bases of ‘delay discounting’, the hypothetical process whereby the efficacy or ‘value’ of a reinforcer decays as a function of delay. Recent work has implicated the core of the nucleus accumbens (AcbC) [1,2,6,8] and the orbital region of the prefrontal cortex (OPFC) [23,26,42] in inter-temporal choice behaviour. However, the interpretation of results obtained using these schedules is complicated by the fact that magnitude and delay of reinforcement are generally manipulated simultaneously. Therefore it is often unclear whether the effect of an intervention on inter-temporal choice behaviour has been brought about by a change in the rate of delay discounting, by a change in the organism's sensitivity to relative reinforcer size, or both [8,19,31]. One approach to overcoming this problem is the application of null equations derived from quantitative models of inter-temporal choice [19,29–31]. For example, according to one such model [19], the overall value of a reinforcer, V, is determined by the multiplicative combination of two hyperbolic expressions that define the effects of delay and magnitude upon reinforcer value:where K is the delay-discounting parameter [29] and Q is a parameter expressing sensitivity to reinforcer size [19]. Faced with a choice between two reinforcers, A and B, the organism is presumed to select the reinforcer with the higher value. However, by experimentally manipulating the sizes and delays of the two reinforcers, it is possible to establish a state of ‘indifference’, in which A and B are selected with equal frequency. Indifference between A and B is taken to imply equality of value, i.e., VA = VB. Using Eq. (1) to define VA and VB and solving for dB(50) (the delay to reinforcer B at the point of indifference), yields the following null equation:in which dB(50) is linearly related to dA. By examining the effect of an intervention on this relation, it is possible to deduce whether the intervention has affected the organism's rate of delay discounting, its sensitivity to reinforcer size, or both. Since K makes no contribution to the slope of the function, a change in slope implies a change in sensitivity to reinforcer size, whereas a change in intercept in the absence of a concomitant change in slope implies a change in the rate of delay discounting [19]. Destruction of the OPFC was found to increase the slope of the linear indifference function, indicating that the lesion affected sensitivity to reinforcer size; the lack of a concomitant increase in the intercept suggested that the rate of delay discounting had also been increased [26] (see Section 4 for further explanation). In contrast, lesions of AcbC reduced the intercept without significantly altering the slope of the function, implying a selective effect on delay discounting [2]. While these observations implicate both the OPFC and the AcbC in delay discounting, they leave unanswered the question of whether the two structures play independent roles in delay discounting, or whether they contribute to an integrated mechanism regulating inter-temporal choice behaviour. The present experiment was intended to address this question: the effect of functional disconnection of the OPFC and AcbC on inter-temporal choice behaviour was examined using the linear indifference relation epitomised by Eq. (2). The principal anatomical link between the OPFC and the AcbC is an ipsilateral excitatory glutamatergic corticofugal pathway, which is believed to comprise one link in a cortico-striato-thalamo-cortical circuit. Inhibitory γ-aminobutyric acid (GABA)ergic efferents from the AcbC project to the internal pallidum and substantia nigra pars reticulata. These structures in turn send inhibitory projections to diencephalic structures, whose excitatory corticopetal projections complete the circuit. In order to effect functional disconnection of the OPFC and AcbC without totally ablating either structure, a ‘disconnection lesion’ [12,14] was employed, which consisted of unilateral excitotoxic destruction of the OPFC in one hemisphere and of the AcbC in the other hemisphere. As recent evidence indicates that transcallosal fibres may make a significant functional contribution to cortico-striatal connections [9], the effects of callosotomy alone, and callosotomy combined with the disconnection lesion were also examined. 2 Methods The experiment was carried out in accordance with UK Home Office regulations governing experiments on living animals. 2.1 Subjects Sixty experimentally naive female Wistar rats (Charles River UK) approximately 4 months old and weighing 250–300 g at the start of the experiment were used. They were housed individually under a constant cycle of 12 h light and 12 h darkness (light on 06:00–18:00 h), and were maintained at 80% of their initial free-feeding body weights throughout the experiment by providing a limited amount of standard rodent diet after each experimental session. Tap water was freely available in the home cages. 2.2 Surgery Anaesthesia was induced with halothane (4% in oxygen), and the rat positioned in a stereotaxic apparatus (David Kopf), with the upper incisor bar set 3.3 mm below the inter-aural line. Anaesthesia was maintained with 2% halothane in oxygen during surgery. Holes were drilled in the skull for introduction of a microinjection cannula or leucotome (see below). Disconnection: sixteen rats received unilateral lesions of the OPFC and the contralateral AcbC (the sides being counterbalanced across rats). The following coordinates (mm, measured from bregma) were used to locate the OPFC: site (i): AP +3.7, L ±1.2, DV −4.8; site (ii): AP +3.7, L ±2.8, DV −4.4. The coordinates for the AcbC were: AP +1.2, L ±1.8, V −7.1. Injections were given via a 0.3-mm diameter cannula connected by a polyethylene tube to a 10-μl Hamilton syringe. In each site, 0.5 μl of a 0.1-M solution of quinolinic acid (2,3-pyridinedicarboxylic acid) in phosphate-buffered 0.9% NaC1 (pH 7.0) was injected at a rate of 0.1 μl per 15 s. The cannula was left in position for 3 min after completion of the injection. Callosotomy: 15 rats underwent a midline leucotomy in order to sever the anterior corpus callosum. The leucotome, similar to that described by Gold et al. [17], was constructed from a 1-μl Hamilton microsyringe. A curved wire could be extruded from the tip of the syringe needle; when fully extended, the wire projected approximately 3 mm in the AP plane (i.e. at right angles to the needle). Two midline cuts were made: (i) the tip of the needle was positioned at AP 0.0, DV −5.0, the wire extruded in a rostral direction, and the tip slowly raised to DV −1.0; the wire was then retracted into the needle, and the needle was withdrawn from the brain; (ii) the tip of the needle was repositioned at AP +2.0, DV −4.0 and the procedure repeated. Disconnection + callosotomy: 15 rats underwent both the disconnection and callosotomy procedures described above. Sham lesion: fourteen rats underwent the same surgical procedures as the disconnection-lesioned group, except that the vehicle solution alone was injected into the target sites. 2.3 Apparatus The rats were trained in standard operant conditioning chambers (CeNeS Ltd., Cambridge, UK) of internal dimensions 25 cm × 25 cm × 22 cm. One wall of the chamber contained a recess into which a peristaltic pump could deliver a 0.6 M sucrose solution. Two apertures situated 5 cm above and 2.5 cm to either side of the recess, through which motor-operated retractable levers could be inserted into the chamber. The levers could be depressed by a force of approximately 0.2 N. A 2.8-W lamp was mounted 2.5 cm above each lever; a third lamp was mounted 10 cm above the central recess. Six red light-emitting diodes were mounted in a row, 4 cm apart, 5 cm above the levers. The operant chamber was enclosed in a sound-attenuating chest; masking noise was generated by a rotary fan. An Acorn microcomputer programmed in Arachnid BASIC (CeNeS Ltd., Cambridge, UK), located in an adjoining room, controlled the schedules and recorded the behavioural data. 2.4 Behavioural training Two weeks after surgery, the food-deprivation regimen was introduced and the rats were gradually reduced to 80% of their free-feeding body weights. They were then trained to press two levers (A and B) for sucrose reinforcement, and were exposed to a discrete-trials continuous reinforcement schedule in which the two levers were presented in random sequence for three sessions. After this initial training, they underwent daily training sessions under the discrete-trials delayed reinforcement schedule for the remainder of the experiment. Each experimental session consisted of six blocks of six trials, except in phases 4 and 5 when sessions consisted of five blocks. The trials were 90 s in duration, with the exception of phase 5, in which the duration was increased to 120 s in order to accommodate the long delay to reinforcement (see below). The six blocks were signalled by illumination of the six light-emitting diodes: in block 1 the first (left-most) diode was illuminated, in block 2 the first and second diodes were illuminated, and so on. The first two trials of each block were forced-choice trials in which each lever was presented alone in random sequence. The other four trials were free-choice trials in which both levers were presented. The beginning of each trial was signalled by illumination of the central light above the reinforcer recess. After 2.5 s the lever or levers (depending on the type of trial) were inserted into the chamber. When a lever-press occurred, the lever(s) were withdrawn, the central light was extinguished, and the light located above the lever that had been depressed was illuminated. This light remained illuminated until the delivery of the reinforcer, and was then extinguished. The chamber remained in darkness until the start of the following trial. If no lever-press occurred within 5 s of the lever(s) being inserted, the lever(s) were retracted and the central light extinguished. (This seldom happened except during the first few training sessions.) A response on lever A initiated a fixed delay dA, following which 50 μl of the 0.6 M sucrose solution was delivered. A response on lever B initiated a variable delay dB, after which 100 μl of the same sucrose solution was delivered. The positions of levers A and B (left vs. right) were counterbalanced across subjects. The experiment consisted of six phases, in which the value of dA was set at 1, 2, 4, 8, 12 and 0.5 s, respectively. In each phase, the value of dA was held constant. In each session the value of dB was set equal to dA in the first block of trials. In subsequent blocks dB was increased in increments of 75%. In phases 4 and 5, when dA was 8 and 12 s, respectively, computing five increments of 75% would have generated a value of dB that was longer than the duration of a trial in the sixth block of trials; therefore the number of blocks was limited to five in these phases. The first phase continued for 100 sessions, phase 2 for 50 sessions, and the remaining phases for 40 sessions. Experimental sessions were carried out 7 days a week, at the same time each day, during the light phase of the daily cycle (between 08:00 and 14:00 h). 2.5 Histology At the end of the behavioural experiment, the rats were deeply anaesthetised with sodium pentobarbitone, and perfused transcardially with 0.9% sodium chloride, followed by 10% formol saline. The brains were removed from the skulls and fixed in formol saline for 1 week. Forty micrometer coronal sections were taken through the regions of the OPFC and AcbC (approximately from AP +5.0 to AP 0.0) using a freezing microtome. 2.5.1 Cresyl violet staining The procedure was similar to that described previously [26]. Alternate sections were mounted on chrome-gelatine-coated slides and air dried, hydrated by successive immersion in 95, 70 and 50% ethanol, stained in 0.25% cresyl violet for 2 min at room temperature, dehydrated by successive immersion in 50, 70, 95, and 100% ethanol and xylene, and mounted with DPX. 2.5.2 Immunocytochemistry In the other sections neurone-specific nuclear protein (NeuN) was labelled as described by Jongen-Relo and Feldon [21]. Our protocol has been described elsewhere [2]. Briefly, freshly sliced sections were rinsed in 0.1 M phosphate buffered saline (PBS) and placed in 0.5% H2O2 in PBS for 30 min. After twice rinsing in PBS, they were placed for 1 h in a blocking solution (10% normal horse serum [Vector Laboratories, Peterborough, UK], 1% bovine serum albumin [BSA, Sigma–Aldrich, Gillingham, UK] and 0.3% Triton X-100 [Sigma–Aldrich] in PBS). They were incubated for 48 h at 4 °C with the primary antibody (monoclonal mouse anti-NeuN serum [1:5000, Chemicon, Chandlers Ford, UK] in 1% normal horse serum, 1% BSA and 0.3% Triton X-100 in PBS), washed twice in PBS, and incubated for 2 h at room temperature in biotinylated horse antimouse serum (Vector Laboratories) (1:1000 in 1% BSA and 0.3% Triton X-100 in PBS). After further rinsing in PBS, they were placed for 2 h in avidin–biotin–horseradish peroxidase complex (1:200, ABC-Elite, Vector Laboratories) in PBS. After two further rinses in PBS, they were placed in a chromagen solution (0.05% diaminobenzidine [Sigma–Aldrich] and 0.01% H2O2 [Sigma–Aldrich]) for 5 min. The reaction was observed visually and stopped by rinsing in PBS. The sections were floated on to chrome-gelatine-coated slides and mounted with DPX. An investigator who was blind to the behavioural results performed the microscopic examination. Drawings of the area of the lesions were superimposed on the appropriate coronal sections in the stereotaxic atlas of Paxinos and Watson [34]. 2.6 Data analysis Data from 5 of the 60 rats were discarded. Histological examination revealed two rats with misplaced excitotoxin-induced lesions; one rat in the disconnection + callosotomy group was discarded because of a failed callosotomy. Data from two further rats were discarded because they showed persistent exclusive responding on one lever. This left 13 rats in the sham-lesioned group, 14 in the disconnection group, 15 in the callosotomy group and 13 in the disconnection + callosotomy group. 2.6.1 Preference functions and linear indifference functions For each rat, the percentage choice of lever B in the free-choice trials (%B) was computed for each block of trials from the pooled data from the last 10 sessions of each phase of the experiment. Plots of %B vs. dB were derived for each rat, and the indifference delay (dB(50): the value of dB corresponding to %B = 50%) was estimated by linear interpolation between the two delays which fell on either side of %B = 50% (i and j) using the formula: dB(50) = dB(i) + ([dB(j)-dB(i)].[%Bi − 50]/[%Bi − %Bj]) [39]. Plots of dB(50) vs. dA were obtained for each rat, and linear functions were fitted by the method of least squares; goodness of fit was expressed as r2, the proportion of the data variance accounted for by the fitted function. The slope and intercept of the linear indifference functions were analysed by two-factor ANOVA (presence/absence of disconnection × presence/absence of callosotomy) followed by multiple comparisons of the lesioned groups with the sham-lesioned group using Dunnett's test. Linear indifference functions (dB(50) vs. dA) were also derived for the group mean data. The slopes and elevations of these functions were analysed by one-factor ANOVA (group) followed by multiple comparisons of the lesioned groups with the sham-lesioned group using Dunnett's test, as described by Zar [43]: the slopes were first analysed, and in the absence of significant between-group variation in slope, a common weighted slope value was adopted in order to make comparisons among the elevations. (‘Elevation’ refers to the y-axis location of the function taking the range of observed data into account, whereas ‘intercept’ refers to the intersection of the function with the y-axis location [43].) 2.6.2 Psychophysical analysis of preference functions Logistic functions were fitted to the group mean %B data and the %B data from each rat in each phase of the experiment: %B = 100/(1 + [dB/dB(50)]ɛ). This function defines a descending sigmoid curve which is symmetrical in semi-logarithmic co-ordinates; dB(50) and ɛ are parameters, dB(50) being the point of intersection of the logistic curve with the indifference line, and ɛ being the slope of the function. These parameters were used to derive the limen ([dB(25) − dB(75)]/2, where dB(25) and dB(75) are the estimated values of dB corresponding to %B = 25 and %B = 75, respectively), and the index of relative precision, the Weber fraction, was defined as limen/dB(50). The Weber fraction was subjected to repeated-measures ANOVA (phase); as no significant effect of phase was revealed, the Weber fractions were averaged across phases and subjected to two-factor ANOVA (presence/absence of disconnection × presence/absence of callosotomy), as described above. 3 Results 3.1 Behavioural data 3.1.1 Preference functions and linear indifference functions Preference functions (%B vs. dB) derived for the four groups in all six phases of the experiment are shown in Fig. 1 (left-hand graphs). In all four groups, preference for lever B declined as a function of the delay to reinforcer B (dB). The horizontal lines in the graphs show the indifference level (i.e., %B = 50); the value of dB at which the preference function crossed this level (i.e. dB(50)) increased as a function of increasing values of dA, reflecting a progressive rightward displacement of the curve. Fig. 2 shows indifference functions (dB(50) vs. dA) for the group mean data. In each group, the linear function accounted for more than 96% of the variance of the group mean data (r2 > 0.96). Comparisons were made between the function derived for each of the lesioned groups and the function derived for the sham-lesioned group [43]. An initial test on the homogeneity of the slopes indicated that there was no significant effect of group upon the slope [F(3,20) = 1.8, P > 0.05], and the common (weighted) slope (3.16) was therefore adopted in statistical comparisons of the elevations of the functions. There was a significant effect of group on the elevation [F(3,23) = 24.4, P < 0.01]. Multiple comparisons (Dunnett's test) showed that the elevations of the functions derived for the disconnection and disconnection + callosotomy groups differed significantly from that of the sham-lesioned group [t(9) = 4.0, P < 0.05, and t(9) = 5.2, P < 0.01, respectively], whereas there was no significant difference between the elevations of the callosotomy and sham-lesioned groups [t(9) = 0.2, P > 0.3]. Linear indifference functions were also fitted to the data from the individual rats. The group mean values (+S.E.M.) of the slope and intercept of the function are shown in Fig. 3. A two-factor analysis of variance (callosotomy × disconnection) showed no significant effect of either factor nor any significant interaction on the slope [all Fs < 1]. There was a significant effect of the disconnection lesion on the intercept [F(1,52) = 8.7, P < 0.005], but no significant effect of the callosotomy [F < 1] and no significant interaction [F < 1]. Multiple comparisons with the sham-lesioned group (Dunnett's test) showed that only the disconnection + callosotomy group differed significantly from the sham-lesioned group. The goodness of fit of the linear function did not vary significantly among the groups [F < 1]; the function accounted for >86% of the data variance for individual rats in all groups [r2 = 0.865 ± 0.022]. 3.1.2 Psychophysical analysis of preference functions The logistic psychometric functions derived for the group mean data (Fig. 1, right-hand panels) accounted for 95% of the data variance (r2 > 0.95) in all cases. The logistic function could be fitted to 311 of the 330 preference functions obtained for the individual rats in the six phases of the experiment (94.2%); functions could be fitted to the data from all six phases in 48 of the 55 rats. Analysis of variance (disconnection × callosotomy) indicated that the goodness of fit was not significantly affected by either the callosotomy [F < 1] or the disconnection lesion [F(1,51) = 1.2, P > 0.2], and there was no significant interaction [F < 1]. The overall mean (±S.E.M.) value of r2 derived from all rats in all phases of the experiment was 0.966 ± 0.005. There was good agreement between the values of dB(50) derived from the logistic functions (dB(50)logist.) and those derived by linear interpolation (dB(50)interp.). The slope of the regression of dB(50)logist. vs. dB(50)interp. (1.02 ± 0.01) did not deviate significantly from unity, and the intercept (−0.16 ± 0.31) did not deviate significantly from zero; the correlation (r) between the two estimates was 0.971. The Weber fraction derived from the logistic function was not systematically related to the value of dA. A single-factor analysis of variance with repeated measures (incorporating data from the 48 rats that generated Weber fractions from all 6 phases) showed no significant effect of phase [F(5,235) = 1.9, P > 0.05]. The Weber fraction was therefore averaged across phases for each rat; the group mean values (+S.E.M.) are shown in Fig. 4 (left-hand histogram). Analysis of variance (callosotomy × disconnection) revealed a significant main effect of the disconnection lesion [F(1,51) = 6.6, P < 0.05], but no significant effect of the callosotomy [F(1,51) = 2.7, P > 0.1] and no significant interaction [F < 1]. Multiple comparisons (Dunnett's test) did not reveal any significant differences between the lesioned groups and the sham-lesioned group. The group mean values of the slope of the logistic functions (ɛ) (+S.E.M.) are shown in Fig. 4 (right-hand histogram). Analysis of variance (callosotomy × disconnection) showed a significant main effect of the disconnection lesion [F(1,51) = 7.2, P < 0.05], but no significant main effect of the callosotomy [F < 1] and no significant interaction [F < 1]. None of the lesioned groups differed significantly from the sham-lesioned group (Dunnett's test). 3.2 Histology Examples of the lesions are shown in Fig. 5. OPFC: injection of quinolinic acid into the OPFC resulted in gliosis and atrophy of the ventral and lateral orbital regions. There was some damage to the medial prefrontal cortex (medial orbital, infralimbic and prelimbic cortices) in some rats. The AP extent of the lesion was from about +3.2 to +4.5; in no case did the lesion extend caudally as far as the anterior margin of the nucleus accumbens. AcbC: coronal sections showed ventricular dilatation and atrophy in the ventral striatal area. The NeuN labelled sections showed that there was extensive neuronal loss in the area of the AcbC of all lesioned animals, with some neuronal loss in the ventral and medial portions of the caudate-putamen in some animals; the shell region of the nucleus accumbens was essentially spared. Callosotomy: the corpus callosum was completely severed between AP +0.5 to +1.6 in most animals. There was some sparing of the corpus callosum posteriorly (caudal to AP +0.5), possibly reflecting the curvature of the leucotome. Destruction of the callosum was generally accompanied by some ventricular dilatation in the vicinity of the lesion, and the mesial surfaces of the cortex overlying the lesion showed some damage. 4 Discussion Injections of quinolinic acid into the OPFC and AcbC produced lesions of similar extent to those seen in previous experiments in which excitotoxins have been used to lesion these structures [OPFC: 13, 23, 25, 26, and 42; AcbC: 1, 2, 3, 5, 6, and 35]. In the case of the OPFC lesion, the main area of damage included the ventral and lateral orbital (VO and LO) regions. There was also some intrusion into the medial orbital (MO), infralimbic (IL) and prelimbic (PrL) regions in some animals. According to the subregional classification recommended by Uylings and van Eden [40] and Kesner [22], the lesion embraced the ventral (VO) and lateral OPFC (LO), with some involvement of the medial PFC (MO, IL and PrL). In the case of the AcbC lesion, the area of destruction was mainly restricted to the target structure. Some additional damage was inflicted to the ventral portion of the caudate-putamen in some rats; and in some cases NeuN staining revealed a band of neuronal loss in the medial caudate-putamen adjacent to the lateral ventricle. The mesial shell region of the nucleus accumbens was spared. The callosotomy lesion was generally successful in severing the corpus callosum anterior to the caudal margin of the AcbC. The discrete-trials schedule used in this study was an adaptation of the progressive delay schedule developed by Evenden and Ryan [10]. The rats in all four groups showed a progressive shift in preference from the larger to the smaller reinforcer as the delay to the larger of the two reinforcers (dB) was progressively increased across successive blocks of trials. This is consistent with previous studies that have used this schedule [2,6–8,10,11,23,24,26,33,42]. As in previous experiments [2,23,24,26], we used a geometric progression to determine the values of dB in successive blocks of trials, thereby allowing the range of values of dB to be adapted to the value of dA, which was systematically manipulated across the six phases of the experiment. The resulting preference functions (relation between %B and dB; see Fig. 1) were used to compute the indifference delays to the larger reinforcer (dB(50)) in each phase. This measure formed the basis of the linear indifference functions (see below). As in previous studies employing Evenden and Ryan's [11] protocol [2,6–8,10,11,23,24,26,33,42], the preference functions seen in this experiment were characterized by a gradual reduction of %B as a function of dB. As noted by Bezzina et al. [2], this is apparently inconsistent with models of inter-temporal choice that are based on the computation of hypothetical ‘values’ of reinforcers. Such models generally assume that organisms should invariably select the more highly valued of two mutually exclusive reinforcers [18,19], leading to the prediction that the ideal preference function should be a step function, %B falling precipitously from near 100% to near 0% around the point at which VA = VB. Bezzina et al. [2] proposed that the gradual decline in preference generated by progressive delay schedules may represent a discrimination gradient for reinforcer value, and adopted a standard psychophysical approach to analyse the preference functions, an analytical approach that was also used in the present study. As in Bezzina et al.'s [2] experiment, a two-parameter logistic function adequately described the group mean preference functions (Fig. 1, right-hand panels) and more than 90% of the preference functions obtained from individual rats. The two parameters of this function define its slope (ɛ) and its locus on the abscissa (dB(50)). Combination of ɛ and dB(50) allows computation of the Weber fraction, the traditional measure of the relative precision of discrimination [15,16,20,28]. The data shown in Fig. 4 indicate that ɛ tended to be lower and the Weber fraction higher in the disconnection-lesioned and disconnection + callosotomy groups than in the sham-lesioned group. This resembles Bezzina et al.'s [2] finding with bilateral AcbC lesions, which induced a robust increase in the Weber fraction, consistent with an impairment of discriminative precision. It appears that functional disconnection of the OPFC from the AcbC had a similar effect on value discrimination as destruction of the AcbC. The locus of the preference function is defined by dB(50), the delay to reinforcer B corresponding to indifference between the two reinforcers. According to hyperbolic models of inter-temporal choice [19,29], indifference implies equality of the values of the two reinforcers, which provides the basis for deriving the linear function expressed by Eq. (2). This equation offers a means of distinguishing between changes in inter-temporal choice behaviour brought about by effects on the hypothetical processes of delay discounting (K) and magnitude discounting (Q). The slope of the linear function reflects the physical magnitudes of the two reinforcers (qA and qB) and Q; a change in slope therefore implies a change in Q. The intercept of the function is influenced jointly by Q and K. Therefore, while an increase in Q causes an increase in both the slope and the intercept, an increase in K simply diminishes the intercept. If both parameters are increased, Q's effect on the intercept is countered by the change in K, whereas its impact on the slope remains unaltered [19]. In the present experiment, the indifference functions of the four groups did not differ significantly in terms of slope; however, there were significant between-group differences in the intercept. The disconnection-lesioned group showed a somewhat lower intercept than the sham-lesioned group. This effect was statistically significant when the functions fitted to the group mean data were compared; however it failed to reach significance when the functions fitted to the data from individual rats in the two groups were compared. A more robust effect on the intercept was seen in the case of the disconnection + callosotomy group, whose intercept was significantly lower than that of the sham-lesioned group, in the case of the functions fitted to the group mean data and also the functions fitted to the individual-subject data. The callosotomy had no significant effect on the intercept. These results indicate that functional disconnection of the AcbC from the OPFC selectively altered the delay discounting parameter, K. The lower intercept seen in the disconnection + callosotomy group implies a higher value of K, in other words, a higher rate of delay discounting, than that of the sham-lesioned group. The fact that callosotomy alone had no discernable effect on performance suggests that ipsilateral cortifugal fibres may be mainly responsible for the functional connection between the OPFC and AcbC; however, the fact that the combined disconnection + callosotomy lesion had a more robust effect on inter-temporal choice than the conventional lesion suggests that inter-hemispheric connections may also make a subsidiary contribution. The existence of inter-hemispheric cortico-striatal connections has been known for some time [32], although their functional importance has only recently been established. Dunnett and his colleagues [9,41] recently found that a combined disconnection + callosotomy lesion was needed to effect complete functional disconnection of the prefrontal cortex and dorsal striatum in the case of delayed alternation performance. Previous experiments have examined the effects of selective lesions of the OPFC and AcbC on inter-temporal choice behaviour. Kheramin et al. [23,26] found that bilateral destruction of the OPFC increased the slope of the linear indifference function without altering the intercept, an effect that was attributed to an increase in the rate of delay discounting combined with increased sensitivity to relative reinforcer size (see also [25]). Cardinal et al. [6,8] first reported that bilateral lesions of the AcbC rendered rats more ‘intolerant’ to delay of reinforcement. This finding was recently confirmed and extended by Bezzina et al. [2], using the indifference-equation approach; these authors reported that destruction of the AcbC lowered the intercept of the linear indifference function without significantly altering the slope, implying an increase in the rate of delay discounting. These findings suggest that both the OPFC and AcbC may contribute to delay discounting. The present results further suggest that delay discounting may be regulated by an integrated mechanism that involves both these structures, and that the integrity of both structures and their connecting fibres may be important for the effective control of behaviour by delayed reinforcers. Interestingly, the present results provide no evidence for an involvement of OPFC–AcbC connections in regulating sensitivity to reinforcer size. This suggests that while the OPFC may integrate information on multiple features of reinforcers, including both size and delay [37,38], its role in delay discounting may be more specifically related to its connections with the AcbC. Data consistent with this notion have recently been obtained using the progressive ratio schedule. Mathematical analysis of performance on this schedule [27] yields a quantitative index of instantaneous reinforcer value that is sensitive to variations of reinforcer size [4,36]. This index has been found to be reduced by lesions of the OPFC [25] but not by lesions of the AcbC [3], suggesting that the OPFC, but not the AcbC, may be involved in determining sensitivity to reinforcer size.

Pediatr_Nephrol-4-1-2413093

How are podocytes affected in nail–patella syndrome?

Nail–patella syndrome is an autosomal-dominant hereditary disease named for dysplastic fingernails and toenails and hypoplastic or absent kneecaps evident in patients with the syndrome. Prognosis is determined by the nephropathy that develops in many such patients. Besides podocyte foot-process effacement, pathognomonic changes in the kidney comprise electron-lucent areas and fibrillar inclusions in the glomerular basement membrane. These characteristic symptoms are caused by mutations in the gene encoding the transcription factor LMX1B, a member of the LIM-homeodomain gene family. Comparable with the human syndrome, homozygous Lmx1b knockout mice lack patellae and suffer from severe podocyte damage. In contrast, however, podocin and the α3 and α4 chains of collagen IV are absent in the glomeruli of Lmx1b knockout mice. Further studies with podocyte-specific Lmx1b knockout mice have confirmed the importance of LMX1B in podocytes, as these mice apparently develop foot processes initially but lose them later on. We therefore conclude that LMX1B is essential for the development of metanephric precursor cells into podocytes and possibly also for maintaining the differentiation status of podocytes. LMX1B can serve as a model system to elucidate a genetic program in podocytes. For many years, the mesangial cell took the forefront of glomerular research (see, for example, [1, 2]), but its pedestal was first slowly shaken by the painstaking morphological investigations of Wilhelm Kriz (confer [3, 4] for an early and late review, respectively) and finally abruptly toppled by irrefutable genetic evidence pointing towards the podocyte as a crucial cell in the glomerulus [5]. Meanwhile, the podocyte is firmly rooted in pathogenetic models of glomerular diseases and it is hard to imagine that it will leave again. Although mutations in WT1, a gene encoding a transcription factor of the Cys2His2-zinc finger family, had been found responsible for the podocytopathies Denys-Drash syndrome [6], WAGR syndrome [7] and Frasier syndrome [8], the article by Karl Tryggvason’s group on the identification of mutations in NPHS1 added another dimension [5]. NPHS1 is mutated in patients suffering from congenital nephrotic syndrome of the Finnish type. It codes for nephrin, a component of the slit diaphragm, and therefore is an essential part of the glomerular filtration barrier [9–11]. Since then several other genes have been cloned that when mutated lead to glomerular disease and which in the kidney are (almost) specifically expressed in podocytes. They are LMX1B [12–14], NPHS2, the gene encoding podocin [15]; ACTN4, the gene encoding α-actinin-4 [16]; CD2AP [17, 18] and TRPC6 [19, 20]. In the following years, an increasing amount of evidence has accumulated on the specific role of these proteins in the podocyte. Nephrin and podocin participate in the formation of the slit diaphragm complex, α-actinin-4 crosslinks actin filaments in podocytes, and TRPC6 belongs to a special class of cation channels. Very little, however, is known about how the podocyte-specific expression of these genes is achieved. The sparse evidence that has been published concerns WT1, which binds to sequences in the promoter regions of the Podxl gene (encoding podocalyxin) [21] and of the NPHS1/Nphs1gene [22, 23]. Although WT1 activates the respective reporter constructs, the induction of the endogenous NPHS1/Nphs1 gene by WT1 was described by one group [23] but not another [22]. Nail–patella syndrome has been known for many decades as a hereditary disease and was one of the first genetic disorders for which linkage was established. In addition to the obvious limb abnormalities, nephrologic symptoms develop in ∼40% of these patients over the course of several decades. On an ultrastructural level, the moth-eaten appearance of the glomerular basement membrane together with fibrillar deposits is considered typical of nail–patella syndrome. In addition, podocytes lose their foot processes (for references, see [24]). In 1998, not only were the first mutations in the LMX1B gene published for patients suffering from nail–patella syndrome [12–14], but a report also appeared on the first characterisation of the Lmx1b knock-out mouse [25]. A more careful analysis of the kidney phenotype in the Lmx1b knock-out mouse revealed pronounced retardation in the development of podocytes that did not elaborate foot processes and slit diaphragms. Consistent with this finding was the splitting of the glomerular basement membrane and the reduced number of endothelial fenestrations, because podocytes synthesise proteins of the glomerular basement membrane and control differentiation of glomerular endothelial cells [26, 27]. Attractive explanations for these morphological defects have come from the observations that the α3 and α4 chains of collagen IV are no longer detected in the glomerular basement membrane ([28] and personal observations), that the Nphs2 gene is no longer expressed in podocytes of homozygous Lmx1b knock-out mice, and that podocytes in homozygous Lmx1b knock-out mice produce less vascular endothelial growth factor (VEGF) [26, 27]. Further molecular analysis demonstrated that LMX1B bound to AT-rich sequences in the first intron of the COL4A4 gene [28] and in the promoter region of the NPHS2 gene [26, 27]. Although the model that LMX1B activates the expression of COL4A4 and NPHS2, and that inactivating mutations in the LMX1B gene secondarily lead to the loss of collagen IV and podocin and therefore to the characteristic alterations in nail–patella syndrome patients certainly is an attractive one, several caveats have to be mentioned as well. Firstly, we could not demonstrate activation of a reporter construct with 4.4 kbp of the NPHS2 promoter by LMX1B [27], although another group showed an approximate twofold activation of the reporter gene controlled by four concatemerised LMX1B binding sites from the NPHS2 promoter [26]. Secondly, when we stably transfected a human cervical carcinoma cell line HeLa cells (which admittedly bears only a very remote similarity to podocytes) with an LMX1B cDNA, we found no upregulation of the endogenous NPHS2 gene [27]. Thirdly, podocin and the α3 and α4 chains of collagen IV were still present in glomeruli from patients with nail–patella syndrome [29]. And fourthly, the constitutive podocyte-specific inactivation of Lmx1b in the mouse does not lead to the loss of podocin or collagen IV [30]. What do these apparently discrepant results mean? If LMX1B already acts at a very early stage of podocyte development, specifically before the NPHS2, COL4A3 and COL4A4 genes are turned on (by other transcription factors?), the podocyte will just not have reached an advanced enough stage of development to produce podocin and collagen IV. In other words, LMX1B may rather exert a permissive influence and, for example, initiate the spreading of the foot processes upon which podocin would be produced and slit diaphragms be elaborated. The constitutive podocyte-specific inactivation of Lmx1b represents a more comparable model for human nephropathy, but the mice only survive for ∼2 weeks after birth [30], again limiting their usefulness. In those animals, podocin and collagen IV α3/α4 are still present, and it appears as if foot processes and slit diaphragms are first elaborated and then lost secondarily (Fig. 1). Does LMX1B therefore play a role not only for the initial development of podocytes but also in the maintenance of their differentiation status? Clearly, more elaborate mouse models with an inducible inactivation of Lmx1b in adult animals are needed to answer this question. Such mice will also have the additional advantage of permitting the isolation of sufficient amounts of glomeruli, which can be used for DNA microarrays and identification of LMX1B target genes. Mouse genetics has already provided us with the verification of LDB1 as an interaction partner of LMX1B [30], and it may in the end help us to identify a genetic hierarchy acting in podocytes by telling us what factors control the expression of LMX1B in podocytes, what other proteins LMX1B interacts with and what genes are regulated by LMX1B in this peculiar cell type. Fig. 1a–f Ultrastructural and immunohistochemical characterisation of mice with podocyte-specific inactivation of Lmx1b. In 11-day-old mice, the podocyte-specific inactivation of Lmx1b leads to the loss of foot processes and to a thickened glomerular basement membrane (arrows in b). However, despite the inactivation of Lmx1b, podocin and the α4 chain of collagen IV are still produced (d, f). +/lox control mice with one wild-type and one floxed Lmx1b allele; lox/lox mice with two floxed Lmx1b alleles; Cre presence of the Cre transgene under control of the human NPHS2 promoter. Bars: 5 μm (a, b), 20 μm (c–f). With permission from [30]

Eur_Spine_J-2-2-1602201

Intraoperative monitoring study of ipsilateral motor evoked potentials in scoliosis surgery

Ipsilateral motor evoked potentials (MEPs) in spinal cord surgery intraoperative monitoring is not well studied. We show that ipsilateral MEPs have significantly larger amplitudes and were elicited with lower stimulation intensities than contralateral MEPs. The possible underlying mechanisms are discussed based on current knowledge of corticospinal pathways. Ipsilateral MEPs may provide additional information on the integrity of descending motor tracts during spinal surgery monitoring. Introduction Intraoperative monitoring (IOM) of the motor pathways is a routine procedure for ensuring integrity of corticospinal tracts during scoliosis surgery. In combination with somatosensory evoked potentials, motor evoked potential (MEP) monitoring is widely utilized in operations with significant risks of spinal cord damage [2, 14]. MEPs are most effectively obtained with multi-pulse cortical electrical stimulation during IOM [3]. However, anesthetic agents, which cause suppression of cortical and spinal motor neuron excitability, affect them [4, 5]. While most IOM protocols involving MEPs utilize total intravenous anesthesia (TIVA), we have previously reported success with desflurane as a halogenated inhalational anesthetic agent [15]. In IOM, MEPs are elicited mostly with contralateral cortical electrical stimulation. Ipsilateral MEP responses have not been adequately studied in this context. Our paper describes observations of ipsilateral and contralateral MEPs with bilateral recordings, in conjunction with TIVA or desflurane during IOM of scoliosis surgery. Methods We studied nine patients (mean age 16.2 years; range 14–17 years; 1 male) over a 6-month period in a prospective manner. The local ethical committee has approved the study protocol. All patients did not have medical conditions contraindicating transcranial electrical stimulation. Apart from scoliosis, they were healthy and had normal neurological examinations. Multi-pulse transcranial electrical stimulation was performed using two constant-current stimulators connected in parallel configuration from a Dantec Keypoint EMG machine (Dantec, Skovlunde, Denmark). A train of five square wave stimuli 0.5 ms in duration was delivered at 4 ms (250 Hz) interstimulus intervals. Stimulating electrodes consisted of 9 mm gold-plated disc electrodes at C3C4 (International 10–20 system) affixed with collodion. Stimulation output was increased from 50 mA in steps of 5 mA until a reproducible MEP was elicited. The intensity was then increased and fixed at 10% above this threshold intensity to obtain a supramaximal MEP response. Each stimulator was capable of delivering a maximum output of 100 mA (200 mA in total). MEP recordings were obtained with 13 mm disposable subdermal needles (Technomed Europe, Beek, Netherlands) in the tibialis anterior (TA) bilaterally. Filter settings were set at 10 Hz and 2 kHz. Input impedance of stimulating and recording electrodes were maintained below 5 kΩ. For induction of anesthesia, sodium thiopentone at 4 mg/kg and fentanyl at 2 mcg/kg was administered; 0.8 mg/kg of intravenous atracurium was used to facilitate endotracheal intubation. No further doses of neuromuscular blocking agents were used subsequently. In the desflurane group, anesthesia was maintained using 60% nitrous oxide in oxygen. Desflurane was introduced through a calibrated vaporizer up to an end-tidal concentration of 2.1–4.3 %, with a mean concentration of 3.4% (approximately 0.5 maximum alveolar concentration). This was measured using an Ohmeda respiratory gas monitor 5250 (BOC Group, Louisville, USA). Closed circuit mechanical ventilation was adjusted to maintain end-tidal carbon dioxide levels between 32 and 35 mmHg. In the TIVA group, anesthesia was maintained using the regime of 10 mg/kg of propofol for the first 10 min, 8 mg/kg for the next 10 min and 5 mg/kg for the subsequent length of operation; 50% air in oxygen was administered. In both groups, morphine was titrated as required for pain relief. Monitoring included electrocardiography, pulse oximetry, capnography and direct radial artery pressures. All patients were kept nornothermic with a warming blanket. Normotensive anesthesia was maintained throughout the operation. After approximately 45 min post-induction, a train of four-twitch assessment was performed using a nerve stimulator (Fischer Paykel NS242, UK). Cortical stimulation was commenced only when the amplitude of the fourth was visibly similar to the first. An interval of 3–5 min was allowed between two trains of cortical stimulation. This alternated with monitoring of somatosensory evoked potentials from posterior tibial nerve stimulation. We measured two parameters: MEP amplitude, onset latency and initial stimulation intensity. Peak to peak amplitudes (between two largest peaks opposite in polarity) and onset latency was utilized for all MEP responses recorded bilaterally. Hence, ipsilateral MEPs refer to MEPs recorded from the TA on the same side as cortical stimulation. Within each patient, ten consecutive supramaximal MEPs obtained before insertion of pedicle screws used as a baseline were averaged to obtain the first two parameters. The initial stimulation intensity was defined as the minimal intensity required to obtain five consistent visible MEP responses at a vertical gain of 20 μV per division. During insertion of pedicle screws and instrumentation, a 50% reduction of MEP amplitude or 10% prolongation of latency was brought to the surgeon’s attention. Statistical analyses using Student’s t-tests were obtained with Microsoft SPSS for Windows Version 10.1. Statistical significance was considered at P < 0.05. Results There were no complaints of headache, seizures or skin burns postoperatively; all patients had normal neurological examination. MEPs were successfully obtained from all patients with TA recordings bilaterally. There were four patients in the TIVA group and five in the desflurane group. Mean ages for desflurane (16.2) and TIVA (15.7) groups were not significantly different (P = 0.6). None of the patients had MEP amplitude or latency changes exceeding our set limits so as to require immediate surgical attention during and after pedicel screw insertion and spinal instrumentation. The ipsilateral MEP amplitudes (standard deviation) were significantly larger than contralateral MEP amplitudes [68.9 (27.1) vs. 52.5 (15.7) μV, P < 0.01, paired t-test]. The initial stimulation intensity to obtain ipsilateral MEPs was significantly lower than for contralateral MEPs [66.9 (12.3) vs. 74.4 (10.1) mA, P < 0.05, paired t-test]. However, there were no significant latency differences for ipsilateral and contralateral MEPs [32.0 (2.1) vs. 31.5 (2.0) ms, P = 0.3, paired t-test). There was no significant difference between use of TIVA or desflurane anesthesia for MEP amplitudes obtained, with ipsilateral (P = 0.06, unpaired t-test) and contralateral (P = 0.09, unpaired t-test) stimulation. Additionally, there was also no significant difference between right and left sided MEP amplitudes, either with ipsilateral (P = 0.9, paired t-test) or contralateral (P = 0.7, paired t-test) stimulation. We consecutively studied an additional 17 subjects monitored for scoliosis surgery using an identical protocol (1 men, mean age 16.1 years, range 14–22). All had the TIVA anesthetic regimen. With right cortex stimulation, mean initial stimulation intensity to obtain ipsilateral MEPs [39.7 (9.9) mA] was significantly lower than to obtain contralateral MEPs [47.1 (11.3) mA, paired t-test, P < 0.0005]. With left cortex stimulation, similar findings were obtained [40.6 (10.7) vs. 50.3 (11.8) mA, paired t-test, P < 0.0005]. With right cortex stimulation, mean ipsilateral MEP amplitudes [107.1 (35.7) μV] were significantly larger than mean contralateral MEP amplitudes [90 (37.1) μV, paired t-test, P = 0.01]. With left cortex stimulation, similar findings were again observed [112.1 (37) vs. 82.3 (30.9) μV, paired t-test, P = 0.0004]. With right cortex stimulation, mean ipsilateral MEP latencies [30.2 (2.4) ms] were not significantly different from mean contralateral MEP latencies [30 (2.3) ms, paired t-test, P = 0.3). With left cortex stimulation, similar findings were again observed [30 (2.1) vs. 30.4 (2.3) ms, paired t-test, P = 0.2]. Examples of MEPs obtained with both ipsilateral and contralateral stimulation are shown in schematically in Fig. 1 and 2. Fig. 1Schematic diagram showing right cortical stimulation, resulting in ipsilaterally and transcallosally conducted corticospinal impulses activating the spinal cord anterior horn cell. The right ipsilateral MEP recording is from the TA. Summation of ipsilaterally conducted and transcallosally generated descending impulses may thus result in larger ipsilateral MEPs from right cortical stimulationFig. 2Actual consecutive MEPs obtained from a patient, showing larger amplitude responses with ipsilateral stimulation. Both recordings were made from the TA at 70% stimulation intensity Discussion The present study showed that ipsilateral MEPs have significantly larger amplitudes and were elicited with significantly lower stimulation intensities than contralateral MEPs. However, onset MEP latencies were not significantly different. The origin of ipsilateral MEPs in humans is not well understood. In animal studies, cat corticospinal neurons have been shown to evoke ipsilateral actions via ipsilaterally descending reticulospinal tracts, as well as via contralaterally descending reticulospinal neurons, both by synapsing spinal interneurons [9]. Tracer studies in rhesus monkeys have quantified ipsilateral corticospinal fibers as approximately 9–12% of the total descending corticospinal projections [14]. Thus, current evidence points to contralateral corticospinal fibers as the predominant pathway for spinal motorneuron activation. The presence of ipsilateral MEPs has mostly been described in pathological conditions. Patients with congenital mirror movements [11], schizencephaly [12], and Kallmann’s syndrome [13] show ipsilateral MEPs, which likely result from abnormal structure and function of ipsilateral corticospinal fibers. However, functional reorganization and unmasking of ipsilateral corticospinal pathways may contribute to the generation of ipsilateral MEPs after adult stroke [6] and congenital hemiparesis [18]. One study involving 50 normal children suggested that presence of ipsilateral MEPs might be a normal state of ontogeny. Their disappearance after 10 years old is likely due to increasing transcallosal inhibitory influences [16]. In our study, all patients were above 10 years of age, and did not have clinical features to suggest presence of underlying conditions mentioned above. Another study comparing healthy adults with stroke patients has suggested ipsilateral MEPs may be conducted via corticoreticulospinal or corticopropriospinal pathways in normal subjects [1]. What are the possible underlying mechanisms, which explain our findings? Firstly, it is possible that ipsilateral MEPs may be solely due to transcallosal stimulation of the contralateral motor cortex. Additionally, the effects of anesthesia on corticospinal excitability may facilitate this, hence resulting in significantly lower initial stimulation intensity to obtain ipsilateral MEPs. While evidence to suggest this is scarce, rat brain studies have demonstrated widespread action of anesthesia at multiple binding sites [8]. Magnetic resonance brain imaging has also demonstrated increased callosal T2 changes with anesthesia, suggesting structural alterations at a molecular level [19]. It is also possible that longstanding scoliosis has led to spinal cord plasticity changes. Motor pathway reorganization and spinal cord plasticity have been well documented in response to cord injury [7] in an activity-dependent manner [10]. Thus, structural and postural changes of longstanding scoliosis may have resulted in reorganization of cortical or subcortical motor pathways, including ipsilateral corticoreticular fibres leading to our observations [17]. However, lack of lateralization of MEP amplitudes with ipsilateral or contralateral stimulation was not supportive of this hypothesis. Additionally, lack of significant ipsilateral and contralateral latency differences suggest bilateral motor cortex stimulation has resulted in ipsilateral MEPs, which may have comprised early ipsilaterally conducted components and late transcallosally stimulated components (Fig. 1). This might also explain the larger amplitudes of ipsilateral MEPs obtained than MEPs derived from contralateral motor cortex stimulation. Further studies clarifying the predominant mechanisms responsible would be interesting. Are ipsilateral MEP responses useful and relevant in clinical settings? Ipsilateral MEPs are readily elicited, as shown in this study. While the relative contributions of ipsilaterally and transcallosal conducted MEPs remain uncertain, bilateral MEP recordings during spinal surgery IOM may provide additional information regarding the integrity of descending motor tracts. Together with the electrophysiological findings presented here, future studies clarifying these aspects would be justified. It thus may be feasible to routinely monitor MEPs bilaterally in future IOM protocols for spinal surgery.

Clin_Rheumatol-4-1-2262146

A 57-year-old man who developed arthritis during R-CHOP chemotherapy for non-Hodgkin lymphoma

Rituximab is a chimeric human-mouse anti-CD20 monoclonal antibody, which is used in the treatment of both B-cell lymphomas and rheumatic diseases. We describe a case of a previously healthy 57-year-old man developing arthritis while being treated with rituximab-CHOP chemotherapy (R-CHOP) for a non-Hodgkin lymphoma. The remittant arthritis developed at successively shorter time-intervals after R-CHOP administration and only improved after rituximab was removed from the chemotherapy schedule, suggesting a rituximab-related phenomenon, as extensive diagnostic testing ruled out any other diagnosis. Introduction Rituximab is increasingly used in the treatment of both B-cell lymphomas as well as rheumatic diseases. Several side-effects of rituximab are known, but until now only two reports described the development of arthritis after rituximab therapy [1, 2]. Here we describe yet another case. Case A 57-year-old man was referred to the out-patient clinic of rheumatology because of arthritis of the right knee and left wrist. The patient had been in perfect health until he noticed a swelling of 2.5 cm under his right jaw. He was referred to a specialized oncology center where he was diagnosed with an aggressive follicular B-cell non-Hodgkin lymphoma (NHL) stage IVA with involvement of submandibular and mesenteric lymph nodes and bone marrow. Treatment was instituted with R-CHOP chemotherapy, which consists of cyclophosphamide, doxorubicin, vincristine, prednisone, and rituximab, given at a 3-week interval. During this treatment, granisetron (Kytril®) was given and occasionally magnesiumoxide and paracetamol. Because this patient suffered from a low tumor burden, no tumor lysis profylaxis, i.e., allopurinol was prescribed. Two and a half weeks after the first course of R-CHOP, the patient developed arthritis of his right knee and left wrist without morning stiffness or involvement of the small hand joints. He had no history of arthralgia, joint swelling, preceding trauma, fever, or infection nor did he have a history of inflammation of the eyes, Raynaud phenomenon, photosensitivity, sicca-syndrome, or inflammatory back pain. The further medical history disclosed several tick bites the year before, with a skin reaction. He rarely used alcohol (maximum of two glasses a day). Family history revealed no rheumatic diseases. Physical examination showed a profusely swollen and painful right knee and left wrist. The other joints were not afflicted. Laboratory testing before the start of chemotherapy showed no abnormalities besides a erythrocyte sedimentation rate (ESR) of 21 mm/h. At the time of presentation with arthritis, there was ESR of 41 mm/h, C-reactive protein 256 mg/l, leucocytes 11.1 × 109/l, a normal liver and kidney function, and an IgM-rheumatoid factor (RF) 7 IE/ml, anti-cyclic citrulline peptide (a-CCP) <1 AU/ml, anti-nuclear factor (ANF) negative, and urate of 0.25 mmol/l. A diagnostic puncture of synovial fluid from the right knee was performed, which revealed an inflammatory arthropathy (cell count 5,300/mm3, 14% lymphocytes, 8% mononuclear cells, and 78% granulocytes) with no signs of malignant cells or crystals. Microbiological testing (Gram-stain, Ziehl–Neelsen, and culture) of the synovial fluid showed no microorganisms. X-rays of both knee and wrist did not show bone involvement or chondrocalcinosis. The patient was treated with naproxen (500 mg, twice daily), with which both pain and swelling of the joints were reduced but not completely resolved. Four days after the second course of R-CHOP, the pain and swelling returned in all its severity and spontaneously regressed after 1 week. Shortly after the third course of R-CHOP, the patient experienced the same painful joint swellings, now also involving the left knee. This was in spite of intra-articular kenacort injections and prednisone maintenance therapy (10 mg daily), which was started between the courses of chemotherapy. Repeated pathological and microbiological testing of synovial fluid showed no new results. After these three courses of R-CHOP, the effect on the NHL was evaluated and disclosed that the intra-abdominal NHL-mass was somewhat reduced, and the remnant of the submandibular lymph node had disappeared. Since the causal relationship between the (activity of) arthritis and the courses of R-CHOP and rituximab being the only component of R-CHOP of which the development of arthritis is described after the administration of this drug [1, 2], the decision was made to give the fourth and fifth course of chemotherapy without rituximab. This time there was no increase in pain or swelling of the joints. Well over a year after the first R-CHOP chemotherapy, the patient is fully recovered. Evaluation of his NHL shows a near complete remission, and no further treatment was indicated. The arthritis has fully resolved. Discussion Arthritis developing during the course of R-CHOP chemotherapy is very unusual. The arthritis developed after successively shorter time-intervals after the R-CHOP administration while no arthritis occurred when CHOP was given without rituximab. This suggests a probably rituximab related, immune-mediated phenomenon. We were not able to detect antibodies against rituximab or immune complexes that might strengthen this diagnosis. However, we did have strong arguments against the other differential diagnoses. The pattern of arthritis in this patient is unusual for a rheumatoid arthritis; moreover, both IgM-RF and a-CCP were negative, while X-rays made of the afflicted joints disclosed no erosions. Spondylarthropathy very rarely develops after the age of 45 years. Moreover, inflammatory back pain was absent, and positive arguments for an infection causing reactive arthritis were lacking. Other autoimmune diseases like systemic lupus erythematosus (SLE) or Sjögren’s syndrome were also very unlikely looking at the clinical presentation, the absence of characteristic signs and symptoms, and the absence of the specific autoantibodies. A crystalarthropathy is a more likely candidate, as treatment of NHL can cause tumor lysis syndrome, which can give rise to gout. However, serum values of urine acid were repeatedly very low, and there where no other signs of tumor lysis syndrome. Moreover, repeated samples of synovial fluid failed to show any crystals, kidney function was normal; the patient barely used alcohol and did not use diuretics. The X-rays of both the right knee and left wrist did not show signs of chondrocalcinosis. Latent infections provoked by the administration of R-CHOP were considered. However, repeated cultures, Gram-stain, and Ziehl–Neelsen of synovial fluid did not disclose any bacteria or yeasts. Serologic testing for Borrelia Burgdorferi, Lues, Chlamydia trachomatis, Parvovirus B19 and PCR on Mycobacteria tuberculosis and Chlamydia trachomatis (both performed in the synovial fluid) proved to be negative. A wide spectrum of autoimmune manifestations, among others arthritis, is described in NHL [3]. However, arthritis is unlikely in this patient as autoimmune manifestation, as it developed during R-CHOP and the NHL responded well to this chemotherapy. Furthermore intra-articular localization of NHL was considered. However, both primary and secondary intra-articular localization of NHL is very rare [4], and the fact that more than one joint is afflicted in this patient makes it even more unlikely. Moreover, magnetic resonance imaging (MRI) of both knees did not show tumor mass, and repeated punctures of synovial fluid of both knees did not show any malignant cells. On top of this, a computed tomography (CT) scan of the abdomen revealed regression of the mesenteric NHL-tumor mass after chemotherapy, which means that in the case of any presumed intra-articular localization of NHL, one would expect relapse of joint involvement in the course of chemotherapy instead of progression. An arthroscopy was considered but not performed because of the risks of infection in this immunocompromised patient, and the low expectancy of achieving the diagnosis by performing the arthroscopy as the presence of NHL intra-articular was considered to be very unlikely. Because our patient received high dosages of prednisone, the diagnosis of avascular bone necrosis was considered but ruled out by a MRI of both knees. None of the other medications our patient used are known to give arthritis. Ruling out the above-mentioned diagnoses and taking into account the relationship in time between the development of arthritis and the gifts of R-CHOP, one or more of the components of R-CHOP seem to be the most likely cause of the arthritis in our patient. Of all of the components of R-CHOP, only rituximab is known to be able to give arthritis [1, 2]. Pijpe et al. described the development of arthritis in three patients who received rituximab for primary Sjögren’s syndrome. All of these patients had developed human antichimeric antibodies (HACAs), which suggests an immune-mediated mechanism. Only one other patient tested positive for HACAs but was asymptomatic. The significance of HACAs being present is actually not quite sure as other studies show HACAs being present without specific clinical manifestations [5, 6] or show HACAs being present but later becoming undetectable [6]. Repeated blood tests of our patient did not show any HACAs. The precise mechanism of action in the development of arthritis in our patient is still unknown. In conclusion, we describe a patient developing severe oligo-arthritis at successively shorter time-intervals after receiving CHOP-rituximab for a NHL. The arthritis is probably caused by an immune-mediated reaction to rituximab. As it is to be expected that rituximab will be used more frequently in future to treat both NHL as rheumatological diseases, we think it is important for physicians to be aware that rituximab may cause severe (oligo-)arthritis as a side effect.

Eur_J_Epidemiol-4-1-2249618

Use of recommended medications after myocardial infarction in Austria

Guidelines recommend long-term use of beta-blockers (BB), statins, and angiotensin-converting-enzyme-inhibitors or angiotensin-receptor-blockers (ACEI/ARB) after myocardial infarction (MI), but data on their use after discharge are scarce. From Austrian sickness funds claims, we identified all acute MI patients who were discharged within 30 days and who survived ≥120 days after MI in 2004. We ascertained outpatient use of ACEI/ARBs, BBs, statins, and aspirin from all filled prescriptions between discharge and 120 days post MI. Comorbidities were ascertained from use of indicator drugs during the preceding year. Multivariate logistic regression was used to evaluate the independent determinants of study drug use. We evaluated 4,105 MI patients, whose mean age was 68.8 (±13.2) years; 59.5% were men. Within 120 days after MI, 67% filled prescriptions for ACE/ARBs, 74% for BBs, and 67% for statin. While 41% received all these classes and 34% two, 25% of patients received only one or none of these drugs. Older age and presence of severe mental illness were associated with lower use of all drug classes. Diabetics had greater ACEI/ARB use. Fewer BBs were used in patients with obstructive lung disease. Statin use was lower in patients using treatment for congestive heart failure (all P < 0.001). We conclude that recommended medications were underused in Austrian MI survivors. Quality indicators should be established and interventions be implemented to ensure maximum secondary prevention after MI. Introduction Cardiovascular disease is among the leading causes of premature death and morbidity in most Western societies. Fortunately, the prognosis of myocardial infarction (MI) has steadily improved over the past decades: case fatality after MI has decreased and long-term survival has increased [1–3], likely due to the introduction and increasing availability and use of acute invasive and non-invasive interventions as well as the introduction of powerful medications for secondary prevention [4]. Among the medications for long-term secondary intervention, aspirin, beta-receptor blockers (BB), 3-hydroxy-3-methyl-glutaryl-co-enzyme-A inhibitors (statins), and angiotensin converting enzyme (ACE) inhibitors have drawn the most attention. More recently, angiotensin receptor blockers (ARB) have become available for use in those patients who develop adverse events from ACE inhibitors; ACE inhibitors and ARBs have been shown to be equally efficacious and effective after MI [5–7]. All these interventions are highly efficacious in reducing the recurrence of MI, the subsequent use of coronary interventions, or the risk of death, and have been demonstrated to be highly attractive economic investments from a societal perspective [8–11]. Practice guidelines recommend the use of these medications for secondary prevention in patients with unstable angina or MI, regardless whether the MI is associated with ST-elevation [12–14]. Considerable variation has been observed, however, in the use of these beneficial interventions [3, 15–17], which has prompted the establishment of quality of care indicators in some countries with the goal to monitor appropriate use of these interventions during hospitalization, at discharge, and during outpatient follow-up after such an event [18–20]. Most relevant in the long run, however, is the outpatient use of these medications after discharge from MI. Several studies have indicated that these drugs are underused for secondary prevention. Most of these investigations were conducted in North America [21–23] and in Europe [24–29]. No such systematic assessment has been conducted in the Austrian health care system except for a small local study [30]. We sought to fill this void by studying proportions of BB, statin and ACE inhibitor or ARB use after MI in a large period cross-section of Austrian MI patients. Methods Data sources Austria is a social welfare state that mandates universal health care to its residents [31]. Patients are assigned membership in one of several sickness funds dependent on their type and location of current or former employment. At the time covered by this study, there were 20 sickness funds in Austria; among those, the 9 provincial sickness funds covering more than three quarters of all residents (6.2 Mio. of 8.2 Mio. overall; data from 2004). Most inpatient and outpatient medical services are covered and each service encounter generates a claim to the health care system. Inpatient services are not paid directly by sickness funds, but by a separate entity, which, however, is partly funded by these. Thus, sickness funds are also informed about each encountered inpatient care. Similarly to most western health care systems, datasets provided by hospitals contain diagnosis codes, admission and discharge dates. Access to prescription drugs is equal across sickness funds and all medications that are deemed efficacious by a national panel are reimbursed. The copayment for medicines per package was €4.25 in 2003, €4.35 in 2004 and €4.45 in 2005. Packages of chronic medications usually contain 28, 30, or 50 pills. Sickness fund premiums, other coinsurance and copayments vary slightly, and indigent patients can apply for waiver of the usually modest copayment for prescription drugs. Few medications are subject to prior authorization by the sickness funds. Prescription claims contain a unique identifier for the specific drug, the dose, and the quantity dispensed. Study population For the purpose of this study, we used the complete claims data of individuals covered by several of these sickness funds: the provincial sickness funds (Gebietskrankenkasse) of Vienna, Lower Austria, Upper Austria, Styria, Burgenland, and Carinthia, as well as the funds covering all federal employees (Bundesversicherungsanstalt, BVA) and farmers (Sozialversicherung der Bauern, SVB), respectively. Cumulatively, these sickness funds cover approximately 6.1 million members of the total Austrian insured population of 8.2 million. Each sickness fund separately identified from their insurance claims all patients who were hospitalized and discharged with a primary diagnosis of acute MI in 2004 (International Classification of Diseases, ICD-9: 410.xx; ICD-10: I21) and provided us with their relevant anonymized health care claims data covering the period from January 1, 2003 to June 30, 2005. We only studied each patient’s first hospitalization for MI (including direct transfers from one hospital to another) and required that the admission date be between January 1, and December 31, 2004. We retained only those patients whose hospitalization exceeded 3 days and who were admitted at an acute care hospital. Including patients who were hospitalized <4 days would open up the possibility of wrongly selecting patients into the study sample who had not experienced an MI (e.g., whose MI was ruled out or patients who were admitted for a diagnostic or therapeutic intervention and still coded for acute MI). Our approach of identifying patients with MI has been validated in similar claims data and found to be highly accurate (positive predictive value: 94%) [32]. Since we were interested in medication use after MI, we generated a uniform outcomes ascertainment window and required all patients to be discharged within 30 days and to have survived for ≥120 days from their initial admission date (=index date). We also excluded those patients whose hospitalization occurred at a non-acute care hospital. Outcomes: use of recommended medications after myocardial infarction From all filled prescriptions, we recorded out-of-hospital use of several medications between date of discharge and 120 days after MI admission: aspirin, beta-blockers, statins, and ACE inhibitors or ARBs. We also assessed the total number of distinct medications that patients received among statin, BB, and ACE inhibitor/ARB (minimum 0; maximum 3). Since aspirin was available for a price that was below the amount of the drug copayment (€4.35), it is possible that insurance claims data may lead to under-ascertainment of aspirin use. Thus, we decided to investigate aspirin separately, and only among patients who had their copayment waived based on income grounds. These indigent patients had a clear economic incentive to fill prescriptions for aspirin via the prescription route, thus generating a claim to the sickness fund. Covariates We created variables indicating each patient’s age in 2004, gender, the length of stay for their index admission, and whether a patient had their prescription copayments waived during the study period (NoCopay vs. Copay). Since claims from outpatient encounters did not contain any diagnosis codes, we ascertained medication use in the 365 days prior to the index date as proxies for several comorbidities using the anatomic therapeutic chemical (ATC) classification system: oral antidiabetic drugs, insulin, antigout medications, several cardiovascular drug classes (alpha blockers, BBs, calcium channel blockers, ACE inhibitors, ARBs, other antihypertensives, diuretics, nitrates, digitalis, vitamin K-antagonist, statins, fibrates), medications indicated for asthma or chronic obstructive pulmonary disease (COPD; inhaled corticosteroids, beta-receptor agonists), drugs reducing gastro-esophageal reflux or ulcer disease (histamine-H2-blockers, proton pump inhibitors, sucralfate, antacids), pain medications (including non-steroidal anti inflammatory drugs, selective COX-2 inhibitors, opioids, and others), oral corticosteroids, and several psychoactive drug classes (benzodiazepines or anxiolytics, antidepressants, antipsychotics). We also recorded the number of hospital days during the year prior to the index date (categorized into: no hospitalization, 1–7, 8–21, and >21 days). Statistical analysis We plotted the unadjusted proportions of medication use for the overall population as well as by copayment status along with the corresponding 95% confidence intervals (CIs). We then used univariate and multivariate logistic regression to estimate the crude and multivariate adjusted odds of receiving a given study medication. Since none of our outcomes were rare, we were able to create full multivariate models that included all variables regardless of their statistical significance. In large datasets where outcomes are not rare, full multivariate models are superior to parsimonious models, because they provide better control for residual confounding compared to more restricted models. Odds ratios (OR) were presented with their 95% CIs. Additionally, we showed the population distribution of the number of different study drug classes received among statin, BB, and ACE inhibitor/ARB (minimum 0; maximum 3). Multivariate ordinal logistic regression and linear regression were used to model the associations between covariates and the number of drugs received. All analyses were conducted in the full final study population as well as after restriction to new users of each study drug, i.e. patients who had not received the respective study drug in the year prior to admission for MI. We used the SAS for Windows (release 9.2) software for all statistical analyses (The SAS Institute, Cary, NC). Results Study population We first identified 8,416 hospitalizations with a discharge diagnosis indicating an MI. After excluding all repeat MI hospitalizations and those that occurred before or after 2004 (N = 1,977; 23.5%), we also excluded those patients who died within 120 days after admission for MI (N = 1,221; 14.5%). After further restricting the sample to those patients whose index hospitalization was ≥4, but ≤30 days and excluding admissions at non-acute care hospitals, we obtained a final study sample of 4,105 patients (48.8% of the original hospitalization sample drawn). The mean age was 68.8 years (standard deviation, SD: ±13.2), and 2,442 (59.5%) patients were men. Age differed considerably by gender: men (median: 66 years, interquartile range: 56–75) were substantially younger than women (median: 76 years, interquartile range: 67–82; P < 0.001). Overall, 654 (15.9%) had their prescription copayments waived on the grounds of low income (Table 1). On average, they had been hospitalized for 6.7 (±14.7) days in the year prior to their MI and the mean length of stay for their index admission was 10.9 (±5.3) days. The proportions of selected prescription drugs used in the year prior to admission for MI are listed in detail in Table 1. Table 1Characteristics of study population (N = 4,105)VariableCount (%) or mean (±SD)Age68.8 (±13.2)    <50 years402 (9.8)    50–69 years1,515 (36.9)    70–89 years1,225 (29.8)    ≥90 years963 (23.5)Male gender2,442 (59.5)Length of stay10.9 (±5.3)Days of hospitalization in prior year6.7 (±14.7)    None2,500 (60.9)    1–7 days595 (14.5)    8–21 days612 (14.9)    ≥21 days398 (9.7)Copayment waived654 (15.9%)Previous medication use    Alpha blocker243 (5.9)    ACE-inhibitor or ARB1,853 (45.1)    Beta-blocker1,454 (35.4)    Calcium channel blocker809 (19.7)    Other antihypertensive734 (18.0)    Diuretic927 (22.6)    Nitrate1,032 (25.1)    Digitalis347 (8.5)    Acetylsalicylic acid1,169 (28.5)    Clopidogrel or Ticlopidine385 (9.4)    Vitamin K-antagonist249 (6.1)    Statin1,043 (25.4)    Fibrate116 (2.8)    Oral hypoglycemic 620 (15.1)    Insulin265 (6.5)    Uric acid lowering drug559 (13.6)    Pain medication1,934 (47.1)    Gastroprotective drug1,584 (38.6)    Asthma/COPD610 (14.9)    Corticosteroid379 (9.2)     Benzodiazepine or anxiolytic491 (12.0)    Antidepressant623 (15.2)    Antipsychotics190 (4.6) Secondary prevention after myocardial infarction Within 120 days of their admission for MI, 3,037 (74.0%; 95% CI: 72.6–75.3%) patients filled a prescription for a BB, 2,745 (66.9%; 95% CI: 65.5–68.3%) for an ACE inhibitor or an ARB, and 2,736 (66.7%; 95% CI: 65.3–68.1%) received a statin (Fig. 1). Indigent patients who were not required to pay a copayment received more ACE inhibitors or ARBs (74 vs. 65.5%; P < 0.001), and slightly fewer statins (63.2 vs. 67.2%; P = 0.03), while the proportion of BB use was similar to patients without such a waiver (P = 0.86). Fig. 1Proportions of medication use after myocardial infarction The distribution of number of distinct drug classes received in this population was as follows: 1,692 patients (41.2%) received a BB, statin, and an ACE inhibitor or ARB, while 1,391 (33.9%) received two of these drugs. Six hundred and sixty (16.1%) patients received only one of the study drugs whereas 362 (8.8%) patients received none of these medications for secondary prevention at all. This distribution did not differ between patients with vs. without copayment requirement (P = 0.33). The number of drugs received was slightly higher among men compared to women (2.1 vs. 2.0; P = 0.02) and decreased with higher age (P < 0.001). Among the 654 patients with waived copayment, 464 (71%) filled a prescription for aspirin within 120 days of admission for MI. Independent predictors of study medication use Age was an independent predictor of recommended medication use after MI: compared to patients younger than 50 years, those between 70 and 89 years had 48% greater odds of receiving ACE inhibitors or ARBs (OR: 1.48), whereas the oldest patients (≥90 years old) were least likely to receive these drugs (OR: 0.73; Table 2). Similarly, these oldest patients were also markedly less likely to receive BBs and statins (Tables 3 and 4). Number of days spent in the hospital in the year preceding the MI admission, a marker of preexisting comorbidity, was consistently and inversely associated with lower use of all these recommended medications (all P for trend <0.001). Patients who had their copayments waived were more likely to fill a prescription for an ACE inhibitor or an ARB after discharge from MI than patients with copayment required (OR: 1.35; 95% CI: 1.10–1.67), but BB or statin use did not differ between these groups. Angiotensin converting enzyme inhibitor or ARB use was further independently associated with previous calcium channel blocker use, use of an oral antidiabetic drug, and prior use of antipsychotic drugs (Table 2). Beta-receptor blockers were more likely to be used in patients who had used other antihypertensive medications prior to MI (Table 3). Lower use of BBs was observed in patients who had received diuretics, vitamin K-antagonist, insulin, antipsychotics, as well as in those patients who received inhaled medications for asthma or COPD (Table 3). Medications that were inversely associated with statin use were diuretics, digitalis, clopidogrel, vitamin K-antagonist, and antipsychotics (Table 4). Table 2Independent determinants of ACE-inhibitor or ARB useVariableAll patients (N = 4,105)New users (N = 2,252)OR95% CIOR95% CIAge<50–Referent–Referent50–691.190.96–1.491.230.96–1.5870–891.481.19–1.851.541.19–2.00≥900.730.59–0.900.790.60–1.03Male gender1.070.91–1.251.190.98–1.44Length of stay1.021.00–1.031.031.01–1.05Hospital days*0–Referent–Referent1–70.680.56–0.840.620.49–0.808–210.790.63–0.990.680.51–0.91>210.510.39–0.680.530.35–0.79Copayment waived1.351.10–1.671.341.03–1.74Alpha-blocker1.070.75–1.531.270.73–2.20ACE-inhibitor or ARB5.674.74–6.78––Beta-blocker1.120.94–1.331.160.93–1.45Calcium channel-blocker1.351.10–1.661.341.01–1.79Other anti-hypertensive agents0.990.80–1.211.120.85–1.47Diuretic0.850.68–1.050.870.65–1.18Nitrate0.880.72–1.070.820.63–1.08Digitalis1.150.84–1.571.040.65–1.65Aspirin0.890.74–1.070.910.70–1.18Clopidogrel0.900.68–1.200.730.46–1.15Vitamin K-antagonist0.990.70–1.411.100.65–1.87Statin1.120.92–1.380.840.63–1.11Fibrate1.210.76–1.931.000.56–1.77Oral antidiabetic1.261.00–1.581.060.79–1.44Insulin1.020.73–1.441.200.68–2.10Asthma/COPD1.070.86–1.341.100.83–1.47Benzodiazepines/anxiolytics1.060.83–1.361.160.83–1.62Antidepressants0.870.70–1.090.850.64–1.13Antipsychotics0.640.45–0.910.650.41–1.04* P for trend <0.001Table 3Independent determinants of beta-blocker useVariableAll patients (N = 4,105)New users (N = 2,651)OR95% CIOR95% CIAge <50–Referent–Referent50–691.250.98–1.601.210.93–1.5970–891.050.83–1.331.140.87–1.47≥900.620.51–0.760.540.43–0.69Male gender1.140.97–1.351.190.98–1.43Length of stay1.000.98–1.011.000.98–1.01Hospital days *0–Referent–Referent1–70.920.74–1.150.850.66–1.088–210.770.61–0.960.780.60–1.02>210.570.44–0.750.610.43–0.84Copayment waived1.090.89–1.351.190.93–1.53Alpha-blocker1.541.07–2.201.731.12–2.67ACE-inhibitor or ARB1.221.02–1.451.160.94–1.41Beta-blocker4.363.55–5.35––Calcium channel-blocker1.291.05–1.591.491.16–1.90Other anti-hypertensive agents0.830.68–1.010.870.69–1.10Diuretic0.800.65–0.980.770.60–0.99Nitrate1.030.84–1.261.040.81–1.33Digitalis0.790.60–1.040.680.48–0.96Aspirin0.960.79–1.160.880.70–1.12Clopidogrel0.790.59–1.070.650.43–0.97Vitamin K-antagonist0.670.49–0.930.440.29–0.66Statin0.940.76–1.160.850.67–1.10Fibrate1.140.70–1.861.210.65–2.25Oral antidiabetic1.060.85–1.321.070.82–1.39Insulin0.720.53–0.980.780.52–1.16Asthma/COPD0.670.55–0.830.630.49–0.80Benzodiazepines/anxiolytics1.331.04–1.711.270.94–1.72Antidepressants0.880.71–1.090.960.74–1.24Antipsychotics0.710.51–1.000.740.50–1.09* P for trend <0.001Table 4Independent determinants of statin useVariableAll patients (N = 4,105)New users (N = 3,062)OR95% CIOR95% CIAge <50–Referent–Referent50–691.381.09–1.761.401.09–1.8170–891.080.86–1.361.090.85–1.39≥900.390.32–0.470.370.30–0.46Male gender1.100.94–1.291.130.95–1.35Length of stay0.980.97–0.990.990.97–1.00Hospital days*0–Referent–Referent1–70.880.71–1.090.820.66–1.048–210.680.55–0.850.680.53–0.87>210.480.37–0.630.470.34–0.65Copayment waived1.090.89–1.341.070.85–1.34Alpha-blocker0.890.65–1.221.040.72–1.51ACE-inhibitor or ARB0.980.82–1.161.040.86–1.25Beta-blocker1.080.91–1.291.070.88–1.30Calcium channel-blocker1.010.83–1.230.980.79–1.23Other anti-hypertensive agents0.970.80–1.180.970.78–1.21Diuretic0.820.67–1.000.790.63–1.00Nitrate0.940.77–1.140.960.77–1.21Digitalis0.600.46–0.800.510.37–0.72Aspirin0.850.71–1.030.740.60–0.92Clopidogrel0.700.51–0.940.610.40–0.94Vitamin K-antagonist0.620.45–0.850.520.35–0.78Statin6.395.03–8.11––Fibrate2.241.36–3.702.311.35–3.97Oral antidiabetic0.950.77–1.180.980.77–1.26Insulin0.760.56–1.040.710.48–1.05Asthma/COPD0.870.71–1.070.840.66–1.06Benzodiazepines/anxiolytics0.890.70–1.120.940.72–1.22Antidepressants1.130.91–1.401.080.85–1.39Antipsychotics0.480.34–0.680.530.36–0.79* P for trend <0.001 Naturally, prior use of a medication class was highly predictive of use of that class after MI in the full study population. Results from the analyses restricted to new users of each study drug were very similar to the full population analyses, albeit with wider confidence intervals (Tables 2–4). Analyses of independent predictors of number of recommended drugs received using linear or ordinal logistic regression confirmed what could be gleaned from the analyses of each drug class: older age, more hospital days in the previous year, indicator drugs for severe heart failure (digitalis, diuretics, vitamin K-antagonist), and antipsychotic use were all independently associated with fewer recommended medications received (results not shown). Gender, however, was not independently associated with number of medications received. Discussion In a large population-based study of patients who experienced an acute MI in Austria, we found that ambulatory use of several recommended medications after discharge was suboptimal. Within 120 days after their MI, only 74% patients received a BB, 67% a statin, and 67% filled a prescription for an ACE inhibitor or ARB. Only 41% of patients received all three interventions, while 25% of patients received only one of these beneficial drugs or even none at all. Underuse of these medications is unfortunate from both a patient’s and a societal perspective: while optimal secondary prevention including these drugs prolongs the expected lifespan of a patient after MI, use of these interventions constitutes an attractive allocation of scarce economic resources. Thus, considerable room for improvement is present in the care of patients after MI in Austria, and interventions ought to be targeted towards increasing the prescribing and use of these medications. While this is the first large-scale study of the quality of post-MI care in Austria, similar investigations have been conducted in other countries. The most relevant comparison can be drawn with results from the EUROASPIRE studies [27–29]. These studies, three waves have been conducted to date, used surveys to evaluate coronary care in 9 (EUROASPIRE I), 15 (EUROASPIRE II), and 22 (EUROASPIRE III) countries, respectively; Austria did not participate in either of these. The most recent data available are from EUROASPIRE II, which covered the years 1999/2000 [28]. Medication use was assessed at admission, discharge (abstracted from charts), and at least 6 months (from patient interview) after the acute admission or procedure. While some variation was observed across countries, data specifically for after MI drug use was only reported in aggregate across countries. Approximately 74% of MI patients received a BB at discharge and 68% >6 months after the event; these proportions are similar to the 74% observed in our study. For ACEI, the proportions at discharge and >6 months after the event were 49 and 45%, respectively, considerably lower than the 67% observed in Austria. It is unclear, however, whether ARB use was regarded equivalent to ACEIs and included in this quality measure in the EUROASPIRE study. Lipid-lowering drugs were prescribed to 42% of patients at discharge and used by 60% at >6 months after the event, compared to 67% in our study [28]. A comparison of these reports, however, needs to be conducted with caution. Considerable differences exist between these and other studies, especially in their population selection criteria, ascertainment of medication use, and the year(s) studied. Several studies, specifically from the United States and Canada, investigated medication use only in older patients (>65 years) [33]. Results from elderly populations cannot be compared with findings from the general population, since age is a strong predictor of preventive medication use. Studies that measured medication use using discharge notes are likely to overestimate medication use, because patients may not have received a prescription or had chosen to not fill it. The increasing use of the study medications also needs to be taken into consideration and, in this respect, older reports considerably vary with recent ones. To this point, Gislason et al. [24] studied MI patients in Denmark and found that the proportion of patients who received an ACE inhibitor increased from 25% in 1995 to 36% in 2002. Similarly, use of BB increased from 38% to 68% during that time. This study differs from ours in that drug use was ascertained within 30 days following the discharge date from MI. While BB use is roughly similar to our findings, the use of ACE inhibitor was substantially lower in that Danish cohort. Statin or aspirin use was not part of that study, but a separate study using the same Danish database focused on this aspect of post MI care. In this report, statin use was ascertained within 6 months of discharge and increased from 13% in 1995 to 61% in 2002 [34]. This percentage is similar to the statin use of 67% in our Austrian sample, given that our data are slightly more recent; we did, however, use a shorter ascertainment window. Another report studied the use of preventive medications after MI in The Netherlands [26]. The authors confirmed a trend towards increased use of all drug classes from 1991 to 2000, with ACE inhibitor being dispensed to 44%, BB to 76%, and statins to 58% in the most recent study year (2000). Use of these medications was ascertained within 30 days after discharge, but patients were not required to survive for a minimum number of days after discharge to be eligible for study. Taking these study design differences into account, it appears that more patients received a BB after discharge in the Netherlands compared to Austria. Interestingly, neither the Dutch, nor the Danish study took ARB use into consideration as an established substitute for ACE inhibitors. Similarly, data from EUROASPIRE I and II confirms increasing use these preventive medications [29]. In the present study, we also identified several predictors of use for the study medications. In general, patients ≥90 years of age were substantially less likely to receive these recommended medications than younger patients. Similar evidence for under-treatment among the elderly was also found in other countries [24]. This behavior may constitute general treatment bias against very old patients, or reflect that these patients may not live long enough to reap the benefits of long-term preventive interventions. Indeed, specific evidence on the efficacy and cost-effectiveness of the study medications in very old patients is lacking. In general, old patients were excluded from participation in virtually all efficacy trials on which current clinical practice recommendations are based. Only the PROspective Study of Pravastatin in the Elderly at Risk (PROSPER) trial has addressed this issue and specifically studied older individuals. PROSPER demonstrated that statins are efficacious in prevention of coronary events in older patients, but the participants in PROSPER were still “only” 70–82 years at enrollment [35]. Further, statin therapy has been shown to be cost-effective in 75–84 years old patients [36]. We found that those indigent patients who had their copayment waived had at least the same level of medication use compared to relatively more affluent patients who were responsible for the copayment portion of their prescription (€4.35), thus indicating that such a waiver successfully removed the economic barriers to filling these prescriptions. Of note, patients whose copayment was waived were more likely to be women and previous recipients of antipsychotic drugs, but otherwise similar to those who were required to pay a copayment (detailed results not shown). The number of hospital days in the year prior to MI was a strong negative predictor of medication use for all classes. This might reflect greater comorbidity or frailty in these patients, which both have been associated with lower use of and persistence with preventive medications. Similarly, patients receiving antipsychotic drugs had a lower likelihood to receive the study medications, likely indicating treatment bias regarding the mentally diseased. Beta-blocker use was significantly lower among patients who received any drugs for inhalation that are indicated in asthma or COPD, a plausible pattern, which may reflect presence of a relative contraindication or intolerance by the patient. Indeed, among patients without previous use of such asthma or COPD drugs (N = 3,495), BB use after MI was 75.4% (rather than 74.0% in the overall population), an only slightly higher proportion with respect to the goal of appropriately treating all patients free from contraindications. Statins were less likely to be used in patients who had previously used diuretics, digitalis, or coumadin, possibly indicating congestive heart failure or atrial fibrillation and thus, worse prognosis. In theory, and barring presence of any absolute contraindications, 100% of patients could receive each of these medications. Clearly, the actual proportion that could be attained in practice is lower than that, since some patients will have an absolute contraindication for any given drug. For ACEI/ARBs, this number would be very low, since the only absolute contraindication (other than pregnancy) is history of angioedema, whose incidence is below 1% among new ACEI users [37]. Beyond this, relative contraindications may exist such as hyperkalemia or advanced chronic kidney disease, but in most patients, at least a low-dose trial of an ACEI/ARB should be attempted with appropriate clinical and laboratory monitoring in place, possibly in combination with a loop diuretic. Similarly, for statins, practically every one should receive this treatment and absolute contraindications are either very rare (active liver disease) or not to be expected in this MI population (pregnancy, lactation). Thus, at least an attempt at using statins and ACEI/ARB in post-MI patients can be expected in probably >95% of patients. The situation is slightly different with BBs, since more contraindications exist whose classification as absolute vs. relative are uncertain [38]. Cardiogenic shock, hypotension, and certain bradycardic arrhythmias untreated with a pacemaker certainly constitute absolute contraindications, whereas a treatment attempt in patients with stable obstructive lung disease should be conducted. From our dataset, it is difficult to ascertain most of these conditions, but when eliminating all patients who had received inhalative corticosteroids or beta mimetics, the percentage of BB users increased slightly to 75.4% from 74.0% in the overall population. This report needs to be read with several limitations in mind. Aspirin was available at a price below the copayment for prescription medications. Thus, aspirin use may be underascertained in non-indigent patients. We were, however, able to assess aspirin use in those patients who had their copayment waived and found that 71% of patients received this drug. While we cannot be certain that medications received in the pharmacy were actually taken by these patients, our way of ascertaining medication use is superior to other methods. Compared to notes in medical charts, patients may not fill the prescriptions they were given. In surveys, patients may report what they perceive as being desirable rather than their actual behavior. Both methods, may lead to inaccurately optimistic estimates of medication use. In our study, comorbidities were not ascertained from diagnosis codes, but rather from typical medications that are given for several comorbidities. It has been shown that the information from diagnosis codes and from medication claims provides only marginally inferior confounding control in administrative datasets [39]. On a similar note, we were unable to ascertain reliable information on contraindications for the study drugs. Since most of these contraindications are relative rather than absolute contraindications [37, 38] this aspect may be of minor importance. While not fully population based, our data differ from other studies in that a wide range of hospitals and typical care settings were studied rather than single departments in predominantly academic medical centers. We consider our data generalizable to the population level, since three quarters of the Austrian population were included in the study, and the preponderance of patients omitted (1.2 Mio.; data from 2004) was excluded because they lived in the three Western provinces (Salzburg, Tyrol, Vorarlberg), whose sickness funds did not participate in our data collection effort. The remaining patients that we were unable to capture were members of very small corporate sickness funds, railroad and mining workers, as well as self-employed individuals; all those, however, received care at the same hospitals and by the same physicians as the enrolled patients. Finally, the analytical dataset did not contain linkable pseudoanonymized hospital identifiers. Thus, we were unable to study variations in treatment behavior across the provider level. It has been shown that provider preference can be a stronger determinant of treatment received than actual patient level indications [40]. In summary, we provide evidence for underuse of several recommended medications after MI as recently as 2004 in the Austrian healthcare system. Our observations are in line with findings from other European and North American healthcare systems, despite the differences in data collection and time period studied. Educational efforts need to be directed at both physicians and patients, and the implementation of quality indicators should be considered. Maximizing secondary prevention after MI is highly desirable from an individual patient and the societal perspective.

Eur_J_Nucl_Med_Mol_Imaging-3-1-1998878

Requirements regarding dose rate and exposure time for killing of tumour cells in beta particle radionuclide therapy

Purpose The purpose of this study was to identify combinations of dose rate and exposure time that have the potential to provide curative treatment with targeted radionuclide therapy applying low dose rate beta irradiation. Introduction Many types of tumour overexpress cell surface-associated antigens or receptors suitable as targets for radionuclide therapy, and many types of targeting agent have been suggested or are already being applied for such therapy. This therapy is currently employed for lymphomas [1, 2] using radiolabelled antibodies and also for neuroendocrine [3–5] and paediatric tumours [6, 7] using radiolabelled somatostatin analogues and meta-iodobenzylguanidine (mIBG), respectively. In the majority of these cases, beta emitters such as 90Y, 131I and 177Lu have been applied. The results have, so far, essentially shown palliative effects [1, 8–12], and there is hope that combinations of beta particle emitters, e.g. 90Y and 177Lu, will improve the therapy results [13]. The cell-killing capacity of low LET radiation, i.e. photons and electrons, is well known when applying high dose rates, typically 0.5–2.0 Gy/min, as in external radiotherapy [14, 15]. However, the extensive experimental and clinical knowledge on the effects of external radiotherapy can be deployed to only a limited extent in understanding the effects of radionuclide therapy. A major difference is that the dose rate in radionuclide therapy is at least two orders of magnitude lower than in external radiotherapy [10, 16–19]. The lower dose rate allows for DNA repair and repopulation during the radiation exposure, which is not the case during high dose rate exposures. Basic radiobiological studies have shown that low dose rates, in the range of 0.1–1.0 Gy/h, give a much lower biological effect (per dose unit) than high dose rates in the range 0.5–2.0 Gy/min [15, 17, 20, 21]. It is also known that an inverse dose rate effect exists in that dose rates of 0.2–0.4 Gy/h can give more cell kill than dose rates in the range 0.7–1.0 Gy/h [15, 22]. Only crude estimates can be made from previous experiments to elucidate which combinations of low dose rate and exposure time can cure a metastasis containing, for example, 105 cells. Cell survival has most often been analysed after a cell cloning 1–2 weeks after the radiation exposure. For example, a total dose of about 30–50 Gy, given with 0.1–1.0 Gy/h, seems necessary to decrease the single cell survival probability to 10−5 [23, 24], and thereby give a reasonable chance of killing 105 tumour cells. Furthermore, targeted radionuclide therapy is complicated, since it is not enough only to consider the macroscopic dose concept; different cellular and intracellular distributions of radionuclides can give different biological effects although the macroscopic dose is the same [25, 26]. One way to obtain solid information on which combinations of low dose rate and exposure time can give curative treatments with beta particles (which also is low-LET radiation) is, of course, through experiments and clinical trials. In this study we used an experimental model with the criterion that the low dose rate beta radiation must kill all 105 tumour cells in a culture dish in order to simulate a successful treatment. The follow-up period was 3 months. The choice of 105 tumour cells is somewhat arbitrary and is based on two arguments. The first is that this number represents a small tumour cell cluster that normally cannot be identified by routine diagnostic procedures such as computed tomography or magnetic resonance imaging (unless the tumour cells cause macroscopic changes in the surrounding normal tissues). Furthermore, this number of tumour cells in most cases does not cause symptoms in the patient. Thus, a cluster of 105 tumour cells in a patient can be considered an “occult” or “subclinical” tumour or metastasis. The second argument is more practical, since the presence of 105 tumour cells in a normal cell culture dish or flask allows enough space for exponential growth and, at the same time, frequent cell–cell contacts. Our ambition was not to simulate the dose rate variations in time and space that occur in radionuclide therapy. In the clinical setting, the dose rate varies with time, not only as a consequence of the physical half-life of the radionuclides, but also due to time-dependent changes in their spatial distribution [16, 17, 24, 26, 27]. Factors of importance are ongoing vascularisation processes, variations in vessel wall leakage and changes in blood flow. There are probably also various diffusion and convection conditions in different areas of tumours, resulting in variable penetration properties of the radiolabelled targeting agents. In addition, there might be variations in the expression of target structures on the tumour cells. All these time-dependent factors make it difficult to establish basic and reproducible dose rate–response relations in vivo. Our experimental model was designed to give reproducible and controllable irradiation conditions, and we applied a model with a rather long physical half-life (32P sources with T1/2=14.3 days), giving only a slow decrease in dose rate during the exposures. Relevant dose rates were selected through the amount of radionuclide placed in the irradiation chambers. The exposure times were selected to correspond to the effective half-lives of the radionuclides delivered by targeting agents of different types. In targeted radionuclide therapy it is, of course, also necessary to consider unwanted side-effects on normal tissues. However, analyses of normal tissue effects were beyond the scope of this study. Hyperradiosensitivity [28, 29] at low doses, bystander effects [30–32] and low dose rate-induced apoptosis [33, 34] are all extensively studied processes. Our model allows these processes to work together, but we did not try to study them separately. The overall goal of the study was to find “dose rate–exposure time” relations that could kill all of the exposed 105 tumour cells, so that no remaining cells would be observed after 3 months. Materials and methods Irradiation chambers The irradiation chambers have been described previously [35], so only a short description is given here. Three identical chambers were used and each was filled with 400 ml distilled water containing 0.74–2.22 GBq 32P. The water was boiled and degassed shortly before adding 32P and filling of the chambers, in order to avoid air bubbles. The beta emitter 32P (orthophosphate) (T1/2=14.3 days) was obtained from Amersham Pharmacia Biotech (Amersham, UK). The upper area of the chambers, where the cells were exposed to beta particles from 32P, was covered with a 0.5-mm thin transparent polycarbonate foil. The radiation protection walls surrounding the side and the bottom of the chambers, as well as the lid above the cell cultures, consisted of 15-mm transparent polycarbonate (Macrolon). 32P could not reach and be incorporated in the cells. The irradiated cells were grown in 3-cm-diameter culture dishes with a plastic bottom thickness of 1 mm. The culture dishes were placed directly on the thin foils above the 32P source. The chambers were kept in cell culture incubators at 37°C (Kebo Assab T304GF, Stockholm, Sweden) and supplied with 5% carbon dioxide. The lid above the cultures had side openings, allowing efficient passage of the incubator atmosphere. Dosimetry The dosimetry has also been described previously [35]. The dose rate for each chamber was controlled by measurements with a thin-walled parallel ion chamber [36]. The ion chamber was calibrated using an external 60Co source with the front wall of the ion chamber at the dose maximum depth, where the dose rate was 0.30 Gy/min. The 60Co source was calibrated according to national standard procedures accepted for radiotherapy purposes. The thin-walled parallel ion chamber accurately measured the dose independent of dose rate down to at least 0.01 Gy/h, if corrections were made for “leakage current” in the instrumentation. The measurements of dose rates were performed with the thin-walled ion chamber placed in cell culture dishes standing on the 32P radiation chambers. This procedure allowed mimicking of the dose rate in the cell environment. All three chambers were measured repeatedly. Measurements were also made with a surface hand detector (RNI 10/R Intensimeter, Nuklex, Uppsala, Sweden), and a calibration curve was constructed to facilitate repeated dose rate determinations during the cell culture periods. Tumour cells The cells used were HT-29 colorectal adenocarcinoma, A-431 cervical squamous carcinoma, SKBR-3 breast cancer, all from the American Type Culture Collection (ATCC), and the two gliomas U-118MG and U-373MG from the Department of Pathology, Uppsala University, Uppsala, Sweden. They were grown in Ham’s F-10 medium supplemented with 10% fetal bovine serum, 2 mmol/l L-glutamine, 100 μg/ml streptomycin and 100 U/ml penicillin, all components from Sigma AB (Stockholm, Sweden). The cells were normally grown in an incubator of type Galaxy S (LabRum Klimat AB, Stockholm, Sweden), and during the irradiations they were grown in a similar incubator (Kebo Assab T304GF, Stockholm, Sweden). Both incubators were run at 37°C and supplied with 5% carbon dioxide. The cells in the study were selected for the following reasons. SKBR-3 cells are often studied since they express large amounts of HER2 receptors and are therefore applied in experiments on HER2-directed radionuclide targeting, using antibodies or affibody molecules [37, 38]. A-431 cells express both large amounts of EGFR and also rather large amounts of HER2 and are therefore often studied for radionuclide targeting with both EGF ligands [39] and anti-HER2 antibodies [40]. Both cell types are planned for experimental therapy using beta particle-mediated radionuclide therapy, especially 177Lu. The two glioma cell lines also express EGFR [41] to some degree but were, together with the HT-29 cells, selected because they have recently been studied at our laboratory with regard to short-term effects after low dose rate irradiations [35]. It was found that U-373MG cells showed radiation-induced apoptosis, while U-118MG cells did not. The HT-29 cells were intermediate in this respect [35]. Furthermore, U-373MG cells have previously been reported not to show hyperradiosensitivity at low doses, while both U-118MG and HT-29 cells do [28]. Choice of dose rate and exposure time In the previous study, with the same 32P chambers, we analysed effects on cells after exposure to initial dose rates of only 0.05–0.09 Gy/h for 7 days [35]. The cells in that study were analysed for effects on cell number, apoptosis and cell cycle block at day 7 of continuous exposure. However, we also allowed cells to continue to grow for longer times in parallel dishes and found that all cultures recovered. Thus, we decided to apply initial dose rates from 0.1 Gy/h up to 0.8 Gy/h in the present study to have a reasonable chance of “simulating” curative treatment. We could not apply higher dose rates because of regulatory rules limiting the amount of 32P that could be handled in the cell culture laboratory. Furthermore, this dose rate range is the same as that applied in published studies on low dose rate effects [21–23], hyperradiosensitivity and low dose rate [29], and radionuclide treatment of gliomas in vitro [42] and in vivo [43], and actually covers the range up to the highest dose achievable in targeted radionuclide tumour therapy [10, 16, 18, 27, 44]. Four cell dishes were placed in each irradiation chamber when the dose rate was as expected and the cells were then kept there for 24, 72 or 168 h (1, 3 or 7 days). The radiation exposure time was chosen not to be longer than a week, since it is known from numerous articles in the field of radionuclide therapy that the tumour cell retention of radioactivity is generally in the range of some days up to about 1 week (in some cases it is only a few hours). Furthermore, it is well known that the biological half-life of tumour targeting agents (ligands, antibodies and antibody fragments) in the systemic circulation is often shorter than a week (in the case of small ligands, it is only a few hours). Thus, the longest exposure time considered to be realistic was 1 week. Cell counting The medium was removed from the cell dishes and the cells were quickly washed with 0.5 ml trypsin-EDTA (0.25% trypsin/0.02% EDTA solution in PBS, VWR, Stockholm, Sweden) and then incubated with 0.5 ml trypsin-EDTA (37°C, 5% CO2) until the cells detached. Next, 1.5 ml medium was added to each dish and the cells resuspended to a single cell solution. For cell counting, 19.5 ml PBS (pH 7.4) was added to 0.5 ml cell suspension and an electronic cell counter was used (Coulter Z2, 7–20 μm, Beckman Coulter, Stockholm, Sweden). Cell culture conditions during and after irradiations Cells were seeded sparsely in culture dishes (diameter 3.5 cm, surface 9.6 cm2, Nunc, Roskilde, Denmark) a few days before the start of the radiation exposure. They were seeded so that each culture dish contained about 105 cells at the start of the exposure. The first cell count was made at the start of each experiment, on representative culture dishes. If the cell number was about 105 per dish, four other, parallel culture dishes were placed in each of the three irradiation chambers. The irradiated cells were grown in the chambers with unirradiated control dishes placed on a near radiation-shielded shelf in the same incubator. The cells did not reach confluence during the irradiation period. After the radiation exposure, the cells were moved from the irradiation chamber incubator to another incubator and the cell growth was followed for several months. At the first subcultivation after the radiation exposure, the cells were transferred from culture dishes to culture flasks (25 cm2, Nunc, Roskilde, Denmark), and these flasks were then used throughout the growth period. The medium in all culture dishes and flasks was replaced three times a week. Cell counting was in most cases performed once a week, followed by reseeding of 105 cells in each new flask. Growth curves The growth curves were constructed as if all cells had been saved at each subcultivation. By calculating how many cells would have been obtained if all cells had been saved, such high cell numbers as 1015 (Fig. 1a) and even up to 1025 (Fig. 1b) were obtained. In reality, repeated dilutions were made to keep the number of cells in each culture flask in the range of 105–106, thus allowing for exponential growth. Four parallel flasks were kept for each experiment. In order to analyse the significance of the differences between the groups, t tests were performed. Cell kill versus regrowth The criterion for killing of a cell culture with 105 cells was that no living cells could be seen in the phase contrast microscope and that no regrowth was observed after at least 3 months of follow-up, also applying electronic cell counting. Growth delay Growth delay was also analysed after 1, 3 or 7 days of continuous low dose rate exposure. The growth curves of the irradiated samples in some cases did not have the same slope as the control curves, so it was not possible to wait and measure growth delay when the recovered cells and the control cells had a similar growth rate. Instead, growth delay was defined as the time it took for the irradiated cells to reach the cell number 1010 in relation to the time it took for the control cultures to reach this number. Literature survey The Medline-based PubMed database was used to survey effects of targeted radionuclide therapy and of low dose rate therapy. Results Growth curves Figure 1 shows examples of growth curves for control cells and cells exposed to continuous low dose rate beta irradiation. In Fig. 1a the initial dose rate was 0.415 Gy/h and the exposure time 72 h. After 72 h the dose rate had decreased to 0.359 Gy/h, and the exposure gave a total dose of about 27.8 Gy. The irradiated HT-29 cells recovered after irradiation and resumed, after about 50 days, a similar growth rate as the controls. In contrast, the U-373MG cells completely ceased to grow after the same radiation exposure. After more than 30 days, only a few giant cells could be identified in the microscope. They were too few to be counted with the electronic cell counter. However, there was a possibility that at least a few of these cells could grow later, so they were observed (with normal medium changes) for up to 3 months. Nevertheless, no recovery could be observed; instead, the cells disappeared. It was then concluded that the irradiation procedure had killed all U-373MG cells. Thus, the HT-29 cultures survived the treatment, while the U-373MG cells died. Fig. 1Examples of cell growth as a function of time after exposure to low dose rate irradiation. a HT-29 and U-373MG cells were exposed to 0.415 Gy/h as the initial dose rate and then continuously exposed for 72 h. b A-431, U-118MG and SKBR-3 cells were exposed to 0.806 Gy/h as the initial dose rate and then continuously exposed for 24 h. Mean values and maximal variations from four parallel samples are given Figure 1b shows examples of growth curves for an initial dose rate of 0.806 Gy/h and an exposure time of 24 h. After 24 h the dose rate had decreased to 0.768 Gy/h and the exposure gave a total dose of about 18.9 Gy. The irradiated A431 cells continued to grow, but at a slower rate than the controls. The growth of irradiated U-118MG cells was arrested for up to about 40 days after the treatment. During those 40 days, resting cells were observed in the culture flasks, but there was no sign of growth. After 40 days a few mitotic cells were observed in the phase contrast microscope, and the number of cells was thereafter measured with the electronic cell counter. A clear regrowth was seen after 60 days. However, their growth rate was slower than that of the corresponding controls. The SKBR-3 cells looked severely damaged (fragmented, decreasing in number and forming giant cells) some days after the exposure and were too few to be counted with the electronic cell counter. Within 1 month, no surviving cells could be observed in the culture flasks. In this case, too, we continued to control the culture flasks for 3 months (with normal medium changes) and found that there were no remaining cells in these cultures. Thus, A-431 and U-118MG cells survived, while the SKBR-3 cells died. The conclusions from the examples in Fig. 1a and b are that four different responses were observed: The HT-29 cells recovered to the control growth rate after a growth delay.The U-118MG cells recovered after a growth delay but continued to grow at a slower rate than the controls.The A-431 cells continued to grow without delay but at a slower rate than the controlThe U-373MG and SKBR-3 cells died. Cell kill and regrowth versus dose rate and exposure time Figure 2 summarises cell-killing and regrowth results from all performed low dose rate experiments. All five analysed cell types were irradiated with different initial low dose rates (0.1–0.8 Gy/h) and were continuously exposed for 1, 3 or 7 days. Fig. 2Summary of all low dose rate experiments carried out for a U-118MG, b U-373MG, c HT-29, d A-431 cells and e SKBR-3 cells. The cells were irradiated with different initial dose rates and were then exposed to the radiation for 1, 3 or 7 days. The figures show at which combinations of dose rate and exposure time all cells were killed (area with no survivors), and at which at least some cells survived and displayed regrowth (the regrowth area). The separation between the two areas is indicated by bold solid lines. The total delivered radiation dose (Gy) is given in parentheses near each point. The 20-Gy isodose curve is indicated by a dashed line Figure 2 parts a–e show a similar and general pattern. When applying 7 days of continuous irradiation, low initial dose rates, about 0.2–0.3 Gy/h, were enough to kill all cells. When cells were exposed for only 3 days, a dose rate in the order of 0.4–0.6 Gy/h was necessary to kill all cells. Only the SKBR-3 cells were killed after 24-h exposure to about 0.8 Gy/h. As mentioned in Materials and methods, higher dose rates could not be used in these experiments, so we do not know how high the dose rates would have to be during 24-h exposure in order to kill the other cell types. It was found that the regrowth pattern was the same in all four flasks. However, even if the results looked rather similar and independent of cell type, there were some cell type-dependent differences. The U-118MG cells (Fig. 2a) required, during 7 days’ exposure, at least about 0.3 Gy/h as the initial dose rate to be completely growth inactivated. The U-373MG cells (Fig. 2b) were completely growth inactivated after an initial dose rate of only about 0.15 Gy/h and 7 days’ exposure. Thus, the U-373MG cells were, in these cases, easier to kill than the U-118MG cells. Considering all cells, it seemed that the U-118MG cells were most resistant and U-373MG and SKBR-3 most sensitive, while A-431 and HT-29 were intermediate. Cell kill and regrowth versus total dose The 20-Gy isodose curve is drawn with a dashed line in Fig. 2a–e. For U-373MG and SKBR-3 cells, at least a total dose of 20 Gy was necessary to prevent regrowth. For HT-29 and A-431 cells an even higher total dose seemed necessary, and for the resistant U-118MG cells it seemed that total doses of at least 40 Gy were necessary to prevent regrowth. Growth delay as a function of dose rate Figure 3a–e shows growth delays versus the initial dose rate after 1, 3 or 7 days of continuous low dose rate exposure. The growth delay curves in Fig. 3a show that the radioresistant U-118MG cells had growth delays, after 24-h exposure, which increased slowly as a function of the initial dose rate. A growth delay of 2 months was obtained at about 0.7, 0.3 and 0.2 Gy/h after 1, 3 and 7 days of continuous exposure, respectively. The growth delay curves for the more radiosensitive U-373MG cells increased more steeply (Fig. 3b), and a growth delay of 2 months was obtained at about 0.5 and 0.2 Gy/h after 1 and 3 days of continuous exposure, respectively. Thus, the U-373MG cells were more sensitive than the U-118MG cells also with regard to growth delay. Fig. 3Growth delay versus the initial dose rate after 1, 3 or 7 days of continuous low dose rate exposure for a U-118MG, b U-373MG, c HT-29, d A-431 and e SKBR-3 cells. The boxes at the top of each figure indicate at which dose rates and exposure times there was no regrowth (infinite growth delay or total kill of the cultures). Mean values and maximal variations from four samples are shown Both the HT-29 cells (Fig. 3c) and the SKBR-3 cells (Fig. 3e) had a biphasic growth delay curve after 1 day’s exposure. The A-431 cells (Fig. 3d) had a slow increase in growth delay as a function of dose rate after 1 day’s exposure. The SKBR-3 cells had a growth delay that increased rather steeply as a function of the dose rate and reached a value of 2 months at about 0.50 and 0.15 Gy/h after 1 and 7 days of continuous exposure, respectively (Fig. 3e). Growth delay as a function of total dose The same growth delay data as in Fig. 3a–e are instead plotted as a function of the delivered total dose in Fig. 4a–c, but here, for clarity, without the maximal variations. In all cases, there was a general increasing growth delay as a function of dose. Growth delays in the order of 10 days were obtained for all cells after delivery of total doses in the range 5–15 Gy, independent of whether the dose was given during 1, 3 or 7 days. Delays around 100 days required total doses of about 15–20 Gy when given during 1 day and 25–40 Gy when given during 3 or 7 days, for all cells. Fig. 4Growth delay versus total dose after a 1, b 3 or c 7 days of continuous low dose rate exposure for U-118MG, U-373MG, HT-29, A-431 and SKBR-3 cells. The growth delay values are the same as in Fig. 3 but, for clarity, without maximal variations Discussion The obtained results can hopefully serve as a guideline for the combinations of dose rate and exposure time necessary to kill tumour cells when applying low dose rate beta irradiation. The shift from regrowth to “cure” (no surviving cells) fell, for each cell type, within a narrow range of combinations. The U-118MG cells were more resistant to the treatments than the U-373MG and SKBR-3 cells. The A-431 and HT-29 cells showed intermediate resistance. However, the differences between the cell lines were not dramatic, and it can be generally stated that when applying 7 days of continuous irradiation, initial dose rates of about 0.2–0.3 Gy/h were enough to kill all cells in the cultures. When exposed for 3 days, an initial dose rate in the order of 0.4–0.6 Gy/h was needed. When the cells were exposed for only 24 h it was not possible to kill all the cells (with the exception of SKBR-3 cells), even if the initial dose rate was as high as about 0.8 Gy/h. The studied dose rates are the highest that can be achieved in targeted radionuclide therapy [10, 16, 18]. The total doses achieved after 1, 3 or 7 days’ exposure also correspond to the highest achievable doses in targeted radionuclide therapy [16], and most often total doses of no more than 10–20 Gy are obtained in targeting of B-cell lymphomas [19]. However, there are indications from preclinical studies that dramatic “dose amplification” per receptor interaction can be achieved by using effective residualising agents [45]. The obtained dose rates in beta particle-based radionuclide therapy are to a large extent a consequence not only of the amount of radionuclides bound to each tumour cell, but also of the cross-fire effect. This means that radionuclides bound to one cell also irradiate nearby cells owing to the long range of the radiation [44, 46]. This can increase the dose rate tenfold, and makes it reasonable to assume that dose rates in the range used in our model experiments can be achieved in tumours in patients. The dose rate will be lower for a single tumour cell considering only the radionuclides bound to that cell [25]. Beta particles with a long range will permit rather uniform dose distributions and hopefully deliver therapeutically relevant radiation doses also to non-targeted tumour cells. Lower dose rates than we applied will probably not lead to curative treatments when beta particles are applied. From the obtained results it is obvious that, at least for the five types of tumour cell tested, we succeeded in finding “dose rate–exposure time” combinations that could distinguish between “cure” (killing of 105 tumour cells) and “relapse” (recovery of tumour cells). There may be cases in which only a fraction of the tumour cells have to be killed directly by radiation, since the remaining tumour cells may be killed through bystander effects [30–32] or other factors (e.g. limited nutrition supply, natural immune response, adjuvant chemo- or immunotherapy). However, our model can be applied under the assumption that 105 tumour cells have to be killed by radiation, even if other tumour cells are killed by other means. When considering radionuclide therapy it is, of course, also important to consider unwanted effects on normal tissues. The tolerance doses for most normal tissues are not known when exposure is to low dose rate irradiation. Targeted radionuclide therapy, using, for example, radiolabelled antibodies, fragments of antibodies or various receptor ligands, is expected to result in highly tumour-specific uptake of the therapeutic radionuclides. Thus, for a curative intent it is reasonable to establish the necessary tumour dose rates and exposure times. Another obvious question is which targeting agent should be tried for each type of tumour and, most importantly, whether the required conditions can be achieved without excessively severe side-effects on normal tissues. However, analyses of effects on normal tissues were beyond the scope of this study. Here we discuss the observed cell type-dependent differences. In a previous study [35] we published data on low dose rate acute effects, using three of the cells in the present study: U-118MG, U-373MG and HT-29. In that study the initial dose rate was only 0.05–0.09 Gy/h and the exposure time, 7 days. As would be expected in view of our new data, all cultures did grow after such treatment. It was shown that the U-373MG cells had, at day 7, the most pronounced reduction in cell number owing to a combination of a G2 block and radiation-induced apoptosis. There were surprisingly low reductions in U-118MG and HT-29 cell numbers. U-118MG had a G2 block but no radiation-induced apoptosis. HT-29 had both a G2 block and some radiation-induced apoptosis, but the amount of apoptosis was smaller than for U-373MG cells. Thus, the results from that study indicated the U-373MG cells to be more sensitive than the other two cell lines owing to a higher degree of apoptosis. This is in agreement with the cell-killing results from the present study. The intrinsic radiosensitivity measured as S2Gy, after exposure to a high dose rate (most often 0.5–2.0 Gy/min) photons has previously been determined for four of our studied cell lines. No determination of S2Gy for SKBR-3 cells was found in the literature. The results are given in Table 1 [28, 47–54]. Table 1Survival at 2Gy after exposure to high dose ratesCell lineSurvival at 2GyReferencesU-118MG0.44–0.70[28], [47]U-373MG0.60–0.62[28], [48], [49]HT-290.55–0.78[28], [50], [51], [52], [53], [54]A-431≈0.52[53] There is also a review article on the intrinsic radiosensitivity, measured as cell survival at the dose 2Gy, S2Gy, for 694 human cell lines, of which 271 were from tumours [55]. The tumour cell lines were grouped according to tumour type, and the S2Gy values for the U-118MG and U-373MG gliomas and the A-431 cervical carcinoma used in this study fell within the expected range of values for the corresponding types of tumour. Thus, these cells can be considered typical for their tumour groups, at least regarding intrinsic radiosensitivity at high dose rate exposure. The HT-29 colorectal carcinoma is an exception, since it seems to be somewhat more radioresistant than most other colorectal carcinomas. Considering the values in Table 1, it is clear that there is no relation between S2Gy and the effects of low dose rate irradiations. For example, the U-118MG cells were found to be most radioresistant to a low dose rate, while their S2Gy values were in the same range as for the other cell lines. The U-373MG cells, which were considered most sensitive to a low dose rate, also had S2Gy values in the same range. One possible explanation of the lack of agreement between intrinsic radiosensitivity measured as S2Gy and the effect of a low dose rate is that there are cell type-dependent differences in repopulation during low dose rate irradiations. Such differences can probably “overshadow” the differences in intrinsic radiosensitivity. Another possible explanation might be cell type-dependent differences in the capacity for low dose rate-induced apoptosis. The latter is supported by our previous study [35], which showed that low dose rate-induced apoptosis was more frequent in U-373MG cells than in HT-29 cells and that no such apoptosis could be observed for U-118MG cells. Furthermore, it seems as if differences in hyperradiosensitivity (measured at low doses but at a high dose rate) are not of great importance, since the U-118MG and HT-29 cells have been reported to show hyperradiosensitivity, while U-373MG cells do not [28]. Had hyperradiosensitivity been of importance under the exposure conditions in this study, then U-373MG cells should have been more resistant than HT-29 and U-118MG cells. It has recently been suggested that variations in radiosensitivity at low dose rates are related to the compactness of chromatin [56], but it is not known whether the cells in our study have any differences in this respect. In another recent experimental study, a good therapeutic effect at a low dose rate was reported; in fact, if the total delivered dose was in the range 1–2 Gy, the effect was as good as that achieved at a high dose rate, although the difference in dose rate was nearly three orders of magnitude [57]. This indicates that there are basic radiation biology aspects of low dose rate radiation that have to be analysed in much more detail. A clue to the molecular factors involved came from a recent report showing that activation or inhibition of the DNA damage sensor ATM is of importance [58]. It was found that DNA damage inflicted at a low rate failed to activate ATM; however, if ATM was activated by chloroquine, the cells survived the low dose rate much better. We conclude that dose rates in the range 0.2–0.3 Gy/h are necessary in order to kill 105 tumour cells during 1 week’s exposure. Higher dose rates, such as 0.4–0.6 Gy/h and >0.8 Gy/h, are necessary if the exposure times are only 3 days and 1 day, respectively. If, in some cases, a good therapeutic effect cannot be obtained with beta particle radionuclide therapy because of too low dose rates and/or too short exposure times, then repeated (fractionated) treatment is a possibility to improve the result. However, the cellular response to repeated low dose rate exposures has to be analysed in further studies, since there is a risk of radiation-induced changes in the growth pattern of the tumour cells surviving the first exposure, as shown in Fig. 1b.

Purinergic_Signal-3-4-2072914

Autoregulation in PC12 cells via P2Y receptors: Evidence for non-exocytotic nucleotide release from neuroendocrine cells

Nucleotides are released not only from neurons, but also from various other types of cells including fibroblasts, epithelial, endothelial and glial cells. While ATP release from non-neural cells is frequently Ca2+ independent and mostly non-vesicular, neuronal ATP release is generally believed to occur via exocytosis. To evaluate whether nucleotide release from neuroendocrine cells might involve a non-vesicular component, the autocrine/paracrine activation of P2Y12 receptors was used as a biosensor for nucleotide release from PC12 cells. Expression of a plasmid coding for the botulinum toxin C1 light chain led to a decrease in syntaxin 1 detected in immunoblots of PC12 membranes. In parallel, spontaneous as well as depolarization-evoked release of previously incorporated [3H]noradrenaline from transfected cells was significantly reduced in comparison with the release from untransfected cells, thus indicating that exocytosis was impaired. In PC12 cells expressing the botulinum toxin C1 light chain, ADP reduced cyclic AMP synthesis to the same extent as in non-transfected cells. Likewise, the enhancement of cyclic AMP synthesis either due to the blockade of P2Y12 receptors or due to the degradation of extracellular neucleotides by apyrase was not different between non-transfected and botulinum toxin C1 light chain expressing cells. However, the inhibition of cyclic AMP synthesis caused by depolarization-evoked release of endogenous nucleotides was either abolished or greatly reduced in cells expressing the botulinum toxin C1 light chain. Together, these results show that spontaneous nucleotide release from neuroendocrine cells may occur independently of vesicle exocytosis, whereas depolarization-evoked nucleotide release relies predominantly on exocytotic mechanisms. Introduction Adenine and uridine nucleotides are present in and released from all types of cells [2] and exert effects via P2X and/or P2Y receptors: P2X receptors are ATP-gated cation channels [25]; P2Y receptors are G protein-coupled receptors, and eight different types have been identified (P2Y1, 2, 4, 6, 11, 12, 13, 14; [1]). P2Y1, 11, 12 and 13 are activated by adenine nucleotides, whereas P2Y6 is activated by uridine nucleotides. P2Y2 and 4 are sensitive to adenine and uridine nucleotides, and P2Y14 is activated by UDP-glucose. In a large number of non-neural tissues, e.g. liver, kidney, bones and blood vessels, released nucleotides subserve autocrine/paracrine functions by activating certain P2Y receptors [32, 16, 38, 12, 30]. In a given cell, this autocrine/paracrine regulation depends on the release of nucleotides, the presence of P2Y receptors and occasionally on the enzymatic activities of nucleotidases: in hepatoma cells [16], for instance, released ATP is degraded to ADP to stimulate P2Y1 receptors; in epithelial [32] and osteoblastic [38] cells, in contrast, ATP itself activates P2Y11 and P2Y2 receptors, respectively. Although ATP has first been described to be released from neurons [15] and to act as a neurotransmitter [6], available information about neuronal autoregulation via released nucleotides and P2Y receptors is limited. Early examples were the feedback inhibition of sympathetic transmitter release via presynaptic P2Y receptors (see [4]) and the ATP-mediated autocrine inhibition of voltage-activated Ca2+ channels in bovine chromaffin cells [9]. More recently, autoregulation via P2Y receptors has been described for the phaeochromocytoma cell line PC12 [24]. There, endogenous P2Y12 receptors were activated by spontaneous nucleotide release to achieve half-maximal inhibition of cyclic AMP (cAMP) synthesis. Depolarization of PC12 cells by 100 mM K+ further enhanced nucleotide release and thereby led to full inhibition of cAMP generation. This latter effect was prevented when voltage-activated Ca2+ channels were blocked by Cd2+ or when extracellular nucleotides were degraded by apyrase. Thus, the inhibition of cAMP synthesis in PC12 cells via P2Y receptors can be viewed as a biosensor for nucleotide release [24]. Nucleotides are released from various cells including fibroblast-like, epithelial, endothelial, glial and neuronal cells [36, 28, 21, 26, 35]. ATP release from non-neural cells can be elicited by different stimuli, such as hypotonic solutions [36], mechanical stimulation [28] or exchange of culture media [21, 26] and is frequently Ca2+ independent and therefore most probably non-vesicular. The mechanisms underlying this non-vesicular ATP release remained largely elusive, but ATP binding cassette transporters, stretch-activated cation channels as well as connexin hemichannels have been implicated [22]. In addition, evidence has been presented that indicates that nucleotide release from non-neural cells may to some extent be vesicular (e.g. [17, 8]). In neurons and neuroendocrine cells, ATP is stored in vesicles either alone or together with classic neurotransmitters and is thus released by exocytosis [27]. Vesicle exocytosis occurs spontaneously at a slow rate and is largely accelerated when Ca2+ entry is triggered by depolarization-induced opening of voltage-activated Ca2+ channels [29]. Accordingly, depolarization-evoked ATP release from neurons and neuroendocrine cells is entirely Ca2+ dependent [35, 13], but it remained unknown whether all of the spontaneous nucleotide release from neuronal sources is also vesicular. Vesicle exocytosis in neurons and neuroendocrine cells involves the SNARE proteins syntaxin, SNAP-25 and VAMP/synaptobrevin. These proteins are cleaved by various serotypes of clostridial neurotoxins which thereby prevent exocytosis. In particular, botulinum neurotoxin C1 cleaves syntaxin, and in some cells also SNAP-25, and thereby impedes spontaneous as well as Ca2+-evoked vesicle exocytosis [5]. Transfection of PC12 cells with a plasmid coding for the botulinum toxin C1 light chain was reported to largely reduce spontaneous as well as depolarization-evoked release of co-transfected human growth hormone [10]. Here, we used an analogous method to investigate whether in PC12 cells the activation of P2Y12 receptors by endogenously released nucleotides relies on vesicle exocytosis. Materials and methods Materials [2,8−3H]adenine (specific activity 32 Ci mmol−1) and levo-[ring-2,5,6-3H]noradrenaline (specific activity 37 Ci mmol−1) were obtained from NEN (Vienna, Austria). Na-ADP, 4-(3-butoxy-4-methoxybenzyl)imidazoline-2-one (RO 20-1724), 3′,5′-cyclic AMP, apyrase (grade VII, with an approximately 1:1 ratio in ATPase and ADPase activity), 2-methylthio-AMP, and 2-p-(2-carboxyethyl)phenethylamino-5′-N-ethylcarboxamido-adenosine (CGS 21680) were purchased from Sigma (Vienna, Austria). Cell culture and transfection PC12 cells were maintained and subcultured as previously described in detail [24]. The cells were plated onto collagen-coated (Biomedical Technologies Inc., Stoughton, MA, USA) culture dishes (NUNC, Roskilde, Denmark) and were kept in OptiMEM (InVitrogen, Vienna, Austria) supplemented with 0.2 mM L-glutamine (HyClone, Aalst, Belgium), 25,000 IU l−1 penicillin and 25 mg l−1 streptomycin (Sigma, Vienna, Austria), 5% fetal calf serum and 10% horse serum (both InVitrogen, Vienna, Austria). Once per week, cell cultures were split; the medium was exchanged twice per week. To investigate the outflow of previously incorporated [3H]noradrenaline under continuous superfusion, PC12 cells were plated onto 5-mm discs, for assays of cAMP accumulation they were plated onto 6-well culture dishes and for membrane preparations for Western blots they were plated onto 100-mm culture dishes. All this tissue culture plastic was coated with collagen (as above). For the generation of PC12 cell clones stably expressing the botulinum toxin C1 light chain (BoNT/C1), 15 μg of a pcDNA3 vector harbouring the botulinum toxin C1 light chain (pBoNT/C1; kindly provided by R.D. Burgoyne, Liverpool, UK) [10] were mixed with 50 μl of the TransFast transfection reagent (Promega, Mannheim, Germany) and added to semiconfluent PC12 cell cultures in serum-free medium. After a 1 h incubation at 37°C, two additional volumes of serum-free medium and the appropriate amount of serum (as above) were added. This medium was exchanged 48 h after transfection for a medium supplemented with 500 μg/ml neomycin (G418) to allow for selection of drug resistance. This selection medium was replaced every 3–4 days until distinct islands of surviving cells were visible. Individual clones of antibiotic-resistant cells were transferred to 24-well plates and grown in medium containing 200 μg/ml G418. Immunoblotting of SNARE proteins Preparation of protein extracts of PC12 membranes was performed as described previously [18]. Materials and methodology used for production and analysis of ECL-immunoblots were also essentially as described therein, with the following modifications: protein extracts in SDS sample buffer were incubated at 95°C for 5 min and subjected to sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) on 5% (stacking) and 10% (separating) gels. After electrophoretic transfer of proteins to nitrocellulose, membranes were cut horizontally along appropriate molecular weight markers in order to separate regions harbouring proteins with apparent molecular weights of 10–20 kD, 21–29 kD and 30–50 kD. These membrane strips were incubated in one of the following monoclonal SNARE antibodies: syntaxin 1 antibody clone 78.3, synaptobrevin 2 antibody clone 69.1 (both obtained from Synaptic Systems, Göttingen, Germany) and SNAP-25 antibody MAB331 (purchased from Chemicon, Temecula, CA, USA). Measurement of [3H]noradrenaline release [3H]noradrenaline uptake in and superfusion of PC12 cell cultures were performed as described previously [23]. Culture discs with PC12 cells were incubated in 0.1 μM [3H]noradrenaline in culture medium supplemented with 1 mM ascorbic acid at 36°C for 1 h. After labelling, culture discs were transferred to small chambers and superfused with a buffer containing (mM) NaCl (120), KCl (3.0), CaCl2 (2.0), MgCl2 (2.0), glucose (20), hydroxyethylpiperazine ethanesulphonic acid (HEPES) (10), fumaric acid (0.5), Na-pyruvate (5.0), ascorbic acid (0.57) and desipramine (0.001), adjusted to pH 7.4 with NaOH. Superfusion was performed at 25°C at a rate of about 1.0 ml/min. Collection of 4-min superfusate fractions was started after a 60-min washout period to remove excess radioactivity. Depolarization-dependent overflow of previously incorporated [3H]noradrenaline was triggered by inclusion of 30 or 100 mM KCl in the superfusion buffer during two 30-s periods 20 min apart (Fig. 2a). At the end of the experiments, the radioactivity remaining in the cells was extracted by immersion of the discs in 2% (v/v) perchloric acid followed by sonication. Radioactivity in extracts and collected fractions was determined by liquid scintillation counting (Packard Tri-Carb 2100 TR). The spontaneous (unstimulated) rate of [3H] outflow was obtained by expressing the radioactivity of a collected fraction as percentage of the total radioactivity in the cultures at the beginning of the corresponding collection period. Values obtained during the 4-min period directly preceding the application of 30 mM K+ were used for the comparison of spontaneous (basal) tritium outflow in wild-type and pBoNT/C1 expressing PC12 cells. K+-evoked tritium overflow was calculated as the difference between the total [3H] outflow during and after stimulation and the estimated basal outflow which was assumed to decline linearly throughout the experiments. Therefore, basal outflow during periods of stimulation was assumed to equate the arithmetic mean of the samples preceding and those following stimulation, respectively. The difference between the total and the estimated basal outflow was expressed as a percentage of the total radioactivity in the cultures at the beginning of the respective stimulation (% of total activity). Determination of cyclic AMP The accumulation of cAMP in PC12 cell cultures was determined as described before [33]. After labelling of cellular purines with tritiated adenine (2.5 μCi/ml for 12 h), the medium was replaced by a buffer (120 mM NaCl, 3 mM KCl, 2 mM MgCl2, 2 mM CaCl2, 20 mM glucose, 10 mM HEPES, 10 mM LiCl, adjusted to pH 7.4 with NaOH) supplemented with 100 μM of the phosphodiesterase inhibitor Ro-20-1724 [4-(3-butoxy-4-methoxybenzyl) imidazolidin-2-one] and 1 U/ml adenosine deaminase. Dishes were then kept at room temperature for 105 min. During the last 15 min of this incubation period, the adenosine A2A receptor agonist 2-p-(2-carboxyethyl)phenethylamino-5′-N-ethylcarboxamido-adenosine (CGS 21680), ADP (10 μM), apyrase (1 U/ml) or 100 mM KCl (NaCl was reduced accordingly) were also included in the medium. When appropriate, the P2Y12 receptor antagonist 2MesAMP (100 μM; [14]) was present for the last 25 min. The incubation was terminated by exchanging the buffer for 1 ml of 2.5% perchloric acid containing 100 μM non-labelled cAMP followed by a 20-min incubation at 4°C. Subsequently, cAMP was separated from the other purines by a chromatographic procedure described before [33]. One tenth of each sample obtained as described above was used for the determination of the total radioactivity. The remaining 900 μl were neutralised by addition of 100 μl 4.2 M KOH and applied to Dowex 50 columns (AG 50W-X4; Bio-Rad, Vienna, Austria), which were then rinsed with 3 ml H2O. The eluate obtained by the subsequent application of 8 ml H2O was directly poured onto alumina columns (Bio-Rad, Vienna, Austria), which were then washed with 6 ml H2O. Finally, cAMP was eluted with 4 ml imidazole buffer (20 mM imidazole in 0.2 M NaCl; pH 7.45). Radioactivity within the samples obtained was determined by liquid scintillation counting. The radioactivity in the fraction of cAMP was expressed as percentage of the total radioactivity incorporated in the cells. Stimulation of PC12 cells, whether they were clonal cell lines or wild-type cells, with the adenosine A2A receptor agonist CGS 21680 caused reproducible increases in the values of cAMP. However, the extent of basal and stimulated cAMP synthesis may vary between different preparations [33]. Hence, to be able to directly compare changes in CGS 21680-induced cAMP synthesis in different cell clones, values obtained in the presence of ADP, apyrase, 2MesAMP or KCl were expressed as percentage of the values obtained in the very same preparation, but in the absence of these agents (% of control). Statistics All data represent arithmetic means±SEM; n represents numbers of culture dishes. Statistical significance of differences between single data points was evaluated by the non-parametric Mann-Whitney test, and p values < 0.05 were accepted as indicators of significance. Results To assess the role of vesicle exocytosis in the release of endogenous nucleotides from neuroendocrine cells, this study employed the autocrine/paracrine control of cAMP synthesis in PC12 cells via P2Y12 receptors. To prevent vesicle exocytosis, we generated cell clones stably expressing pBoNT/C1. Depending on the cell types being investigated, this toxin cleaves one or more SNARE proteins and thereby prevents vesicular transmitter release or hormone secretion [5]. Cleavage of SNARE proteins by the expression of the BoNT/C1 light chain The major target of BoNT/C1 is syntaxin 1, which is cleaved by this toxin in all types of preparations [5]. Therefore, several pBoNT/C1 expressing clones were initially tested in immunoblots of harvested membranes for a lack of full-length syntaxin. Two of these pBoNT/C1 expressing clones (BT2 and BT7) were selected for further analysis. In bovine chromaffin cells, BoNT/C1 was reported to cleave not only syntaxin 1, but also SNAP-25, even though the toxin failed to cleave purified and recombinant SNAP-25 [11]. To evaluate the efficiency of SNARE protein cleavage by the BoNT/C1 light chain expressed in PC12 cells, the immunostaining of syntaxin and SNAP-25 in Western blots was compared with that of synaptobrevin, which is undoubtedly insensitive to BoNT/C1 [11, 5]. As shown in Fig. 1, the stable expression of pBoNT/C1 markedly reduced or even abolished the staining of a protein with an apparent molecular weight of 37 kD by an anti-syntaxin antibody. In contrast, there were no clear-cut reductions in the staining of a 25-kD band by an anti-SNAP-25 antibody, and occasionally the immunostaining of this protein appeared to be even enhanced in pBoNT/C1 expressing clones (Fig. 1a). To quantify these changes in SNARE protein expression, the immunoblots were quantified by densitometry and the values obtained for the anti-syntaxin and anti-SNAP-25 immunoreactive bands were set in relation to those of the anti-synaptobrevin immunoreactive bands (Fig. 1b). This procedure revealed a significant reduction of syntaxin in the pBoNT/C1 expressing clones when compared with wild-type PC12 cells. SNAP-25 immunoreactivity, in contrast, was not significantly altered in these cell clones. Fig. 1Identification of SNARE proteins in membrane extracts from either wild-type (WT, untransfected) PC12 cells or from two PC12 cell clones stably expressing pBoNT/C1 (BT2 and BT7). SDS-membrane extracts were separated on SDS-PAGE and transferred to nitrocellulose membranes. The membranes were cut horizontally to separate regions harbouring proteins with apparent molecular weights of 1–20 kD, 21–29 kD and 30–50 kD and immunostained with anti-syntaxin 1 antibody clone 78.3 (Stx), anti-SNAP-25 antibody MAB331 (S25) and anti-synaptobrevin 2 antibody clone 69.1 (Sb), respectively. a Representative blot obtained as described above. b The relation between the anti-syntaxin and the anti-synaptobrevin immunoreactivities (Stx/Sb) as determined by densitometric analysis (n = 3). c The relation between the anti-SNAP-25 and the anti-synaptobrevin immunoreactivities (Stx/Sb) as determined by densitometric analysis (n = 3). *Significant differences versus the results obtained with membranes from wild-type cells at p < 0.05 Inhibition of transmitter release by the expression of the BoNT/C1 light chain Previously, it has been shown that the transient expression of pBoNT/C1 reduced stimulated vesicular catecholamine release from PC12 cells [10]. To evaluate whether catecholamine release was reduced in the pBoNT/C1 expressing clones, clonal and wild-type PC12 cells were labelled with [3H]noradrenaline and the basal and depolarization-dependent outflow of radioactivity was determined. After washout of excess radioactivity during a 1-h washout period, all PC12 cell cultures steadily released small amounts of radioactivity in the superfusion buffer (Fig. 2a). However, this spontaneous outflow of radioactivity was significantly less in the two pBoNT/C1 expressing clones than in untransfected PC12 cells (Fig. 2b). Fig. 2Basal and depolarization-evoked [3H]noradrenaline release from either wild-type (WT, untransfected) PC12 cells or from two PC12 cell clones stably expressing pBoNT/C1 (BT2 and BT7). PC12 cell cultures were labelled with [3H]noradrenaline and superfused. Subsequent to a 60-min washout period, 4-min fractions of superfusate were collected. Tritium overflow was stimulated twice (S1 after 72 min and S2 after 92 min of superfusion) by the presence of KCl (30 or 100 mM, each for 30 s). a Comparison of the time courses of fractional [3H] outflow from the three different kinds of cell cultures (n = 5 to 6). b The amount of basal [3H] outflow from the three PC12 cell cultures (n = 8 to 9). c The amount of [3H] overflow from the three PC12 cell cultures triggered by 30 mM KCl (n = 8 to 9). d The amount of [3H] overflow from the three PC12 cell cultures triggered by 100 mM KCl (n = 8 to 9). *, **, ***Significant differences versus the results obtained with wild-type cells at p < 0.05, p < 0.01 and p < 0.001, respectively In wild-type cells, the presence of 30 mM K+ (Na+ was reduced accordingly) for 30 s caused a reproducible increase in the outflow of radioactivity (Fig. 2a), but such an effect was hardly observed in the pBoNT/C1 expressing clones. Accordingly, the overflow of radioactivity triggered by 30 mM K+ was significantly reduced in the pBoNT/C1 expressing clones in comparison with wild-type cells (Fig. 2c). This confirms that the expression of the BoNT/C1 light chain reduced spontaneous noradrenaline release and almost abolished release due to mild depolarizations. Large intracellular Ca2+ increases have been reported to rescue transmitter release from cells poisoned with BoNT/C1 [7]. Therefore, PC12 cultures were also exposed to 100 mM K+ (Na+ was again reduced) for 30 s. Although this stimulation paradigm caused unequivocal increases in the outflow of radioactivity even from pBoNT/C1 expressing cells (Fig. 2a), the amount of tritium overflow was still reduced when compared with the values obtained in untransfected PC12 cells (Fig. 2d). Thus, the expression of the BoNT/C1 light chain efficiently reduced spontaneous as well as depolarization-evoked vesicular transmitter release. Changes in the regulation of cAMP synthesis via P2Y receptors due to the expression of the BoNT/C1 light chain In PC12 cells, activation of endogenous P2Y12 receptors by the exposure to nucleotides causes an inhibition of adenylyl cyclase [33]. Furthermore, spontaneously released nucleotides are also sufficient to partially activate these receptors, and depolarization of the cells fully activates the receptors through enhanced nucleotide release to maximally suppress cAMP synthesis [24]. In accordance with these previous data, exposure of wild-type PC12 cells to 10 μM ADP significantly reduced the amount of cAMP generated due to the activation of A2A adenosine receptor by 1 μM CGS 21680 present for 15 min. In the two pBoNT/C1 expressing clones, the inhibition of cAMP accumulation was as pronounced as in the wild-type cells (Fig. 3a). Thus, the adenylyl cyclase inhibiting P2Y receptors were operating equally well in transfected as well as untransfected PC12 cells. Fig. 3Changes in cAMP synthesis caused by exogenous ADP or endogenously released nucleotides in either wild-type (WT, untransfected) PC12 cells or in two PC12 cell clones stably expressing pBoNT/C1 (BT2 and BT7). After loading with [3H]adenine, PC12 cells were incubated in RO 20–1724 (100 μM) for 105 min. During the last 15 min of this incubation period, 1 μM CGS 21680 was present either alone or together with 10 μM ADP (a), 1 U/ml apyrase (b), 100 μM MeSAMP (c) or 100 mM KCl (d). The amount of radioactivity retrieved within the fraction of cyclic AMP (cAMP) was calculated as percentage of the total radioactivity extracted from the cell cultures. In order to compare the results of different cell cultures, values obtained in the presence of ADP, apyrase, 2MesAMP or KCl were expressed as percentage of the values obtained in the very same preparation, but in the absence of these agents (% of control). a Results obtained in the presence of 10 μM ADP are expressed as percentage of the results obtained in its absence (% of control; n = 6 to 10). b Results obtained in the presence of 1 U/ml apyrase are expressed as percentage of the results obtained in its absence (% of control; n = 9 to 10). c Results obtained in the presence of 100 μM MeSAMP are expressed as percentage of the results obtained in its absence (n = 6 to 9). d Results obtained in the presence of 100 mM KCl are expressed as percentage of the results obtained in its absence (n = 9; ***p < 0.001 versus the results obtained in wild-type cells; #, ###p < 0.05 and p < 0.001 versus the results obtained in the absence of KCl) In the presence of apyrase (1 U/ml) to degrade extracellular nucleotides, the cAMP accumulation triggered by CGS 21680 was almost doubled. Again, this effect was not different between wild-type and pBoNT/C1 expressing cells (Fig. 3b). In accordance with these latter results, the blockade of the receptors by the P2Y12 receptor antagonist 2-MeSAMP (100 μM; [14]) also enhanced the stimulation of cAMP about twofold and there were no significant differences between the three types of PC12 cultures (Fig. 3b). Thus, the activation of the P2Y receptors negatively linked to adenylyl cyclase by spontaneously released nucleotides was the same whether pBoNT/C1 was expressed or not. When wild-type PC12 cells were depolarized by 100 mM K+, the CGS 21680-induced cAMP accumulation was reduced to about the same extent as by 10 μM ADP. However, in one (BT2) of the two pBoNT/C1 expressing clones, the depolarization of PC12 cells failed to cause any alteration in the CGS 21680-induced cAMP accumulation; and in the other clone (BT7), this inhibition was significantly weaker when compared with the values obtained in wild-type PC12 cells (Fig. 3d). Thus, depolarization-dependent inhibition of the cAMP synthesis via P2Y12 receptors was greatly impaired in the pBoNT/C1 expressing cells. Discussion In non-neural cells, release of nucleotides is mostly non-vesicular and involves, for instance, ATP binding cassette transporters, stretch-activated cation channels or connexin hemichannels [22]. In neurons, however, nucleotide release is generally believed to rely on vesicle exocytosis [27]. Here, we employed the autocrine/paracrine activation of P2Y12 receptors in PC12 cells to investigate whether nucleotide release from neuroendocrine cells might involve a non-vesicular component. Vesicle exocytosis involves the SNARE proteins syntaxin, SNAP-25 and synaptobrevin, and cleavage of one or more of these proteins by clostridial neurotoxins reduces or even abolishes exocytotic transmitter release [5]. The primary target of BoNT/C1 is syntaxin, and its inhibitory effect on transmitter release has been correlated with the cleavage of this protein [3]. However, in cultures of bovine chromaffin cells [11], spinal cord [37] or hippocampal neurons [7], BoNT/C1 was reported to cleave not only syntaxin, but also SNAP-25. In contrast, in sensory neurons the toxin was reported to act exclusively on syntaxin [31]. Furthermore, in cell-free assays BoNT/C1 was found to leave either native SNAP-25 after subcellular fractionation or recombinant SNAP-25 unaltered [11, 37]. Thus, the effects of BoNT/C1 on SNARE proteins appear to vary between different preparations. In this study, expression of the BoNT/C1 light chain reduced syntaxin, but not SNAP-25 immunoreactivity detected in PC12 cells. Likewise, in the insulin secreting β-cell line HIT-T15 expression of the BoNT/C1 light chain only affected syntaxin, even though application of recombinant toxin to permeabilized cells did lead to a reduction in SNAP-25 [20]. Hence, heterologous expression of the toxin’s light chain in neuroendocrine cells does not entirely reproduce the effects of toxin application to these cells. Syntaxin was reported to be more sensitive to BoNT/C1 light chain than SNAP-25 [20] and a 1000-fold higher light chain concentration of BoNT/C was required for SNAP-25 cleavage when compared with the other SNAP-25-targeting serotypes BoNT/A or BoNT/E [34]. Thus, the resistance of SNAP-25 to heterologously expressed BoNT/C1 light chain is likely due to the relatively low toxin affinity and insufficient expression levels of the peptide. Nevertheless, in functional experiments, expression of the BoNT/C1 light chain significantly reduced spontaneous as well as depolarization-evoked release of [3H]noradrenaline. On the one hand, this confirms that the cleavage of syntaxin alone is sufficient to reduce transmitter release [3]. On the other hand, this result corroborates that catecholamines are released from PC12 cells through vesicular mechanisms and these are largely impaired by the expression of the BoNT/C1 light chain [10]. Considering that the heterologous expression of the BoNT/C1 light chain did efficiently reduce vesicle exocytosis, the question arises as to how this manipulation altered the release of endogenous nucleotides. In PC12 cells, endogenously released nucleotides activate P2Y12 receptors, which mediate a decrease in the synthesis of cAMP triggered by the activation of A2A adenosine receptors [24]. In accordance with these previous results, exposure of non-transfected PC12 cells to the P2Y12 receptor antagonist 2-MesAMP enhanced the cAMP synthesis. An equivalent effect was achieved when the cells were incubated in the nucleotide degrading enzyme apyrase, thus indicating that spontaneously released nucleotides continuously activated the P2Y12 receptors. To be able to evaluate whether the generation of PC12 clones stably expressing the BoNT/C1 light chain might interfere with the above autocrine/paracrine signalling loop only by affecting exocytosis, we had to establish that the function of the P2Y12 receptor was not compromised by the toxin’s light chain or by the clonal selection procedure. Accordingly, 10 μM ADP, a saturating concentration of the nucleotide in the inhibition of cAMP synthesis in PC12 cells [33], were applied to non-transfected cells and to the two chosen cell clones (BT2 and BT7): the inhibition of cAMP accumulation was not different in the three cultures. Thus, the control of cAMP synthesis via the P2Y12 receptor was not altered by pBoNT/C1 expression or by clonal selection. Previously, we had found that the control of voltage-gated Ca2+ channels via this receptor subtype was the same in a number of PC12 cell clones expressing various types of membrane receptors [19]. Knowing that the P2Y12 receptors operated independently of the BoNT/C1 light chain, the autocrine/paracrine activation of the receptors was assessed. In the toxin expressing clones, the degradation of extracellular nucleotides by apyrase and the blockade of P2Y12 receptors by 2-MesAMP had identical effects on cAMP accumulation as in non-transfected cells. This demonstrates that the spontaneous release of nucleotides was not significantly affected by the BoNT/C1 light chain. Enhancement of nucleotide release by depolarization of PC12 cells fully activates P2Y12 receptors and thereby reduces cAMP synthesis to about the same extent as 10 μM ADP. This phenomenon is based on Ca2+-dependent vesicle exocytosis as it is suppressed when Ca2+ influx is prevented by the blockade of voltage-gated Ca2+ channels or when extracellular nucleotides are degraded by apyrase [24]. In support of this conclusion, the inhibition of cAMP synthesis through a depolarization of PC12 cells by 100 mM K+ was either significantly attenuated or entirely abolished in the pBoNT/C1 expressing clones. Hence, the K+-evoked and thus Ca2+-dependent release of nucleotides was largely reduced by the BoNT/C1 light chain. Taken together, the present results show that the expression of the BoNT/C1 light chain impairs vesicular transmitter release from PC12 cells, but does not alter the autocrine/paracrine activation of P2Y receptors by spontaneously released nucleotides. Thus, the spontaneous nucleotide release from neuroendocrine cells may occur independently of vesicle exocytosis. The depolarization-evoked nucleotide release, in contrast, relies predominantly on exocytotic mechanisms.

Eur_Arch_Otorhinolaryngol-4-1-2279159

Dilatation tracheoscopy for laryngeal and tracheal stenosis in patients with Wegener’s granulomatosis

Wegener’s granulomatosis (WG) frequently involves the subglottis and trachea and may compromise the upper airway. The objective of this study is to evaluate retrospectively the effect of treatment of subglottic stenosis (SGS) and tracheal stenosis (TS) by dilatation tracheoscopy (DT) in patients with WG. We performed a cohort study on all patients who underwent DT between February 2001 and September 2005 in our institution. From this cohort we identified a total of nine WG patients. In all patients, clinical, serological and histopathological data had been prospectively collected by a standardized protocol from the time point of diagnosis. In the nine patients that were identified with SGS or TS due to WG (eight women and one man), a total of 22 DT’s were performed. Two patients needed a tracheostoma (one temporarily). The mean follow-up after the first DT was 25.4 ± 14.1 months. Two patients did not experience a recurrence of SGS or TS. Six patients required a second DT without recurrence of local disease. The remaining patient underwent 8 DT's in a 4-year period. DT can offer a simple and repeatable solution to SGS and TS due to WG. Seven of the nine patients required more than one dilatation and some patients experience a functional restriction. One patient has a definitive tracheostoma. Introduction Wegener’s granulomatosis (WG), first described by Friedrich Wegener in Stuttgart in 1936, is multi-system disease characterized by a necrotizing granulomatous arteritis of the upper and lower respiratory tract and a necrotizing crescentic glomerulonephritis [1]. Vasculitis may also affect other organs such as the eyes, skin, joints, heart and the nervous system [1]. Upper respiratory tract manifestations—particularly sinusitis, ulcerations of the nasal mucosa and epistaxis—are common and debilitating presentations. Subglottic stenosis (SGS) and tracheal stenosis (TS), however, are potentially life threatening presentations of WG. SGS has been found to occur in approximately 16–23% in patients carrying a diagnosis of WG [2, 3]. This narrowing of the upper airway at the level of cricoid cartilage and/or upper tracheal rings presents a management dilemma. Dilatation tracheoscopy (DT) is one of the possible options for treatment of SGS and TS. The long-term effects of DT in patients with WG have never been reported. DT is a minimally invasive self-standing procedure. It can be used for elective and emergency intervention in patients with subglottic and tracheal stenosis. This has been described elsewhere (G.B. Halmos, F.G. Dikkers, Dilatation tracheoscopy in treatment of subglottic and tracheal stenosis, submitted for publication). Patients suffering from WG present differences (age, sex, response to treatment) compared to other aetiologies, which justifies separate publication. The objective of this study is therefore to evaluate the effect of the treatment of SGS and TS by DT in patients with WG. Methods We retrospectively identified all patients from our institution that underwent a DT between February 2001 and September 2005. The total cohort of benign, grade II (Myer-Cotton) subglottic or tracheal stenosis consisted of patients with a history of posttraumatic tracheal stenosis, thermal tracheal stenosis, posttracheotomy cicatricial stenosis, tracheal rupture, postintubation stenosis, and WG, amongst others. As stated before, the cohort is described elsewhere. From the cohort we identified a total of nine WG patients. These nine patients had been diagnosed with WG between May 1990 and August 2003. During this period a total of 195 patients in our institution were diagnosed as having WG (including the nine patients who underwent DT). From this cohort, one additional patient underwent a DT for SGS in 1992. Three other patients were diagnosed with relatively mild SGS or TS that did not necessitate DT. The reference group for the comparison of demographic and serological data consisted of the total cohort, minus the patients who underwent DT and minus the three patients with mild tracheal stenosis who did not undergo DT. Differences in age between the DT group and the reference group were tested with the Mann Whitney U test; differences in the male/female ratio and antineutrophil cytoplasmatic autoantibodies (ANCA) specificity were tested with chi-square test. Although this was a retrospective study, the clinical, serological and histopathological data of both the DT group and the control group had been prospectively collected by a standardised protocol starting at the time of the diagnosis of Wegener’s granulomatosis. Extrarenal organ involvement was categorized as described previously [4]. ANCA was assayed with indirect immunofluorescence. Sera from all patients were assayed for the presence of antibodies against proteinase 3 (PR3–ANCA), myeloperoxidase (MPO–ANCA) and elastase as described previously [5]. DT was performed when patients complained of progressive dyspnoea in combination with decrease of peak flows. DT is an endoscopically performed intervention. An intubation laryngoscope, a Groningen optical dilatation tracheoscope (Karl Storz 1033R) (Fig. 1), telescope and suction tubes are required for this procedure. The tracheoscope has a length of 30 cm. The proximal end of the tracheoscope is designed in such a way that customary ventilation tubes and a 30 cm Hopkins® straight forward telescope (Karl Storz 27005AA) can be connected. The distal end of the instrument contains numerous lateral tiny openings, which enable air to come through in the centre of the stenosis. The dilatation tracheoscope is available with a diameter of 8 and 12 mm. The appropriate size of the instrument is determined by the size of the patient’s larynx and the healthy portion of the trachea. Fig. 1Groningen dilatation tracheoscope The intervention is carried out under general anaesthesia. Following the administration of the anaesthesia, with ventilation taking place via an anaesthesia mask, the dilatation tracheoscope is introduced under endoscopic control. The stenosis is then visible through the vocal cords (Fig. 2). The bevelled design of the tip, which can be advanced forward through the stenosis, ensures that the ventilation is maintained during the process. The conical construction of the tip enables the instrument to be advanced up to the wider section of the tracheoscope (Fig. 3), after which the tracheoscope remains in place for 5 to 10 min. Fig. 2Patient nr 9 pre dilatation. During laryngotracheoscopy a subglottic stenosis is clearly visible. The vocal cords can be seen bilaterallyFig. 3Patient nr 9 during DT. The dilatation tracheoscope is introduced through the stenosis. Parts of the tissue protrude through the tiny distal openings of the tracheoscope The constellation of the tracheoscope is suitable for most grade II (Myer-Cotton) subglottic or tracheal stenosis. The intervention can be repeated after any time interval. Typically, no antibiotics or corticosteroids are used. The use of mitomycin-C was considered in each case, but it was never used, because mitomycin-C should be used in fresh wounds. Patients are dismissed the day after DT. Peakflows were measured three times (Respironics Healthscan Inc., Cedar Groove, NJ, USA) in an upright position. The mean of the three measurements was taken. Results Patients The total cohort of patients who underwent DT because of SGS and TS was 25. Nine of them were identified as patients with WG. Eight were female, one was male. The other 16 patients did not have vasculitis as cause of their stenosis. The causes in these 16 patients were posttraumatic [5], postintubation [4], idiopathic [4], post tracheotomy [2] and post thermal [1] injury. The mean age of the patient group at first presentation of WG was 41.6 ± 11.2 years (Table 1). The interval between the presentation of WG and the first symptoms of laryngeal and tracheal stenosis varied between 0 and 122 months (mean 48.9 ± 44.1). The average age at of the time of presentation of the stenosis was 45.6 ± 11.9 years (Table 1). Table 1Patient characteristics of nine patients with Wegener’s Granulomatosis (WG) and subglottic stenosis (SGS) or tracheal stenosis (TS). Clinical and laboratory findings. ANCA antineutrophil cytoplasmatic autoantibodies, PR3 proteinase 3, MPO myeloperoxidasePatient No.SexAge at presentation of WGOrgans involved at first presentation of WGANCA specificityNumber of relapses of WGOrgans involved at first relapse of WGFollow-up after diagnosis of WG (years)Age at first presentation of SGS or TSInterval between diagnosis of WG and first treatment of TS or SGS (months)1M44JointsPR312Joints, skin1751922F26Lungs, TracheaPR31Lungs, kidney142603F43TracheaMPO1Lungs124304F57JointsPR32Ear, eye, kidney, joints12671225F26TracheaPR31Kidney, eye1131606F51MastoidPR32Nose (concha inferior), mastoid1058817F42Ear drumAtypical0Lung, ear drum845358F52NosePR31Nose (septum and concha), joints, eye756489F33Nasal vesselsPR30Vessels4332 At diagnosis, all WG patients had additional organ involvement outside the otorhinolaryngologic area. The kidneys, lungs and eyes were affected three times. The joints were affected two times. The skin was affected once (Table 1). As shown in Table 1, the number of relapses of WG varied between 0 and 12. There was no relation between the number of relapses and the occurrence of TS or SGS. In only one of the DT procedures there was endoscopic and/or histopathological evidence of vasculitis activity. The TS or SGS occurred in 95% in periods where the disease appeared to be less active. In Table 2 the patient characteristics of the 9 patients who underwent DT for SGS or TS are compared with those patients who were diagnosed with WG in the same time frame but who did not have evidence of SGS or TS. In the DT group, 89% were female whereas 43% of patients in the reference group were female (P < 0.01). In the DT group the age at diagnosis of WG was significantly lower in comparison with the reference group (P < 0.05). Table 2Patient characteristics of nine patients with Wegener’s granulomatosis who underwent DT for SGS or TS compared with the reference group of 182 patients who were diagnosed with WG in the same period without evidence of SGS or TSWG patients who underwent DT (n = 9)WG patients without evidence of SGS or TS (n = 182)P valueMale/ female number (%)1/8 (11%/89%)104/78 (57%/43%) <0.01Age at diagnosis of WG (years, mean ± SD)41.6 ± 11.253.3 ± 17.0<0.05ANCA specificityPR3-ANCA: n = 8 (89%)MPO-ANCA: n = 1 (11%)PR3-ANCA: n = 160 (88%)MPO-ANCA: n = 13 (7.1%)HNE-ANCA: n = 1 (0.5%)ANCA-negative: n = 8 (4.4%)Not significantWG Wegener's granulomatosis, DT dilatation tracheoscopy, SGS subglottic stenosis, TS tracheal stenosis, ANCA antineutrophil cytoplasmatic autoantibodies, PR3 proteinase 3, MPO myeloperoxidase, HNE human neutrophil elastase Follow-up of the DT-group In the nine patients identified with SGS or TS due to WG, a total of 22 DT’s were performed (Table 3). The mean follow-up after treatment was 25.4 ± 14.1 months (Table 3). Two patients (numbers 7 and 8) did not experience a recurrence of significant stenosis. Six patients (numbers 1, 2, 3, 4, 6, 9) required a second DT without recurrence of local disease activity after the second DT. One of these patients (patient number 6) had acute WG in the trachea, proven by biopsy (Fig. 4). She was treated with high dose prednisolone (intravenous methylprednisolone 1000 mg on three consecutive days) and intubation for a week. She could be detubated a week later (Fig. 5). Two patients needed a tracheostoma (patient numbers 2 and 5). One patient (number 5) developed a cricoid stenosis 5 years after the diagnosis of WG. Initially the stenosis was treated with two DT’s. During a pregnancy the patient required an emergency tracheotomy because of a threatened airway. After pregnancy she was treated with a CO2 laser and three DT’s. Unfortunately, however, she required a definitive tracheostomy. Currently the process of decannulation is taking place. One patient (number 2) became respiratory insufficient at diagnosis of WG and had a tracheotomy for 3 months. She subsequently developed a cicatricial SGS, but she has enough lumen to live without a tracheostomy (Table 3). Table 3Follow-up after last DTPatient numberLocation of stenosisNumber of DT’sFollow-up after last treatment (months)Tracheotomy required1Subglottic219No2Subglottic221Yes, temporary3Trachea243No4Trachea214No5Cricoid812Yes6Subglottic216No7Subglottic154No8Subglottic126No9Subglottic224NoFig. 4Patient nr 6 with acute WG during biopsy of a subglottic stenosisFig. 5Patient nr 6 after pulse therapy and intubation. The picture was taken one week after Fig. 4 The effect of DT on peakflows was evaluated in five patients. In these patients the mean (±SD) peakflow increased from 164 ± 45.7 l/min before DT to 226 ± 69.6 l/min after dilatation. An example of the effect of DT on the peakflow is shown in Fig. 6. Two months after this patient had been diagnosed with WG she developed complaints of dyspnoea. Tracheoscopy revealed a SGS with signs of active vasculitis. After treatment with high-dose corticosteroids she had a regression of the WG activity. Fourteen and 28 months after first diagnosis of SGS a dilatation tracheoscopy was performed successfully. Fig. 6Peak flow measurements after dilatation of patient number 9. The arrows indicate the DT procedures Discussion There are multiple causes of benign laryngeal or tracheal stenosis, the most common being traumatic. However, if there is no prior history of tracheal trauma, the aetiology of the stenosis may be obscure and difficult to determine, necessitating a systematic approach to make the diagnosis. Excluding trauma, the differential diagnosis of TS can be subdivided into four categories: congenital, neoplastic, infectious, and inflammatory. Congenital TS is really quite rare and is often the result of posterior fusion of the tracheal rings, thereby forming complete rings. Other causes of congenital stenosis include vascular rings and other congenital cardiovascular anomalies such as an anomalous subclavian artery. These patients typically present at young age. Primary benign tumours of the trachea such as chondromas, fibromas, squamous papillomas, hemangiomas, and granular cell tumours are also unusual causes of stenosis. In addition, extrinsic compression of the trachea can occur by thyroid neoplasms and goiters. A number of infections of the bronchopulmonary tree can lead to TS. Fungal infections such as histoplasmosis and blastomycosis should always be considered when the aetiology of the stenosis is unclear. Serologic testing and histopathologic examination can be helpful in this regard. Other infectious causes of TS include rhinoscleroma, tuberculosis, syphilis, and diphtheria. Non-infectious, inflammatory causes of TS include sclerosing mediastinitis, primary amyloidosis, and sarcoidosis. WG and relapsing polychondritis can also cause TS, but they are almost always seen in combination with other, more classic hallmarks of these diseases. A laryngeal or tracheal stenosis is optimally diagnosed via tracheal visualization, which is generally performed by an otorhinolaryngologist. Indirect and fiberoptic laryngoscopy are non-invasive examination techniques that can be performed in the office, but usually do not show the entire trachea. A subglottic stenosis is not always visible, therefore direct rigid tracheoscopy in general anaesthesia is indicated in cases suspect of WG. Recurrent SGS is a well recognized but uncommon feature of WG. Patients with WG who develop laryngotracheal disease have usually already been diagnosed as such because of the presence of disease in other organs at the same time or at a previous occasion [6]. Isolated involvement of the subglottic larynx is conspicuously rare—only a few cases have been reported in the literature [2, 7, 8]. In our series, however, three patients had SGS or TS as presenting symptom of their WG. We do not have an explanation for this rare presentation. Another particular finding is that eight of our nine patients were female. This contrasts significantly with the male/female ratio in the reference group that consisted of WG patients that were diagnosed in the same period as the DT group but who did not have evidence of SGS or TS. It is difficult to speculate on the reason for this female preponderance. Anatomically the female airway is narrower than the male, making it more prone for post-intubation stenosis. However, we even had females presenting with a SGS or TS. More research in larger series has to be performed to explain this phenomenon. In cases where the diagnosis of WG is suspected, ANCA testing should be part of the routine laboratory evaluation [9]. C-ANCA has been shown to be a specific marker for WG with rare false-positive results; p-ANCA testing is much less specific. Only the particular laboratory findings of c-ANCA reacting with PR3 and p-ANCA reacting with myeloperoxidase (MPO) are specific for the autoimmune vasculitides [9, 10]. Notably, seven of our nine patients that underwent DT were PR3-ANCA positive and one patient was MPO-ANCA positive. All patients had involvement of additional organs. In three patients the disease presented as a stenosis, and laryngotracheoscopy revealed the size and site of the lesion. Organs involved showed no pattern in which there should be additional suspicion for the development of SGS or TS. There was no relation between the number of relapses and the development of TS or SGS. The TS or SGS occurred more often in periods where the disease appeared to be less active. Remarkably, TS or SGS predominantly became manifest in periods in which WG appeared to be inactive. Interestingly, seven out of our nine patients who developed TS or SGS had not been diagnosed previously with tracheal involvement of WG, neither at first presentation nor at relapse(s). We speculate that during active disease a subclinical tracheal involvement occurs which may subsequently heal with scar formation. To examine this hypothesis in patients, elective tracheoscopy should be performed in all patients with WG. The time interval between presentation of clinical stenosis in these seven patients (mean 63 months, median 60, range 2–122) favours watchful waiting. DT with the Groningen dilatation tracheoscope is a safe, minimally invasive procedure for the treatment of Cotton-Myer grade II subglottic or tracheal stenosis of various origin. It is an elegant, self-standing surgical intervention, where no additional interventions are needed. For an overview of the effect of DT the reader is referred elsewhere (G.B. Halmos, F.G. Dikkers, Dilatation tracheoscopy in treatment of subglottic and tracheal stenosis, submitted for publication). Nowadays, mitomycin-C, an alkylating agent that inhibits cell division, protein synthesis, and fibroblast proliferation [11], is increasingly used as adjuvant treatment in the management of selected cases of laryngeal and tracheal stenosis, for example luminal obstruction in fresh circular sutured wounds. However, the laryngeal and tracheal stenoses of WG patients are almost always of older age and display advanced scarring at the time of diagnosis. Dilating such “mature” stenoses will inevitably lead to damage of tracheal epithelium. However, this damage is considerably different from granulomatous scarification in sutured lumina and, therefore, we do not expect an additional favourable response to mitomycin. Therefore, we have not applied mitomycin-C in our WG patient cohort. In cases where TS or SGS develops, there is always the question whether or not to intervene, and the question of timing of the intervention. There are two main parameters in these patients: complaints, and physical signs. One might see this as a two-by-two table, with complaints on the x-axis, and peakflow values on the y-axis. Progressive complaints and progressive decline of peakflow values (+/+) indicate intervention by DT. Absence of progressive complaints and unchanged peakflow values (−/−) indicate watchful waiting. Absence of progressive complaints with decline of peakflow values (−/+) need to be addressed when the values reach a critical level, in which it can be expected that a common cold might lead to severe stenosis of the airways. Finally, progressive complaints with unchanged peakflow values (+/−) indicate that pulmonary function tests should be performed, and, if negative, indicate that the patient has an incorrect perception of his physical potentials. This can then be addressed. Unfortunately, in only five of the nine patients in this retrospective study, we have peak flow measurements. A reason is that two patients have or had a tracheostomy. We have registered no complications or deaths during or because of DT. We have three reasons to regard dilatation tracheoscopy as a minimally invasive intervention. We have experienced no complications related to dilatation tracheoscopy. It requires short hospitalization (generally 3 days). The intervention is not straining for the patients. A variety of surgical techniques has been used to treat SGS or TS of other aetiologies [12, 13]. The success of these surgical techniques in upper airway stenosis related to WG has been variable [3, 14]. In our series two out of nine patients were treated once for upper airway stenosis with dilatation without any re-stenosis because of WG. Unfortunately, two patients required a tracheostomy. One was temporary, but one patient needed a permanent tracheotomy (11%): this was the pregnant woman needing an emergency procedure. This number equals that of Gluth et al. [9]. Conclusion WG as such is a rare disease, and SGS and TS are rare symptoms in patients with WG. DT can offer a simple and repeatable solution to this very serious symptom. Patient complaints and monitoring of peakflow values offer simple tools for the decision whether or not to intervene. However, a causative solution to WG should be the ultimate goal.

Environ_Manage-4-1-2359833

Public-Private Partnerships in China’s Urban Water Sector

During the past decades, the traditional state monopoly in urban water management has been debated heavily, resulting in different forms and degrees of private sector involvement across the globe. Since the 1990s, China has also started experiments with new modes of urban water service management and governance in which the private sector is involved. It is premature to conclude whether the various forms of private sector involvement will successfully overcome the major problems (capital shortage, inefficient operation, and service quality) in China’s water sector. But at the same time, private sector involvement in water provisioning and waste water treatments seems to have become mainstream in transitional China. Introduction In the wake of the United Kingdom’s water privatization in the 1980s, the 1990s witnessed the spreading of privatization and a variety of public-private partnership (PPP) constructions in developing countries, especially following the promotion and push by international development agencies such as the World Bank, the International Monetary Fund (IMF), the Asian Development Bank (ADB) and others (Nickson 1996, 1998; Kikeri and Kolo 2006). It was believed that private sector participation in the water sector would bring in much needed investment and improve service coverage, quality, and efficiency by replacing conventional public-sector systems suffering from under-investment and inefficiencies due to excessive political interference and rent-seeking behavior by vested state and bureaucratic interests (Hall and others 2005). During the past decades, a wide literature in economics, governance, and public management has provided theoretical and empirical arguments and evidence in favor of further private sector involvement in what used to be public utilities. At the same time, however, debate continues on the different partnership constructions, the division of tasks and responsibilities between public and private sectors, and the social effects coming along with these developments. Topics, such as the relationships between ownership (public or private) and efficiency (Vining and Boardman 1992; Spiller and Savedoff 1997; Birchall 2002; Afonso and others 2005; Anwandter and Jr.Ozuna 2002; Hart 2003), the classification of various public-private constructions and their characteristics (World Bank 2004; Seppälä and others 2001; US National Research Council 2002), the consequences of privatization for governmental regulation (Nickson and Vargas 2002; Pongsiri 2002) and questions of equity and equality are still heavily debated, in particular with respect to the water sector and less so regarding other utilities. Although private sector participation in the water sector is one of the more controversial topics in public utility management today, this wave also spread to China at the turn of the millennium, where the government started to reform public sectors (water, electricity, roads, etc.) via introducing market functions. The so-called marketization reform expected to address the increase of several water problems (water shortage, insufficient infrastructure, water pollution, etc.) to meet the requirement posed by accelerated urbanization and high economic growth. As a late comer in this field of private sector involvement in the provision of water services, China is able to learn from numerous experiences of other countries, such as the United Kingdom, France, United States, Chile, Philippines, Mexico, Argentina, and Bolivia. Since the earlier attempts of applying the Build-Operate-Transfer (BOT) approach in the water sector in the 1990s and the full development of marketization reform in public sectors in 2002, China has applied different models of private sector involvement in over 300 water supply and wastewater projects. This marketization reform emphasizes the importance of the market, investment and financial liberalization, deregulation, decentralization, and a reduced role of the state in the water sector (also see Robison and Hewison 2005; Prasad 2006). Tariff reform with full-cost recovery, competitive bidding procedures, changing ownership structures (e.g., public and private, Sino and foreign), and restrictive fiscal policies are part of it. This article reviews developments in private sector involvement in China’s water management and assesses whether expected results of marketization in the Chinese water sector have been met: raising investment for infrastructure, increasing service coverage and improved efficiency in China’s water supply and wastewater treatment. After interpreting further private sector involvement in China’s urban water management in terms of modernizing water governance, this article provides a country-wide overview of current privatization developments in the Chinese water sector, and subsequently makes an in-depth investigation in three distinct cases with respect to the new roles and functions of the governments and private parties. The final section assesses the current status of privatization programs in China’s water management and its implications of future research on water governance reform. Private Sector Participation as Part of Modernizing Urban Water Governance In the debate on private sector participation in environmental governance in general, and urban water governance in particular, we can identify three — sometimes interrelated — discourses. First, private sector participation goes back to the literature on state failure in the early 1980s. State failure refers to the notion that the nation-state falls short in the provisioning of collective goods, in this case environmental services and quality. Some of the key publications in this regard come from Germany. Martin Jänicke’s (1986) Staatsversagen analyzed the fundamental inability of the nation-state to protect the environment in the 1980s, and called for an innovation or modernization of environmental politics, later to be labeled political modernization (e.g., Tatenhove and others 2000; Mol 2002): a reorientation towards a more preventive, pro-active and flexible strategy using new instruments and closer cooperation with and participation of non-state actors. With a similar analysis of the environmental state’s fundamental inabilities, Joseph Huber (1985) came to slightly different solutions with his strong plea for involving the private sector into environmental services and protection. Finally, around the same time Ulrich Beck (1986) formulated his Risk Society hypothesis and identified subpolitical arrangements (i.e., arrangements for environmental protection and service provision without and beyond the public state) as an alternative for the conventional environmental politics of the nation-state. Inspired by these and several other authors and ideas, from the mid 1980s onward environmental social science scholars started to develop ideas, investigate practices, and formulate theories on governing environmental problems, in which the environmental state was given a less dominant and monopolistic position. Around the same time (the second half of the 1980s) ideas of further private participation and involvement in the provisioning of environmental services (water, waste, energy, etc.) started to develop, especially in the United States and the United Kingdom. While also here the fundamental idea is involving the private sector in tasks traditionally fulfilled by the public sector, the orientation and literature is slightly different. The majority of the literature comes from the management and organization sciences and the orientation is less focused on state failures and governance, but rather on efficiency, the bringing in of new capital and the introduction of market logics. The dominant form of organizing urban infrastructure (water, energy, waste, transport) by state agencies has been replaced in many places by various PPP constructions, with different reasons put forward to legitimate such new constructions (cf. Linder 1999). At the same time, these partnerships led to considerable debate, most significantly on issues of equity and equality: who is involved in these partnerships, for who are these constructions bringing more effective and efficient services, are local governments able to balance the power of private capital coming in (especially in situations of Transnational Companies (TNCs) in developing countries) (e.g., Oppenheim and MacGregor 2004), and what does private sector involvement mean for affordability of environmental services for the poor? Thirdly, in the 1990s, following the United Nations Conference on Environment and Development (1992, Rio de Janeiro), and even stronger after the Rio+10 conference (2002, Johannesburg), ideas and practices of public private partnerships started to emerge forcefully on the national and global agenda (cf. Mol 2007). In this literature, the emphasis is strongly on transnational partnering of public and private entities, with a strong focus on the role of civil society organizations. The main reason behind the recent attention to private sector participation in environmental protection and service delivery is related to tendencies of globalization and governance complexities. As Davies (2002) correctly summarizes, in this interpretation the notion of partnership has a positive rhetoric referring to inclusiveness, transparency, participation and dialogue, redistribution of power, and equity. And not so much to ideas of efficiencies, capital investment, market logics, and increased service coverage. In reviewing the arguments and legitimacy of the push for private sector involvement in China’s urban water governance, there is a strong relation to the second discourse on efficiency, capital investments and service coverage, while ideas of state-failure and political modernization incidentally emerge. By the same token, the Chinese discourse on private sector participation in urban water management hardly draws upon ideas of wide cross-sectoral partnerships and the positive logics of transparency, democracy, participation, and dialogue. Discussions on China’s urban water governance reform argue for the advantages of effectiveness and efficiency, and debate the best organizational modes, division of responsibilities, and coordination structures. Potential negative outcomes of private sector participation — so strongly emerging in and dominating western debates — are much less emphasized: loss of decision-making autonomy of states and governments; unequal power relations and information asymmetry in public-private partnerships; problems around equity, access for the poor, participation and democracy in decision-making (e.g., Hancock 1998; Poncelet 2001; Miraftab 2004). According to the World Bank, China, Chile and Colombia are the only countries that remain active in water privatization after 2001 (Izaguirre and Hunt 2005). How to explain that, while the activities of water sector privatization intend to shrink in an increasing number of countries and international development agencies, such as the World Bank, start to slow down such privatization programs, China is actively promoting private sector involvement in urban water governance?Two interdependent arguments elucidate this. First, China’s urban water management comes from a radically different starting position, where market principles and logics were almost absent. Water management was not just completely publicly organized but also highly inefficient, with large capital shortages, poor coverage, no economic incentives and demand side management, and highly centralized. This is a fundamental, rather than marginal, difference with most of the public utility systems in OECD countries before the privatization discourses and practices of the 1980s and 1990s. Under such Chinese conditions, private sector involvement in water management means more that just handing water business over to for-profit private companies. It most of all means building economic incentives and logics, safeguarding enough financial capital for infrastructure investments, and widening the service area. Second, private sector participation in China’s urban water management is not just a matter of privatization. It is part of a much wider and complex modernization program in urban water governance, involving some of the critical issues that emerged in the privatization debates in OECD countries. The modernization of urban water governance also includes (see OECD 2003, 2004a, 2004b, 2004c, 2005a, 2005b):water tariff reforms, where costs of drinking water increasingly include full costs (also of wastewater treatment), but come along with safeguards for low income households to continue access to drinking water;transparency, accountability and control of the government;public participation in for instance water tariff setting, complaint systems on water pollution and corruption, public supervising committees on utility performance, public and media debates on water governance, disclosure of information to non-governmental actors (cf. Zhong and Mol 2007); anddecentralization of water tasks and responsibilities to the local level.In exploring the degree, nature, and forms of private sector participation in China’s urban water governance in the following sections; we have to leave these wider — related — developments aside. Privatization Policy in China’s Water Sector In China, the term “private sector” has been regarded as politically sensitive since 1949 when China started to establish a socialist regime characterized by the nationalization of ownership. The first breakthrough of the development of “private sector,” which was officially defined as “economic organizations that aim at making profit, in which assets are privately owned and which have eight or more employees” (Provisional Regulations of Private Enterprises in PRC, the State Council, June 25 of 1988), took place mainly in competitive sectors in accordance with the launch of China’s economic reform in the late 1970s. The government remained in control of public sectors such as water services, energy provisioning, waste management, and public transport. In the mid 1990s, Chinese Government attempted to introduce the BOT approach into the field of urban infrastructures (thermal power, hydropower, highway, water supply, etc.) via promulgating the Circular on Attracting Foreign Investment through BOT Approach (No.89 Policy Paper of 1994, the former Ministry of Foreign Trade and Economic Cooperation, January 16 of 1995) and the Circular on Major Issues of Approval Administration of the Franchise Pilot Projects with Foreign Investment (No.208 Policy Paper of Foreign Investment, the former National Development and Planning Commission, the Ministry of Electric Power Industry, and the Ministry of Communications, 1995). These two policy papers formed the first legal ground for private sector involvement and foreign capital investment in Chinese urban infrastructure. Subsequently, the National Development and Reform Commission firstly approved three BOT infrastructure projects in 1996, including Chengdu No.6 Water Supply BOT Plant (B), Guangxi Laibin Power BOT Plant, and Changsha Wangcheng Power BOT Plant (failed). The earlier experiences of BOT projects brought in needed capital and investment to develop China’s urban water infrastructure. But it illustrated also many problems. The issue of the fixed investment return to investors was one of these problems. After intensifying control over foreign exchanges and loans in the late 1990s, the General Office of the State Council promulgated a specific circular in 2002 to correct foreign investment projects with fixed investment returns, by modifying the relevant contract terms, buying back all shares of foreign investors, transferring foreign investment into foreign loans, or dismantling contracts with often severe losses. The full-fledged commitment of the Chinese government to private involvement in the water and other utility sectors dates from late 2002. In the December of 2002, the Opinions on Accelerating the Marketization of Public Utilities (No.272 Policy Paper of the MOC, 2002) started the marketization reform of water and other public sectors by opening public utilities to both foreign and domestic investors: multi-financing approaches, concession right and concession management, pricing mechanism, reduction of governmental monopolies and roles ended the traditional policies of public utilities. The subsequent Measures on Public Utilities Concession Management (No.126 Policy Paper of the MOC, 2004; in this policy, “concession management” refers to all forms of private sector participation.) of 2004 specifies the procedure of how to involve the private sectors in public utilities through awarding concession right, but still relies heavily on BOT modes. These steps proved more than just giving the private sector a permission to enter public utilities. It is a complex process involving among others ownership reforms, redefinition of the role of governments and operators, restructuring the tariff mechanism, reforming governmental regulation, and designing public participation. In the early years of marketization, the emphasis was especially on market opening and financing issues. With Opinions on Strengthening Regulation of Public Utilities (No. 154 Policy Paper, the MOC, 2005) the neglect of governmental regulation and the public good character of water in the previous policy papers was corrected. This policy paper emphasizes that the water sector provide basic public and social goods and that the governmental regulation remains essential (Fu and Zhong 2005). However, there is still a lack of a systematic and comprehensive regulatory framework for the Chinese urban water sectors in practice. The MOC is attempting to introduce and develop a competitive benchmarking system that might be helpful for further regulation, but this is not yet in place. During the authors’ field surveys, the local officials of relevant water authorities are laboring under the lack of effective measurements for regulatory framework, have too much freedom of (non)regulation, and have sometimes an incorrect perception of the government role as a regulator. Fu and colleagues (2006) also refer to the fact that the government has paid some attention to assets regulation while restructuring ownership in the water sector, but neglected regulating water service quality. Compared to the exponential growth of water projects with significant private sector involvement, the legal basis under privatization developed quite slow and is still underdeveloped in China. Different from some water privatization forerunner countries (e.g., England and Wales, Philippines), which enacted specific laws before entering into privatization, the marketization reform of and private participation in the Chinese water sector is conducted under various governmental policy papers, but without specialized legislation. The current legal codification of public-private partnering in water services is largely a reactive process, where various policy papers address specific problems in the reform process due to the lack of a well-established legal framework. Thus, much room for improvement remains in the current legal basis, for instance on further economic regulation, stronger legislative sanctions, and public participation (cf. Tong 2005; Zhang 2006; Fu and Zhong 2005; Fu and others 2005). As implementation problems were slowly or not adequately addressed or resolved at the national level, local governments started to issue local policy papers on specific water projects. For instance, the Interim Provision on Administrating Concession Right of Chengdu (No.131 Policy Paper of Chengdu Municipality, 2001) was issued for implementing the BOT project of Chengdu No.6 Water Supply Plant (B), which was the first water BOT pilot project approved by the NDRC. And the Measures on Public Utilities Concession Management of Shenzhen (No.124 Policy Paper of Shenzhen Municipality, 2003) guided the reform of Shenzhen Water Group, the largest water project with private sector involvement in China to date. The Current Landscape of Private Sector Involvement In China’s water supply and wastewater services, four major types of private corporations are active (Fu and others 2006): (1) the water transnational corporations (e.g., VEOLIA and SUEZ); (2) Chinese investment developers (e.g., Beijing Capital Group and Tianjin Capital Environmental Protection Co. Ltd.); (3) liberalized water companies (e.g., Shenzhen Water Group and Beijing Sewerage Group); and (iv) environmental engineering corporations (e.g., Beijing Sound Group and Tsinghua Tongfang Water Engineering Corporation). In December of 2004, the Ministry of Construction called provincial-level authorities to summarize marketization of public sectors (e.g., water and wastewater, solid waste, gas, and public transportation). In July of 2005, a follow-up field survey was organized by the MOC, in which the authors have participated. All reported data of this section come from the reports of provincial-level authorities, supplemented by surveys of the Water Policy Research Center of Tsinghua University (in which authors participated). According to the MOC surveys, various forms of private sector participation can be identified in both water supply and waste water treatment: (1) commercialization of public utilities: it is the transformation of a public agency/utility into an independent corporation; (2) management contract (or namely operations and maintenance contract): it refers to a contractual arrangement in which a private operator manages and maintains the service in a given period but does not have investment obligations; (3) lease contract: it is a short-term contract in which a private operator pays an agreed-upon fee to the government for the right to manage the facility; (4) Greenfield contract (such as BOT, TOT, BOOT, etc.): it means the government commits new investment projects to a private company, within the contract duration, the private operator manages the infrastructure and the government purchases the water by a contracted price (this price isn't necessarily determined by the actual water tariff); (5) concession contract: it is a long-term contract in which a private operator bears responsibilities for operations and maintenance and also assumes investment and service obligations; (6) Joint Venture: it is not a contract but, rather, an arrangement whereby a private company forms a legal entity with the public sector, in which both the private and the public parts share responsibilities and (investment) obligations; and (7) full sale (or full divesture): it is the sale of public assets to the private sector. Table 1 summarizes the various forms of private sector participation and their characteristics. Until July 2005, a total of 152 water supply projects and 200 wastewater treatment projects involved private participation. The total water production capacity of the 152 water supply projects equaled about 17% of national water production capacity of 2004. The treatment capacity of the 200 wastewater projects was over 30 million m3 per day, equaling 67% of the national total wastewater treatment capacity of 2004. Table 1Different forms of private sector participation in China’s water sectorForm of private sector participationAsset ownershipCapital investmentOperations & maintenanceContract periodCommercialization of governmental enterprises/utilitiesPublicPublicPublicIndefiniteManagement contractPublicPublicPrivate3–5yrLease contractPublicPublicPrivate8–15yrGreenfield (BOT-type)Private/ publicPrivatePrivate20–30yrConcessionPublicPrivatePrivate25–30yrJoint ventureSharedSharedSharedIndefiniteSale or full divesturePrivatePrivatePrivateIndefinite Figure 1 shows the prevalence of different forms of private sector participation in water supply and in wastewater projects. The joint venture approach (including the Sino-foreign joint ventures) has the largest share in the water supply sector with 51% of the 152 privatized projects. The Greenfield modes of private sector participation (including the BOT and TOT contracts) dominated in the wastewater sector, with 59% of the 200 projects. The commercialization of governmental utilities also plays an important role in both water supply (16% of 152 projects) and wastewater (13% of 200 projects). The differences in prevalence of private sector participation forms between water supply and wastewater have a close relation with the level of infrastructure development and with tariff levels. Compared to urban water supply (with service coverage of 88.8%), urban wastewater treatment lags behind, with a service coverage of 45.6% in 2004 (MOC 2005). Direct investment demand for urban wastewater infrastructure (including wastewater treatment, sewers, and sludge treatment) in China is expected to be over 30 billion US dollars between 2006 and 2010, to meet the objective of 60% municipal wastewater to be treated. Accordingly, local governments prefer direct private sector investment in and building of new wastewater infrastructure, resulting in high levels of the Greenfield modalities. In addition, the current low wastewater treatment charges result in a preference for Greenfield modes. In these modes, financing is based on negotiated prices between the government and the private sector and is less dependent to the user fee or charge; drinking water supply costs are much better represented in prices, making joint ventures more likely (Zhong and others 2006). Fig. 1Public sector participation in water: distribution over modalities (2005) Figure 2 categorizes public sector participation into five groups, according to project capacity. The joint venture approach leads the reform of water supply sector in all size-categories, while the Greenfield approach dominates in wastewater sector, except for projects over 500,000 m3 per day. This might also be related to the different financial risks. Larger projects require much more direct capital investment from the private sector, increasing the financial risk for private investors and moving, then, rather toward joint venture approaches. Furthermore, the full sale/divesture approach occurred more in the field of water sector and mainly in small projects in specific provinces (see Fig. 3). And commercialization is more often found among larger projects. This might be related to not only the larger capital demands of bigger projects, but also huge labor redundancies within such large projects. Existing large water projects are traditionally run by state-owned enterprises with high levels of superfluous workers. For private investors it is often difficult to improve efficiency, because government contracts often do not allow firing existing workers following a commercialization process. Fig. 2Distribution of private sector participation in water projects by capacitiesFig. 3Distribution of private sector participation in water projects by provinces Figure 3 visualizes the provincial distribution of water projects with private sector participation. At least 25 provinces have private sector participation experience in water supply and 23 provinces in wastewater treatment. The form of private sector participation is determined by the level of development of water/wastewater infrastructure, as well as the local economic, social and political conditions. With richer markets, more open economic policies and higher payment capacity of local residents, the southern coastal (e.g., Guangdong and Fujian) and the eastern coastal (e.g., Jiangsu) provinces witnessed high levels of reform in their water sector. Over 60% of foreign private sector investment in water supply projects and about 50% foreign private sector investment in wastewater projects were implemented in these coastal regions, according to the MOC survey. In the meanwhile, the first national BOT pilot project of Chengdu Water Supply (Sichuan Province) has triggered a wave of private sector participation in and around Sichuan Province (including Chongqing and Yunnan). Furthermore, the special environmental protection policies related to “The Three Gorges” dam might have impelled private sector participation in wastewater sector of Sichuan Province and Chongqing. As shown in Fig. 3, in water supply the joint venture approach dominates in 19 provinces. In the wastewater sector, Greenfield projects (including BOT and TOT) dominate in 12 provinces. The commercialization of traditional state-owned water enterprises was adopted more widely in inland provinces (such as Gansu, Heilongjiang, Jilin, Sichuan, Xinjiang, Yunnan) than in coastal provinces. A joint venture approach for private sector involvement in the wastewater sector was only adopted in provinces with high wastewater treatment charges, such as Beijing, Fujian, Jiangsu, Zhejiang, and Shanghai. Three Case Studies of Public-Private Partnerships The reported growing involvement of the private sector has led to radical changes in China’s water management institutions. In this section, we report on fieldwork of three case studies with distinct modes of private sector involvement (a joint venture, a concession, and a Greenfield contract) to analyze in detail the new institutions and relationships between actors in these constructions. During fieldwork in Maanshan and Shanghai, we carried out face-to-face semi-structured interviews with relevant local officials (from the construction authority, price authority, planning and reform authority, state-owned assets administration authority, and environmental protection bureau) and managers of water service providers (water treatment plants/companies, wastewater treatment plants/companies). In the performance assessment project of Macau Water Company Ltd., the managers of relevant departments as well as the representative of Macau Government were interviewed. In total, around 30 interviews were held. While these three cases represent different forms of private sector involvement, they cannot be held representative. All three cases have been assessed positively by the Chinese government and independent researchers (see Fu and others 2006), making them rather best practices than representative cases. But together they illustrate the institutional transformations that come along private sector involvement. Joint Venture: Maanshan Water Supply Maanshan City is an industrial, prefecture-level city of 1686 square kilometers, and a population of 1.24 million (2004), of whom 46.8 per cent lives in urban areas. According to the 2004 MOC statistics, 88.7 per cent of the urban population has access to water supply. Water resources are abundant in Maanshan City due to its advantageous location on the south bank of the Yangtze River and abundant annual rainfall (1062–1092 mm). Maanshan Construction Commission (MASCC) is not only the competent authority for water supply and wastewater treatment and as such, plays a leading role in the water sector reform. It is also, as a so-called “Big Construction Commission,” the main governmental agency responsible for urban planning, construction, and management (cf Wu 2003). In 2002, following the call of Central Government and Anhui Provincial Government, MASCC embarked upon marketization reform in water and other public utilities (e.g., gas and public transport), widely inviting business actors to become active and invest. The director of MASCC, Mr. Xu, argued that changing the current water institutions and increasing service quality were the most important reasons and objectives for embarking on marketization in the water sector in Maanshan, rather than bringing in nongovernmental capital (personal communication 2004). Marketization was expected to impel and accelerate the reform of converting the old Maanshan Water Supply Company (MASWSC, established in 1958 as state-owned and state-subsidized company with total assets of 4.37 million RMB in 2002, ca. 0.528 million US$ at the exchange rate of 1US$ = 8.277RMB) into a new institutional lay-out. After negotiating with several private companies, MASCC first started — as a kind of trial — a joint venture with Beijing Capital Group (BCG) for one water supply plant (WTP, BCG owning 60% of shares). This joint WTP sold purified water to MASWSC and performed significantly better than other WTPs managed by MASWSC alone. In 2004, MASCC expanded the joint venture cooperation with BCG to all WTPs of Maanshan City, in which BCG obtained a 60% share by bringing in 90 million RMB (ca. 10.875 million US$ at the exchange rate of 1US$ = 8.276RMB). The new joint venture company (MAS-BCWLC) was awarded a 30-year concession right. Both BCG (private sector) and MASWSC (public sector) bear responsibility of investment, operation, and maintenance of the WTPs (excluding the pipe networks) and service obligations (see Fig. 4). With respect to the pipe networks, MAS-BCWLC manages and maintains the existing (pre-2004) network by signing a lease contract with MASWSC, which remained owner of the assets and bears the financial obligations (debts). In the meanwhile, MAS-BCWLC is requested to invest in new pipe infrastructure in new development areas and in nonpiped neighborhoods. Fig. 4Organizational structure of Maanshan water supply system Within the new joint venture structure, the board of MAS-BCWLC (4 members from BCG and 3 from WASWSC) is the current decision-maker regarding planning (within the objectives set by the municipal master planning), investment and financing, partly replacing the tradition of government decision structures. According to the contract, the general manager of the joint venture company comes alternately from MASWSC and BCG. Taking into account the social dimensions of water provisioning, the government claimed three key conditions in the agreement with the concessionaire: first, the concessionaire (MAS-BCWLC) must ensure sufficient and safe water provision and the government can take over all facilities without any indemnity if the concessionaire fails; second, the concessionaire cannot change the public and social nature of water and should include relevant social responsibilities as governmental requirements (e.g., employing all personnel from the old water company, providing free water for firefighting, reducing/subsidizing water bills of the poor); third, the government controls the water price. In order to ensure high-quality water and service, MASCC regulates the performance of MAS-BCWLC via assessing annually the specified objectives approved by both the MAS-BCWLC board and MASCC. For instance, MAS-BCWLC was requested to achieve 12 key objectives in 2004: (1) investment of 18 million RMB (ca. 2.175 million US$ at the exchange rate of 1US$ = 8.276RMB); (2) selling 48 million cubic meter water or more and reclaiming >90% of water bills; (3) fulfilling indicators of water service quality (for instance, >99% of the control points should reach the required water quality standards; >98% control points should reach standards for water pressure; a maximum of 30% water loss; burst pipes repairs within maximum time limits); (4) fulfilling all MASCC indicators for safe work; (5) construction of the main body of the No.4 WTP and 25 kilometer new pipes; (6) fulfilling client service indicators (for instance, 100% good client service; >90% public satisfaction); (7) fulfilling the reconstruction of Xiangshan Town water supply system; (8) elaboration and submitting a water supply plan; (9) achieving the relevant objectives of National Civilized City Assessment System (which was proposed by Central Cultural and Ideological Building Commission in 2004; it includes 119 indicators); (10) submitting water supply plans to Municipal People’s Congress and Municipal People’s Political Consultative Conference; (11) responding adequately to complaints and reporting this information to the government; and (12) take anti-corruption measures. After establishing the joint venture in 2002, the total length of pipes and the volume of water provision have increased (see Fig. 5) and MAS-BCWLC has been in compliance with all requirements of the government, according to interviews with local officials. From 2004 to 2005, MAS-BCWLC has invested about 90 million RMB (ca. 10.875 million US$ at the exchange rate of 1US$ = 8.276RMB) for building new infrastructure, updating old facilities and aged pipes, and establishing a customer service system. In the meanwhile, the government has stopped subsidizing WTPs after the involvement of BCG and the joint venture even turned over about 18.7 million RMB (ca. 2.260 million US$ at the exchange rate of 1US$ = 8.276RMB; including 2 million RMB of the rent fee for pipe networks, 4.7 million RMB of dividends, 7.7 million RMB of corporate income tax, 3 million RMB of value added tax, and 1.3 million RMB of the expense of other taxation; the total taxation of 12 million RMB is about 25% of the total turnover of MAS-BCWLC in 2004) to the local government in 2004. The improved service quality of water provision not only satisfied the consumers, but also resulted in government (and the price public hearing; cf. Zhong and Mol 2007) support for the first tariff reform after private sector involvement in 2004. Maanshan Government increased the water tariff from 0.83 to 1.08 RMB/m3 (ca. 0.10 to 0.13 US$/m3 at the exchange rate of 1US$ = 8.276RMB; rate for household consumers) and indirectly subsidized MAS-BCWLC by moving the additional tax of water provision (e.g., 0.05 RMB/m3 for household consumers) to the income of the joint venture water company. In 2004, the per capita annual income of urban households of Maanshan was 10,189 RMB (ca. 1231.15 US$ at the exchange rate of 1US$ = 8.276RMB), of which around 1.16% was spent on water services (calculated based on daily household water use of 300 liters per capita). Fig. 5Total length of pipes and annual water provision in Maanshan (1995–2004) Obviously, the involvement of BCG has brought in additional capital to develop Maanshan’s water supply sector. But more importantly it has changed the institutional structure, improved the water service quality and quantity, as well as reduced the governmental input in this field (see Fig. 6). In this structure, the government benefits both from the taxations and dividends of the joint venture company, while transferring part of the financial, building, and operational risks to the private sector. Following this model of Maanshan City, BCG has successfully expanded its activities to other cities, such as Huainan (Anhui Province), Baoji (Shanxi Province), and Yuyao (Zhejiang Province). Fig. 6Monetary flow within Maanshan water supply However, this private sector involvement practice of Maanshan is argued to have a (potential) political risk due to the lack of a sound legal basis. In transitional China, in particular, policies are perceived to be instable and insufficiently law-based. Until now, details on measures and rules to regulate private utility companies are still missing in current national and Anhui provincial policy papers (Maanshan has no legislation right). This is a common problem in Chinese marketization practices in the water sector, as argued by many lawyers and academics. For instance, Shenyang water supply has experienced several failed marketization practices due to the constantly changing policies and decisions of the local government during 1995–2000 (field survey 2004). Concession Contract: Macau Water Supply Macau is one of the two Special Administrative Regions of China, together with Hong Kong. Administrated by Portugal until 1999, it was the oldest European colony in China, dating back to the 16th century. As a small territory of 28 km2 on the southern coast of China, consisting of a peninsula and the islands of Taipa and Coloane, it has a population of 508,000 (2006). Macau has a long history in the private provision of drinking water, since the earliest Macau Water Company Ltd. (MWC) was founded in 1932 as a full private capital company invested by individuals. Three years later, MWC was taken over by a British Electricity Lighting Company for 10 years and since 1946 by the president of the Macau Economic Department and other individual shareholders. Due to lack of capital and advanced technologies, Macau had an inadequate water supply service with poor water quality and discontinuous water provision during the 1970s. In 1985, Macau Government, learning from the concession management practices in the French water sector, awarded a consortium of two private companies, NWS Holding Limits (Hong Kong) and SUEZ Environment (France), a 25-year concession contract. Macau Government remained owner of the existing, pre-1985, assets (plants and pipe networks), while the private Macau Water Supply Ltd. (MWSL, the former MWC) bears responsibilities for operations and maintenance of these assets, as well as for new investments and service obligations (see Fig. 7). This concession contract is not only the first private sector participation construction in Chinese water sector, but also the first contract that seems to end with a positive result. Fig. 7Organizational structure in Macau water supply Distinct from the previous private owners, who had little experience in the field of water provision, SUEZ (France) brought in advanced water knowledge and technology. According to the concession contract, MWSL must provide high-quality water supply service, as well as bear several obligations, such as planning, investment, construction, operation, and maintenance of the infrastructure under the supervision of Macao Government. In practice, Macau Government has delegated tasks, responsibilities and obligations to a very large degree to MWSL. Coming to the end of the 25-year concession contract, MWSL has fulfilled almost all terms of the initial contract. It has, among others, considerably improved water service quality by increasing service access and provision, decreased the loss of water leakage (see Fig. 8), and kept water tariff (corrected for inflation) at a stable level (see Fig. 9). Fig. 8Annual water demand-provision and water loss in Macau (1982–2005)Fig. 9Water tariff rates of Macau (1982–2005) In the concession contract, the government did not specify conditions and safeguards for the poor. But in practice, MWSL not only reduced the water bill for low-income, disabled and other vulnerable groups. For instance, MWSL has launched the “Elderly-In-Needs” water subsidy program in 2001, which offers those aged over 55 free water consumption of 5 m3 per month. Since May 2005 the “Water for All” program offers free water consumption to other categories of people in needs, such as single-parent families and disabled. But also in addition, it built two potable “Wallace fountains” (a special public fountain with potable water) in Macau, providing free potable water to tourists and citizens. MWSL has also been active in various social welfare and charity activities, providing total donations of 2.08 million MOP (1MOP = 0.965RMB, 2007; ca. 0.26 million US$ at the exchange rate of 1US$ = 8.276RMB) during 2002–2005. During 1985–2005, MSWL also charged discounted water tariffs for governmental agencies, and handed in over 260 million MOP (1MOP = 0.965RMB, 2007; ca. 32.56 million US$ at the exchange rate of 1US$ = 8.276RMB) of taxes and about 56 million MOP (1MOP = 0.965RMB, 2007; ca. 7.012 million US$ at the exchange rate of 1US$ = 8.276RMB) of concession fees to the government. In both Maanshan and Macau, the water tariff is the main financial source for water companies, while governmental subsidies have been abandoned. Accordingly, whether the water tariff can cover the costs is significant. In the case of Macau, the Macau Government owns the pre-concession infrastructure assets, which demands a smaller first investment from the Consortium. The water tariff could easily cover the cost of operation and maintenance (and not the huge capital costs of existing assets). Unlike the joint venture construction in the Maanshan case, Macau Government leaves all financial responsibilities to the private sector after the concession, and benefits from taxes, concession fees and discounts on government water bills (see Fig. 10). Due to the limited initial investments of the private consortium, sharp water tariff increases were avoided after privatization (often one of the major reasons for public resistance and failed private sector participation in other countries). The local government still owns part of the infrastructure assets, in particular the pipes system, with huge sunk-costs. Fig. 10Monetary flow within Macau water supply Macau is also an interesting case because of the unique regulatory system, which includes the water quality regulator (IACM), and a unique Government Delegate. IACM is in charge of the water quality regulation, and monitors and controls drinking water quality by random sampling and analysis of over 70 water samples around Macau everyday. The Government Delegate is not a government official, but an individual working in another public utility company and appointed by the government. Following Macau laws, Mr. Lin Runzhong, the Government Delegate for water supply, was appointed for a period of five years by the Macau Government, and is not only the regulator of MWSL, but also an important linkage between MWSL and the government. He participates at all MWSL board meetings and reports relevant information and documents to the government. The Government Delegate decides which information is considered relevant. He is also in charge of assessing the performance of MWSL, and comments on the five-year plans and tariff plans before MWSL sends these to the government for approval. The Macau government generally follows the comments and assessments of the Government Delegate. In this sense, the nongovernmental Government Delegate is defined a specified role and powerful position in governing the water sector. This institutional arrangement relates to the small size of Macau Government, where only a limited state capacity (in quantitative and qualitative terms) is available for numerous public tasks. In conclusion, it can be argued that after 1985 the Macau government has played a meager role in the drinking water management. Greenfield Contract: Shanghai Wastewater The Greenfield contract (e.g., BOT, TOT) is the dominant form of private sector participation in wastewater sector reform throughout the country. Shanghai Zhuyuan No.1 WWTP project is one of the most famous Greenfield projects in China. It is presently one of the largest WWTP in China, with a treatment capacity of 1.7 million m3 per day and an advanced primary treatment, serving an area of 107 km2 and about 23.5 million inhabitants. But it also has become famous for the lowest service price: 0.22 RMB (ca. 0.0266US$ at the exchange rate of 1US$ = 8.276RMB) per cubic meter treated wastewater. In 2002, the Youlian Consortium (consisting of Youlian Development Company with 45% shares, Huajin Information Investment Ltd. Company with 40% shares, and Shanghai Urban Construction Group with 15% shares) won the open tender for Zhuyuan No.1 WWTP project by bidding the lowest treatment costs. A Project Company (Shanghai Zhuyuan Youlian No.1 Wastewater Treatment Ltd. CO.) was established and awarded a 20-year concession agreement by Shanghai Water Authority. A service management contract was signed with Shanghai Sewerage Company (a fully state-owned company administrated by the government) including details of rights and obligations. Two years later, Youlian Development Company withdrew from this project by transferring the shares and obligations to InterChina Holdings Group (see Fig. 11). Fig. 11Private sector involvement in Shanghai wastewater treatment According to the agreement between Shanghai Water Authority and the private company, Shanghai Water Authority should minimize its interventions in the construction, operation, and maintenance of WWTP and limit them to safeguarding public health and safety. All conditions and objectives with regard to water service quality are defined in the service contract between Shanghai Sewerage Company and the private company. Among others, the private company has to install an on-line monitoring system and is requested to invite an authorized third party for regular monitoring (on indicators such as BOD5, CODcr, SS, NH4-N, and phosphate). This should be paid by the private company, while reporting to the Shanghai Sewerage Company and should take place within five days. Shanghai Sewerage Company may conduct random water examination at any time. According to the local officials, Shanghai Zhuyuan WWTP has fulfilled all responsibilities and obligations required by the contract up till now, including meeting the water quality standards. In the case of Shanghai Zhuyuan Greenfield project, the government has transferred its traditional responsibilities of investment, construction, operation, and maintenance (for the contract period) to the private Project Company, accompanied by paying a service fee (see Fig. 12). Different from the joint venture construction in Maanshan and the concession construction in Macau, in which corporate profits directly depend on the water tariff, the private operator within a Greenfield contract is paid a service price negotiated between the government and the private sector. This service price depends on the investments and agreed performance levels, rather than on the user fee level, and which provides the private sector with the financial risks. Accordingly, the low service price of Zhuyuan No.1 WWTP (which was 42% less than the projected costs by government) presented in the public bidding, was argued to have a close relation to earlier governmental input in this project. Shanghai Water Assets Management Development CO. Ltd., a fully public-owned company, was in charge of the pre-phase design and invested about 30 million US dollars in the fixed infrastructure of this project, while the government provided the land free of charge to the operator. Strictly speaking, Shanghai Zhuyuan No.1 WWTP Greenfield project is a quasi-BOT project, due to the fact that part of the investment comes from the government. Fig. 12Monetary flow within Zhuyuan Greenfield project The experience of Shanghai is an example of full governmental delegation of the daily management of WWTP to the private sector, while financial support via subsidies and preferential policies (e.g., land use) facilitate privatization with low service prices. It is, however, too early to fully assess the success of this project. Some BOT WWTP projects in other cities have met problems following gaps in the current national policy documents. For instance, projects in Foshan (Guangdong Province) could not run properly due to conflicts over current land use right. And projects in Beijing were delayed during the financing process because the domestic private actors met difficulties in obtaining loans from domestic banks due to the lack of a sound loan policy. The commercial banks couldn’t provide long-term loans as required for BOT-types projects as their credit policies are restricted for the private sector (Zhong and Fu 2005), while the China Development Bank can provide long-term loans for BOT-types projects only for a limited number of clients (Chang and others 2006). Conclusions With the emergence and blossoming of various forms of private sector involvement in the Chinese water sector, the traditional structure of full governmental provision of water supply and wastewater treatment has changed dramatically. The analysis in this article has provided evidence of the contribution of these new modes to increased capital investment, and especially of more efficient operations and improved service provision. In that sense, the original goals of the Chinese government to embark upon private sector involvement in water provisioning and treatment have been met. However, the early stage that most contracts are in, and the not yet crystallized forms and modes of privatization, prevents us from drawing any final conclusions on the impact of private sector involvement in the Chinese water sector. From the three casestudy projects with private sector participation, we can draw some lessons for how to successfully involve the private sector into the provision of water services. Firstly, a balance between the water tariff level, profits of investor and governmental subsidies is required. As Hall and Lobina (2005) state, most practices of water privatization fail due to public resistance following sharp price increases and job losses. In China, this has not (yet) been the case, due to large increases in efficiencies and governmental support to fixed infrastructure assets, reducing financial risk of the private sector and limiting the need for large water tariff increases. At the same time, the significant economic growth levels enables local residents to cope with some tariff increases, the poor and disadvantaged have been subsidized by the government, job losses have been minimized following social policies, and public hearings have contributed to higher levels of legitimacy. This all contributed strongly to a relatively smooth transformation of China’s water sector. Secondly, the selection of the PPP form has a close relation with the level of local water tariff. As illustrated by this article, Greenfield projects appear to be applied when tariffs are not sufficient, especially in the wastewater sector (see also Zhong and others 2006), while Joint Venture approaches are often used in cities with sufficiently high water tariff, in particularly in the water supply sector. Thirdly, it is crucial to accelerate the establishment of systematic and comprehensive governmental regulatory framework, as the current ad hoc, fragmented and diverse regulatory system endangers efficiency in water service development and certainty and stability for foreign investors. Experiences in many countries have proven that regulation is a key aspect in successful privatization in the water sector and a competitive benchmarking system is regarded as useful in an effective regulatory approach. In late 2006, the MOC attempted to develop a Chinese water supply benchmarking system, which is still ongoing. However, the current private sector involvements in the Chinese water sector still face many legal and regulatory uncertainties. Too often local authorities experiment with systems of governmental regulation and control, or — as in Macau — seem to become marginalized. According to interviews with local officials during our fieldwork, the importance of establishing a workable regulatory and legal system is essential. Guaranteeing sufficient and safe water service to the public is jeopardized by the fact that governments can no longer fully control the planning, operation, and management of water services as before private sector participation. This might only be signs of uneasiness with the new water institutions and division of tasks and responsibilities, but can also be the heralds of an emerging debate on privatization in the Chinese water sector. Finally, but not least, it is important to identify the differences in risk allocations in the water (service) market between the public and private sectors within different modes of PPP. As Table 1 and the three case-study projects illuminate, with the various forms of privatization, the government often transfers (smaller or larger parts of) financial risks, building risks, and operation and maintenance risks to the private sector. Meanwhile, in the end the government can always take over all facilities without paying an indemnity to the private sector if a concessionaire fails in obtaining the goals as formulated by governmental authorities, or some conflicts emerge in the further policies (e.g., the terminated contracts that are regarded as providing the private sector a fixed investment return). In that sense, the still unstable legal base in transitional China provides a major political and transfer risk for private investors.

Pain-2-1-2258319

How does the self-reported clinical management of patients with low back pain relate to the attitudes and beliefs of health care practitioners? A survey of UK general practitioners and physiotherapists

Guidelines for the management of low back pain (LBP) have existed for many years, but adherence to these by health care practitioners (HCPs) remains suboptimal. The aim of this study was to measure the attitudes, beliefs and reported clinical behaviour of UK physiotherapists (PTs) and general practitioners (GPs) about LBP and to explore the associations between these. A cross-sectional postal survey of GPs (n = 2000) and PTs (n = 2000) was conducted that included the Pain Attitudes and Beliefs Scale (PABT.PT), and a vignette of a patient with non-specific LBP (NSLBP) with questions asking about recommendations for work, activity and bedrest. Data from 1022 respondents (442 GPs and 580 PTs) who had recently treated patients with LBP were analysed. Although the majority of HCPs reported providing advice for the vignette patient that was broadly in line with guideline recommendations, 28% reported they would advise this patient to remain off work. Work advice was significantly related to the PABS.PT scores with higher biomedical (F1,986 = 77.5, p < 0.0001) and lower behavioural (F1,981 = 31.9, p < 0.001) scores associated with advice to remain off work. We have demonstrated that the attitudes and reported practice behaviour of UK GPs and PTs for patients with NSLBP are diverse. Many HCPs held the belief that LBP necessitates some avoidance of activities and work. The attitudes and beliefs of these HCPs were associated with their self-reported clinical behaviour regarding advice about work. Future studies need to investigate whether approaches aimed at modifying these HCP factors can lead to improved patient outcomes. 1 Introduction Low back pain (LBP) is common, affecting 38% of adults in any one year, of whom 1 in 4 experience significant disability [37]. Only 25% of patients consulting in primary care will be symptom free 12 months later [18]. The last two decades have also seen dramatic rises in work loss and sickness benefit payments, attributed to recurrent and persistent LBP [16,36]. Guidelines for the clinical management of patients with LBP encourage health care practitioners (HCPs) to advise patients to stay active, avoid bed rest, stay at or return to work, and stress simple messages about self-management [3,31,45,47,49,50]. Previous studies have identified that HCPs do not always follow guideline recommendations for LBP [10,20,24,26] and so despite the abundance of guidelines for practice, the management of LBP poses considerable challenges and frustrations for both patients and practitioners [14] and it is increasingly clear that it is insufficient to study patient factors alone [25,48]. A potentially important but relatively unexplored influence on patients’ pain experiences is the attitudes and beliefs of the HCPs with whom they come into contact. HCPs are frequently asked to provide advice and recommendations about physical activities, work, and rest and HCPs’ attitudes and beliefs may be an integral part of the health care process, influencing the success or failure of treatment. HCPs hold a range of attitudes and beliefs about back pain [17,19,28,32,39,42–44], and these attitudes appear to be associated with the work and activity recommendations that HCPs give to patients [17,28,42,44]. In the UK, approximately 98% of the population is registered with a National Health Service general practitioner (GP) [13]. GPs serve as gatekeepers to secondary care, selecting and referring patients for specialist investigations and treatment services. Physiotherapy is one of the most common services to which patients are referred, or which patients seek out privately [36], and LBP accounts for more than half of physiotherapists’ workload in the UK [24]. Few studies have explored HCP factors in the UK, but it has been shown that many physiotherapists (PTs) continue to advise limitations of work and activity levels, despite identifying when patients with LBP are at risk of chronicity [11] and an important proportion of therapists continue treating patients with LBP even when they fail to improve [41]. The aim of this study was to measure, in national random samples, the attitudes, beliefs and reported clinical behaviour of GPs and PTs about LBP, explore their associations and evaluate the implications for both clinical practice and future research. 2 Methods 2.1 Design and setting We conducted a cross-sectional, nationwide postal survey of UK GPs and PTs, involved in the management of patients with LBP, between April and November 2005. Ethical approval for the study was obtained from the West Midlands Multi-centre Research Ethics Committee (MREC). Written consent was not sought from each participant for use of survey data, but consent of respondents was assumed if they completed and returned the questionnaire. 2.2 Questionnaire sample and mailing process We used simple random sampling to obtain details of GPs (n = 2000) and PTs (n = 2000) from national databases (Binleys database for GPs, n = 46,000 GPs on the list; Chartered Society of Physiotherapy membership database, n = 32,000 PTs on the list). In the UK, all GPs working in the National Health Service are included on the Binleys database [7], which is produced in conjunction with the Royal College of General Practitioners. The Chartered Society of Physiotherapy (CSP) is the professional, educational and trade union body representing the UK’s chartered physiotherapists and 98% of all PTs are members of the CSP. A sample size calculation indicated that a sample of 900 responders (450 GPs and 450 PTs) was required to allow us to find a minimum difference of 10% in the proportion of respondents with ‘helpful’ to ‘unhelpful’ beliefs by important practitioner characteristics at a significance level of 0.05 and a power of 90% [2]. A questionnaire package containing the questionnaire, a cover letter, an information sheet and a pre-paid envelope was mailed to each HCP. A single reminder was sent to all non-responders four weeks after the first mailing. In order to allow assessment of non-response bias within the survey estimates, a brief questionnaire was mailed to a random sample of non-responders. No incentives for completing the questionnaire were offered. 2.3 Questionnaire A filter question was used to identify those HCPs who had treated at least one patient with non-specific LBP (NSLBP) in the previous six months, so that only respondents with recent experience of managing patients with LBP were included in the analysis. 2.3.1 Demographics and practice information A number of demographic and practice questions, relevant to each profession, were included. Some items were pertinent to both professions: gender; years since qualification; postgraduate training in LBP; clinical interests/speciality and personal experience of back pain. Data gathered exclusively from GPs included whether they worked only in general practice and whether the practice was a single-handed or a group practice. Data gathered exclusively from PTs included how much of their clinical practice was based in the NHS, what proportion of their caseload was primary care patients, whether they worked alone or in a team, and grade of current job. 2.3.2 Attitudes and beliefs measure The Pain Attitudes and Beliefs Scale (PABS.PT [28,39]) was included as a measure of HCPs’ attitudes about LBP. This was selected following a systematic review of available tools for assessing the attitudes and beliefs of HCPs about LBP [12], in which the PABS.PT fared well on pre-defined quality criteria [34]. This tool was originally developed for use in physiotherapists, but more recently has been applied to a cohort of Dutch general practitioners [30]. In addition, the members of a multi-disciplinary clinical advisory group confirmed face and content validity of the PABS.PT for both GPs and PTs after recommending that the term ‘therapy’ was changed to ‘treatment’ in two of the items of the PABS.PT. The resulting minimally amended PABS.PT was used for both GPs and PTs. The PABS.PT assesses the strength of treatment orientation on two subscales, ‘biomedical’ and ‘behavioural’. The biomedical orientation is described as one in which the HCP believes in a biomechanical model of disease, where disability and pain are a consequence of a specific pathology within the spinal tissues and treatment is aimed at treating the pathology and alleviating the pain. The behavioural orientation is where the HCP believes in a biopsychosocial model of disease in which pain does not have to be a consequence of tissue damage, and can be influenced by social and psychological factors. We used the amended PABS.PT [28], which consists of 19 items, each rated on a six point Likert scale (‘Totally disagree’ = 1 to ‘Totally agree’ = 6), with ten items on the biomedical subscale (score range: 10–60) and nine on the behavioural subscale (score range: 9–54). Higher scores on each subscale indicate a stronger biomedical or behavioural treatment orientation, respectively. 2.3.3 Clinical behaviour measures Clinical behaviour was elicited by asking the HCPs about diagnostic investigations and for their recommendations about work, activity levels, and bedrest, for a patient with NSLBP described in a vignette. The vignette described a patient with uncomplicated NSLBP who was not at work as a result of their symptoms (Appendix A). Vignettes have been shown to be a useful measure of clinicians’ practice behaviour and a more accurate assessment of clinical behaviour than data extracted from case notes when measured against the gold standard of standardised patients [40]. The clinical behaviour question regarding work was as follows: “The patient described in the vignette asks what your advice would be about her work. I would recommend this patient to: (Please tick the one response that best describes what you would recommend this patient to do)a.Be off work until pain has completely disappearedb.Return to part time or light dutiesc.Be off work for a further … weeks (please state number of weeks)d.Return to normal worke.Be off work until pain has improved” Responses for each of the work, activity and bedrest questions were subsequently classified by the authors as being ‘strictly in line with guideline recommendations’, ‘broadly in line with guideline recommendations’ and ‘not in line with guidelines’. For the work question given above, we considered option ‘d’ to be strictly in line with guideline recommendations, option ‘b’ to be broadly in line with guideline recommendations and options ‘a’, ‘c’ and ‘e’ to be not in line with guideline recommendations. This classification was based on a previously published expert consensus carried out on similar practice recommendations in a postal survey of physiotherapists, osteopaths and chiropractors in the UK [22]. 2.4 Brief questionnaire The brief questionnaire sent to a sample of non-responders contained the filter question to ensure that respondents recently involved in the management of patients with LBP could be identified. Alongside key demographic questions, we included four items from the PABS.PT (two from each subscale chosen on the basis of factor loadings described by the tool’s developers and data from a pilot study), the vignette patient and the clinical behaviour questions related to work, activity and bedrest. 2.5 Statistical analysis Scores for the PABS.PT were calculated according to methods specified by the questionnaire developers, i.e. a simple summation of the items in each subscale [39]. No method for dealing with missing data on this measure has been published so a pragmatic decision was made that if one value was missing from a subscale, a mean score based on the remaining values was substituted. If more than one value was missing the score for the whole subscale was classed as missing. A Pearson’s correlation coefficient was calculated between the scores on the two subscales of the PABS.PT as previous work has shown that they are not totally independent [28,39]. We used descriptive statistics to summarise, by professional group, demographic, and practice data for both subscales of the PABS.PT. In addition, in response to the reviewer’s suggestions, we conducted a subgroup analysis of work, activity and bedrest recommendations for those respondents who had high biomedical scores and low behavioural scores and vice versa. Unless differences occurred by profession, analyses were performed on the combined GP and PT dataset. The relationship between attitudes and beliefs and clinical behaviour was examined using ANOVA to test for an overall relationship with clinical behaviour and, when appropriate, for a linear trend across clinical behaviour groups (strictly in line, broadly in line and not in line with guidelines). The effect of non-response was examined by comparing responses from all responders to the full questionnaire to those completing the brief questionnaire. All analyses were carried out using the Statistical Package for Social Scientists for Windows (SPSS Inc., Chicago, IL, version 13). 3 Results The overall response rate was 38% (n = 1534), 22% (n = 443) for GPs and 55% (n = 1091) for PTs. Of the respondents, 580 PTs and 442 GPs reported treating at least one patient with LBP in the previous six months and were included in the analysis. 3.1 Characteristics of respondents The demographic and professional characteristics of the respondents are summarised in Table 1. The majority of GPs worked exclusively in general practice, within group practices and had at least one specialist clinical interest. The majority of PTs worked within the NHS, with other HCPs, were of senior clinical grade or above, and had a patient caseload of more than 50% primary care patients. The PTs were qualified for a shorter length of time than GPs, were more likely to be female and to have postgraduate training in LBP. 3.2 Attitudes and beliefs Scores for both of the PABS.PT subscales could be calculated for the majority of the 1022 responders (biomedical n = 1010, behavioural n = 1004). Mean (standard deviation, range) score for the biomedical subscale was 31.0 (6.4, 12–50) overall: GPs 30.9 (5.3); PTs 31.1 (7.2), and for the behavioural subscale was 33.0 (4.6, 15–48) overall: GPs 33.7 (4.2); PTs 32.5 (4.8). For both subscales and both professional groups, the mean observed scores were in the middle of the possible ranges. The Pearson’s correlation coefficient (r = −0.38; p < 0.0001) showed a statistically significant level of dependence between the two subscales, suggesting that respondents who score higher on one subscale tend to score lower on the other subscale. 3.3 Diagnostic investigations In response to the vignette patient, most HCPs reported that they would not want the patient described to have any diagnostic investigations. Of the GPs 33% (n = 142) reported that they would request at least one investigation, compared with 24% (n = 134) of PTs (Table 2). GPs were more likely to want laboratory tests and PTs were more likely to want an X-ray or special imaging procedure such as an MRI. 3.4 Clinical behaviour The responses to the clinical behaviour questions were classified according to whether these were ‘strictly in line’, ‘broadly in line’ or ‘not in line’ with guideline recommendations and the responses and the classifications are summarised in Table 3. The majority of respondents reported advice that was either ‘strictly in line’ or ‘broadly in line’ with guideline recommendations’. Very small proportions of respondents reported they would provide advice that was ‘not in line’ with guideline recommendations for activity and bedrest, however, this figure was considerably higher for recommendations regarding work, with 28% of respondents reporting that they would recommend the patient in the vignette to remain off work. The summary of responses of the two subgroups of high biomedical and low behavioural scores (n = 187) and low biomedical and high behavioural scores (n = 137), compared to the total sample, is also presented in Table 3. The proportion of practitioners recommending that the patient in the vignette remain off work, i.e. not in line with guideline recommendations, was substantially higher in those with high biomedical and low behavioural scores (44.9%) than those with high behavioural and low biomedical scores (11.9%). Similar differences were also seen for recommendations regarding activity and bedrest. 3.5 Relationship between attitudes and beliefs and clinical behaviour Given the very small proportion of respondents whose advice was ‘not in line with guidelines’ for both activity and bedrest, associations with the PABS.PT scores were not examined. Fig. 1 shows the distributions of the PABS.PT biomedical and behavioural subscale scores for each of the reported work recommendation groups. With increasing disparity with guidelines, biomedical scores increased (mean scores: 28.3, 30.6, 33.5) and behavioural scores decreased (mean scores: 34.1, 33.3, 31.8). These associations were shown to have a significant linear trend for both the biomedical (F1,986 = 77.5, p < 0.001) and behavioural (F1,981 = 31.9, p < 0.001) subscale scores. 3.6 Effect of non-response In order to assess the impact of non-response bias within the survey estimates, a brief questionnaire was mailed to a random sample of non-responders (GPs n = 414, PTs n = 243), and responses were received from 14% of GPs (n = 59) and 17% PTs (n = 40). For the GPs, gender mix and years in practice were similar for those completing the full and brief questionnaire. For the PTs, those completing the brief questionnaire were slightly less experienced (mean of 12 years experience versus 15 years) and more likely to be male compared to the full questionnaire responders (25% vs. 19% male). Responses to both behavioural subscale PABS.PT items and one of the two items from the biomedical subscale were similar to those for the full questionnaire. Responders to the brief questionnaire, from both professions, were more likely to agree with the statement that ‘patients with back pain should preferably practice only pain free movements’, indicating a more biomedical orientation. The responses to the items regarding work and activity advice were similar for responders to the full and brief questionnaires. GPs responding to the brief questionnaire reported bedrest advice that was less in line with guideline recommendations than the responders to the full questionnaire (19.3% strictly in line with guidelines compared to 38.4%, respectively), whereas the PTs completing the brief questionnaire reported bedrest advice that was more in line with guideline recommendations than the initial responders (35.0% strictly in line with guidelines compared to 21.8%, respectively). 4 Discussion 4.1 Main findings This is the first national UK survey of LBP related attitudes, beliefs and reported clinical behaviour of GPs and PTs and results show that responses are diverse. The majority of respondents reported advice that was strictly or broadly in line with guideline recommendations about activity and bedrest, however, over a quarter of HCPs recommended that the vignette patient with NSLBP should remain off work. Reasons why adherence to guideline recommendations for work is lower than for activity and bedrest are unclear, but may be due to the complex nature of the clinical consultation, and previous studies have shown that GPs see sickness certification as a potential threat to the doctor–patient relationship [15,29]. The attitudes and beliefs of HCPs were significantly associated with reported work advice for the patient described, i.e. HCPs with stronger biomedical and weaker behavioural treatment orientations were more likely to report advice, regarding work, which was ‘not in line with clinical guidelines’. The subgroup analysis supports this, although only a third of respondents could be categorized into these subgroups i.e. high biomedical and low behavioural scores on the PABS.PT or vice versa. The differences in the PABS.PT scores were small, and although statistically significant, no guidance is currently available to suggest whether these represent a clinically relevant difference. A considerable proportion of HCPs in the UK continue to provide advice to patients about work that is not in line with guideline recommendations. The associations between attitudes, beliefs and reported clinical behaviour suggest that some HCPs continue to practice predominantly within a biomedical model, placing most importance on the severity of tissue damage when determining a patient’s level of pain and functional disability. Others have adopted a more behavioural approach to management, embracing the notion that the level of pain and functional loss may be influenced by psychological and social factors in addition to biomechanical factors. 4.2 Comparison to other studies HCPs in this study had similar attitudes and beliefs to therapists in the Netherlands [28], with Dutch therapists having similar mean biomedical scores (29.5 vs. 31.0), but slightly higher behavioural scores (35.6 vs. 33.0) on the PABS.PT. Direct comparison of subscale scores with studies using the original PABS.PT is not possible due to a different number of items [30,39]. The attitudes and beliefs of HCPs were significantly associated with reported work advice for the vignette patient. Respondents reporting advice ‘strictly in line with guidelines’ demonstrated stronger behavioural and weaker biomedical orientations than those reporting advice ‘not in line with guideline recommendations’. Using a variety of measures, previous studies have demonstrated that advice to restrict work or activities is also associated with a biomedical treatment orientation [28], patho-anatomical focus of training courses [39], higher fear avoidance beliefs of HCPs [17,32,42] and a strong belief that pain and impairment are invariably linked [28,44]. Our study adds to this body of literature by showing a significant association between attitudes and beliefs and reported work advice in HCPs in the UK. 4.3 Implications for clinical practice and future research The results suggest that the attitudes and beliefs of HCPs are linked to clinical practice and the recommendations provided to patients. These practitioner factors are thus part of the dynamic interaction within LBP care episodes, along with the LBP problem itself and the patient’s own perceptions about their problem. This may help explain patient outcomes, although the mechanisms behind this are likely to be complex. It is probable that HCPs’ attitudes and beliefs are expressed to patients in a variety of ways, with a range of possible consequences. By restricting activities and work, HCPs may reinforce patient’s unhelpful illness perceptions and increase spinal vigilance. Alternatively, they may over-direct the patient by providing strict advice to perform only specific activities and exclude others, encouraging an over-reliance on the HCP [35], which may make it difficult to foster the patients’ self-management skills, something recommended as part of best practice for patients with LBP. The reported clinical behaviour of HCPs illustrates that the majority would provide advice that is strictly or broadly in line with guideline recommendations, however, nearly 30% reported they would advise the described patient to remain off work. Staying at work or an early return to work with NSLBP is recommended [50], as the longer someone is off work the likelihood of them returning steadily diminishes, with a 20% risk of long term disability for those off work for four to six weeks [51]. Although the management of LBP, in terms of advice about activity and bedrest, seems to be broadly in line with guideline recommendations, our results show that adherence about advising early return to work is suboptimal. Attitudes and beliefs held by HCPs may help explain why implementation of current LBP guidelines has been slow and difficult [6,8,20,23,33]. Changing clinical behaviour is recognised to be a challenge [27]. Evidence from recent clinical trials suggests that although modest intervention strategies can result in moderate changes in reported adherence to guideline recommendations [8], this does not lead to a corresponding improvement in patient outcomes [9,21,30]. A better understanding of the attitudes and beliefs of HCPs, what influences these and how these relate to outcomes of patients with LBP is needed to inform development of future implementation strategies. Future work should further test the psychometric properties of the PABS.PT to assess responsiveness and determine appropriate cut offs for ‘high’ and ‘low’ scores on the subscales and what constitutes a clinically relevant change. Methods to assess HCP attitudes, beliefs and behaviours warrant further study. For example, the validity of using methods to measure implicit attitudes about LBP, such as those employing automatic responses, could be explored in an attempt to overcome potential social desirability bias in survey responses as HCPs become more aware of clinical guidelines. 4.4 Strengths and limitations The strengths of this study include the large sample sizes, simple random sampling of UK GPs and PTs, use of a validated beliefs measure, and investigation of potential non-response bias. The response rate of GPs was low, but comparable to other postal surveys of GPs in the UK [4,5,38]. The sample size calculation took this into account and yielded the required sample size for the planned analyses. The response rate of PTs was in keeping with other studies [11,39,41]. Responses to the brief questionnaire were broadly similar to those completing the full questionnaire in terms of attitudes, and recommendations for work and activity. However, some differences in the advice for bedrest suggest that we cannot rule out non-response bias in our survey. Responses to one PABS.PT item showed a stronger biomedical treatment orientation for responders to the brief questionnaire. Also, GPs responding to the brief questionnaire reported advice for bedrest that was less in line with guideline recommendations than responders to the full questionnaire, so for GPs, where the potential for non-response bias is greatest, our survey may underestimate the strength of a biomedical treatment orientation and the numbers providing advice not in line with guideline recommendations. This study captured self-reported behaviour rather than real clinical practice, which is very difficult to measure. To provide a context for the clinical behaviour questions we used a vignette of a patient with NSLBP, an approach shown previously to have acceptable validity [40,46]. Although we used established tools to assess attitudes, beliefs and clinical behaviours, there may be some overlap in the constructs they measure. We attempted to address this by the wording of instructions and the order of the tools within the questionnaire. The PABS.PT attitudes measure came first with instructions to respond to the general attitudinal type statements. The vignette and the behaviour questions came later with the instruction to consider the specific management of the patient described. 5 Conclusion This study shows the diversity of the attitudes and self-reported practice behaviour of UK GPs and PTs for patients with NSLBP. Many HCPs believed LBP necessitates some avoidance of activities and the need to be off work. For a patient with a history of being off work since onset of LBP four weeks previously, over a quarter of HCPs recommended further time off work. The attitudes and beliefs of HCPs were associated with their advice about return to work. Future studies need to investigate the associations between HCP factors and patient outcomes, and test if approaches aimed at modifying attitudes, beliefs and clinical behaviours of HCPs can be successful.

J_Immunol_Methods-2-1-2225449

Quantitative imaging assay for NF-κB nuclear translocation in primary human macrophages

Quantitative measurement of NF-κB nuclear translocation is an important research tool in cellular immunology. Established methodologies have a number of limitations, such as poor sensitivity, high cost or dependence on cell lines. Novel imaging methods to measure nuclear translocation of transcriptionally active components of NF-κB are being used but are also partly limited by the need for specialist imaging equipment or image analysis software. Herein we present a method for quantitative detection of NF-κB rel A nuclear translocation, using immunofluorescence microscopy and the public domain image analysis software ImageJ that can be easily adopted for cellular immunology research without the need for specialist image analysis expertise and at low cost. The method presented here is validated by demonstrating the time course and dose response of NF-κB nuclear translocation in primary human macrophages stimulated with LPS, and by comparison with a commercial NF-κB activation reporter cell line. 1 Introduction Nuclear factor kappa B (NF-κB)/rel represent a family of transcription factors, present in all eukaryotic cells, that regulate inducible expression of wide ranging genes involved in immune responses and cell-cycle regulation. Therefore activation of NF-κB provides a key molecular switch that is relevant to many aspects of cellular immunology research. In immune cells NF-κB is most abundant either as a heteromeric complex of two components, p65 (rel A) and p50, or as a p65/p65 homodimer. The p65 component contains the main transactivating domain responsible for NF-κB transcription factor function. Regulation of NF-κB activity is dependent upon cytoplasmic sequestration in association with an inhibitory molecule, IκBα. As a consequence of intracellular kinase signalling cascades IκBα is phosphorylated, and this leads to its degradation, allowing nuclear translocation of p65/rel A and hence so-called “activation” of NF-κB (Ghosh et al., 1998). Conventional methods for testing NF-κB nuclear translocation utilise a semi-quantitative electromobililty gel-shift assay. This involves incubation of nuclear extracts with 32P-labelled oligonucleotides of NF-κB binding sites and separation from unbound probe by electrophoresis in a non-denaturing polyacrylamide gel. However, this assay is principally limited by sensitivity, and requires large scale cell culture (typically > 10 × 106 cells), thus precluding its use with primary cells. Additional disadvantages are the labour-intensive protocol and the need for radioisotopes. Cell lines transfected with NF-κB promoter–reporter gene constructs are used widely to study cell signalling pathways, and are easy to quantitate with colorimetry or luminometry, but are not suitable for most primary cell culture studies. Commercial kits for enzyme linked immunosorbent assays of NF-κB quantitation within nuclear extracts have also become available, but their expense may be prohibitive. An attractive new strategy is to image NF-κB translocation from the cytoplasm to the nucleus, using immunofluorescence staining. This can be performed on a small scale and at a single-cell level. Fluorescence microscopy, laser scanning cytometry and flow cytometry to quantify NF-κB nuclear translocation have all been reported (Deptala et al., 1998; George et al., 2006; Rogers and Fuseler, 2007; Fuseler et al., 2006). Drawbacks of these previous reports are the need for expensive imaging equipment and/or image analysis software. Here we present an alternative simple method for quantitative detection of NF-κB rel A nuclear translocation which uses standard confocal immunofluorescence microscopy and the public domain Java image processing program, ImageJ. We suggest that the general availability of all three components—fluorescence microscopy, immunostaining reagents, and the analytic protocol-provides a readily accessible method for the study of NF-κB nuclear translocation in primary cell cultures. For this study we have chosen to use a cell system that has been difficult to analyse previously, primary monocyte-derived macrophages that have been isolated from healthy human volunteers. In vivo macrophages are resident tissue mononuclear phagocytic cells derived from circulating monocytes. They function both as sensory cells of innate immunity, and as effectors, initiating early non-specific host defences, both by local recruitment of other immune cells and by induction of an acute phase response, which leads to systemic priming of the immune system. In vitro work on these cells often includes stimulation by model innate stimuli, now known to be Toll-like receptor (TLR) ligands, such as lipopolysaccharide (LPS) (TLR-4) and Pam3CSK4 (TLR-2). Activation of the NF-κB pathway is a common downstream component of the cellular response to many different innate immune stimuli and is used frequently in these cells as a biochemical detection and quantification method to study innate immune cellular activation. Therefore analysis of the NF-κB pathway in macrophages, and in related mononuclear phagocytic cells such as dendritic cells, is an important area of research interest. However, analysis in these cells has been hampered by the lack of suitable methods. Myeloid leukaemic cell lines (e.g. U937, THP-1) are the standard for these experiments, but unless they are very carefully differentiated (which itself may involve NF-κB activation) they are not ideal models. This necessity for an accurate and reproducible primary macrophage system stimulated us to adapt the NF-κB assay as described here. 2 Methods 2.1 Macrophage culture and innate immune stimulation Human blood samples were obtained from healthy volunteers. The study was approved by the joint University College London/University College London Hospitals NHS Trust Human Research Ethics Committee and written informed consent was obtained from all participants. Peripheral blood mononuclear cells (PBMC), consisting of monocytes and lymphocytes, were prepared by density-gradient centrifugation of heparinised blood with Lymphoprep™ (Axis-Shield) according to the manufacturer's instructions. PBMC were resuspended in RPMI 1640 with l-glutamine (Gibco, Invitrogen) (NM) containing 5% heat inactivated (56 °C for 30 min) type AB normal human serum (NHS) (Sigma Aldrich) (107 cells /ml) and seeded on to 13 mm (No 1.5) glass coverslips (VWR) using 2 × 105 cells/cover slip. After 1 h at 37 °C non-adherent cells (lymphocytes) were removed by sequential washes with Hanks buffered saline solution (HBSS) (Gibco, Invitrogen). Adherent cells (monocytes) were incubated in NM containing 10% autologous heat-inactivated HS supplemented with 20 ng/mL macrophage-colony stimulating factor (M-CSF) (R&D systems) for 3 days. Any remaining non-adherent cells were removed by further washes with HBSS, and NM/HS was replaced, but without additional M-CSF. This protocol yields adherent macrophages by day six with less than 5% contamination by lymphocytes (data not shown). Ultra-pure LPS (Invivogen) and Pam3CSK4 (Invivogen) were used as prototypic innate immune stimuli at different concentrations as outlined below. Polymyxin B (Invivogen) was used as a specific inhibitor of LPS bioactivity. 2.2 Immunofluorescence staining For immunostaining, rabbit polyclonal affinity purified antibody to rel A (C-20) (Santa Cruz Biotechnology) was used (2 μg/ml) with a secondary antibody, Alexa-Fluor (AF)633—conjugated F(ab')2 goat anti-rabbit IgG (Invitrogen) used at 4 μg/ml. 10% normal goat serum (Sigma Aldrich) used as blocking buffer, eliminated all non-specific binding of the secondary antibody (data not shown). Cells cultured on glass coverslips were fixed with 3.7% paraformaldehyde (15 min, room temperature) and washed with Tris-buffered saline (TBS). Each coverslip was inverted on to 50 μl of solution placed on impermeable Nesco film (VWR). All reagents were diluted in TBS, and coverslips were washed by immersion into TBS between each staining step. Coverslips were incubated sequentially with 0.2% Triton-X100 (Sigma) (10 min, room temperature), blocking buffer (30 min, room temperature), primary antibody diluted in blocking buffer (overnight, 4 °C) and secondary antibody diluted in blocking buffer (1 h, room temperature). Nuclei were counterstained with 2 μg/ml of the nuclear stain DAPI (Sigma Aldrich) for 5 min. Coverslips were mounted on to glass slides (VWR) using Vectashield hard-set mounting media (Vector). 2.3 Image acquisition and analysis Fluorescence images were captured on a Leica SP2 confocal microscope. DAPI (excitation 405 nm, emission 400–450 nm) and AF633 (excitation 633 nm, emission 650–700 nm) fluorescence were captured using sequential acquisition to give separate image files for each (Fig. 1A). A pin hole of 1 Airy (114.5 μm), scan speed of 400 Hz and 4 frame averaging was used. Photomultiplier tube gain and offset were adjusted to give sub-saturating fluorescence intensity with optimal signal to noise ratio. Image analysis was performed using ImageJ v3.91 software (http://rsb.info.nih.gov/ij). Five high power fields were selected for analysis of each stain. To avoid being biased by the NF-κB staining, each field was selected by viewing nuclear (DAPI) staining only to identify near-confluent cells and thereby maximise the cell numbers included in the analysis. Preliminary experiments demonstrated that this approach provided data on at least 500 cells, and that the variance of the data was not changed by increasing the sample size further (data not shown). For each high power field, binary image masks were created of rel A and DAPI positive staining to define regions of interest (ROI) for analysis. This was done by applying a median filter (3 × 3 pixel radius) to remove noise and to approximate the distribution of staining intensity to a median value (Fig. 1B). Automatic thresholding, using the Isodata algorithm (Ridler and Calvard, 1978) was then used to convert the image to a binary mask (Fig. 1C) that included all fluorescence data above background. The DAPI staining mask was used to define the nuclear ROI. Using the image calculator, the DAPI mask was subtracted from the rel A mask to create a staining mask defining the cytoplasmic ROI. Each of these ROI masks were then applied, by the image calculator, to the original rel A (AF633) staining images to separate nuclear and cytoplasmic staining within each high power field (Fig. 2A). Quantitative fluorescence data were exported from ImageJ generated histograms into Microsoft Excel software for further analysis and presentation (Fig. 2B). Measurement of nuclear fluorescence alone does not distinguish NF-κB nuclear translocation from increased background levels of NF-κB expression or artefactual differences in staining intensity. Therefore nuclear and cytoplasmic staining intensities were compared to give the nuclear:cytoplasmic ratio as a relative measure of rel A nuclear localisation. Nuclear and cytoplasmic histogram data were first normalised for the total number of data points included in the analysis and then comparison was made of the sum of staining intensities. In this way, NF-κB nuclear translocation is represented by an increase in nuclear:cytoplasmic ratio of rel A staining. Multiple images can be batch processed simultaneously in this way by converting individual images into DAPI and NF-κB rel A stacks. 3 Data presentation and discussion To validate this methodology and analysis, time course and dose response studies of LPS induced NF-κB nuclear translocation were performed. Macrophages were stimulated with 0–100 ng/ml LPS in NM with 10% HS for up to 1 h. NF-κB nuclear translocation was evident by 30 min and maximal at 60 min (Fig. 3). The expected increase in nuclear:cytoplasmic ratio was also evident across the LPS dose range (Fig. 4). In addition we tested the effect of Polymyxin B (PMB), a polypeptide that binds and neutralises the bioactive lipid A component of LPS. PMB completely abrogated NF-κB nuclear translocation in response to LPS specifically, and had no effect on stimulation with the TLR2 ligand, Pam3CSK4 (Fig. 5). To further evaluate the quantitation provided by this assay the results were compared with a commercial NF-κB reporter gene assay using the HEK-293 cell line transfected with TLR2 and a secreted alkaline phosphatase NF-κB reporter gene construct (Invivogen). Transfected cells were seeded on to glass coverslips as described above and allowed to adhere for 48 h. They were then stimulated with Pam3CSK4 (dose range 0–2 μg/ml) diluted in the manufacturer's detection media. Cells were fixed and stained as above for immunofluorescence staining of NF-κB after 1 h stimulation. Duplicate wells were allowed to incubate for 6 h to allow the reporter gene and substrate reaction to take place. Cell culture supernatants were then harvested to quantify the colorometric reaction spectrophotometrically at 630 nm. Quantitative comparison of NF-κB activation using the reporter gene expression assay and NF-κB nuclear translocation by confocal microscopy showed statistically significant correlation (Fig. 6). Importantly, reporter gene expression provides a measure of NF-κB function that is dependent on its nuclear translocation, but not exclusively regulated by it. The difference between measurement of nuclear translocation and NF-κB function is acknowledged and may in part explain the imperfect regression analysis (r2 0.83) between the two methods. The method proposed here is most suitable for adherent cell cultures with relatively large cytoplasmic:nuclear area ratios that allow clear distinction between nuclear and cytoplasmic NF-κB staining. It requires relatively few cells and can be used to study NF-κB nuclear translocation at single-cell level or in mixed cultures. It can be readily applied to the study of NF-κB activation in macrophages, dendritic cells, epithelial and endothelial cells, and fibroblastic cells. We have also been able to apply this method to monocytic cells in suspension (THP-1 and K562 cell lines) by air drying them onto coverslips for immunostaining (data not shown), albeit their typically small nuclear:cytoplasmic area ratio may limit the accuracy of quantitation. This generic image analysis methodology may be applied to quantitative analysis of other transcription factors and signalling events in which assessment of sub-cellular localisation is necessary. Where confocal microscopy facilities are available, this method overcomes the problems related to sensitivity, use of radioisotopes and cost. It can be easily adopted in current cellular immunology research, and given the ready accessibility of the public domain image analysis software, with further validation this methodology may serve as a universal standard that allows better comparison of data from separate experiments and different research groups.

Environ_Manage-3-1-1866215

Early Vegetation Development on an Exposed Reservoir: Implications for Dam Removal

The 4-year drawdown of Horsetooth Reservoir, Colorado, for dam maintenance, provides a case study analog of vegetation response on sediment that might be exposed from removal of a tall dam. Early vegetation recovery on the exposed reservoir bottom was a combination of (1) vegetation colonization on bare, moist substrates typical of riparian zones and reservoir sediment of shallow dams and (2) a shift in moisture status from mesic to the xeric conditions associated with the pre-impoundment upland position of most of the drawdown zone. Plant communities changed rapidly during the first four years of exposure, but were still substantially different from the background upland plant community. Predictions from the recruitment box model about the locations of Populus deltoides subsp. monilifera (plains cottonwood) seedlings relative to the water surface were qualitatively confirmed with respect to optimum locations. However, the extreme vertical range of water surface elevations produced cottonwood seed regeneration well outside the predicted limits of drawdown rate and height above late summer stage. The establishment and survival of cottonwood at high elevations and the differences between the upland plant community and the community that had developed after four years of exposure suggest that vegetation recovery following tall dam removal will follow a trajectory very different from a simple reversal of the response to dam construction, involving not only long time scales of establishment and growth of upland vegetation, but also possibly decades of persistence of legacy vegetation established during the reservoir to upland transition. Introduction Although more than 450 dams have been removed in the United States in the past century, dam removal has only recently received significant attention from the scientific community (Beyer 2002, Hart and Poff 2002, Graf 2003). Reasons for dam removal include unsafe conditions or loss of function associated with aging or sediment-filled structures and, more recently, environmental restoration (Hart et al. 2002, Pohl 2002). Dam removal decisions involve a tradeoff of multiple socioeconomic and ecological costs and benefits (Stanley and Doyle 2003). Potential environmental consequences of dam removal include (1) benefits from restoring more natural flow and sediment regimes (Poff et al. 1997, Kondolf 1997) and (2) removing barriers that block fish passage (Lenhart 2003) and fragment the river corridor (Nilsson et al. 2005). Potential negative effects include the impacts of releasing stored, and possibly contaminated, sediment and enhancing dispersal of undesirable species (Bednarek 2001). The aspect of dam removal examined here is the ecological fate of the land under the former reservoir pool. Trajectories of vegetation response on lands exposed by dam removal influence higher-order responses such as human and wildlife use and biogeochemical processes (Shafroth et al. 2002). In some cases, restoration of pre-dam vegetation may be a management goal. The extent to which this can be accomplished by natural colonization and subsequent vegetation change may significantly affect project costs. In some cases, natural colonization may be the default management action because of budgetary constraints, limited mandates, or lack of interest in restoration. Even in the absence of a specific restoration target for vegetation on the exposed surfaces, there are concerns about rapid dominance of these barren areas by undesirable, weedy, non-native species and the need to provide stabilizing vegetation to minimize erosion. Several sources of information support projections of likely future vegetation on land exposed by a dam removal. The first source includes studies of the plant communities in the surrounding upland landscapes, in the riparian zones of rivers, and in the margins of lakes and reservoirs. The description of these communities and the controls on their composition, especially vegetation dynamics on disturbed, bare ground sites, provide a coarse identification of possible states for the former reservoir pool. There is also a significant literature examining vegetation dynamics within periodically exposed lake or reservoir shorelines illustrating the importance of intra- and inter-annual water level fluctuation (Keddy and Rznicek 1982, 1986, Hill et al. 1998) and positive relations between species richness and both total cover and substrate fineness (Nilsson and Keddy 1988, Nilsson et al. 1997). Studies of floodplain vegetation colonization and dynamics provide information on likely pioneer species and subsequent changes associated with fluvial processes and geomorphic surfaces within river bottomlands in many regions. For example, Friedman et al. (1996) described patterns of vegetation change on the floodplain of Plum Creek in eastern Colorado, where species richness peaked at intermediate ages, older and higher surfaces were increasingly dominated by rhizomatous perennials, and the overall species list was 36% non-native. In the western United States, much of the focus on relations between streamflow and riparian vegetation has centered on cottonwood, which is the structurally dominant native tree. Populus deltoides subsp. monilifera (plains cottonwood) is a pioneer species with a relatively narrow regeneration niche. The requirements for a bare, moist surface with limited drawdown following germination have been represented in a formal recruitment box model describing the floodplain locations and patterns of water stage where establishment is likely for P. deltoides subsp. monilifera and other species of cottonwood and willow with similar establishment requirements (Mahoney and Rood 1998, Rood et al. 2005). This model is a clear example of how expectations derived from the study of riparian plant distributions and life history requirements can be used to inform an assessment of recolonization of the former reservoir pool following dam removal. The second general source of information is observations from actual dam removals (Bednarek 2001, Stanley and Doyle 2002). In relatively few cases have environmental effects been evaluated following dam removal, and these studies were all of dams less than 17 m tall in relatively humid settings (Hart et al. 2002). Quantitative analysis of vegetation response to actual dam removal is rare, although two recent studies examine vegetation colonization and succession within the former reservoir pools of small dams removed in Wisconsin (Lenhart 2000, Orr and Stanley 2006). A final source of information is case studies from alterations at least partially analogous to dam removal, including breaching of beaver and debris dams, accidental human dam failures, and dam maintenance activities. For example, the episodic release of a large sediment pulse from dam maintenance has been analyzed as a surrogate for the downstream effects of the type of sediment pulse that might be produced by dam removal (Wohl and Cenderilli 2000, Zuellig et al. 2002). In this study, we examine vegetation colonization and early dynamics on areas exposed when a reservoir was drained for four years to facilitate dam repairs. We describe vegetation pattern in terms of time since exposure in this analog to dam removal, in order to supplement the sparse empirical database available to scientists, resource managers, and policy makers involved in dam removal evaluations. Study Site Horsetooth Reservoir is located in the foothills of the Rocky Mountains, 7 km west of Fort Collins, Colorado, in the transition zone between two physiographic provinces: the Colorado Piedmont subdivision of the Great Plains to the east and the southern Rocky Mountains to the west (Fenneman 1931). At an elevation of 1,655 m asl, the study site is in the rain shadow of the Rocky Mountains, approximately 70 km east of the Continental Divide; mean annual precipitation ranges from 36–40 cm, more than 70% of which falls between April and September. Summer in the study area is typically hot, with a mean July maximum of 29°C. Fall is cool and typically dry, punctuated occasionally by wet and sometimes heavy upslope snowstorms. Winter is characteristically dry and cool to cold with a mean minimum January temperature of −3.2°C and extreme winter minimum temperatures as low as −40°C (Hansen et al. 1978). The transition zone between these physiographic regions, described by Marr (1961) as the Grassland-Lower Montane ecotone, is characterized by a rapid change in elevation and a shift from grassland to forest. In this transition zone, localized differences in soil primarily determine the dominant vegetation type at a given location. Grasslands characteristically dominate deeper, finer textured soils, transitioning to shrublands and open stands of ponderosa pine (Pinus ponderosa) on shallow, rocky soils and fractured rock outcrops (Marr 1961). Existing vegetation on the steep slopes and ridges above the high water line of Horsetooth Reservoir, matches the transition from shrublands dominated by Cercocarpus montanus and Rhus trilobata (Rhus aromatica, Great Plains Flora Association 1986) to a Pinus ponderosa community type described by Peet (1981) for rocky slopes below 1,700 m. This open, xeric forest type is characterized by widely scattered Pinus ponderosa with a grass-dominated understory. Despite xeric site conditions, understory species diversity and cover are relatively high. Cercocarpus montanus, Rhus trilobata, and Yucca glauca dominate the shrub layer, while Stipa comata, Bromus tectorum, Helianthus pumilus, Sporobolus cryptandrus, Bouteloua hirsuta, and Verbascum thapsus are important in the herbaceous layer (Peet 1981). As seen in portions of the valley not inundated by Horsetooth Reservoir, grasslands dominate toeslopes above the valley margins as well as the valley floor, where finer-textured soils accumulate to greater depths. These grasslands are characteristically dominated by Agropyron smithii, Andropogon scoparius, Bouteloua curtipendula, B. gracilis, Bromus tectorum, and Stipa comata (Hansen and Dahl 1957). Narrow valley floors typically support scattered stands of riparian shrubs and trees, including Salix irrorata, Betula occidentalis, Populus deltoides subsp. monilifera, and Salix amygdaloides (Marr 1961). The reservoir is situated between two sharp ridge crests or hogbacks, formed by steeply dipping layers of shales and sandstones. The reservoir is approximately 10 km long, and is formed by four large, earth-filled dams; Horsetooth Dam closes the northern end of the valley, and Soldier Canyon, Dixon Canyon, and Spring Canyon Dams close breaches in the eastern hogback ridge created by pre-existing cross-valley drainages. The structural heights of the dams are 47, 69, 73, and 67 m, respectively. Construction of the four dams creating Horsetooth Reservoir occurred between 1946 and 1949. With a total off-channel storage capacity of approximately 1.9 × 108 m3, Horsetooth is one of 12 storage reservoirs built as part of the Colorado-Big Thompson Project, which stores, regulates, and diverts approximately 3.2 ×108 m3 of water annually from the Colorado River headwaters on the west slope of the Continental Divide to the more heavily populated east slope. The project provides water for irrigated agriculture, municipal and industrial use, hydroelectric power generation, and water-based recreation. Discovery of sinkholes and increased seepage from the reservoir prompted a dam modernization project and reservoir drawdown that began in the fall of 2000 (B. Boaz personal communication, U.S. Bureau of Reclamation 2005). Methods Field Sampling We sampled vegetation along 13 transects that began 5 m into upland vegetation (above the reservoir’s high water mark of 1654.6 m above sea level) and extended down slope (perpendicular to the shoreline) to an elevation of 1621.5 m asl. Below 1621.5 m asl, the reservoir begins to separate into distinct pools with different water surface elevations. Transects were located randomly along the entire length of the shoreline, with the exception of the following excluded areas: small, shallow coves on the west side of the reservoir; the four dams; and 200 m on either side of each dam. Along each transect, we estimated the percent cover of every species present in 1-m2 plots in mid-September of 2001 and 2002. Plants were identified to species when possible using local and regional floras (Great Plains Flora Association 1986, Weber 1990). Plots were spaced variably depending on the steepness of the slope so that they were evenly distributed along the elevational gradient. On steep slopes, a plot was sampled every meter; on progressively gentler slopes, plots were sampled at intervals of 2, 3, or 4 m in order to achieve as close as practicable to 3 plots per m of elevation change. Transect lengths ranged from 89 to 382 m. Following exclusion of plots without clear hydrologic history as described below, the numbers of plots analyzed per transect ranged from 80 to 142 with a total of 1,345 plots each year. The proportion of surface area at each plot that was occupied by cobble-sized particles or larger (>64 mm diameter) was estimated in the field. A single elevation of each plot was determined using a total station surveying instrument, registered to the water surface elevation and tied to the gage measuring long-term reservoir water levels. Substrate characteristics and topography were measured in 2001. General observations in 2002 suggested that there had been little change in substrate or topography between years. Records of water level fluctuation during the study period were combined with plot elevations to estimate when plots were exposed. Using this information, we divided the transects into four basic zones (from highest elevation to lowest): UPL, an upland zone above the level of reservoir inundation; TOP, former reservoir bottom exposed during the 1999 growing season and not inundated again during the study; MID, former reservoir bottom exposed during the 2000 growing season and not inundated again during the study; and BTM, former reservoir bottom exposed temporarily during the 2000, 2001, and 2002 growing seasons (Figs. 1 and 2). Areas excluded from analysis consisted of (1) portions of transects below 1621.5 m asl where separate pools began to form in various portions of the reservoir; (2) a wave action zone, above the high water mark of the reservoir, but still disturbed; and (3) a narrow band between the MID and BTM zones that was exposed in 2000 and 2001 but was not re-exposed in 2002. The wave action zone and the zone exposed only in 2000 and 2001 each spanned less than 0.5 m of the elevation gradient and had too few plots for meaningful analysis. Fig. 1Reservoir water surface elevation and drawdown zones. Vegetation analysis is limited to four distinct elevation zones: UPL was never inundated; TOP was first exposed in 1999 and not subsequently; MID was first exposed in 2000 and not subsequently; BTM was first exposed in 2000 and was subsequently re-flooded and re-exposed in both 2001 and 2002Fig. 2Horsetooth Reservoir when sampled in September 2001. At this time, the TOP01 zone was in the third year of exposure, MID01 in the second year, and BTM01 in the first year. Second-year seedlings of Populus deltoides subsp. monilifera are evident in the MID01 zone and several mature individuals are present near the full pool elevation of the reservoir before drawdown. Melilotus spp. strongly dominates the UPL01 zone here Data Analysis Plant characteristics and communities Plant characteristics are summarized by elevational zone and sampling year with each combination denoted by the zone (BTM, MID, TOP, and UPL) subscripted by the last two digits of the year of sampling (2001 and 2002). The resulting eight zone-year combinations are grouped into classes based on the number of years since the zone was last exposed or drawn down (1, 2, 3, 4 years and Upland). There was no preexisting, rooted vegetation in any of these zones prior to their first exposure and no apparent survival of vegetation in the BTM zone between the two successive years it was drawn down. Fractional nativity, duration, and wetland index were based on the aggregate species list for each zone-year-transect combination. Nativity was calculated as the fraction of species classified as native and duration was calculated as the fraction of species classified as perennial based on McGregor et al. (1986) and USDA-NRCS (2004) For example, a value of 0.6 for nativity would mean 60% of the species were classified as native and a value of 0.4 for duration would mean 40% of the species were classified as purely perennial (as opposed to annual, biennial or with a mixed duration). Wetland indicators (Reed 1988, USDA-NRCS 2004) for individual plants of Obligate (OBL), Facultative Wetland (FACW), Facultative (FAC), Facultative Upland (FACU), and Upland (UPL) were assigned numeric values of 1 to 5, respectively (Tiner 1999). Thus, values of the wetland index range from 1.0 for a hydric community composed entirely of obligate wetland species to 5.0 for a xeric community composed entirely of upland species. Total vegetative cover within each zone-year-transect combination was adjusted for differences in substrate by analysis of covariance using rockiness (proportion of surface occupied by cobble or larger particles) as a covariate. This analysis (Proc Mixed, SAS 2003) fit separate relationships between an arcsine square root transformation of cover and a square root transformation of rockiness for each zone-year combination. We report total vegetative cover means and 95% confidence intervals for each zone-year combination adjusted to the grand mean of the covariate. Direct comparisons of observed species richness across zones and transects were not possible because of the different areas and numbers of plots sampled. We pooled all plots sampled in each zone-year combination, without regard to transect, in order to develop comparable estimates of species richness. We used two procedures. A first-order jackknife estimate of total richness is based on the number of observed species, the number of sampled plots, and the number of species observed in only one plot (McCune and Grace 2002). We also used a bootstrap procedure to estimate the total number of species as the asymptote of a fitted species-area curve. For each zone-year combination, we drew 50 random samples of each number of plots (1 to n = total number of plots), averaged the total number of species for each number of plots, and then fit a curve to the data [Michaelis-Menten equation following Inouye (1998)]. Differences in dominant species were evaluated by calculating species cover relative to the total cover within each zone-year-transect combination. Differences in overall species composition between zone-year combinations were evaluated using relative Sorensen distance (McCune and Meford 1999, McCune and Grace 2002) expressing a percent dissimilarity of the distribution of cover across species, normalized to the total cover within a zone-year combination. This index was based on cover values for each zone-year obtained by first averaging all plots within a zone-year-transect and then averaging across transects within each zone-year combination. Cottonwood seedling establishment We examined a priori expectations derived from Mahoney and Rood’s (1998) recruitment box model concerning (1) the position of new seedlings relative to the water’s edge during the period of seed availability and (2) rates of water level decline that could be survived by new seedlings. For these analyses, we pooled plots across transects and years. We then calculated average cover of first-year Populus deltoides subsp. monilifera seedlings for each year in elevational zones relative to the elevations of the water surface each year during the seed release and germination window. We used the period of June 1 to July 7 for seed release and germination based on Segelquist et al. (1993) and our antecdotal observations in Fort Collins from 1990 to 2005. To examine the effects of drawdown rate, we focused on plots within the elevational zone corresponding to the location of the water’s edge during the germination window. These plots were both predicted and observed to have the highest probabilities of establishment. For each of these plot-year combinations, we estimated the drawdown rate as the difference between the water surface elevation on the last day the plot was inundated and the water surface elevation 45 days later. In some cases, especially in 2002 when drawdown was very rapid, the water surface elevation 45 days later was below the limit of 1621.5 m asl at which pools formed in the reservoir bottom and reliable estimates of water surface at individual transects were not possible. In these cases, we used 1621.5 m asl as a lower bound of water surface elevation, thus producing a conservative (low) estimate of the rate of decline experienced by seedlings. Average cover of Populus deltoides subsp. monilifera seedlings was calculated for classes of drawdown rate by pooling observations of current year germinants in both 2001 and 2002. Results Plant Characteristics and Communities The percentage of native species was relatively constant over time with median values ranging from 56% for BTM02 exposed for one year to 40% for TOP02 exposed for four years (Fig. 3). There was no suggestion of a trend for the early colonizing communities to be shifting towards the higher proportions of native species associated with the upland that had median percentages of native species of 75% and 71% in the two years of sampling. In contrast, both duration and wetland index showed clear trends of increasing over the first four years of exposure toward values observed in the upland zone (Fig. 3). Median percentages of perennial species were 20% and 21% in the first year of exposure (BTM01 and BTM02); 31% after two years (MID01); 32% and 38% after three years (MID02 and TOP01); and 46% after four years (TOP02). The median percentage of perennial species in the upland zone was 86% in both 2001 and 2002. Median values for wetland index in the first year of exposure were 2.4 and 2.6 (BTM01 and BTM02), which are in the range of wetland vegetation communities (Tiner 1999). Wetland index increased steadily over time to 4.2 in the fourth year (TOP02) approaching the xeric value of 4.7 for both 2001 and 2002 in the upland zone (Fig. 3). Fig. 3Vegetation characteristics by drawdown zone and time exposed. The unit of replication is a transect within each zone-year combination. Wetland index, fraction native, and fraction perennial are calculated for each plot based on species presence and then averaged across plots for each transect within a zone-year combination. Tops and bottoms of boxes represent 75th and 25th percentiles, the horizontal line is the median, whiskers (vertical lines) include the minimum of 1.5 times the inter-quartile range and the range of the data, and values outside the whiskers are represented by asterisks Patterns in both species richness and cover were complicated by differences among sampling years in precipitation and among zones in substrate and sampled area. The second year of sampling, 2002, was drier than 2001, which contributed to lower total 2002 cover in the upland zone and lower numbers of identifiable species. We recorded a total of 124 distinct taxa. In some cases these were combinations of species that could not be reliably separated in the field at the time of sampling. Examples among the more dominant species include conflation of Bromis japonicus and B. tectorum, and Melilotus spp. which included Melilotus alba and M. officinalis. Differences in estimation procedures for total species richness did not appreciably change the relative patterns (Fig. 4). Based on the first-order jackknife estimate, species richness was 52 and 53 species in the first year following exposure (BTM01 and BTM02) compared to 78 and 53 species in the upland zone (UPL01 and UPL02). There was some suggestion of an intermediate peak in richness in the second (96 for MID01) and third (93 for MID02 and 59 for TOP01) years of exposure. Fig. 4Species richness and adjusted total cover by drawdown zone and time exposed. In the top graph, asterisks are the asymptotic richness from Michaelis-Menten species-area curves fit to bootstrapped sets of 1-m2 plots pooled across transects (Inouye 1998); open circles are first-order jack-knife estimates of richness (McCune and Grace 2002). In the bottom graph, solid circles are estimates of mean total cover adjusted to the grand mean substrate value from an analysis of covariance using transect within a zone-year combination as the unit of replication. The vertical lines from the solid circles are 95% confidence intervals for these estimated means The substrate area composed of large particles (>64 mm diameter) was a significant predictor of total cover (P < 0.001), with more rocky substrate associated with lower total cover (Fig. 4). Total plant cover, adjusted with analysis of covariance to the mean fraction of large particles, was 21–36% in the first year of exposure (BTM01 and BTM02). There was some suggestion of a decline in cover within the first four years of exposure, rather than a trend in the direction of the generally higher (37–58%) upland cover values. Species composition was most similar between zone-year combinations within a given time of exposure (Table 1). Species composition of exposed areas became progressively more similar to the upland composition with increasing time of exposure. Composition at one year of exposure was 99% dissimilar to the uplands, and composition at four years of exposure was 83–84% dissimilar to the uplands. There was, however, substantial turnover of species within the first four years, with generally large dissimilarities between the different years (e.g., 94% dissimilarity between one and four years exposed). Table 1Community dissimilarity matrix. Distance measure is a Sorensen percent dissimilarity using mean cover values for each zone-year combination and relativized to zone-year unit totals (McCune and Grace 2002). Values range from 0% for identical species cover composition to 100% for no similarity of species cover composition. Matrix is symmetrical around main diagonal. Main diagonal entries are indicated by dashes and are 0 by definintionYears exposedZone yearRelativized Sorenson distance (% dissimilarity)Years exposed1234UplandBTM01BTM02MID01MID02TOP01TOP02UPL01UPL021BTM01—15588087949999BTM0215—6579889499992MID015865—51587998983MID02807951—58799897TOP0187885858—5091924TOP029494795450—8483UplandUPL01999998979184—19UPL0299999897928319— There were also substantial shifts in the individual dominant species related to time exposed (Table 2). The introduced annual Chenopodium glaucum, with a FACW wetland indicator value, strongly dominated the first-year communities, was the second most dominant species two years after exposure, but was strongly reduced in the third year and absent from fourth-year and upland zones. Panicum capillare, a native annual, was the second most dominant species in the first year of exposure, increased to the most dominant species in the second year, declined in the third and fourth years, and was absent from the upland zone. The native Rorippa curvipes is classified as an obligate (OBL) wetland plant and was important in the first year of exposure, declined to a very minor presence in the second year, and was absent in the third-year, fourth-year, and upland communities. Table 2Relative cover of selected species by elevational zone and time exposed. Relative cover is calculated at the transect level for each zone-year combination and then averaged across transects. All species with cover ranks among the top four in any zone-year combination are included. Dashes indicate absenceSpeciesCharacteristicsRelative Cover (%) by Years Exposed and Sampled ZoneDurationNativityWetland indicator1234UplandBTM01BTM02MID01MID02TOP01TOP02UPL01UPL02Amaranthus albusANFACU2.27.13.72.6————Ambrosia tomentosaPNUPL<0.1<0.14.69.3————Bromus inermisPEUPL—<0.1<0.1<0.1—0.313.011.0Bromus japonicus - B. tectorumAEUPL——<0.10.21.74.324.718.7Cercocarpus montanusPNUPL————0.62.310.417.1Chenopodium glaucumAEFACW57.856.115.80.4————Cirsium arvensePEFACU0.20.11.813.30.83.51.00.7Ericameria nauseosaPNUPL——<0.10.210.39.21.71.8Lactuca serriolaAEFAC——2.02.912.011.7——Melilotus spp.A/BEFACU0.20.19.55.915.10.70.5—Panicum capillareANFAC10.410.326.27.28.01.2——Populus deltoides subsp. moniliferaPNFAC4.41.82.43.50.40.5——Rhus aromaticaPNUPL——————13.29.2Rorippa curvipesA/B/PEFACW5.30.9<0.1————Salsola collinaAEUPL——0.915.12.913.3—0.3Suckleya suckleyanaANFACW3.57.20.10.2————Verbascum thapsusBEUPL<0.1<0.11.32.511.111.7——Verbena bracteataA/PNUPL0.20.29.610.39.94.0<0.1—A = annual, B = biennial, P = perennial, N = native, E = exotic, W = wetland, FACW = facultative wetland, FAC = facultative, FACU = facultative upland, UPL = upland Some of the dominant species in the second through fourth years of exposure were substantially less important or absent from both the first year of exposure zone and the upland zone. These included Cirsium arvense, Ericameria nauseosa, Lactuca serriola, Salsola collina, Verbascum thapsus, and Verbena bracteata. The conflated Bromus japonicus and B. tectorum, both introduced, dominated the upland zone. This group was absent in the first year of exposure and gradually increased in relative cover in the second through fourth years. The other most dominant species in the upland, Bromus inermis, Cercocarpus montanus, and Rhus aromatica, were unimportant or absent in essentially all of the first four years of exposure (Table 2). Cottonwood Seedling Establishment Plains cottonwood (Populus deltoides subsp. monilifera) was the dominant tree species colonizing the exposed surfaces (Table 2). Mature individuals occurred in the upland near the margin of the full-pool reservoir, although none were sampled in the upland plots on randomly located transects. Seedlings of Salix amygdaloides, S. exigua, Populus angustifolia, P. tremuloides, and Tamarix ramossima were present, but rare in the drawdown zones along sampled transects. First-year cottonwood (Populus deltoides subsp. monilifera) seedlings were generally distinguishable from previous year germinants and root sprouts by their cotyledons and absence of bud scale scars when sampled in mid-September. New seedlings were strongly concentrated in, or slightly above, the elevational band occupied by the water’s edge during the period of seed release of June1 to July 7 (Fig. 5). However, smaller numbers of new seedlings were spread over a considerable range of elevations from 2 m above the zone exposed during the germination window to more than 5 m below the optimum zone. Drawdown rate had a strong influence on the cover of first-year seedlings as sampled near the end of the growing season (Fig. 6). For seedlings established in the optimum band of the wetted edge during the germination window, cover decreased substantially at drawdown rates greater than 4–8 cm/day. However, the range of drawdown rates survived by seedlings was wide, with new seedlings present with measurable cover at rates of greater than 24 cm/day. The zone of maximum cover of current year germinants was 10.2 to 12.2 m above the water level on September 1 in 2001 and 7.2–11.7 m above in 2002, using the conservative 1621.5 m asl elevation at which separate pools formed in the reservoir. Actual observed water levels at the dam on September 1 were 1618.8 m asl in 2001 and 1618.1 m asl in 2002. Average elevations of all plots containing older than first-year Populus deltoides subsp. monilifera seedlings in 2001 were 9.6 m above the maximum water level in that year and 21.8 m above the conservative 1621.5 m asl value for September 1 water level; in 2002, older seedlings were 7.5 m above the maximum and 19.5 m above the conservative September 1 level. Fig. 5Elevational distribution of first-year Populus deltoides subsp. monilifera seedlings in relation to the elevations exposed during the germination window of June 1 to July 7. Means ± 1 SE are based on plots in each elevational class pooled across transects and sampling yearsFig. 6Cover of first-year Populus deltoides subsp. monilifera seedlings (established in the zone of the wetted edge during the germination window) in relation to rate of drawdown. Means ± 1 SE are based on plots in each drawdown class pooled across transects and sampling years. Drawdown is calculated for the 45-day period following the date a plot was last exposed Discussion Nilsson et al. (1997) distinguished land as pre-upland or pre-riparian in an analysis of vegetation response to inundation from dam construction. This distinction is similarly useful in describing the response of vegetation to dam removal. Large fractions of the land under the reservoir pools of the shallow dam removal sites in Wisconsin examined by Orr and Stanley (2006) and Lenhart (2000) were pre-riparian. Contingent on changes in topography from sediment accumulated behind the dam and the geomorphic response to dam removal, pre-riparian areas might be expected to support marshland or riparian vegetation following dam removal (Shafroth et al. 2002). In contrast, we sampled slightly more than 30 meters of drawdown at Horsetooth Reservoir. Essentially all of the area exposed was pre-upland. If this had been a permanent dam removal, the exposed area would not be expected to support a wetland or riparian plant community in the long term. The patterns of vegetation change we observed are a combination of the initial colonization of a riparian, bare ground disturbance patch and the vegetation response to a shift from mesic to xeric site conditions. The mesic to xeric transition is reflected in the wetland index value, which systematically increased from a hydric or wetland community in the first year of exposure to upland or xeric value in the fourth year of exposure, very similar to the upland values of the surrounding landscape vegetation. In contrast, Lenhart (2000) reported wetland index values at sites 3–5 years following removal of shallow dams in Wisconsin that were still indicative of hydric or wetland communities. The species composition at the Horsetooth site changed dramatically over the first four years of exposure (Tables 1 and 2), but had not yet approached that of the surrounding upland plots. The initial colonizing community of a bare surface might be expected to have higher fractions of short-lived and perhaps non-native species than later vegetation on the site. At the Horsetooth site, the fraction of perennials steadily increased through four years of exposure, whereas the fraction of native species was below that of the surrounding uplands and did not change appreciably or consistently. Substrate-adjusted mean values for total cover were lower than the upland and appeared to decline somewhat in the second through fourth years of exposure. A possible explanation for these patterns in cover and richness is that the mesic conditions in the first year of exposure produce a reasonably high plant cover composed heavily of annual species. This is followed by a period of reorganization, replacing annuals with deeper rooted perennials. Species richness peaks during this transition, whereas total cover tends to drop in the drier conditions of years two through four following exposure and only gradually increases to the ultimately higher cover of a slowly developing perennial community present on the dry upland sites. The cottonwood recruitment box model of Mahoney and Rood (1998) expresses the regeneration niche of these disturbance-dependent, pioneer, riparian tree species and makes predictions about where recruitment will occur along rivers in central and western North America in relation to water surface elevations (Rood et al. 2005). Cottonwood seedling establishment requirements are well known from multiple laboratory (Fenner et al. 1984, Mahoney and Rood 1991, Segelquist et al. 1993, Amlin and Rood 2002) and field studies (Rood et al. 1998, Auble and Scott 1998, Shafroth et al. 1998, Rood and Mahoney 2000, Rood et al. 2005). Seedling recruitment tends to occur on bare ground sites that are moist during an early summer period of seed dispersal and germinability, and that remain moist enough for seedlings to survive drought stress. More specifically, for Populus deltoides subsp. monilifera, the model describes suitable establishment sites as (1) bare ground, largely created by fluvial disturbance; (2) wetted by water surface elevations during a 3- to 6-week window of seed release, dispersal, and germinability; (3) subject to water table declines of no more than 2.5 cm/day; and (4) generally occurring at elevations from 60 to 200 cm above the water surface elevation of late summer base flow (Mahoney and Rood 1998). The Horsetooth drawdown provided bare, wetted surfaces over a much greater elevational range than would be associated with riparian floodplain sites or removal of a shallow dam. Predictions of the recruitment box model were met in two respects. First, the highest cover of first-year P. deltoides subsp. monilifera occurred in the elevation zone corresponding to the water’s edge during the estimated June 1 to July 7 period of estimated maximum seed release and germinability for the area (Fig. 5). Second, cover of first-year seedlings dropped sharply at locations subject to drawdown rates >8 cm/day (Fig. 6). However, there was appreciable seedling establishment well outside the bounds predicted by the recruitment box model. Wave action, capillary rise, and plot heterogeneity in elevation are likely explanations for the substantial establishment in locations 0–1 m above the water’s edge. The establishment at much lower elevations as far as 5–6 m below the water’s edge on July 7 is more difficult to explain. It appears that the extreme drawdown rates created suitable bare moist sites far below the elevation of the optimum, but close enough to the period of maximum seed availability to allow some of the tail end of the distribution of germinable seed to establish. We observed maximum cover at drawdown rates of 4–8 cm/day compared to the upper limit of 2.5 cm/day used in the recruitment box model, and we observed some seedlings surviving first-year drawdown rates of greater than 24 cm/day. Some survival of seedlings at drawdown rates of 4 and 8 cm/day has been reported in laboratory experiments (Mahoney and Rood 1991) and the value of 2.5 cm/day is probably better viewed as the upper end of the most suitable range of rates rather than as an absolute limit. The rapid Horsetooth drawdown rates produced seedling establishment at high elevations relative to late summer water levels. The elevation zone with maximum cover of first-year seedlings was at least 10 m above the water level on September 1 in 2001 and at least 7 m above the September 1 level in 2002. This was not purely a first-year phenomenon as the average elevations of older than current-year seedlings were well above the range of current year water levels, more than 7 m above the maximum level of the current year and more than 19 m above the level on September 1 of the current year. Mahoney and Rood (1998) discuss several factors that may support establishment at high elevations relative to the water’s edge including capillary rise, lateral subsurface water movement, and advantageous sequences of precipitation events. Given that some exposed surfaces at Horsetooth Reservoir were on the sides of sharp ridges with very limited contributing drainage area, we believe the most likely explanation for the high elevation establishment we observed is local variation in particle size and subsurface drainage (i.e., fine sediment in cracks between boulders) that create pockets of adequate moisture retained from inundation and channeled from rainwater-derived percolation. Implications for Dam Removal Decisions Given the multiple dimensions in which dams differ (Poff and Hart 2002) and the relative paucity of long-term dam removal data sets, evaluations and decisions about dam removal in the near future will be based largely on case studies and site assessments rather than on statistical conclusions from a sampled population of previous removals. A first-order approximation of future vegetation within a former reservoir pool can be obtained by considering topography of the new surface in relation to the new water surfaces; substrate characteristics; the pre-dam vegetation and its character as pre-upland or pre-riparian; and typical wetland, riparian, and upland vegetation communities in the region (Shafroth et al. 2002). For shallow dams constituting the vast majority of dams removed to date (Hart and Poff 2002), the exposed surfaces will be at elevations close to the surface of the new river and a large fraction of the exposed area would have been riparian or wetland before inundation. For tall dams and the drawdown case we report, the vast majority of the exposed land is pre-upland, not significantly influenced by a river before dam construction, substantially above the water surface following removal or exposure, and likely to support upland vegetation typical of the region. There are two primary ways such a first approximation could be misleading. The first is that the post-dam topography and substrate characteristics may have been altered substantially by the period of inundation. This is especially true if the dam has accumulated substantial quantities of trapped sediment that have altered topography, stored nutrients or contaminants, or altered organic content or particle size distributions. Our results showed a significant effect of substrate size proportions on total vegetation cover; however, Horsetooth Reservoir was not accumulating substantial sediment because of the diversion source of most of the impounded water. The second way a first approximation might be misleading is that the character of the transition between extended inundation to post-dam site conditions will influence vegetation response over some time scale, at least on the order of the life spans of the initially colonizing plants. Erosion of accumulated sediment may produce substantial changes in post-dam topography, likely resulting in a topography intermediate between the pre-dam and immediately exposed conditions. This was not a major factor at Horsetooth Reservoir because of the relative absence of accumulated sediment. On the other hand, extended inundation is an abnormal source of disturbance to initiate a bare ground plant succession for pre-upland surfaces such as those exposed by the Horsetooth Reservoir. Thus, the initially colonizing plant community was considerably more mesic (lower wetland indicator score) than the community that might colonize a dry, bare-ground surface created by a disturbance such as fire. In semi-arid landscapes, establishment of pioneer riparian tree species such as P. deltoides subsp. monilifera is almost always associated with hydrologic disturbance and bare ground sites near the water’s edge, although mature trees can often persist on much drier sites. Thus, successful establishment of riparian cottonwood during the drawdown may leave a transient legacy of vegetation on the exposed reservoir surfaces that would be atypical of the sites either before reservoir construction or over the long term following dam removal. These patterns are very much a function of the specifics of reservoir drawdown and are thus potentially amenable to management as part of a dam removal. For example, cottonwood would be unlikely to colonize the bulk of pre-upland first-year exposure sites if the drawdown occurred in winter and the area was dry during the early summer period of seed release. On the other hand, the summer drawdowns observed at Horsetooth probably increased total plant cover on the exposed surfaces, which might be desirable from the perspective of erosion mitigation. Weed control on the former reservoir pool is a consideration in many dam removal evaluations. Initial colonization of the reservoir pools at Horsetooth was by a ruderal community (Menges and Waller 1983, Grime 2001) with high fractions of short-lived species, including invasive, non-native weeds. This is a reasonable expectation of a riparian bare ground disturbance patch and generally for reservoir pool surfaces that might be exposed by dam removal where non-native species are common in the regional species pool (e.g., Orr and Stanley 2006, Lenhart 2000). Effective weed control is complicated in the case of a tall dam. We observed a dramatic turnover of dominant species during the first four years and substantial changes in wetland character of the vegetation in response to drier site conditions. This suggests that (1) species-focused weed control will have to be maintained over multiple years with shifting targets as some of the weeds are diminishing naturally and being replaced by other species; and (2) planting of a desired or transition crop to preempt space and resources from invading weeds will have to consider rapidly changing site conditions to be effective in both the initial drawdown and immediately subsequent years. Stanley and Doyle (2003) suggested that dam removal is best viewed as a new disturbance with a consequent set of positive and negative effects, rather than a simple reversal of the original impacts of dam construction. Sediment transport is one clear example of hysteresis, where the pulse release of accumulated sediment can produce a post-removal response that is not a simple mirror image or reversal of the response to dam construction. Vegetation of the former reservoir pool is another such asymmetrical response. The existing plant community is eliminated quickly (weeks to months) following inundation of the reservoir pool from dam construction. Recovery follows a different path on a different time scale. Lenhart (2000) suggested that the initial establishment of stinging nettle (Utrica dioica) and rice-cut grass (Leersia oryzoides) on sediment of the former reservoir pool may be suppressing the development of woody vegetation on shallow dam removal sites in Wisconsin. We saw little indication of this type of effect, as the dominant species continued to shift rapidly in the first four years of the Horsetooth Reservoir drawdown. However, the changes we observed in the first four years suggest that recovery of the vegetation community will be on the time scale of years to decades. Vegetative cover, composition of dominant species, and proportions of native and perennial species were still changing after four years and were substantially different from the upland community representing the likely long-term vegetation. Furthermore, the establishment and survival of cottonwood in the drawdown zone suggests that the particular nature of the transition from decades of continuous inundation to a xeric condition may leave a legacy signature in the vegetation community at least on the decades to century time scale corresponding to the lifespan of cottonwood.

Histochem_Cell_Biol-4-1-2386534

The human keratins: biology and pathology

The keratins are the typical intermediate filament proteins of epithelia, showing an outstanding degree of molecular diversity. Heteropolymeric filaments are formed by pairing of type I and type II molecules. In humans 54 functional keratin genes exist. They are expressed in highly specific patterns related to the epithelial type and stage of cellular differentiation. About half of all keratins—including numerous keratins characterized only recently—are restricted to the various compartments of hair follicles. As part of the epithelial cytoskeleton, keratins are important for the mechanical stability and integrity of epithelial cells and tissues. Moreover, some keratins also have regulatory functions and are involved in intracellular signaling pathways, e.g. protection from stress, wound healing, and apoptosis. Applying the new consensus nomenclature, this article summarizes, for all human keratins, their cell type and tissue distribution and their functional significance in relation to transgenic mouse models and human hereditary keratin diseases. Furthermore, since keratins also exhibit characteristic expression patterns in human tumors, several of them (notably K5, K7, K8/K18, K19, and K20) have great importance in immunohistochemical tumor diagnosis of carcinomas, in particular of unclear metastases and in precise classification and subtyping. Future research might open further fields of clinical application for this remarkable protein family. Introduction Most eukaryotic cells contain in their cytoplasm a more or less elaborated cytoskeletal system consisting of intermediate filaments (IF), which are chemically very stable long and unbranched filaments of ~10 nm in diameter. Among the various families and subfamilies of IF proteins, that of the keratins is outstanding due to its high molecular diversity. The keratin gene family consists of the highest number of members in humans with 54 distinct functional genes. IF proteins are expressed in a highly cell type-specific manner, and herein keratins represent the typical IF category of epithelial cells. In some but not all epithelia, keratin filaments are conspicuously bundled as tonofilaments. Figure 1 shows these keratin filament bundles at the light microscopical (Fig. 1a, b) and the electron microscopical level (Fig. 1c, d). Inside the cell they braid the nucleus (Fig. 1a), span through the cytoplasm and are attached to the cytoplasmic plaques of the typical epithelial cell–cell junctions, the desmosomes (Fig. 1b, d; for a recent review, see Waschke 2008). This feature already suggests that keratins play a major functional role in the integrity and mechanical stability of both the single epithelial cells and, via cell–cell contacts, of that of the epithelial tissues. Consequently, they are inherent part of the continuum of stability from the single cell to the tissue formation. Evidence for this main function of keratin filaments has been amply provided by the recognition of various hereditary keratin diseases and transgenic mouse models. In addition, however, various regulatory functions have been discovered more recently (for recent reviews, see Magin et al. 2007; Oshima 2007; Uitto et al. 2007; McLean and Irvine 2007).Fig. 1Cytoskeleton of epithelial cells. a Immunofluorescence staining of keratin K18 (red, nuclei stained in blue by DAPI) in PLC (liver carcinoma) cells in vitro. b Keratin filaments (in red) and the desmosomal component desmoplakin (in green) are labeled in cultured keratinocytes of line HaCaT. c Electron microscopic image of tonofilament (keratin) bundles (arrowhead) of HaCaT keratinocytes. d Keratin intermediate filaments (black arrowhead) insert at desmosomes (white arrowhead) at cell–cell contact sites of keratinocytes of the epidermal stratum spinosum (electron microscopy) Historically, keratin research started with studies of sheep hair (wool) keratins (Crick 1952; Powell and Rogers 1986; Oshima 2007). Several important discoveries were made in the 1970s of the last century. One was the finding of the spontaneous self-assembly and polymerization of keratin filaments from denatured, soluble keratin proteins by dialysis in vitro (Steinert et al. 1976). Further milestones were the findings that antibodies against keratins from epidermis-type epithelia such as the bovine muzzle (“prekeratin”) react with tonofilaments in various epithelial cells including non-stratified “simple” epithelia of inner organs (Franke et al. 1978), and that keratins of various mammalian species exhibit a high degree of molecular diversity with differentiation-specific expression (Franke et al. 1981). Systematic protein biochemical analyses of human cells and tissues by one- and two-dimensional gel electrophoresis, Western blotting and peptide mapping disclosed the diversity of human (cyto)keratin polypeptides (Moll et al. 1982b; Tseng et al. 1982; Wu et al. 1982). From these data, in 1982, the catalog of human cytokeratins including 19 members was proposed (Moll et al. 1982b) which, although intended as provisional, has been widely accepted and used. Along with these studies and subsequently, the principle of separation of these proteins into type I (“acidic”) and type II (“basic to neutral”) keratins (see below) also emerged. Another unique property of the keratins is that in contrast to the other IF proteins they only can constitute their filamentous stage by heteropolymeric pair formation of type I and type II (1:1) molecules. Later on, several new keratins were identified and added to the cytokeratin catalog, the most notable of these being the simple-epithelial keratin 20 (K20; Moll et al. 1990, 1992) and several keratins specific for distinct epithelia such as keratin K2e in the upper epidermis (appendix “e”; now K2), K2p in the upper hard palate epithelium (similar to K2e but with appendix “p” for palate; now K76) (Collin et al. 1992a, b), or several keratin K6 (K6a–h) isoforms (Takahashi et al. 1995). Simultaneously, informations about the keratin gene sequences were revealed. Moreover, within the last 10 years a large number of hair follicle-specific epithelial keratins were discovered. This series started with K6hf (appendix “hf” stands for “hair follicle expression”, now K75), which was expressed in the hair follicle companion layer (Winter et al. 1998). K75 was the first epithelial keratin specifically expressed in the hair follicle. Surprisingly, there were much more epithelial keratins with hair follicle specificity, namely the type II keratins K6irs1, K6irs2, K6irs3 and K6irs4 (now K71–K74) and type I keratins K25irs1, K25irs2, K25irs3 and K25irs4 (now K25–K28), all of them specifically expressed in and closely restricted to the various compartments of the hair follicle inner root sheath (Langbein et al. 2002, 2003, 2006; for review see Langbein and Schweizer 2005). Besides the variety of epithelial (“soft” or “cyto-”) keratins, hairs and nails are built up from a somewhat separate subfamily of “hard” or “trichocytic” keratins, commonly designated as hair keratins (Heid et al. 1988a; Langbein et al. 1999, 2001, 2004; Langbein and Schweizer 2005; Schweizer et al. 2007). They differ from the epithelial keratins by their considerably higher sulfur content in their non-α-helical head and tail domains, which is mainly responsible for the extraordinary high degree of filamentous cross-linking by keratin-associated proteins (KAPs) (for review, see Rogers et al. 2006). Very recently, the “Keratin Nomenclature Committee” established the novel consensus nomenclature for mammalian keratin genes and proteins (Schweizer et al. 2006), relying upon and systematically extending the aforementioned 1982 catalog. This nomenclature is now in accordance with the nomenclature of the Human Genome Organization (HUGO) for both the gene and protein names. Following the unified new principles, several parts of the former nomenclature were implemented; the hair keratins (e.g. “Ha” and “Hb”) and the special epithelial keratin designations (e.g. K2p, K6hf, K6irs, K25irs) were equally integrated (see Table 1), and the nomenclature system—although now complete for humans—is open to application in other mammalian species by following the same principles. Among human keratins, the new consensus nomenclature (Table 1) comprises the type I keratins K9–K10, K12–K28, and K31–K40 (including K33a and K33b) and the type II keratins K1–K8 (including K6a, K6b and K6c) and K71–K86. Thus, there are 28 type I keratin genes (17 epithelial keratins and 11 hair keratins) and 26 type II keratin genes (20 epithelial keratins and 6 hair keratins). All in all, out of the 54 human keratin genes, at least 26 (~50%) are specifically expressed in the hair follicle. In the human genome, the keratin genes are clustered at two different chromosomal sites: chromosome 17q21.2 (type I keratins, except K18) and chromosome 12q13.13 (type II keratins including K18). The keratin genes are designated as KRT1, KRT2, KRT3, etc. (Schweizer et al. 2006; Fig. 2a).Fig. 2Human keratins. a Organization of the human keratin genes in the genome. The type I and type II keratin gene subdomains are located on chromosomes 17 and 12, respectively. The type I keratin K18 is located in the type II cluster on chromosome 17 (arrow). b Two-dimensional catalog of the human keratin proteins according to molecular weights (MW) and isoelectric points (IEP) as calculated from amino acid sequences. The keratin genes are designated according to the new keratin nomenclature (Schweizer et al. 2006)Table 1The new human keratin nomenclature (Schweizer et al. 2006)For K1–K20 (in gray), the numbering of the original catalog (Moll et al. 1982b, 1990) has been maintained. The respective gene names (“KRT”) by the human genome consortium utilize the same numbers, e.g. “KRT20” All these keratins belong to the family of IF proteins and therefore share common protein-structural characteristics. They contain a central rod domain of ~310 amino acids with α-helical conformation flanked by non-helical head and tail domains of variable length. The head domain consists of subdomains V1 and H1. The central α-helical rod domain is composed of subdomains 1A, 1B, 2A, and 2B connected by the linkers L1, L12, and L2. The tail domain then consists of subdomains H2 and V2 (Lane and McLean 2004; Parry et al. 2007; Geisler and Weber 1982). The molecular weight of human keratins ranges from ~44 to ~66 kDa (Fig. 2b). A unique feature of keratins, including the hair keratins, is their pairing, i.e. the obligate formation of heterodimers between one type I keratin and one type II keratin. This occurs by association of the corresponding rod domains in α-helical coiled-coil conformation. The resulting heterodimers and -tetramers form the basic building units of the keratin filaments. Single keratin proteins deviating from equimolar type I/type II amounts are rapidly degraded (Lu and Lane 1990). As keratin filaments are important structural stabilizers of epithelial cells, there is unabatedly high interest in keratins in biology, embryology, pathology, and dermatology. Notably, this main cytoskeletal function transcends the single cell level. Typically, keratin filaments insert at desmosomes (Fig. 1b, d) and hemidesmosomes. Thus, they contribute not only to the stability between epithelial cells itself but also to basement membrane attachment and insofar to the connective tissue compartment of a given epithelium. In the non-stratified (simple) epithelia of internal parenchymatous organs, which experience little mechanical stress, only very few keratin members form sparse and loosely distributed keratin filaments in the cytoplasm. Otherwise, considerably more members take part on the IF cytoskeletal composition of squamous epithelia which increases in the cornified stratified epithelia such as in the epidermis lining the outer body surface where they are abundant and densely bundled as tonofilaments. The loose filaments in the former case are composed of “simple-epithelial keratins” like K8/K18 (and K19), while the bundled filaments (tonofilaments) in the latter case (Fig. 1c, d) are built up from keratinocyte-type keratins such as K5/K14 in the basal layer and—with even more pronounced bundling—K1/K10 in the suprabasal layers and K2/K10 in the uppermost ones. The “rule” that the “stronger/harder” the epithelial structure the more keratin members are involved culminates in the hair fiber where 17 keratins are sequentially expressed. This clearly underscores the importance of the keratins for the tissue integrity and the relevance of the molecular diversity of keratin proteins. The important mechanical function of stratified epithelial- and epidermis-type keratins is evident and proven not only through knock-out mouse models but also through various human hereditary keratin diseases. Thus, point mutations of distinct keratin genes now widely explain the pathogenesis of several autosomal-dominant familial diseases, many of which are blistering skin diseases. The most well known of these inherited skin fragility disorders is epidermolysis bullosa simplex (EBS), the various variants of which are caused by a spectrum of point mutations of K5 or K14 (Lane and McLean 2004; McLean and Irvine 2007; Uitto et al. 2007). Nineteen different keratin genes including hair keratins and hair follicle-specific epithelial keratins have up to now been identified as being involved in pathogenic keratin mutations (Lane and McLean 2004; Schweizer et al. 2007); they will be discussed in the descriptions of the individual keratins below. Updated details may be retrieved from an Internet database (Human Intermediate Filament Database; http://www.interfil.org). Notably, the knock-out experiments and the genetic diseases demonstrated that mutations in keratin genes often (depending on the locus of mutation within the keratin molecule) cause more severe defects than the complete loss of a keratin gene whose failure might be compensated—if available at this site—by another/other keratin/s. Moreover, it has been recognized that keratins are not simply static intracellular skeletal structures but rather are highly dynamic. Along with this view, besides their mechanical function new functional roles of keratins have been defined and still emerge under special physiological conditions (Magin et al. 2007; Oshima 2007). These include the protection of the placental and trophoblast barrier function (K8/K18/K19: Jaquemar et al. 2003; Hesse et al. 2000), the protection from apoptosis (K8: Caulin et al. 2000; Ku et al. 2003b; K17: Tong and Coulombe 2006), the protection of the liver against stress and from injury (K8/K18: Zatloukal et al. 2000; Ku et al. 2003a), and the regulation of protein synthesis and cell size during wound healing involving intracellular signaling pathways (K17: Kim et al. 2006). Keratins may also play a role in epithelial polarity and membrane traffic (Oriolo et al. 2007). Thus, keratins obviously exert widely varying signaling functions beyond their mechanical roles. Beyond their biological functions, keratin expression patterns not only characterize cells as “epithelial”, they are also characteristic for distinct—including the terminal—stages during cellular epithelial differentiation from embryonal to adult or of the internal maturation program during development. Epithelial tumors—including metastases—most widely retain the keratin patterns of their (normal) epithelial origin; thus, the determination of the keratin patterns of tumors are being widely exploited for cell and tumor typing. Therefore, keratins have evolved to be one of the most potent epithelial differentiation and tumor markers in cell biology, embryology, and surgical pathology. Specific antibodies against several keratins are routinely used world wide in pathology laboratories for immunohistochemical typing of carcinomas in tumor diagnostics. Numerous papers published since 1980 deal with the application of keratins as marker proteins in tumor pathology (Oshima 2007), and several previous review articles (e.g. Lane and Alexander 1990; Nagle 1994; Schaafsma and Ramaekers 1994; Moll 1998; Chu and Weiss 2002b) cover this field of application, which also will be especially considered in this review. Another clinical application is the detection of soluble keratin protein fragments derived from K8, K18, and K19 in the circulation of cancer patients; such fragments—released by carcinoma cells—are increasingly used to monitor tumor load and disease progression in the case of certain carcinomas such as non-small cell lung cancer (Barak et al. 2004; Linder 2007). Through analysis of different K18 fragments in the serum it is also possible to assess the type of chemotherapy-induced tumor cell death and distinguish between apoptosis and necrosis, due to the fact that K18 is cleaved at specific sites during apoptosis and a monoclonal antibody (M30) specific for caspase-cleaved forms of K8 is available (Leers et al. 1999; Linder et al. 2004; Linder 2007). Human keratins and their expression patterns In the following, the different human keratins and keratin pairs will be discussed together with their cell type and tissue distribution (summarized in Table 2) and their functional significance in relation to transgenic mouse models and human hereditary keratin diseases. Furthermore, characteristic expression patterns in human tumors (summarized in Table 3) and their possible diagnostic relevance will be considered.Table 2Characteristic expression patterns of typical keratins in selected human normal epithelial tissuesKeratins of simple epitheliaKeratins of stratified epitheliaType II keratins:K8K7K5K6K1K2K4Type I keratins:K18K19K10K14K15K16K17K10K9K13Parenchymatous epitheliaaAll cellsDuctal epithelia of parenchymatous organsbAll cellsK7 and K19 in all cellsSparse cellsSparse cellsK4 heterogeneously in pancreatic ductsGastrointestinal epitheliaAll cellsK19 in all cellsGastric foveolar epithelium, intestinal epitheliumcK4 heterogeneously in luminal cells; K13 in sparse luminal cellsRespiratory epitheliumPredominantly luminal cellsK7 in luminal cells; K19 in all cellsPredominantly basal cellsK6 in basal cellsPredominantly basal cellsUrotheliumAll cellsK7 and K19 in all cellsLuminal (umbrella) cellsBasal cellsFew basal cellsK13 in all basal and intermediate cellsNon-keratinizing stratified squamous epitheliaSome basal cellsK19 in many basal cellsBasal cell layer (predominantly)Basal cells layerSuprabasal compartmentAt some sites focal expression in suprabasal cellsSuprabasal compartmentEpidermisPredominantly basal cell layerBasal cell layerSuprabasal compartmentK2 in upper spinous and granular layer; K9 in palmoplantar epidermisdNot included are the corneal keratins K3/K12, the gingival/hard palate keratin K76, the eccrine sweat gland-specific keratin K77, and the hair follicle-specific epithelial and hair keratins (for keratin expression in eccrine sweat glands and in the hair follicle, see the schematic drawings in Figs. 5 and 6)aIncluding hepatocytes, acinar cells of pancreas, proximal tubular cells of kidneybBile ducts, pancreatic ducts, renal collecting ductscMost villus- and surface-lining cells; scattered cells in cryptsdHeterogeneous expression in the suprabasal compartmentTable 3Characteristic expression patterns of typical keratins in selected human carcinomasOnly diagnostically relevant data as detectable by monoclonal antibodies widely established in clinical pathology are presented (for references, see Moll 1998; Chu and Weiss 2002b; and text). Explanation of symbols: filled circle, extended staining of most tumor cases; open dotted circle, focal/heterogeneous staining of some but not all cases; open circle, no stainingaIn rare cases focal staining may be observedbFocal or extended staining in a subpopulation of tumor cases, corresponding to the basal-like phenotypecNon-mucinous typesdPreferentially/more extended in poorly differentiated caseseIn rare cases focal staining may be observed; however, squamous cell carcinomas of the cervix uteri may express K7 extendedly Simple (one-layered) epithelia K8/K18: primary keratins of simple epithelial cells The keratins K8 and K18 typically are co-expressed and constitute the primary keratin pair of simple epithelial cells, including various parenchymatous epithelia (Franke et al. 1981; Moll et al. 1982b; Owens and Lane 2003). They are the first keratins to appear in embryogenesis, as early as in pre-implantation embryos (Jackson et al. 1980), and also seem to be the oldest keratins during phylogenesis (Blumenberg 1988). In some epithelial cell types, K8 and K18 are the sole keratins present. The classical example is the liver, with K8/K18 representing the characteristic and only keratin pair of normal hepatocytes. The same is true for other highly specialized parenchymatous epithelia such as acinar cells of the pancreas, proximal tubular epithelial cells of the kidney, and certain endocrine cells such as pancreatic islet cells. Ultrastructurally, keratin filaments of this simple composition are loosely distributed within the cytoplasm and show little bundling. In other simple, one-layered epithelia such as duct-lining cells, intestinal cells, and mesothelial cells, additional simple-epithelial keratins (K7, K19, and/or K20; see below) are present in addition to the primary pair K8/K18. Furthermore, K8/K18 occur—together with other keratins—in various pseudostratified (e.g. respiratory) and complex (e.g. glandular) epithelia and in the urothelium; in these composite epithelial tissues, K8 and K18 are often most prominent in the lumen-lining cells. Even in non-keratinizing stratified squamous epithelia, K8 and K18 may be focally expressed in the basal cell layer, together with K19 and the constitutive stratified-epithelial keratins (Bosch et al. 1988; Moll 1993). Thus, K8 and K18 are widely distributed among normal epithelial tissues although they are absent in differentiating keratinocytes. It should be noted that K8 and K18 are not strictly epithelium-specific since expression of K8 and K18 may occur in rare mesenchymal cells (more frequently in fetal stages) such as certain smooth muscle cells and fibroblastic reticulum cells of lymph nodes as well as various mesenchymal tumors including rhabdo- and leiomyosarcomas (Huitfeldt and Brandtzaeg 1985; Franke and Moll 1987; van Muijen et al. 1987; Jahn and Franke 1989; Knapp and Franke 1989; Knapp et al. 1989; Jahn et al. 1993; Gould et al. 1995; Kuruc and Franke 1988; Langbein et al. 1989), where they are co-expressed with other intermediate filament types, notably vimentin and desmin. While highly specialized parenchymatous epithelial cells in their normal state, such as proximal tubular cells of the kidney, only express K8 and K18, this may change in reactive conditions. Upon various types of injury such as inflammation or atrophy these cells may additionally switch on K7 and K19, sometimes also K17 (as well as vimentin) and thus express four to five instead of two keratins (Moll et al. 1991). This increased keratin expression appears to parallel the reduction in the degree of differentiation. Thus, the keratin pattern of a given epithelial tissue may be modulated to some extent in the course of reactive changes, frequently resulting in higher complexity of keratin composition. Already the tissue distribution of K8/K18—mainly in internal epithelia—suggests that structural and mechanical functions are not their key roles, although their absence or dysfunction may be associated with hepatocyte and trophoblast fragility (for references, see Magin et al. 2007). Instead, genetic knock-out experiments have revealed distinct regulatory functions of these keratins (Magin et al. 2007; Oshima 2007). They play a role in protecting the placental barrier function (K8: Jaquemar et al. 2003) and protecting cells—in particular liver cells—from apoptosis (K8: Caulin et al. 2000; Ku et al. 2003b), against stress, and from injury (K8/K18: Zatloukal et al. 2000; Ku et al. 2003a), possibly by functioning as a phosphate “sponge” for stress-activated kinases (K8: Ku and Omary 2006). Interestingly, K8 and K18 may play a role in the regulation of the cell cycle, whereby phosphorylation of these keratins and binding of 14-3-3 adaptor proteins seem to be involved (Toivola et al. 2001; Ku et al. 2002; Margolis et al. 2006; Galarneau et al. 2007; Magin et al. 2007). In human pathology, defects in K8 and K18 may predispose to liver diseases, in particular cryptogenic liver cirrhosis (Ku et al. 2003a; Zatloukal et al. 2004), as well as to chronic pancreatitis and inflammatory bowel disease (for references, see Owens and Lane 2004). Altered K8 and K18 proteins, together with several stress proteins, in particular ubiquitin and p62, constitute the hyaline protein aggregates of hepatocytes of several (e.g. alcoholic) liver diseases now known as Mallory–Denk bodies (Zatloukal et al. 2007). In regard to malignant tumors, K8 and K18 are expressed in most carcinomas except for some differentiated squamous cell carcinomas. Therefore, K8 and K18 antibodies strongly stain most adenocarcinomas, hepatocellular carcinomas, renal cell carcinomas, and neuroendocrine carcinomas. Since highly sensitive monoclonal antibodies against these keratins are available, such as the classical mouse monoclonal CAM5.2 clone against K8 (Makin et al. 1984) and clone Ks18.04 against K18 (Bártek et al. 1991), these keratins may be helpful in diagnostic immunohistochemistry in cases of carcinomas with low keratin content such as small-cell lung cancer, to prove their epithelial nature. Regarding carcinoma subtyping, negative or weak/focal immunostaining for K8 and K18 may indicate squamous cell differentiation, although strong expression of these keratins can occur particularly in poorly differentiated squamous cell carcinomas (see below, chapter “Keratins as diagnostic markers in tumor pathology”). In the case of breast carcinomas, certain publications have reported a correlation between the level of K8 or K18 immunostaining and a favorable prognosis for the patients (see below, chapter “Keratins as diagnostic markers in tumor pathology”). Another clinical application of K8/K18 is the monitoring of fragments of these keratins in the serum as serological tumor markers to monitor cancer load, cancer progression, and response to therapy. Among the oldest of these markers are tissue polypeptide antigen (TPA) and tissue polypeptide-specific antigen (TPS) which have been recognized to correspond to a mixture of K8, K18, and K19 (Weber et al. 1984) and to K18 (Rydlander et al. 1996), respectively (for review, see Linder 2007). More recently, an apoptosis-specific fragment of K18 as detected by monoclonal antibody M30 (Leers et al. 1999) has become of increasing interest for distinguishing between necrosis and apoptosis and for the evaluation of the chemotherapy response of carcinomas by investigation of cancer patient serum, e.g. in prostate, breast, and lung cancer (Linder et al. 2004; Linder 2007). K7/K19: secondary keratins of simple epithelial cells Apart from K8/K18, keratins K7 and K19 are “additional” (secondary) and also widely distributed simple-epithelial keratins which are frequently but not always co-expressed. They typically occur as a keratin pair in simple ductal epithelia such as bile and pancreatic ducts (“ductal-type” keratins). However, in several epithelia lacking K7 such as intestinal epithelium, the type I keratin K19 must form a pair with the sole type II keratin K8. The type I keratin K19 is the smallest keratin and is exceptional since it widely lacks the non-α-helical tail domain typical for all other keratins (Bader et al. 1986). It may have evolved from keratinocyte keratins (Stasiak et al. 1989). As detectable by several specific and well-tested monoclonal antibodies (Karsten et al. 1985; Bártek et al. 1986; Nagle et al. 1986), K19 exhibits a rather broad tissue distribution. It is expressed in most simple epithelia (excluding parenchymatous cells such as hepatocytes, pancreatic acinar cells, and renal proximal tubular cells), notably in various ductal epithelia, in small and large intestinal epithelium, in gastric foveolar epithelium, and in mesothelium. Furthermore, it is present in most cells of pseudostratified epithelia and urothelium as well as in basal cells of non-keratinizing stratified squamous epithelia. Functionally, keratin K19 is dispensable since K19 knock-out mice were viable, fertile, and appeared normal (Harada et al. 1999). This is apparently due to functional compensation by K18, since only mice, double deficient for K18 and K19 exhibited a severe phenotype with trophoblast fragility and early embryonic lethality (Hesse et al. 2000). No mutation of the human K19 gene causing a disease has yet been found (Owens and Lane 2004). The expression of K19 may be induced in certain epithelia that normally lack this keratin by pathological alterations. One example is damage to renal proximal tubular epithelia by various types of injury as discussed above (Moll et al. 1991). K19 induction is also observed in suprabasal stratified squamous epithelial cells of oral mucosa with epithelial dysplasia (Lindberg and Rheinwald 1989; for further references, see Moll 1998), but also with inflammation (Bosch et al. 1989; Moll 1993), so that K19 cannot be used as a specific marker for dysplasia in oral mucosa. In carcinomas, K19 is widely expressed in both adenocarcinomas and squamous cell carcinomas and therefore is not extensively used as an immunohistochemical marker for carcinoma subtyping. One example for such application may be, in liver tumors, the distinction of hepatocellular carcinomas, which show little expression of K19, from cholangiocarcinomas and adenocarcinoma metastases, which strongly stain for this keratin (Balaton et al. 1988; Goldstein and Bosler 2006; see Table 3). The detection of soluble K19 fragments in the serum released by carcinoma cells by the CYFRA 21-1 assay has found broad clinical application as a marker to monitor treatment and evaluate response to therapy and has proven particularly useful in the case of squamous cell carcinomas of the lung (for review, see Barak et al. 2004, Gu and Coulombe 2007). The type II keratin K7, another “ductal-type” keratin, has a basically similar but somewhat more restricted tissue distribution as compared to K19 (Moll et al. 1982b; Ramaekers et al. 1990). Like K19, it is expressed in several simple ductal epithelia, in mesothelium, in pseudostratified epithelia (preferentially in luminal cells), and in urothelium but absent in parenchymatous cells such as hepatocytes. However, K7 is sparsely expressed or absent in gastric foveolar epithelium, intestinal epithelium, and stratified squamous epithelia. Human and mouse K7 genes have been characterized (Glass and Fuchs 1988; Smith et al. 2002), but mutations or disease associations have not yet been reported (Owens and Lane 2004). Several monoclonal antibodies against K7 have been described, some of which (e.g. Ks7.18, OV-TL12/30) are well reactive with formalin-fixed, paraffin-embedded tissues. Of these, clone OV-TL12/30 (van Niekerk et al. 1991) appears to elicit the broadest span of immunoreactivity and has found wide application in diagnostic tumor pathology, especially in cases where primary tumors or metastases are uncertain. Since the majority of carcinomas are K7 positive, negative reactions are of particular diagnostic significance. One main point of diagnostic utility of this keratin is the negative (or weak/focal) K7 immunostaining in colorectal adenocarcinomas (see Fig. 3c), in contrast to the strong staining in most other adenocarcinomas (see Fig. 3e; Moll et al. 1992, 1993b; Chu et al. 2000; Chu and Weiss 2002b; Tot 2002). This is especially valuable for classifying adenocarcinoma metastases with regard to their possible primary tumor. Low K7 expression (negative or weak immunostaining, or staining of a minor proportion of tumor cells) is also a characteristic feature of conventional (clear-cell) renal cell carcinomas (as opposed to papillary and chromophobe carcinomas) (Moll 1998; Skinnider et al. 2005) and of (non-cervical) squamous cell carcinomas. The validity in predicting the primary tumor in cases of unclear metastases is significantly increased when K7 is used in combination with K20 (see below) since many carcinomas exhibit characteristic K7/K20 phenotypes (Moll et al. 1992, Moll et al. 1993b; Chu et al. 2000; Chu and Weiss 2002b; Tot 2002; Dabbs 2006).Fig. 3Keratins in simple epithelia and adenocarcinomas (paraffin sections of human tissues; avidin–biotin complex peroxidase staining). Keratin K20 is a characteristic and prominent keratin of the foveolar epithelium of the gastric (a) and colorectal (b) mucosa. The K7−/K20+ phenotype of the normal mucosa is mostly maintained in primary and metastatic colorectal adenocarcinomas. This is shown here for a skin metastasis (on the head) of a poorly differentiated adenocarcinoma: the tumor cells are negative for K7 (c) but positive for K20 (d, left portion of the figure), even including a tumor cell cluster that has invaded a lymphatic capillary (d, right upper corner). This phenotype is strongly suggestive of colorectal origin. A primary tumor in the right colon was detected later. Liver metastasis of a ductal adenocarcinoma of the pancreas with typical keratin pattern, showing extended expression of K7 (e) and staining of scattered tumor cells for K20 (f). Magnifications: a, d ×80; b, c, e, f ×140 K20: keratin of gastrointestinal epithelium, urothelium, and Merkel cells K20 is the simple-epithelial keratin with the most restricted expression pattern. Although it appeared in our early cytoskeletal preparations of intestinal epithelial cells as a quite prominent protein spot of ~46 kDa, where we tentatively designated it as “IT protein” (from intestinal; Moll et al. 1982b), we succeeded rather late in identifying it as a type I keratin (Moll et al. 1990). For pair formation, its type II partner usually appears to be K8. The remarkable expression spectrum of K20 among normal tissues comprises gastric foveolar epithelium (Fig. 3a) and small and large intestinal epithelium (Fig. 3b; “gastrointestinal-type” keratin) and, in addition, the urothelium and certain neuroendocrine cells, in particular Merkel cells of the skin. In human embryogenesis, it appears in the small intestinal epithelium at embryonic week 8 (Moll et al. 1993b). Within the gastrointestinal and urothelial epithelial tissues, K20 is absent from the stem cell compartment and appears to be switched on during terminal differentiation and thus is most prominent in small intestinal villus-lining and large intestinal surface-lining epithelia (Fig. 3b) and in urothelial umbrella cells. Although K20 is a keratin typically expressed in simple epithelia it is also found in the lone basally located Merkel cells of the epidermis and hair follicle outer root sheath (Moll et al. 1992, 1995). Our knowledge about the function of K20 still is limited. Mutations of human K20 or associated diseases have not yet been described (Owens and Lane 2004). Transgenic experiments suggest a role for K20 in maintaining keratin filaments in intestinal epithelia (Zhou et al. 2003). In mouse small intestinal epithelium, K20 phosphorylation on serine 13 is induced during apoptosis and tissue injury and thus may serve as stress marker (Zhou et al. 2006). Among the K20-specific monoclonal antibodies available, clone Ks20.8 reacts well on routine paraffin sections (Moll et al. 1992) while clone Ks20.10 reacts with the K20 homologue of rodents (Moll 1993). K20 is a potent immunohistochemical marker in tumor pathology since its peculiar expression spectrum is essentially maintained in the corresponding primary and metastatic carcinomas (Moll et al. 1992), and the K20 clone Ks20.8 has become part of the routine antibody panel in most pathology laboratories. Most colorectal adenocarcinomas (Fig. 3d), the majority of gastric adenocarcinomas, the majority of transitional cell carcinomas, as well as most Merkel cell carcinomas are K20-positive (Moll et al. 1992, 1993b; Miettinen 1995; Chu et al. 2000; Chu and Weiss 2002b; Goldstein and Bosler 2006). In a few other carcinoma types, variable and focal K20 expression is seen, notably in ductal adenocarcinomas of the pancreas (Fig. 3f) and in adenocarcinomas of the biliary tract including cholangiocarcinomas of the liver (Moll et al. 1992; Miettinen 1995; Chu et al. 2000). Among ovarian carcinomas, K20 is mainly detected in the mucinous type. Most other carcinomas, including adenocarcinomas, irrespective of their morphology, are essentially negative for K20. Thus, significant K20 positivity of a metastatic adenocarcinoma is predictive of a primary tumor in the gastrointestinal or pancreaticobiliary tract. It should be noted that the diagnostic value is increased when the markers K20 and K7, are applied in combination. For example, a K7−/K20+ phenotype of an adenocarcinoma metastasis (Fig. 3c, d) strongly favors a colorectal origin. Some colorectal adenocarcinomas co-express K7 in addition to K20, but as a general rule, the level of K20 immunostaining exceeds that of K7 (see below, chapter “Keratins as diagnostic markers in tumor pathology”). Using RT-PCR analyses, K20 mRNA can be detected in cell preparations from bone marrow and peripheral blood from some colorectal cancer patients, indicating the presence of disseminated tumor cells that maintain K20 expression, and a number of clinical studies have shown that this is correlated with a worse prognosis (Soeth et al. 1996; Funaki et al. 1998; Wyld et al. 1998; Koch et al. 2005; Katsumata et al. 2006; Friederichs et al. 2007). In addition to patients with metastatic colorectal carcinoma, K20 expression in the peripheral blood was also detected in patients with metastatic gastric and pancreatic adenocarcinoma but hardly in patients with metastatic lung carcinoma (Chausovsky et al. 1999), underlining the tumor type-specific expression of this keratin. Stratified epithelia K5/K14: major keratins of basal keratinocytes The type-II keratin K5 and the type-I keratin K14 form the primary keratin pair of the keratinocytes of stratified squamous epithelia, including the epidermis as well as mucosal non-keratinizing stratified squamous epithelia (Moll et al. 1982b). They are strongly expressed in the undifferentiated basal cell layer containing the stem cells and are down-regulated in the differentiating suprabasal cell layers (Fig. 4a; Fuchs and Green 1980). Otherwise, in the widely well stratified follicular outer root sheath, K5 and K14 are uniformly expressed throughout all layers. Moreover, the follicular companion layer, which is directly adjacent to the outer root sheath (and formerly considered as “innermost layer of the outer root sheath”), is completely negative for both of these keratins and expresses an own special keratin, K75 (see below). Ultrastructurally, K5/K14 keratin filaments are bundled as tonofilaments and attached to desmosomes and hemidesmosomes. The mouse K5 promoter is frequently used in transgenic experiments to promote epidermis-specific expression of transgenes (e.g. Oki-Idouchi and Lorenzo 2007). In addition to their occurrence in keratinocytes, K5 and K14 are expressed in basal and myoepithelial cells of complex and glandular epithelial tissues (Purkis et al. 1990). On the other hand, these keratins are absent from most simple/one-layered epithelia, with very few exceptions, notably the mesothelium lining serous cavities (Moll et al. 1989) and the amnion epithelium.Fig. 4Keratins in stratified squamous epithelia and squamous cell carcinomas (paraffin sections of human tissues; avidin–biotin complex peroxidase staining). In the epidermis as an example of a normal stratified squamous epithelium, the basal cell layer contains abundant keratin K5 (a) whereas the differentiating suprabasal compartment strongly stains for K10 (b; note the negative basal cell layer). Lymph node metastasis of a squamous cell carcinoma of the head and neck region, expressing K5 (c; more intensely in the peripheral tumor cell layers) as well as K6 (d; particularly strongly in central tumor cells) as signs of their keratinocyte origin. Keratin K5 is also maintained in a lymph node micrometastasis of a squamous cell carcinoma of the head and neck region (e) and in a lymph node metastasis of an undifferentiated nasopharyngeal carcinoma with dissociated growth pattern of the tumor cells (f), in these examples being a diagnostically helpful feature. Magnifications: a, b ×160; c–e ×80; f ×140 Several monoclonal antibodies (MAbs) specific for K5 and K14 have been described that helped to reveal their exact tissue distribution. These antibodies include MAb AE14 (Moll et al. 1989) against K5 and MAbs LL001 and LL002 against K14 (Purkis et al. 1990). The best performance on paraffin sections is displayed by MAb D5/16B4 (Lobeck et al. 1989; Demirkesen et al. 1995) which—although being often regarded as “K5/K6 antibody”—specifically recognizes K5 (Böcker et al. 2002). The functional importance of K5 and K14 for the physical stability of the epidermis has become clearly evident by the recognition that dominant-negative mutations of the K5 or the K14 gene cause the hereditary blistering skin disease epidermolysis bullosa simplex (EBS) (for reviews, see Omary et al. 2004; Lane and McLean 2004). Most of these mutations are missense or small in-frame deletion mutations, primarily affecting the keratin rod domain. The presence of mutated K5 or K14 results in increased fragility of the basal keratinocytes so that even mild physical trauma leads to intraepidermal cytolysis of basal cells and the formation of fluid-filled blisters. These patient-related findings were preceded by the experimental demonstration that expression of mutant K14 in transgenic mice causes abnormalities similar to EBS (Vassar et al. 1991). The expression spectrum of K5 and K14 in tumors corresponds well to the patterns in normal epithelia. Thus, most squamous cell carcinomas (Fig. 4c, e) as well as malignant mesotheliomas strongly express these keratins whereas little, focal, or no expression is found in adenocarcinomas (Moll et al. 1982b, 1989; Moll 1998; Chu and Weiss 2002a, b). Hence, these keratins, in particular K5, have found several lines of diagnostic application in pathology, which has been aided by the availability of a highly sensitive and specific, robust, paraffin-suited MAb (D5/16B4; see above). Pertinent examples are the recognition and diagnosis of poorly differentiated squamous cell carcinomas, including micrometastases in lymph nodes (Fig. 4e), of undifferentiated nasopharyngeal carcinomas which may be diagnostically difficult due to their dissociated growth pattern (Fig. 4f), and of malignant mesotheliomas. Thus, K5 immunostaining allows the distinction of the small cell type of squamous cell carcinoma of the lung, which is K5+, from a small cell carcinoma or a poorly differentiated adenocarcinoma, both of which are K5− (Moll 1998; Chu and Weiss 2002a), and the distinction of a malignant mesothelioma of the pleura (K5+) from a pulmonary adenocarcinoma with pleural involvement (K5−) (Moll et al. 1989; Ordonez 1998; Yaziji et al. 2006). In well and moderately differentiated squamous cell carcinomas, K5 is preferentially localized in the peripheral layers of the tumor cell formations (Fig. 4c), corresponding to the K5 expression in the basal cell layer of normal stratified squamous epithelia. Focal K5 expression may be observed in certain adenocarcinoma types, notably in adenocarcinomas of the endometrium, the ovary, and the pancreas, which seems to be related to their potency for focal squamous differentiation (Moll 1998; Chu and Weiss 2002a, 2002b). Much interest has evolved regarding the role of K5 in breast pathology in several aspects, including the identification of myoepithelial cells, the classification of proliferative lesions (Otterbach et al. 2000), and the recognition of a certain subtype of invasive ductal breast carcinoma (see below, chapter “Keratins as diagnostic markers in tumor pathology”). In prostate pathology, the diagnosis of prostatic adenocarcinoma is supported by the immunohistochemical demonstration of absence of K5-positive basal cells (Abrahams et al. 2002, 2003). K15: basal keratinocyte keratin and hair follicle stem cell “marker” K15 was first identified as a minor keratin of human epidermis, by gel electrophoresis of cytoskeletal preparations (Moll et al. 1982b, c). Its sequence demonstrated its assignment to the type I keratins (Leube et al. 1988). Only later its cellular distribution was recognized, and it was uncovered that K15 is a specific basal cell component of the epidermis (Moll et al. 1993a; Lloyd et al. 1995) and other stratified squamous epithelia (Waseem et al. 1999) (see below for its significance in the hair follicle). Frequently, K5 and K14 can also be detected in the lower suprabasal cell layers (see above and Fig. 4a). Whereas the mRNA synthesis of these keratins is restricted to the basal layer, the K5 and K14 proteins remain integrated in the complex keratin cytoskeleton for some time when cells leave the basal compartment. Thus, they may be stained by immunohistochemistry in more or less suprabasal layers depending on the epitope of the antibody used. In comparison, K15 seems completely restricted to the basal cell layer of stratified squamous epithelia (Lloyd et al. 1995; Waseem et al. 1999) where it can form heteropolymeric filaments with K5 (Lloyd et al. 1995). MAbs recognizing K15 include clone LHK15 (Waseem et al. 1999) and clone C8/144B (Lyle et al. 1998, 1999). The latter in fact is a surprising and exciting MAb originally made against the lymphocyte antigen CD8 which cross-reacts with K15 especially in the basal keratinocytes of the hair follicle bulge region, where the respective stem cells are assumed. Also polyclonal K15 antibodies have furthermore been described (Moll et al. 1993a; Kurzen et al. 2001, Langbein et al. 2008). K15 expression in basal cells is downregulated in activated epidermal keratinocytes such as in organotypic cultures and in hyperproliferation (Waseem et al. 1999) or upon wounding (Porter et al. 2000). Human mutations or knock-out mice for K15 have not yet been described. As to the occurrence of K15 in tumors, the literature is inconclusive as to whether there is differential expression of K15 among cutaneous tumors (benign follicular tumors versus basal cell carcinomas), possibly due to the use of different antibodies1 (see also above), and thus a putative diagnostic application of K15 immunohistochemistry in dermatopathology is still open (Kanitakis et al. 1999; Jih et al. 1999; Porter et al. 2000; Kurzen et al. 2001). Currently, the most interesting feature with K15 is the discovery that within the hair follicle at least one antibody against this keratin detects putative stem cells residing in the hair follicle bulge and thus might be used as a stem cell marker in hair follicle biology (Lyle et al. 1999). K6/K16: keratins of hyperproliferative keratinocytes inducible in “activated” epidermis By gel electrophoresis, the type-II keratin K6 and the type-I keratin K16 have been identified in epidermis only in samples from plantar glabrous skin while in hairy skin these keratins appeared to be absent in interfollicular epidermis but were clearly present in hair follicle outer root sheath (Moll et al. 1982b, c; Langbein and Schweizer 2005) and companion layer (Winter et al. 1998; Langbein and Schweizer 2005; Gu and Coulombe 2007). In nail epithelia, K6 and K16 are constitutive components (Heid et al. 1988b; Lane and McLean 2004; Perrin et al. 2004). K6 and K16 are also consistently expressed in non-keratinizing stratified squamous epithelia (Moll et al. 1982b). They are often but not always co-expressed as a keratin pair (Moll et al. 1982b; Langbein et al. 2005). Molecular genetic studies have revealed that in humans three isoforms of K6 exist, K6a, K6b, and K6c, encoded by distinct genes (Rogers et al. 2005; Schweizer et al. 2006). MAb KA12 is an antibody which most probably stains at least keratin K6a isoform and reacts well with paraffin sections (Demirkesen et al. 1995; Langbein et al. 2003; Schmelz et al. 2005) although isoform-specific antibodies against K6a, K6b or K6c will hardly be made because of their extremely high peptide sequence homology. Using this MAb, plantar epidermis shows extended albeit heterogeneous expression of K6a, while interfollicular epidermis is negative or exhibits only some positive suprabasal cell groups (Demirkesen et al. 1995; Swensson et al. 1998). Non-keratinizing stratified squamous epithelia express K6 uniformly in all suprabasal cell layers. A MAb against K16 has also been described (Leigh et al. 1995). In early immunohistochemical studies using keratin group-specific MAbs, Weiss et al. (1984) demonstrated the induction of K6 (56-kd keratin) and K16 (48-kd keratin) in various hyperproliferative epidermal disorders, suggesting that these keratins may be molecular markers for hyperproliferative keratinocytes. More recent experimental studies showed that after skin wounding, K6 and K16 are rapidly induced within 6 h in human keratinocytes at the wound edge, before migration and regeneration begins (Paladini et al. 1996). The particular cell biological properties of these keratins may confer to activated keratinocytes on the one hand a moderate level of mechanical scaffolding and on the other hand sufficient plasticity required during migration and re-epithelialization (Wawersik et al. 2001). K6a knockout mice showed delayed re-epithelialization after skin wounding (Wojcik et al. 2000). Upon K6a/K6b double knock-out, mice exhibit epithelial disintegration and white plaques in the dorsal tongue epithelium (Wong et al. 2000; Wojcik et al. 2001). The absence of a hair and nail phenotype in these mice—as might be expected from the occurrence of K6 in these appendages—has been shown to be due to the presence and compensatory function of other K6-related keratins in hair follicles and nails, such as mK6hf (Wojcik et al. 2001; now mK75) or mK6irs1—mK6irs4 (now mK71–mK74; cf. Langbein et al. 2002, 2003). In man, mutations in K6a or K16 have been proven to give rise to the hereditary disorder pachyonychia congenita type 1 (Jadassohn–Lewandowsky form) that manifests with thickened nails, palmoplantar hyperkeratosis, and oral leukoplakias (McLean et al. 1995; for further references, see Lane and McLean 2004). Notably, the pathologic changes affect those tissues that constitutively express K6a and K16 (see above) but not the interfollicular epidermis. Thus, K6/K16 are constitutive keratins of stratified epithelia built up by keratinocytes of relatively high proliferative state such as mucosal tissues, palmoplantar epidermis, and certain skin appendages. On the other hand, they are “stress-inducible” keratins in interfollicular epidermis, being rapidly switched on e.g. after injury and UV-irradiation or being present also in inflammation and in hyperproliferative disorders. Expression of these keratins is not restricted to stratified squamous epithelia but may also be observed in certain glandular structures. Thus, K6 (most probably K6a) and K16 are expressed in ductal luminal cells as well as in some secretory cells of human eccrine sweat glands (Fig. 5; Demirkesen et al. 1995; Langbein et al. 2005). K6 has been detected in subpopulations of luminal cells of the mouse mammary gland (Grimm et al. 2006) and the human mammary gland; in the latter, also in ductal myoepithelial cells (Hesse 2003; Langbein et al. 2005). Recently, a population of K6-positive cells in the prostate gland with high potential for proliferation and differentiation has been described (Schmelz et al. 2005).Fig. 5Keratin K77 in eccrine sweat glands and adnexal tumors. K77 mRNA and protein by in situ hybridization (ISH, a, a′) and indirect immunofluorescence (IIF, b-b′′) microscopy. This keratin is specifically expressed in the luminal cells (lc) of the intraglandular (igd), the intradermal (idd), the sweat duct ridge (sdr) and the intraepidermal/acrosyringial (ied/ac) duct of eccrine sweat glands of plantar skin. Peripheral duct cells are negative. The secretory portions (asterisks) are K77-negative. sb str. basale, ss str. spinosum, sg str. granulosum, sc str. corneum, gl glandular region with secretory and intraglandular duct portions. c Expression scheme of keratins of the eccrine sweat gland. d-d′′ Immunoperoxidase staining of K77 in the luminal cells of the eccrine sweat gland duct of the foot sole epidermis. K77 is also expressed in the tubular structures (arrows) of eccrine tumors such as syringoma (e) and cylindroma (f). Bars 100 μm In the pseudostratified epithelia of respiratory mucosal tissues, K6 and K16 are highly upregulated in squamous metaplasia (Leube and Rustad 1991; Stosiek et al. 1992). Among tumors, K6 and K16 are typically and strongly expressed in squamous cell carcinomas of different sites (Moll et al. 1982b; Moll 1998), preferentially in inner, maturing layers of the tumor cell nests (Fig. 4d). Although low expression of these keratins may be found in adenocarcinomas such as occasionally in adenocarcinomas of the uterine cervix (Smedts et al. 1993) and in less than 20% of invasive breast carcinomas (Wetzels et al. 1991), K6 as detected by MAb KA12 may be suitable—in addition to K5—as another immunohistochemical marker of squamous differentiation in poorly differentiated squamous cell carcinomas (Moll 1998). K17: keratin of basal/myoepithelial cells and inducible in “activated” keratinocytes The type I keratin K17 was identified in our early gel electrophoretic studies as a major keratin of basal cell carcinomas of the skin which was also present in the normal pilosebaceous tract but not in normal epidermis (Moll et al. 1982c). Further protein analyses showed its presence in squamous cell carcinomas of various origins as well as in normal glandular tissues (such as sweat glands and breast) but its apparent absence also from non-keratinizing stratified squamous epithelia (Moll et al. 1982b, 1983). The unique cell type distribution of this keratin became apparent after establishment of a specific MAb, clone E3 (Troyanovsky et al. 1989). Broad tissue screening revealed its selective expression in basal and myoepithelial cells of complex tissues, including various glands, respiratory epithelium, and urothelium (Troyanovsky et al. 1989, 1992). Thus, K17 may be regarded as a “basal-/myoepithelial cell keratin”. In the hair follicle, confirming and extending previous biochemical findings (Moll et al. 1982c), K17 has been localized as a prominent component of the suprabasal cell layers of the outer follicular root sheath (Winter et al. 1998; Langbein and Schweizer 2005; Langbein et al. 2006; Tong and Coulombe 2006). It is also present in nail bed and nail matrix epithelia (McGowan and Coulombe 2000; Perrin et al. 2004). Immunohistochemistry also confirmed the essential absence of K17 from adult interfollicular epidermis, but, interestingly, revealed its specific expression in the specialized epidermal keratinocytes of the sensory Merkel cell-associated “haarscheiben” organs (Moll et al. 1993a); this keratin thus may be applied as a sensitive haarscheiben marker in studies of cutaneous neurobiology. In contrast to the adult, K17 is a prominent component of fetal epidermis (Moll et al. 1982a) as well as of cultured epidermal cells (Weiss et al. 1984). Another interesting feature of K17 is its inducibility after skin injury: after K6/K16 (see above), K17 is switched on in regenerating and migrating epidermal keratinocytes upon wound healing (Paladini et al. 1996). Its functional importance in wound healing is suggested by the observation that K17 knockout mouse embryos show a delay in the closure of surface ectoderm wounds (Mazzalupo et al. 2003). Recent transgenic experiments have shown that K17 can bind to the adaptor protein 14-3-3σ and influences cell growth and size of mouse keratinocytes by regulating protein synthesis (Kim et al. 2006). The aforementioned expression of K17 in the pilosebaceous tract (in the follicular outer root sheath, companion layer, medulla and sebaceous gland; for review, see Langbein and Schweizer 2005) has also proven to be functionally relevant. K17 null mice develop transient severe alopecia in early postnatal life, correlating with hair fragility and apoptosis in hair matrix cells (McGowan et al. 2002). The same group later showed that K17 modulates hair follicle cycling by delaying apoptosis, whereby K17 is functionally linked with TNFα signaling (Tong and Coulombe 2006). Hereditary human diseases due to K17 mutations have been identified (for references, see Lane and McLean 2004), most notably pachyonychia congenita type 2 (Jackson–Lawler form). The phenotype of this genodermatosis includes thickened nails and pilosebaceous cysts. Another condition related to K17 mutations is steatocystoma multiplex, in which patients present with multiple hair follicle-associated cysts. These genodermatoses obviously are related to the expression and functional importance of K17 in pilosebaceous and nail (Perrin et al. 2004; Langbein and Schweizer 2005) epithelia. Since in keratinocytes K17 is—like K6 and K16 (see above)—an inducible keratin upon stress, injury, or inflammation, it is not surprising that squamous cell carcinomas consistently express these three keratins (Moll et al. 1982b, 1983; Chu and Weiss 2002b). As most normal stratified squamous epithelia lack K17, its presence in the corresponding tumors may be regarded as neo-expression during tumorigenesis. In the uterine cervix, K17 is expressed in cervical intraepithelial neoplasia but since it is already present in endocervical reserve cells (Weikel et al. 1987) and immature squamous metaplasia, it has not yet become a routine diagnostic marker (Martens et al. 1999). Among adenocarcinomas, focal K17 expression seems to be particularly characteristic of pancreatic ductal adenocarcinomas (Real et al. 1993; Moll 1998), which may become useful for their distinction from other adenocarcinomas (Chu and Weiss 2002b). In ductal breast carcinomas, K17 is expressed in a minor subset of tumor cases (Malzahn et al. 1998), now recognized to correspond to the basal-like subtype as defined by global gene expression data (see below, chapter “Keratins as diagnostic markers in tumor pathology”). K1/K10: major keratins of keratinocyte differentiation and keratinization In the epidermis, the transition of keratinocytes from the proliferative basal cell layer to the postmitotic suprabasal spinous cell layers in the process of terminal differentiation and keratinization is characterized by a profound change in keratin expression. This involves a switch from expression of the basal cell keratins (K5, K14, K15) to the suprabasal epidermal keratins, the type II keratin K1 and subsequently the type I keratin K10 (Fig. 4b; Fuchs and Green 1980; Moll et al. 1982b; Tseng et al. 1982, Weiss et al. 1984; Roop 1987, Stoler et al. 1988). This is one of the classical examples for the carefully regulated differentiation-specific expression of keratin proteins. Ultrastructurally, keratin filaments composed of the pair K1/K10 form particularly dense bundles which are so characteristic of suprabasal epidermal keratinocytes (Fig. 1d). Clearly, this imparts mechanical integrity to the cells and the whole epidermis. In addition, however, there seem to exist further functional roles, as experimental data have demonstrated that K10 specifically inhibits proliferation and cell cycle progression of keratinocytes (Paramio et al. 1999; Koch and Roop 2004) and loss of K10 leads to increased keratinocyte turnover (Reichelt et al. 2004; for review, see Magin et al. 2007). The importance for epidermal integrity is underscored by the fact that point mutations in K1 and K10 are associated with the blistering disorder epidermolytic hyperkeratosis/bullous congenital ichthyosiform erythroderma (BCIE), initially presenting with skin blisters but later with thickened ichthyotic skin (for reviews, see Lane and McLean 2004; Omary et al. 2004). As expected, the suprabasal cells become fragmented easily and, in addition, the epidermis becomes hyperproliferative and hyperkeratotic. K1 mutations are notably heterogeneous and may result in diverse overlapping, often relatively mild phenotypes (Lane and McLean 2004). In transgenic mouse experiments, mutation or knock-out of K10 results in a phenotype similar to human epidermolytic hyperkeratosis (Fuchs et al. 1992; Porter et al. 1996). Despite their association with terminal epidermal differentiation and keratinization, K1 and K10 may be focally expressed in suprabasal cells of internal noncornifying stratified squamous epithelia (for references, see Moll 1998). They are also a typical component of cells of eccrine sweat gland ducts (Fig. 5; Langbein et al. 2005). Surprisingly, although the differentiated parts of the hair follicle, such as suprabasal outer root sheath, upper companion layer, upper inner root sheath or the hair fiber, are heavily keratinized structures, all of them are free of K1/K10. Both typical epidermal keratins are completely lost in the infundibulum. Among various antibodies against K1 and K10 described in the literature, MAb DE-K10 against K10 is particularly suitable for application with paraffin-embedded tissues (Ivanyi et al. 1989). In squamous cell carcinomas focal expression of K1 and K10, usually in relation to maturation and keratinization, can be observed regardless of whether they are derived from the skin or from internal organs (for references, see Moll 1998). They are more sparse in poorly differentiated tumors but may still be detected in about 50% of cases of oral and pharyngeal squamous cell carcinomas (Moll 1998). However, quantitatively, squamous cell carcinomas rather embark on an alternative maturation pathway characterized by abundant expression of K6 and K16 (see above; Fig. 4d). Overall, K1 and K10 can be regarded as “keratinization markers” of keratinocytes. These keratins have not yet been routinely applied to tumor diagnosis except for some special aspects of skin tumors (Yuspa et al. 1991). K9: palmoplantar epidermal differentiation keratin The type I keratin K9 is a highly specific keratin of terminally differentiating keratinocytes of palmoplantar epidermis where it is abundantly, albeit heterogeneously expressed (Moll et al. 1987; Langbein et al. 1993). At other body sites, there may be extremely sparse and focal expression in upper epidermal layers (Moll et al. 1987). Thus K9, forming a pair with K1, appears to reflect a special program of keratinocyte differentiation associated with particular mechanical reinforcement (Swensson et al. 1998). Therefore it is not surprising that mutations in the K9 gene are associated with a disorder of the skin of the palms and soles, epidermolytic palmoplantar keratoderma, which manifests itself as cytolysis and epidermal thickening (Reis et al. 1994; Torchard et al. 1994; for review, see Lane and McLean 2004). In its sequence, K9 is most closely related to K10 (Langbein et al. 1993). MAbs against K9 (clones Ks9.70, Ks9.216; PROGEN, Heidelberg, Germany) are available (Langbein et al. 2005, 2006, 2008). Immunostaining for K9 has significance for characterization of palmoplantar keratinocyte direction of transplants (Compton et al. 1998; Stoner and Wood 1999) and of special genodermatoses (McLean and Irvine 2007) but there is no relevance of K9 in tumor diagnosis. It should be noted that the occurrence of keratin K9 (and K1 and K10) may give rise to problems in biochemical practice as biochemicals or buffers are sometimes contaminated with these keratins derived from abraded “horny particles”, i.e. terminally differentiated skin keratinocytes (“laboratory dust”; Clark et al. 1971; Fox et al. 2008). K2: keratin of highly differentiated, advanced epidermal keratinocytes K2 (formerly K2e, see before and Table 1) is another keratin specific for the advanced terminal differentiation process of epidermal keratinocytes. Being widely distributed over most body sites, this type II keratin is expressed late, at an advanced stage of differentiation, in the uppermost epidermal layers (upper stratum spinosum, stratum granulosum) to a variable extent (Collin et al. 1992a). K2 is not expressed in follicular skin adnexal structures like the late compartments of outer or inner root sheath. Correspondingly, mutations in K2 have been associated with ichthyosis bullosa of Siemens, a blistering disease showing cytolysis in superficial epidermal layers (for references, see Lane and McLean 2004). MAbs specific for K2 (clones Ks2.342.7.1, Ks2.398.3.1; PROGEN, Heidelberg, Germany) have been described (Langbein et al. 2005, 2006, 2008). K3/K12: keratins of the corneal epithelium The K3 (type II)/K12 (type I) pair is the cell type-specific and differentiation-related keratin pair of the corneal epithelium. These keratins are expressed in all corneal epithelial cell layers, whereas in the limbus corneae only suprabasal cells are positive and the basally located corneal stem cells are K3/K12 negative (Schermer et al. 1986; Pitz and Moll 2002). Mutations in these keratins give rise to Meesmann’s corneal dystrophy characterized by intraepithelial microcysts in the corneal epithelium (Irvine et al. 1997; for further references, see Lane and McLean 2004). K12 knock-out mice have a mechanically fragile, easily detachable corneal epithelium (Kao et al. 1996). Antibodies against these corneal keratins described include MAb AE5 (PROGEN, Heidelberg, Germany) which reacts with K3 and additionally with the related K76 (formerly K2p; see below; Collin et al. 1992b) and MAb AK12 which recognizes K12 (Chaloin-Dufau et al. 1993). K4/K13: keratins of mucosal stratified squamous epithelial cells In internal stratified squamous epithelia which mostly are non-keratinizing, a highly characteristic keratin pair indicates the mucosal path of keratinocyte differentiation, i.e. the type II keratin K4 and the type I keratin K13 (Moll et al. 1982b; Cooper et al. 1985). Immunohistochemical studies using specific MAbs—such as MAb 6B10 against K4 (van Muijen et al. 1986) and MAbs 1C7, 2D7 (van Muijen et al. 1986) and Ks13.1 (Moll et al. 1988b) against K13—revealed the presence of K4 and K13 in the entire suprabasal compartment of mucosal stratified squamous epithelia, whereas the basal compartment is positive for K5/K14. Interestingly, K4/K13 is completely absent in the epidermis and adnexal structures. Keratin K13 is also expressed in the urothelium as a major component of the basal and intermediate cells whereas it is lost in the superficial umbrella cells (that switch on K20 expression; see above). Keratin K4 is also—heterogeneously—expressed in columnar luminal cells of the pseudostratified respiratory epithelium and in the non-stratified simple epithelial cells of e.g. pancreatic ducts (van Muijen et al. 1986; Moll 1993). Functionally, K4 and K13 appear to be important particularly as components of mucosal stratified squamous epithelia. Mutations in these keratins, lying in the helix initiation or termination motifs (HIM or HTM, respectively), have been shown to cause the hereditary disorder white sponge nevus of Cannon (for references see Lane and McLean 2004). This mucosal disorder presents with white plaques mainly on the buccal mucosa, histologically showing thickened spongy epithelium with hydropic swelling of suprabasal epithelial cells. Here again, the clinical manifestation of pathological alterations of keratins well reflects their tissue distribution. It is not surprising that squamous cell carcinomas derived from the epidermis essentially lack K4 and K13 (Kuruc et al. 1989; for further references see Moll 1998). However, in contrast to what might be expected, they are not major components of, but are only focally and variably expressed in squamous cell carcinomas of the head and neck, with more pronounced expression in poorly differentiated cases (Moll 1998). Instead, the predominant maturation-associated keratins expressed by these tumors are the hyperproliferative keratins K6 and 16 (see below, chapter “Keratins as diagnostic markers in tumor pathology”). Corresponding to its characteristic expression in normal urothelium, K13 is maintained—at least focally—in most transitional cell carcinomas of the urinary tract (Moll et al. 1988b; for further references, see Moll 1998). K13 may be part of a panel of markers (which also includes K20) useful in the histological diagnosis of metastatic transitional cell carcinomas (see below, chapter “Keratins as diagnostic markers in tumor pathology”). As to adenocarcinomas, it is noteworthy to mention the frequent expression of K4 in ductal adenocarcinomas of the pancreas (Schüssler et al. 1992; Real et al. 1993; Moll 1998) and in a subpopulation of poorly differentiated invasive ductal breast carcinomas (Malzahn et al. 1998). K76, K77: keratins with very special expression sites K76 (previously designated K2p) is specifically expressed in suprabasal cell layers of oral masticatory epithelium, i.e. the slightly orthokeratinized stratified squamous epithelium lining the gingiva and the hard palate (Collin et al. 1992b). Because of the failure of specific antibodies for long time, which are now available (PROGEN, Heidelberg, made by L.L.), no tumor studies have been done yet. The expression pattern of keratin K77 (previously designated K1b) was very surprising and extremely restricted. This keratin is exclusively expressed in and restricted to the luminal cells of eccrine sweat gland ducts (Fig. 5a–d). All other epithelia and glands tested so far, including apocrine sweat gland, were negative (Langbein et al. 2005). Based on the extensive investigation of the keratin pattern, new aspects of eccrine sweat gland differentiation could be achieved (Langbein et al. 2005 and supplemental data therein). The high specificity of expression makes this keratin recommendable for using as an “eccrine duct marker” in tumor diagnostics. In a first study, this assumption could be confirmed by the investigation of “eccrine” adnexal tumors such as syringoma (Fig. 5e), poroma (Langbein et al. 2008) and cylindroma (Fig. 5f). K25, K26, K27, K28, K71, K72, K73, K74, K75: hair follicle-specific epithelial keratins Only recently it has become clear that some of the epithelial root sheaths of the hair follicle, the inner root sheath and the companion layer, are unique by their expression of a number of very special keratins (for review, see Langbein and Schweizer 2005; Langbein et al. 2006; Schweizer et al. 2007). Over the years, these keratins were not detected before by biochemical methods because of their quantitative “under-representation” when compared to the masses of hair or epidermal keratins in the tissue. The first of these new special keratins described was K75. This keratin was originally called K6hf, with “hf” indicating its expression site in the “hair follicle”, not knowing that this aspect of designation would be unfeasible later. K75 (K6hf) is a type II keratin and was, although closely related in its peptide sequence to K5, following the principles of the former keratin designation (cf. Collin et al. 1992a, b) by using the electrophoretic properties, designated as a “K6” keratin (Winter et al. 1998). K75 is specifically expressed in the companion layer of the hair follicle (Fig. 6a, h, i), a thin layer between the outer and the inner epithelial root sheath (Winter et al. 1998). As the expression of this keratin as monitored by its mRNA synthesis using in situ hybridization starts from the bulbar matricial compartment of the hair follicle and the protein is still existent in the upper differentiated part where the K75 mRNA is no longer synthesized, it was one first but doubtless indication that this structure is an own individual compartment of the hair follicle and not the “innermost layer of the outer root sheath” (see Winter et al. 1998; Langbein and Schweizer 2005). The only further structures in which K75 has been detected are the hair medulla of, e.g. beard hairs, the nail bed, and fungiform papillae of the tongue (Wang et al. 2003; Perrin 2007; Langbein and Schweizer 2005). A mutation in K75 appears to predispose to the common hair disorder pseudofolliculitis barbae, which is characterized by ingrown beard hairs with inflammation, induced by shaving (Winter et al. 2004; Schweizer et al. 2007), and to the loose anagen hair syndrome (Chapalain et al. 2002). Immunostaining for K75 has been reported in trichoblastomas and basal cell carcinomas (Kurzen et al. 2001) and has recently been observed in some squamoid cells of pilomatricomas (M. Divo, L. Langbein and R. Moll, in preparation), indicating special focal differentiation in these diverse cutaneous tumors. We have recently found sparse and focal expression of K75 in certain squamous cell carcinomas of inner organs (M. Divo, L. Langbein and R. Moll, in preparation).Fig. 6Immunofluorescence labeling of hair follicle-specific and hair keratins. The hair follicle-specific epithelial keratin K75 (a) is specifically found in the hair companion layer (cl) and in the medulla (med) of sexual (e.g. beard) hairs. K71 (b) is expressed in all compartments and K72 (c) in the cuticle (icu) of the hair inner root sheath (IRS). The hair keratin K85 (d) expression is found from the hair matrix to the upper cortex and the hair cuticle (cu), whereas K82 (e) is restricted to the hair cuticle. K86 (f) is an example for hair keratins expressed in the mid-to-upper hair cortex (co). g Hair keratin K81 is also expressed in the upper transitional cells of pilomatricoma. co cortex, dp dermal papilla, ORS outer root sheath. h, i Summary schemes of the expression of all hair and hair follicle-specific keratins in the human hair follicle. **K37 is found in the cortex of vellus hairs and medulla of sexual hairs. *K38 is heterogeneously expressed in the cortex. The keratin genes are designated according to the new keratin nomenclature (Schweizer et al. 2006) A set of four type I keratins (K25–K28; previous designations K25irs1–K25irs4, see Table  1) and four type II keratins (K71–K74; previous designations K6irs1–K6irs4, see Table 1) is highly specific for the inner root sheath (IRS) of the hair follicle (Fig. 6h, i). These IRS keratins are differentially and partially sequentially expressed in the various IRS compartments, the Henle layer, the Huxley layer, and the IRS cuticle. The keratinocytes of all three compartments synthesize the IRS keratins K71 (Fig. 6b, h, i), K25, K27 and K28. K74 is restricted to the Huxley layer, whereas K73, K72 (Fig. 6c, h, i) and K28 are sequentially expressed in the IRS cuticle (Langbein et al. 2002, 2003, 2006; Langbein and Schweizer 2005; Fig. 6h, i). Otherwise, no “classical” epithelial keratins were evidenced without any doubts in the IRS and earlier reports most probably showed (at that time not expected) cross-reaction of antibodies with at least one of these IRS keratins (cf. Langbein et al. 2006). Some of the IRS keratins—together with many others—have also been found in the hair medulla (see Schweizer et al. 2007 and Langbein et al., in preparation). In mice, spontaneous hair disorders due to mutations in K71 have been identified (for references, see Schweizer et al. 2007). Human hair disorders related to the IRS keratins have not yet been discovered. Monospecific antisera against all of these keratins are available (PROGEN, Heidelberg, Germany; Langbein et al. 2004, 2006). K31, K32, K33a, K33b, K34, K35, K36, K37, K38, K39, K40, K81, K82, K83, K84, K85, K86: keratins of the hair fiber (hair keratins) It is long known since the early period of keratin research that in the (hard) material of hairs, wool, nails, claws and feathers, the tremendous masses of keratin filaments (selection of the early literature: Odland 1953; Fraser et al. 1959; Rogers and Clarke 1965; Orfanos and Ruska 1968) are embedded in a matrix of cross-linking specialized keratin associated proteins (KAPs) with more than 85 genes in humans (for review, see Rogers et al. 2006). The special, more sulfur-rich keratin proteins constituting these filaments are the “hard” or “trichocytic” keratins. Originally, eight “major” (type I: Ha1-4, type II: Hb1-4) and two “minor” (Hax, Hbx) keratins were distinguished (Heid et al. 1986, 1988a, b). In the last 10 years many more, namely 17 members of this keratin subfamily have been identified first at their gene level (Rogers et al. 2004, 2005 and references therein) which are now generally referred to as the hair keratins (Langbein et al. 1999, 2001, 2007; Langbein and Schweizer 2005; Schweizer et al. 2006; see also Table 1). The conspicuous abundance of these proteins comprises eleven type I hair keratins (K31–K40; previous designations Ha1–Ha8, Ka35, Ka36; Langbein et al. 1999, 2001, 2007, Langbein and Schweizer 2005; Schweizer et al. 2006; see also Table 1) and six type II hair keratins (K81–K86; previous designations Hb1–Hb6; Langbein et al. 1999, 2001; Schweizer et al. 2006; see also Table 1; Fig. 6h, i). In the hair, they exhibit differential, complex and in many cases sequential expression patterns within the cuticle and the cortex (Langbein et al. 1999, 2001, 2007; Langbein and Schweizer 2005; Schweizer et al. 2007; for medulla: Langbein et al., in preparation). K35 and K85 (Fig. 6d, h, i) are already expressed in the hair-forming matrix of the cortex and the hair cuticle. The other hair keratins [type I: K31, K33a, K33b, K34, K36, K38 (focally) and K39 (including hair cuticle); type II: K81 and K86] are sequentially switched on upon differentiation in the lower hair cortex and in particular the large “bulk” of hair keratins are expressed in the middle cortex (“keratinizing zone”) of the ascending hair fiber (Fig. 6f, h, i). Furthermore, K32, K83, K82 (Fig. 6e, h, i) and K40 are sequentially expressed and restricted in the hair cuticle (Langbein and Schweizer 2005; Langbein et al. 2007). As exceptions, K37 was only found in the cortex of vellus hairs and K84, although a typical hair keratin was not detected in the hairs but specifically in the filiform papillae of the tongue (Langbein and Schweizer 2005). One has to keep in mind that out of the 54 human keratins at least 26 (~50%) are expressed in the hair follicle. Therefore, it is most amazing that this extraordinary complex structure—built up by a stratified outer root sheath, the companion layer, the inner root sheath comprising Henle, Huxley and IRS-cuticle layers, the hair cuticle, cortex, and sometimes a medulla—differentiate from cells of an undifferentiated and pluripotent “germinative cell pool” (Fig. 6a, h, i). There must exist an incredible “fine-tuning” in the regulation of gene expression when e.g. one cell differentiates into the IRS-cuticle and the directly neighboring cell builds up the hair cuticle, both structures being only one cell wide. Unfortunately, up to now, our knowledge on this most complex regulation of hair keratin gene expression is rather fragmentary. Hair keratins are also prominently expressed in the nail matrix and nail bed and contribute to the formation of the hard tissue of the nail plate (Perrin et al. 2004; Perrin 2007). Moreover, in the 1980s hair keratins have been additionally detected by immunohistochemistry in filiform papillae of the tongue and, intriguingly, in reticulum cells and Hassall’s corpuscles of the thymus (Heid et al. 1988b). Some human mutations affecting hair keratin genes have been recognized. The most well-known of these is the congenital hair disease monilethrix, characterized by deformed hair shafts with a beaded appearance. Causative mutations for this disorder have been identified in the hair keratins K86, K81, and rarely in K83 (Winter et al. 1997; for further references, see Schweizer et al. 2007), all of which are expressed in the cortex of hair shafts and thus indicating that monilethrix is a disease of the hair cortex. A further, rare disease recently found to be related to a distinct mutation in the type II hair keratin K85 is ectodermal dysplasia of hair and nail type, characterized by total alopecia and severe nail dystrophy (Naeem et al. 2006). The expression of K85 already in the hair matrix from which hair fiber formation starts (Schweizer et al. 2007) may explain the severe hair phenotype of these patients. As to tumors, hair keratins have—up to now—only been detected in pilomatricomas (Moll et al. 1988a; Régnier et al. 1997) which are regarded as originating from hair matrix and undergoing true hair differentiation. The complex hair keratin expression in benign and malignant pilomatricoma (Cribier et al. 2001, 2004, 2006) confirms this notion at the molecular level. Monospecific antisera against all of these keratins are available (PROGEN, Heidelberg, Germany; Langbein et al. 2004, 2006). K23, K24, K78, K79, K80: keratins with still unknown expression pattern These five, very different keratins complete the family of human keratin proteins. In principle, mainly the gene and cDNA sequences for the type I keratins K23 (Zhang et al. 2001) and K24 and the type II keratins K78 (formerly K5b), K79 (formerly K6l), and K80 (formerly Kb20) (Rogers et al. 2004, 2005) are known up to now. Unfortunately only very vague and preliminary (or no) expression data are available from Northern blot analyses, suggesting their expression in tongue (K78, K80) and skin (K79), respectively (Rogers et al. 2005). As they were obviously “overlooked” in former studies, their expression might be restricted to very special epithelia, distinct cellular differentiation stages or even transient expression phases. The investigation of the detailed expression pattern of these keratins will be one of the most urgent studies in this field. Keratins as diagnostic markers in tumor pathology One of the important fields of “application” of keratins (making use of the knowledge of their high number of gene family members combined with their special and characteristic expression patterns in distinct cell types, differentiation stages or functional states) is their use—aided by specific antibodies—as immunohistochemical markers in diagnostic tumor pathology. Epithelial tumors maintain—at least widely—specific features of the keratin expression patterns of the respective cell type of origin. Thus, in cases which on the basis of clinical data and conventional histopathology remain unclear, keratin typing may help to correctly identify and classify the tumor entity present. Keratin profiling is especially valuable for carcinomas of poorly differentiated histology, for carcinomas spreading over several organs, and in particular for metastases of an unknown primary tumor. Of the 54 human keratins, only a relatively small panel has attained diagnostic importance up to now, which might grow further with time and respective studies by including the keratins described within the last years. The carefully selected use of diagnostically relevant keratin antibodies—if appropriate, as part of a panel together with other tumor type markers—has become diagnostic standard in state-of-the-art clinical pathology (for recent overviews, see Chu and Weiss 2002b; Dabbs 2006). In the previous chapter on the individual keratins, some data on their expression patterns in tumors have already been presented. In the following, key diagnostic features of keratin typing concerning important tumor entities will be briefly summarized. Adenocarcinomas Adenocarcinomas are one of the largest groups of human malignant tumors and may arise in many organs and tissues. They comprise more than one-half of cases of cancer with unknown primary tumor. The identification of the specific origin—e.g. colon, ovary, pancreas, or lung—has become therapeutically important since effective chemotherapy schedules may be vastly different. As a group, adenocarcinomas are characterized by the predominance of simple-epithelial keratins, notably K8, K18 and K19, whereas K7 and K20 are variably expressed. It is this variability which may be diagnostically exploited, and thus staining for both K7 and K20 is the current practice resulting in different K7/K20 phenotypes (Table 3). The highest diagnostic significance of keratin typing is true for colorectal adenocarcinomas which—like the normal mucosa—almost always remain K20-positive (Fig. 3d). Most cases, including metastases, exhibit a K7−/K20+ phenotype (Fig. 3c, d) or express K7 at lower level as compared to K20 (Moll et al. 1992; Miettinen 1995; for further references, see Moll 1998, Chu and Weiss 2002b). K7 co-expression has been detected more frequently in advanced colorectal cancers (Hernandez et al. 2005). Although so characteristic of colorectal tumors, K20 may in certain situations be reduced. Thus, reduced expression of K20 has been found in a specific subset of colorectal carcinomas with high levels of microsatellite instability (McGregor et al. 2004); notably, K7 expression remained low in these tumors. Since K20 is a differentiation marker its reduction or even loss may also be due to dedifferentiation. In a large tissue microarray study, reduced expression of K8 and K20 was found to be associated with shorter patients’ survival, possibly on the basis of epithelial-mesenchymal transition (Knösel et al. 2006). Adenocarcinomas of the stomach usually show heterogeneous expression of both K7 and K20, the latter being rather variable but yet expressed in the majority of cases (Moll 1998, Chu and Weiss 2002b). No systematic differences between intestinal and diffuse/signet ring cell carcinomas have been described. In adenocarcinomas of the pancreas and the biliary tract, there is usually a clear predominance of K7 (Fig. 3e) together with variable and focal (often minor) expression of K20 in up to 75% of cases (Fig. 3f, Moll 1998; Chu and Weiss 2002b). Ductal adenocarcinomas of the pancreas in addition often express certain stratified-epithelial keratins, most notably K4 and K17 (Schüssler et al. 1992; Real et al. 1993; Moll 1998). Finally, a K7+/K20− phenotype is characteristic of adenocarcinomas of the ovary (except for the mucinous type), the endometrium and the lung. Endometrial adenocarcinomas often focally co-express certain stratified-epithelial keratins including K5, reflecting the potential of the tumor cells for stratification to develop squamous metaplasia. Like other adenocarcinomas, most breast carcinomas constitutively express K8, K18 and K19 (Altmannsberger et al. 1986; Malzahn et al. 1998). In some studies, however, a high level of K8 or K18 immunostaining, as detected by certain monoclonal antibodies, has been correlated with favorable prognosis and reduced or absent staining was associated with unfavorable outcome (Takei et al. 1995; Schaller et al. 1996; Woelfle et al. 2004). Microarray-based expression profiling of breast carcinomas has led to the definition of distinct subgroups. One of these has been designated as basal-like group (Sørlie et al. 2001) which is characterized by relatively poor prognosis. Interestingly, the typical expression profile of the basal-like cancers includes the basal cell-typical keratins K5, K14 and K17. Several older reports have already described the presence of such keratins in a subpopulation of breast carcinomas (Moll et al. 1983; Wetzels et al. 1991), and some did already point to a negative prognostic significance of basal cell keratin expression in invasive ductal breast cancers (Dairkee et al. 1987; Malzahn et al.1998). This has now been confirmed and extended by the growing body of recent microarray data, allowing to define a subgroup of sporadic breast cancers which exhibit bad prognosis and association with BRCA1 mutation or dysfunction (Gusterson et al. 2005; Diaz et al. 2007). For the recognition of this subgroup, immunostaining for keratins such as K5 may achieve importance (van de Rijn et al. 2002; for further references, see Gusterson et al. 2005). The different subtypes of renal cell carcinomas (RCCs) exhibit some characteristic features in keratin expression which may support the precise classification of these tumors (Thoenes et al. 1988; Moll 1993; Chu and Weiss 2002b; Skinnider et al. 2005; Liu et al. 2007). Thus, conventional (clear cell) RCCs express a simple keratin pattern of mainly K8/K18 together with variable, mostly minor expression of K19. In contrast, the papillary subtype of RCC is characterized by strong K19 expression as well as K7 expression in addition to the basic pair K8/K18. Chromophobe RCCs, on the other hand, typically express K7 (but little K19) in addition to K8/K18 (Thoenes et al. 1988). The benign oncocytomas, which histologically may resemble chromophobe RCC, are mostly K7-negative (Liu et al. 2007). The peculiar co-expression, together with keratins, of the mesenchymal intermediate filament protein vimentin in conventional (clear cell) and papillary RCCs but not in chromophobe RCCs and oncocytomas is another feature valuable in differential diagnosis. Another tumor category with characteristic co-expression of keratins and vimentin are malignant mesotheliomas. The epithelial type may be difficult to distinguish histologically from adenocarcinomas (e.g. pleural involvement by pulmonary adenocarcinomas or pleural/peritoneal metastases of adenocarcinomas of various origins). In contrast to most adenocarcinomas, epithelial mesotheliomas consistently express keratinocyte-type keratins, notably K5, in addition to keratins K8, K18, K19, and K7 typical for simple epithelia (Moll et al. 1989; Ordonez 1998; Chu and Weiss 2002a; Yaziji et al. 2006). Thus, in recent years, K5 has been included in the battery of useful mesothelioma markers. Neuroendocrine tumors Neuroendocrine tumors, a large and heterogeneous tumor group, are generally characterized by the expression of keratins typical of simple epithelia (notably K8, K18, and—more variably—K19) and the complete lack of keratins K5/K14 typically found in stratified epithelia. Particularly interesting in a diagnostic respect is the well-proven value of K20 as a consistent marker of Merkel cell carcinomas of the skin, allowing their delineation from metastatic small cell neuroendocrine carcinomas arising at other sites such as small cell lung carcinoma which—although morphologically similar—consistently lack K20 (Moll et al. 1992; Cheuk et al. 2001). Another interesting recent issue concerns endocrine tumors of the pancreas. Several studies suggest that the expression of K19 in these tumors may be correlated with a poor prognosis (Schmitt et al. 2007; La Rosa et al. 2007). Transitional cell carcinomas The urothelium exhibits a unique, complex pattern of keratin expression. K8, K18, K19, and K7 are expressed in all cell layers. K5 and K17 are restricted to the basal cell layer. Particularly characteristic are K13 expressed in the basal and intermediate cell layers and K20 specific for the superficial (umbrella) cell layer. This urothelial keratin pattern is relatively well conserved in noninvasive and invasive transitional cell carcinomas (TCCs) (for detailed reviews, see Moll 1998; Southgate et al. 1999). K20 has been found to be retained in ~80% of TCCs (Moll et al. 1992). Indeed, the combined presence of keratins K8/K18 (as well as K7 and K19) typical for simple epithelia together with K13 and K20 in a tumor is characteristic of urothelial origin, although K13 may be reduced or lost in poorly differentiated TCCs. Squamous metaplasia may modify the keratin expression pattern. Particularly interesting are the recent, prognostically relevant findings that noninvasive papillary TCCs may exhibit a normal K20 pattern (predominantly superficial) or an abnormal K20 pattern (all cell layers or negative) and that the normal pattern is predictive of tumor non-recurrence (Southgate et al. 1999). Squamous cell carcinomas Squamous cell carcinomas of different sites of origin are generally characterized by a predominance of stratified-epithelial/keratinocyte-type keratins but may co-express certain simple-epithelial keratins (for details, see Moll 1998). Most of these tumors strongly express the keratins K5 (Fig. 4c, e), K14 and K17 normally found in the basal layer as well as the keratins K6 (Fig. 4d) and K16 characteristic for hyperproliferative keratinocytes. Focally, there may be expression of K1/K10 (particularly in higher differentiated tumor cells which can end in the formation of horn pearls), and—to a lesser extent—K4 and K13. The co-expression of simple epithelia-typical keratins comprises K8, K18, and K19, and different studies have suggested that this co-expression seems to be more pronounced in poorly differentiated squamous cell carcinomas (for references, see Moll 1998). Recently it has been demonstrated that in squamous cell carcinomas of the oral cavity the expression of K8 and K18 is an independent prognostic marker and indicates a decreased overall and progression-free survival (Fillies et al. 2006). In summary, stratified-epithelial keratins, in particular K5 and K6, are useful as general markers for squamous cell carcinomas in histologically uncertain, poorly differentiated, or metastatic tumor cases. Although certain differences between the keratin expression patterns of squamous cell carcinomas from different sites of origin have been noted, it is not yet possible to use keratins as specific site markers in cases of unclear metastases (Moll 1998). Perspectives In the past 10 years, our knowledge on keratins has tremendously increased, regarding their molecular and cell biology as well as their application as markers in pathological diagnosis. The introduction of a new consensus nomenclature (Schweizer et al. 2006), which for the common (non-hair follicle specific) epithelial keratins widely conserves the older names established in the literature, makes this complex field much clearer than before and will facilitate future research. Today much is known about the structural functions of keratins, as proven by a wealth of human hereditary keratin diseases and transgenic mouse models. However, numerous questions still need to be answered, particularly concerning the regulatory functions of keratins. As reviewed in this article, recent experimental studies have pointed to newly recognized roles of certain keratins in apoptosis, cell growth, tissue polarity, wound response, and tissue remodeling. First, specific signaling pathways have already been described which seem to be associated with distinct keratins. The question of whether keratins might play a role in malignant transformation is particularly exciting, but at present a causal role in tumorigenesis has not been established for any keratin (for review, see Magin et al. 2007). However, keratins may in fact be linked to the cell cycle machinery (Margolis et al. 2006). Future research will hopefully answer the question why the human organism actually needs 54 different keratin proteins, all of which in the first instance form the seemingly “primitive”, uniform structure of a 10 nm intermediate filament—although with peculiarities in the head and tail domains of their single molecular components. One may also expect for the future that the diagnostic application of keratins, although already established in tumor pathology, will be further refined and extended. Examples of very recent new developments in tumor classification are the recognition of the basal-like subtype of breast cancer on the basis of the expression of K5 and the prognostic relevance of K19 in endocrine pancreatic tumors. At the moment, the panel of keratins introduced into routine diagnosis as histopathological tumor markers is distinctly small, essentially consisting of K5, K7, K8/K18, K19, and K20. Furthermore, the expression patterns of a battery of special keratins were detected only in the last few years and these keratins are just started to be involved into tumor diagnostic studies. Not to forget that at the time 5 out of the 54 human keratins are still uncharacterized in their expression sites and patterns and possibly, their expression in neoplasia might have a potential as diagnostic markers. Since all keratins have an exquisite cell type- and differentiation stage-specific regulation and expression pattern, it might well be that upon future research further keratins will be introduced as markers for certain diagnostic questions in pathology to widen and to refine the diagnostic potential of keratins.

Bioinformation-1-7-1891696

Database of cell signaling enzymes

This paper describes a database for cell signaling enzymes. Our web database offers methods to study, interpret and compare cell-signaling enzymes. Searching and retrieving data from this database has been made easy and user friendly and it is well integrated with other related databases. We believe the end user will be benefited from this database. Background A cell communicates with its neighbors and environment by sending and receiving information in the form of chemical signals. These signals are converted into intracellular second messenger signals that ultimately make the cells respond appropriately by dividing, moving or even dying. The external signals may enter the cell through enzymes, G-protein coupled receptors, hydrophobic molecules and ion channels. When the receptor sensing the signal is a catalyst such as enzyme, the response is amplified. Thus cell signaling governs the basic cellular activities and coordinates the cell action. Errors in cell signaling is the cause for many serious diseases/disorders including cancer, autoimmune diseases, diabetes etc. By understanding cell signaling we can treat these diseases effectively and potentially. [1] However, no attempts have been made so far to curate and catalog the enzymes involved in cell signaling. DCSE, the Database of Cell Signaling Enzymes covers a gamut of cell signaling enzymes, which includes kinases, phosphatases, adenylyl cyclases, caspases, phosphodiesterases, phospholipases, prenyltransferases etc. As mentioned above the defect in cell signaling mechanism is the major cause for diseases and hence the cell signaling enzymes are considered to be potential target in rational drug design approach. Methodology DCSE was developed using MySQL [ 2], a relational database management system, at the back-end for storing data. The database can be regularly updated. The data for the database were collected from SwissProt [3 ], the repository of protein sequence information. PHP [4], a widely used general purpose scripting language that is especially suited for web development, was used to design the database interface. The database can be accessed over the Internet. A screenshot of the home page is shown in Figure 1.The database can be searched by specifying keywords such as name, Enzyme Commission Number and species. For each enzyme in the database a unique identifier called DCSE ID has been assigned. The ID consists of two parts. The first part tells about the enzyme class and the second part indicates the number of that enzyme in that class. The two fields are separated by an underscore (_). Each enzyme has been defined with its name, Enzyme commission number, and the species from which it has been sequenced. Crossreference details to other databases are also provided. The functional and other sequence related informations are provided by SwissProt, the domain classification and function is provided by InterPro [5], the protein family classification is provided by Pfam [ 6] and the structural details are provided by PDB. [ 7] A hyperlink has been provided to corresponding entry page in the abovementioned databases. The sequence is also displayed in raw-text format. One can retrieve the sequence in FASTA format using the ‘Retrieve in Fasta format’ option available with each entry. An advanced search can also be performed by filling an advanced query form that takes input as DCSE id, name, EC number, Species, other database id such as SwissProt, InterPro, Pfam and PDB. The fields in the search form are joined together by AND operator. BLAST (Basic Local Alignment Search Tool) [8 ] is an algorithm for comparing biological sequences. Given a library or database of sequences, a BLAST search enables a researcher to look for sequences that resemble a given sequence of interest. A standalone BLAST has been installed and it can be used for similarity search purpose. BLAST can be done against SwissProt, DCSE or PD. Usability/Accessibility There are three ways in which a user can query the database. The first is the ‘keyword search’ that can be done by specifying exact or likely keywords such as name, Enzyme Commission Number and species. The second search option is the ‘Advanced Search’ wherein the user can fill a form by specifying the details for various fields such as DCSE id, Enzyme name, Enzyme commission number, SwissProt accession number, InterPro domain ID, Pfam id or PDB id. The fields are joined together by AND operator. Alternatively the users can also browse the database by the 21 different categories of cell signaling enzymes. Whenever the database is searched the search returns the result with number of hits found for that query along with a summary of details for each entry with its id, name and species. The user can then select the appropriate hit by following the link on it. This displays a page with all the available details of the enzyme in the database such as name, EC number, sequence information and cross-reference to already existing databases as shown in Figure 2. A hyperlink has been provided to corresponding entry page in the cross-referenced databases for easy access. The sequence is displayed in raw-text format. One can retrieve the sequence in FASTA format using the ‘Retrieve in Fasta format’ option available with each entry. A BLAST search can also be done to search for similar sequences within the database or in general. A link has been provided in the tools page. The BLAST input page and result page are shown in figure 3 and 4. Utility to the biological community Since DCSE gather together all the information on biological molecules, sequences, structures, functions, and biological reactions, which transfer the cellular signals, this database has the potential of becoming a major hub of resource for the biological community. This database provides a mechanism by which researchers and students can transform information about interactions between biomolecules into knowledge about a cellular process. Caveats The database needs to be updated from time to time as the data has been obtained from other sources. Despite our efforts to collect information from various sources and check them for consistency, the quality of the information depends heavily on the original source. Future development The coiled-coil structure is important for protein interaction. Since the cell signaling also involves protein interaction, a tool to predict coiled-coil structure is of importance. There are few tools for this purpose like MultiCoil. The Multicoil program just gives the probability for a residue to be in coiled-coil region. We intend to develop a coiled-coil prediction tool based on artificial neural networks that could tell whether a residue is in coiled-coil region if it crosses a threshold computed from a training set consisting of proteins with known coiled-coil structure.

J_Occup_Rehabil-3-1-1915618

Are Pain Intensity and Pain Related Fear Related to Functional Capacity Evaluation Performances of Patients with Chronic Low Back Pain?

Introduction: Pain related fear and pain intensity have been identified as factors negatively influencing Functional Capacity Evaluation (FCE) performances in patients with CLBP. Conflicting results have been reported in the literature. The objective of this study was to analyze the relationships between pain intensity and pain-related fear on the one hand, and performances during an FCE on the other hand in two samples of patients with chronic low back pain (CLBP). Methods: Two cross sectional observation studies were performed with two samples of patients with CLBP (study 1: n = 79; study 2: n = 58). Pain related fears were operationally defined as the score on the Tampa Scale of Kinesiophobia in study 1, and the Fear Avoidance Beliefs Questionnaire (FABQ) in study 2. Pain intensity was measured with a Numeric Rating Scale in both studies. Avoidance behavior observed during FCE was in both studies operationally defined as the unwillingness to engage in high intensity performance levels of three different functional activities: high intensity lifting, prolonged standing in a forward bend position, and fast repetitive bending at the waist. Results: A total of 25 correlations between pain and pain related fear, and performance variables were calculated, out of which 7 were significant (p < 0.05). The strength of these significant correlations ranged from r = −0.23 to r = −0.50. Multivariate linear regression analyses revealed non-significant relationships in most instances. Pain and pain related fear contributed little if any to these models. Conclusions: The relation between pain and pain related fear and FCE performance is weak or non-existent in patients with CLBP. Introduction Functional Capacity Evaluations (FCEs) are batteries of tests designed to assess a person's functional capacity related to work [1–3]. To establish functional capacity, the patient is asked to perform a number of activities. Combined, these performances reflect a patient's ability to perform work-related activities. A patient's performance is determined by biological, psychological and social factors [4]. Pain related fear, also referred to as fear avoidance beliefs or fear of movement and reinjury [5], and pain [6] have been identified as examples of psychological factors influencing a patient's performance during an FCE. Pain related fear refers to a condition in which the patient has an excessive, irrational, and debilitating fear of physical movement and activity, resulting in feelings of vulnerability to painful injury or reinjury [7, 8]. People who experience pain-related fear will avoid activities they associate with increased risk for pain or (re)injury. As such, pain-related fear should have a negative effect on the results of performance testing [5, 9]. It has been stated that ‘a valid assessment of functional capacity cannot be carried out without controlling for fear avoidance beliefs’ [8]. Several studies were performed to study the relationship between pain-related fear and different types of avoidance behavior in patients with chronic non-specific low back pain (CLBP). The strength of this relationship varied between the studies from r=0.06 to r=0.49 [8–10, 11]. The relationship between pain related fear and performance appeared stronger in studies where patients were observed under strictly controlled conditions [8, 10], and weaker in studies where patients were observed in a less controlled environment [11]. In a previous study performed in an admission phase for a pain rehabilitation program, no relationship was found in a sample of patients with CLBP, between lifting performance and pain-related fear [12]. Although the evidence from other studies did not overwhelmingly support the strength of the relationship, our results, presented in the Journal of Occupational Rehabilitation in 2003, were in contrast with other studies. It could not be determined whether the results found were weaker because of the operational definitions of pain-related fear (score on Tampa Scale for Kinesiophobia (TSK)) or avoidance behavior (maximum lifting performance). It was recommended that future studies should use additional operational definitions for pain-related fear and avoidance behaviors. In a recent review pain intensity has been identified to consistently associate negatively with patient performances during FCEs [6]. The strength of the significant associations between pain intensity and FCE performances in patients with chronic pain vary from r = −0.25 to r = −0.56 [6]. In the study mentioned above, the relation between pain intensity and lifting performance was (somewhat) weaker and non-significant (r = −0.21) [12]. Aim of this study was to further analyze the relationships between pain intensity and pain-related fear on the one hand, and avoidance behaviors on the other hand. Two studies were performed with separate samples of patients with CLBP. Pain related fears were operationally defined as the score on the TSK in study 1, and the Fear Avoidance Beliefs Questionnaire (FABQ) in study 2. Avoidance behavior was in both studies operationally defined as the unwillingness to engage in high intensity performance levels of three different FCE tests: high intensity lifting, prolonged standing in a forward bend position, and fast repetitive bending at the waist. Non-significant or weak relationships were considered falsifications of the hypotheses that pain and pain-related fears would have clinically relevant impact on FCE performances in patients with CLBP. Materials and methods Patients Two separate cohorts of patients who were referred to an outpatient rehabilitation program of the Center for Rehabilitation of the Groningen University Medical Center, The Netherlands, were included. Both study samples consisted of patients diagnosed with CLBP, aged between 18 and 60 years, with symptoms lasting longer than 3 months. Excluded were patients with comorbidity with negative consequences for functioning (i.e. severe depression needing a psychiatric referral), and patients with specific pathology related to the lumbar spine (i.e. disc herniations, tumors, spondylolisthesis grade 3 or 4, etc.). Patients were not selected based on (high or low) levels of pain or pain related fears. Selection process of both study samples were equal, with the exception that patients in study sample 1 explicitly agreed to participate in a larger study in addition to regular clinical care [13]. The patients of sample 2 received regular clinical care only. Procedures Prior to the medical intake patients filled out a TSK [7, 8], a numeric rating scale to assess pain intensity (NRS), a Roland Morris Disability Questionnaire [14] (RMDQ) to assess disability, and a study consent form. The patients underwent the FCE approximately 2 weeks after the medical intake. Prior to the FCE, all patients were asked whether their pain and functional status was different compared to their status during medical intake. In study 2, patients filled out the FABQ [15] prior to the FCE. For study 1, approval was granted by the medical ethics committee of the University Medical Center Groningen, The Netherlands. The data of the patients of study sample 2 were collected as part of regular clinical procedures, for which institutional approval was received. Measures Patients rated their current pain intensity on a NRS ranging from 0 to 10; 0 being no pain at all and 10 being the worst pain imaginable. Self reported disability was assessed by the RMDQ [14]. Scores on the RMDQ can range from 0 to 24; 0 indicating no disability and 24 indicating severe disability. The psychometric properties of the RMDQ are good [14, 16]. Pain related fear was assessed with a different questionnaire for each sample. In study 1, pain related fear, called fear of movement and reinjury, was assessed by the TSK [8]. The TSK consists of 17 items. Each item is provided with a 4-point Likert scale with scoring alternatives ranging from “strongly disagree” to “strongly agree.” Although the scale consists of 2 subscales, activity avoidance and somatic focus, a total score is usually presented. The total score ranges from 17 to 68. The reliability of the TSK is fair [8]. Criterion validity was established from correlations with other self-reported measures of fear (of bodily injury), anxiety, depression, and catastrophizing [8]. In study 2, pain related fear, called fear avoidance beliefs, was assessed by the FABQ [10, 15]. The FABQ consists of 16 items. Each item is provided with a 7-point Likert scale with scoring alternatives ranging from 0 to 6. The FABQ consists of 2 scales. The activity scale consists of 4 items (total score 0–24), the work scale consists of 6 items (total score 0–36). The remaining 4 items are not used. The reliability of the FABQ is good in patients with acute and subacute back pain [17], but has not been established in patients with CLBP. Avoidance behavior was assessed during an FCE. The patients were asked to perform 14 different activities to their maximum abilities (3 types of lifting, carrying, pushing, pulling, overhead work, stooping, crouching, kneeling, standing, walking, sitting, stair climbing), according to the protocols of Isernhagen Work Systems [18]. Selection of activities was based on the Dictionary of Occupational Titles (DOT), thus assuming construct validity with regard to work [19]. Three activities tested during the FCE were selected for this study, because they might be potentially harmful for patients with CLBP [20]. These activities were lifting, static bending and dynamic bending. Lifting was assessed by means of a standardized lifting task consisting of lifting a receptacle with incremental weights from a table (74 cm) to the floor and vice versa. ‘Ergonomically correct’ body mechanics were not emphasized [21]. The patient's maximum was reached in four to five increments. The maximal amount lifted five times within 90 seconds was recorded (kilograms). Test-retest reliability of the lifting task is good in patients with CLBP [22–24]. During lifting, heart rate was measured using a heart rate monitor. In study 2, the intensity level of lifting was observed by the evaluator, and recorded by means of a Borg CR-10 scale. The reliability and validity of this procedure is good [25]. These observations were not recorded in study 1. During the static bending capacity test the patient was asked to stand as long as possible with his / her trunk 30–60° flexed forward, while performing a simple manipulation task. The time this position was held was recorded (seconds). Test-retest reliability of this test in patients with CLBP is good [22]. For the dynamic bending test, the patient is asked to bend at the hips and back as fast as possible, pick up a small object from the floor, and to remove this object to a shelf at crown level (top of head). This was repeated 20 times. The number of repetitions completed and the time needed to complete this activity was recorded (seconds). Test retest reliability of this test in patients with CLBP is moderate to good [22]. The construct validity of the tests has been established [26, 27]. Before testing, patients were instructed regarding termination of the test. Testing was terminated when one of the following occurred: the patient stated verbally that he/she wished to terminate the activity, the heart rate reached 85% of the age-related maximum, indicated by a heart rate monitor, or the evaluator deemed further testing to be unsafe. No verbal reassurance was given during the testing procedures. The evaluator was blinded to the questionnaire scores. Analysis Descriptive statistics were used to describe the study samples. All variables of interest were analyzed at interval level. A student t-test was used to analyze differences between males and females. When differences were significant, further analyses were performed separately for males and females. Pearson product moment correlations were calculated to express linear associations between pain intensity and measures of pain related fear on the one hand and performance variables on the other hand. Correlations were interpreted as follows: 0.25 or less little if any relationship, 0.26–0.49 poor relationship, 0.50–0.69 moderate relationship, 0.70–0.89 strong relationship, 0.90–1.00 very strong relationship [28]. Pain intensity and pain related fear, as well as gender, age and self-reported disability were entered as predictor variables in a multivariate linear regression model (method: enter) and performances as outcome variables. A p-value of <0.05 was considered statistically significant for all analyses. Results The characteristics of the patients participating in study 1 and 2 are presented in Table 1. All patients declared that their pain and functional status had not changed between medical intake and the FCE. Scores on the TSK subscale in study 1 are: activity avoidance males mean 19.1 (SD 3.8), females mean 18.9 (3.3); somatic focus males mean 9.7 (2.5), females mean 9.2 (2.6). The 20 repetitions of the dynamic bending test were not completed by 10 (of 79, 13%) patients in study 1 and by 6 (of 58, 10%) patients in study 2. To correct for differences in number of repetitions, the time needed to complete one repetition was calculated and presented (time/number of repetitions). The observed level of intensity of the maximal lifting performances in study 2 were for males mean 7.2 (SD 2.2) and for females 7.4 (SD 1.6) (Borg CR-10 scale). Observational data was unavailable in study 1. Within the study samples, differences between males and females were non-significant, with the exception of the maximum lifting performance. Between the study samples, none of the differences were significant. Table 1Age, pain intensity, self-reported disability, pain related fear (TSK or FABQ), and performance variables of 2 samples of patients with chronic low back painStudy 1 (n = 79)Study 2 (n = 58)Males (n = 49)Females (n = 30)Males (n = 39)Females (n = 19)Mean (SD)Mean (SD)Mean (SD)Mean (SD)Age (years)37.8 (9.0)37.8 (8.8)40.4(8.6)35.6 (8.3)Pain intensity (NRS; 0–10)4.7 (2.3)5.0 (1.6)4.5 (2.4)4.9 (2.2)Self-reported disability (RMDQ; 0–24)12.5 (4.4)12.4 (4.7)11.6 (4.4)11.2 (5.9)TSK (17–68)37.5 (5.3)36.0 (5.6)N/AN/AFABQ activity scale (0–24)N/AN/A13.4 (4.1)13.2 (5.6)FABQ work scale (0–36)N/AN/A19.3 (9.5)15.1 (11.6)Lifting performance (kg)*31.7 (14.7)18.9 (8.1)32.3 (3.8)20.5 (6.3)Lifting maximum heart rate (BPM)126.7 (15.8)129.7 (23.5)128.2 (14.5)128.8 (19.3)Static forward bend (sec)187.4 (148.9)230.6 (146.3)188.0 (115.8)237.3 (154.8)Dynamic forward bend (sec/rep)2.8 (0.7)3.2 (1.3)3.2 (1.6)2.9 (0.8)SD: standard deviation; NRS: Numeric Rating Scale; RMDQ: Roland Morris Disability Questionnaire; TSK: Tampa Scale for Kinesiophobia; FABQ: Fear Avoidance Beliefs Questionnaire; BPM: beats per minute; N/A: not assessed. *: Difference between males and females significant in both studies (p < 0.05).Note. Differences between the study samples were non-significant (p < 0.05). A total of 25 correlations between pain intensity, TSK (study 1) and FABQ (study 2) and performance variables are presented in Table 2. The strengths of the correlations were moderate in one occasion and otherwise weak/poor or non-significant. Pain intensity explained 25% of the variance of the performances of male lifting performance in study 2 (r = −0.50). In all other analyses over both studies, the explained variance between pain intensity and performance variables was 8% (r = −0.29) or less. The FABQ work scale explained at best 13% of the variance of the performances (male lifting performance in study 2; r = −0.37). The correlation between either TSK subscales with any of the performance variables was non-significant. No differences were observed between patients who did or did not complete 20 repetitions of the dynamic bending test concerning their relationship with the TSK or the FABQ (all non-significant). The correlation between observed intensity of lifting and FABQ work scale was r = −0.27 (significant). The correlation between observed lifting intensity and FABQ activity scale was r = −0.07 (not significant). Table 2Relationships between performance variables and pain intensity, the Tampa Scale for Kinesiophobia (TSK, study 1; n=79) and the Fear Avoidance Beliefs Questionnaire (FABQ, study 2; n = 58) in patients with chronic low back pain. All relationships are expressed in Pearson's r (95% CI)Study 1Study 2PainTSKPainFABQ activity scaleFABQ work scaleLifting performance males (kg)−0.26 (−0.57 to 0.11)−0.12 (−0.46 to 0.25)−0.50* (−0.71 to −0.22)−0.13 (−0.43 to 0.19)−0.37* (0.06 to 0.61)Lifting performance females (kg)−0.06 (−0.34 to 0.23)−0.17 (−0.43 to 0.12)0.09 (−0.52 to 0.38)−0.07 (−0.51 to 0.40)−0.03 (−0.48 to 0.43)Lifting maximum heart rate (BPM)0.05 (−0.17 to 0.27)−0.10 (−0.31 to 0.12)−0.27* (−0.49 to −0.01)−0.24 (−0.47 to 0.02)−0.24 (−0.47 to 0.02)Static forward bend (sec)0.18 (−0.04 to 0.39)−0.23* (−0.43 to −0.01)−0.29* (−0.51 to −0.03)−0.33* (−0.54 to −0.08)−0.25 (−0.48 to 0.01)Dynamic forward bend (sec/rep)−0.10 (−0.31 to 0.12)0.14 (−0.08 to 0.35)0.03 (−0.23 to 0.29)0.07 (−0.19 to 0.32)0.30* (0.05 to 0.52)Note. TSK: Tampa Scale for Kinesiophobia; FABQ: Fear Avoidance Beliefs Questionnaire; BPM: beats per minute; 95% CI: 95% confidence Interval.*p < 0.05. Analyses were performed on the subgroups of patients scoring in the highest quartile of the pain intensity spectrum (study 1, NRS ≥ 6, n = 22; study 2, NRS ≥ 7, n = 17), the TSK (score ≥ 40, n = 22), and the FABQ (activity scale ≥ 17, work scale ≥ 25, n = 16). Analyses involving lifting performances were performed for males and females separately. Correlation coefficients between pain intensity or pain related fears and performance variables ranged between r = 0.00 and r = 0.23 in study 1, and between r = 0.00 and r = 0.26 in study 2. All correlations were non-significant. Scatter plots of the distribution of pain intensity or pain related fears and performance variables were created to visually analyze patterns in the datasets. As an example, the distribution of pain related fear (TSK) and lifting performances is presented in Fig. 1. Other distributions between pain intensity or pain related fears (TSK or FABQ) and performance variables were similar, but they are not presented.Fig. 1Plot of maximum lifting performance and scores on the Tampa Scale for Kinesiophobia (TSK) in 79 patients with chronic low back painTable 3Multivariate linear regression models predicting variance in performance in patients with chronic low back painβ (95% CI)Betapr2 changeModel r2Study 1 (n = 79)Dependent variable: lifting •Gender (female = 0, male = 1)14.45 (7.36 to 21.54)0.48<0.0010.230.28 •Constant17.01 (−9.60 to 43.62)–0.205–Dependent variable: static forward bend •Model r2 = 0.11; None of the independent variables contributed significantly to the regression equationDependent variable: dynamic forward bend •Model r2 = 0.12; None of the independent variables contributed significantly to the regression equationStudy 2 (n = 58)Dependent variable: lifting •Gender (female = 0, male = 1)14.94 (6.15 to 23.73)0.480.0010.230.37 •Pain intensity−1.83 (−3.65 to −0.01)−0.290.0490.08 •Constant (kg)45.70 (25.67 to 65.73)–<0.001–Dependent variable: static forward bend •Constant (sec)549.15 (336.09 to 762.21)–<0.001–0.27 •None of the other independent variables contributed significantly to the regression equationDependent variable: dynamic forward bend •Model r2 = 0.18; None of the independent variables contributed significantly to the regression equation Results of the multivariate regression analyses are presented in Table 3. Study 1: Gender was the only variable that independently contributed significantly to the regression equation to predict lifting performance. The explained variance was 28%. Both models to predict the other two outcome variables, static forward bend and dynamic forward bend, were non-significant. Study 2: Gender and pain intensity (to a lesser extend) contributed significantly to the regression equation to predict lifting performances. The explained variance was 37%. In the regression analysis to predict static forward bend performances, none of the independent variables contributed significantly to the regression equation (explained variance 27%). In the regression analysis to predict dynamic forward bending performances none of the independent variables contributed significantly to the regression equation. Collinearity diagnostics were performed for the independent variables. In both studies the average variance inflating factors (VIF) were greater than 1, and none of the VIF-values were greater than 10. Specifically, the relationship between pain intensity and measures of pain related fear were analyzed. In study 1, the strength of the relation between pain and the TSK was r = −0.04 (ns). In study 2, the strength of the relation between pain intensity and FABQ activity scale was r = 0.07 (ns), and between pain intensity and FABQ work scale was r = 0.20 (ns). Discussion This study has demonstrated that the associations between pain intensity and pain related fears on the one hand and FCE performances on the other hand were generally weak or non-significant. Correlations between pain intensity and pain related fears and performances were significant in only 7 out of 25 analyses (Table 2). The strength of these significant relationships ranged from r = −0.23 to r = −0.50, explaining 5 to 25% of the variance (r2). Multivariate regression analyses were non-significant in 3 out of 6 instances. When significant, pain intensity or pain related fear contributed little if any to these models. Relationships are consistent throughout the intensity spectra (upper quarter analyses and Fig. 1). Thus, overseeing all the results of this study, it appears that the relationships between pain and pain related fear and performances in an FCE are generally weak or non-existent. Although we realize that interpretations of these results are open for discussion, we interpret the magnitude of the relationships between pain and pain related fears on the one hand and performances in an FCE on the other hand of limited clinical relevance. In comparison to our previous study [12], we have now used two new cohorts, two questionnaires instead of one to measure a wider range of pain related fears, and three performance tests instead of one to measure a wider range of avoidance behaviors. Additionally, we have added heart rate and observations as indices for effort level in lifting. Both pain intensity and pain related fear were poorly associated with indices for effort, explaining 7% or less of the variance. The results of the 2 new studies presented here are generally consistent with our previous study, which adds to the robustness of the results. There may be several explanations for the weakness or non-existence of the associations between pain intensity and pain related fears on the one hand, and avoidance behaviors on the other hand. Pain intensity and pain related fears might have been inappropriately operationally defined by the NRS, the TSK and the FABQ, avoidance behavior might have been inappropriately operationally defined by the three performance measures, our study samples differ from samples reported elsewhere, or the variables were appropriately operationally defined, but the relationship was mediated by one or more currently unknown variables. Additionally, a combination of the above may explain our findings. The NRS, TSK and the FABQ may be considered among the standards to measure pain intensity and pain related fears. The reliability and the validity of the measures for avoidance behavior and pain related fears have been established. Both the performance measure ‘heavy lifting’ and a postural tolerance test have been used previously in different studies as measures for avoidance behavior [8, 9, 12]. The activities selected are potentially harmful for patients with CLBP [20]. Thus, it seems unlikely that the operational definitions were inappropriate for the constructs measured. Both our study samples are similar to samples reported in other studies reporting on patients with CLBP concerning age, pain intensity, self-reported disability, and amount of fear avoidance beliefs [8, 10]. This similarity may rule out the plausibility the results of our study are caused by the selection of study samples. While the TSK and the FABQ may be considered among the consented standards to measure pain related fears, it may be possible that the domain measured might be too large and general to be of value in predicting specific behaviors such as the ones used in this study. This in itself could explain the findings of this study and the lack of consistent findings in previous research. However, if the measures are considered valid representations of the constructs pain intensity, pain related fear, and avoidance behavior, then a different explanation should be considered for our findings. Other variables may serve as a mediator between pain variables and avoidance behavior. As suggested in recent pain literature, two variables should be considered as mediators or confounders: motivation [29] and acceptance [30]. Within the motivational models [29, 31, 32], it is suggested that the value of the goal and the patient's belief in his ability (self-efficacy) to attain this goal is predictive for his behavior. When applied to the results of this study, patients were apparently willing to perform the tests regardless of their pain related fears. They were willing to perform, because they may have been convinced that this would have helped them to reach a valued goal (for example return to work). Contrarily, patients without pain related fears may perform poorly on the tests, because they do not see it to be of any value. Alternatively, a poor performance could be explained if the patient's goal would be to use the test results for financial gain (disability allowance) [33]. In that case, pain related fears would be of little importance to determine a patient's behavior. Within the acceptance model [30], it is suggested that patients may be able to accept their pain or fears as an inevitable part of their lives. Patients would report pain or have cognitions such as pain related fears, but are able to function normally with their pain or fears. Some empirical evidence is found to support the applicability of this model within chronic pain [34, 35]. If a mediating effect or confounding of one or more different variables is present, then poor or non-existent associations between pain intensity, pain related fears and test performance are very well imaginable. Future studies should be conducted to find empirical evidence for the proposition that constructs as motivation and acceptance are more important predictors for functional capacity than pain or pain related fears. Additionally, other currently unknown variables may have played a mediating role in the relationship between pain variables and avoidance behaviors. Both self-report and performance measures assess effort-related performance [36]. As such, they cannot be defined independently of the person's behavior. Pain-related disability is a matter of human performance, whether it is observed/measured or reported [36] and this may also apply to patient's behavior in an admission phase of a rehabilitation program. In this stage patients need not only reveal their cognitions and perception of disability, but at the same time present an image to the practitioner [37], to justify the need for treatment [38, 39] and perhaps to justify the fact that they are off work due to CLBP. The scores on the questionnaires may thus be an overrating of the ‘real’ pain or pain related fears. An FCE is a measure of demonstrated ability. Pain behaviors may prevent the patients to perform to their maximal physical capacity. The ratings of effort level indicate a sub maximal performance, suggesting that this behavior has occurred in study 2 (and probably also in study 1, but data were unavailable). Without these pain behaviors, the FCE results might have been higher. On average, the performance-based scores presented in this study may be considered an underrating of the physical abilities of the patients. Consequently, clinical interpretation of the outcomes of effort related assessments should be made with care. We propose that the scores on the questionnaires are interpreted as ‘the patient reports that … (he is fearful), instead of ‘the patient is/feels/perceives … (fearful/fear). For the FCE, we propose that the scores are interpreted as the patient's performance, instead of the patient's capacity (FPE: Functional Performance Evaluation). The performance of a patient depends on his capacity and his willingness to produce [25]. The current study and our previous study show that report of pain and pain related fears explain little of the patient's performances during FCE. Statements that ‘a valid assessment of functional capacity cannot be carried out without controlling for fear avoidance beliefs [8] and ‘pain related fear is more disabling than pain itself’10 are not supported by the results of this study. The results of this study may not be generalizable to other FCEs or other groups of patients. From the results of the current study it should be concluded that future research aimed at unrevealing ‘determinants of performance’ in chronic pain should not be restricted to pain and pain related fears only, but include other avenues such as self efficacy [40, 41], motivation and acceptance as well.

Acta_Neuropathol-4-1-2386158

Peroxiredoxin 6 in human brain: molecular forms, cellular distribution and association with Alzheimer’s disease pathology

Peroxiredoxin 6 is an antioxidant enzyme and is the 1-cys member of the peroxiredoxin family. Using two-dimensional electrophoresis and Western blotting, we have shown for the first time that, in human control and brain tissue of patient’s with Alzheimer’s disease (AD), this enzyme exists as three major and five minor forms with pIs from 5.3 to 6.1. Using specific cellular markers, we have shown that peroxiredoxin 6 is present in astrocytes with very low levels in neurons, but not detectable in microglia or oligodendrocytes. In control brains, there was a very low level of peroxiredoxin 6 staining in astrocytes that was confined to a “halo” around the nucleus. In AD, there were marked increases in the number and staining intensity of peroxiredoxin 6 positive astrocytes in both gray and white matter in the midfrontal cortex, cingulate, hippocampus and amygdala. Confocal microscopy using antibodies to Aβ peptide, tau and peroxiredoxin 6 showed that peroxiredoxin 6 positive astrocytes are closely involved with diffuse plaques and to a lesser extent with neuritic plaques, suggesting that plaques are producing reactive oxygen species. There appeared to be little astrocytic response to tau containing neurons. Although peroxiredoxin 6 positive astrocytes were seen to make multiple contacts with tau positive neurons, there was no intraneuronal colocalization. In brain tissue of patients with AD, many blood vessels exhibited peroxiredoxin 6 staining that appeared to be due to the astrocytic foot processes. These results suggest that oxidative stress conditions exist in AD and that peroxiredoxin 6 is an important antioxidant enzyme in human brain defenses. Introduction Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by the progressive decline in memory, language, cognition and motor functions. The neuropathological hallmarks of AD are the accumulation of extracellular amyloid plaques containing the amyloid beta peptide (Aβ) and intraneuronal neurofibrillary tangles containing hyperphosphorylated microtubule-associated protein tau. Although the initiating molecular event(s) in AD are not known, oxidative stress produced by the generation of reactive oxygen species (ROS) appears to be a fundamental process contributing to the neuropathophysiology [10]. Whether oxidative stress has a role in the cause or affect in AD has been debated for many years, but recent research has suggested that it occurs prior to the onset of symptoms [35]. The histopathological evidence for oxidative stress in AD includes lipid, protein and DNA oxidation. There is evidence to suggest that oxidative stress is generated peripherally [26] with the brain being particularly susceptible, but there is also a body of evidence to suggest that amyloid β peptide (Aβ42) may be linked to lipid peroxidation [5, 6], which may then be linked to the phosphorylation of tau [24]. Two reactive products of Aβ42-induced lipid peroxidation are the production of four hydroxynonenal and acrolein [5], which have been shown to disrupt membrane asymmetry [7] and may be involved in the conformational changes to tau, promoting the formation of neurofibrillary tangles [24]. Oxidative damage occurs as part of the aging process, but it occurs at a much greater rate in most neurodegenerative diseases [3]. The hallmark of protein oxidation is the generation of protein carbonyls, which are markedly elevated in AD [3, 16]. There is overwhelming evidence to implicate oxidative stress in the pathogenesis of AD, and this is most likely an early event in the disease process. Less is known however of how the brain responds to oxidative insults. There are a number of cellular antioxidant defenses whose primary role is to convert the ROS into unreactive compounds. These defense mechanisms include superoxide dismutase, which works together with selenium-dependent glutathione peroxidase, catalase and more recently the peroxiredoxin family. These enzymes transform the superoxide radical to hydrogen peroxide, which in turn is converted to water. The distribution and changes in expression of the antioxidant enzymes have not been well studied as the other aspects of the disease process. Cellular selenium-dependant glutathione peroxidase (GPx-1) has long been regarded as the major cellular antioxidant enzyme, but there are conflicting reports regarding the changes in activities of this enzyme in AD. Early reports suggest that there is no change in GPx-1 activities in AD brains [15, 20], while other researchers suggest that there is a decrease in GPx-1 activity [8, 27, 32]. There are no reports indicating compensatory increase in GPx-1 activity; however, cell models overexpressing GPx-1 are more resistant to Aβ-mediated toxicity, suggesting that increased levels of GPx-1 is protective [2]. There is considerable evidence to implicate Aβ toxicity and the generation of ROS and in particular hydrogen peroxide as key steps in contributing to the neurotoxicity in AD. We have previously reported that peroxiredoxin 6, also termed 1 Cys-peroxiredoxin or nonselenium glutathione peroxidase, is upregulated in Parkinson’s disease (PD) [30]. This enzyme has no amino acid homology to any known sequence of the glutathione peroxidase enzymes and has both antioxidant and phospholipase A2 activity and can reduce hydroxyperoxides [9]. Given that the brain is particularly vulnerable to oxidative damage due to its high energy and oxygen requirements, high levels of transition metals and peroxidizable fatty acids suggests that peroxiredoxin 6 could play a key antioxidant role in neuroprotection. There are six members of the peroxiredoxin family at this stage with peroxiredoxins 1–5 being the 2-Cys members while peroxiredoxin 6 is the sole 1-Cys member [36]. This enzyme is bifunctional with 80% of its function as a peroxidase and the remainder as a phospholipase A2. Peroxiredoxin 6 has not specifically been examined in brain tissue of AD patients, and the aims of this study are to examine the molecular forms, cellular distribution and association with AD pathology. This study focused on the cingulate, amygdala, hippocampus and cortical regions in six AD and six control brains. Materials and methods Antibodies Primary antibodies Peroxiredoxin 6 (rabbit) was obtained from Antibody Technology Australia Pty, Ltd, Adelaide, Australia. Tau (mouse) was obtained from Novacastra Laboratories Ltd, Newcastle, UK. Tau (goat) was obtained from Santa Cruz Biotechnology Inc., California, USA. Amyloid beta (mouse) was obtained from Novacastra Laboratories Ltd, Newcastle, UK. Glial fibrillary acidic protein (GFAP) (mouse) was obtained from Novacastra Laboratories Ltd, Newcastle, UK. Human neuronal protein HuC/HuD (mouse) was obtained from Molecular Probes Inc., Eugene, USA. Human HLA (MHC2) (mouse) was obtained from Dakocytomation, Denmark A/S. Myelin basic protein (MBP) (mouse) was obtained from Novacastra Laboratories Ltd, Newcastle, UK. Secondary antibodies Donkey anti sheep Cy3 was obtained from Jackson Immunoresearch, West Grove, USA. Donkey antirabbit Cy3 and Cy5 were obtained from Jackson Immunoresearch, West Grove, USA. Goat antimouse Alexa 488 was obtained from Molecular Probes Inc., Eugene, USA. Goat antirabbit alkaline phosphatase was obtained from Sigma Chemical Company, St Louis, USA. Brain tissue The brain tissue was obtained from the National Health and Medical Research Council South Australian Brain Bank. Brains were removed at autopsy from clinically diagnosed AD cases and confirmed pathologically postmortem. The control cases were obtained from people who died of unrelated causes without diagnosed neurological pathology that was confirmed postmortem. The brains were bisected and one half snap-frozen at −80°C and the other immersion fixed in buffered formaldehyde. Tissue blocks were processed and embedded in paraffin and 8 μm sections were cut and mounted onto gelatin-coated slides. Table 1 provides a list of the brain regions examined and the case details. Table 1List of cases used in cell countingCase numberSexAge (years)DiagnosisRegionPMI (h)Braak stageNIA ReaganAD1M65ADC, A, H, MC 35/6HighAD2F84ADC, A, H, MC 165/6HighAD3M63ADC, A, H, MC 335/6HighAD4M59ADC, A, H, MC205/6HighAD5F81ADC, A, H, MC195/6HighAD6F69ADC, A, H, MC385/6HighC1F79ADLC, A, H, MC4LowC2M69UnkC, A, H, MC48LowC3F61CFC, A, H, MC8LowC4F84SSC, A, H, MC15LowC5F86DMC, A, H, MC17LowC6F86HVDC, A, H, MC6LowPMI postmortem index, AD Alzheimer’s disease, NIA Reagan high or low likely-hood of AD, M male, F female, C cingulate, A amygdala, H hippocampus, MC midfrontal cortex, C control, ADL adeno lung carcinoma, Unk unknown, CF cardiac failure, SS septic shock, DM disseminated malignancy, HVD hypertensive vascular disease Brain homogenate Frozen tissue from midfrontal cortex of three AD brains and three control cases were homogenized in 50 mM Tris, 5 mM EDTA, 0.1% sodium azide, 1 μl/ml pepstatin and leupeptin and 0.3 mM of phenylmethylsulfonyl fluoride. Five milliliters of homogenization buffer was used per gram of brain tissue and homogenized using six passes of a motorized Wheaton Teflon pestle tissue grinder. Homogenates were centrifuged (500×g) for 10 min to remove particulate matter and snap-frozen at −80°C. Table 2 provides the details of these cases. Table 2List of cases used for 2D PAGE and Western blotsCase numberSexAge (years)DiagnosisRegionPMI (h)Braak stageNIA ReaganAD7F69ADMC225/6HighAD8F79ADMC185/6HighAD9F76ADMC235/6HighC6F86HVDMC6LowC7F73ABDMC24LowC8M63SSMC38LowPMI postmortem index, AD Alzheimer’s disease, F female, M male, MC midfrontal cortex, C control, HVD hypertensive vascular disease, ABD anoxic brain damage, SS septic shock Sample preparation Aliquots (100 μl) of Alzheimer’s and control brain homogenates underwent a 2D clean up as per the Amersham Biosciences protocol. The samples were then resolubilized in 300 μl of TUC pH 4–7 buffer containing 7 M urea, 2 M thiourea, 4% CHAPS, 0.5% dithiothreitol, 0.5% IPG buffer pH 4–7. Samples were then assayed for total protein content. Prior to rehydration of the immobilized pH gradient strips, trace amounts of bromophenol blue was added to samples. Table 2 provides a list of cases used in the 2D electrophoresis. 2D polyacrylamide gel electrophoresis (PAGE) Isoelectric focusing Isoelectric focusing was performed on 13 cm immobilized pH gradient strips (pH 4–7NL, Amersham Pharmacia) using an IPGphor (Amersham Pharmacia) isoelectric focusing unit. Samples were applied in gel rehydration (50 V, overnight) using 7 M urea, 2 M thiourea, 4% CHAPS, 0.5% dithiothreitol, 0.5% pharmalyte pH 4–7 with a trace of bromophenol blue as the re-swell buffer with a total loading volume of 280 μl including sample. Stained gels received a 400 μg protein load and Western blotting gels received a 200 μg load. IEF was performed for ∼60 kV h (500 V 30 min, 1,000 V 30 min, 1,000–8,000 V 30 min, 8,000 V 5–7 h, 1,000 V hold to end). SDS PAGE SDS PAGE was performed using 12.5% acrylamide gels. Immobilized pH gradient strips were incubated in SDS equilibration buffer containing 1% w/v dithiothreitol for 15 min followed by SDS equilibration buffer containing 4% w/v iodoacetamide for 15 min. Strips were then placed above the second dimension gels and overlaid with 1% low melting point agarose in Tris glycine gel running buffer. Electrophoresis was performed at 350 V for 2.5–3 h at 10°C. Protein gels were visualized using SYPRO ruby fluorescent stain (incubated for 2 h) and scanned using a Typhoon 9400 Variable Mode Imager (Amersham Biosciences). Western blotting Separated proteins were transferred onto low fluorescence PVDF membrane by semi-dry electroblotting at 0.8 mA/cm2 for 2 h. After blocking, the membranes were incubated in affinity-purified rabbit antiperoxiredoxin 6 antibodies (1/2,000) overnight at 4°C followed by goat antirabbit alkaline phosphatase antibodies (1/10,000) (Sigma) for 1 h and visualized using ECF substrate and a Typhoon 9400 Variable Mode Imager. Immunohistochemistry Localization of peroxiredoxin 6 Immunohistochemical staining of AD and control tissue was carried out as previously described [30]. The cellular distribution of peroxiredoxin 6 was compared with the staining obtained with antibodies to specific cellular markers. Antibodies to GFAP (astrocytes), HLA (MHC2) (microglia), MBP (oligodendrocytes) and HuC/HuD (neurons) were used to map the distribution of peroxiredoxin 6. Colocalization of peroxiredoxin 6, Aβ and Tau Brain sections were incubated with rabbit antiperoxiredoxin 6, mouse anti Aβ peptide and goat anti tau antibodies overnight and detected with secondary antibodies conjugated to fluorescent fluorophores Cy5, Alexa 488 and Cy3, respectively. Sections were examined using a Bio-Rad Confocal laser scanning microscope and Bio-Rad software package, as previously described [30]. Cell counting Cell counting was carried out blind on masked sections, and peroxiredoxin 6 positive astrocytes were counted at 20× magnification with a graticule eye piece (0.0625 mm2) using an Olympus HO-2 microscope on brain tissue from six AD and six control cases. In each case, five regions were randomly selected to give a uniform representation of the white and gray matter in the cingulate, hippocampus, amygdala and midfrontal cortex. The number of positive cells from each of the five areas was averaged and divided by the area of the graticule eyepiece and expressed as cells/mm2. The data was then analyzed using a one-way ANOVA that examined the interactions listed in Table 3. The significance was determined at 0.01 level of probability. Table 3Interactions examined in cell counting dataInteractions examinedF valueCritical F (P = 0.01)Significance (>0.01)Are there differences in AD (C, A, H, MC) gray matter cell counts?1.004.94NSAre there differences in AD (C, A, H, MC) white matter cell counts?1.234.94NSAre there differences in control (C, A, H, MC) gray matter cell counts?0.624.94NSAre there differences in control (C, A, H, MC) white matter cell counts?0.254.94NSAre there differences between gray and white matter cell counts in AD tissue?89.37.21SAre there differences between gray and white matter cell counts in control tissue?14.57.21SAre there differences in gray matter cell counts between control and AD tissue?2997.21SAre there differences in white matter cell counts between control and AD tissue?28.07.21SAD Alzheimer’s disease, NS not significant, S significant, C cingulate, A amygdala, H hippocampus, MC midfrontal cortex Results Molecular forms of peroxiredoxin 6 in human brain We have previously shown that this antibody is specific for a 26 kDa monomer in rat lung and human brain, and the N-terminal sequence is identical to peroxiredoxin 6 [30, 31]. Two-dimensional PAGE and Western blotting of AD and control brain homogenates indicated that this protein exists as three major and five minor forms as shown in Fig. 1a–d. Although the loadings were consistent, the five minor forms could not be detected on all gels. For comparison, the spots were labeled 1–8 (Fig. 2). Fig. 1a, b A two-dimensional PAGE gel and blot of AD brain tissue (Case AD7) and a corresponding control brain gel and blot panels (c, d) (Case C7). pI 4–6 is indicated between the gel and blots and the molecular weight of the standards (Std) in kilo Daltons is shown on the Y-axis. The gels were stained with SYPRO ruby fluorescent stain. The box indicates the location of the molecular forms of peroxiredoxin 6. Some spots are shown on the blots to indicate the range of peroxiredoxin 6 pIsFig. 2A composite figure at higher magnification showing the range of molecular forms in three cases of AD tissue (a AD7, AD8 and AD9) and three cases of control tissue (b C7, C6 and C8). The range of molecular forms 1–8 is shown, but not all forms are present in all cases. Although there are slight variations between blots, the respective pIs are as follows: 1 (5.3), 2 (5.4), 3 (5.6), 4 (5.75), 5 (5.85), 6 (5.9), 7 (5.95), 8 (6.1) The respective pIs of the eight forms are 5.3, 5.4, 5.6, 5.75, 5.85, 5.9, 5.95 and 6.1. The three major forms were 3, 6 and 7 and the average pIs of these were 5.59, 5.88 and 5.96 for AD tissue and 5.59, 5.80 and 5.86 for control tissue. These major forms were consistent in all homogenates. The intensity of the three major forms was analyzed using the Typhoon imager and the intensity of spots 6 and 7 was reversed in the AD and control homogenates (Fig. 2). Minor forms 1 and 2 were present in AD homogenates but absent in the control tissue, while minor form 8 was present in control tissue but absent in AD tissue (Fig. 2). Cellular distribution of peroxiredoxin 6 in control and AD brain tissue Using immunohistochemical markers for specific human brain cell types, we have shown that peroxiredoxin 6 is abundant in reactive astrocytes (Fig. 3a–c) in AD tissue and present in very low levels in neurons (Fig. 3j–l). We were not able to detect peroxiredoxin 6 in either microglia (Fig. 3d–f) or oligodendrocytes (Fig. 3g–i). Fig. 3Confocal localization of peroxiredoxin 6 (P6) with GFAP (astrocyte marker) (a P6, Cy2-green; b GFAP, Cy3-red; c merged image; bar 25 μM), MHC2 (microglia marker) (d P6, Cy2-green; e MHC2, Cy3-red; f merged image; bar 25 μM), MBP (oligodendrocyte marker) (g P6, Cy2-green; h MBP, Cy3-red; i merged image; bar 10 μM), Hu (neuronal marker) (j P6, Cy2-green; k Hu, Cy3-red; l merged image; bar 25 μM). Cell localization was performed using AD tissue There was a marked contrast between the staining of control and reactive astrocytes in AD. Reactive astrocytes in AD were hypertrophied with abundant staining of peroxiredoxin 6, while in control astrocytes the staining was confined to a “halo” around the nucleus (Fig. 4). The increased staining of peroxiredoxin 6 was seen in all regions examined and in both gray and white matter, but the staining was not evenly distributed and some regions on the same section were heavily labeled but adjacent regions showed little staining (Fig. 5). Fig. 4Oil immersion image (×1,000) of typical astrocyte seen in AD cortex (a) and in control cortex (b)Fig. 5Light immunohistochemistry (DAB substrate) of peroxiredoxin 6 staining of astrocytes in AD gray matter (a), AD white matter (b), control tissue gray matter (c) and control tissue white matter (d). All images at ×200 magnification Cell counting In AD and control tissue, peroxiredoxin 6 was colocalized with GFAP, indicating that the upregulation is occurring in astrocytes, which is similar to what was found in PD and dementia with Lewy bodies [30]. Cell counting was conducted on midfrontal cortex, cingulate, amygdala and hippocampus brain regions and in gray and white matter to determine if specific areas or matter type were specifically affected. These are summarized in Fig. 6. A one-way ANOVA was performed comparing the cell counts between AD and control tissue in different regions by tissue and matter type. The results of the interactions examined are shown in Table 3. Fig. 6A bar chart showing the range of cell counts in different brain regions in gray and white matter in control tissue and in AD tissue. WM white matter, GM gray matter, Cing cingulate, Amy amygdala, MFC midfrontal cortex, Hip hippocampus. Counts are shown as mean ± SE, N = 6 in all cases Colocalization of peroxiredoxin 6 and AD pathology The major defining pathology in AD is the presence of extracellular plaques and intracellular hyperphosphorylated tau in the form of tangles. Plaques are present in brain tissue at different stages of development, ranging from diffuse plaques to fibrillar and neuritic plaques. It is generally considered that diffuse plaques do not contain tau, but neuritic plaques contain both Aβ and tau. Colocalization of peroxiredoxin 6 and the Aβ peptide, a major component of plaques, and tau was carried out to determine if peroxiredoxin 6 positive astrocytes were associated with AD pathology. Tau pathology Tau deposits in the form of neurofibrillary tangles were observed in many neurons and neurites, and were often surrounded by peroxiredoxin 6 astrocytes, which made multiple contacts (Fig. 7a–c). Although astrocytes made many contacts with neurons, most neurons had very low levels of peroxiredoxin 6 and it did not appear to be upregulated or to be colocalized with tau in AD. Fig. 7a–c Confocal localization of tau within neurons (a Cy2-green) and peroxiredoxin 6 positive astrocytes (b Cy3-red) and (c) the merged image in AD brain tissue (bar = 10 μM). d–f Confocal localization of Aβ peptide (d Alexa 488, green) and peroxiredoxin 6 positive astrocytes (e Cy3-red) and (f) the merged image in a diffuse plaque in AD brain tissue (bar = 50 μM). g–j Confocal localization of Aβ peptide (Alexa 488, green) and peroxiredoxin 6 positive astrocytes (Cy5-light blue) in four different plaques in AD brain tissue (bar = 50 μM). k Immunohistochemical staining (DAB substrate) showing peroxiredoxin 6 staining of astrocytes and astrocyte foot processes in the walls of blood vessels in the midfrontal cortex in Alzheimer’s disease (×400). Arrows indicate peroxiredoxin 6 positive astrocytic processes Plaques Diffuse plaques contained strongly stained aggregations of Aβ peptide. Peroxiredoxin 6 staining was confined to astrocytes and their projections that were in close proximity to plaques (Fig. 7d–h). Some astrocytes were observed within plaques, but generally the astrocyte cell bodies were on the periphery with their processes projecting into the plaques. Some plaques contained the beginnings of a dense core of amyloid and took on a “wagon wheel” appearance (Fig. 7i, j). Diffuse plaques and those with Aβ beginning to aggregate appeared to elicit a stronger peroxiredoxin 6 astrocytic response than those with a dense core of Aβ. Peroxiredoxin 6 staining of blood vessels In many regions of the AD brain, there appeared to be strong peroxiredoxin 6 staining in the walls of many blood vessels, which could be seen to have multiple contacts with strongly staining activated astrocytes (Fig. 7k). Discussion Peroxiredoxin 6, also known as 1-Cys peroxiredoxin or non-selenium glutathione peroxidase, is a human brain antioxidant enzyme that is abundant in activated astrocytes and present in low levels in neurons. On exposure to hydrogen peroxide, the N-terminal peroxiredoxin 6 corresponding to Cys 47 is readily oxidized, but as yet its redox partner has not been identified. Glutathione, lipoic acid and cyclophilin have all been implicated [13, 23, 29]. In addition to its peroxidase activity, peroxiredoxin 6 also has phospholipase A2 activity that might be significant in the brain, which has a high lipid content [9]. Peroxiredoxin 6 staining of astrocytes is markedly elevated in many brain regions in PD and dementia with Lewy bodies [30]. We now show that it is upregulated in the astrocytes in AD and describe for the first time the range of molecular forms of peroxiredoxin 6 in human brain in both control and AD tissue. Three major variants and five minor forms were identified with pIs ranging from 5.3 to 6.1. There were three major differences in the peroxiredoxin 6 variants between control and AD tissue. The AD tissue had the two more acidic forms 1 and 2 that were not present in control tissue; the intensity of the major forms 6 and 7 were reversed between AD and control tissue and minor form 8 was absent in AD tissue. We can only speculate on the significance of these findings at this stage, but different forms may have different affinities for different substrates that might be more abundant under various pathological conditions or the more acidic forms are the result of oxidation [11]. At this stage, we cannot determine whether these differences are true isoforms or post-translational modifications of the same protein, but would suspect the latter. These findings are in agreement with two previous proteomic studies in an AD mouse model and in human brain AD tissue, which reported elevated levels of peroxiredoxin 6 [33, 34]. A more recent study using PC12-resistant and PC12 cells treated with Aβ also reported increased levels of peroxiredoxin 6 [11]. Aβ-treated cells contained an extra acidic form, which was suggested to be the result of oxidation of the catalytic cysteine to a sulfenic, sulfinic or sulfonic derivative. We found two minor acidic forms in AD tissue, which is also consistent with this report. Tissue selection is also a variable that needs to be considered. While the AD tissue had all the hallmarks of AD pathology and was easy to define, age-matched control tissues are somewhat more variable. All the control tissue was from people without clinically defined AD, but most brains in this age group usually contain some neurodegenerative pathology. The age span of the two groups was similar, 59–84 years for the AD group and 61–86 years for the control group, but the mean age for the control group was 7 years older. We feel confident that, although some control tissue may have had some elements of subclinical AD pathology, the reduced level of astrocyte activation and the clear differences in the cell counting would indicate that all the control tissues were not experiencing high levels of oxidative stress as compared to the AD tissue. Previously, we had shown that peroxiredoxin 6 was abundant in astrocytes in PD and dementia with Lewy body tissue [30]. We have now used cellular markers to other brain cell types to show that in addition to astrocytes there is a low level in the cytoplasm of most neurons. Unlike astrocytes, neurons do not appear to have increased peroxiredoxin 6 staining in AD tissue. We could not detect any peroxiredoxin 6 staining in microglia or oligodendrocytes. Cell counting of peroxiredoxin 6 positive astrocytes indicated that there was a marked increase in both the gray and white matter in AD tissue compared with control tissue. Gray matter had more peroxiredoxin 6 positive astrocytes than white matter in both AD and control tissue, and this difference was more pronounced in AD tissue with F values of 89.0 and 14.5, respectively. Considering that the major pathology (plaques and tangles) is located in the gray matter, it would be expected that oxidative stress is more pronounced in this area. Cell counting indicated that the peroxiredoxin 6 positive astrocyte numbers were elevated in all regions examined, suggesting that the insult activating astrocytes is wide spread and not confined to any specific region in either the gray or white matter. Indeed, extracellular plaques were observed in all regions examined in AD tissue. Although all regions examined showed astrocyte activation, this was not uniformly distributed through all areas, with some areas showing extensive astrocyte activation while adjacent areas were similar to controls. There are two defining pathologies in AD. These are the presence of extracellular plaques comprising abundant deposits of Aβ peptides, which later aggregate to form amyloid [25] and intracellular hyperphosphorylated tau aggregations in the form of paired helical filaments, which are termed tangles [14, 22]. The two well-described forms of plaques have been termed diffuse plaques and neuritic or senile plaques [12, 18], although other intermediate forms such as fibrillar plaques have been described [12]. Diffuse plaques are spherical structures of less dense Aβ deposits and are considered an early stage of plaque formation. In diffuse plaques, aggregations of Aβ exists as nonamyloid deposits, but as the plaque progresses, the Aβ and other protein aggregations change from the normal conformation into a β-sheet structure, termed amyloid, and take up specific dyes like Congo red and thioflavin [4]. These plaques do not generally contain tau; however, in many confocal images, diffuse plaques contain focal points of tau, which are probably enlarged neurites. Neuritic or senile plaques contain a dense core of amyloid enriched with Aβ peptide with a corona of less dense amyloid. The diffuse plaques appeared to elicit a stronger astrocytic response based on the level of peroxiredoxin 6 staining than the neuritic plaques. Our images suggest that once the Aβ forms amyloid and condenses, it is less stimulating toward astrocytes and perhaps the formation of amyloid is a protective mechanism to inactivate the toxicity of soluble Aβ. These observations are consistent with the findings of Kawaguchi-Niida et al., who reported that the highly reactive carbonyl crotonaldehyde generated during lipid peroxidation is localized to reactive astrocytes, microglia and diffuse plaques but undetectable in amyloid cores [19]. Although the activation of astrocytes and upregulation of peroxiredoxin 6 would be considered a protective response, in cell culture, soluble Aβ has been shown to stimulate nitric oxide synthetase and the production of nitric oxide, which is very damaging to neurons [17]. This response is further upregulated by other inflammatory cytokines such as interleukin 1β and tumour necrosis factor-α, which are reported to be released from astrocytes [1]. At this stage, we are not sure of the net benefit to neural tissue of activated astrocytes with elevated levels of peroxiredoxin 6, concomitant with the increased production of nitric oxide. Neurofibrillary tangles comprising hyperphosphorylated tau were observed in many regions in AD tissue. Two antibodies were used to localize tau, a monoclonal antibody (Novacastra) and one specific for paired helical filaments (Santa Cruz) and both produced a similar staining pattern. Neurons containing tau were not surrounded by activated astrocytes as were the diffuse plaques and to a lesser extent the neuritic plaques, suggesting that these cells are not secreting toxic or activating products that are stimulating to astrocytes. In PD, Lewy bodies contained a dense core of peroxiredoxin 6 staining suggesting that it was trying to detoxify oxidative stress produced by Lewy bodies, but we did not observe any interaction between tau and peroxiredoxin 6 in AD tissue. Although there is considerable evidence to implicate oxidative stress in the hyperphosphorylation of tau, this does not appear to upregulate peroxiredoxin 6 in tau positive neurons. In some regions, the walls of blood vessels exhibited strong peroxiredoxin 6 staining, which also elicited a strong astrocytic response in surrounding astrocytes. These activated astrocytes had many processes in contact with the endothelial cells. Although vascular amyloid deposits were seen in some sections and is a key feature of AD [28], the upregulation of peroxiredoxin 6 appeared to occur in regions without distinct Aβ staining. Aβ has been shown to induce vascular vasoconstriction in rat skin, which can be reversed by superoxide dismutase and catalase, indicating that Aβ is capable of generating ROS within blood vessels [21]. The blood brain barrier consists of the capillary endothelial cells, and the foot processes of the astrocytes and the fused basal lamina of both cells. The peroxiredoxin 6 staining within the capillaries appears to be in the foot processes of the astrocytes. Whether this increased staining is due to ROS generated from within neural tissue or reflects a more systemic pathology derived from the circulation is unclear. In conclusion, we have shown for the first time the range of peroxiredoxin 6 variants, which are either isoforms or post-translational modifications in human brain tissue. We have also shown that peroxiredoxin 6 is primarily an astrocytic enzyme with very low levels in neurons and is not detectable in microglia or oligodendrocytes. This enzyme is markedly elevated in astrocytes in both white and gray matter in AD. Strongly staining peroxiredoxin 6 astrocytes appear to be involved in the detoxification of diffuse plaques and to a lesser extent in neuritic plaques, but is not associated with tau aggregations within neurons. From this work and previous work on PD and dementia with Lewy bodies, we suggest that peroxiredoxin 6 is a major antioxidant enzyme in human neural tissue.

Qual_Life_Res-3-1-1915604

Quality of life in couples living with Huntington’s disease: the role of patients’ and partners’ illness perceptions

Research suggests that chronically ill patients and their partners perceive illness differently, and that these differences have a negative impact on patients’ quality of life (QoL). This study assessed whether illness perceptions of patients with Huntington’s disease (HD) differ from those of their partners, and examined whether spousal illness perceptions are important for the QoL of the couples (n = 51 couples). Partners reported that their HD-patient spouses suffered more symptoms and experienced less control than the patients themselves reported. Illness perceptions of patients and partners correlated significantly with patient QoL. Partners’ beliefs in a long duration of the patients’ illness and less belief in cure, were associated with patient vitality scores. Suggestions for future research emphasize the importance of qualitative research approaches in combination with cognitive-behavioural approaches. Quality of life research is increasingly being directed at both the impact of the social environment on the quality of life (QoL) of chronically ill individuals, and at the toll the illness exacts on the QoL of close family members [1] One of the most influential members of the social network of chronically ill individuals is the spouse. Research conducted in chronically ill individuals and their healthy spouses revealed that the spouses’ role problems [2, 3], their unsupportive behaviour [4–6], and the marital relationship itself [7] are some of the factors influencing patients’ QoL. Studies examining adaptation in caregivers have identified negative social support [8], patients’ cognitive impairment [9, 10], and the caregivers’ overestimation of the patient’s functional disabilities [11] as some of the factors influencing the partners’ QoL. Recent findings suggest that patients’ QoL is also dependent on beliefs fostered by the social environment regarding the patient’s illness. Illness beliefs or illness perceptions can be subdivided into five components: identity (the label the patient places on the illness and the symptoms he/she experiences), cause (the personal ideas patients have about the cause of their illness), timeline (the duration of the illness according to the patient), consequences (the expected effects and outcome of the illness), and cure/control (the curability or controllability of the illness according to the patient) [12]. In the area of physical health problems, illness perceptions have been found to be related to the ways patients react and cope with their illness. A strong illness identity combined with a perceived long duration and perceived more severe consequences of the disease for daily life have been associated with poor physical and psychological well-being. Stronger beliefs regarding the curability or controllability of a disease have been found to have positive effects on patients’ QoL [13, 14]. Research shows that illness beliefs of spouses significantly affect patients’ beliefs, the coping mechanisms they adopt to deal with their disease, and ultimately their functioning and well-being [15, 16]. Heijmans and colleagues found that dissimilarities between the illness perceptions of chronically ill patients (patients with Addison’s disease and patients with Chronic Fatigue Syndrome) and those of their healthy spouses were associated with higher impairments in predicted patients’ adaptive outcome [15]. In another study, Figueiras and Weinman explored whether the degree of similarity in patients’ and spouses’ illness perceptions was related to recovery following myocardial infarction [16]. Few studies have focused on the role of illness perceptions for caregiver outcomes, and none have examined the role of patients’ illness perceptions on the QoL of their healthy partners. Barrowclough et al. [17] and Fortune et al. [18] studied illness perceptions in carers of schizophrenia patients. Their findings suggest that carer perceptions about identity, consequences and control may have important implications for carer outcomes in schizophrenia. In this study we will focus on Huntington’s disease (HD) which is an inherited neurodegenerative disorder, characterized by clearly defined clinical features such as involuntary movements and hypokinesia, dementia, and personality changes. The first symptoms of HD typically manifest between the ages of 35 and 45, and the disease has a mean duration of 16 years [19, 20]. At present, there is no cure for HD. Little empirical data has been gathered on the psychosocial aspects of living with HD, but some research findings suggest that HD contributes to marital breakdown [21–23]. One of the most important reasons for this is believed to be the changes HD brings about in the relationship. Spouses of HD patients often consider their partners to be lost [23], particularly when dementia and changes in personality and behaviour develop [22, 24]. In addition, over time the spouse takes on an increasing nursing role, which creates a psychological distance between the patient and his/her spouse [22]. HD greatly impacts the patients’ physical and psychosocial well-being [25, 26], and places a heavy burden on families [27–29]. Interestingly, some research has indicated that there is a discrepancy between the aspects of HD the patient finds most disturbing and those the spouse or partner finds most disturbing. Partners are most disturbed by mental and personality changes in the patients [30, 31]. In all studies examining the role of spousal illness perceptions for patient outcomes, researchers have concentrated on the degree of similarity / dissimilarity in patients’ and spouses’ illness perceptions as a predictor of patient outcome, thereby ignoring alternative possible relationship patterns between spouses’ illness beliefs and patient outcome. Results from Figueiras and Weinman [16] showed that not only similar positive perceptions in couples, but also conflicting representations were predictive of lower levels of disability (as compared to similar negative perceptions). This suggests that it might be more important that at least one member of a couple has positive perceptions. However, the results obtained in the Heijmans et al. [15] study show that better patient adjustment in Addison’s Disease is related to spouses’ negative perceptions about timeline, while better patient adjustment in Chronic Fatigue Syndrome is related to spouses’ positive perceptions of a short illness duration. From these studies it seems that both contrasting and concordant perceptions in couples can be related to better patient adjustment, and further study is needed to evaluate the extent to which spouses’ illness perceptions can influence outcomes which have been found to be related to the patient’s own perceptions. Thus, the first aim of this study was to examine if patients and partners hold similar views about HD. Our second aim was to compare the relationship patterns between patients’ and partners’ own illness perceptions and QoL. The third aim was to contribute to the understanding of the cognitive factors related to QoL in couples dealing with HD, by examining which spousal illness beliefs are important in both patients’ and partners’ quality of life. Method Sample and procedure Participants in this study were 51 couples (HD patients and their partners) who were recruited from the outpatient clinic of the Department of Neurology of the Leiden University Medical Centre (LUMC; n = 14 couples), and the Dutch Huntington Association (n = 37 couples). HD patients from the LUMC were selected if they had received a clinical diagnosis of HD at least 1 year prior to commencement of the study, and if they were capable (mentally, emotionally, and physically) of participating in an interview lasting approximately 2 h. The selected patients (n = 75) and their spouses were invited by post to participate. A total of 36 patients and 19 spouses agreed to participate. The main reason for not participating were the verbal communication difficulties that patients were experiencing, and having participated in other research projects before. Non-participants did not differ from participants with respect to age, sex, or duration of HD. HD patients and their spouses recruited from the Dutch Huntington Association were invited to participate by means of a letter sent to all the members of the association (n = 1,450). The selection criteria for participation were the same as those applied to the patients selected from the Department of Neurology of the LUMC. A total of 41 patients and 71 spouses agreed to participate. No information could be gathered on non-participants due to the fact that members of the Dutch Huntington Association are not registered on the basis of their patient status. Members include HD patients and their partners, as well as their family members, friends, and researchers in the field of HD. A total of 51 couples (married or living together) were identified. Patients and their partners were interviewed separately at their homes by a psychologist. Measures Demographic variables—HD patients and their partners were asked their age, sex, marital status, the duration of their relationship, number of children, their employment status, and the duration of HD. Unified Huntington Disease Rating Scale [32]—We used the motor section of the UHDRS to assess HD patients’ motor performance. The motor section is composed of 20 items rating ocular motor function, dysarthria, chorea, dystonia, gait, and postural stability. The Total Motor Score (TMS) is the sum of all the individual items, higher scores indicating worse motor performance (maximum score = 124). Mini-Mental State [33]—We used this scale to assess HD patients’ cognitive performance. It comprises 11 items covering a number of cognitive domains including: orientation, registration, attention, memory, language and visuo-constructional abilities. The maximum score is 30, lower scores indicating worse performance. Scores of 20 or less have been associated with dementia, delirium, schizophrenia or affective disorder [33]. The Illness Perception Questionnaire [34]—This questionnaire was used to assess the illness perceptions of HD patients and those of their partners (not yet validated in Dutch). The latter were interviewed by means of a partner-version of the IPQ [35]. The IPQ consists of the following five subscales: “Identity”, “Timeline”, “Cause”, “Consequences”, and “Cure/Control”. In this study we divided the last subscale into “Cure” and “Control”, thus creating a sixth scale. This is in line with current revisions of the IPQ in which these scales are generally separated [36]. The “identity” scale was composed of 24 items, each corresponding to a symptom commonly reported in HD. Patients are asked to rate whether or not they have experienced each symptom since their illness began, and if they believe the symptom to be specifically related to their illness (yes or no). Partners were asked whether or not their partner (the patient) had experienced each symptom since onset of their illness, and to report if they believed the symptom to be specifically related to their partners’ illness (yes or no). The summed yes-rated items on the second questions were divided by the number of items to form the illness identity scale, with higher scores indicating a stronger belief that the experienced symptoms are part of the patient’s illness. Cronbach’s alpha for patients and partners was 0.88 and 0.82, respectively. For the remaining scales, patients and their partners were asked to indicate whether they agreed with statements on a five-point scale, ranging from “strongly agree” to “strongly disagree”. The “Timeline” scale contained two statements about the perceived duration of the disease, with high scores indicating strong beliefs in a chronic long-term disease. Cronbach’s alpha for patients and partners was 0.45 and 0.72, respectively. “Consequences” consisted of five items assessing beliefs about the impact of HD on everyday life. High scores indicate stronger beliefs in serious consequences of the disease. Cronbach’s alpha for patients and partners was 0.69 and 0.61, respectively. High scores on the “Cure” scale (two items) indicate strong beliefs in the effectiveness of treatment. Cronbach’s alpha for patients and partners was 0.77 and 0.66, respectively. “Control” contained two items pertaining to the degree to which patients and partners believe they have the ability to influence the course of illness. Cronbach’s alpha for patients and partners was 0.85 and 0.83, respectively. The weight of all the items per scale were summed, and divided by the number of items, with the exception of those of the “Cause” subscale. Each item in this subscale was considered individually because each causal item represents a specific causal belief. The Medical Outcome Study 36-item Short Form Health Survey [37]—We assessed the QoL of HD patients, and the QoL of partners by means of seven subscales of the MOS SF-36 (“Physical functioning”, “Role functioning-physical”, “General health”, “Vitality”, “Social functioning”, “Role functioning-emotional”, and “Mental health”). The raw scores are transformed in order to obtain a 0–100 scale, with higher scores indicating a better outcome. Analysis plan First, the samples of HD patients and partners from the outpatient clinic and the Dutch Huntington Association involved in this study were compared for demographic and illness related variables (age, sex, marital status, duration of the relationship, number of children, employment status, duration of HD, patients’ TMS, and patients’ scores on the MMS) by means of t-tests. To examine the extent of (dis)agreement that HD patients and their partners held with regard to their beliefs about HD, Pearson correlations and paired sample t-tests were computed. To assess whether spousal illness perceptions are related to patients’ and partners’ QoL, we conducted hierarchical regression analyses with scores on the MOS SF-36 as criteria. For patients, illness related variables (disease duration, TMS, and MMS scores) were entered as control variables, prior to the steps containing the patients’ and partners’ scores on the IPQ (step 2 and 3, respectively). For partners, parallel analyses were conducted, but no control variables were entered. The variables for the regression analyses were selected based on an examination of bivariate correlations between the illness perception dimensions and the QoL scales. Results Sample characteristics Couples recruited from the Department of Neurology of the LUMC did not differ significantly in demographic or illness related variables from couples from the Dutch Huntington Association. Further analyses were thus conducted on the combined data. In total, 51 HD patients (28 males, 23 females) and their partners (23 males, 28 females) were included in this study. The mean age of both patients and partners was 51 years (SD = 10). Forty-seven couples were married and four were living together. The mean duration of the relationship was 25 years (range: 2–48 years), and the mean number of children was 2 (range: 0–6). Eight couples did not have children. Thirty-nine patients (76.5%) were unemployed, of which 28 (54.9%; 19 males, 9 females) stopped working because of HD, whereas 10 (19.6%; 6 males, 4 females) were still working, and 2 (3.9%; males) were retired and receiving pension. Thirty-four partners (66.7%; 19 males, 15 females) were employed, 15 (29.4%) were unemployed (for different reasons including marriage and childbearing; 5 females had never had gainful employment), and 2 (3.9%; males) were retired and receiving a pension. The mean duration of HD was 7 years (SD = 5; range: 1–24). Patients’ mean Total Motor Score (TMS) was 37.8 (SD = 29.5), and their mean score on the MMS was 23.5 (SD = 3.9). Patient-partner differences in illness perceptions HD patients and their partners did not differ significantly on the “Timeline”, “Consequences”, “Cure” or “Cause” subscales of the IPQ (see Table 1). Both groups of respondents reported perceiving HD as having a long duration and having many consequences for their daily lives, without either of them believing in a cure for HD. Patients and partners (with the exception of four patients and one partner) attributed HD to genetic causes. Six HD patients (11.8%) reported believing that stress was an important contributing factor to the development of the symptoms of HD they were experiencing. Table 1Means, standard deviations (SD), correlations and t-test statistics comparing Huntington’s Disease (HD) patients and their partners on the Illness Perception Questionnaire (IPQ) IPQ subscalesHD patientsPartnersPatient-partner correlation coefficientaPatient-partner differencebMean (SD)Mean (SD)tIdentity.44 (.22).55 (.22).57***−3.59***Timeline4.70 (.63)4.50 (.81).101.44Consequences3.60 (1.19)3.82 (1.01).29*−1.16Cure1.72 (1.77)1.69 (1.69).90***.27Control3.11 (1.72)2.36 (1.58).30*2.73**aPearson correlation. b Paired t-test* P < .05; ** P < .01; *** P < .001 Partners reported that the HD patients were suffering from significantly more symptoms (on the “Identity” subscale) of HD than the patients did themselves, and reported experiencing significantly less control over HD than the patients did. Correlations between illness perceptions and QoL Tables 2 and 3 display the bivariate correlations between the study variables. As expected, examination of the correlations between patients’ and partners’ illness perceptions and patients’ QoL (presented in Table 2) shows that patients’ QoL was most strongly associated with their own illness perceptions. A higher QOL correlated with a less strong illness identity, a longer perceived illness (and thus life) duration, less perceived consequences, more control, and less belief in treatment. With regard to partners’ perceptions, the same associations (albeit fewer) were found, with the exception of partners’ cure perceptions. Partner, but not patient, belief in cure through treatment was significantly related to patients’ vitality and social functioning ratings. Also, partners’ identity and consequences perceptions were not significantly related to patients’ vitality, social functioning, and mental health ratings. Except for the MOS subscales “Role Functioning - emotional” and “Mental Health”, illness perceptions of both patients and partners correlated significantly with patient QoL. Thus, regression analyses were conducted on the remaining MOS subscales, including only the illness perceptions dimensions that correlated significantly with QoL. Table 2Pearson correlations between Illness Perceptions (IPQ) and patients’ QoL (MOS SF-36)IPQ patientsIPQ partnersIdentityTimelineConsequencesControlCureIdentityTimelineConsequencesControlCureMOS subscalesPhysical functioning−.51***.48***−.31*.37**−.33*−.43**.28*−.31*.39**−.39**Role functioning (physical)−.31*−.29*General health−.39**−.31*−.31*Vitality−.44**.34*−.35*.33*−.35*Social functioning−.32*−.38**−.31*Role functioning (emotional)Mental health−.40**−.39**Note: Only significant correlations are depicted* P < .05; ** P < .01; *** P < .001 Bivariate correlations between illness perceptions and partners’ QoL are presented in Table 3. The quality of life of partners was associated with their own beliefs about the patients’ illness, with a stronger belief in a long duration of the patients’ illness being related to better physical functioning and more vitality, and less perceived consequences being related to better physical role functioning, more vitality, and to better mental health. Partners’ vitality and mental health were also associated with illness perceptions of patients. Stronger patient beliefs in control over the illness, and less serious perceived consequences were related to better partner QoL. Interestingly, the only significant correlations between partners’ ratings of general health and emotional role functioning, and illness perceptions were correlations with patient beliefs in control over the illness. Table 3Pearson correlations between Partners’ and Patients’ Illness Perceptions (IPQ), and Partners’ QoL (MOS SF-36)IPQ partnersIPQ patientsIdentityTimelineConsequencesControlCureIdentityTimelineConsequencesControlCureMOS subscalesPhysical functioning.29*Role functioning (physical)−.38**General health.28*Vitality.33*−.29*.30*Social functioningRole functioning (emotional).33*Mental health−.40**−.33*Note: Only significant correlations are depicted* P < .05; ** P < .01; *** P < .001 Importance of spousal illness perceptions for patients’ QoL To determine the relative extent that partners’ illness perceptions are predictive of patients’ QoL, hierarchical stepwise multiple regression analyses were conducted for patients’ physical functioning, role functioning (physical), general health, vitality and social functioning. The results of the regressions are summarized in Table 4. After controlling for patients’ illness related variables and their own scores on the IPQ, partners’ IPQ scores added a significant amount of explained variance (13%) to patient scores on the “Vitality” subscale only. A stronger partner belief in a long duration of the patients’ illness (ß .30, P < .05), and less belief in cure through treatment (ß −.33, P < .05) both added to the prediction of higher patient vitality. Table 4Hierarchical regressions examining if spousal illness perceptions explain additional variance in patients’ QoLStep and variablesAdj. R²R² changeF for R² changeMOS physical functioning1. Control variablesa.49.5217.05***2. Illness perceptions (patient): identity, timeline, consequences, control, cure.59.143.29*3. Illness perceptions (partner): identity, timeline, consequences, control, cure.64.082.06MOS role functioning (physical)1. Control variablesa.01.071.232. Illness perceptions (patient): cure.11.095.83*3. Illness perceptions (partner): cure .09.000.01MOS general health1. Control variablesa.10.152.84*2. Illness perceptions (patient): identity, consequences.18.113.24*3. Illness perceptions (partner): identity.16.000.08MOS vitality1. Control variablesa.12.183.35*2. Illness perceptions (patient): identity, timeline, consequences.29.204.65**3. Illness perceptions (partner): timeline, cure.41.13 5.62**MOS social functioning1. Control variablesa.01.050.832. Illness perceptions (patient): identity, consequences.11.154.18*3. Illness perceptions (partner): cure.16.063.51aDisease duration, TMS, and MMS scores * P < .05; ** P < .01; *** P < .001 The variance in patients’ scores on the other subscales of the MOS SF-36 was explained mainly by patients’ own IPQ scores, with the amount of added explained variance ranging from 9 to 20%. Less perceived consequences, a less strong illness identity, more control, and less belief in treatment contributed significantly to better QoL. Importance of patients’ illness perceptions for partners’ QoL To determine the relative extent that patients’ illness perceptions are predictive of partners’ QoL, hierarchical stepwise multiple regression analyses were conducted for partners’ vitality and mental health. The results of the regressions are summarized in Table 5. Patients’ IPQ scores added a significant amount of explained variance (8%) to partner scores on the “Vitality” subscale only. Stronger patient beliefs in control over the illness (ß .29, P < .05) added to the prediction of higher partner vitality. Table 5Hierarchical regressions examining if spousal illness perceptions explain additional variance in partners’ QoLStep and variablesAdj. R²R² changeF for R² changeMOS vitality1. Illness perceptions (partner): timeline, consequences.17.216.26**2. Illness perceptions (patient): control.24.08 5.05*MOS mental health1. Illness perceptions (partner): consequences.14.169.04**2. Illness perceptions (patient): consequences.17.053.11* P < .05; ** P < .01; *** P < .001 Most variance was explained by partners’ own IPQ scores, with amount of explained variance ranging from 16 to 21%. A longer perceived illness (and thus life) duration, and less perceived consequences contributed significantly to better QoL. Discussion Our results indicate that HD patients and their partners did not differ significantly in their beliefs regarding the duration, consequences, causality, and curability of the disease. HD was attributed, in all but four HD patients and one partner, to a genetic cause. There were however, significant differences between the illness identity of HD patients and that of their partners. Partners attributed significantly more symptoms to HD than patients. In addition, HD patients and their partners differed significantly in the degree of control they believed they had over the disease process. HD patients perceived their disease as being more controllable than their partners. In general, as was the case in couples dealing with Addison’s disease [15], patients held more positive beliefs about HD than their partners did. Examining the relationships between spousal illness perceptions and patients’ QoL, the results indicate that none of the partners’ perceptions (whether consonant or dissimilar) are relevant in being associated with patients’ quality of life in the areas of mental health and emotional role functioning. Partners’ identity and consequences perceptions (whether consonant or dissimilar) are not associated with patients’ quality of life in the areas of social functioning and vitality. Partners’ consequences perceptions (whether consonant or dissimilar) are not relevant for patients’ quality of life in the area of general health. Also, less partner (but not patient) belief in cure through treatment was related to higher patients’ vitality and social functioning ratings. Thus, our results suggest that except for patients’ physical functioning, it is of limited use to investigate (only) dissimilarities in illness perceptions of patients and partners as determinants of patient outcomes in HD. With regard to the third aim of this study, although patients’ and partners’ own illness perceptions explained the largest amount of variance in QoL, spousal illness perceptions were related to patients’ and partners’ quality of life, but only to their vitality ratings. After controlling for patients’ illness related variables and patients’ own beliefs, a stronger partner belief in a long duration of the patients’ illness and less belief in cure through treatment both added to the association with higher patient vitality. The results are in line with those of Heijmans et al. [15], who found positive relationships between spousal maximization of illness duration and patients’ vitality ratings in Addison’s disease. Stronger patient beliefs in control over the illness added to higher vitality scores in partners. From both our results and those of Heijmans et al., it appears that it is most beneficial to have a spouse who is realistic (albeit negative) about the possibilities for cure, and who expects the illness to be long-lasting. Research on how realistic optimism may impact on vitality in healthy persons and patient samples supports these findings [38, 39]. The results obtained in this study must be regarded in the light of some limitations. The lack of information on the non-response rate in the sample of members of the Dutch Huntington Association makes selection bias likely. Also, the patients involved in this study were in the early to middle stages of HD, as shown by their scores on the UHDRS and MMS. This prevents extrapolation of our results to other samples of patients with Huntington’s disease. On the other hand, our results indicate that the patients in our study did not differ significantly from those patients with Huntington’s Disease involved in other studies with regard to important disease related characteristics such as motor and cognitive functioning [40]. Our results could, therefore, be instrumental in future research on quality of life in patients (and their partners) who are in these stages of HD. The relatively small number of couples included and the relatively large number of variables entered into the regression analyses must be taken into account as well. More research is justified on the unique contributions of the spouses’ perceptions of the illness that may be associated with outcome in chronically ill patients and their partners. Examining these associations further and studying which psychological mechanisms may be involved, for instance by interviewing pairs in whom these associations are clearly discernable, are areas for future research. From a clinical perspective, an understanding of the cognitive factors that are related to quality of life in couples dealing with chronic illness will help to guide family interventions. Given the relative paucity of research on biopsychosocial aspects of Huntington’s disease, qualititative research on quality of life may be helpful in future research, e.g., in the work by Brouwer-Dudokdewit et al. [41], where qualitative research in a case-study format that was carefully embedded in a theoretical framework helped explore quality of life issues in pre-symptomatic testing for HD. These researchers emphasize the relevance of adding an existential and/or spiritual approach in exploring quality of life issues in HD patients [42, 43]. Empirical studies in other neurological disorders support these suggestions: Hodgson et al. [44], for example, describe the lives of 10 couples living with Parkinson’s disease, and outline how these couples preferred a multidisciplinary approach to their treatment and believed in taking an active role in their health care. Qualitative approaches to assessing QoL in Huntington’s disease spousal carers are described in a recent paper by Aubeeluck & Buchanan [29], where visual representations of QoL were gathered by using ‘Photovoice’: spousal carers photographed and described elements of their life. Finally, cognitive-behavioural approaches in patients with dementia and their caregivers were recently shown in a randomized controlled trial to result in improvements in patients’ daily functioning and reduced burden in the caregiver. These studies and ours, illustrate possible directions for future research in patients (and their partners) who experience an extreme negative impact on quality of life [45].

Sleep_Breath-4-1-2270921

In-home evaluation of efficacy and titration of a mandibular advancement device for obstructive sleep apnea

There is increasing evidence that mandibular advancement devices (MADs) can be an effective treatment for some patients with obstructive sleep apnea, a highly prevalent chronic disease. In this study, the objectives were to objectively assess the effectiveness of MAD therapy using a limited channel recorder, and to develop a model for identifying patients who may be appropriate for MAD therapy as the initial treatment option. Thirty patients were prospectively recruited and studied at two independent dentist offices and the participants’ homes. Subjects wore the ARES Unicorder for two nights before insertion of the MAD, and again when the dentist determined that the patient had reached the titration endpoint. Self-reported measures of depression, sleepiness, and quality of life were obtained pre- and posttreatment. The reviewer was blinded to the study status while the physiological signals were being visually inspected. Significant reductions in the apnea/hypopnea index (AHI), hypoxemia measures, and snoring level were observed posttreatment. Twenty-seven of the 30 (90%) patients had a posttreatment AHI (using a 4% desaturation for hypopneas) below a clinical cut-off of 10. All but one patient (97%) exhibited at least a 50% decrease in AHI or had a posttreatment AHI ≤ 10. Significant differences in body mass index, weight, and neck circumference in patients with posttreatment AHIs above and below a clinical cut-off of five were identified. The linear regression used to predict the posttreatment AHI using pretreatment data resulted in an R2 of 0.68. The model correctly predicted two patients who were unable to obtain a posttreatment AHI of 10 or less. This study was designed to demonstrate two models of collaboration between a dental sleep medicine specialist and a sleep medicine physician in the monitoring of a patient treated with a MAD. The outcome data suggest that the limited channel recording system can be used as an alternative to laboratory polysomnography to reduce the cost of MAD treatment, and to improve the quality and consistency of posttreatment patient care. Introduction Obstructive sleep apnea (OSA) is the most common disorder observed in the practice of sleep medicine and is responsible for more mortality and morbidity than any other sleep disorder [1]. Although characterized over 40 years ago [2, 3], OSA has only recently gained recognition as one of the world’s most prevalent, underdiagnosed disorders [4, 5]. Due to associated morbidity, OSA has been identified as a major public health concern [6]. The disorder is characterized by frequent loud snoring and recurrent failures to breathe adequately during sleep (termed apneas or hypopneas), usually as a result of collapse of the upper airway. Mandibular advancement devices (MADs) are being increasingly recognized as a treatment alternative to continuous positive airway pressure (CPAP), particularly for patients with mild to moderate OSA [7–12]. MADs are designed to protrude the mandible and thus the tongue and epiglottis during sleep, preventing airway occlusion [13]. Comparisons of MAD to CPAP have revealed that although MADs are less efficacious than CPAP in reducing the RDI they are, on average, used more frequently, preferred by more patients and more readily accepted than CPAP [14]. Recent studies demonstrated that both MADs and CPAP can reduce the 24-h diastolic blood pressure by an average of 3 mmHg after 4 weeks of treatment [15–17]. Similarly to CPAP, self-reported improvements in quality of life and decrease in sleepiness are reported in the majority of patients [14]. The primary goal of this study was to use a self-applied limited channel recorder to objectively assess the effectiveness of MAD therapy when the dentist determined the patient had reached the titration endpoint. A second goal was to determine factors which might predict successful treatment outcomes and provide a more refined method for identifying patients who may be appropriate for MAD therapy as the initial treatment option. Finally, we wanted to demonstrate two potential models of collaboration between the dentist and sleep medicine physician in monitoring MAD treatment outcomes. Materials and methods Eleven females and 19 males were recruited from two dental practices before treatment with an oral appliance and enrolled in the study. Twenty-five of the patients had failed CPAP therapy. After obtaining an informed consent (approved by the BioMed IRB, San Diego, CA, USA) patients completed a two-night pretreatment in-home study with the Apnea Risk Evaluation System (ARES TM) Unicorder TM (Advanced Brain Monitoring, Carlsbad, CA, USA). This pretreatment recording was conducted between the time the dental impressions were taken and when the TAP II Mandibular Advancement Device (MAD) (Airway Management, Dallas, TX, USA) was fabricated and ready for insertion. The TAP is a custom-made oral appliance with separate upper and lower appliances joined by a titration or advancement mechanism on the upper and a transverse bar or socket on the lower. The titration mechanism uses a hook to engage the bar or socket on the lower once each device is placed in the mouth. A jackscrew controls the position of the hook and thus the amount of protrusion. The patient can self-titrate the device using a removable hex key which engages the screw. On the day of insertion of the MAD, patients completed the Beck depression index, Epworth sleepiness scale, and the Flemons’ quality of life questionnaire. At the insertion appointment, both study sites attempted to achieve a starting MAD titration position whereby the patient could just hook the lower tray with the upper tray using active protrusion with both trays in place. In the rare occasion that this level of advancement was not tolerated by the patient, the starting protrusion was reduced. Patients were instructed to begin adjusting the MAD in one-half turn increments as soon as it was tolerable, until a cessation in snoring or the symptoms had resolved. As a result, the titration endpoint was determined by the dentist based on the patient’s self-report. At the follow-up appointment, which was typically scheduled 3 to 4 weeks subsequent to the MAD insertion, the assessment questionnaires were completed again and the ARES Unicorder study was repeated. Twenty-seven of the 30 patients reached their endpoint within 34 days. The other three patients completed their endpoint in 40, 61, and 75 days; the delay in reaching the endpoint was due to patient illness unrelated to this study. In between the time that treatment was initiated and the titration endpoint, each patient maintained a daily journal that recorded the time the appliance was inserted each night, and the time it was removed in the morning. The pre- and posttreatment studies were conducted with the ARES Unicorder. From a single site on the forehead, the wireless recorder measures oxygen saturation, pulse rate, airflow, respiratory effort, snoring levels, head movement, and head position [18] (Fig. 1). Reflectance oximetry is used to obtain the SpO2 and pulse rate signals. Respiratory effort is derived from the measurement of changes in forehead venous pressure acquired using a combination of photoplethysmography and changes in surface pressure of the forehead reflectance oximetry sensor. Airflow is obtained via a cannula and nasal pressure transducer. A calibrated acoustic microphone is used to acquire quantified snoring levels (dB). Actigraphy is used to measure head movement and derive head position. The recorder was designed to be affixed by the patient, and provide alerts during the study if poor quality airflow or SpO2 is detected so the device could be adjusted. Fig. 1Patient wearing a unicorder The description and validation of this device in 284 valid comparisons of the in-laboratory simultaneous PSG and ARES and 187 valid comparisons of the in-laboratory PSG with a separate two nights unattended self-applied ARES Unicorder has been previously published [18]. Using a diagnostic AHI cutoff of >10, the concurrent in-lab comparison yielded a sensitivity of 97.4, a specificity of 85.6, a positive predictive value of 93.6, and a negative predictive value of 93.9. The in-home comparison sensitivity, specificity, positive predictive value, and negative predictive value were 91.5, 85.7, 91.5, and 85.7, respectively. The failure rate was 2%. In an independent validation of 40 subjects (13 healthy controls and 27 patients with symptoms of EDS and/or snoring) using a clinical cut-off of 15, the sensitivity of the ARES compared to concurrent PSG was 100% (25/25 patients) and specificity 67% (10/15 subjects). The sensitivity for two-nights of ARES in-home compared to the PSG was 96%; specificity was 80%. The failure rate was 5% [19]. Automated scoring algorithms were applied off-line to detect sleep disordered breathing. The AHI was computed using a time-in-bed measure based on recording time with acceptable signal quality minus periods when the patient was upright or presumed to be awake based on actigraphy. Apneas, based on a 10-s cessation of airflow detected by the automated algorithms, were included in all apnea–hypopnea indexes (AHI). Hypopnea events required a 50% reduction and recovery in airflow and were further stratified based on the depth of desaturation. The AHI-4% criteria required a minimum 3.5% reduction in SpO2 and at least a 1.0% recovery. Hypopneas included in the AHI-3% and AHI-1% criteria required SpO2 desaturation and resaturation using stepped thresholds. For the AHI-3%, if the SpO2 at the point of maximum saturation before the event was greater than or equal to 95% then a 2.2% reduction and a 2.2% recovery in SpO2 was required. For maximum saturations of between 95–93% the required SpO2 change was a 2.5% reduction and 2.5% recovery; between 93–91% a 3.0% reduction and 2.7% recovery; between 91–88% a 3.5% reduction and 3.0% recovery; and below 88% a 4.0% reduction and 3.5% recovery. For the AHI-1%, if the point of maximum saturation before the event was greater than 93%, then a 1.0% reduction and 1.0% recovery was required; and for events with a starting SpO2 between 93–91%, a 1.2% reduction and 1.2% recovery was required. For an AHI-1% event to be called, the change in flow and desaturation needed to be associated with a behavioral arousal (i.e., an abrupt change in pulse rate, snoring sound or a head movement). After the automated scoring was applied, the full disclosure recordings were visually inspected by a sleep medicine physician to confirm the accuracy of the automated scoring, and to reclassify as central and/or exclude autodetected events if necessary. At the time of the review, the clinician was blinded to the study status (i.e., pre- or posttreatment). The physiological data, including AHI values, percent time with SpO2 below 90, 85, and 80%, and percentage time snoring greater than 30, 40, 50, and 60 dB were then calculated. Responses to the Beck depression index and Flemons’ quality of life questionnaire were tallied in accordance with published methods [20, 21]. Paired t tests were used to identify significant changes in the pre- and posttreatment physiological data and questionnaire responses. To identify anthropomorphic factors that may impact MAD treatment outcomes, patients were stratified into two groups. Group 1 included all patients with a posttreatment AHI-4% ≤ 5 (n = 18). The balance of patients with an AHI-4% > 5 were assigned to group 2 (n = 12). Paired t tests were used to identify significant group differences. To develop and validate the prediction of the posttreatment AHI using pretreatment data, patients were paired and assigned into either the model development or cross validation group based on similarities in the pre- and post-4% AHI and 1% AHI. Correlation analysis was used to identify anthropomorphic variables and measures of obstructive breathing before treatment which might be useful in estimating the posttreatment 4% AHI (post-T 4%). Variables with significant correlations were then used in a linear regression to derive predicted posttreatment values (predict AHI). Results Overall effects of MAD treatment The mean ± SD and minimum pre- and posttreatment valid recording times were 9.6 ± 3.6 and 3.5 h; and 10.3 ± 2.4 and 4.6 h, respectively.Paired t tests revealed significant changes (p < 0.001) in all pre- vs posttreatment sleep disordered breathing measures, including: apnea–hypopnea index with a 4% (AHI-4%) and 1% (AHI-1%) desaturation (Fig. 2). Fig. 2Mean + SE changes in pre- and posttreatment AHI-4% and AHI-1%A decrease in hypoxemia measures pre- and posttreatment were also observed including the percentage of time below 90% SpO2 (T% < 90%) (p < 0.05) and the mean percentage change in SpO2 across all AHI events (% Dip)(p < 0.001) (see Fig. 3). Fig. 3Mean + SE changes in pre- and posttreatment percentage of time the SpO2 was <90%, and the percentage change in SpO2 resulting from sleep disordered breathingSignificant reduction in the average snoring levels at 30 and 40 dB across the valid recording time were also observed (p < 0.001) (Fig. 4). Fig. 4Mean + SE changes in pre- and posttreatment snoring loudness above 30 and 40 dBBased on conventional outcome measures, the MAD therapy was highly efficacious. Twenty-seven of the 30 (90%) patients had a posttreatment AHI-4% < 10 (Fig. 5a). All but one patient (97%) exhibited at least a 50% decrease in AHI-4% or had a posttreatment AHI-4% ≤ 10. Fig. 5Individual changes in pre- and posttreatment sleep disordered breathing using: a AHI-4% desaturation and b AHI-1% desaturationUsing an alternative measure of AHI-1% (used to identify residual sleep disordered breathing/upper airway resistance) 21 of the 30 patients (70%) had a posttreatment AHI-1% ≤ 15 (Fig. 5b). Eighty-seven percent (26 of 30) of the patients exhibited at least a 50% decrease in AHI-1% or had a posttreatment AHI-1% ≤ 15.Paired t tests applied to the pre- and posttreatment scores revealed statistically significant differences for Beck depression index, Flemons’ QOL, Epworth sleepiness score (all at the p ≤ 0.001 level) (Fig. 6). Seventy-six percent (23/30) of the patients showed a decrease in Epworth score; 60% had an Epworth score reduction of three or more. Seventy-three percent (22/30) of the patients reported a posttreatment reduction in depression; 53% had a Beck depression index score reduction of three or more. Eighty-seven percent of the patients reported an increase in the Flemons’ QOL index; 60% showed in improvement score of one or greater. All patients reported some level of subjective improvement on at least one of these subjective measures. Fig. 6Pre- and posttreatment changes in Beck depression index, Epworth sleepiness scale, and Flemons’ quality of life indexThe correlation between the posttreatment percentage time snoring above 30 dB and the AHI-4% was computed because snoring is a principal measure used to assess the MAD titration endpoint. The results in Fig. 7 suggest that snoring is a good predictor of outcome when objectively measured. Fig. 7Regression plot predicting posttreatment AHI-4% based on the posttreatment percentage of time the snoring loudness was above 30 dBDaily self-reported compliance for the MAD revealed that out of a total of 693 nights in bed recorded across all subjects, only 20 nights were reported as not wearing the MAD. The mean usage per night was 7.34 ± 1.3 h. Three of the 30 patients had incomplete paperwork due to either confusion on the part of the patient or problems in collecting the paperwork at the study sites. Although all three patients said that they had worn the MAD for the duration of the study, there was no written record available that allowed the investigators to assess compliance/usage. Factors that affect treatment outcomes Analysis of variance was used to compare demographic and anthropomorphic measures of patients who had a posttreatment AHI-4% ≤ 5 and those who did not respond optimally to treatment (Table 1). Significant differences in the two groups were observed in the body mass index (p < 0.01), weight (p < 0.01), and neck circumference (p < 0.02). Table 1Characteristics of group that was treated optimally and the group that was not treated optionally with a MAD (mean + SE) Treated optimally post-T AHI-4% < 5Did not treat optimally post-T AHI-4% > 5GenderFemales74Males118Age (years)48.3 + 2.550.9 + 3.4BMI (kg/m)**27.6 + 0.833.1 + 2.1Weight (kg)**84.4 + 3.0100.2 + 5.0Neck circumference (cm)*40.0 + 0.743.2 + 1.0*p < 0.05**p < 0.01 Prediction of successful treatment outcome Correlation analysis was applied to the entire data set (n = 30) identify pretreatment variables that had a significant relationship with the posttreatment AHI-4% (see Table 2). Table 2Correlations between posttreatment AHI-4% and pretreatment measuresPearson r (p < 0.01)Pearson r (p < 0.05)Pretreatment 4% AHIr = 0.53Snoring >30 dBr = 0.44Pretreatment 3% AHIr = 0.44Neck circumferencer = 0.45Pretreatment 1% AHIr = 0.54Body mass indexr = 0.45Snoring >40 dBr = 0.52Weightr = 0.45% Time <90% SpO2r = 0.56Various combinations of variables were submitted to linear regression using the model development data set in an effort to predict the posttreatment AHI-4%. Limiting the number of variables to four (due to the small sample size), the variables AHI-4%, AHI-3%, AHI-1%, and percent time snoring >30 dB resulted in an R2 of 0.68. The coefficients derived from the linear regression were applied to the model development and cross-validation data sets.Twenty-seven of the 30 patients had predict-AHI suggesting a successful outcome would be attained using a clinical cutoff post-T 4% AHI ≤ 10 (Fig. 8). Two of the three patients were accurately predicted as unable to achieve a post-T AHI≤10. The Bland–Altman plots comparing the predicted-AHI and posttreatment AHI (Fig. 9) suggest a reasonable fit of the data given the small sample size used for the model development and cross validation data sets. Fig. 8Comparison of pretreatment, posttreatment, and predicted posttreatment AHI-4% for a the model development, and b the cross validation groupsFig. 9Bland–Altman plot of the difference between the predicted posttreatment AHI-4% vs the actual posttreatment AHI for a the model development and b the cross validation groups Discussion Consecutive patients who were referred to the dentist for a MAD were provided the opportunity to enroll in the study. The only exclusion criteria applied were ages less than 18 or older than 70 years. The inclusion of 25 subjects who had previously failed CPAP therapy contributed to the wide OSA severity range: nine of the subjects (30%) were considered to have severe OSA (i.e., AHI-4% ≥ 30) and an additional 30% were considered to have moderate OSA (AHI-4%>15 and <30). Seventy-seven percent of the patients (23/30) had at AHI-1%>25, a clinical cut-off which was considered at least moderate. In this study, a treatment efficacy rate for the MAD was 90% using an AHI-4% clinical cut-off of 10, and 97% when a 50% reduction in AHI was included in determining a successful outcome. Eighty-seven percent of the patients showed improvement using an AHI-1% with a clinical cut-off of 15 or a 50% reduction in AHI. Eighty percent of the patients (24/30) had a posttreatment AHI-1% ≤ 16. A recent evidence-based review summarized results from 87 published studies (over 2,000 patients evaluated) including 15 Sackett Level I and Level II randomized controlled trials [14]. Using a definition of treatment success that required achieving a posttreatment respiratory disturbance index below 10, success was achieved on average across all studies for 52% of patients treated with oral appliances. When the more liberal criteria of a 50% reduction in RDI were applied to assess outcome, 65% of patients were shown to have a positive result with MAD. In an attempt to compare these previous finding to this data set, a clinical cut-off of 10 was applied to the AHI-1%. The differences between the AHI-1% and RDI include a 50% reduction in flow vs 30%; a 1% desaturation vs no desaturation; arousal indicators based on changes in snoring, pulse rate, and head movement vs cortical arousals; and the use of time-in-bed vs total-sleep-time. Using this alternative clinical cut-off, 33% of the patients had a successful outcome. When a 50% reduction in AHI-1% was included, 77% had a successful outcome. Results are more consistent with the previous research. The findings in this study were consistent with previous reports which found self-reported improvements in quality of life and decrease in sleepiness in the majority of patients [14]. No effort was made in this study to obtain objective measures of improvement. To date, only four studies have utilized objective measures of alertness or neurocognitive function to assess treatment outcomes with MAD therapy, with mixed results [22–25]. Further investigations need to be conducted to assess the effectiveness of MAD therapy in ameliorating cognitive impairments in OSA. The strong correlation (r = 0.63) between posttreatment % time snoring >30 dB and the posttreatment AHI-4% suggests that the current clinical practice of relying on residual snoring to asses MAD titration is valid. It is interesting to note that there was no apparent correlation between snoring levels and pretreatment AHI values. Given the discrepancy between the self-reported and objectively measured snoring at the titration endpoint, it appears that the bed-partner or patient’s observation may not provide the most accurate means for titrating a MAD. Several studies have reported increased efficacy for MADs in patients with supine positional OSA [7, 26]; however, the association has not been widely confirmed. This is due in part to the fact that the majority of studies on MAD therapy do not report the number of abnormal respiratory events by sleep position despite the recommendation of the ASDA Practice Parameters for separate calculations of supine and lateral RDIs [27, 28]. Supine positionality did not appear to influence the treatment outcome in our study, possibly, because there were so few MAD treatment failures. This may be a result of the appliance which was used or that the small sample size was not representative of the population treated with oral appliances. A recent evidence-based review reported that median MAD treatment compliance across patients was 77% after 1 year [14]. In this study, compliance was only monitored during the titration period which may have contributed to the favorable finding (i.e., 97%). Other factors that may have influenced this outcome were a relatively small sample size and/or patient compensation being provided. Given that 80% of the study participants were CPAP failures, the influence of prior CPAP use on MAD compliance should be explored. While an objective measure of MAD compliance would eliminate potential bias contributed by self-reported use, practical methods are not currently available. The suggested decrease in MAD efficacy with increasing body mass index (BMI) was confirmed [27, 28]. Weight and neck circumference also appeared to influence the posttreatment apnea/hypopnea index. These variables make sense: the upper airway tends to be narrower in patients with more fatty tissue around the neck and the additional mass combines with gravity to contribute to greater collapsibility when sleeping supine. The successful treatment outcome of patients with severe sleep apnea suggests that a more quantitative approach should be investigated to identify candidates appropriate for a MAD therapy. The results from the predictive model, once fully validated, could provide the guidance needed for sleep medicine physicians to recommend an oral appliance as an initial treatment option for more severe patients. Alternatively, substantial differences between the predicted and actual posttreatment AHI could help dentists determine when a patient has not been fully titrated. Given the small sample size of the model development data set (n = 15), only four variables were included in the regression model even though the correlations presented in Table 2 suggested additional variables would be predictive. It is expected that the error between the predicted and actual posttreatment AHI can be reduced with larger data sets. In addition to expanding the database used for the predictive model, future investigations should be conducted to determine if the accuracy of the predictive model is influenced by the type of MAD. Dentists represent an important access point for identifying and treating patients with undiagnosed OSA. This study was designed to demonstrate two models of collaboration between a dental sleep medicine specialist and a sleep medicine physician. As recommended by the AASM, only patients with mild to moderate OSA were offered MAD therapy as the initial treatment choice in this study. Patients with severe sleep apnea were enrolled only after failure of CPAP therapy. In one model, the dentist referred the patient to the sleep medicine physician and his staff to obtain and review the pre- and posttreatment physiological data. In the second model, the dentist acquired the data and transmitted it to the sleep medicine physician for review. In both models, the physical and history was made available to the physician for interpretation of the data [29]. A follow-up PSG is generally not affordable in cases where it is not covered by an insurance company or health ministry; sleep centers do not tend to offer less expensive level III studies as an alternative. This creates a situation whereby treatment outcomes are simply not assessed, or the dentist conducts the study independently, without the assistance or oversight of the sleep medicine professional. Left on their own, many dentists rely upon oximetry and other level IV devices to monitor treatment outcomes, in part because of their limited experience in interpreting the more sophisticated signals obtained by level III devices. Although several level III devices have automated scoring algorithms, the physiological data should still be reviewed by a trained professional. In two patients, the review of the full disclosure recording allowed recognition of complex sleep apnea (i.e., central sleep apnea was revealed posttreatment after the obstructive breathing was resolved) suggesting the importance of having an experienced professional review the data. This study suggests an alternative approach for assessing MAD treatment outcomes that is based on a collaborative relationship between dentists and sleep medicine physician using a limited channel recording system.

Diabetologia-3-1-1820754

Patients with type 2 diabetes have normal mitochondrial function in skeletal muscle

Aims/hypothesis Insulin resistance and type 2 diabetes are associated with mitochondrial dysfunction. The aim of the present study was to test the hypothesis that oxidative phosphorylation and electron transport capacity are diminished in the skeletal muscle of type 2 diabetic subjects, as a result of a reduction in the mitochondrial content. Introduction It has been reported that insulin resistance is associated with mitochondrial dysfunction in several tissues. While mitochondria are considered central to altered metabolic pathways, leading to pathogenic processes in type 2 diabetes, the mechanisms by which mitochondrial function contributes to the disease remain to be elucidated. Whether it is insulin resistance per se, chronic hyperglycaemia, or accumulation of intracellular lipid and associated alterations in metabolic pathways that affect mitochondrial function, or vice versa, is also unclear. In skeletal muscle, evidence for reduced oxidative capacity in type 2 diabetes has been provided by findings of reduced oxidative enzyme levels [1–4] and mismatches between glycolytic and oxidative enzyme activities [3, 5]. The activities of rotenone-sensitive NADH:O2 oxidoreductase and citrate synthase have been shown to be reduced by 40% in the skeletal muscle of type 2 diabetic patients, and transmission electron microscopy has revealed that these mitochondria have a reduced size and an altered morphology [1, 6]. A decrease in the expression of genes involved in oxidative phosphorylation has also been reported in muscle of type 2 diabetic patients [7, 8] and men on a high-fat diet [9]. It has been stated that impaired muscle mitochondrial function is linked to excess intramuscular lipid accumulation and reduced fatty acid oxidation defined by rates of oxidative phosphorylation and ratios of inorganic phosphate:phosphocreatine determined by magnetic resonance studies [6]. These studies suggest mitochondrial dysfunction in type 2 diabetes; however, to date, no direct measurements of mitochondrial O2 flux capacity in intact cells in human type 2 diabetes have been reported. We hypothesised that oxidative phosphorylation and electron transport capacity are diminished in the skeletal muscle of type 2 diabetic subjects, and that these changes are attributable to a reduction in muscle mitochondrial content. In the present study we tested these hypotheses by using high-resolution respirometry to quantify oxidative phosphorylation and electron transport capacity in permeabilised muscle fibres from biopsy samples of the quadriceps in healthy subjects and patients with type 2 diabetes. Subjects and methods Subjects Informed consent was obtained from all subjects. The study was conducted in accordance with the principles of the Declaration of Helsinki, and was approved by the local ethics committee for Frederiksberg and Copenhagen County. Mitochondrial respiration was measured in permeabilised skeletal muscle fibres obtained from needle biopsies of the vastus lateralis in men with (n = 11) or without (control; n = 8) type 2 diabetes. The characteristics of the subjects are provided in Table 1 and Fig. 1. All subjects were in good health but classified as living a typical Westernised sedentary lifestyle, participating only in routine activities of daily living (walking, gardening, etc.) and not engaged in regular structured or individualised aerobic or strength training programmes or athletics. None of the control subjects had a family history of diabetes and none was receiving treatment for a disease. The diabetic patients were treated for their diabetes with diet or oral glucose-lowering medicine. All medications were withheld 24 h prior to the experiment. The patients with type 2 diabetes had no clinical signs of long-term diabetic complications and were representative of patients treated in the primary care sector. Table 1Characteristics of the subjects Type 2 diabetic subjects (n = 11)Control subjects (n = 8)Age (years)62 ± 258 ± 1Height (cm)177 ± 3179 ± 1BMI (kg/m2)32 ± 2*28 ± 1Time since diagnosis (years)5 ± 2–Fasting insulin (pmol/l)61 ± 9*34 ± 6Fasting glucose (mmol/l)9.0 ± 0.5*5.4 ± 0.1Complex I activity (nmol min−1 mg protein−1)50.8 ± 6.058.3 ± 4.7 Citrate synthase activity (pmol mg−1 s−1)1.6 ± 0.12.0 ± 0.2mtDNA (copies/μg tissue) ×103119 ± 7*147 ± 12mtDNA/genomic DNA2,773 ± 2523,030 ± 185Data are means±SEM. *p < 0.05 vs control subjectsFig. 1Glucose (a) and insulin (b) concentrations in venous plasma before (t = 0 min) and during an OGTT. The patients with type 2 diabetes had higher fasting glucose levels and were severely insulin resistant compared with healthy control subjects (*p < 0.05). Black and white symbols represent healthy control subjects and patients with type 2 diabetes, respectivelySubjects were fasted overnight prior to the experiment. A catheter was inserted into an antecubital vein for blood sampling. After local anaesthesia of the skin and the subcutis, a muscle biopsy was taken (Tru-Core; PBN-Medicals, Stenløse, Denmark) and then a 120-min OGTT (75 g glucose dissolved in 300 ml of water) was performed. At t = 30 min, a second muscle biopsy was taken.A portion of the obtained muscle tissue was frozen immediately in liquid nitrogen and stored at −80°C for later analysis (see below), and a smaller piece (2–6 mg) was placed onto a Petri dish on ice with 1 ml of relaxing solution containing Ca2+/EGTA buffer (10 mmol/l), free calcium (0.1 μmol/l), imidazole (20 mmol/l), K+/4-morpholinoethanesulfonic acid (MES) (50 mmol/l), dithiothreitol (DTT; 0.5 mmol/l), MgCl2 (6.56 mmol/l), ATP (5.77 mmol/l), phosphocreatine (15 mmol/l), pH 7.1, and individual fibre bundles were separated with two pairs of sharp forceps, achieving a high degree of fibre separation. The fibre bundles were permeabilised for 30 min in 3 ml of ice-cold relaxing solution containing saponin (50 μg/ml) [10]. After rinsing in respiration medium (MiR05; Oroboros, Innsbruck, Austria) containing sucrose (110 mmol/l), potassium lactobionate (60 mmol/l), EGTA (0.5 mmol/l), MgCl2.6H2O (3 mmol/l), taurine (20 mmol/l), KH2PO4 (10 mmol/l), HEPES (20 mmol/l), sucrose (110 mmol/l), BSA (1 g/l), pH 7.1, the muscle bundles were blotted and measured for wet weight in a balance controlled for constant relative humidity, so that all biopsy samples were hydrated to the same degree. The muscle bundles were then immediately transferred into a respirometer (Oxygraph-2k; Oroboros) containing air-saturated respiration medium at 37°C.The Oxygraph-2k is a two-chamber titration-injection respirometer with a limit of oxygen flux detection of 1 pmol s−1 ml−1. The instrumentation allows for O2 flux measurements with only 0.04 mg of mitochondrial protein or 1.5 mg of muscle fibres (wet weight). Standardised instrumental and chemical calibrations were performed to correct for back-diffusion of O2 into the chamber from the various components, leak from the exterior, O2 consumption by the chemical medium, and sensor O2 consumption [11]. O2 flux was resolved by software capable of converting nonlinear changes in the negative time derivative of the oxygen concentration signal. Analysis of muscle tissue Citrate synthase activity and complex I activity were measured spectrophotometrically at 37°C. Citrate synthase activity was determined as described previously [12], and complex I activity was assessed by measuring the oxidation of NADH (300 μmol/l) using ubiquinone 1 (100 μmol/l) as the acceptor. The complex I rotenone-sensitive activity was measured by the addition of rotenone (1 μmol/l). The protein content, needed to calculate the specific activity, was measured using a commercially available assay (BCA, Sigma Chemicals, St Louis, MO, USA). For measurement of mitochondrial DNA (mtDNA) content, DNA was isolated from muscle biopsy samples (∼10 mg) by proteinase K digestion at 55°C for 3 days. The 100-μl digestion mix contained 50 mU proteinase K (PCR grade, Roche, Basel, Switzerland), 20 mmol/l Tris-HCl (pH 8.4) and 50 mmol/l KCl. After incubation at 80°C for 45 min, the remains were spun down and the supernatant fraction diluted ×200 in triethanolamine titanate (TE) plus 1 ng/μl salmon sperm DNA (Sigma). 5 μl of this dilution was amplified in a 25 μl PCR reaction containing 1×Quantitect SYBR Green Master Mix (Qiagen, Hilden, Germany) and 100 nmol/l of each primer. The amplification was monitored real-time using the MX3000P Real-time PCR machine (Stratagene, La Jolla, CA, USA). The primers were designed to target genomic DNA (Forward: AGG TGC TGT CAG GAA GCA AGG A, Reverse: TAG GGG GAG GAG GGA ACA AGG A) or mtDNA (Forward: CCC CTG CCA TAA CCC AAT ACC A, Reverse: CCA GCA GCT AGG ACT GGG AGA GA). The threshold cycle (Ct) values were related to a standard curve made with the cloned PCR products. Respirometry protocol All measurements of respiration were made in duplicate, simultaneously. Resting, routine respiration (state 2, absence of adenylates) was assessed by the addition of malate (1.5 mmol/l) and glutamate (19 mmol/l) as the complex I substrate supply, and then state 3 respiration was assessed by the addition of ADP (4.8 mmol/l). The addition of succinate (9.5 mmol/l) provided state 3 respiration with parallel electron input to complexes I and II. The integrity of the outer mitochondrial membrane was established by the addition of cytochrome c (19 μmol/l); no stimulation of respiration was observed. We examined ADP control of coupled respiration and uncoupling control through addition of the protonophore carbonylcyanide-4-(trifluoromethoxy)-phenylhydrazone (FCCP) (0.7 μmol/l). The addition of rotenone (0.1 μmol/l) resulted in inhibition of complex I for examination of O2 flux with complex II substrate alone, while antimycin A (12 μmol/l) was added to inhibit complex III to observe non-mitochondrial respiration with small contributions from electron leak in the uncoupled state. The concentrations of substrates and inhibitors used were based on prior experiments conducted for optimisation of the titration protocols. Data analysis All values are given as means±SEM for all experiments, run in duplicate or triplicate. For all statistical evaluations, a p value of less than 0.05 was considered significant. Statistical analysis of differences in oxygen flux between healthy control subjects and patients with type 2 diabetes was carried out with a two-way ANOVA for repeated measures. In the case of a significant main effect and interaction between the variables, the Holm-Sidak method was used for post hoc analysis. All other comparisons between the two groups were performed using the unpaired Student’s t test. SigmaStat version 3.11 (Systat software, Richmond, CA, USA) was used in all analyses. Results The sequential addition of substrates to the muscle tissue, obtained from both groups before an OGTT, always (p < 0.05) resulted in a stepwise increase in state 3 O2 flux (Fig. 2). Notably, the addition of succinate (stimulating parallel electron input from complexes I+II) resulted in a marked increase in O2 flux in both groups. The O2 flux per muscle mass was significantly (p < 0.05) lower in the patients compared with the healthy subjects during complex I and complex I+II respiration (Fig. 2). Further increases in flux capacity and preserved significant differences between the groups were observed with uncoupling by FCCP (109 ± 8 vs 86 ± 4 pmol mg−1 s−1 in control and diabetic subjects, respectively; p < 0.05). Subsequent inhibition of complex I and III with rotenone and antimycin A blunted the O2 flux (Fig. 2). The addition of cytochrome c did not result in significant increases in O2 flux (data not shown). Fig. 2O2 flux in permeabilised skeletal muscle fibres from patients with type 2 diabetes and healthy control subjects. Data are shown as O2 flux per mg of tissue (a) and further normalised to the number of copies of mtDNA per μg of tissue ×10,000 (b). When data are expressed relative to mtDNA, any difference between the groups disappears. Data are means±SEM (*p < 0.05). Black and white bars represent healthy control subjects and patients with type 2 diabetes, respectively The number of copies of mtDNA and citrate synthase activity (Table 1) indicated a lower mitochondrial density in the patients with type 2 diabetes. The O2 flux data were therefore recalculated relative to mtDNA content (Fig. 2b) and citrate synthase activity (data not shown). All differences in O2 flux between patients with type 2 diabetes and healthy control subjects disappeared following either normalisation procedure (Fig. 2b). The increase in O2 consumption when ADP was added (complex I respiratory control ratio [state 3:state 2 respiration]) was not different between healthy subjects and patients with type 2 diabetes (Fig. 3). Fig. 3a Respiratory control ratio for complex I (NADH supply from substrates glutamate + malate) measured as the ratio of O2 flux with (state 3) and without (state 2) ADP. b Electron transport capacity measured as O2 flux after FCCP-induced uncoupling relative to coupled O2 flux at state 3 with malate + glutamate + ADP + succinate (parallel electron input into both complex I and II). No significant difference between the groups was noted. Data are means±SEM Mitochondrial respiration may be influenced by the prevailing level of glucose and/or insulin. However, none of the O2 flux rates measured with different substrates and inhibitors displayed significant correlations with these parameters, or the changes seen during the OGTT (data not shown). The glucose and insulin concentrations were markedly increased in both groups at 30 min into the OGTT (Fig. 1). However, the measured O2 flux in the muscle tissue obtained at this time point deviated to a lesser extent (mean +2.4 ± 3.5%) from the O2 flux measured in the biopsies obtained during fasting (data not shown). The activity of complex I was affected by rotenone to a similar extent in the two groups (Table 1). Furthermore, the biochemically measured activity of complex I was significantly (p < 0.02) correlated with O2 flux measured by respirometry during rotenone inhibition (r2 = 0.37). Discussion The primary novel findings in this study are that: (1) ADP-stimulated state 3 mitochondrial O2 flux capacity with electron flux through either complex I or II, or with parallel electron input through both complexes I and II, is substantially reduced in type 2 diabetic patients when expressed per unit mass of skeletal muscle; and (2) when O2 flux is normalised for mitochondrial DNA content or citrate synthase activity, levels of both oxidative phosphorylation and electron transport capacity are similar to those observed in age-matched healthy control subjects. These results provide direct experimental evidence for normal function of muscle mitochondria in type 2 diabetes and do not support other investigations reporting mitochondrial dysfunction in diabetes. The reduced mitochondrial capacity per unit muscle mass observed in this study is consistent with the concept of reduced mitochondrial content and volume, oxidative enzyme levels, mtDNA and decreased levels of co-regulators of mitochondrial biogenesis-such as peroxisome proliferator-activated receptor-γ coactivator-1 (PGC-1), nuclear respiratory factor (NRF-1 and NRF-2) and mitochondrial transcription factor A (mtTFA)-in insulin-resistant states, as found in some [7, 8] but not all studies [13]. Therefore, it could be argued that specific cellular signals that alter levels of mitochondria, and thus reduce electron transport (and oxidative phosphorylation capacity) per unit muscle mass, contribute to a variety of aberrant metabolic pathways, including intracellular fat accumulation, insulin resistance and glucose intolerance. The various substrate and inhibitor titrations employed in this study permitted an examination of various steps of oxidative phosphorylation and electron transport under resting respiration and maximally ADP-stimulated O2 flux by ADP (state 3). Both resting (state 2) and ADP-stimulated (state 3) coupled respiration were substantially reduced (18–28%) in diabetic subjects, as was uncoupled O2 flux with parallel electron supply from both NADH (complex I) and FADH2 (complex II). These responses indicate a blunting of oxidative phosphorylation linked to ATP synthase and a decreased maximal electron flux capacity in the uncoupled state induced by addition of FCCP, respectively. The key finding, however, is that these values are comparable with those in healthy control subjects when they are normalised for mtDNA content or citrate synthase activity, both of which are used as indices of mitochondrial density per unit muscle mass. The 30% reduction in state 3 O2 flux capacity per mg of muscle in diabetic subjects with parallel electron input (glutamate + malate + succinate) suggests an attenuation of cellular VO2max. While the prevailing view is that O2 delivery is a factor that influences whole-body maximal O2 consumption, it remains to be determined if, and to what extent, the observed decrement in muscle state 3 mitochondrial O2 flux capacity in diabetic patients contributes to the lower systemic VO2max and substrate utilisation. The patients with type 2 diabetes were in a chronic hyperglycaemic state and were clearly insulin resistant, but even so, mitochondrial function/mtDNA was not impaired. Furthermore, mitochondrial function was also measured in the biopsy samples obtained 30 min into the OGTT, at a time when both glucose and insulin were markedly increased (Fig. 1). This acute metabolic perturbation did not have any significant effect on mitochondrial respiration in any of the subjects (data not shown). It has been proposed that chronic hyperglycaemia associated with insulin resistance results in the alteration of several metabolic pathways [14]. A central hypothesis involving the mitochondria focuses on an effect of hyperglycaemia providing increased reducing equivalents to the electron transport chain, resulting in a higher membrane potential, with consequent flow of electrons between coenzyme Q and complex III, forming superoxides in the mitochondrial matrix. Excess superoxide production can induce damage to mitochondrial structures, including several electron transport complexes, the mitochondrial lipid bilayer and mtDNA. A role for excess superoxide production is supported by findings of reduced glutathione and metallothionein antioxidant defence systems in type 2 diabetic subjects [15]. The results of this study do not preclude these concepts, but rather suggest the possibility that many pathogenic pathways associated with mitochondrial function in cellular energetics may result from conditions leading to a reduced number of mitochondria, which in turn could place the existing mitochondria under stress with subsequent production of reactive O2 species, impaired metabolism of intracellular lipids, and glucose uptake. It is well known that exercise training increases mitochondrial content in skeletal muscle [16–19], and recent work has provided evidence for increased levels of transcriptional regulators of mitochondrial biogenesis in response to exercise [20, 21]. In addition, physical training can play a significant role in the prevention of insulin resistance and type 2 diabetes [22–25]. Physical training also substantially improves skeletal muscle insulin sensitivity in patients with overt type 2 diabetes [26], and many cellular adaptations responsible for the effect of training on insulin sensitivity in skeletal muscle have been described [27–31]. The cellular and mitochondrial changes in response to exercise training occur in parallel. In obese insulin-resistant subjects, exercise training has been shown to increase the percentage of skeletal muscle fibre volume occupied by mitochondria [32]. Thus, the present data support the hypothesis that type 2 diabetes is, to a large extent, a lifestyle disease, with insufficient exercise-induced gene expression and a surplus of energy intake contributing to its pathogenesis. Accordingly, studies on the effects of ‘de-training’ have provided evidence of reduced mitochondrial content. After cessation of an endurance training program, citrate synthase and succinate dehydrogenase activities in human muscle have been shown to decline with a half-time of only 12–14 days [33]. Cytochrome c protein concentration in rat muscles declines with a half-life of only 7–8 days [34], and the activity in human muscles has a half-time of similar magnitude. These decreases in the activities of cytochrome c and succinate dehydrogenase do not exactly follow the detraining decline in VO2max [35]. Taken together, these studies support the notion that lack of physical activity lowers mitochondrial concentration. The independent influences of exercise training and detraining, hyperglycaemia, intramuscular lipid accumulation and obesity on mitochondrial function remain to be elucidated. In this study, the ability of the mitochondria to respond with increased O2 consumption following the addition of ADP, represented by the respiratory control ratio, was preserved in the diabetic muscle (Fig. 3). Thus, this index of mitochondrial phosphorylation capacity and coupling of electron transport to phosphorylation indicates that the respiratory chain of the mitochondria in type 2 diabetic subjects functions in a similar manner to that in the mitochondria in control subjects. This view is supported by similar data reported 40 years ago, albeit mostly from patients with ‘juvenile diabetes’ [36]. Similarly, an increase (1.3–1.4 fold) in O2 flux was seen in response to uncoupling by FCCP (Fig. 3). The fact that the increase was similar in the two groups also testifies that electron transport capacity is not impaired in type 2 diabetes, and that the phosphorylation system (adenine nucleotide transporter, phosphate transporter and ATP synthase) exerts control over electron transport in patients and control subjects to the same degree. In conclusion, the results of the present study provide the first direct evidence of normal mitochondrial function in the skeletal muscle of type 2 diabetic subjects. An apparent impairment of oxidative phosphorylation and electron transport capacity is fully accounted for by a diminished mitochondrial content in the diabetic muscle.

FEMS_Microbiol_Ecol-1-_-2121141

Multivariate analyses in microbial ecology

Environmental microbiology is undergoing a dramatic revolution due to the increasing accumulation of biological information and contextual environmental parameters. This will not only enable a better identification of diversity patterns, but will also shed more light on the associated environmental conditions, spatial locations, and seasonal fluctuations, which could explain such patterns. Complex ecological questions may now be addressed using multivariate statistical analyses, which represent a vast potential of techniques that are still underexploited. Here, well-established exploratory and hypothesis-driven approaches are reviewed, so as to foster their addition to the microbial ecologist toolbox. Because such tools aim at reducing data set complexity, at identifying major patterns and putative causal factors, they will certainly find many applications in microbial ecology. Introduction Microbial ecology is undergoing a profound change because structure–function relationships between communities and their environment are starting to be investigated at the field, regional, and even continental scales (e.g. Hughes Martiny et al., 2006; Ramette & Tiedje, 2007a, b). Because DNA sequences are being accumulated at an unprecedented rate due to high-throughput technologies such as pyrosequencing (Edwards et al., 2006a, b), single-cell genome sequencing (Zhang et al., 2006), or metagenomics (Venter et al., 2004; Field et al., 2006; Gill et al., 2006), future challenges will very likely consist of interpreting the observed diversity patterns as a function of contextual environmental parameters. This would help answer fundamental questions in microbial ecology such as whether microbial diversity responds qualitatively and quantitatively to the same factors as macroorganism diversity (Horner-Devine et al., 2004; van der Gast et al., 2005; Green & Bohannan, 2006; Hughes Martiny et al., 2006). Most obstacles encountered by microbial ecologists when they try to summarize and further explore large data sets concern the choice of the adequate numerical tools to further evaluate the data statistically and visually. Such tools, which have been developed by community ecologists to work on distribution and diversity patterns of plants and animals, could be readily applied in microbial ecology. Although multivariate analyses of community diversity patterns are well described in the literature, microbial ecologists have used multivariate analyses either rarely or mostly for exploratory purposes. A brief survey of the literature confirms this trend (Table 1; Fig. 1). Table 1 indicates that bacterial studies rank third after plant and fish studies for their use of multivariate analyses. Complex data sets are mostly explored via principal component analysis, or cluster analysis, and hypothesis-driven techniques such as redundancy analysis, canonical correspondence analysis (CCA), or Mantel tests are more rarely used (Fig. 1). Axis 1 (horizontal) clearly differentiates microscopic (bacteria, microorganisms, fungi) from macroscopic (fish, bird, plant, insect) life, and this may be related to the use of more exploratory methods (e.g. cluster analysis, PCA) in the first group. It is important to state that the figures presented in Table 1 and Fig. 1 have to be taken with caution because many articles do not include a description of statistical approaches in their titles or abstracts, and so the table is certainly biased and incomplete. However, the point of the table was both to identify some general trends in the literature and to give one example of the usefulness of multivariate analysis to analyze a data table. Table 1 Usage (%) of multivariate methods in different fields Exploratory analysis Hypothesis-driven analysis Keywords† Cluster PCA MDS PCoA CCA RDA manova Mantel anosim CVA Total number‡ Bacter* 48.5 38 4.5 0.4 3.2 1.8 1.3 0.4 0.9 1.1 1141 Microb* 45.8 40.2 3.9 1.1 2.2 2.2 1.1 1.7 0.6 1.1  179 Plant* 40.3 28.5 4.6 1.7 15.5 3.7 1.9 2.3 0.6 0.9 3335 Fung* 54 27.2 2.8 1.1 8.5 2.8 0.9 1.1 0.2 1.4  563 Fish* 30.1 33.7 9.8 0.3 13.5 2.7 3.6 2.9 2.3 1.2 1464 Bird* 41 20.5 5.4 0.7 21.2 3.5 2.1 4.2 0.5 0.9  429 Insect* 54.3 13.7 6.1 0.8 11.5 4.4 3.5 3 1.1 1.7  637 A literature search was performed with the Thomson ISI research tool with the following parameters (Doc type, all document types; language, all languages; databases, SCI-EXPANDED, SSCI, A&HCI; Timespan, 1900–2006) on December 13, 2006 in the titles and abstracts of the articles only. † Asterisks were placed at the end of each keyword to accommodate for variations. Each keyword was additionally combined with the following technical designations: cluster, cluster analysis; PCA, principal component analysis; MDS, multidimensional scaling; PcoA, principal coordinate analysis; CCA, canonical correspondence analysis; RDA, redundancy analysis; Mantel, Mantel test, or CVA, canonical variate analysis. ‡ Total number refers to the total number of publications identified by each keyword and all its combinations. The ordination based on correspondence analysis of the raw number is depicted in Fig. 1. Fig. 1 Correspondence analysis of method usage in various scientific fields. In this symmetrical scaling of CA scores, the first two axes explained 47.3% and 35.8% of the total inertia of Table 1, respectively. The gray areas were drawn to facilitate the interpretation. Complete row names (scientific fields; full circles) and column names (methods; white triangles) are given in Table 1. Methods (triangles) located close to each other correspond to methods often occurring together in studies. The distance between a scientific field point and a method point approximates the probability of method usage in the field. This review aims at presenting some common multivariate techniques in order to foster their integration into the microbial ecologist's toolbox. Indeed, ‘it is no longer possible to gain a full understanding of Ecology and Systematics without some knowledge of multivariate analysis. Or, contrariwise, misunderstanding of the methods can inhibit advancement of the science’ (James & McCulloch, 1990). Such a review is ambitious because it tries to provide a few guidelines for a very vast discipline that is still under development. For this reason, it cannot be exhaustive and does not pretend to offer in-depth coverage of all selected topics. The review is largely inspired by descriptions, comments, and suggestions originating from multiple, highly recommended sources (ter Braak & Prentice, 1988; James & McCulloch, 1990; Legendre & Legendre, 1998; Leps & Smilauer, 1999; ter Braak & Smilauer, 2002; Palmer, 2006), where detailed information about each technique can be obtained. In the first part, data type and preparation are reviewed as a necessary basis for subsequent multivariate analyses. Second, common multivariate methods (i.e. cluster analysis, principal component analysis, correspondence analysis, multidimensional scaling) and a few statistical methods to test for significant differences between groups or clusters are described, focusing on the methods' main objectives, applications, and limitations. Beyond the mere identification of diversity patterns, microbial ecologists may wish to correlate or explain those patterns using measured environmental parameters, and this approach is addressed in the third part. Special emphasis is placed on a few methods that have proven useful in ecological studies, namely redundancy analysis, CCA, linear discriminant analysis, as well as variation partitioning. The final part provides practical considerations to help researchers avoid pitfalls and choose the most appropriate methods. Data types and data preparation Data sets The initial multivariate data set may consist of a table of objects (e.g. samples, sites, time periods) in rows and measured variables for those objects in columns. This table structure is the standard used in the present review. When the latter variables are biological taxa, the columns will simply be designated as ‘species’ thereafter. It is critical to clearly identify what corresponds to objects and variables in the data set. Indeed, objects in one study may be species or operational taxonomic units (OTU) for which catabolic profiles, gene presence or polymorphism, etc. are measured. In another study where samples from different sites are compared based on, for instance, community fingerprinting techniques, objects can now be samples and species variables. This distinction is important because procedures that analyze relationships among objects or among variables are different. Objects are defined a priori by the sampling strategy before making observations and variable measurements. Besides, most multivariate analyses assume independence between objects (or samples), i.e. observations made on an object are not a priori dependent on those made on another object. Variables, however, can be found to be intercorrelated to various degrees, but this is not necessarily known in advance. Initial data sets can also consist of distance matrices where pairwise dissimilarities between objects are calculated. The original table of raw data is not always available, e.g. for DNA–DNA hybridization values, phylogenetic distances, and thus specific multivariate techniques have to be considered to deal with data matrices. Data transformations In multivariate data tables, measured variables can be binary, quantitative, qualitative, rank-ordered, classes, frequencies, or even a mixture of those types. If variables do not have a uniform scale (e.g. environmental parameters measured in different units or scales) or an adequate format, variables have to be transformed before performing further analyses. Each qualitative variable has to be recoded as a set of numerical variables that replace it in the numerical calculations. One way to do so is to create a series of ‘dummy’ variables that correspond to all the states of the qualitative variable. For instance, if the variable ‘season’ has to be recoded, four associated variables will be constructed, and for each object the value 1 will be given to the corresponding season when it occurs, and 0 for the three other seasons when it is absent. Many statistical packages automatically perform this recoding. Standardization provides dimensionless variables and removes the undue influence of magnitude differences between scales or units. A common procedure is to apply the z-score transformation to the values of each variable. For each variable, it consists of (1) computing the difference between the original value and the mean of the variable (i.e. centering) and of (2) dividing this difference by the SD of the variable. Normalizing transformations aim at correcting the distribution shapes of certain variables, which depart from normality. One thus tries to obtain a homogeneous variances for variables, conditions under which multivariate procedures often perform better. Different mathematical transformations can be used to normalize the x values of a variable: for instance, the arcsin (√x) transformation can be applied to percentages or proportions, log(x+c) to variables departing strongly from a normal distribution, and √(x+c) to less problematic cases, with c being a constant that is added to avoid mathematically undefined computations. The c constant is generally chosen so that the smallest nonzero value is obtained by computing x+c in the former functions. The constant should also be of the same order of magnitude as the variable (Legendre & Legendre, 1998). To make community composition (either presence–absence or abundance) data containing many zeros suitable for analysis by linear methods such as principal component analysis (PCA) or canonical redundancy analysis (RDA), the Hellinger transformation [Eq. (1)] is one of five transformations that give good results (Legendre & Gallagher, 2001). The chord transformation is a useful transformation that also gives less weight to rare species in the species table [Eq. (2)]. The transformations are given by where yij is the original species value for site i and species j, yi+ represents the sum of all species values for site i (i.e. sum per row), p is the number of species in the table (number of columns), and represents the resulting, transformed species value (Legendre & Gallagher, 2001). These transformations are particularly recommended when rare species are not truly rare, i.e. when they mostly occur because the sampling was performed blindly, as generally done in soil or marine microbial ecology. Further data transformations can be found in Sokal & Rohlf (1995) and Legendre & Legendre (1998). The way to deal with missing data is a discipline on its own (Legendre & Legendre, 1998). Briefly, one can either delete rows or columns containing the missing value(s), or try to replace the missing values by mathematical estimates inferred from values obtained from other objects in the data set. In the latter case, it is still difficult to provide ecologically meaningful explanations for these estimates. In any case, the specific handling of missing data should be reported by the investigator. When dealing with matrices, it is possible to change a similarity matrix (S) into a dissimilarity matrix (D) by applying the following transformations: D=1−S, D=√(1−S), or D=√(1−S2). To normalize any D matrix to the interval [0–1], one can compute D/Dmax, or (D−Dmin)/(Dmax−Dmin), where Dmax and Dmin represent the highest and lowest values of D, respectively (Legendre & Legendre, 1998). Exploratory analyses Visualization and exploration of complex data sets The basic aim of ordination and cluster analysis is to represent the (dis)similarity between objects (e.g. samples, sites) based on values of multiple variables (columns) associated with them, so that similar objects are depicted near from each other and dissimilar objects are found further apart from each other. Exploratory multivariate analyses are thus useful to reveal patterns in large data sets, but they do not directly explain why those patterns exist. This latter point is addressed in the third part of the review. Cluster analysis and association coefficients Cluster analysis encompasses several multivariate techniques that are used to group objects into categories based on their dissimilarities. The aim is both to minimize within-group variation and maximize between-group variation in order to reveal well-defined categories of objects, and therefore reduce the dimensionality of the data set to a few groups of rows (James & McCulloch, 1990; Legendre & Legendre, 1998). This approach is thus generally recommended when distinct discontinuities instead of continuous differences (i.e. gradients) are expected between samples (objects) because cluster analysis mostly aims at representing partitions in a data set (Legendre & Legendre, 1998). Because distance matrices that are based on differences in DNA or amino acid sequences are commonly used to describe microbial diversity, cluster analysis has become very popular in microbial ecology (Table 1; Fig. 1). This is not surprising because the grouping of organisms based on their phenotypic or genotypic similarities in order to infer their taxonomic positioning is generally and historically based on cluster analysis (or at least based on a tree-like representation) and, as such, is central to biology and evolution (Avise, 2006). Typical microbial ecology questions that are addressed by cluster analysis are whether the clustering patterns of molecular sequences reflect sample origin or sampling time in order to reveal specific biogeographical or temporal patterns, respectively (Whitaker et al., 2003; Acinas et al., 2004). Those factors are generally hypothesized to be of a discontinuous nature, but the rationale of generally representing molecular differences as discontinuous clusters in microbial ecology and microbial genomic studies has only started to be questioned (Konstantinidis et al., 2006). Another common application consists of sorting out clones from environmental samples based on specific criteria (e.g. genetic or phenotypic markers) because clones or variants are expected to form tight clusters around their parental strains and to be more distinct from other lineages (Acinas et al., 2004). In microarray data analysis, cluster analysis has helped identify common expression patterns of groups of genes, which may shed light on functionally related genes or pathways (Eisen et al., 1998). Cluster analysis of a data table proceeds in two steps. First, a relevant association coefficient has to be chosen to measure the association (similarity or dissimilarity) among objects or among variables. Second, the calculated association matrix is represented as a horizontal tree (hierarchical clustering) or as distinct groups of objects (k-means clustering), based on specific rules to aggregate objects. For ecologists, the power of cluster analysis derives from the existence of different types of (dis)similarity coefficients. The choice of appropriate and ecologically meaningful association coefficients is particularly important because it directly affects the values that are subsequently used for the categorization of objects. The analysis of similarities among objects (rows) is designated as Q mode analysis, whereas when relationships among variables (columns) are the focus of the study, this is referred to as R mode analysis (Legendre & Legendre, 1998). Noticeably, the two modes of analysis do not generally use the same association coefficients. Although it is not possible to give a full review of all association coefficients here, it is useful to known that, for comparing objects (rows) based on their column attributes in Q mode analysis, coefficients may be chosen as a function of data type (quantitative, qualitative, ordinal, or mixed data, normalized data, presence-absence), importance given to rare species, weight given to each object, and calculation of associated probability levels. For comparing objects in a sample-by-environment table (e.g. water, soil chemistry), selection of appropriate coefficients generally depends on data type and unit homogeneity of the measured variables. In R mode analyses, in addition to the previously cited criteria, the choice of a coefficient may also depend on how the variables are related to each other (e.g. linearly, monotonically, qualitatively, ordered), and on how species absence is handled in the calculations. In most ecological studies, the absence of a species at two sites being compared is not considered as a measure of similarity between those sites. Indeed, a simultaneous species absence at two sites may be due to different reasons, e.g. the sites offer different physical–chemical conditions and the species cannot exist under both conditions, and so there is no straightforward conclusion about site similarity that can be drawn in this case. Asymmetric coefficients are coefficients that do not take into account cases of double absences of species (‘double zeros’) in the calculation of pairwise similarities among sites. Moreover, in microbial ecology where environmental communities are generally far from being exhaustively sampled, a double absence of an OTU has to be regarded more as a lack of information rather than a sign of common structure among samples, and asymmetric coefficients such as Jaccard (1901) or Sørensen (1948) should be preferred. More details about the calculation of association coefficients and their appropriateness can be found, for instance, in Chapter 7 of (Legendre & Legendre, 1998). When an association matrix is calculated, the relationships between objects or variables can be represented following specific aggregation rules. Three general approaches are commonly used: hierarchical clustering, k-means partitioning, and two-way joining. In hierarchical clustering, a linkage rule to form clusters and the numbers of clusters that best suit the data have to be determined a priori. Clusters, which are nested rather than mutually exclusive here, are either formed by progressively agglomerating objects from high to low similarity cutoff values (forward clustering), or using the converse strategy, i.e. grouping all cases together and progressing from low to high cutoff values in order to merge objects and clusters (backward clustering). These two strategies do not necessarily yield the same clusters. The merging of clusters is visualized using a tree format (generally horizontal) and is successful when well-defined clusters are identified in the data set (Sneath & Sokal, 1973). Common linkage rules are, e.g. nearest neighbor (the distance between two clusters is the distance between their closest neighboring points), furthest neighbor (the distance between two clusters is the distance between their two furthest objects), and the widely used unweighted pair-group method using averages (UPGMA; Sneath & Sokal, 1973), where the distance between two clusters is the average distance between all intercluster pairs. When within-cluster homogeneity is desired, Ward's method, which merges clusters only if they increase the within-cluster variation the least, is recommended (Legendre & Legendre, 1998). Finally, equal weight can also be given to clusters that are expected to be of different sizes using the weighted arithmetic average clustering (WPGMA), which consists of giving less weight to the original similarities of the largest groups (Legendre & Legendre, 1998). In k-means clustering, objects are assigned to k clusters (k being defined in advance), based on their nearest Euclidean distance to the mean of the clusters. The mean of the cluster is iteratively recalculated until no more assignments are made and cluster means fall below a predefined cut-off value or until the iteration limit is reached. Different means for each cluster are ideally obtained for each dimension used in the analysis, as indicated by high F-values from the respective analyses of variance. Unlike hierarchical clustering, k-means clustering does not require prior computation of dissimilarity matrix among objects and is therefore more adapted to large data sets (e.g. few thousand objects) where computing power is an issue. However, the method is quite sensitive to outliers, which are usually removed before performing the analyses (Legendre & Legendre, 1998). Two-step cluster analysis may be useful to group objects into clusters when one or more of the variables are categorical (not interval or dichotomous). Objects are first grouped based on the categories, which are themselves hierarchically clustered as single cases. Because neither a proximity table nor iterative steps are required, the method is particularly suited for the analysis of very large data sets (Eisen et al., 1998). Principal component analysis (PCA) PCA has been applied to numerous phenotypic and genotypic (e.g. fingerprinting patterns) data sets, and it is one of the most popular exploratory analyses (Table 1), perhaps because the technique is generally the first multivariate approach to be explained in most data analysis manuals. However, this choice may not always be justified in ecology and recommendations for appropriate applications are provided at the end of this section and in the ‘Practical considerations’ part of the present review. Examples of use in microbial ecology concern the identification of patterns of microbial community change over seasons or geographic areas (e.g. Merrill & Halverson, 2002), or as those patterns relate to different plant compartments at different plant developmental stages (Mougel et al., 2006), or the reduction of the complexity of data sets involving hydrochemistry data, bacterial, and archeal community profiles in order to visualize and interpret complex multivariate data sets onto two-dimensional geographic maps of contaminated sites (Mouser et al., 2005). The PCA procedure basically calculates new synthetic variables (principal components), which are linear combinations of the original variables (for instance, the species of a sample-by-species table), and that account for as much of the variance of the original data as possible (Hotelling, 1933). The aim is to represent the objects (rows) and variables (columns) of the data set in a new system of coordinates (generally on two or three axes or dimensions) where the maximum amount of variation from the original data set can be depicted. In practice, PCA is either performed on a variance–covariance matrix or on a correlation matrix. The first approach is followed when the same units or data types are used (e.g. abundance of different species). The aim is then to preserve and to represent the relative positions of the objects and the magnitude of variation between variables in the reduced space. PCA on a correlation matrix is rather used when descriptor variables are measured in different units or on different scales (e.g. different environmental parameters) or when the aim is to display the correlations among (standardized) descriptor variables. The two approaches lead to different principal components and different distances between projected objects in the ordination; hence, the interpretation of the relationships must be made with care (Table 2). Indeed, for correlation matrices, variables are first standardized (i.e. they become independent of their original scales), and so distances between objects are also independent from the scales of the original variables. All variables thus contribute to the same extent to the ordination of objects, regardless of their original variance. Table 2 Interpretation of ordination diagrams Linear methods (PCA, RDA) PCA, RDA RDA Scaling 1 Scaling 2 Samples Species ENV NENV Focus on sample (rows) distance Focus on species (columns) correlation ✓ Euclidean distances among samples – ✓ – Linear correlations among species ✓ Marginal effects of ENV on ordination scores Correlations among ENV ✓ Euclidean distance between sample classes – ✓ ✓ Abundance values in species data ✓ ✓ – Values of ENV in the samples ✓ ✓ Membership of samples in the classes ✓ ✓ Linear correlations between species and ENV ✓ ✓ Mean species abundance within classes of nominal ENV ✓ ✓ – Average of ENV within classes Unimodal methods (CA, CCA) CA, CCA CCA Focus on sample (rows) distance and Hill's scaling Focus on species (columns) distances ✓ Turnover distances among samples χ2 distances between samples ✓ - χ2 distances among species distributions ✓ Marginal effects of ENV Correlations among ENV ✓ Turnover distances between sample classes χ2 distances between sample classes ✓ ✓ Relative abundances of the species table Relative abundances of the species table ✓ ✓ – Values of ENV in the samples ✓ ✓ Membership of samples in the classes ✓ ✓ Weighted averages – the species optima in respect to particular ENV ✓ ✓ Relative total abundances in the sample classes ✓ ✓ – ENV averages within sample classes The interpretation of ordination diagrams depends on the focus of the study, because sample scores are rescaled as a function of the scaling choice. Approximate relationships between and among the different elements represented in biplots and triplots as species (represented as dots or arrows), samples (dots), environmental variables (ENV; arrows), and nominal (qualitative) environmental variables (NENV; dots). A meaningless interpretation (“–”) happens when the suggested comparison is not optimal because of inappropriate scaling of the ordination scores. Adapted from ter Braak (1994); Leps & Smilauer (1999); ter Braak & Smilauer (2002). PCA results are generally displayed as a biplot (Jolicoeur & Mosimann, 1960), where the axes correspond to the new system of coordinates, and both samples (dots) and taxa (arrows) are represented (Fig. 1a). The direction of a species arrow indicates the greatest change in abundance, whereas its length may be related to a rate of change. Depending on whether a distance or a correlation biplot is chosen, different interpretations can be made from the ordination diagram (Table 2). The interpretation of the relationships between samples and species differs and is directly affected by the scaling chosen, i.e. whether the analysis mainly focuses on intersample relationships (scaling 1) or interspecies correlations (scaling 2). For instance, in scaling 1, the distances between objects are an approximation of their Euclidean distances in the multidimensional space, but this approximation is not valid if scaling 2 is chosen (Table 2). Projecting an object at a right angle on a species arrow in the ordination diagram approximates the position of the object along that species descriptor. The length of the species descriptor indicates its contribution to the formation of the ordination space. For correlation biplots, the length of the orthogonal projection of a species arrow on the axes approximates its SD on the respective axes. Angles between species arrows reflect their correlations, e.g. putative interactions between species (scaling 2). An erroneous interpretation of the biplot would be to use the proximity of an object point and the tip of a species arrow to deduce a relationship between them. Indeed, only right-angle projections of samples onto species arrows are correct to derive approximated species abundance in the samples. PCA should generally be used when the objects (sites or samples) cover very short gradients, i.e. when the same species are mostly identified everywhere in the study area (i.e., when samples mostly differ in species abundances), and when species linearly respond to environmental gradients. Because those conditions are often not met in ecological studies, other multivariate approaches have been progressively preferred over PCA (as also suggested by Table 1) such as correspondence analysis or multidimensional scaling. PCA is successful when most of the variance is accounted for by the largest (generally the first two or three) components. The amount of variance accounted for by each principal component is given by its ‘eigenvalue.’ The mathematical description of eigenvalue calculation steps goes beyond the aim of the present review but can be found in most linear algebra manuals. Eigenvalues derived from a PCA are generally considered to be significant when their values are larger than the average of all eigenvalues (Legendre & Legendre, 1998). The cumulative percentage of variance accounted for by the largest components indicates how much proportion of the total variance is depicted by the actual ordination. High absolute correlation values between the synthetic variables (principal components) and the original variables are useful to identify which variables mainly contribute to the variation in the data set, and this is referred to as the loading of the variables on a given axis. However, because the synthetic and original variables are linearly correlated (i.e. they are not independent), standard tests to determine the statistical significance of the correlations between them cannot be used. Principal coordinate analysis (PCoA) The technique is more rarely used by microbial ecologists (Table 1), despite its usefulness at reducing and representing patterns present in distance matrices displaying dissimilarities among objects (Gower, 1966). Its objectives are very similar to those of PCA in that it uses a linear (Euclidean) mapping of the distance or dissimilarities between objects onto the ordination space (i.e. projection in a Cartesian space), and the algorithm attempts to explain most of the variance in the original data set. In microbial ecology, PCoA has been used, for instance, to test whether virulence profiles (i.e. presence or absence of specific genes) arising from pathogenic strains could differentiate either healthy or contaminated hosts (Chapman et al., 2006), or to determine whether PCoA axes could separate groups of Staphylococcus aureus isolates into bovine and human hosts when genetic relationships among them had been established by random amplified polymorphic DNA-PCR analysis (Reinoso et al., 2004). As opposed to PCA, PCoA works with any dissimilarity measure and so specific association coefficients that better deal with the problem of the presence of many double zeros in data sets can be surmounted. Moreover, PCoA does not provide a direct link between the components and the original variables and so the interpretation of variable contribution may be more difficult. This is because PCoA components, instead of being linear combinations of the original variables as in PCA, are complex functions of the original variables depending on the selected dissimilarity measure. Besides, the non-Euclidean nature of some distance measures does not allow for a full representation of the extracted variation into a Euclidean ordination space. In that case, the non-Euclidean variation cannot be represented and the percent of total variance cannot be computed with exactness. The choice of the dissimilarity measure is thus of great importance, and subsequent transformation of the data to correct for negative eigenvalues is sometimes necessary (see Legendre & Legendre, 1998, section 9.2.4. for how to correct for such negative eigenvalues). Objects are represented as points in the ordination space. Eigenvalues are also used here to measure how much variance is accounted for by the largest synthetic variables on each PCoA synthetic axis. Although there is no direct, linear relationship between the components and the original variables, it is still possible to correlate object scores on the main axis (or axes) with the original variables to assess their contribution to the ordination. Correspondence analysis (CA) A basic question that ecologists may want to address when facing a multidimensional table of sites (or samples) by species is whether certain species occur at specific sites, as a measure of their ecological preferences. CA has generally been used in microbial ecology to determine whether patterns in microbial OTU distribution could reflect differentiation in community composition as a function of seasons, geographic origin, or habitat structure (Olapade et al., 2005; Edwards et al., 2006a, b; Kent et al., 2007). The overall aim of the method is to compare the correspondence between samples and species from a table of counted data (or any dimensionally homogenous table) and to represent it in a reduced ordination space (Hill, 1974). Noticeably, instead of maximizing the amount of variance explained by the ordination, CA maximizes the correspondence between species scores and sample scores. Several algorithms exist and the most commonly described one is reciprocal averaging, which consists of (1) assigning arbitrary numbers to all species in the table (these are the initial species scores), (2) for each sample, a sample score is then determined as a weighted average of all species scores (this thus takes into account the abundance of each species at the site and the previously determined species scores), (3) for each species, a new species score is then calculated as the weighted average of all the sample scores, (4) both species scores and sample scores are standardized again to obtain a mean of zero and a SD of one, and (5) steps two to four are repeated until species and site scores converge towards stable solutions in successive iterations (Hill, 1974). The overall table variance (inertia) based on χ2 distances is decomposed into successive components that are uncorrelated to each other, as in the PCA or PCoA procedures. For each axis, the overall correspondence between species scores and sample scores is summarized by an eigenvalue, and the latter is thus equivalent to a correlation coefficient between species scores and sample scores (Gauch, 1982). The technique is popular among ecologists because CA is particularly recommended when species display unimodal (bell shaped or Gaussian) relationships with environmental gradients (ter Braak, 1985), as it happens when a species favors specific values of a given environmental variable, which is revealed by a peak of abundance or presence when the optimal conditions are met (this can be visualized by plotting species abundance against the environmental parameter). The unimodal model that supports the concept of ecological niches has also been shown to be of the right order of complexity for the ordination of most ecological data (ter Braak & Prentice, 1988). Although examples of unimodal distributions along variables or environmental gradients exist with macroorganisms (ter Braak, 1985), the shape of the distribution of the abundance of microbial species along environmental parameters or gradients has not been extensively investigated (but see Ramette & Tiedje, 2007a, b). This may arise from the fact that, in microbial surveys, environmental sampling is mostly performed blindly in relation to environmental heterogeneity, and the abundance of target species is generally determined without systematically analyzing associated environmental parameters. Finally, another important feature of CA for microbial ecologists is that the reciprocal averaging algorithm disregards species double absences because the relationships between rows and columns of the table are quantified using the χ2 coefficient that excludes double absences (Legendre & Legendre, 1998). Both samples and taxa are often jointly depicted in the ordination space (i.e. joint plot; Fig. 2b), where the center of inertia (centroid) of their scores corresponds to the zero for all axes. Depending on the choice of the scaling type, either the ordination of rows (samples) or the columns (species) is meaningful, and can be interpreted as an approximation of the χ2 distances between samples or species, respectively (see Table 2 for more details about interpretation). Sample points that are close to each other are similar with regard to the pattern of relative frequencies across species. It is important to remember that in such joint plots, either distances between sample points or distances between species points can be interpreted, but not the distances between sample and species points. Indeed, these distances are not simple Euclidean distances computed from the relative row or column frequencies, but rather they are weighted distances. The proximity between sample and species points in the plot can thus be understood as a probability of species occurrence or of a high abundance in the samples in the vicinity of a species point. Fig. 2 Ordination diagrams in two dimensions. (a) In a PCA biplot representation, samples are represented by dots and species by arrows. The arrows point in the direction of maximal variation in the species abundances, and their lengths are proportional to their maximal rate of change. Long arrows correspond to species contributing more to the data set variation. Right-angle projection of a sample dot on a species arrow gives approximate species abundance in the sample. (b) In a CA joint plot representation focusing on species distance, both samples and species are depicted as dots. Species dots correspond to the center of gravity (inertia) of the samples where they mostly occur. Distances between sample and species points give an indication of the probability of species composition in samples (see Table 2 for more details about diagram interpretation). In scaling 2 (i.e. focus on species), species points found at the center of the ordination space should be carefully checked with the raw data to clarify whether the species ordination really corresponds to the optimal abundance or occurrence of the species, or whether the species is just badly represented by the main axes, as it is the case when other axes are more appropriate to represent the species. Rare species contribute little to the total table inertia (i.e. they only play a minor role in the overall table variance) and are hence positioned at the edges of the plot, next to the site(s) where they occur. In general, only the species points found away from the ordination center and not close to the edges of the ordination have more chances to be related to the ordination axes, i.e. to contribute to the overall variance (Legendre & Legendre, 1998). When the species composition of the sites progressively changes along the environmental gradient, sample positions may appear in the ordination plot as nonlinear configurations called ‘arch’ (Gauch, 1982) (or ‘horseshoe’ in the case of PCA), which may impair further ecological interpretation. In CA, the arch effect may be mathematically produced as a side-effect of the CA procedure that tries to obtain axes that both maximally separate species and that are uncorrelated to each other (ter Braak, 1987): when the first axis suffices to correctly order the sites and species, a second axis (uncorrelated with the former) can be obtained by folding the first axis in the middle and bringing its extremities together, thus resulting in an arch configuration. Further axes can be obtained by further dividing and folding the first axis into segments (Legendre & Legendre, 1998). To remove the arch effect in CA, a mathematical procedure, detrending, is used to flatten the distribution of the sites along the first CA axis without changing their ordination on that axis. The approach is then designated as detrended correspondence analysis (DCA). The review of different detrending algorithms such as using segments or polynomials goes beyond the scope of this review, but more information can be obtained in (ter Braak & Prentice, 1988; Legendre & Legendre, 1998). Some authors have also argued that the arch effect may not be an artifact but an expected feature of the analysis, especially when species turnover is high along environmental gradients (James & McCulloch, 1990). In that case, if the samples are meaningfully positioned along the arch, the ordination should be accepted as a valid result. Nonmetric multidimensional scaling (NMDS) NMDS is generally efficient at identifying underlying gradients and at representing relationships based on various types of distance measures. Not surprisingly, NMDS has found an increasing number of applications in microbial ecology (Table 1). The technique has been generally applied to identify patterns among multiple samples that were subjected to molecular fingerprinting techniques. For instance, NMDS was used to analyze and to compare the reproducibility of various fingerprinting techniques such as ribosomal internal spacer analysis (RISA), terminal fragment length polymorphism (T-RFLP), and denaturing gradient gel electrophoresis (DGGE) between different laboratories when applied to samples chosen from a salinity gradient (Casamayor et al., 2002). NMDS was also used to compare diversity patterns of microbial communities (as determined by length heterogeneity-PCR) from samples undergoing different land management practices (Mills et al., 2006). Another example is the analysis of the bacterioplankton communities of four shallow eutrophic lakes that differed in nutrient load and food web structure using DGGE profiling, so as to determine the specificity of community signatures in each lake (Van der Gucht et al., 2005). The NMDS algorithm ranks distances between objects, and uses these ranks to map the objects nonlinearly onto a simplified, two-dimensional ordination space so as to preserve their ranked differences, and not the original distances (Shepard, 1966). The procedure works as follows: the objects are first placed randomly in the ordination space (the desired number of dimensions has to be defined a priori), and their distances in this initial configuration are compared by monotonic regression with the distances in the original data matrix based on a stress function (values between 0 and 1). The latter indicates how different the ranks on the ordination configuration are from the ranks in the original distance matrix. Several iterations of the NMDS procedure are generally implemented so as to obtain the lowest stress value possible (i.e. the best goodness of fit) based on different random initial positions of the objects in the ordination space. For sample-by-species tables, simulations have shown that before applying NMDS, a standardization of each species by its maximum abundance, followed by the computation of distances between samples based on the Steinhaus or Kulczinski similarity coefficients yielded informative ordination results (Legendre & Legendre, 1998, p. 449). In NMDS ordination, the proximity between objects corresponds to their similarity, but the ordination distances do not correspond to the original distances among objects. Because NMDS preserves the order of objects, NMDS ordination axes can be freely rescaled, rotated, or inverted, as needed for a better visualization or interpretation. Because of the iterative procedure, NMDS is more computer intensive than eigenanalyses such as PCoA, PCA, or CA. However, constant improvement in computing power makes this limitation less of a problem for small- to medium-sized matrices. Testing for significant differences between groups In addition to representing objects in an ordination plot or as clusters of similar objects, another objective may be to test whether differences between groups of objects (rows) in a multivariate table are significantly different based on the set of their attributes (columns), i.e. to test whether similarities within groups are higher than those between groups. Here, nonparametric multivariate anova (npmanova) and analysis of similarities (anosim), which are commonly found in standard statistical packages, are briefly reviewed. It is also possible to use canonical analyses (‘Testing for significant differences between groups’) to test for significant differences between groups of objects. These statistical tests, however, must not be used to assess the statistical difference among groups that were derived from a previous cluster analysis on the same variables because, under those conditions, the two approaches would not be independent from each other. Indeed, the groups derived from cluster analysis (which are themselves made to fit the data) would then be used for testing the null hypothesis that there is no difference among the groups. This hypothesis would then not be independent of the data used to test it, and would nearly always produce significant differences between the groups even if it is not the case (Legendre & Legendre, 1998). npmanova The method can be used to test for significant differences between the means of two or more groups of multivariate, quantitative data (Anderson, 2001). The null hypothesis of equality of means is tested based on Wilks' Λ (lambda) statistic, which replaces the F-test normally used in univariate anova. When only two groups are compared, Hotelling's T2 test is more appropriate. The latter test can also be used, as a post hoc test, to assess the significance of pairwise comparisons statistically between groups, following an overall significant Wilks' test. Significance is generally computed by permutation of group membership, with several thousand replicates, alleviating concerns about multinormality of the data. Because multiple pairwise comparisons are made, the significance level of the pairwise Hotelling's tests needs, however, to be corrected. With the Bonferroni correction, for instance, the P-value usually chosen for significant differences between groups (i.e. 0.05) is replaced by a smaller P-value calculated by dividing the original P-value by the total number of pairwise comparisons that are performed. For instance, for 10 pairwise comparisons, the corrected P-value becomes 0.005. This correction is often judged to be rather conservative as it leads to significance for fewer pairwise comparisons (Legendre & Legendre, 1998). anosim This nonparametric procedure tests for significant difference between two or more groups, based on any distance measure (Clarke, 1993). It compares the ranks of distances between groups with ranks of distances within groups. The means of those two types of ranks are compared, and the resulting R test statistic measures whether separation of community structure is found (R=1), or whether no separation occurs (R=0). R values >0.75 are commonly interpreted as well separated, R>0.5 as separated, but overlapping, and R<0.25 as barely separable (Clarke & Gorley, 2001). The test makes fewer assumptions than manova because it is based on the ranks of distances, and it is often used for sample-by-species tables, where groups of samples are compared. All groups should have comparable within-group dispersion to avoid finding falsely significant results (Legendre & Legendre, 1998). Applications in microbial ecology include testing for spatial differences, temporal changes, or environmental impacts on microbial assemblages. For instance, Kent et al. (2007) determined whether bacterial communities from the same lake were more similar in composition to each other than to communities in different lakes. The bacterial composition and diversity of samples from different geographic origins, habitats, and avian hosts were also compared using anosim based on a length heterogeneity (LH)-PCR (Bisson et al., 2007). Another example is the application of anosim to terminal restriction fragment length polymorphism (T-RFLP)-generated data to determine the impact of B and NaCl on soil microbial community structure in the wheat rhizosphere (Nelson & Mele, 2007). Environmental interpretation Exploratory analyses may reveal the existence of clusters or groups of objects in a data set. When a supplementary table or matrix of environmental variables is available for those objects, it is then possible to examine whether the observed patterns are related to environmental gradients. Typical objectives may be, for instance, to reveal the existence of a relationship between community structure and habitat heterogeneity, between community structure and spatial distance, or to identify the main variables affecting bacterial communities when a large set of environmental variables has been conjointly collected. The significance of the relationships between species patterns and environmental variables can generally be assessed by permutation techniques such as Monte Carlo permutation tests, which infer statistical properties from the data themselves. The order of data (generally the rows of one matrix) is permuted and the relationships between the observed patterns and environmental variables can be assessed for randomness. This approach is particularly suitable when variables do not follow a normal distribution (which is often the case with environmental or ecological data), as generally required by traditional statistical procedures (Legendre & Legendre, 1998). Indirect gradient analyses Ordination axes or clusters can be interpreted based on additional environmental variables (i.e. variables not used in the ordination or cluster analysis) that provide ecological knowledge about the studied sites or species ecological characteristics. When using exploratory ordination approaches on a sample-by-species table, samples are displayed along the axes of main variation in species composition. These axes are thus constructed without reference to environmental characteristics, but they can be hypothesized to represent underlying environmental gradients (e.g. environmental parameters, spatial or temporal variables, chemical gradients), which need to be subsequently identified. Such an approach is designated as ‘indirect,’ because synthetic variables (i.e. the axes) are first constructed and thereafter related to environmental variation. For instance, the scores of the objects on PCA or CA main components (axes) can be further related by standard statistical procedures (e.g. anova, regression analysis) to environmental variables. Likewise, in PCoA or NMDS, it is possible to statistically compare the ranks obtained by the objects on each axis and the ranks of those objects on additional environmental variables, using Spearman's rank correlation coefficients (Legendre & Legendre, 1998). A convenient method of interpretation is to represent the additional environmental variables as fitted arrows directly on the ordination diagram. These variables are added to the existing ordination by linear regression of their values onto the existing ordination axes. This procedure is implemented in various statistical packages (e.g. canoco, R). Hence, it is possible to assess the direction and magnitude of the most rapid change in the environmental variables and to determine whether they correspond to the observed patterns among objects (Oksanen, 2007). In cluster analysis, the magnitude of the absolute correlation value between an ordered clustering solution and environmental variables may also provide clues about putative environmental causes for the observed discontinuities in the data set. Another convenient way of displaying additional information to help interpret the ordination is to use site symbols whose sizes are proportional to the values of the additional variable. Hence, one can visually assess whether the ordination of objects (samples, sites) matches specific trends in the additional variable. This strategy was, for instance, used on NMDS ordination plots inferred from DGGE profiles on which the values of five additional environmental variables were individually mapped as proportional circles in order to identify the main environmental factors related to the bacterial community structure in four freshwater lakes (Van der Gucht et al., 2005). Direct gradient analyses (constrained analyses) In constrained (canonical) ordination analyses, only the variation in the species table that can be explained by the environmental variables is displayed and analyzed, and not all the variation in the species table. Gradients are supposed to be known and represented by the measured variables or their combinations, while species abundance or occurrence is considered to be a response to those gradients. Constrained ordinations are mostly based on multivariate linear models relating principal axes to the observed environmental variables, and the different techniques depend on data types (matrix or table), and on the hypothesis underlying species distribution in the gradients (i.e. linear or unimodal). Their aim is to find the best mathematical relationships between species composition and the measured environmental variables, and to assess whether, statistically, such a relationship could have been produced due to chance alone using permutation tests. The resulting ordination diagrams display samples, species, and environmental variables so that ‘fitted species × samples’ and ‘species × environment’ relationships can be derived as easily as possible from angles between arrows or distances between points and arrows (Table 2). Redundancy analysis (RDA) In microbial ecology, RDA has been applied, for instance, to test whether the occurrence of biocontrol bacteria with specific carbon source utilization profiles was related to their origin from different root samples (Folman et al., 2003), to determine which environmental factors were the most significant to explain variation in microbial community composition in undisturbed native prairies and cropped agricultural field (McKinley et al., 2005), to examine the effects of sampling locations (longitude, latitude, altitude) on genetic diversity of plant pathogenic bacteria (Kolliker et al., 2006), or to assess the influence of season, farm management, and soil chemical, physical, and biological properties on nitrogen fluxes and bacterial community structure (Cookson et al., 2006). This method can be considered as an extension of PCA in which the main axes (components) are constrained to be linear combinations of the environmental variables (Rao, 1964). Two tables are then necessary: one for the species data (‘dependent’ variables) and one for the environmental variables (‘independent’ variables). Multiple linear regressions are used to ‘explain’ variation between independent and dependent variables, and these calculations are performed within the iterative procedure to find the best ordination of the objects. The interest of such an approach is to represent not only the main patterns of species variation as much as they can be explained by the measured environmental variables but also to display correlation coefficients between each species and each environmental variable in the data set. When the data set consists of a matrix of distances between objects, distance-based RDA (db-RDA; Legendre & Anderson, 1999) can be applied to determine how well additional environmental parameters can explain the variation among objects in the matrix. The technique first applies a PCoA on the distance matrix to convert it back to a rectangular table containing rows of objects by columns of PCoA coordinates. Those new, uncorrelated coordinates thus correspond to synthetic ‘species’ variables that are then related to additional environmental parameters using a classical RDA. For instance, db-RDA was successfully used to determine how the variation in matrices of genomic distances among environmental strains could be explained by factors such as soil parameters, host plant species, and spatial scale, each factor being taken alone or in combination (Ramette & Tiedje, 2007b). Most software outputs provide the total variation in species composition as explained by the environmental axes, the cumulative percentage of variance of the species–environment relationship, and the overall statistical significance of the relationships between the species and environmental tables. RDA can be represented by a triplot of samples (dots), species (arrows), and environmental variables (arrows for quantitative variables and dots for each level of qualitative or nominal variables), or by any combinations thereof (i.e. biplots) (ter Braak, 1994). Depending on the scaling chosen, i.e. whether the analysis mainly focuses on intersample relationships or interspecies correlations, the interpretation of the relationships between samples, species, and environmental variables differs (Table 2). Canonical correspondence analysis (CCA) The approach is very similar to that of RDA, except that CCA is based on unimodal species–environment relationships whereas RDA is based on linear models (ter Braak, 1986). CCA can be considered as the constrained form of CA in which the axes are linear combinations of the environmental variables. CCA uses the unimodal model to model species response to the environmental variation as a mathematical simplification to enable the estimation of a large number of parameters and the identification of a small number of ordination axes. This species model seems, however, to be robust even when some species display bimodal responses, unequal ranges, or unequal maxima along environmental gradients, and the technique is thus considered to be the method of choice by many ecologists (ter Braak & Smilauer, 2002). It is therefore particularly adapted for the environmental interpretation of tables of abundance and occurrence of species, and accommodates well the absence of species at certain sites in the data set. CCA is sensitive to rare species that occur in species-poor samples, and down-weighting of such species help reduce the problem (Legendre & Legendre, 1998). Software outputs are very similar to those of RDA and as for RDA, triplot and biplot representations and interpretation depend on the choice of the scaling type (Table 2). The same interpretation of the relationships between sample and species points is found in CA and CCA. Right-angle projection of these points on the environmental arrows leads to the correct approximation of the ranking of the points along environmental variables. CCA has been used in an increasing number of publications dealing with microbial assemblages in marine and soil ecosystems. Typical questions that are addressed concern the identification of environmental factors that influence the diversity of bacterial assemblages among large sets of candidate environmental parameters measured for the same samples, when the diversity is determined by culture-independent, genetic fingerprinting techniques such as automated ribosomal intergenic spacer analysis (ARISA) (Yannarell & Triplett, 2005), DGGE (Salles et al., 2004; Sapp et al., 2007), or T-RFLP (Córdova-Kreylos et al., 2006; Klaus et al., 2007). Another interest in the technique comes from the possibility of determining the specific species or OTUs that respond to particular environmental variables, and as such that can be identified as candidate indicator species. Those species can then be subjected to further experiments so as to confirm their status of indicator species. For instance, the relationships, as determined by CCA, between bacterial community composition and 11 environmental variables for 30 lakes in Wisconsin, revealed that patterns in bacterial communities were best explained by regional- and landscape-level factors, as well as by specific seasons, pH, and water clarity (Yannarell & Triplett, 2005). CCA was also successfully used to demonstrate that former land use management affected the composition of the targeted soil microbial community (Burkholderia) to a larger extent than did plant species (Salles et al., 2004). Another interesting example in the marine ecosystem is the study of the interactions between various abiotic parameters and phytoplankton community data (biotic parameter) to explain bacterioplankton dynamics in the North Sea and the subsequent identification of the bacterial phylotypes responding more specifically to the factors (Sapp et al., 2007). Another example of using CCA to identify some microbial communities as pollution indicators can be found in (Córdova-Kreylos et al., 2006). Partial ordination, variation partitioning When the effects of a particular environmental variable need to be tested after elimination of possible effects due to other (environmental) variables, partial ordination may be used (e.g. partial CCA, partial RDA). Such an approach is also referred to as ‘partialling out’ or ‘controlling for’ the effects of specific variables, which are specified as covariables in the constrained analysis. For instance, in a study dealing with the effects of environmental and pollutant variables on microbial communities, Córdova-Kreylos et al. (2006) observed that variation in microbial communities was more due to spatial variation than to pollutants. The use of partial CCA to account for spatial variation in the biological data set revealed that metals had a greater effect on microbial community composition than organic pollutants. This idea of controlling for the effects of specific variables can be extended to evaluate the effects of all the different sets (factors) of environmental variables present in a study so as to determine the relative contribution (amount of variation explained) and significance of each variable set on the total biological variance. The so-called variation partitioning procedure (Borcard et al., 1992) partitions the total variance of the species table into the respective contribution of each set of environmental variables and into their covariations using both standard and partial constrained ordinations (Fig. 3). Two methods have traditionally been used to partition the variation of community composition data, i.e. canonical partitioning and regression on distance matrices based on Mantel tests (Legendre & Legendre, 1998). The canonical approach has been shown to be more appropriate to partition the β diversity correctly among sites and to test hypotheses about the origin and maintenance of its variation (Legendre et al., 2005). Fig. 3 Partitioning biological variation into the effects of two factors. The large rectangle represents the total variation in the biological data table, which is partitioned among two sets of explanatory variables (a, b). Fraction 4 shows the unexplained part of the biological variation. Fractions 1 and 3 are obtained by partial constrained ordination or partial regression, and can be tested for significance. For instance, fraction 1 corresponds to the amount of biological variation that can be exclusively explained by (a) effects when (b) effects are taken into consideration (i.e., when b is considered as a covariable). Fraction 2 [i.e., variation indifferently attributed to (a) and (b) or a covariation of (a) and (b)] is obtained by subtracting fractions 1 and 3 from the total explained variance, and cannot be tested for statistical significance. Applications of variation partitioning in microbial ecology include, for instance, the study by Ramette & Tiedje (2007b), which applied the technique in the context of RDA to disentangle the effects of space, environmental soil parameters, and plant species on Burkholderia community abundance and diversity. By quantifying the amount of biological variation that is left unexplained when all environmental variables had been considered, the study suggested that much less of the biological variation could be predicted at the intraspecific level compared with higher taxonomic levels. Another interesting example is the study of seasonal changes in bacterial community composition in shallow eutrophic lakes, in which top-down regulation (grazers) of bacterial community composition was examined after accounting for bottom-up regulation (resources) (Muylaert et al., 2002). Linear discriminant analysis (LDA) When groups or clusters of objects have been obtained by exploratory analyses for instance, LDA can be used to identify linear combinations of additional environmental variables that best discriminate those groups. In that respect, LDA can be seen as an extension of manova for two or more groups, in which environmental variables that specifically explain the grouping of objects are identified. Another application consists of assigning new objects to previously defined groups for prediction or classification purposes based on the calculated discriminant function. For instance, Fuhrman et al. (2006) used the technique to evidence the existence of repeatable temporal patterns in the community composition of marine bacterioplankton over 4.5 years. The technique is mostly recommended for multinormal data for which attribute data are linearly related and for which variances and covariances of the variables are good summary statistics. A visual representation of LDA can be performed, and in the resulting ordination, the axes are then the discriminant functions. The distances between objects, which correspond to Mahalanobis distances that take into account the correlations among descriptors (Mahalanobis, 1936), are independent of the scale of measurement of the various descriptors and are mostly used to compare groups of sites or objects with each other (Legendre & Legendre, 1998). Selection of variables in regression models In the previous constrained methods where linear combinations of environmental (explanatory) variables are used, the inclusion of too many explanatory variables to describe species distribution may lead to difficult ecological interpretations and to lower predictability of the models, due to intercorrelations among the explanatory variables (i.e. multicollinearity). Multicollinearity has the effects of inflating the variance of the regression coefficients in the models, leading to reduced precision in the prediction of the response variables (Legendre & Legendre, 1998). In order to only include in the model the environmental variables that mostly and significantly contribute to the variation of the species table, automatic selection procedures (forward selection, backward elimination, or stepwise selection) are often used. The selection depends on whether the partial correlation coefficients of the variables fall below a given significance level, the latter being generally assessed by Monte Carlo permutation tests. In forward selection, the construction of the regression model starts with the variable that explains the most variation in the dependent variables (generally the species table). What remains of the biological variation to explain after fitting the first environmental variable (i.e. of the residual variation) is then used to choose the second environmental variable. The process of selection goes on until no more variables significantly explain the residual variation. In backward elimination, the construction of the regression model starts with all environmental variables and the least significant ones are excluded from the model, one at a time until a group of only ‘significant’ variables is obtained. To take advantage of the two approaches, stepwise regression mixes forward selection with backward elimination by performing a forward selection, but excluding the variables that no longer become significant after the introduction of new variables into the regression model. Despite the clear advantages of these variable selection strategies, most authors still caution that researchers should not blindly rely on automatic selection procedures to choose the relevant environmental variables in regression models because ecologically irrelevant models may also be obtained, or other variable combinations could also yield better models to explain species variation (Legendre & Legendre, 1998). Noticeably, the three selection strategies do not necessarily yield the same set of significant environmental variables, because they may be seen as heuristic methods to identify a significant model when all possible combinations of significant models are not possible to evaluate computationally. Another approach is thus to combine variables into biologically or environmentally meaningful sets, instead of relying on automatic selection procedures, and then to examine all possible regression models based on the reduced number of variable sets (James & McCulloch, 1990). For instance, before applying variation partitioning to different groups of variables representing spatial scales (15 variables), host species (four variables), and soil parameters (10 variables), Ramette & Tiedje (2007b) applied forward selection within each group to determine the variables significantly explaining the variation of microbial diversity and abundance at different taxonomic levels. Mantel test This test is appropriate to compare two matrices that were calculated for the same objects but that are based on two independent data sets (e.g. a species dissimilarity matrix and an environmental dissimilarity matrix for the same samples) (Mantel, 1967). It calculates the correlation coefficient between corresponding positions in the two matrices, and assesses its significance based on permutations of the objects in one of the matrices. In microbial ecology, the Mantel test has become popular especially for testing the relationships between molecular and geographic distance matrixes for a same set of organisms or to relate community diversity to environmental heterogeneity (e.g., Parker & Spoerke, 1998; Cho & Tiedje, 2000; Horner-Devine et al., 2004; Scortichini et al., 2006). Another interesting application, called a goodness-of-fit Mantel test, corresponds to the case where one matrix is recoded to represent ecological hypotheses to be tested on the other matrix (Legendre & Legendre, 1998). For instance, if a matrix of molecular data is available for a set of strains and their habitat of origin is known, it is possible to determine whether the genetic distances are related to habitat type using the (goodness-of-fit) Mantel test. The matrix representing the ecological hypotheses should then consist of a series of 1 and 0 for isolates found in the same or different habitats, respectively. The Mantel test can thus determine whether the posited habitat distribution can significantly explain the structure of the molecular matrix. This test cannot be used, however, to test a hypothesis matrix that would be based on the results of a cluster analysis, for instance. Indeed, as indicated in ‘Testing for significant differences between groups,’ there would be a lack of independence between the hypothesis being tested and the data used to test the hypothesis. Note that the Mantel test is also used to compute Mantel correlograms, which are often found in biogeographical studies (e.g. Mantel correlograms are usually used to detect spatial structure in species assemblages based on grouping of the response data into specific spatial distance classes). Mantel tests are then applied to each group in order to detect significant correlations at a given scale, i.e. the scales at which the data are autocorrelated (Legendre & Legendre, 1998). Practical considerations Choice of an ordination method (Fig. 4) Linear methods such as multiple regression, LDA, PCA, or RDA are generally meant to be applied to continuous data. Their use is thus sometimes limited in Ecology where species generally display nonlinear, nonmonotone responses to environmental variables (ter Braak & Prentice, 1988; Legendre & Legendre, 1998). Different approaches can be undertaken to choose the most appropriate ecological model. Plot of species abundances along ordination axes or explanatory variables (also called coenocline) may help visualize whether species responses are linear or unimodal (ter Braak & Smilauer, 2002). Besides, the choice of linear (PCA, RDA) or unimodal (CA, CCA) species response models can be made on the basis of whether the underlying gradient length is short or long, respectively. Gradient length, as measured in SD units along the first ordination axis, can be estimated by DCA for unconstrained ordination and by detrended CCA (DCCA) for constrained ordination in, e.g. the software canoco (ter Braak & Smilauer, 2002). It is recommended to use linear methods when the gradient length is <3 SD, unimodal methods when it is >4 SD, and any method for intermediate gradient lengths (ter Braak & Smilauer, 2002). Fig. 4 Relationships between numerical methods. Exploratory tools such as PCA, CA, PCoA, NMDS, or cluster analysis can be applied to a sample-by-species table to extract the main patterns of variation, to identify groups or clusters of samples, or specific species interactions. Sample scores on the main axes of variation can be related to variation in environmental variables using indirect gradient analyses. When a constrained analysis is desired (i.e. direct gradient analysis), RDA, db-RDA, CCA, or linear discriminant analysis can be used as extensions of the unconstrained methods. Mantel tests are appropriate to test the significance of the correlation between two distance matrices (e.g. one based on species data and the other on environmental variables). Raw data may be transformed, normalised or standardised as appropriate before analysis. Data type is also another important criterion. To represent absolute abundance values, linear-based methods (PCA, RDA), which produce weighted summations, are appropriate, whereas unimodal techniques (CCA, CA) are rather used to model relative abundances (because species scores are weighted averages of the samples scores, and vice versa), i.e. they model the dissimilarities between samples (β diversity). They also accommodate well the presence of many zeros in the species table, in contrast to linear-based methods for which double zeros lead to inadequate estimates of sample distances. Cluster analysis is the method of choice when relationships between objects are expected to be discontinuous and where defined categories or groups of objects are expected. On the contrary, ordination would be more useful when the variation between objects is posited to be continuous. Although NMDS is more computer intensive than PCoA, it is generally better at compressing the distance relationships among objects into a few dimensions. This is because NMDS can always lead to a Euclidean representation even for non-Euclidean embeddable distances (Legendre & Legendre, 1998). NMDS and PCoA can be compared using Shepard diagrams to decide which technique better represents the original distances. If one assumes that species do not have a linear response to environmental gradients, NMDS is more appropriate than PCA. CA may also be an alternative to PCA when many zeros populate the data set and one strong gradient is present. With long ecological gradients, however, CA may produce the arch effect that can be corrected for using DCA. In terms of the underlying species model, the main difference between DCA and NMDS is that the former is based on a specific model of species distributions (unimodal model), while NMDS is not. Thus, DCA may be favored by ecologists who assume that the niche theory better fits their data set, while NMDS may be a method of choice if species composition is determined by factors other than position along a gradient (for instance if the habitat is known to be fragmented). In constrained and unconstrained ordinations, all species are posited to react to different extents to the same composite gradients of environmental variables, whereas in a multiple regression approach, a different gradient could be modeled for each species separately. Because most species do not respond linearly to environmental gradients, fitting nonlinear models to individual species may be difficult, especially when dealing with a huge data set. Constrained ordinations thus provide a good summary of species–environment relationships and can be very successful in ecological data analysis (ter Braak & Prentice, 1988). It is also useful to note that RDA is very similar to manova, but in contrast to the latter, RDA allows the consideration of any number of species (columns) (Legendre & Legendre, 1998). Constrained and unconstrained (exploratory) methods should be used in parallel (Fig. 4) because, with the former, only the biological variation that can be explained by the available environmental variables is represented on the main axes, whereas with unconstrained methods, the highest amount of variance is extracted from the biological data alone and represented on a few axes. If the constrained and unconstrained approaches yield the same ordination of the samples (objects), it thus means that the measured environmental variables explain most of the biological variation. In order to compare the results of different ordinations, a useful technique is Procrustes analysis (Gower, 1975), which estimates the concordance of scores in two ordinations after rotating, translating, and dilating them in order to obtain the best fit. A permutation procedure can also be used to test for the significance of the concordance between ordinations or matrices (Peres-Neto & Jackson, 2001). Cluster analysis and ordination techniques can be combined to provide powerful visualization tools. For instance, hierarchical clustering can help obtain a better interpretation of ordination diagrams (Fig. 5). Because ordination diagrams represent most of the data set variation into a dimensionally reduced space, some relationships among objects can be distorted because only a few projection axes are considered. The addition of linkage results obtained from cluster analysis may help identify objects belonging to the same clusters even if their relative position in the ordination diagram is not ideal (Legendre & Legendre, 1998). Fig. 5 Combination of ordination and cluster analysis. On a same distance matrix, NMDS or PCOA can be applied to represent the major axes of variation among objects in a two-dimensional space. The superimposition of the results of cluster analysis (primary connections) onto the ordination diagram can help identify the structure in the data set as discontinuities (clusters) into a continuous space (ordination). Adapted from Legendre & Legendre (1998). Ordination and diversity indices The measurement of diversity is generally performed using indices such as the Shannon or Simpson indices. The latter are often applied to measure different components of the diversity such as α, β, and γ diversity, corresponding to diversity within a particular site or ecosystem, to change in species composition from site to site (i.e. species turnover), and to the diversity at the landscape scale, respectively (Whittaker, 1972). The ordination approach sounds similar, in that variation among samples is compared based on their within-sample composition in species assemblages, and so some of the α and β diversity should be depicted on ordination diagrams. Because diversity indices pool the multispecies information into a single value for each observation, before comparing them, it is not surprising that complex diversity patterns may not be identified sometimes. For instance, Hartman & Widmer (2006) did not find significant changes in soil bacterial communities submitted to various soil managements when using diversity indices, while community structures were shown to have changed using community fingerprinting analysis. To obtain a consistency between ordination techniques and diversity index measurements, two numerical strategies have been proposed: for species occurrence data, the CA-species richness strategy adapted for data set rich in rare species, and the Nonsymmetric CA – Simpson strategy, which is more appropriate for tables dominated by abundant species (Pelissier et al., 2003). These strategies attribute specific weights to the species data so that simple or constrained ordinations of the new species table represent the total inertia as α and β diversity, and would thus be consistent with the measures obtained by common diversity indices. Misconceptions about multivariate analyses It is essential to reiterate that multivariate statistical procedures may suggest causes or factors, but investigators should bear in mind that the synthetic variables, axes, or clusters derived do not necessarily correspond to biological or ecological entities in nature (James & McCulloch, 1990). One should thus not overinterpret the data by relying on unjustified causality, especially in the absence of real experimentation. In theory, it would be necessary to validate the inferences and models made about pattern formation and putative causes by analyzing new data, but this is rarely performed in practice. Moreover, whether the originally collected data are typical of the situation to be described is most of the time not even questioned. Another common misconception is that multivariate analyses alone can sort out all solutions of complex multivariate studies. Although exploratory analyses may help reveal interesting patterns in data sets, the interpretation and explanation of the observations ultimately rely on the researcher's hypotheses and previous knowledge of the ecological situation. Microbial ecologists themselves need to formulate ecologically sound hypotheses and test them. Conclusions Exciting questions in Ecology typically consist of determining whether community patterns are structured across space or time, of explaining how those patterns can be related to environmental heterogeneity, and of quantifying how much still remains unexplained when all significant, measured variables have been considered. Such questions can now start to be addressed in microbial ecology because numerical tools may help explore and test such ecological hypotheses. These are indeed exciting times because even larger and more complex databases are being created and in parallel, computing power gradually becomes less of an issue. If microbial ecologists want to test numerical methods, develop new ecological theories, or validate existing ones for the microbial case, access to diversity data and above all, to the relevant associated environmental parameters, becomes a central issue. It would thus be of great interest to make such complex data sets publicly available, such as microbial ecological databases, so that microbial diversity can be studied in its environmental context. This would indeed be a step toward making microbial ecology a central discipline in Ecology.

Eur_J_Appl_Physiol-4-1-2358938

Characteristics of fast voluntary and electrically evoked isometric knee extensions during 56 days of bed rest with and without exercise countermeasure

The contractile characteristics of fast voluntary and electrically evoked unilateral isometric knee extensions were followed in 16 healthy men during 56 days of horizontal bed rest and assessed at bed rest days 4, 7, 10, 17, 24, 38 and 56. Subjects were randomized to either an inactive control group (Ctrl, n = 8) or a resistive vibration exercise countermeasure group (RVE, n = 8). No changes were observed in neural activation, indicated by the amplitude of the surface electromyogram, or the initial rate of voluntary torque development in either group during bed rest. In contrast, for Ctrl, the force oscillation amplitude at 10 Hz stimulation increased by 48% (P < 0.01), the time to reach peak torque at 300 Hz stimulation decreased by 7% (P < 0.01), and the half relaxation time at 150 Hz stimulation tended to be slightly reduced by 3% (P = 0.056) after 56 days of bed rest. No changes were observed for RVE. Torque production at 10 Hz stimulation relative to maximal (150 Hz) stimulation was increased after bed rest for both Ctrl (15%; P < 0.05) and RVE (41%; P < 0.05). In conclusion, bed rest without exercise countermeasure resulted in intrinsic speed properties of a faster knee extensor group, which may have partly contributed to the preserved ability to perform fast voluntary contractions. The changes in intrinsic contractile properties were prevented by resistive vibration exercise, and voluntary motor performance remained unaltered for RVE subjects as well. Introduction Previous research has shown that exposure to actual or simulated spaceflight leads to pronounced muscle atrophy in humans. The associated muscle weakness significantly impairs the performance of various motor tasks (for reviews see Adams et al. 2003; Desplanches 1997; Edgerton and Roy 2000; Fitts et al. 2000). Muscle function (e.g., maximal isometric force) is often further impaired by adaptations in the central motor control system, as indicated by reductions in the amplitude of the surface electromyogram (EMG) during maximal voluntary contractions (Berg et al. 1997; Deschenes et al. 2002; di Prampero and Narici 2003; Gondin et al. 2004; Schulze et al. 2002). Compared to steady-state contractions, much higher levels of neural activation are needed for contractions where torque develops as rapidly as possible. This is seen for both voluntary (de Ruiter et al. 2004) as well as for electrically evoked contractions (de Haan 1998; de Ruiter et al. 1999). In daily life, elderly individuals who lack sufficient motor speed or possess poor lower extremity strength have an increased risk of fall-related bone fractures (Shigematsu et al. 2006). The same may hold for astronauts suffering from muscle atrophy, neural deconditioning and increased bone fragility following space missions (Smith and Heer 2002). The adaptive physiological responses in the human body as a consequence of real or simulated space flight may be offset by means of efficient countermeasures. With respect to the neuromuscular system, it appears that resistance exercise (strength training) is effective to maintain, or at least to minimize changes in muscle mass and strength during bed rest (Akima et al. 2001; Ferrando et al. 1997; Kawakami et al. 2001). Previously, we have shown by means of the twitch interpolation technique, as well as by means of EMG recordings that neural deconditioning was absent during 56 days of bed rest for maximal steady-state contractions, regardless of whether subjects participated in an exercise countermeasure program (Mulder et al. 2006, 2007). Nonetheless, based on the above, it can be hypothesised that the neural activation of explosive isometric contractions is more deteriorated by bed rest than the activation of steady-state contractions (Mulder et al. 2007). The first aim of the study was to test the hypothesis that, in the absence of an exercise countermeasure, the maximal rate of voluntary isometric torque development would show a greater decrease during 56 days of bed rest than the maximal steady-state torque, due to reduced neural drive during the fast voluntary contractions. Such an effect was expected to be prevented by combining daily resistive exercise training with vibration training, which was hypothesized to provide a better protection against musculoskeletal deconditioning during bed rest than heavy-resistance training only (Rittweger et al. 2006). Although the individual merits of resistance training versus vibration training cannot be quantified with such a study design, the efficacy of this combined countermeasure to preserve neural activation and maximal steady-state torque is reported elsewhere (Mulder et al. 2007). Apart from neural activation, the rate at which muscle force develops under voluntary command is also determined by peripheral factors, such as the intrinsic contractile muscle fibre speed (Andersen and Aagaard 2006) and the stiffness of the series elastic component (Bojsen-Moller et al. 2005). Both factors are known to be altered by exposure to actual or simulated space flight, but their effect is opposite. Whereas faster intrinsic contractile speed characteristics (Talmadge 2000) can partly or fully compensate for the effect of atrophy on power output of single muscle fibres (Widrick et al. 1998), the power output of whole muscles (with intact tendons) would be diminished by decrements in tendon stiffness (Kubo et al. 2000; Reeves et al. 2005). The second aim of the study was to assess whether changes in peripheral factors influenced the voluntary rate of torque development during bed rest. The functional change in peripheral factors, i.e., the combined effect of intrinsic muscle and tendon characteristics, were investigated by applying percutaneous muscle stimulation. This is a frequently used methodology to assess muscle properties irrespective of central neural influences (Binder-Macleod et al. 1995; de Haan et al. 2000; Gerrits et al. 2001; Harridge et al. 1996). We hypothesized that the muscle–tendon complex of the knee extensors would acquire the intrinsic contractile properties of a faster muscle, which would be prevented by the current countermeasure design, conceivably due to the large number of contraction–relaxation cycles during resistive vibration exercise (Blottner et al. 2006). Methods Subjects A total of 16 subjects participated in the present study. All subjects were in good health and were involved in normal physical activity before participation in the large-scale Berlin Bed Rest study (Rittweger et al. 2006). At the start of the study the subjects were randomly assigned to an experimental group or an inactive control group. The experimental group (RVE, n = 8; mean age, height and body mass ± SD: 33.0 ± 1.9 years, 1.84 ± 0.03 cm and 79.5 ± 3.8 kg) participated in a progressive resistive vibration exercise (RVE) training program during the bed rest. The subjects of the inactive control group (Ctrl, n = 8; mean age, height and body mass ± SD: 34.3 ± 2.5 years, 1.82 ± 0.02 cm and 76.8 ± 1.8 kg, respectively) were restricted to bed rest without exercise countermeasure. All subjects were familiarized with the concepts of the experiments, procedures, and the equipment during a familiarization session that was scheduled 3 days prior to the start of bed rest. The local Ethics committee of the Charité, Campus Benjamin Franklin Berlin approved the study and all participants gave their written informed consent. General design All subjects underwent 56 days of strict horizontal bed rest at the Charité Benjamin Franklin Hospital, Berlin, Germany. During the bed rest, the subjects were not allowed to stand up, to lift their trunk in bed more than to 30° of trunk flexion, to move their legs briskly, or to elicit large forces with their legs muscles other than during testing sessions or during training sessions. Adherence to this protocol was controlled for by continuous video surveillance and by force transducers in the frames of the bed. The diet was balanced using the Harris–Benedict equation and ingestion of alcohol or nicotine, excessive doses of caffeine, as well as the regular intake of any drug or medication was prohibited (details in Rittweger et al. 2006). Exercise countermeasure RVE subjects performed resistive exercises on a vibration system that was specifically developed for application under microgravity and bed rest conditions (Galileo Space; Novotec, Pforzheim, Germany). The applied equipment and protocol for countermeasure exercise are described in detail elsewhere (Rittweger et al. 2006). In short, the training device consists of a vibration platform, which is vertically suspended on a trolley. Elastic springs were attached to the trolley for the subjects to attach themselves through belts with their shoulders, hips, and hands. During bed rest, RVE subjects trained in the supine position, two times daily for 6 days/week. In each training session, four resistive exercises were performed in the following order: squats, heel raises, toe raises and explosive squats. All exercises were performed while the platform was vibrated at a frequency of 19 Hz. Vibration frequency was progressively increased during the 56-day bed rest period to ∼26 Hz at the end of bed rest (Rittweger et al. 2006). Experimental set-up Isometric force recordings were made from voluntary and electrically evoked contractions of the knee extensor group of the right leg. Subjects were tested in the supine position using the same equipment as previously described (Mulder et al. 2006). Force signals were digitized using a sampling rate of 1 kHz and stored on disc for immediate and off-line analysis. Torque (N m) was off-line calculated as the product of force and external moment arm. Experimental procedures During bed rest, all subjects participated in seven experimental sessions, which were scheduled at days: 4, 7, 10, 17, 24, 38 and 56, the latter being the last day of the bed rest period. The baseline experiment was conducted on the fourth day of bed rest (BR4) for logistical reasons. Although this prevents us to address rapid initial changes in strength and contractile characteristics of the quadriceps (Berg and Tesch 1996), which might result in an underestimation of the effect of bed rest on the measured parameters, it does not compromise the comparison between groups during bed rest, because the RVE subjects started their exercise training program on the fourth day of bed rest, which was scheduled after the completion of the functional testing conducted on that day. In addition, in this way all experiments were conducted under methodologically similar conditions. That is, at the baseline experiment subjects were already minimally 72 h bedridden, each subject was tested at the same time of day, and subjects of the RVE group were always tested before their morning training session. To minimize potential damage to muscle and tendon structures due to maximal and explosive contractions after a long period of strict muscle inactivity, subjects started each experimental session by performing a standardised warm-up set that consisted of eight to ten unloaded dynamic contractions (right leg not yet strapped to the force transducer), followed by eight sub-maximal isometric contractions at 70° of knee flexion. Further force recordings were obtained at the individually determined pre-bed rest optimal knee flexion (either 60° or 70°, whereby a knee flexion angle of 0° corresponds to full knee extension; see Mulder et al. 2006 for details). Subjects were asked to perform two to three maximal voluntary contractions (MVC) of 2–4 s in duration. Attempts were interposed with 2 min of rest. After this procedure, the quadriceps muscle was stimulated through two self-adhesive surface electrodes (model 283100, Schwa-Medico, Nieuw Leusden, The Netherlands) of 80 mm × 130 mm. The stimulation intensity was progressively increased until 40% of the MVC torque was obtained during a 700 ms tetanic contraction at 150 Hz (Mulder et al. 2006). The quadriceps muscle was then electrically stimulated with trains of the same single pulse intensity at low (7 pulses at 10 Hz) and high (24 pulses at 300 Hz) stimulation frequency. The trains were applied in this order and interposed with 1 min of rest. Following the electrically evoked contractions, the stimulation electrodes were removed and the skin over the lateral vastus muscle was re-prepared for the positioning of a high-density surface EMG system (HD-sEMG, Active One, BioSemi Inc., Amsterdam, The Netherlands). The system consisted of 130 densely spaced skin-surface electrodes, arranged in a rectangular 10 × 13 matrix with 5 mm inter-electrode distance (Blok et al. 2002). Before mounting the grid to the skin, the skin was scrubbed with alcoholic pads and slightly rubbed with electrode paste. Prior to each test, the skin-electrode impedance was checked and, if necessary, the skin was re-prepared. The grid was positioned over the distal (third), anterio-lateral part of the right vastus lateralis muscle, such that the columns of 13 electrodes were aligned parallel to the muscle fibre orientation of the muscle and with the motor endplate zone around the centre of the columns of the grid (Mulder et al. 2007). The pre-amplified 130 monopolar signals (referenced to the patella) were bandpass filtered (16–400 Hz) and simultaneously AD-converted (16 bits with a resolution of 1 μV/bit at a rate of 2 kHz/channel). Data were stored on hard disk for subsequent off-line processing. With the sEMG system properly positioned, each subject performed another MVC of 2–4 s. If the torque deviated more than 5% from the highest torque attained during the previous MVC-task (i.e., the highest value of the attempts before the application of electrical stimulation), another attempt was made. Lastly, each subject performed three fast (explosive) isometric contractions to assess the neural control of these contractions, as well as the initial rate of torque development during these attempts. Subjects were instructed to contract “as fast and forcefully as possible” on a given signal from the test leader (3-2-1 “Go”). Subjects were required to reach a minimum of 80% of the current MVT and to maintain torque at the highest attained level for approximately 1 s, i.e., ‘kicks’ were always disqualified. Attempts with an initial countermovement (identified by a drop in the torque signal exceeding 1 N just before the onset of torque development) were also disqualified (Aagaard et al. 2002; de Ruiter et al. 2004). Attempts were interposed with 2 min of rest. Data analysis Fast voluntary isometric knee extensions The neural activation of muscle fibres at the start of a contraction greatly determines the performance of specific voluntary motor functions, such as the rate at which muscle force develops during fast and forceful voluntary isomeric contractions (de Ruiter et al. 2004). Even so, the maximal rate of isometric torque development did not differentiate subjects according to their ability to generate high neural activation levels at the very start of the contraction. Instead, the time torque integral, calculated as the area under the time torque curve over the first 40 ms after the onset of torque development, showed to be more sensitive to the initial level of neural activation (de Ruiter et al. 2004). Based on these findings, we calculated the voluntary time torque integral (vTTI40) as an estimate of maximal isometric tension development under voluntary command (Fig. 1). The onset of torque development was thereby defined as the point at which the torque curve exceeded baseline torque by more than three standard deviations. Fig. 1Voluntary torque (thick line) and rectified surface EMG of the vastus lateralis muscle (thin line) time traces, obtained from a representative subject during an isometric voluntary knee extension performed as fast and forcefully as possible. The arrow and diamond at time 0 ms indicate the start of voluntary torque development. The shaded area under the voluntary torque trace reflects the vTTI40, calculated the first 40 ms after onset of torque development. The horizontal bar indicates the 40 ms immediately preceding the onset of torque development (i.e., from −40 to 0 ms) for which the root mean square (RMS) of the surface EMG was calculated. Subsequently, both vTTI40 and RMS−40–0 were normalized to the steady state maximal isometric knee extension condition at the day of testing Voluntary neural activation during the fast isometric knee extensions was assessed by averaging the amplitude (based on root mean square, RMS) of the monopolarly recorded sEMG signals (Fig. 1) over 40 ms before the onset of torque development (RMS−40–0). Unlike a single bipolar recording, the HD-sEMG system allowed for the assessment of monopolar recordings, and allowed for the spatial selection of the grid column with the highest amplitude. The latter was based on the mean of all electrodes within one column. To investigate how neural activation related to the relative maximal rate of torque development for each individual, vTTI40 values were corresponded to RMS−40–0 data, whereby each accepted contraction of each session was included. For the evaluation of both RMS−40–0 and vTTI40 as a function of bed rest duration, only the data of one single fast contraction per session were incorporated into the analyses. The contraction with highest level of EMG activity was selected for this purpose. Since we were interested to compare the neural activation during fast voluntary contractions with the neural activation during voluntary contractions where torque is developed more slowly, the contractile and electromyographic data were normalized to the steady-state maximal isometric knee extension condition at the day of testing. These data are also part of another study and are reported elsewhere (Mulder et al. 2007). Electrically evoked contractions For each experimental session, the peak torque (T) attained during the 10 Hz (T10) train was expressed relative to the maximal tetanic torque reached during the 150 Hz tetanus (T150), i.e., expressed as a T10/T150 ratio. The force profiles of the 10 Hz tetanus showed clear oscillations (Fig. 2). The force oscillation amplitude (FOA) relative to the mean force was also calculated and used as a measure of the degree of force-fusion (Gerrits et al. 1999; Vøllestad et al. 1997). Fig. 2Methods used for evaluation of contractile properties of the quadriceps femoris muscle evoked by electrically evoked muscle stimulation at 10 Hz. The torque elicited at 10 Hz stimulation was first expressed as a percentage of the maximal torque evoked at 150 Hz (T150). T10 was the peak value of the 10 Hz torque trace. The force oscillation amplitude (FOA) was determined by expressing the mean amplitude of the torque oscillation (Os) between the fourth and seventh stimulus as a fraction of the mean torque (Tm) during this time The contractile characteristics at high pulse frequency stimulation (i.e., 300 Hz) were quantified by assessing the time to peak tension from the start of the evoked contraction (TPT300). The start of contraction was defined as the instant the first pulse of the 80 ms train was delivered. Half-relaxation time was determined for each session as the time needed for the elicited torque to decay to half the maximal value following the last pulse of the 150 Hz tetanus (HRT150). Force data were filtered using a fourth-order, 50 Hz low-pass filter. This filter was found not to affect the course of torque development; it only removed high-frequency noise from the signal. Statistical analysis Values are expressed as mean ± SE (standard error of the mean). Independent-samples t tests were used to determine whether the absolute values of variables related to muscle strength, neural activation and contractile properties of the quadriceps femoris muscle differed at baseline (BR4). Changes in muscle strength and contractile properties after 56 days of bed rest were assessed by means of linear regression, and expressed as a percentage change with respect to the value at BR4. One-sample t tests were used to determine whether the normalized slope (slope/intercept on y-axis) of the linear regression was significantly different from zero. Independent-sample t tests were used to determine whether the groups differed in their response to bed rest. Pearson’s correlation coefficients were calculated to establish significance of correlation. The level of significance was set at P < 0.05. Results Fast voluntary isometric knee extensions Data of neural activation and steady state isometric strength are presented elsewhere (Mulder et al. 2006). Briefly, whereas neural activation remained unaltered for Ctrl, maximal voluntary isometric steady-state strength significantly declined after 56 days of bed rest by about 17%. For the RVE group neural activation increased by approximately 30%, whereas voluntary steady-state torque was maintained. The ability to perform fast voluntary contractions showed a substantial variability during bed rest. There was considerable variation in vTTI40 as well as in RMS−40–0 during these contractions, both within one session as well as across sessions. Nonetheless, significant positive linear relationships (P < 0.01) between RMS−40–0 and the vTTI40 were obtained for 12 of the 16 subjects in the present study (4 RVE subjects and 8 Ctrl subjects). Significant Pearson’s correlation coefficients (r) ranged from 0.52 to 0.90; with a median value of 0.74. When individual data were pooled per group, each group had a significant relationship between normalized EMG amplitude and normalized vTTI40 (RVE, r = 0.572; P < 0.01; Ctrl, r = 0.758; P < 0.01). However, as can be seen in Fig. 3, for neither group did RMS−40–0, or vTTI40 change during the course of the bed rest. Fig. 3Mean values (±SE) of the voluntary time torque integral over the first 40 ms (TTI40) after torque development (vTTI40; a) and the sEMG amplitude 40 ms before the onset of torque development (RMS−40–0; b) obtained during 56 days of bed rest (BR). For each session, both vTTI40 and RMS−40–0 are expressed as a percentage of the corresponding values at the steady-state maximal voluntary torque (MVT) of that session Intrinsic contractile properties The contractile properties obtained from electrically evoked contractions during bed rest are shown in Fig. 4. The course of FOA during bed rest was significantly different for Ctrl and RVE subjects (P < 0.05). For Ctrl, the FOA increased substantially (by 47.5 ± 10.0%, P < 0.01) during the bed rest study (Fig. 4a). In contrast, no change in the FOA was observed for RVE. The groups also differed in their response to bed rest with respect to contraction time as indicated by TPT300 (P < 0.01). TPT300 declined significantly by 6.8 ± 0.9% (P < 0.001) for Ctrl, whereas no changes were observed for RVE (Fig. 4b). Furthermore, in the Ctrl group the HRT150 tended to be somewhat shortened by 2.5 ± 1.1% (P = 0.056) after 56 days of bed rest, whereas for RVE it again remained unaltered (Fig. 4c). Group differences in HRT150 did, however, not reach significance. Interestingly, the T10/T150 ratio increased for both Ctrl (by 15.0 ± 5.5%; P < 0.05) and RVE (by 40.6 ± 17.1%; P < 0.05) and these changes were not significantly different between groups (Fig. 4d). Fig. 4Mean values (±SE) of the force oscillation amplitude (FOA; a) time to peak tension at 300 Hz stimulation (TPT300; b), half relaxation time at 150 Hz stimulation (HRT150), and peak torque at 10 Hz stimulation expressed as a fraction of the maximal torque obtained during tetanic stimulation at 150 Hz (T10/T150) obtained during 56 days of bed rest (BR) Discussion Contractile characteristics of electrically evoked knee extensions The rate at which muscle force develops during a fast voluntary contraction is influenced by both peripheral factors and neural activation properties. In order to separate these influences, the contractile response of the knee extensor group was also assessed by means of percutaneous sub-maximal muscle stimulation. This is a reliable method, which has been used to assess intrinsic muscle characteristics in various human populations (Binder-Macleod et al. 1995; de Haan et al. 2000; Gerrits et al. 2001; Harridge et al. 1996). Although it is not possible to assess the specific characteristics of the portion of the muscle fibre population that is recruited during the sub-maximal stimulation procedure of the present study, the activation of about 40% of the muscle fibres suffices to represent the global contractile characteristics of the quadriceps muscle. Previous studies have shown that the torque frequency relationship, as well as the contractile properties of the twitch are not intrinsically influenced by the absolute force level, provided that a force between 20 and 50% MVC is reached during maximal tetanic stimulation (Binder-Macleod et al. 1995). In the absence of the exercise countermeasure, the knee extensors exhibited characteristics of a faster muscle following 56 days of bed rest. The degree of fusion at 10 Hz stimulation was decreased, the time to reach peak torque at 300 Hz stimulation was reduced and relaxation after tetanic stimulation at 150 Hz tended to be faster. The possibility of an error seems unlikely, since we found complementary changes towards enhanced contractile speed at all frequencies studied. In addition, these changes were observed only for the inactive control group. The stiffness of the series elastic component is amongst the factors known to affect the rate of torque development (Bojsen-Moller et al. 2005). Although not measured in the present study, from previous other studies it appears that with muscle unloading tendon stiffness often decreases (Kubo et al. 2000; Reeves et al. 2005). This would tend to result in a reduced rate of torque development of the whole muscle–tendon complex, opposite to what we observed. The increased rate of torque development, as well as the tendency towards a faster rate of relaxation in the present study could, however, be explained by an elevated rate of cross-bridge cycling (Unsworth et al. 1982). Although at odds with some previous findings (e.g., Davies et al. 1987; Gondin et al. 2004; Narici et al. 2003) such interpretation would be consistent with documented elevations in maximal unloaded shortening velocity (Caiozzo et al. 1994; Widrick et al. 2001; Yamashita-Goto et al. 2001) potentially linked to shifts in muscle fibre phenotype from slow to fast as a direct consequence of muscle unloading (Fitts et al. 2000; Gerrits et al. 2003; Ohira et al. 1999; Talmadge 2000; Trappe et al. 2004). Because the assessed muscle torque in the present study is the resultant of the entire muscle–tendon complex, and given a likely increase in tendon compliance by bed rest, the observed enhanced contractile speed characteristics, might even underestimate the underlying intrinsic changes in the present study. Despite faster contractile properties, torque production at 10 Hz stimulation relative to maximal tetanic stimulation was increased after bed rest. Although consistent with previous research (Seki et al. 2001), the higher relative torque responses at low frequency stimulation did not result from a significant enhancement of twitch summation, as previously opted (Gondin et al. 2004). In contrast, the level of force fusion substantially decreased during bed rest period in the present study. At present, the exact processes responsible for these anomalous findings remain unclear, but a similar phenomenon was reported in the paralyzed muscles of individuals with spinal cord injury (Gerrits et al. 1999), which may be considered as an extreme model for muscle unloading. In part, our data supports the hypothesis postulated by Rittweger et al. (2006) that the large number of contraction–relaxation cycles during resistive vibration exercise (Blottner et al. 2006) may be effective in preserving muscle fibre contractile properties. Indeed, no changes in time to reach peak torque at 300 Hz stimulation, the rate of relaxation after tetanic stimulation at 150 Hz, or the level of force fusion at low stimulation frequency (i.e., the FOA) were observed in the exercise trained subjects. The significant difference between some baseline values, e.g., FOA and T10/T150 ratio (Fig. 4a, d), and the tendencies for TPT300 (P = 0.098) and HRT150 (P = 0.057) to be lower in RVE compared to Ctrl at baseline deserves attention, since it may point towards a difference between groups with respect to muscle fibre type at the start of the study, with RVE exhibiting a faster muscle. Nonetheless, at least for the FOA it has been demonstrated that it is still much higher (i.e., 0.65 in Gerrits et al. 1999) in paralyzed muscles of people with spinal cord injury. This makes it unlikely that the preservation of FOA in the present study resulted from a ceiling effect for RVE. Despite the observation that speed characteristics were unaltered for the exercise-trained group, the relative peak torque at low stimulation frequency also increased for this group. As the level of force fusion remained unaltered during bed rest for RVE, other factors are likely involved. Once possibility is that the peak torque during 10 Hz stimulation increased as a consequence of an enlarged twitch response. Although not directly measured in the present study, the torque developed during the first response of the 10 Hz tetanus increased during the course of the bed rest by about 26%, which was comparable to the elevation in 10 Hz peak torque (∼40%) for RVE. Interestingly, the relative torque production of the first response of the 10 Hz contraction also increased (by about 30%) for the inactive control group. However, the reduced fusion of successive individual twitches diminished the increase in peak torque at 10 Hz stimulation to about 15%. Despite the differences in contractile speed characteristics, the percent change in 10 Hz peak torque after 56 days of bed rest was not different between groups. At present, the selectivity of the exercise countermeasure paradigm to prevent changes in contractile speed properties, but not in the torque response at low frequencies of stimulation, remains difficult to explain. Indeed, more research is needed to determine the effectiveness of resistive vibration exercise as a countermeasure, since the individual merits of resistance training versus vibration training could not be quantified in the present study. Fast voluntary isometric knee extensions at maximal effort Adequate preservation of the rate at which muscle torque develops during a forceful volitional contraction is imperative for astronauts, as neuromuscular deconditioning, coupled to a weakened load-bearing skeleton, increases the risk of fall-related bone fractures after prolonged space missions. To add to the concern, compared to steady-state contractions, much higher levels of neural activation are needed for contractions where torque develops as rapidly as possible (de Haan 1998; de Ruiter et al. 1999). As such, we hypothesised that the ability to perform fast and forceful voluntary contractions would be more deteriorated by bed rest confinement than the ability to perform maximal steady-state contractions. Surprisingly, and at odds with the finding of others (di Prampero and Narici 2003; Koryak 1998), we found no evidence for such bed rest induced functional impairment in either group (Fig. 3). For the same inactive control subjects as used in the present study, we previously reported an absence of neural deconditioning for maximal voluntary steady-state contractions, whether examined by the twitch interpolation technique, or by the assessment of electromyographic activity of the quadriceps femoris muscle (Mulder et al. 2006, 2007). We concluded that the preservation of neural activation of this specific motor task was likely associated with the repeated functional retesting sequence employed during bed rest. The observation that the percent reduction in maximal isometric knee extension strength of the left leg, which was not repeatedly retested during bed rest, exceeded the level of atrophy by a factor of two after 8 weeks of bed rest, strengthened this notion (Mulder et al. 2006). Based on the findings of the present study, we are inclined to suggest that the repeated retesting regime also served as a contributory factor in maintaining neural activation during fast isometric knee extensions. Although such an effect was not intended at the time, this supposition is important since it suggests that neural activation of different isometric motor tasks can be maintained during bed rest without rigorous exercise training regimes. In part, the absence of loss of muscle functionality following bed rest might have resulted from changes in the contractile properties of unloaded muscles. Widrick et al. (1998) reported significant atrophy of single human soleus muscle fibres after 17 days of spaceflight. Absolute peak power of these fibres was, however, partly or fully preserved by an elevated contraction velocity. In the present study, the examined muscle group in the inactive control group also acquired mechanical characteristics of a faster muscle during the course of the bed rest. As selective neural deconditioning could not be demonstrated for the fast voluntary contractions, the expectation might arise of an increased rate of voluntary torque development for the inactive control group. Such a systematic change was not observed. Although it seems difficult to explain this absence based on the changes in intrinsic contractile characteristics, it is clear that neural activation and the subsequent mechanical response during voluntary contractions are much more variable than the activation and response when contractions are electrically evoked. To overcome this variability, larger alterations than those observed in intrinsic contractile characteristics would have been required to allow for detectable changes for the fast voluntary isometric actions. Interestingly, an increase in the rate of torque development could have also been expected for the exercise trained RVE group. For this group we previously found that the amplitude of the surface EMG during the maximal steady-state contractions was substantially increased at the end of bed rest (by ∼30%; Mulder et al. 2007). The latter was suggested to result from an increase in the mean motor unit firing rate, due to a change in the excitability of the alpha motoneurons (Mulder et al. 2007). Such modulation has been associated with increased rate of force development after resistance training (Holtermann et al. 2007). However, in the present study the neural activation during the fast voluntary contractions and the subsequent initial rate of torque development remained unaltered for the exercise trained group. The current exercise-training regime consisted mainly of relatively slow dynamic loaded contractions while vibration was simultaneously applied to the feet. Such a motor task is quite different from the isometric contractions performed during the testing of the subjects. Based on the selectivity of training, and the fact that the number of exercise training sessions by far outweighed the number of testing sessions (89 vs. 7), it is likely that neural activation strategies employed by the subjects were more guided towards optimal performance during the training sessions, than optimal performance during the fast voluntary isometric actions during testing. In conclusion, in the subjects who were confined to 8 weeks of bed rest without preventive measures the knee extensor muscle group acquired intrinsic contractile properties of a faster muscle. Resistive vibration exercise proved effective to counteract these changes at the muscle level. An unexpected finding of the present study was that neither group showed deterioration in the capacity to maximally activate the knee extensors at the very start of a voluntary contraction performed as fast and forcefully as possible. For the RVE group this might indicate an effective countermeasure design. However, considering that neural activation and voluntary muscle function were also maintained in the Ctrl group, it is also conceivable that the multiple retesting of the subjects resulted in or at least contributed to these preservations.

Cancer_Immunol_Immunother-4-1-2335290

Local therapy of cancer with free IL-2

This is a position paper about the therapeutic effects of locally applied free IL-2 in the treatment of cancer. Local therapy: IL-2 therapy of cancer was originally introduced as a systemic therapy. This therapy led to about 20% objective responses. Systemic therapy however was very toxic due to the vascular leakage syndrome. Nevertheless, this treatment was a break-through in cancer immunotherapy and stimulated some interesting questions: Supposing that the mechanism of IL-2 treatment is both proliferation and tumoricidal activity of the tumor infiltrating cells, then locally applied IL-2 should result in a much higher local IL-2 concentration than systemic IL-2 application. Consequently a greater beneficial effect could be expected after local IL-2 application (peritumoral = juxtatumoral, intratumoral, intra-arterial, intracavitary, or intratracheal = inhalation). Free IL-2: Many groups have tried to prepare a more effective IL-2 formulation than free IL-2. Examples are slow release systems, insertion of the IL-2 gene into a tumor cell causing prolonged IL-2 release. However, logistically free IL-2 is much easier to apply; hence we concentrated in this review and in most of our experiments on the use of free IL-2. Local therapy with free IL-2 may be effective against transplanted tumors in experimental animals, and against various spontaneous carcinomas, sarcomas, and melanoma in veterinary and human cancer patients. It may induce rejection of very large, metastasized tumor loads, for instance advanced clinical tumors. The effects of even a single IL-2 application may be impressive. Not each tumor or tumor type is sensitive to local IL-2 application. For instance transplanted EL4 lymphoma or TLX9 lymphoma were not sensitive in our hands. Also the extent of sensitivity differs: In Bovine Ocular Squamous Cell Carcinoma (BOSCC) often a complete regression is obtained, whereas with the Bovine Vulval Papilloma and Carcinoma Complex (BVPCC) mainly stable disease is attained. Analysis of the results of local IL-2 therapy in 288 cases of cancer in human patients shows that there were 27% Complete Regressions (CR), 23% Partial Regressions (PR), 18% Stable Disease (SD), and 32% Progressive Disease (PD). In all tumors analyzed, local IL-2 therapy was more effective than systemic IL-2 treatment. Intratumoral IL-2 applications are more effective than peritumoral application or application at a distant site. Tumor regression induced by intratumoral IL-2 application may be a fast process (requiring about a week) in the case of a highly vascular tumor since IL-2 induces vascular leakage/edema and consequently massive tumor necrosis. The latter then stimulates an immune response. In less vascular tumors or less vascular tumor sites, regression may require 9–20 months; this regression is mainly caused by a cytotoxic leukocyte reaction. Hence the disadvantageous vascular leakage syndrome complicating systemic treatment is however advantageous in local treatment, since local edema may initiate tumor necrosis. Thus the therapeutic effect of local IL-2 treatment is not primarily based on tumor immunity, but tumor immunity seems to be useful as a secondary component of the IL-2 induced local processes. If local IL-2 is combined with surgery, radiotherapy or local chemotherapy the therapeutic effect is usually greater than with either therapy alone. Hence local free IL-2 application can be recommended as an addition to standard treatment protocols. Local treatment with free IL-2 is straightforward and can readily be applied even during surgical interventions. Local IL-2 treatment is usually without serious side effects and besides minor complaints it is generally well supported. Only small quantities of IL-2 are required. Hence the therapy is relatively cheap. A single IL-2 application of 4.5 million U IL-2 costs about 70 Euros. Thus combined local treatment may offer an alternative in those circumstances when more expensive forms of treatment are not available, for instance in resource poor countries. Systemic IL-2 therapy of cancer Rosenberg and co-workers were the first to treat cancer with IL-2. They showed that IL-2 renders white blood cells cytotoxic in vitro. These cells were coined Lymphokine Activated Killer cells (LAK cells). Rosenberg and co-workers treated transplanted lung metastases in mice. Injection of IL-2 plus LAK cells clearly reduced the number of metastases in the lung [82]. Next these authors treated 25 consecutive patients with different types of advanced cancer for whom no effective treatment was available. In 11 patients, objective responses were obtained after treatment with IL-2 and LAK cells [81]. Obviously this paper attracted worldwide attention, as it was a breakthrough in immunological treatment of cancer. In further studies, systemic application of IL-2 with or without LAK cells appeared useful in patients with metastasized renal cell carcinoma and metastasized melanoma. In both conditions about 20% objective responses were obtained; that is about 6% complete and about 14% partial tumor regressions [57]. Repeated cycles of systemic IL-2 administration were required to achieve systemic tumor-inhibitory effects. Grande et al. [46] recently reviewed the therapeutic effects of systemic IL-2 therapy. Systemic IL-2 application required bolus injections of IL-2 given every 8 h at a dose of 105 IU/kg body weight for at least 5 days. These high doses were very toxic, as systemic IL-2 therapy induces a generalized vascular leakage syndrome [7, 83]. In addition the preparation and application of LAK cells was cumbersome indeed. Supposing that the tumor infiltrating leucocytes have to be stimulated by IL-2, we and other groups decided to focus on local IL-2 therapy, that is IL-2 application at the site of the tumor (peritumoral = juxtatumoral, intratumoral, intra-arterial, intracavitary infusion, or inhalation). Since local IL-2 application requires smaller doses of IL-2 than systemic treatment, less complications were expected. Locally applied relatively small doses of IL-2 leads to much higher IL-2 concentrations at the site of the tumor and to much lower concentrations elsewhere in the body. Conclusion The development of systemic IL-2 therapy was a break-through in cancer immunotherapy. We supposed that the drawbacks of systemic IL-2 application might be overcome by local IL-2 application. Effectiveness of local versus systemic IL-2 therapy Subject The therapeutic effectiveness of local versus systemic IL-2 application. Data Silagi et al. [86] studied mice with melanoma or sarcoma using combined cyclosphosphamide and IL-2. Cyclophosphamide was always applied systemically. Cyclophosphamide alone had no therapeutic effect but combined treatment of cyclophosphamide and IL-2 showed a synergistic effect. When either tumor was implanted s.c. at day 0, and IL-2 treatment was given at the site of the tumor beginning 1–3 days later, 87–100% of the mice were cured. Only 35–50% were cured when IL-2 was administered i.p. Conversely, with i.p. treatment of i.p. tumors, 60–83% of the mice were tumor-free on day 50, as compared with only 17% with s.c.treatment. Vaage [94] tested the therapeutic effects of 12 daily injections of 100–300,000 U of human IL-2 against the syngeneic, immunogenic mammary carcinoma MC2 implanted s.c. into C3H/He mice. A local therapeutic effect was observed after injecting tumors even with doses as low as 300 U of IL-2 per injection. Systemic IL-2 treatment however required at least 5,000 U per injection for obtaining discernable results. Belardelli et al. [10] treated mice with s.c. transplanted highly metastatic Friend leukemia, non-metastatic Friend leukemia, RBL-5 lymphoma, and HeJ16 fibrosarcoma. In all these tumor models, peritumoral injections of IL-2 were more effective in inhibiting tumor growth than systemic treatment. Anderson et al. [2] studied the therapeutic effects of IL-2 in C57BL/6 mice with MCA-106 sarcoma pulmonary metastases. When mice were treated once daily with free IL-2 on days 5, 6, and 7 after tumor inoculation, the intrathoracic route was superior to the i.p. or s.c. routes. Maas et al. [68] treated DBA/2 mice bearing i.p. and s.c. SL2 lymphoma. If IL-2 was injected i.p., then in 70% of the mice i.p. tumors regressed completely, and in 50% tumors regressed completely both i.p. and s.c. When mice had only a s.c. growing tumor and IL-2 was injected i.p., then in only 7% of the mice tumors regressed completely. Dubinett et al. [33] transplanted s.c. Line 1 alveolar carcinoma cells in Balb/c mice. On the third day following tumor implantation, mice received injections of IL-2 twice daily, either by i.p. or intratumoral injection, 5 days/week for 3 weeks. Intratumoral injection of IL-2 significantly reduced the tumor volume, increased the median survival time, and resulted in a 23.5% cure rate. However no long-term survivors were among the i.p. treated mice. Jacobs et al. [55] investigated human patients with nasopharyngeal carcinoma. Local IL-2 treatment was combined with standard irradiation therapy. Sixty three % of the patients showed a disease free survival during the next 5 years, whereas only 8% of the controls treated with irradiation alone were still disease free after 5 years. Interestingly, these results contrast with those of Chi et al. [23] who applied single modality systemic IL-2 therapy. They concluded that no response was observed. So, apparently, local IL-2 application is more effective than systemic application. It is important that these results of Jacobs et al. and Chi et al. were obtained in spontaneously occurring tumors, not with transplanted tumors. Local IL-2 should even preferably be given intratumorally as IL-2 application adjacent to the tumor is far less effective. Jacobs et al. [56] compared the effect of peritumoral and intratumoral IL-2 therapy in mice with s.c. transplanted SL2 lymphoma. Intratumoral IL-2 was significantly more effective than peritumoral IL-2. Krastev et al. [61] treated 16 patients with various gastrointestinal tumors with intratumoral and/or intraperitoneal IL-2 therapy. Six patients had a clinical response. All six belonged to the group of seven patients who received intratumoral therapy. No objective responses were obtained in patients treated only with intraperitoneal IL-2. In an early review Bernsen et al. already concluded that locoregional IL-2 treatment was more effective than systemic treatment [13]. Since then an overwhelming amount of research has corroborated this conclusion. Conclusion The data show that local IL-2 therapy and in particular intratumoral IL-2 application is more effective than systemic IL-2 therapy. Local IL-2 therapy leads to systemic therapeutic effects even curing metastatic disease Subject Obviously a major problem of cancer therapy is inducing regression of metastases. Immunotherapy of cancer is an attractive concept, as systemic immunity may indeed cure metastases. Consequently, many groups have concentrated on the systemic therapeutic effects of local IL-2 therapy. Data Maas et al. [67] have shown that i.p. injections of IL-2 can cure DBA/2 mice with a large burden of i.p. transplanted and greatly disseminated SL2 lymphoma. This implied that tumor metastases can be successfully treated with local IL-2 therapy. In more detailed studies SL2 tumor cells were injected in mice both i.p. and s.c. on the flank resulting in i.p. and s.c. tumors [67]. About 50% of the mice treated i.p with IL-2 rejected both the i.p. tumor and the large distant s.c. tumor. In contrast, similar i.p. treatment cured only 7% of the mice bearing only a s.c. SL2 tumor. Thus, it was shown that IL-2 can induce systemic tumor rejection when injected at the site of tumor growth. This SL2 tumor rejection was specific, as mice that were rejecting i.p. and s.c. SL2 lymphoma did not reject P815 mastocytoma. Vaage [95] tested the therapeutic effects of IL-2 against intramammary implants of an immunogenic, syngeneic C3H mammary carcinoma. Peritumoral injected IL-2 had almost equal local and systemic therapeutic effects, whereas systemically injected IL-2 was not therapeutically effective at all. Immunity also explains the long lasting absence (>54 months) of bladder carcinoma recurrences in two IL-2 treated patients who previously had a 7 and 11 years’ history of recurrent bladder cancer [29]. Van Es et al. [96] showed that peritumoral injections of IL-2 in a transplanted rabbit carcinoma model can induce complete regression of the treated tumors as well as untreated contralateral tumors in four out of 12 rabbits. Also metastases in the draining regional lymph nodes of both treated and untreated primary tumors regressed in three of these animals. So, local treatment of a tumor led to systemic effects curing untreated tumors at a distant site as well as metastases in the draining lymph nodes of both the treated and the untreated tumors. A second challenge with tumor cells of the cured animals was rejected. Jacobs et al. [56] transplanted in mice tumors at two different sites. Rejection of the intratumoral IL-2 treated tumors was stronger than rejection of the untreated tumors. Systemic immunity is also likely in patients with nasopharyngeal carcinoma treated with radiotherapy and intratumoral IL-2 application [55]. Addition of IL-2 to the standard radiotherapy reduced the number of loco regional and distant recurrences. In line with these observations is the finding that local or generalized effector dysfunction of the immune system can be reversed by IL-2 exposure in patients with advanced cancer [70]. Conclusion Local IL-2 therapy can also cause systemic therapeutic effects, probably due to immune reactivity [68, 96]. Therapeutic effects of local IL-2 therapy against transplanted tumors in laboratory animals (Tables 1, 2) Subject For studying the potential therapeutic effects of anticancer drugs one has to start in animal models using transplanted tumors, as well-performed and well-interpreted studies in animals have predictive value regarding the therapeutic effectiveness of a drug in human cancer patients [32]. Table 1Successful therapeutic effects of local IL-2 application on transplanted tumors: results published by other groupsTumorsSpeciesRoute of IL-2 applicationTherapeutic effectReferencesCarcinomasMammary carcinomasMousePeritumoral+[94, 95]Transitional cell carcinomaMouseIntratumoral+[87]Lung carcinomaMouseIntratumoral++[33]Prostate carcinomaRatIntratumoral+[50]SarcomasMC-induced sarcomasMousePeritumoral+[16–18]HeJ fibrosarcomaMousePeritumoral++[10]MCA106 sarcomaMousePeritumoral+[2]LymphomaRBL5 lymphomaMousePeritumoral+[10]MiscellaneousMyeloma X5563MouseLocal site++[69]MelanomaMouseTumor site++[86]Friend leukemiaMousePeritumoral+[10]HPV associated tumorMouseTumor site+[21]We have included data of all papers that we have found on therapeutic effects of local application of free IL-2 to transplanted cancer in animalsA well-recognized problem with a Table like this one is that usually positive results are published, in contrast to negative results. So, there is a positive publication bias.The message of Table 1 is not that all transplanted tumors are sensitive, but that positive therapeutic effects have been obtained in many different models and many different tumor types+ Clear therapeutic effect++ Many cures were obtainedTable 2Therapeutic effect of local IL-2 application on transplanted tumors: results from our groupTumorOrgan /typeSpeciesPredictive modelRoute of applicationTherapeutic effectReferencesCarcinomasLine 10LiverGuinea pig+Intratumoral+[6, 72]M8013BreastMouse  Peritumoral+[58]MOTTeratomaMouse Intratumoral-[14]MC-ASNBreastMouse+Peritumoral±[73]MC-BCBreastMouse+Peritumoral±[73]MC38ColonMouse+Peritumoral+[64]5D04StomachMouse Intratumoral- VX2 carcinomaHead and NeckRabbit+Peritumoral+[96]Recently arisen carcinomasX5BreastMouse Intratumoral±[42]X6BreastMouse Intratumoral±[42]X9BreastMouse Intratumoral±[42]SarcomaMecho FibrosarcomaMouse Intratumoral+ LymphomasSL2LymphomaMouse+Intratumoral++[12, 67, 68, 70]L5178YLymphomaMouse Intratumoral++[67]L1210LymphomaMouse Intratumoral+ RBL5LymphomaMouse Intratumoral+ EL4LymphomaMouse Intratumoral− MiscellaneousTLX9ThymomaMouse Intratumoral− P815MastocytomaMouse+Intratumoral++[68]This Table presents all results obtained with models tested by our group; so there is no selection bias (in contrast to Table 1)± therapeutic effects just measurable, + therapeutic effects are clear; ++ animals can be cured from extensive tumor loadsThe four negative models were only tested in a few experiments, until we were convinced that therapeutic effects were not obtained in our hands. Three of the four negative models were not published for obvious reasons. On the other hand, SL2 lymphoma in syngeneic DBA/2 mice is our standard model since the 1989 paper of Maas et al. [68]. Numerous studies were performed with this model. The therapeutic effect of local IL-2 application was always clear Data As early as in 1983, Bubenik and co-workers [18] established that peritumoral injections of rat lymphoid IL-2 suppressed or markedly inhibited the growth of methylcholanthrene-induced sarcomas in syngeneic mice. An equally effective inhibition of murine sarcoma transplants in syngeneic recipients could be obtained with crude lymphoid rat IL-2, with purified IL-2 of murine lymphoid origin, and with molecularly homogeneous human recombinant IL-2 [16–18]. Maas et al. [67] injected DBA/2 mice i.p. with SL2 cells and 10–14 days later these mice were treated with i.p. IL-2 injections. At the time of IL-2 injection the transplanted SL2 tumor had greatly expanded by growth, infiltration and metastasis. A mouse of 25 gram developed in 10 days a tumor load of at least 5 g, about 5 × 109 tumor cells. Nevertheless about 25% of these mice were cured by IL-2. This was a very important step forward in immunotherapy of cancer. The therapeutic effects were dose-dependent. These data were confirmed in more elaborated studies by Bernsen et al. [12] and by Everse et al. [37]. Tables 1 and 2 summarize the therapeutic results of local IL-2 application to transplanted tumors. Table 1 displays the results obtained by other groups and Table 2 results by our group. Both Tables show that local IL-2 therapy may be effective in a broad range of tumor types such as carcinomas [6, 33, 42, 50, 58, 64, 72, 73, 87, 94, 96], sarcomas [10, 16–18, 86], a myeloma [69], lymphomas [10, 66, 67], leukemia [10], a mastocytoma [68], and HPV associated tumor [21]. Obviously not all cancers are sensitive to IL-2 therapy. Our group studied the effect of local IL-2 therapy in 19 models of transplanted tumors (Table 2). In 15 models positive therapeutic results were obtained. IL-2 applied in breast cancers was only moderately effective. In four models no therapeutic effects were obtained, namely murine 5D04 stomach carcinoma [27], murine MOT teratoma [14], murine EL4 lymphoma [27] and murine TLX9 lymphoma [27]. Why some tumors are sensitive to local IL-2 therapy and other tumor models do not show any response, remains hitherto an enigma. The data summarized in Tables 1 and 2 demonstrate that local IL-2 therapy has the capability to destroy tumor cells and to cure the hosts. An animal model using transplanted tumors has predictive value for the therapeutic effect in human cancer if the model tumor comprises more than 1% of the body weight of the host and if this tumor is metastasized [32]. Eight models mentioned in Table 2 fulfil these requirements, viz. the models marked in column 4 with +. So, the positive therapeutic results obtained with these models predict that local IL-2 therapy can be therapeutically effective in human patients with metastatic cancer. Conclusion In models with transplanted tumors in laboratory animals there is overwhelming evidence showing the therapeutic effect of local application of free IL-2. This therapy is effective against a broad range of tumors. In addition the magnitude of the therapeutic data (for instance [67]) suggests that this form of therapy can also induce objective therapeutic responses in human cancer patients. N.B. Not every transplanted tumor or tumor type is sensitive to local IL2 application. For instance transplanted EL4 lymphoma or TLX9 lymphoma are not sensitive in our hands (Table 2). Local IL-2 tumor treatment in veterinary patients (Table 3) Subject Positive results in veterinary patients with spontaneous cancer are an important intermediate between experiments with transplanted cancer in laboratory animals and clinical application in human cancer cases. If therapeutic effects are positive in well-performed, well-interpreted experiments with transplanted cancer in laboratory animals as well as in spontaneous cancer in veterinary patients, then one can be almost certain that such a therapy will also be effective in human cancer patients. Table 3Therapeutic effect of local IL-2 application veterinary cancer patientsType of cancerHostTherapeutic effectReferencesBovine ocular squamous cell carcinomaCattle++[30, 31, 84, 89, 90]Vulval papilloma and carcinoma complexCattle+[51]SarcoidsHorses+[88]FibrosarcomaDogs+[100]This Table summarizes to our knowledge all published studies on local application of free IL-2 in veterinary cancer patients+ Detectable effect++ Many CR cases Data Table 3 summarizes all published studies that we know on local IL-2 application in veterinary cancer patients. Bovine Ocular Squamous Cell Carcinoma (BOSCC) BOSCC originates in the cornea, the third eyelid (membrana nictitans), the lower or the upper eyelid. Ultimately the tumor covers the whole eye. It also metastasizes to the draining lymph nodes. This tumor occurs frequently in tropical countries with intense solar radiation, particularly at high altitudes. BOSCC is for various reasons a very useful veterinary tumor model: The tumor is readily visible and can be directly treated with peritumoral or intratumoral IL-2 injection in field studies. In The Netherlands BOSCC is a very rare disease. So our first tests of IL-2 sensitivity of BOSCC were performed in only five cows with BOSCC [84]. The results showed that BOSCC can be sensitive to local IL-2 therapy. In Zimbabwe about 10% of the cows are affected by BOSCC. This allows large-scale studies [31]. Added to this BOSCC causes an enormous economical burden. In our most extensive study in Zimbabwe [86] we treated 174 BOSCC cases with tumor areas ranging from 20 to 2,800 mm2. Peritumoral injections of various doses of IL-2 were applied during 2 × 5 days. Nine months after treatment, the daily doses of 5 × 103, 2 × 104, 2 × 105, 5 × 105, 1 × 106, 2 × 106 U IL-2 had induced complete tumor regression in 82, 81, 56, 15, 44, and 35% of the animals, respectively. In the control animals the tumors had completely regressed in only 14% of the cases. After 20 months the comparable figures were 55, 52, 58, 50, 69 and 52%, respectively, and there had been no change in the control group. The tumors on the third eyelid and limbus were the most responsive [89]. Similar results were obtained in other studies [30, 90]. Even large BOSCC tumors of up to 66 mm can regress completely by local IL-2 therapy [31]. Bovine Vulval Papilloma and Carcinoma Complex (BVPCC) BVPCC is a common disease in Bos taurus breeds of cattle kept at high altitude with high levels of solar radiation in Africa [19, 51]. It also occurs in other countries and continents with similar high levels of solar radiation. Burdin [19] originally described the pathogenesis of this neoplasm. Hill et al. [51] adapted this description to develop a useful system for clinical staging of the tumors. In Zimbabwe BVPCC occurs in about 10% of the cattle. It causes much animal suffering. In addition BVPCC usually proceeds to a more advanced stage. So BVPCC forms a real economical burden for the farmer and on a national scale. Twenty three papillomas and carcinomas of the bovine vulva were treated with local IL-2 therapy. Sixteen partial remissions and three complete remissions add to a tumor reduction in 83% of the treated cows [Stewart et al. to be published]. Remissions were striking in papillomas with a massive lymphocytic infiltrate in particular in those epithelial areas that showed marked dysplasia or (pre-)malignant changes. Sarcoids Sarcoids are fibro-epithelial skin tumors of horses, donkeys, and mules. Infiltrative growth is prominent but they seldom metastasize. After surgical removal they usually recur. Sarcoids were treated by intratumoral IL-2 injections for 5 or 10 days. There were 36 and 50% objective responses, respectively after 12 months [88]. Fibrosarcomas in dogs Tumors often occur in cats and dogs as owners care for pets even into advanced age as members of the family. Like ageing humans they show a large variety of tumors in their later years. Preliminary data demonstrated that fibrosarcomas in dogs are sensitive to local IL-2 therapy [100]. Conclusion Local IL-2 therapy can be effective against spontaneous veterinary tumors. Local IL-2 application has an enormous economical impact as BOSCC as well as BVPCC occur in about 10% of cattle in Zimbabwe and probably also in other tropical countries. Local IL-2 treatment of BOSCC leads to CR in the majority of the cases, and led to tumor reduction in the majority of BVPCC cases. For ethical reasons specific immunity cannot be tested in veterinary patients. However specificity and systemic effects have been shown in mice [68] and rabbits [96] with transplanted tumors. Effective local IL-2 treatment greatly improves the quality of life by reducing suffering. These therapeutic effects in veterinary cancer patients make a strong case for the development and acceptance of local IL-2 therapy with free IL-2 in human cancer patients. Results of local IL-2 tumor treatment in human cancer patients (Table 4) Subject The final step in the experimental chain is the local application of free IL-2 to human cancer. Many research groups have pioneered in this field. We now summarize and discuss the results. Table 4Therapeutic effect of local IL-2 application on human cancer patientsType of cancerNumber of patientsCRPRSDPDReferencesBasal cell carcinoma12831–[59]Bladder carcinoma633––[79]Bladder carcinoma51–4–[53]Bladder carcinoma143–11–[44, 45]Bladder carcinoma99–––[41]Bladder carcinoma108–2–[29]Gastro-intestinal cancer16–6–10[61]Hepatocellular carcinoma5–131[85]Lung cancer; pleural effusions2176–8[71]Melanoma265858[35]Melanoma23155–3[80]Mesothelioma21–4710[43]Mesothelioma2211137[5]Mesothelioma31161014[22]Mesothelioma11–––[62]Ovarian carcinoma3563719[34]Ovarian carcinoma1763–8[91]Neoplastic effusions1446–4[65]Total numbers28878655392 Percentage (%)10027231832 CR complete regression, PR partial regression, SD stable disease, PD progressive disease Data Basal cell carcinoma Kaplan et al. [59] treated basal carcinoma of the skin. A total of 12 tumors were treated in eight patients. Overall response rates were : complete response in 8 of 12 treated tumors, partial response in 3 out of 12 treated tumors, stable disease with no improvement in one tumor site. Bladder carcinoma Pizza et al. [79] obtained tumor regressions after intralesional injections of IL-2 in bladder cancer. Repeated injections of IL-2 under cystoscopic control resulted in complete regression of the tumor in three out of six patients and in partial regression of another three patients. Huland and Huland [53] obtained histologically confirmed complete remission lasting more than 6 months in one out of five patients with urinary bladder carcinoma after continuous IL-2 perfusion of the bladder for 5 days. Gomella et al. [44, 45] treated 14 patients with superficial bladder carcinoma. Patients were treated first with transurethral resection leaving a marker lesion, followed by intravesical IL-2 instillation. There were three complete responses, one lasting more than 9 months. Ferlazzo et al. [40, 41] treated superficial bladder cancer cases with intravesical infusions of IL-2. This gave similar clinical results as obtained by vesical instillation with BCG after transurethral resection. Den Otter et al. [26, 29] treated patients with recurrent bladder carcinoma stage T1, grades 1 to 2 with incomplete transurethral resection leaving a marker tumor of 0.5–1.0 cm. Two days after resection IL-2 was instilled for 2 h. Patients were asked to turn over regularly in order to ensure maximal exposure of IL-2 to the bladder wall. This procedure was repeated on five consecutive days. Two months later the effect was measured by cystoscopy examination. In eight out of ten patients the marker tumor had regressed completely. Four patients were still tumor free after 30 to 54 months. In one patient with a 7-years’ history of bladder cancer requiring 23 cystoscopies, the marker had only partially regressed after 2 months. After removal of the remainder of the marker this patient was tumor free during the follow-up of 54 months. Also, a patient with an 11 years’ history of recurrent bladder cancer remained tumor free during the whole follow-up period [29]. The finding that these patients remained tumor free for >54 months suggests that these patients were (locally) immune to the tumor after tumor regression. Grasso et al. [47] treated 27 patients with transitional bladder carcinoma Ta/T1-G1–2 with intravesical instillations of IL-2 for 1 h during 5 days. After 2 months none of the lesions had disappeared or were clearly reduced. This result seems to contrast with the findings by Den Otter et al. [26, 29]. The different treatment protocols may be essential: Den Otter et al. started with an incomplete TUR followed by IL-2 instillations, whereas Grasso et al. did not perform a TUR before IL-2 instillation. TUR causes tumor cell damage, which may induce tumor immunity. Interestingly, Grasso et al. obtained 33.3% relapses after a median follow-up of 12 months; this contrasted to the restrospective analysis in which the historical recurrence rate per year was 95%. Gastrointestinal tumors Shirai et al. [85] treated five patients with hepatocellular carcinoma with intratumoral IL-2 injections. In two patients 32 and 57% tumor regression was observed. Krastev et al. [61] treated patients with different forms of stage III and IV gastrointestinal malignancies (primary or metastatic) for whom no further treatment options were available. With locoregionally applied IL-2 a modest but clinically worthwhile improvement was obtained in six out of 16 patients; remarkably these six all belonged to the group of seven patients that were treated with intratumoral IL-2. Melanoma Radny et al. [80] treated patients with skin and soft-tissue melanoma metastases with intralesional injection of IL-2. A total of 24 patients with AJCC stage III or IV melanoma and single or multiple skin and soft-tissue metastases were included. IL-2 was administered intralesionally into all cutaneous and soft-tissue metastases accessible from the skin, 2–3 times weekly, over 1–57 weeks. Response evaluation was confined to the intralesionally treated tumors. CR of the treated metastases was achieved in 15 patients, the longest remission lasting 38 months to date. In five patients a PR was achieved and in a further three PD (one patient was not assessable). A total of 245 metastases were treated. There was CR in 209 (85%), and PR in 21 (6%). The therapy was generally well tolerated; the observed adverse events were mainly of grade 1–2 severity. Pfohler and coworkers [78] treated two patients with multiple cutaneous metastases of malignant melanoma with intra and perilesional application of interleukin-2 and achieved complete regression of these metastases. Mesothelioma Goey et al. [43] treated patients with pleural mesothelioma stage I-IIA with continuous daily intrapleural infusion of IL-2. PR occurred in four out of 21 evaluable patients with a median time to progression of 12 months (range 5–37). SD occurred in seven patients with a median time to progression of 5 months (range 2–7). There were no CRs. The median overall survival time was 15.6 months (range 3.0–43). Astoul et al. [5] treated 22 patients with malignant pleural mesothelioma. The response rate was evaluated 36 days after treatment. There were one CR, 11 PR, three SD, and seven PD. The median survival time of responders differed significantly from that of the non-responders (28 versus 8 months). Castagneto et al. [22] treated 31 consecutive patients with unresectable malignant pleural mesothelioma with pleural effusion with intrapleural instillation of IL-2. In 90% of the patients there was no further or minimal asymptomatic pleural fluid collection. Median overall survival was 15 months whereas the expected survival range of patients with involvement of the visceral pleura is 9–12 months. Krastev et al. [62] treated a patient with a large abdominal mesothelioma with intratumoral IL-2 injections and IL-2 instillation in the peritoneal cavity. The tumor regressed completely; the patient was cured and is still healthy and working 6 years after publication. Neoplastic effusions Masotti et al. [71] treated neoplastic pleural effusions in 21 patients with non-small cell lung cancer with intrapleural administration of IL-2. CR was obtained in seven patients and PR in six patients. Lissoni et al. [65] treated 14 patients with neoplastic effusions from a variety of solid tumors. There were four CR and six PR with a median duration of 4 months. Castagneto et al. [22] treated pleural effusions of mesothelioma as described in the previous section. Ovarian carcinoma Edwards et al. [34] treated patients with ovarian carcinoma with infusions of IL-2. Eligibility criteria included six or more courses of prior platinum-based chemotherapy and laparotomy-confirmed persistent or recurrent ovarian cancer. Among 35 assessable patients, there were six laparotomy-confirmed CRs and three PRs. The median survival time of the cohort was 13.7 months and the overall 5-year survival probability was 13.9%. For the nine patients who demonstrated responses, the median survival time had not been reached at the time of publication (range 27 to 90+ months). Taylor et al. [91] treated patients with advanced ovarian cancer with intraperitoneal IL-2. Nine out of 17 patients showed an objective response. Renal cell carcinoma (lung metastases) Huland et al. [52] introduced IL-2 inhalation therapy for lung metastases of renal cell carcinoma. Progressive pulmonary metastases responded dramatically in 15% of the patients for a median of 15.5 months and were stabilized in 55% of patients for a median of 6.6 months. The overall median response duration was 9.6 months. Median survival was 11.8 months; expected survival according to risk analysis was 5.3 months [52]. The PortugeseSpanish Inhaled IL-2 Group [36] studied the effect of inhaled IL-2 on pulmonary metastases of renal cell carcinomas. They found 13.7% Objective Responses (OR), a median progression free survival of 8.6 months and an overall survival of 23 months. Quantification of therapeutic data Therapeutic data of Table 4 were further analyzed regarding CR, PR, SD, and PD. This analysis was possible in 288 cancer patients treated with locally applied IL-2. There were 27% CR, 23% PR, 18% SD, and 32% PD. These data may be too optimistic as positive results are published more frequently than negative results. Toxic effect of local IL-2 application Table 5 shows the toxic effect of local IL-2 application to cancer patients. There were 24 papers with an abstract mentioning toxic side effects. In 20 papers there were no or minor side effects, one paper [41] reports some side effects, and three papers [22, 34, 43] mention more serious side effects. The latter papers have in common that very high IL-2 doses (see Table 5) were used for intrapleural, intraperitoneal or subcutaneous administration. In essence these massive doses act in a pseudo-systemic manner. Much of these massive doses of IL-2 will be absorbed and as a consequence will give generalized effects similar to systemic treatment. Hence, toxicity mentioned in the latter three papers form the exception to the rule that local IL-2 application causes no or minimal side effects. In reference [43] side effects are due to the very high daily doses of 36 × 106 IU IL-2, as this was a Phase I-IIA study with escalating doses of IL-2. In this case IL-2 treatment may give systemic toxicity complications. In paper [22] patients were treated with repeated intrapleural instillations with 9 × 106 IU twice weekly for 4 weeks; in addition, in non-progressing patients 3 × 106 IU IL-2 were administered s.c. thrice weekly for up to 6 months. Obviously such high intrapleural doses are readily resorbed and reach the general circulation. Hence the toxicity (grade 3 fever and grade 3 cardiac toxicity) in 7/31 patients is not surprising. In paper [34] dose-limiting toxicity was seen in patients treated with 7-days’ infusions with the maximum tolerated dose; obviously there was toxicity per definition. Table 5Toxic effects of local IL-2 application on human cancer patientsReference Toxic side effects[59]Local pain, swelling, erythema, in one patient fluelike symptoms[79]No early or late adverse clinical side effects[53]No evidence of side effects[44, 45]Toxicity from IL-2 given intravesically was minimal. One patient malaise for 24 h after each treatment. Two patients developed asymptomatic lower urinary tract infections[41]2/9 patients had hematuria after the end of treatment; one patient had fever (grade I), and seven patients had hypotension (grade I-II)[29]No toxic effects[61]Negligible adverse effects[85]Abstract did not contain toxicity data[71]Treatment of pleural effusions of lung cancer was well tolerated[35]Side effects of treatment were minimal[80]Therapy was generally well tolerated; the observed adverse events were mainly of grade 1–2 severity[43]Intrapleural administration of IL-2 was associated with acceptable toxicity. Dose limiting toxicity was observed at 36 × 106 IU daily, and consisted of catheter infection, fever and flue-like symptoms.[5]Intrapleural administration of IL-2 was well tolerated[22]Patients with malignant pleural mesothelioma received intrapleurally 9 × 106 U IL-2 twice weekly for 4 weeks. In non progressing patients 3 × 106 U IL-2 were subcutaneously administered thrice weekly for up to 6 months. Toxicity (WHO criteria) with intrapleural IL-2 consisted of grade 3 fever in 6/31 patients and cardiac toxicity (failure) grade 3 in one patient.[62]Abstract did not contain toxicity data[34]Significant locoregional dose-limiting toxicity was seen with 7 day infusionsb (including bowel perforation) with 6 × 105 IU/m2/day as the maximum tolerated dose, but catheter infection was the only significant complication seen with 24 hrs infusions[91]Abstract did not contain cytotoxicity data[65]No important toxicity in 14 patients with neoplastic effusions after intracavitary administration of IL-2[36]aInhaled IL-2: most common toxicities were cough (40% of cycles) and fatigue (7%). Grade 1 or 2 toxicities [40]aLocoregional perfusion of the bladder with IL-2 is safe[52]a116 patients. Toxicity associated with exclusive inhalation of IL-2 was local and consisted mainly of cough[93]aIntravesical IL-2 instillation. No evidence of toxicity[97]aRegionally administered IL-2 was well tolerated in patients with advanced squamous cell carcinoma of the head and neckToxicity as described in the abstracts of the papers mentioned in Table 4aToxicity data from Abstracts that did not contain information for Table 4bAlternating continuous 7 day infusions followed by 7-day intervals Dose–response Figure 1 shows the total applied doses of IL-2 and the percentage of objective responses. Results suggest that (a) good therapeutic results can be obtained with low doses of IL-2 (103 to 105 IU); (b) that there is no dose-response effect in the range of total dose of 106 up to 109 IU IL-2. Of course, in this figure there are only a few data for different tumors treated slightly different by different groups. More detailed analysis with homogeneous groups of mice with transplanted SL2 lymphoma in DBA/2 mice showed a dose response effect in the dose range of 5,000–200,000 IU IL-2 given on five consecutive days [12]. It is not surprising that this was not reproduced with the very heterogeneous data in Fig. 1. Nevertheless these data of Fig. 1 are remarkable as total doses of 106–109 IU of IL-2 can result in 20–100% objective responses. We have analysed whether high and low percentages of objective responses are obtained with sensitive and non-sensitive tumor types, respectively. This seems not to be the case as bladder carcinoma, mesothelioma, and ovarian carcinoma were reported as sensitive (>60% OR) as well as non-sensitive (<60% OR) (Table 4; Fig. 1). Another hypothesis is offered by the work of Jacobs et al. [55]. He found that intratumoral IL-2 application is more effective than peritumoral IL-2 application. We therefore further analyzed whether the different therapeutic effects can be ascribed to differences in the localization of IL-2 application. Also this was not the case. Another possibility is that intratumoral IL-2 application leads to more intratumoral edema formation/leakage causing hemodynamic stagnation and additional tumor necrosis. This may stimulate the antitumor response and consequently lead to superior therapeutic effects. Obviously this hypothesis is difficult to study in human cancer patients for logistical and ethical reasons. A more simple explanation of the absence of a dose-response effect is that different tumors have different sensitivity for local IL-2 therapy, e.g., due to different tumor associated antigenicity. Those that are well-responsive can be cured by lower doses, those that are less responsive are (automatically) treated with higher doses. Hence it is impossible to evaluate dose-response effects using different tumors. Fig. 1% Objective response versus total applied dose of IL-2. The data are derived from the abstracts of the papers mentioned in Table 4 Conclusions Local IL-2 therapy can be effective against a variety of human tumors. In some studies results suggest that metastases are cured and that (systemic) immunity is obtained [29]. There is no obvious dose–response with regard to therapeutic effects. Good therapeutic effects can be obtained with total doses of 103 to 105 IU IL-2; these doses cause no or negligible toxic effects. High total IL-2 doses of 108 IU and higher may exert toxic side effects as are described in systemic treatment. Combined therapy modalities (Table 6) Subject Combination of various therapeutic modalities, like surgery, irradiation, chemotherapy, hormone therapy, etc., are standard in tumor treatment. For this reason several groups have also used local IL-2 therapy with free IL-2 combined with other treatment modalities. Table 6Synergism of local IL-2 therapy with other therapeutic modalitiesTherapyTumorHostTherapeutic effectReferencesSurgery FibrosarcomaDogCR for 12 months vs no CRa[100]Radiotherapy LymphomaMouse93% vs. 17% disease free survival; P < 0.0001a[38, 58] Nasopharyngeal carcinoma Man63% vs 8% disease free 5-years’ survival a[55]Cytokine therapyInterleukin-1Friend leukemiaMouse Synergistic effect with IL-2, 60% survival[10]ChemotherapyCyclophosphamideMelanoma Mouse87–91 vs 67% curesa[86]Cis-platinumTerato carcinomaMouse20 vs 0% cures; 50% survival at ca 50 vs ca 25 daysa[14]DoxorubicinM109 adenocarcinomaMouseSynergistic effect with long term survivors[20]IfosfamideTC1MouseSubstantial inhibition of tumor residuab[54]Cis-platinumSarcoidsHorse53% CR, 27% PR vs 10–18% CR, 10–18% PRc[77]ImiquinodMelanoma metastasesMan 40.7% OR[48]aComparison of therapeutic effect of combined therapy and therapy without IL-2bTumor residua left after ifosfamide treatmentcCis-platinum/IL-2 vs IL-2 only Data Surgery Ziekman [100] described cases of incomplete surgical removal of a tumor in veterinary patients, followed by intra-operative local IL-2 application. This led to complete tumor regression for instance in dogs with fibrosarcomas. IL-2 instillation in the bladder to treat bladder tumors seems only effective after a preceeding TUR (see above). Radiotherapy Everse et al. [38] treated mice with s.c. growing tumors with radiotherapy combined with IL-2 therapy. The combined therapy was more effective than radiotherapy or IL-2 therapy alone. Similar results were obtained by Jűrgenliemk-Schulz et al. [58]. Early nasopharyngeal carcinoma is not clinically apparent and hence these tumors are often diagnosed in an advanced stage. The standard treatment is irradiation, often complemented with chemotherapy. In The Netherlands usually WHO type 1 nasopharynx carcinoma occurs. Type 1 is very therapy resistant. That is, only about 10% of the patients treated with irradiation have a 5 years’ tumor free survival [55]. In this study tumors of 10 patients were irradiated daily for 7 weeks. In weeks 2, 4, and 6 the tumors were (transnasally) injected with low doses of IL-2 from Monday through Friday. The 5 years’ tumor free survival was 63% [55]. Cytokine therapy Belardelli et al. [10] described that combined treatment with IL-1 beta and IL-2 produced a synergistic antitumor effect: 60% of mice injected with highly metastasized Friend Leukemia Cells survived. T cells appeared to be essential for IL-1/IL-2 therapy. Chemotherapy Enk et al. [35] treated melanoma patients with pulmonary metastases with dacarbazine and concurrent inhalation of IL-2. The patients’ condition previously had progressed on chemotherapy, predominantly on dacarbazine based regimens. Five of the 27 patients experienced a complete pulmonary remission. Eight patients achieved a partial pulmonary remission, and five patients experienced stabilization of the disease. Four of the five patients with a complete response and seven of the eight patients with a partial response were previously treated with dacarbazine and progressed. The complete responses all persisted during a follow-up of 12 months, whereas patients with partial responses or stable disease progressed when IL-2 was discontinued. Silagi et al. [86] found synergism between cyclophosphamide and IL-2 in the treatment of mice with sarcoma or melanoma. Bernsen et al. [14] treated mice bearing Murine Ovarian Teratocarcinoma (MOT) with cis-platinum and IL-2. The effect of the combined treatment was greater than either therapies alone. Similarly in horses with sarcoids (a characteristic equine fibro-epithelial tumor of the skin) the therapeutic effect of cis-platin and local IL-2 was greater than the therapeutic effect of either therapies alone [88]. Indrova et al. [54] studied mice with HPV-16 associated tumors. Peritumoral IL-2 administration could substantially inhibit growth of tumor residua after chemotherapy with cyclophosphamide or ifosfamide derivatives. Cabanes et al. [20] treated mice bearing M109 pulmonary adenocarcinoma with a combination of liposomal doxorubicin and free IL-2. Both the tumor and the IL-2 application were i.v. or both were i.p. There were synergistic effects with long-term survivors. Green et al. [48] treated melanoma cases with multiple cutaneous and/or subcutaneous metastases. Metastases were treated with imiquinod daily for 4 weeks before the introduction of intralesional IL-2. This was injected up to three times a week. 182 lesions in a total of 10 patients were treated. A clinical response was seen in 50.5% of the lesions with 40.7% being CR. Furthermore patients with clinically responding cutaneous lesions experienced a marked slowing of the appearance of new lesions. No cutaneous lesions that responded reappeared on cessation of the treatment. Imiquimod alone is often sufficient to elicit a response in purely cutaneous lesions. The addition of intralesional IL-2 however greatly increased the response rate in subcutaneous lesions, and in otherwise refractory cutaneous melanomas. Conclusion Combined therapy of locally applied IL-2 and surgery, radiotherapy or chemotherapy may lead to a synergistic therapeutic effect. The durable complete responses of melanoma metastases in the lung after combined chemotherapy and IL-2 suggested that the metastases were eradicated [35]. Mechanism of tumor regression due to local IL-2 application Original hypothesis on the mechanism of IL-2 in tumor destruction The original hypothesis about the function of systemically applied IL-2 in cancer therapy was that IL-2 induced Leucocyte Activated Killer cells (LAK cells; [15, 81, 82]. LAK cells were thought to invade the tumor and to kill tumor cells.The following data are incompatible with the cytotoxic LAK-cell hypothesis or any kind of cytotoxic cell hypothesis:Hardly any intravenously injected LAK-cell reaches the tumor [9].In the DBA/2-SL2 lymphoma model it is impossible to induce cytotoxic LAK cells [67].Yet IL-2 therapy is very effective in this model.No general correlation was established between in vitro sensitivity to the cytolytic activity of LAK cells and the antitumor effects observed in vivo [10].Various types of leucocytes may dominate the tumor infiltrate after IL-2 application: macrophages, T cells, NK cells, NKT cells, plasma cells, neutrophilic, eosinophilic cells etc., [70]. The assumption that all these different types of cytotoxic cells are directly induced by IL-2 in different tumors is too complex to be true. In contrast IL-2 should induce a similar mechanism against a wide range of (sensitive) cancers.In addition, when SL2 lymphoma is growing both i.p. and s.c., and if IL-2 is injected i.p. at day 10 after tumor transplantation, then in about 50% of the cases the s.c. tumor regresses completely at about day 17, this in spite of the tumor containing only 0–2% leucocytes [70]. Furthermore, effector/target cell ratios of cytotoxic T-cells are usually very low (in the order of 1:50). It is very unlikely that such a huge tumor is rejected by these few leukocytes. So, cytotoxic cells are not the major cause of tumor regression in this experiment [68].As this hypothesis was at least at some points wanting to explain the mechanism of the antitumor activity of IL-2, we started to study the histopathology of IL-2 induced tumor regression as study of the histopathology of the IL-2 induced reaction would be essential. IL-2 induced histopathology in normal tissue [60] The study of the histopathological events during IL-2-induced tumor regression proved to be extremely difficult. The breakthrough came by serendipity. Professor Hennink and his group at the Faculty of Pharmacy, Utrecht University, are developing slow release systems (microspheres), among others for IL-2. To study the biocompatibility of these microspheres, we injected microspheres loaded with IL-2 subcutaneous in rats [60]. We discovered that the microspheres could be stained with PAS and thus these stained microspheres marked the IL-2 release site. This possibility of a precise localisation of IL-2 was a great tool to further analyse the function of the released IL-2. In a series of sections obtained during the IL-2 induced process we could trace a sequence of reactions. First a localized edema developed, related to swelling of the endothelial lining of the capillaries. After 3 days necrosis of the connective tissue was observed in the center of this edematous area. About a week later there were three zones around this necrotic area. Near the necrotic border was a zone of mixed inflammatory cells. The second zone was a large area of edematous connective tissue. The outer zone showed angiogenesis, a massive proliferation of macrophages around the newly formed blood vessels and also an influx of lymphocytes. About 3–5 week later these macrophages invaded the necrotic material showing features of phagocytosis. Finally also plasma cells and occasionally eosinophils appeared in the peri-necrotic inflammatory area. The plasma cells indicate the induction of an immune reaction. In short two main phases could be ascertained in this process:an almost immediate marked edema due to leakage of the local blood vessels;a later and more complex reaction consisting of the induction of angiogenesis, a macrophage reaction, migration of the infiltrate into the seminecrotic tissue and finally clearance of tissue in combination with granulomatous processes.Next we again studied the histopathology of the mechanism of tumor regression induced by free IL-2. Of course the histopathological events after injection of IL-2 in or around a tumor are far more complex than the histopathological events after subcutaneous injection of IL-2 loaded microspheres in normal tissue in rats. But the events after injection of IL-2 at the site of tumors were similar to those after injection of IL-2 loaded microspheres in normal tissue. It is remarkable, however, that IL-2 induced tumor regression may be fast, requiring only some days, or slow, requiring several months, or intermediate. These differences can be related to histopathological differences of the tumor tissues. Before we describe these differences we have to pay attention to some hemodynamic features in tumors. Some hemodynamic features in tumors In tumors the interstitial fluid pressure is higher in the center of the tumor mass than in surrounding tissue due to vessel leakage [98]. The increased vascular leakage in tumors is explained as follows:The tumor vessels usually lack a continuous basal membrane [26].Endothelial injury in ischemic tissue [49].Tumors have poor homeostatic control of the circulation due to lack of smooth muscle and lack of innervation [25].This vascular leakage causes increase of interstitial pressure. This pressure will cause vascular compression, particularly of the post-capillary venules. This leads to stagnation of the blood flow and vascular distension of the prestenotic vessels and sometimes thrombosis in smaller blood vessels. Stagnation of blood causes necrosis, in particular in the tumor center. A peripheral rim of vital tumor tissue often survives as it is just outside the area of deficient blood circulation. Local IL-2 application into/near the tumor appeared to induce additional edema in and around the tumor, just as in normal tissue. Increased edema exerts extra pressure that causes further stagnation of the blood flow and of the lymph drainage. Obviously, this leads to tumor necrosis and thrombosis within only a few days. This early edema is illustrated with photographs [8, 24, 60, 64]. Fast tumor regression of well-vascularized tumors (Fig. 2) [8] It is remarkable that IL-2 induced tumor regression is very quick in the case of (very) fast growing tumors like SL2 lymphoma and P815 mastocytoma [67, 68]. Similarly, relatively fast growing cases of mammary carcinoma [73] or BOSCC are more responsive to IL-2 than slower growing BOSCC [91]. In fast growing tumors the intratumoral vascular tree is relatively poorly developed and also the endothelial lining of the microvessels is abnormal. Local IL-2 application will further stimulate vascular leakage. So we assume that much fluid will leak from the tumor vessels, leading to edema, microthrombosis and extensive hemorrhages in the tumor. It is noteworthy that soon after IL-2 injection the tumor becomes firmer, indicating swelling of the tumor mass by edema. This leads to extensive necrosis. Fig. 2Mechanism of tumor regression caused by local IL-2 application of cancer Of course, this large necrotic tumor mass will induce an acute clearance reaction to remove this necrotic debris, which causes an early reduction in size of the tumor mass. This induces a marked immune reaction. It is also noteworthy that this tumor necrosis is accompanied by marked angiogenesis. Adjacent to many of the newly formed smaller blood vessels there is a cuff of proliferating macrophages, which ultimately move into the necrotic debris. As different tumors differ in tumor associated antigen make-up, they will induce also different types of immune reactions, dominated by macrophages, T cells, plasma cells, or eosinophilic cells. This phenomenon can be observed in biopsies taken late at the margin of the disappearing tumor. As a result of the increased tumor immunity isolated tumor strands may be enveloped by granulomatous inflammatory tissue destroying tumor cells. Specific antitumor immunity induced by IL-2 has been described by Maas et al. [67, 68]. After an animal is cured, a second implant of the same tumor is rejected [6, 68, 96]. Our present hypothesis (Fig. 2) seems now most straightforward to explain the fast tumor regression induced by IL-2 in fast growing tumors: The primary function of locally applied IL-2 is the induction of vascular leakage. This leads to an acute massive tumor necrosis and clearance of necrotic tumor material. As a consequence of massive liberation of antigenic tumor material an immune reaction develops. The different types of dominant cytotoxic cells (CTL, eosinophils, macrophages etc.,) depend on the character of the tumor-associated antigens released in the necrotic tumor debris. According to this view, maximal therapeutic results will be obtained with maximal edema formation within the tumor. In line with this view is the finding that IL-2 is more effective when injected directly into the tumor than when injected peritumorally [56]. It is interesting that it was originally thought that IL-2 killed tumor cells through induction of cytotoxic LAK cells and that the vascular leakage seemed a nasty side effect [7]. According to our present views, it is just the other way round: vascular leakage causing edema is the primary effect of IL-2; the development of systemic immunity [68, 96] seems the secondary effect. An intriguing point was for many years that in the DBA/2-SL2 model local IL-2 therapy is not effective 1–10 days after tumor cell transplantation, whereas it can be effective after 10 days until about 2 days before the expected death [37, 67]. Similarly, Maekawa et al. [69] found that local IL-2 therapy was only effective when IL-2 therapy was given seven or more days after transplantation of myeloma X5563. This can be explained by the observation that these days are required for developing a corona of angiogenesis around the tumor; this is essential for an effective local IL-2 therapy. Slow tumor regression of less vascularized tumors A number of tumors show a slow IL-2 induced regression. Examples are BOSCC [30, 84] and sarcoids in the skin of horses [88]. This regression pattern more closely follows the histological changes as observed in the model, in which the IL-2-releasing microspheres were deposited in normal subcutaneous tissue [60]. Also in these slowly growing tumors edema will be induced in the tumor, but this is more focal at the site of the injection and does not involve the entire tumor. Also the necrotic focus develops slower and the necrosis is less hemorrhagic. In these tumors the complete development of the IL-2 induced response takes at least some weeks. So, the tumor immunity will develop slower and consequently tumor regression will be slower. Moreover in BOSCC the tumor contains only about 50% of tumor cells; the other 50% is stroma. When a the tumor is reduced in size by 50%, then usually hardly any or no tumor cell is present anymore; the remaining 50% is stroma. After killing of the tumor cells, the final regression of the stroma requires several additional months. Of course both processes (edema and leukocyte infiltration after neoangiogenesis) are always present as a result of local IL-2 application. But in fast-regression models edema is more prevailing, in slow-regression models leukocyte and particularly macrophage infiltration dominate. Solid subcutaneous SL2 tumors expand by infiltrating the surrounding tissue. These tumor strands are the first targets of macrophages, as the tumor strands are in close proximity to surrounding tumor stroma and are progressively destroyed [8]. So, the body of the tumor and the infiltrating tumor strands are destroyed in two different ways, namely by edema due to vascular leakage and a granulomatous antitumor response originating in the tumor surrounding stroma, respectively. Obviously IL-2 may stimulate the existing inflammation as often present in tumors and hence attracts more macrophages and inflammatory cells, and as a consequence intensifies the already existing cellular response to the tumor. Usually, there is in the perivascular zones a marked increase of varied types of inflammatory cells. The reduction of the tumor can be accelerated by using IL-2 in combination with a tumor necrosis inducing agent, as is demonstrated by local cis-platin treatment of mice with MOT tumors [14] and sarcoids in horses [88]. Everse et al. [38] showed that irradiation stimulated the therapeutic effect of IL-2 therapy in mice. We assume that the necrotic material is phagocytosed, thus boosting an immune reaction. A remarkable finding is that IL-2 injected into the primary tumor can induce regression not only of the primary tumor but also the metastasis in the draining lymph node [6, 96]. However, after surgical removal of the primary tumor IL-2 cannot induce regression of the lymph node metastasis [6]. We assume that this depends on the differences between the histological structure of the primary tumors. Usually in primary tumors there is extensive vascularization and leucocyte infiltration and the tumor is in close contact with the surrounding stroma. In contrast, early tumors in lymphatic tissue often have a scanty vascularization and are situated within the lymph nodes without a direct stroma contact. Tumor tolerance and local IL-2 In the previous sections histopathological techniques were used to study the mechanism of IL-2 induced tumor regression. In this section we will also pay some attention to immunological data. Besides stimulating immune responses, IL-2 may also inhibit immune responses. This is clear as deficiency of IL-2 leads to autoimmunity[1]. IL-2 inhibits the immune response by stimulation of CD4+CD25+ T regulatory cells (Treg). These cells suppress immune reactivity including antitumor immune responses [3, 75]. The presence of antigen [77] and IL-2 are important in the maintenance of CD4+CD25+ regulatory T lymphocytes. Depletion of CD4+CD25+ Treg is important for tumor rejection [99]. Depletion of Treg with anti-CD25 antibody stimulates tumor rejection [76]. Because of this tolerance inducing role of IL-2 through Treg, some authors have suggested to use IL-15 for cancer immunotherapy. IL-15 is a cytokine that is functionally closely related to IL-2, but it does not induce tolerance [74]. In 1995, the first animal studies for anti-cancer therapy were performed with IL-15. Although IL-15 therapy still lacks therapeutic successes in human clinic, some authors still suggest that IL-15 should replace IL-2 for immunotherapy of cancer [3]. Importantly and in contrast to IL-15, IL-2 is capable of breaking tolerance [4, 11, 39, 70]. This may be crucial for effective anti-cancer immunotherapy. This reversal of tolerance is mediated through the activation of immature dendritic cells [63]. Activation of intratumoral dendritic cells and reversal of tolerance exerted through local regulatory CD4+CD25+ T lymphocytes, could be the link between local effects of intratumoral IL-2 therapy and systemic immunity. Stimulation of systemic immunity by local IL-2 therapy is also suggested by clinical data on cytokines in treated human patients. Pro-inflammatory cytokines (IFN-gamma, IL-5) are more increased after local IL-2 therapy than anti-inflammatory cytokines. In contrast, anti-inflammatory cytokines (IL-10) are increased after systemic IL-2 therapy [92]. These data highlight mechanistic differences between local and systemic IL-2 therapy. The discussion above may lead to the question, whether the specific immune system is involved in IL-2 induced tumor rejection. Pleiotropic effects of high local concentrations of IL-2 could lead to aspecific activation of T cells, activation of NK cells and the generation of LAK cells. Nevertheless, the immune system must be involved as local IL-2 treatment causes systemic immunity, that resides in CD3+ lymphocytes or CD4+ and CD8+ lymphocytes [69, 70]. Conclusions Local IL-2 application seems to induce severe vascular leakage in well-vascularized and fast growing tumors. This leakage in fast growing tumors induces a massive hemorrhagic necrosis which results in early clearance of the tumor tissue. In less-vascular tumor tissue the edema formation is more limited. Neoangiogenesis allows the arrival of especially macrophages that move to the tumor site. This leads to tumor cell killing and a specific immune reaction, resulting in tumor regression. The character of this immune reaction probably depends on the different types of tumor-associated antigens present in the seminecrotic tumor debris. According to our views the primary effects of local IL-2 application to a tumor are the vascular phenomena followed by a host of complex histological events clearing the tumor necrosis. The immune reactivity seems to be a secondary and presumably an indirect effect. Of course, this is an analysis at the histopathological level. The whole process is far more complicated as immune regulatory events––including cells and cytokines are involved. Characteristics of local IL-2 therapy Local IL-2 treatment of cancer often cures also the metastases [67, 68, 96] This systemic effect is probably the result of IL-2 induced systemic immunity [67, 68]. Local IL-2 therapy can cure large local tumors [31, 67, 68, 88] and even large metastasized tumor burdens [6, 67, 68, 96]. Local IL-2 therapy is effective against a large variety of spontaneous carcinomas, melanoma, fibrosarcoma, equine sarcoid, mesothelioma, lung metastases of renal cell carcinoma and melanoma (references see Tables 3, 4). Of course, local IL-2 application is no panacea for all tumors and all types of cancer all the time. Intratumoral IL-2 application is more effective than peritumoral (juxtatumoral) application [55] or application at a more distant site [68]. A single IL-2 application may be sufficient [28]. This can be advantageous during surgery when a tumor cannot be excised completely. Local IL-2 therapy is particularly effective in fast growing tumors [67, 68, 73, 91]. Most of the tumors that well respond to local IL-2 treatment are those with an already (pre-)existing peritumoral leukocytic infiltrate. Local treatment of the primary tumor may result in regression of both the primary tumor and a metastasis in the draining regional lymph nodes [6, 96]. IL-2 induced tumor regression can be fast [24, 67, 68], but it is often a slow process requiring months [30, 31, 88]. The fast tumor regression is often observed in fast growing highly vascularized tumors in which IL-2 causes vascular leakage, leading to edema and intravascular thrombosis followed by necrosis. This is also a consequence of the often abnormal endothelial lining of the intratumoral vascular tree. Slow tumor regression is often observed in poorly vascularized tumors with prominent macrophage infiltration after IL-2 induced neovascularization and an activation of the existing round cell infiltrate. Prior induction of tumor necrosis and local inflammation may accelerate the effect of IL-2. Local IL-2 therapy can exert a synergistic therapeutic effect with surgery, radiotherapy and chemotherapy (references see Table 6). Thus local IL-2 treatment can be a valuable addition to the standard oncotherapy. Local IL-2 has minor side effects and is generally well tolerated (references see Table 5). It can usually be applied without difficulty. It does not require complex technical procedures. Local IL-2 does not hamper or interfere with the standard oncotherapy. The local IL-2 therapy findings suggest to revisit our views on the relation between local and systemic oncotherapy at least as far as IL-2 therapy concerns Systemic chemotherapy/hormone therapy, and later immunotherapy, were initially particularly reserved for patients in which adequate tumor treatment by surgery and radiotherapy was not possible. In most of these cases the tumor was inoperable or disseminated and often the patient was in the terminal phase of the disease. Now with modern therapeutic regimes the indications of systemic treatment have been extended, whereas local tumor treatment is becoming rare. Data described in this paper show that cancer treatment with local IL-2 application can lead to cure of metastases. We stress that there are good reasons to consider IL-2 tumor treatment also in early cancer. Surgical treatment remains the golden standard in malignant tumors. Local IL-2 therapy may be considered in local tumors if these tumors can not be resected completely. Occasionally a previously inoperable lesion might become fit for surgical interventions due to the reduction of the tumor size. In some cases local IL-2 application may be considered as an adjunct to the standard treatment particularly if recurrences are expected. An example of the latter is local IL-2 treatment in TUR-treated bladder tumors. Local IL-2 therapy may also be used as an adjunct to treat a primary tumor showing metastases. In advanced tumor cases local IL-2 treatment of the primary tumor may induce a specific immune response and might also contribute to cure or decrease tumor size and sometimes also may have a positive effect on the metastases. IL-2 does not interfere with the standard oncotherapy and might be an useful adjunct to diminish suffering and expand lifespan. Dose at the tumor site Only about 0.1% of the systemically applied dose reaches the tumor. In other words the same dose injected locally results at the site of the tumor in a 1,000 times higher concentration than a systemically applied dose. Reduced side effects An advantage of local treatment is furthermore that side effects, both local and systemic, are modest if present at all, since the required local dose is only a fraction of the dose required in systemic treatment. No immunologic overflow feed-back Systemic IL-2 therapy causes immuno suppression as indicated by systemic IL-10 production. This is probably the body’s reaction to a generally activated immune system (cytokine storm). Local IL-2 avoids this down-regulating feedback, and leads to systemic increase of pro-inflammatory cytokines, like IFN-gamma and IL-5. Reduced costs of locally applied low doses Systemic chemotherapy and immunotherapy in cows and horses is no option. Cattle and horses are so large that systemic oncotherapy is out of order; e.g., required doses of IL-2 would be very large and hence far too expensive in animal husbandry. This problem can be circumvented by local therapy. The results in cattle and horses show that local IL-2 therapy in these large animals is feasible, readily to apply, economic, and therapeutically effective. Local IL-2 application: doses and duration of the treatment The number of IL-2 injections required for optimal treatment is still a point of debate. A single injection containing 2 up to 16 million U IL-2 is sufficient in the DBA/2–SL2 model [28], whereas doses of only 5,000 U IL-2 were required in case daily injections on five consecutive days were given [67]. However, Kusnierczyk et al. [64] cured C57BL mice with s.c. solid colon carcinoma with multiple IL-2 treatments, whereas no cure was obtained by a single IL-2 injection. Treating spontaneous BOSCC with a single dose of 2 million U IL-2 [28] gave comparable results as daily IL-2 injection of the same dose on ten consecutive working days [30]. For logistic reasons most clinicians (both in human and in veterinary medicine) strongly prefer to apply one single dose instead of daily injections over five or ten consecutive working days. A single large dose can also be appropriate for tumor treatment during surgery when the tumor mass cannot completely be removed. But such a single dose should be much higher (about 4.5 × 106 IU IL-2) than doses on five or ten consecutive working days (ca 5,000 IU per day in mice). Very different IL-2 doses, time schedules and routes of application were used. In a human patient we might consider <106 U IL-2 as low dose, (1–18) × 106 U as intermediate dose, and >18 × 106 U IL-2 high dose for local IL-2 application. But a systemically applied dose is far more toxic than the same dose after local application. In addition, the same dose given i.p. in a mouse can be regarded to be high, whereas this dose is low when it is given intravenously in a human patient. A systemically applied dose of 106 U IL-2 is intermediate. But if this dose is given daily for 100 days, then it is a large total dose. Nowadays we usually treat cancer with a single local injection of 4.5 × 106 U IL-2. This dose is based on data obtained in mice, cattle and dogs [28]. It may be astounding that in local treatment the same dose is effective in a mouse as well as in cattle. In fact similar local processes have to be induced in these different species. In systemic application however the dose required by a cow has to be about 20,000 fold of that used in mice. We preferably inject IL-2 inside the tumor. This leads to better therapeutic results than peritumoral IL-2 application [56]. Website Our website on local IL-2 therapy is available at http://cancerimmunotherapy.net/ We provide information targeted on medical professionals. Readers are invited to co-operation. Please reach us at w.denotter@uu.nl.

Pflugers_Arch-3-1-2082653

Restitution analysis of alternans and its relationship to arrhythmogenicity in hypokalaemic Langendorff-perfused murine hearts

Alternans and arrhythmogenicity were studied in hypokalaemic (3.0 mM K+) Langendorff-perfused murine hearts paced at high rates. Epicardial and endocardial monophasic action potentials were recorded and durations quantified at 90% repolarization. Alternans and arrhythmia occurred in hypokalaemic, but not normokalaemic (5.2 mM K+) hearts (P < 0.01): this was prevented by treatment with lidocaine (10 μM, P < 0.01). Fourier analysis then confirmed transition from monomorphic to polymorphic waveforms for the first time in the murine heart. Alternans and arrhythmia were associated with increases in the slopes of restitution curves, obtained for the first time in the murine heart, while the anti-arrhythmic effect of lidocaine was associated with decreased slopes. Thus, hypokalaemia significantly increased (P < 0.05) maximal gradients (from 0.55 ± 0.14 to 2.35 ± 0.67 in the epicardium and from 0.67 ± 0.13 to 1.87 ± 0.28 in the endocardium) and critical diastolic intervals (DIs) at which gradients equalled unity (from −2.14 ± 0.52 ms to 50.93 ± 14.45 ms in the epicardium and from 8.14 ± 1.49 ms to 44.64 ± 5 ms in the endocardium). While treatment of normokalaemic hearts with lidocaine had no significant effect (P > 0.05) on either maximal gradients (0.78 ± 0.27 in the epicardium and 0.83 ± 0.45 in the endocardium) or critical DIs (6.06 ± 2.10 ms and 7.04 ± 3.82 ms in the endocardium), treatment of hypokalaemic hearts with lidocaine reduced (P < 0.05) both these parameters (1.05 ± 0.30 in the epicardium and 0.89 ± 0.36 in the endocardium and 30.38 ± 8.88 ms in the epicardium and 31.65 ± 4.78 ms in the endocardium, respectively). We thus demonstrate that alternans contributes a dynamic component to arrhythmic substrate during hypokalaemia, that restitution may furnish an underlying mechanism and that these phenomena are abolished by lidocaine, both recapitulating and clarifying clinical findings. Introduction Correlations between beat-to-beat alternations in electrocardiographic QT interval (QT or T-wave alternans) and cardiac arrhythmia have been described in classical work [13, 25] and subsequently confirmed in both clinical studies [2, 28, 38] and experimental models [7, 33, 52]. Indeed, the presence of such alternans provides a stronger predictor of arrhythmic risk than is offered by signal-averaged electrocardiography and a prediction of a similar value to that offered by invasive programmed stimulation procedures in current use [10] (but see [45]). The association between alternans and arrhythmogenicity is particularly marked at high heart rates (short baseline cycle lengths, BCLs) [15, 49] and in conditions of congenital and acquired electrocardiographic QT, and therefore action potential, prolongation [1, 34, 39, 46, 47, 64]. The relationship between heart rate and alternans has previously been analysed by constructing restitution curves which plot relationships between action potential duration (APD) and preceding diastolic interval (DI), as BCL is varied [12]. Clinical and experimental studies then associated alternans and arrhythmia observed at short BCLs with steeply sloping restitution curves with gradients greater than unity [14, 21, 24, 32]. The latter would reflect disproportionate decreases in mean APD with DI. Furthermore, manoeuvres which decrease the slopes of restitution curves, such as hyperkalaemia in a tachycardic canine endocardial preparation [22] and certain cardiotropic drugs in canine and porcine preparations [8, 31, 37], have been reported to both suppress alternans and exert anti-arrhythmic effects. In contrast, other studies have attributed such alternans to alteration in Ca2+ cycling involving the sarcoplasmic reticulum [35, 44, 54]. Whatever the underlying mechanism, alternans results in temporal variations in the wavelengths of propagating action potentials that could potentially lead to reentry and arrhythmogenesis [55]. Hypokalaemia is an important clinical cause of acquired QT prolongation and is associated with arrhythmia of various degrees of spatial organisation [12, 30, 57] suppressible by lidocaine [6, 36, 48]. The present study uses the monophasic action potential (MAP) technique [19, 23] and an established Langendorff-perfused murine model [18, 40–42] to explore the effects of hypokalaemia and of lidocaine on alternans and arrhythmogenicity for the first time in any species. These investigations of temporal heterogeneities in recovery after depolarization are performed on hearts paced at short BCLs, thereby complimenting recent work [40, 41] which contrastingly examined the contribution of spatial heterogeneities at long BCLs. This approach successfully demonstrates such phenomena in parallel with the clinical situation and proceeds for the first time to apply Fourier analysis to MAP waveforms and to recordings obtained during hypokalaemia. This analysis demonstrates evolution from a monomorphic to a polymorphic pattern as has previously been reported in other species [55, 58, 63], thereby further validating the murine model. The presence or absence of alternans and arrhythmogenicity is then related to alterations in the slopes of restitution curves. Thus, while the relationship between APD and BCL has previously been determined [20], the present study reports restitution curves relating APD to the preceding DI for the first time in a murine system. Use of this murine system further permits the comparison of findings with results obtained from genetically modified examples [26]. Taken together, these novel findings implicate a dynamic component to arrhythmic substrate under conditions of hypokalaemia for which restitution may furnish a possible underlying mechanism. Materials and methods Experimental animals Mice were housed at 21 ± 1°C with 12-h light/dark cycles, were fed sterile chow (RM3 Maintenance Diet, SDS, Witham, Essex, UK) and had free access to water. Wild-type 129 Sv mice aged 3–6 months were used in all experiments. All procedures complied with the UK Animals (Scientific Procedures) Act 1986. Solutions Solutions were based on bicarbonate-buffered Krebs–Henseleit solution (mM: NaCl 119, NaHCO3 25, KCl 4, KH2PO4 1.2, MgCl2 1, CaCl2 1.8, glucose 10 and Na-pyruvate 2; pH adjusted to 7.4) and were bubbled with 95% O2/5% CO2 (British Oxygen Company, Manchester, UK). Hypokalaemic (3.0 mM K+) solutions were prepared by reducing the amount of KCl added. Lidocaine-containing solutions were prepared by adding lidocaine (Sigma-Aldrich, Poole, UK) to a final concentration of 10 μM. Preparation A Langendorff perfusion protocol previously adapted for murine hearts [3] was used. In brief, mice were killed by cervical dislocation [Schedule 1: UK Animals (Scientific Procedures) Act 1986]. Hearts were then quickly excised and placed in ice-cold bicarbonate-buffered Krebs–Henseleit solution. A short section of aorta was cannulated under the surface of the solution and attached to a custom-made 21-gauge cannula filled with the same solution using an aneurysm clip (Harvard Apparatus, Edenbridge, Kent, UK). Fresh Krebs–Henseleit solution was then passed through 200- and 5-μm filters (Millipore, Watford, UK) and warmed (37°C) using a water jacket and circulator (Techne model C-85A, Cambridge, UK) before being used for constant-flow retrograde perfusion at 2–2.5 ml/min using a peristaltic pump (Watson-Marlow Bredel model 505S, Falmouth, Cornwall, UK). Hearts were only regarded as suitable for experimentation if they regained a healthy pink colour and began to contract spontaneously on rewarming. Electrophysiological measurements An epicardial MAP electrode (Hugo Sachs, Harvard Apparatus) was placed against the basal left ventricular epicardium. A small access window was created in the interventricular septum to allow access to the left ventricular endocardium [5]. A custom-made endocardial MAP electrode comprising two twisted strands of high-purity Teflon-coated silver wire of 0.25-mm diameter (Advent Research Materials, UK) was constructed. The Telflon coat was removed from the distal 1 mm of the electrode, which was then galvanically chlorided to eliminate DC offset, inserted and placed against the septal endocardial surface. MAPs were amplified, band-pass-filtered (0.5 Hz to 1 kHz: Gould 2400S, Gould-Nicolet Technologies, Ilford, Essex, UK) and digitised at a sampling frequency of 5 kHz (micro1401, Cambridge Electronic Design, Cambridge, UK). Analysis of MAPs in both the time and frequency domains was performed using Spike II (Cambridge Electronic Design). Experimental protocol A bipolar platinum stimulating electrode (1 mm inter-pole spacing) was placed on the basal surface of the right ventricular epicardium. Square-wave stimuli (Grass S48 stimulator, Grass-Telefactor, Slough, UK) of 2-ms duration and amplitudes of twice the excitation threshold were initially applied to hearts at a constant baseline cycle length of 125 ms until MAPs showed stable baselines, rapid upstroke phases that reached consistent amplitudes and smooth repolarisation phases [19] and for at least 10 min. Hearts were then exposed to test solutions for 20 min while stimulation was continued before subsequent recordings were made. In initial experiments, MAPs were recorded during regular stimulation at baseline cycle lengths of 130, 100 and 70 ms. Hearts were then subjected to an adapted dynamic pacing protocol [9]. This comprised cycles each consisting of 100 stimuli delivered over a range of BCLs. Steady states were consistently reached during the first 50 responses in each cycle, and thus, mean epicardial and endocardial APD90 values and DIs were calculated from the final 50 action potentials of each cycle. With each successive cycle, BCL was decremented by 5-ms steps from an initial value of 175 ms. Cycles were continued until a reproducible sequence of consistently shaped waveforms was no longer obtained. These data were then used to construct restitution curves. All data are presented as means ± standard errors of the means and include the number of hearts studied. Comparisons between data sets used analysis of variance (significance threshold set at P ≤ 0.05). Curve fitting of particular function to data sets used a Levenberg–Marquardt algorithm (OriginPro 7.5, OriginLab, MA, USA). Results The experiments explored the consequences of increases in heart rate for arrhythmogenicity in hypokalaemic murine hearts, thereby extending a previous report that considered the effect of bradycardia [41]. They thus studied identical experimental groups, first exposing hearts to normokalaemic (5.2 mM K+) and hypokalaemic (3.0 mM K+) test solutions before and after addition of lidocaine (10 μM) for 20 min during regular stimulation at a BCL of 125 ms. Experiments then proceeded to study electrical activity through a range of steady state BCLs to assess propensity to spontaneous arrhythmia. Electrical traces from arrhythmic hearts were then quantitatively analysed in the frequency domain to empirically characterise their kinetics. Finally, hearts were subjected to a dynamic pacing protocol which explored the effect of varying BCLs [21] and permitted the analysis of alternans and its relationship to arrhythmogenicity. Alternans and arrhythmic activity occur in hypokalaemic hearts paced at a reduced baseline cycle length The first series of experiments recorded MAPs [23] from the epicardia of hearts paced at BCLs of 130, 100 or 70 ms over 30-min recording periods (five hearts in each case; Fig. 1). This demonstrated a tendency to both alternans, as reflected in alternating short-long-short sequences in action potential duration, and arrhythmogenesis at the shortest BCLs, corresponding to the highest heart rates, specifically in the hypokalaemic hearts and not in the other groups. Thus, MAPs obtained from normokalaemic hearts whether paced at BCLs of 130, 100 or 70 ms demonstrated consistent waveforms and stable rhythms during 30-min recording periods (five hearts in each case; Fig. 1a). In contrast, while recordings from hypokalaemic hearts paced at BCLs of 130 and 100 ms demonstrated stable rhythms (five out of five hearts), alternans leading to periods of arrhythmic activity occurred during pacing at a BCL of 70 ms in all five cases (P < 0.01 as compared to normokalaemic hearts; Fig. 1b). However, neither alternans nor arrhythmic activity occurred in either normokalaemic or hypokalaemic hearts treated with lidocaine during pacing at any BCL (five hearts in each case; P < 0.01 as compared to hypokalaemic hearts not treated with lidocaine; Fig. 1c and d, respectively), confirming the anti-arrhythmic effect of lidocaine described in earlier studies [6, 41, 42, 48]. Fig. 1Effect of baseline cycle length on presence of alternans and initiation of arrhythmic activity. Epicardial monophasic action potential recordings obtained from hearts exposed to normokalaemic (5.2 mM K+, a) and hypokalaemic (3.0 mM K+, b) test solutions, and normokalaemic (c) and hypokalaemic (d) test solutions containing lidocaine (10 μM) during regular stimulation at baseline cycle lengths of 130 ms (A), 100 ms (B) and 70 ms (C). Vertical lines indicate the timing of stimuli Electrophysiological waveforms after the initiation of arrhythmic activity When arrhythmic activity was observed in those hypokalaemic hearts paced at a BCL of 70 ms, this showed progressive degeneration from a regular monomorphic to a disorganised polymorphic pattern (four hearts), in common with earlier reports from the larger, porcine and human, hearts [16, 58, 63]. This trend was quantitatively apparent in a spectral analysis performed using a fast Fourier transform method [4] on ∼10-s sequences of MAP waveforms recorded from the epicardia and endocardia of hypokalaemic hearts to which a Hanning window had first been applied. These waveforms had been sampled at a rate of 5 kHz with low- and high-frequency filter cutoffs of 0.5 Hz and 1 kHz, respectively. Each transform was performed on 16,384 (=216) points to give a spectral resolution of 0.30 Hz up to a maximum, Nyquist, frequency of 2.5 kHz. Relative root mean square powers were then extracted from the real and imaginary parts of the transforms. Figure 2 thus compares typical waveforms in both the time (a) and frequency (b) domains. Waveforms recorded during the first ∼5 min after the initiation of arrhythmic activity were monomorphic in appearance (A in Fig. 2a). This was confirmed by spectral analysis which similarly demonstrated a single identical dominant frequency peak of 18.9 ± 1.8 Hz in both epicardial and endocardial recordings with harmonics whose frequencies were integral multiples of a dominant fundamental frequency and any given heart showed at least two such peaks (∼38.4, ∼57.6, ∼76.8 Hz, A in Fig. 2b). Fig. 2Progression from monomorphic to polymorphic arrhythmic activity. Epicardial and endocardial monophasic action potential recordings obtained from hearts exposed to hypokalaemic (3.0 mM K+) test solution 5 min (A) and 10 min (B) after initiation of arrhythmic activity in the time (a) and frequency (b) domains In contrast, waveforms recorded ≥10 min after the initiation of arrhythmic activity were consistently polymorphic in appearance (B in Fig. 2a), with spectral analysis (B in Fig. 2b) further demonstrating dissimilar epicardial and endocardial frequency spectra. The dominant frequency peak was then at 27.6 ± 5.6 Hz in the epicardium, while that in the endocardium was 26.5 ± 5.8 Hz. Furthermore, the higher order peaks, whether recorded from the epicardium or the endocardium, occurred at irregular frequency intervals that were not multiples of the dominant frequency. Alternans occurs in the epicardia and endocardia of hypokalaemic hearts paced at a baseline cycle length of 70 ms As illustrated in C in Fig. 1b, hypokalaemic hearts paced at a BCL of 70 ms demonstrated alternans. Figure 3 goes on to show mean APD90 values of the alternating, odd-numbered (filled bars) and even-numbered (open bars) action potentials, and the difference between these values, giving the magnitude of any alternans (hashing), recorded from the epicardia (a) and endocardia (b) under normokalaemic (A) and hypokalaemic (B) conditions and under such conditions during treatment with lidocaine (C and D, respectively). Fig. 3Alternans during stimulation at a baseline cycle length of 70 ms. Durations of successive odd (filled bars) and even (open bars) numbered action potentials (at 90% repolarisation, APD90) and the difference between these values, giving the magnitude of alternans (hashing) in the epicardia (a) and endocardia (b) of hearts exposed to normokalaemic (5.2 mM K+, A) and hypokalaemic (3.0 mM K+, B) test solutions and normokalaemic (C) and hypokalaemic (D) test solutions containing lidocaine (10 μM) during regular stimulation at a baseline cycle length of 70 ms In normokalaemic hearts (five hearts), alternans was not observed, the difference between the mean APD90 values of odd-number and even-numbered action potentials not reaching significance (P > 0.05) in either the epicardium (4.7 ± 2.7 ms) or the endocardium (5.6 ± 1.7 ms). In contrast, alternans did occur in hypokalaemic hearts (five hearts), the difference between the mean APD90 values of odd-number and even-numbered action potentials being statistically significant (P < 0.01) in both epicardium (14.9 ± 3.8 ms) and endocardium (15.6 ± 1.8 ms). Treatment with lidocaine had no significant (P > 0.05) effect on normokalaemic hearts (five hearts): The difference between the mean APD90 values of odd-number and even-numbered action potentials remained statistically insignificant (P > 0.05) in both epicardium (2.7 ± 3.4 ms) and endocardium (2.5 ± 4.6 ms). In contrast, treatment with lidocaine eliminated alternans in hypokalaemic hearts (five hearts). Hence, the difference between the mean APD90 values of odd-number and even-numbered action potentials did not reach significance (P > 0.05) in either epicardium (4.1 ± 4.6 ms) or endocardium (2.4 ± 3.2 ms). The dynamic pacing protocol initiates alternans in hypokalaemic hearts at short baseline cycle lengths The above findings concerning the presence or absence of alternans at the shortest BCL studied above prompted a detailed exploration of the effect of BCL upon the alternans phenomenon. To assess this relationship, hearts were subjected to a dynamic pacing protocol [21] consisting of cycles each lasting 100 stimuli at through a range of BCLs. In these procedures, steady states were consistently reached within the first 50 responses. Accordingly, mean values of epicardial and endocardial APD90 and DI were obtained from the final 50 action potentials. BCL was decremented in 5-ms steps with each cycle from an initial value of 175 ms until a reproducible sequence of consistently shaped action potential waveforms could no longer be obtained. Figure 4a–d shows typical recordings obtained from the epicardia of individual hearts under the four sets of experimental conditions during the dynamic pacing protocol at BCLs of 170, 130, 90, 85, 80, 75 and 70 ms when such steady states had been achieved. Figure 5a–d shows the corresponding relationships between the durations of successive odd- and even-numbered action potentials obtained from these same hearts during the dynamic pacing protocol over the full range of BCLs studied. Thus, alternans was not observed under normokalaemic conditions even at the shortest BCL studied (Fig. 4a), resulting in points falling on the line y = x (Fig. 5a). Conversely, alternans occurred under hypokalaemic conditions at short BCLs (Fig. 4b), resulting in such points falling below the line of equality (Fig. 5b). In contrast, alternans did not occur in either normokalaemic or hypokalaemic hearts during exposure to lidocaine (Fig. 4c and d), again resulting in points falling on the line of equality (Fig. 5c and d). Fig. 4Effect of baseline cycle length on action potential waveforms. Epicardial monophasic action potential recordings obtained from hearts exposed to normokalaemic (5.2 mM K+, a) and hypokalaemic (3.0 mM K+, b) test solutions and normokalaemic (c) and hypokalaemic (d) test solutions containing lidocaine (10 μM) during stimulation at baseline cycle lengths of 170, 130, 90, 85, 80, 75 and 70 ms. Vertical lines indicate the timing of stimuliFig. 5 Relationships between durations of successive action potentials over a range of baseline cycle lengths. Relationships between the durations of successive odd- and even-numbered epicardial monophasic action potentials (at 90% repolarisation, APD90) in the same hearts exposed to normokalaemic (5.2 mM K+, a) and hypokalaemic (3.0 mM K+, b) test solutions and normokalaemic (c) and hypokalaemic (d) test solutions containing lidocaine (10 μM). These were obtained at baseline cycle lengths decremented in 5-ms steps from 175 ms until stimulation no longer resulted in a reproducible sequence of waveforms. Filled and open circles indicate data obtained at baseline cycle lengths at which arrhythmogenesis was and was not observed, respectively. Arrow indicates departure of points from the line y = x (broken line) Alterations in the slopes of restitution curves correlate with pro- and anti-arrhythmic effects The findings described above prompted the construction and analysis of restitution curves relating APD90 and the preceding DI given by the difference between the BCL and the preceding APD90 [21, 29, 32, 55]. Alternans and arrhythmia have previously been associated with increases in the slopes of such restitution curves reflecting disproportionate decreases in mean APD90 with DI. Such curves (data points, left ordinate) were plotted using the APD90 (left ordinate) and DI data obtained from the epicardia (circles, Fig. 6A) and endocardia (squares, Fig. 6B) of hearts under these conditions (a–d, respectively). Fig. 6Restitution curves obtained from epicardia and endocardia. Restitution curves plotting action potential duration (at 90% repolarization, APD90) against preceding diastolic interval (DI) obtained from the epicardia (circles, A) and endocardia (squares, B) of hearts exposed to normokalaemic (5.2 mM K+, a) and hypokalaemic (3.0 mM K+, b) test solutions and normokalaemic (c) and hypokalaemic (d) test solutions containing lidocaine (10 μM). Curves are fitted with mono-exponential growth functions obtained by least-squares fitting to the experimental values of APD90 and DI (solid lines, left ordinates). Gradients were obtained by differentiation of the fitted functions (broken lines, right axes). Shaded boxes indicate ranges of DI values at which such gradients exceed unity Under all conditions, both epicardial and endocardial APD90 decreased as DI decreased. However, the slope of this relationship was greater under hypokalaemic (A and B in Figs. 6b) than under normokalaemic conditions (A and B in Figs. 6a). This paralleled both the alternans and arrhythmic activity that were observed under the hypokalaemic conditions. Such increases in the slopes of restitution curves were observed even when the DI was so short such that stimuli were delivered before 90% repolarisation was achieved. This would tend to result in the underestimation of APD90, and hence, the underestimation of the slopes of restitution curves. Exposure to lidocaine had little effect on the slopes of the corresponding curves obtained under normokalaemic (A and B in Figs. 6c), but markedly decreased the slopes of curves obtained under hypokalaemic conditions (A and B in Figs. 6d), in precise agreement with its corresponding effects in suppressing alternans and arrhythmia. Restitution curves yield parameters predictive of arrhythmogenicity The restitution curves obtained above proved amenable to quantitative analyses that has been described on previous occasions. For example, the simplest, mono-exponential growth function that has been used to describe data of this kind [13] takes the form: where y represents APD90, x represents DI and y0 and A and τ are constants obtained by least squares fitting to the experimental values of APD90 and DI in each case. Such curve-fits are used merely to empirically describe the data and should not be taken to imply any particular underlying mechanistic process. The corresponding gradient is then given by: and assumes its maximum value at the shortest BCL. Figure 6 plots the original APD90 and DI values and superimposes optimisations of Eq. 1 to these data (solid lines, left-hand ordinates). Such optimisations gave reduced χ2 values indicative of better fits to data points obtained under normokalaemic than hypokalaemic conditions. It also shows the corresponding gradients derived from constants obtained from the optimisations (Eq. 2: broken lines, right-hand axes). Shaded regions where present indicate the range of DIs below a critical DI (DI < critical DI) at which these gradients exceed unity, obtained from Eq. 2: Such a phenomenon is associated with an alternans initiated by incomplete recovery of one action potential and the consequent shortening of its subsequent DI. This would result in a disproportionate shortening of the subsequent action potential which would consequently permit its full recovery and a lengthening of the DI that follows, thereby reinitiating the cycle and resulting in a progressive decrease in mean APD90 relative to mean DI with decreasing BCL [29]. The resulting alternans would assume an amplitude determined by the gradient of the restitution curve at the given BCL, whether the latter is determined by altered external pacing rates or local ventricular arrhythmic activity. A BCL that gives a gradient of >1 will then initiate a shortening not only of mean APD90 but also of mean APD90 relative to mean DI. The resulting shortening of action potential wavelength would then predispose to reentry and arrhythmogenesis. These parameters successfully predict the pro-arrhythmic effect of hypokalaemia Figure 7 goes on to show these maximum gradients (a) and values of the critical DI (b) derived from the computed A and τ values in epicardia (filled bars) and endocardia (open bars). Under normokalaemic conditions (Fig. 7A) A and τ took values of 56.3 ± 2.4 ms and 58.4 ± 9.8 ms in the epicardium and 73.8 ± 1.7 ms and 65.2 ± 8.4 ms in the endocardium, with reduced χ2 values from least-squares regression of  = 1.4 and 1.2, respectively. This gave maximal gradients of 0.55 ± 0.14 in the epicardium and 0.67 ± 0.13 in the endocardium and critical DI values of −2.14 ± 0.52 ms and 8.14 ± 1.49 ms, respectively. None of the experimental DI values realised the critical DI in either the epicardium or the endocardium, paralleling the absence of arrhythmic activity (Fig. 2a). Fig. 7Maximum gradients and critical diastolic intervals obtained from restitution curves. Maximum gradients (a) and critical diastolic intervals at which gradients equalled unity (b) obtained from epicardial (filled bars) and endocardial (open bars) restitution curves shown in Fig. 6 under normokalaemic (5.2 mM K+, A) and hypokalaemic (3.0 mM K+, B) conditions and under normokalaemic (C) and hypokalaemic (D) conditions during exposure to lidocaine (10 μM). Asterisks indicate values that are significantly (P < 0.05) larger than those recorded in normokalaemic hearts Hypokalaemia (Fig. 7B) significantly (P < 0.05) increased the slopes of both epicardial and endocardial curves, concurring with the arrhythmogenic findings. Thus, A and τ took values of 166.7 ± 5.63 ms and 29.3 ± 8.29 ms in the epicardium and 133.3 ± 8.0 ms and 29.8 ± 2.1 ms in the endocardium, with values of 8.2 and 2.5, respectively. This gave significantly (P < 0.05) increased maximum gradients of 2.35 ± 0.67 in the epicardium and 1.87 ± 0.28 in the endocardium and significantly (P < 0.05) increased the critical DI values of 50.93 ± 14.45 ms and 44.64 ± 5 ms, respectively. The critical DI was realised by the observed data points in both epicardium and endocardium. Furthermore, the DIs at which this occurred were in full agreement with the corresponding DIs at which alternans was observed. Thus, in the epicardium, alternans occurred during pacing at BCLs of ≤95 ms, corresponding to DIs of 46.7 ± 3.6 ms, while in the endocardium, alternans occurred during pacing at BCLs of ≤75 ms, corresponding to DIs of 44.0 ± 5.7 ms. These parameters also predict the anti-arrhythmic effect of lidocaine Lidocaine had no significant (P > 0.05) effect on these values when applied to normokalaemic hearts (Fig. 7C). Thus, A and τ took values of 81.1 ± 3.05 ms and 56.92 ± 10.29 ms in the epicardium and 104.47 ± 13.69 ms and 97.17 ± 36.22 ms in the endocardium, with values of 5.37 and 10.47, respectively. This gave maximum gradients of 0.78 ± 0.27 in the epicardium and 0.83 ± 0.45 in the endocardium and the critical DI values of 6.06 ± 2.10 ms and 7.04 ± 3.82 ms, respectively, which were not realised even at the shortest BCLs studied. In contrast, treatment of hypokalaemic hearts with lidocaine (Fig. 7D) significantly (P < 0.05) reduced the slopes of both epicardial and endocardial restitution curves, in parallel with its anti-arrhythmic effect. Thus A and τ took values of 109.90 ± 11.95 ms and 37.36 ± 5.97 ms in the epicardium and 231.72 ± 44.70 ms and 27.60 ± 4.44 ms in the endocardium, with values of 6.83 and 15.68, respectively. This gave significantly decreased (P < 0.05) maximal gradients of 1.05 ± 0.30 in the epicardium and 0.89 ± 0.36 in the endocardium and significantly decreased (P < 0.05) critical DI values of 30.38 ± 8.88 ms and 31.65 ± 4.78 ms, respectively. The critical DI was realised only at the shortest BCLs studied and then only in the epicardium, in parallel with the substantial reduction in arrhythmogenicity produced by this drug. Slopes of restitution curves thus precisely correlate with the presence or absence of alternans and arrhythmogenesis under all the conditions explored in the hypokalaemic Langendorff-perfused murine heart. Discussion Electrical alternans, beat-to-beat alternation in action potential duration in turn reflected in alternation in electrocardiographic QT interval (QT or T-wave alternans), is most commonly seen at high heart rates and has been associated with arrhythmogenesis in both clinical [2, 25, 28, 38] and experimental [7, 33, 52] studies. This association is especially marked in situations of preexisting action potential, and therefore electrocardiographic QT, prolongation [39, 46, 47, 49, 64] as might occur in the congenital long QT syndromes. Alternans phenomena have classically been analysed by the construction of restitution curves relating action potential duration to the preceding diastolic interval as heart rate is varied [29]. Restitution curves of slopes greater than unity have been associated with alternations in action potential durations of progressively increasing magnitude [14, 21, 24, 29, 32]. In contrast alternans has more recently been associated with disruption of normal cellular Ca2+ homeostasis leading to alternating short-long-short pulses of release from sarcoplasmic reticular stores [35, 44, 54]. Irrespective of the underlying mechanism, any spatial discordance in either phase or magnitude of alternans between myocardial regions would lead to spatial repolarisation gradients that might cause reentry and arrhythmogenesis [53, 56]. Furthermore, any resulting decrease in action potential wavelength, given by the product of action potential duration and conduction velocity, to less than a critical value, might break up the propagating wave of excitation and lead to reentry and arrhythmogenesis [55]. Relationships between restitution curve slopes, alternans and arrhythmogenicity at high heart rates have been established in canine preparations made to model congenital long QT syndrome type 2 (LQT2) by application of the IKr blocking agent E-4031 [27, 61]. Correspondingly, there is a known association between sudden arousal, presumably resulting in increased heart rate, and arrhythmogenesis in human LQT2 [59]. This association has also been established in clinical hypokalaemia [11]. This common and pathophysiologically important condition similarly results in decreased repolarising K+ currents, together with action potential, and therefore electrocardiographic QT, prolongation [17, 62]. The experiments reported recorded MAP technique [19, 23] from an established intact hypokalaemic murine model [18, 40–42] and explored for alternans and arrhythmogenicity and the relationship between these before proceeding to study the effect of lidocaine upon these phenomena. Such studies of temporal heterogeneities in recovery were made at short BCLs in contrast to recent work [40, 41] which examined spatial heterogeneities at long BCLs. The murine system utilised facilitates the introduction of genetic modifications [26] and may therefore permit future comparisons with models replicating congenital arrhythmic syndromes. Indeed, alternans, in association with arrhythmogenicity, has previously been demonstrated in a murine model of LQT5 [50]. While the murine heart is well established as a model for human disease, it must be noted that important interspecies differences exist. Thus, differences between humans and mice in the kinetics of the ion channels carrying the key repolarising currents (the delayed rectifier and slowly-inactivating delayed rectifier currents, IKr and IKs, in humans and the transient outward current, Ito in mice) result in shorter ventricular AP and the absence of a plateau phase in mice [2]. However, in both species, these key repolarising currents are K+-sensitive [6, 11, 16], making the murine heart well suited to modelling clinical hypokalaemia. Furthermore, in both species, depolarisation is rapid and attributable to the same fast Na+ current (INa,f) [4], making our model well suited to the study of Na+-channel-blocking agents such as lidocaine. While both species share similar differences in AP duration between epicardium and endocardium [7], M cells appear to be absent in the murine heart [1]. Nonetheless, relationships between APD and refractory period are similar between the two species [3, 8], and transmural conduction velocities are almost identical [5, 10]. With these caveats in mind, similarities between human and murine hearts make our murine model well suited for the study of the basic mechanisms of arrhythmogenicity. First, we confirm that alternans is absent in control normokalaemic hearts and that this is associated with the maintenance of stable rhythms even at short BCLs. Secondly, we report alternans in association with arrhythmogenesis at short BCLs for the first time in hypokalaemic murine hearts, in full agreement with previous clinical and experimental studies in other species [2, 7, 33, 38, 52]. Thirdly, we associate such alternans with increases in the slopes of restitution curves above unity [29, 55]. Fourthly, we demonstrate for the first time in any experimental model that exposure to the class I agent lidocaine reduces the slopes of restitution curves and suppresses alternans in association with its anti-arrhythmic effect. Taken together, these observations implicate a restitution mechanism in driving alternans in these hypokalaemic preparations. Furthermore, this extends previous reports that the anti-arrhythmic effects of both class III and IV agents are associated with decreases in the slopes of restitution curves [8, 31, 37]. This contrasts with a previous report that found lidocaine to have no effect on restitution curves obtained using an extrasystolic stimulation procedure or alternans phenomena in a normokalaemic excised canine papillary muscle preparation [43]. Fifthly, we show that such alternans and steeply sloping restitution curves are associated with the temporal evolution of the normal regular pattern of cardiac excitation into arrhythmic activity. The latter, in turn, was processed by a spectral Fourier transform analysis of MAP waveforms performed for the first time in any cardiac system. Arrhythmic MAP waveforms were initially monomorphic, and their corresponding spectra contained a single dominant frequency and integral multiple harmonics identical in epicardial and endocardial recordings. Such monomorphic arrhythmic waveforms subsequently degenerated into polymorphic waveforms whose spectra showed dissimilar epicardial and endocardial dominant frequencies and larger higher order peaks that were not integral multiples of the dominant frequency in full agreement with previous results from the porcine heart [63]. Spectra with similar features have previously been obtained from volume-conducted electrograms and pseudoelectrograms (as opposed to MAPs) from murine hearts during both monomorphic and polymorphic arrhythmic activity [51]. Taken together, these results recapitulate in a murine system arrhythmogenic patterns classically described in clinical situations [16, 58] and implicate a dynamic component to arrhythmic substrate under conditions of hypokalaemia for which restitution may furnish a possible underlying mechanism. Furthermore, they disprove previous suggestions that small hearts are unable to sustain such arrhythmias [9, 60], and thus further validate the murine heart as a model for the study of clinical arrhythmia. The present findings might therefore form a basis for further explorations of the relationships between heart rate, restitution slopes, alternans and arrhythmogenicity in other states of QT prolongation, whether attributable to congenital or acquired factors.

Breast_Cancer_Res_Treat-3-1-2092407

Nuclear morphometric features in benign breast tissue and risk of subsequent breast cancer

Certain nuclear morphometric features measured in breast tumor tissue have been shown to predict the prognosis of breast cancer patients. However, the application of these features to predicting risk of breast cancer development has received little attention. We conducted a case-control study to evaluate nuclear morphometric features in benign breast tissue in association with subsequent breast cancer risk. The study was nested within a cohort of 4,888 women with a histopathologic diagnosis of benign breast disease (BBD) and involved 61 cases and 71 controls, amongst whom there were 53 matched case-control sets. Conditional logistic regression models were fitted to assess various measurements of nuclear size and nuclear shape factors in relation to subsequent breast cancer risk. In multivariate analysis, subsequent breast cancer risk was positively associated with a nuclear shape factor that takes the shortest nuclear axis and the longest nuclear axis into consideration simultaneously (highest quartile versus lowest 3 quartiles: odds ratio = 3.07, 95% confidence limits = 1.61, 5.84). In contrast, there was no alteration in subsequent breast cancer risk in association with nuclear size features and other shape factors. In conclusion, our study results suggest that the shape factor that takes both the shortest nuclear axis and the longest nuclear axis into consideration might be of value to predict subsequent development of breast cancer among women with BBD. Introduction Benign breast disease (BBD), in addition to certain hormonal, anthropometric, and lifestyle factors, is a well-established risk factor for breast cancer [1, 2]. However, BBD comprises a broad spectrum of histological entities [3]. Both epidemiologic and experimental studies suggest that non-atypical and atypical proliferative changes represent successive steps preceding the development of in situ cancer and then invasive carcinoma of the breast [4]. However, only a small fraction of women will eventually develop breast cancer after their diagnosis of BBD [5]. Therefore, it is important to differentiate BBD patients with a high risk of subsequent development of breast cancer from those with a low risk. Our understanding regarding this issue, however, is rather limited, although previous studies have suggested that factors such as type of histological subtype (e.g., atypical hyperplasia), menopausal status, and family history of breast cancer, might modify breast cancer risk among women with BBD [6]. Computerized image analysis and morphometry can quantify a number of nuclear morphometric features such as nuclear size, nuclear shape, and chromatin texture [7]. The evaluation of these features may facilitate the diagnosis and management of breast cancer patients [8–10]. Indeed, certain nuclear morphometric features measured in breast tumor tissue have been shown to predict the prognosis of breast cancer patients [11–15]. Furthermore, a study by Mommers et al. [16] observed that normal breast tissue or usual ductal hyperplasia harbored nuclear morphometric changes that might be used to predict subsequent development of breast cancer. In the study reported here, we conducted a nested case-control study to evaluate whether nuclear morphometric features as evaluated in tissue sections of BBD may be related to the risk of subsequent breast cancer among patients with BBD. Methods Study population The present investigation was undertaken using histological sections from a previous case-control study nested within the cohort of 4,888 women in the Canadian National Breast Screening Study (NBSS) who were diagnosed histopathologically with BBD during the active follow-up phase of the NBSS [17]. The NBSS is a multi-center randomized, controlled trial of screening for breast cancer among 89,835 women aged 40–59 years at recruitment. The design of the NBSS and population characteristics have been described in detail elsewhere [18, 19]. Recruitment took place between 1980 and 1985, and study subjects were followed actively until 1988. Eligibility for the study was restricted to women with no history of breast cancer (in situ or invasive). The NBSS was approved by the appropriate Institutional Review Boards, and the study described here involved the analysis of material and data from that study in accordance with the approved study design. Informed consent was obtained from all study participants. Diagnosis of BBD In the NBSS, women who had clinical or radiologic evidence of breast lesion underwent either a needle aspiration or a biopsy. Diagnosis of BBD was performed by a reference pathologist. Our study was restricted to women who had no evidence of either in situ or invasive breast cancer on their initial surgical biopsy. Women with a history of BBD were not excluded from the analyses. During the follow-up period, we identified 4,888 women with a histopathologic diagnosis of BBD, who were followed up for the subsequent development of breast cancer. Selection of cases and controls Incident cases of breast cancer were ascertained by record linkage with the provincial cancer registries, and death clearance was performed by linkage to the Canadian National Mortality Database [18, 19]. The dates of the linkages varied by province, ranging from late 1988 to early 1991. A total of 16 subjects with ductal carcinoma in situ and 76 subjects with invasive carcinoma were ascertained among the cohort of women with BBD. Potential control subjects were women with BBD who had not developed breast cancer (but were alive at) by the date of diagnosis of the corresponding case subject. Five controls were selected randomly (with replacement) for each case from those non-cases available within strata defined by screening center, NBSS study arm, year of birth (if possible to the nearest year, and mostly within 2 years), and age at diagnosis of BBD. For the study reported here, 61 case subjects and 71 control subjects (including 53 matched case-control sets) were included. Questionnaire Upon enrollment in the NBSS, all participants completed a questionnaire that sought information on demographic characteristics and risk factors for breast cancer, including menstrual and reproductive histories and family history of breast cancer. Morphometry Morphometric measurements were performed on H&E stained slides, using the QPRODIT interactive video-overlay system (Leica, Cambridge, UK). About 50 nuclei were selected in the most representative areas of the slide (selected by a breast pathologist), and their contours were traced manually using a 100× objective (final magnification about 3,000×) [20]. Mean and standard deviation of nuclear area, perimeter, diameter, shortest axis, longest axis, and axis ratio were calculated, as well as different shape factors. The shape factors were calculated by the following formulas: Form_AR = (1/4) * pi * longest axis * shortest axis; Form_PE = 4 * pi * area/(perimeter squared); Form_NCI = perimeter/sqrt (area); Contour ratio = perimeter squared/4 * pi * area; and Roundness = perimeter/(2*sqrt (pi * area)). All morphometric assessments were performed by one observer without knowledge of patient outcome. Statistical analysis Morphometric measurements were first compared between cases and controls using Student’s t-test. Subsequently, the measurements were categorized by quartiles and then odds ratios (OR) and 95% confidence limits (CLs) were calculated for the risk of breast cancer for those in the highest quartile level compared to that for those in the lowest 3 quartile levels using conditional logistic regression. In multivariate analyses, we controlled for age at menarche (<13, 13, 14+), age at first live birth (nulliparous, <23, 23–26, 27+), menopausal status (pre-, peri-, post-), oral contraceptive use (ever versus never), postmenopausal estrogen use (ever vs. never), body mass index (<25, 25+), family history of breast cancer, and the presence of hyperplasia in the benign tissue. All statistical analyses were performed in SAS 9.1 (SAS Institute, Cary, NC). P-values were two-sided. Results Table 1 summarizes the distribution of selected characteristics among the cases and controls. Overall, few differences between the cases and controls were observed for age at menarche, age at first live birth, menopausal status, oral contraceptive use, postmenopausal estrogen use, body mass index, family history of breast cancer, and the presence of hyperplasia in benign tissue. Table 1.Distribution of selected characteristics among breast cancer cases and non-cases N (%)P-valueCasesControlsAge at menarche<1330 (49)26 (37)0.291313 (21)22 (31)14+18 (30)23 (32)Age at first live birthNulliparous11 (18)9 (13)0.84<2322 (36)29 (41)23–2619 (31)23 (32)27+9 (15)10 (14)Menopausal statusPre-30 (49)31 (44)0.71Peri-9 (15)14 (20)Post-22 (36)26 (36)Ever used oral contraceptivesYes35 (57)42 (60)0.76No26 (43)28 (40)Missing01Ever used postmenopausal estrogensYes15 (25)15 (22)0.70No46 (75)54 (78)Missing02Body mass index (kg/m2)<2532 (53)41 (58)0.4225– < 3027 (44)25 (35)30+2 (3)5 (7)Family history of breast cancerYes23 (38)28 (39)0.84No38 (62)43 (61)Hyperplasia in benign tissueAbsent34 (59)47 (68)0.27Present24 (41)22 (32)Missing32 There was little difference between the cases and controls with respect to nuclear morphometric features including mean area, standard deviation (SD) of area, perimeter, diameter, shortest axis, and longest axis, as well as such shape factors as Form_PE, Form_NCI, contour, and roundness (Table 2). In contrast, the shape factor Form_AR was greater among cases than among controls. Furthermore, subjects with hyperplasia had greater measures of some nuclear size features including mean area, SD of area, perimeter, diameter, and longest axis, and the shape factor Form_AR than did subjects without hyperplasia (data not shown). Table 2.Comparison of nuclear morphometric features in benign breast tissue between breast cancer cases and non-casesMorphometric measurementsMean (standard deviation)P-valueCases (n = 61)Controls (n = 71)Mean nuclear area (μm2)26.8 (7.5)25.3 (7.2)0.25SD of nuclear area (μm2)5.2 (1.8)5.0 (1.6)0.43Nuclear perimeter (μm)19.7 (2.7)19.3 (2.7)0.37Nuclear diameter (μm)5.8 (0.8)5.6 (0.8)0.23Shortest nuclear axis (μm)4.8 (0.7)4.6 (0.7)0.16Longest nuclear axis (μm)7.1 (1.0)7.0 (1.0)0.53Axis ratio1.5 (0.1)1.6 (0.2)0.15Form_AR0.984 (0.005)0.981 (0.007)0.0089Form_PE0.844 (0.037)0.831 (0.045)0.083Form_NCI3.874 (0.095)3.909 (0.122)0.071Contour1.198 (0.061)1.221 (0.080)0.068Roundness1.093 (0.027)1.103 (0.034)0.071 Quartile analyses revealed that subsequent breast cancer risk was increased in association with the shape factor Form_AR, but not with the other nuclear morphometric measurements (Table 3). Compared to BBD subjects with Form_AR equal to or less than 0.986, subjects with Form_AR greater than 0.986 had a more than three-fold increased risk of developing breast cancer subsequently (OR = 3.07, 95%CL = 1.61, 5.84). When the analyses were repeated using unconditional logistic regression, which enabled all the available cases and controls to be included, the results did not change substantially. Table 3.Risk of Subsequent development of breast cancer in association with nuclear morphometric featuresaMorphometric measurementsCut-off valueOR (95% CL)Model 1bModel 2cMean nuclear area (μm2)31.21.28 (0.73, 2.25)0.94 (0.50, 1.78)SD of nuclear area (μm2)6.11.33 (0.76, 2.31)1.11 (0.59, 2.07)Nuclear perimeter (μm)21.41.14 (0.70, 1.93)0.85 (0.47, 1.55)Nuclear diameter (μm)6.31.29 (0.73, 2.27)0.95 (0.50, 1.79)Shortest nuclear axis (μm)5.21.62 (0.92, 2.86)1.18 (0.62, 2.26)Longest nuclear axis (μm)8.01.34 (0.75, 2.39)0.95 (0.50, 1.81)Axis ratio1.60.59 (0.30, 1.17)0.71 (0.33, 1.54)Form_AR0.9862.45 (1.42, 4.22)3.07 (1.61, 5.84)Form_PE0.8671.22 (0.71, 2.08)1.57 (0.83, 2.97)Form_NCI3.9351.07 (0.58, 1.98)1.18 (0.61, 2.27)Contour1.2361.13 (0.61, 2.10)1.22 (0.63, 2.35)Roundness1.1101.07 (0.58, 1.98)1.18 (0.61, 2.27)a Analyses were conducted among 53 matched case-control sets by comparing the highest quartile versus the lowest 3 quartiles in conditional logistic regression modelsb Adjusted for matching variablesc Adjusted for matching variables, age at menarche (<13, 13, 14+), age at first live birth (nulliparous, <23, 23–26, 27+), menopausal status (pre-, peri-, post-), oral contraceptive use (ever vs. never), postmenopausal estrogen use (ever versus never), body mass index (<25, 25+), family history of breast cancer, and the presence of hyperplasia in the benign tissue Discussion We found that breast cancer risk in women with BBD was positively associated with the shape factor Form_AR, a measurement that takes the shortest nuclear axis and the longest nuclear axis into consideration simultaneously. In contrast, there was no alteration in risk in association with nuclear area, SD of nuclear area, nuclear perimeter, nuclear diameter, shortest nuclear axis, longest nuclear axis, and other shape factors. Although subjects with hyperplasia had greater measures of Form_AR than did subjects without hyperplasia, we adjusted for hyperplasia, suggesting that the association with Form_AR is independent of that due to the presence of hyperplasia. Shape is one of the factors that pathologists use in assessing nuclear atypicality. Shape factors have been shown to have prognostic value in breast cancer [21–23], renal cell cancer [24], colorectal cancer [25], squamous cell carcinoma of the larynx [26], melanoma [27], and rhabdomyosarcoma [28]. Apparently, alterations in nuclear shape can already be present at the earliest stages of carcinogenesis. This has in the breast also been shown for nuclear chromatin patterns [29]. To date, only one study has been published that assessed morphometric features in association with subsequent development of breast cancer among women with BBD [16]. That study found positive associations for mean nuclear area, nuclear diameter, nuclear perimeter, and the longest nuclear axis, but no associations for SD of the nuclear area and the shortest nuclear axis; shape factors were not evaluated. However, potential confounding factors were not controlled for. In contrast to these findings, nuclear size features were not associated with risk in the present study, which may perhaps be explained by differences in tissue processing procedures. Our case-control study was nested in a cohort of patients with histopathologically confirmed BBD and our findings are likely to be internally valid. Biased measurement of the study exposures was not likely a source of error, given that the morphometric features were assessed without knowledge of the patient outcome status. Our study power, however, was limited by the relatively small sample size, due to which we were not able to evaluate modifying effects by well-documented risk factors of breast cancer. Moreover, residual confounding might still exist, although to minimize confounding we controlled for menstrual and reproductive history, exogenous estrogen use, body mass index, and family history of breast cancer in multivariate analyses. In conclusion, our study results suggest that the shape factor that takes both shortest nuclear axis and longest nuclear axis into consideration might be of value to predict subsequent development of breast cancer among patients with BBD. Given the limitations of our study, larger studies are warranted to confirm our study results.

Purinergic_Signal-1-3-2096543

Tri-nucleotide receptors play a critical role in epithelial cell wound repair

The cornea plays a major role in the refraction of light to the retina. Therefore, the integrity and transparency of the corneal epithelium are critical to vision. Following injury, a combination of rapid signal transduction events and long-term cell migration are essential for wound closure. We have demonstrated previously that injury resulted in the release of nucleotides that induce the propagation of a Ca2+ wave to neighboring cells. This suggests that nucleotides and their receptors are critical components of wound healing. Epidermal growth factor (EGF) and integrins also have been shown to play a role in injury. In this study, we demonstrate that pretreatment of cells with ATP and UTP inhibited the immediate wound response, while BzATP, ADP, and UDP did not affect this response. Tri-nucleotide pretreatment also reduced the EGF induced Ca2+ response. Additionally, lower EC50 concentrations of ATP and UTP triggered migration of cells that was enhanced further with EGF and was inhibited by the tripeptide, RGD. Results indicate that the desensitization induced by ATP and UTP was specific. While ADP and UDP cause a homologous desensitization of their own signal, they did not cause an inhibition of the wound response nor does BzATP. Neither Ca2+ wave propagation nor cell migration occurred in response to β,γ-MeATP. Together these results lead us to hypothesize that corneal epithelial wound repair is mediated by both P2Y2 and P2Y4 receptors. Introduction Wound healing is a series of complex biological responses requiring cytoskeletal and extracellular matrix remodeling, signal cascades, and gene regulation. Initial and long-term signals after injury are communicated to the cells and promote wound healing. These signals and events must be understood to advance our knowledge of normal wound repair. Investigators have demonstrated that nucleotides are released from cells by mechanical stimulation, ligand binding, exocytotic release, or injury [1–5] Nucleotide stimulation can also lead to an increase in intracellular calcium. The increase in cytosolic calcium can be generated via activation of purinergic (P2Y or P2X) receptors through distinct mechanisms. In addition, purinergic receptors can promote cell migration, a critical component of wound repair. The exact role and identity of the various receptors and their importance in different systems has yet to be elucidated. Nucleotides have been shown to accelerate wound closure [6]. Nucleotides can act as a chemotactic stimulus for cell migration in many cell types including corneal epithelial [7], arterial smooth muscle [8], microvascular endothelial [9], and immature dendritic cells [10]. P2 receptors also have been shown to play a role in repair after ischemia and renal tubular epithelial wound healing [11, 12]. Previously, we demonstrated that ATP plays a critical role in intercellular communication following mechanical injury. ATP induced cell migration, and Reactive Blue-2, a P2 antagonist, inhibited wound closure [7]. In addition, nucleotides induced calcium waves and phosphorylation of paxillin (pY118) and ERK1/2 in a transient manner [7, 13]. Furthermore, neither occurred in the presence of β,γ-MeATP [7, 13]. Our goal is to determine how epithelial cells respond to nucleotides in order to understand the events that occur after injury. P2Y and P2X receptors are expressed in a number of cell types. As G-protein coupled receptors (GPCR), P2Y receptors are seven pass transmembrane receptors coupled to heterotrimeric G-proteins (review in [14]). P2Y receptors couple to Gαq and activate phospholipase C β (PLCβ), which cleaves phosphatidylinositol 4,5-bisphosphate (PIP2) to diacylglycerol (DAG) and phosphoinositol tri-phosphate (IP3). IP3 binds to Ca2+ channels on the endoplasmic reticulum and causes an increase in intracellular Ca2+. This family of receptors consists of eight members: P2Y1, P2Y2, P2Y4, P2Y6, P2Y11, P2Y12, P2Y13, and P2Y14. P2X receptors are ligand gated ion channels that open upon agonist binding allowing extracellular calcium to enter the cell. This family of receptors has seven members: P2X1–7. Each receptor has a specific agonist potency profile for different nucleotides with some showing preference for a specific nucleotide and others preferentially responding to tri-nucleotides or adenine based nucleotides [15]. We have shown that corneal epithelial cells express a number of P2YR subtypes [7], and an intracellular Ca2+ increase is seen in response to both nucleotides and injury [5]. Epidermal growth factor receptor (EGFR) also plays an important role in injury response. The EGFR family has four receptor family members, ErbB1-4. These receptors can form both homo- and hetero-dimers when stimulated by the epidermal growth factor (EGF) or other ligands [16]. The EGFR becomes activated upon injury, and inhibition of ErbB phosphorylation limits proper wound closure in vitro [7]. EGF itself promotes cell migration and wound healing in vivo [17] and can up-regulate integrin expression [18]. Gross mechanical injury induces a Ca2+ wave that propagates out from the wound site. EGF has been shown to increase the intensity of the Ca2+ wave but is not required for propagation [5]. Recently, growth factor receptors and GPCRs have been demonstrated to participate in cross talk connecting signaling pathways through trans-activation, which can occur via enhancement or down-regulation of downstream signal cascades [19]. In addition, specific P2 receptors can associate with other signaling proteins. For example, P2Y2 has an arginine-glycine-aspartic acid (RGD) domain on its first extracellular loop that can associate with integrins [20]. Currently, no other P2Y receptor has an identified RGD domain that interacts with integrins. In addition, this RGD domain promotes signaling and interactions with epidermal growth factor receptors (EGFR) [21]. Our goal is to determine if specific P2 receptors play a role in the injury response and wound healing in the cornea. Improper wound repair can alter refraction of light, one of the major roles of the cornea. Immediately after injury, corneal epithelial cells display an intracellular Ca2+ wave that propagates from the site of injury to neighboring cells [5]. This wave does not propagate via gap junctions but is mediated by the release of extracellular nucleotides activating P2Y receptors that lead to activation of intracellular signaling pathways such as ERK1/2 [13]. We hypothesize that injury induces an immediate localized event that stimulates later events such as release of growth factors and long-term signals. In this paper, we provide evidence that specific subtypes of purinergic receptors regulate the injury response. Pretreatment of cells with ATP and UTP inhibited the propagation of the injury induced Ca2+ wave while BzATP, ADP, and UDP did not participate in the desensitization. In addition, pre-stimulation with tri-nucleotides, but not BzATP or di-nucleotides, resulted in a decrease in the intracellular Ca2+ release induced by EGF. ATP and UTP preferentially enhanced later cellular events, including migration. Cellular migration was enhanced further when cells were co-stimulated with EGF. In addition, the ATP mediated migration was reduced by the tri-peptide, RGD. The results indicate that immediate and long-term components of the wound response are mediated by tri-nucleotide receptors. Materials and methods Reagents The nucleotides [adenosine 5′-triphosphate (ATP), uridine 5′-triphosphate (UTP), adenosine 5′-diphosphate (ADP), uridine 5′-diphosphate (UDP), β,γ-methyleneadenosine 5′-triphosphate (β,γ-MeATP), and 2′-3′-O-(4-Benzoylbenzoyl) adenosine 5′-triphosphate (BzATP)], adenosine, apyrase, and Reactive Blue 2 were purchased from Sigma Chemical Company (St. Louis, MO). Human recombinant epidermal growth factor (EGF) and penicillin-streptomycin were purchased from Invitrogen Corporation (Carlsbad, CA). The fluorescent Ca2+ indicator dye Fluo-3AM and pluronic acid were from Molecular Probes, Inc. (Eugene, OR). 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic (BAPTA) was purchased from Calbiochem (La Jolla, CA). RGD and RGE peptides were purchased from American Peptide (Sunnyvale, CA). Cell culture Primary rabbit corneal epithelial cells and an established human corneal cell line were used. Human epithelial cells transformed (HCE-Ts) were established by Araki-Sasaki et al. [22], using an SV-40 recombinant adenovirus vector. Primary corneal epithelial cells were isolated and cultured from freshly extracted whole rabbits eyes (Pel-Freeze Biologicals (Rogers, AR) [7]. Ca2+ imaging HCE-T cells and primary cells were grown to confluency on 22 mm2 coverslips and quiesced 18 to 24 h before experiments in unsupplemented serum-free keratinocyte medium (K-SFM). Cells were washed in a HEPES-buffered saline solution containing 137 mM NaCl, 5 mM KCl, 4 mM MgCl2, 3 mM CaCl2 · 2H2O, 25 mM glucose and 10 mM HEPES [23]. Cells were incubated 20–30 min in 5 µM Fluo-3AM resuspended in HEPES buffer supplemented with 0.02% pluronic acid in DMSO [5, 7, 13]. Coverslips were mounted in a flow through perfusion apparatus connected to inlet tubes with varying solutions and an outlet pump (Warner Instruments). The mounted coverslip was placed on the stage of a Zeiss Axiovert LSM 510 laser scanning confocal microscope, and cells were scanned with an Argon laser every 786 ms [7]. Cells were perfused with HEPES buffered saline at a rate that did not elicit a response, and baseline fluorescence was obtained. Cells were stimulated either by perfusing nucleotides or EGF prepared in HEPES buffered saline and/or by manually perturbing cells [5]. Ca2+ data analysis Ca2+ data was recorded by LSM software as average fluorescence intensity (F) for a region of interest over time. The region was either a 460 \smm \sx 460 \smm field or single cells, over time. Raw data was entered into KaleidaGraph. Percent change in average fluorescence was calculated using the following formula where F0 is initial average fluorescence for the region of interest [5, 7]. Data was either graphed as average maximal fluorescence of at least three different experiments or as percent change in average fluorescence for time course. For the injury experiments, the area cleared by the wound was excluded from calculations. Migration assay HCE-T cells were quiesced 18 to 24 h in unsupplemented K-SFM. Cells were washed in phosphate buffered saline (PBS), trypsinized, and the reaction was stopped with filtered soybean trypsin inhibitor as described previously [7]. Cells were resuspended at a concentration of 125,000 cells/100 µl in binding buffer (unsupplemented K-SFM, 0.05% gelatin and 25 mM HEPES). Binding buffer, 600 µl, with or without added stimuli was placed in the wells of a 24 well plate. Costar Transwell inserts (6.5-mm diameter polycarbonate membranes, 8 mm pore size) were placed into the wells, and 100 µl of the cell suspension was added to the inserts. Experiments were conducted for 8 h at 37 °C. To collect conditioned wound medium, cells were washed and incubated in binding buffer. Cells were scrape wounded as described previously [13]. The wound medium was collected immediately and used in migration assays. Unwounded medium was defined as binding buffer added to cells and collected without injury to cells. The conditioned wound medium was also treated with apyrase prior to the migration assays. To determine migration, upper and lower chambers were aspirated and rinsed with PBS. Cells were fixed at room temperature for 10 min, rinsed with PBS, and non-migrated cells were removed. Migrated cells were stained with propidium iodide (5 µg/ml) (Molecular Probes, Inc., Eugene, OR) for 5 min and rinsed with PBS. Membranes were mounted onto a slide with SlowFade Antifade (Molecular Probes, Inc., Eugene, OR). For each membrane (33.2 mm2), six random 10× fields (one field = 1.37 mm × 1.08 mm, or 1.48 mm2) were photographed. Cells were counted and averaged for each membrane and experiments were performed in triplicate. Statistical analysis Values were given as the mean \+- SD. Statistical comparisons were made using Student’s t-test with a stringency of P > 0.01 for significance. Data were fit using KaleidaGraph to calculate concentration for half the maximal response (EC50) and maximal possible change in fluorescence for a given agonist (\gD Fmax). A generalized single-site binding model was used as follows: , where \gDF is the percent change in fluorescence and [L] is the concentration of the agonist. Results Nucleotide dose response curves Ca2+ dose response curves were determined for ATP, UTP, ADP, UDP, and BzATP over a concentration range of 10−7 to 10−3 M (Figure 1). The response to ATP and UTP was greater than that of ADP and UDP over the entire range. The BzATP response was similar to ADP and UDP at concentrations ≤10−4 but lower at higher concentrations. In addition, adenosine, an agonist for P1 receptors, and β,γ-MeATP, an agonist for P2X receptors, were evaluated. Neither of these agonists induced a change in intracellular Ca2+ (data not shown). The EC50 for ATP and UTP were found to be (3.4 ± 1.8) × 10−6 M and (3.4 ± 2.0) × 10−6 M respectively. The EC50 for ADP and UDP were higher than the tri-nucleotides, (1.8 ± 0.9) × 10−4 M and (1.83 ± 0.12) × 10−4 M. respectively. The EC50 for BzATP was (8.6 ± 5.9) sx 10−5. R values for line fit were 0.96 for ATP, UTP, and UDP, 0.93 for ADP, and 0.91 for BzATP. Figure 1Nucleotide induced Ca2+ dose response curves. HCE-T cells were incubated in 5 µM Fluo-3AM for 30 min and imaged every 786 ms using a flow through apparatus on an LSM 510 confocal. Cells were stimulated with the indicated concentration of nucleotide for 60 s. Maximal percent change in average fluorescence of a 460 µm × 460 µm field was determined. Data was fit to a generalized single-site binding model. Graph represents a minimum of three experiments at each concentration tested. To verify that P2 receptors caused the intracellular Ca2+ increase, cells were pre-incubated for 30 min in 100 µM Reactive Blue 2 prior to experimentation. Cells were placed in a flow through system and washed for at least 50 s in HEPES buffered saline. Cells were stimulated with 100 µM of the indicated nucleotide for 100 s and washed again with HEPES buffer. Non-pretreated cells were used as control. Reactive Blue 2 lowered the Ca2+ response to nucleotides. The ATP and UTP responses were reduced by one-fifth, and the ADP and UDP responses were reduced by half. Role of nucleotides in injury response We have demonstrated that epithelial cells respond to nucleotides in a saturable manner. However, these results did not identify which P2 receptors were involved in the injury response. To evaluate the wound response, cells were placed in a flow through apparatus. Background images were collected in HEPES buffered saline, and the cells perturbed manually (Figure 2a). Percent change in average fluorescence of individual cells adjacent to the wound (Figures 2b, 2c (#1–5)) are shown inside the large white circle and graphed (Figure 2c). These depict a rapid increase in fluorescence after injury that is followed by attenuation. The rate of attenuation of injured cells was variable. While these cells did respond they were not a part of the calcium wave that propagated from the site of injury [5]. We also examined cells not immediately adjacent to the wound, cells involved in the wave (Figures 2b, 2d). When individual cells that were not adjacent to the wound were evaluated [(Figures 2b, 2d (#6–10) (outside of white circle)], the percent response was lower on average and the response time was delayed (Figure 2d). The delay in response time could be accounted for by the time it took the Ca2+ wave to travel the distance to nonadjacent cells. Figure 2Pre-stimulation with ATP inhibits injury induced Ca2+ wave. Primary corneal epithelial cells were incubated in 5 µM Fluo-3AM for 30 min and imaged in a flow through apparatus on an LSM 510 confocal. Cells were washed in HEPES buffered saline with Ca2+ for at least 30 s, and stimulated by wounding or with ATP. Individual cells were analyzed for percent change in average fluorescence. Intensity scale is shown in (a) with red indicating highest Ca2+ levels and blue indicating lowest Ca2+ levels. The horizontal white bar in (a) represents 100 µm. a) Cells were washed in HEPES buffer containing Ca2+ and wounded. A series of images taken from a time course of a representative experiment of a wound (shown at asterisk) is presented. b) A single image taken after wounding is shown. Cells (#1–5) immediately adjacent to the wound (c) and cells away (#6–10) from the wound (d) were analyzed. e) Cells were washed in HEPES buffer containing Ca2+, stimulated with 100 µM ATP (80 s), and wounded (157 s). A series of images taken from a time course of a representative experiment is presented (wound shown at asterisk). f) A single image taken after wounding is shown. Cells (#1–5) immediately adjacent to the wound (g) and cells away (#6–10) from the wound (h) were analyzed. Images are representative of at least 10 independent experiments. To evaluate what role ATP played in the initial injury response, cells were stimulated by ATP and wounded. To perform the experiment, cells were placed in the flow through apparatus, equilibrated with HEPES buffered saline, and stimulated with HEPES buffered saline containing 100 µM ATP. The response to ATP was allowed to attenuate (approx. 80 s), and then the cells were wounded (Figure 2e). The cells were analyzed for average change in percent fluorescence for individual cells adjacent to the wound (Figures 2f, 2g) and those involved in wave propagation (Figures 2f, 2h). As predicted from Figure 2a and d, cells adjacent to the wound responded (Figures 2f, 2g). The Ca2+ wave was absent when it followed pretreatment with ATP (Figures 2f, 2h). In its place was a minor inflection at the time of injury that resembled the wound response induced in the absence of intracellular Ca2+ [5] (see Figure 5). These responses indicate that the injury response was inhibited by the desensitization of specific P2Rs by ATP. Figure 5Ca2+ free media inhibits an injury response after ATP stimulation. Primary corneal epithelial cells were incubated in 5 µM Fluo-3AM for 30 min and imaged in a flow through apparatus on an LSM 510 confocal. Cells were washed in HEPES buffered saline for at least 30 s, and stimulated by wounding or ATP. Individual cells were analyzed for percent change in average fluorescence. Intensity scale is shown in (a) with red indicating highest Ca2+ levels and blue indicating lowest Ca2+ levels. The horizontal white bar in (a) represents 100 µm. a) Cells were pre incubated in BAPTA (100 µM) washed in HEPES buffer containing Ca2+ and wounded. A series of images taken from a time course of a wound (shown at asterisk) of a representative experiment is presented. b) A single image taken after wounding is shown. Cells (#1–5) immediately adjacent to the wound (c) and cells away (#6–10) from the wound (d) were analyzed. e) Cells were washed in Ca2+ free HEPES buffer containing EGTA, stimulated with 100 µM ATP, and wounded. A series of images taken from a representative time course is presented (wound shown at asterisk). f) A single image taken after wounding is shown. Cells (#1–5) immediately adjacent to the wound (g) and cells away (#6–10) from the wound (h) were analyzed. Images are representative of at least 10 independent experiments. The series of images are taken from Movie 2 (see online version of article at www.springeronline.com). To determine which nucleotides could desensitize the Ca2+ release in response to wounding, cells were pretreated with nucleotides and injured (Figures 3 and 4). Both ATP and UTP pre-stimulation inhibited the wound-induced wave (Figure 3), while ADP, UDP, and BzATP did not inhibit the injury induced Ca2+ wave (Figure 4). Together, these results indicate that both ATP and UTP desensitize receptors and inhibit the injury response at an equivalent molar concentration. Figure 3Tri-nucleotides attenuate the injury induced Ca2+ wave. Primary corneal epithelial cells were incubated in 5 µM Fluo-3AM for 30 min and imaged in a flow through apparatus on an LSM 510 confocal. Cells were washed in HEPES buffered saline with Ca2+ for at least 30 s and stimulated with the indicated nucleotide and wounded. Intensity scale is shown in (a) with red indicating highest Ca2+ levels and blue indicating lowest Ca2+ levels. The horizontal white bar in (a) represents 100 µm. a, b) Cells were washed in HEPES buffer with Ca2+, stimulated with 100 µM of the indicated nucleotide, and wounded. A series of images taken from a time course of a representative experiment is presented (wound shown at asterisk). c, d) Percent change in average fluorescence for the whole field (460 µm × 460 µm) was calculated and graphed over the time course for each experiment. Images are representative of at least 10 independent experiments. Series of images are taken from Movie 1 (see online version of article at www.springeronline.com).Figure 4Di-nucleotides and BzATP do not attenuate the injury induced Ca2+ wave. Primary corneal epithelial cells were incubated in 5 µM Fluo-3AM for 30 min and imaged in a flow through apparatus on an LSM 510 confocal. Cells were washed in HEPES buffered saline with Ca2+ for at least 30 s and stimulated with the indicated nucleotide and wounded. Intensity scale is shown in (a) with red indicating highest Ca2+ levels and blue indicating lowest Ca2+ levels. The horizontal white bar in (a) represents 100 µm. (a–c) Cells were washed in HEPES buffer with Ca2+, stimulated with 100 µM of the indicated nucleotide, and wounded. A series of images taken from a time course of a representative experiment is presented (wound shown at asterisk). (d–f) Percent change in average fluorescence for the whole field (460 µm × 460 µm) was calculated and graphed over the time course for each experiment. Images are representative of at least 10 independent experiments. Series of images are taken from Movie 1 (see online version of article at www.springeronline.com). To evaluate the inflection detected in Figures 2e–h, cells were pretreated with BAPTA (100 µM) for 30 min and individual cells were analyzed (Figures 5a–5d). Cells responded to the injury in a manner similar to the cultures that had been pretreated with ATP (Figures 3 and 4). These results indicate that the cellular response adjacent to the wound did not occur via P2Y receptor signaling since the Ca2+ increase came from extracellular stores (Figures 5a–5d). This agrees with previous results using thapsigargin [5]. To evaluate the requirement for extracellular Ca2+, cultures were placed in the flow through apparatus in Ca2+ free HEPES buffer containing EGTA. These cells were stimulated with 100 µM ATP made in Ca2+ free HEPES buffered saline. Once the response had attenuated (approx. 100 s), the cells were wounded (Figure 5e), and individual cells were analyzed for average change in percent fluorescence in cells adjacent to the wound margin (Figures 5f, 5g) and those involved in wave propagation (Figures 5f, 5h; as in Figure 2). When the experiment was performed in Ca2+ free media, the inflection detected previously in the injury response was absent, completely eliminating any Ca2+ response to injury (Figures 5e–5h). Nucleotides affect EGF induced Ca2+ response Cross-talk between the EGF and ATP signaling pathways occurs in many cell systems. To determine if this was a property unique to ATP, the response to EGF was evaluated after pre-stimulation with other nucleotides. HCE-Ts were perfused with HEPES buffered saline to establish baseline fluorescence, and the cells were stimulated with 100 µM ATP, UTP, ADP, UDP or BzATP for 100 s. Cells were washed with HEPES for 100 s and stimulated with 8 nM EGF for 100 s. Maximal percent change in fluorescence of the entire 460 µm × 460 µm field was calculated (Figure 6a). The results were compared with cells washed with HEPES buffer and stimulated with 8 nm EGF alone for 100 s (− pretreatment). Cells exhibited an attenuated Ca2+ response upon EGF stimulation when they were pretreated with either ATP or UTP. In contrast, when the cells were pre-stimulated with BzATP, ADP, or UDP, there was no detectable desensitization (Figure 6a). The reverse experiment was performed where cells were pre-stimulated with EGF and then by specific nucleotides (data not shown). Pretreatment with EGF did not desensitize the cells indicating that EGFR stimulation with EGF did not inhibit the response to nucleotides. Figure 6Injury and tri-nucleotides decrease subsequent EGF induced Ca2+ response. HCE-T cells were incubated in 5 µM Fluo-3AM for 30 min and imaged every 786 ms in a flow through apparatus on an LSM 510 confocal. a) Cells were stimulated with 100 µM of the indicated nucleotide for 100 s, washed in HEPES buffered saline for 100 s, and then stimulated with 8 nM EGF or stimulated with EGF without pretreatment. Maximal percent change in average fluorescence of a 460 µm × 460 µm field was determined. b) Cells were either wounded and stimulated with 8 nM EGF or stimulated with EGF without injury. Individual cells were selected for analysis. In the wound model, cells that responded to the wound were selected. (t-test, * P < 0.0005) Data represent a minimum of three independent experiments. As desensitization occurred after cells were pre-stimulated with ATP or UTP and wounded or stimulated with EGF, our goal was to determine if desensitization of the EGF receptor occurred in our wound model. Cells were placed in the flow through apparatus, perfused with HEPES buffered saline to establish baseline fluorescence, and injured. Once fluorescence returned to the base line, the cells were stimulated with EGF. The analysis was performed on individual cells that responded to both the injury and EGF (Figure 6b). Results were compared to individual cells stimulated with EGF alone. The Ca2+ response to EGF was significantly lowered after injury. The high standard deviation was due to the variable size of the wound. Nucleotides and EGF induce cell migration The long-term effects of P2 receptors on wound healing were examined using Transwell migration assays. Previously we showed that P2 receptors were necessary for wound repair in directed migration scratch assays [7]. Furthermore, we demonstrated that migration was optimal at 1 µM for ATP and 0.16 nM for EGF in Transwell assays [7]. In these experiments, migration to other nucleotides and adenosine were evaluated over a range of concentrations from 0.1 to 100 µM (Figure 7). Optimal migration for ADP and BzATP occurred at 10 µM (Figures 7a, b), while optimal migration for ATP and UTP occurred at 1 µM ([7] and data not shown). UDP and β,γ-MeATP did not induce migration at any concentration (Figures 7c, d). To assess preferential migration to any one particular nucleotide or adenosine, the nucleotides and adenosine were evaluated and compared. The cellular response was tested at 1 µM, the optimal concentration for ATP and UTP. In addition, ADP was tested at 10 µM, its optimal concentration. ATP, UTP, and ADP induced comparable migration at their optimal concentrations (162 ± 39, 170 ± 70, and 188 ± 47 cells/1.48 mm2, respectively). However, ADP at 1 µM, UDP, adenosine, or binding buffer did not induce cell migration (23 ± 7, 18 ± 7, 31 ± 6 and 29 T 11 cells/1.48 mm2, respectively) (Figure 7e). Figure 7Cell migration is nucleotide specific. Transwell migrations were performed for 8 h at 37 °C with the indicated nucleotide. HCE-T cells were stained with propidium iodide, counted in six randomly chosen fields (1.46 mm2), and averaged. Dose response curves for a) ADP, b) BzATP, c) UDP, and d) β,γ-MeATP are shown. Experiments were performed at least three times and representative curves were chosen. e) Cells were stimulated with binding buffer (neg), 1 µM ATP, UTP, ADP, UDP, or Adenosine (Ado), or 10 µM ADP. Experiments were performed in triplicate. We evaluated the role of factors released into the wound media on cell motility and compared these effects to stimulation with nucleotides and EGF. Wound media was collected immediately after injury and added to migration assay chambers. Luciferase assays performed on wound media demonstrated presence of ATP [5]. Cell migration was 3.7-fold greater when cells were stimulated with wound media compared to wound media treated with apyrase or unwounded control media (Figure 8a). Previously we established that epithelial cells migrate to ATP or EGF in a dose dependent manner and optimal migration was determined to be 1 µM and 0.16 nM respectively [24]. To evaluate the combined effect of ATP and EGF, cells were co-stimulated with the optimal concentrations of each and compared. In a representative experiment an average of 369 (±34) cells migrated/1.48 mm2 in response to 0.16 nM EGF while an average of 314 (±31) cells migrated/1.48 mm2 in response to 1 µM ATP (Figure 8b). When cells were co-stimulated with 1 µM ATP and 0.16 nM EGF, migration increased more than threefold with an average of 1043 (±66) cells/1.48 mm2 (Figure 8b). Co-stimulation experiments were performed with UTP, ADP and UDP at their respective optimal concentrations (Figures 7 and 8c). Cells exposed to ADP or UDP (1 µM) and co-stimulated with EGF (527 ± 61 and 637 ± 51 cells/1.48 mm2, respectively) showed migration similar to that of EGF alone (457 ± 115 cells/1.48 mm2). When cells were co-stimulated with UTP and EGF, migration was enhanced in an additive manner (801 ± 133 cells/1.48 mm2). In addition when cells were exposed to 10 µM ADP and EGF, migration was greater than either alone (753 ± 173 cells/1.48 mm2). Figure 8Wound media, nucleotides, and EGF mediate cell migration. Transwell migrations were performed for 8 h at 37 °C with the indicated nucleotides and/or EGF. HCE-T cells were stained with propidium iodide, counted in six randomly chosen fields (1.46 mm2) and averaged. a) Migration of cells stimulated with wound media were compared to those stimulated with unwounded control media, wound media treated with apyrase and binding buffer (negative). b) A representative experiment is shown where cells were stimulated with optimal concentration of ATP (1 µM) and EGF (0.16 nM) and compared to co-stimulation with ATP and EGF and negative binding buffer. c) Cells were simulated with binding buffer (neg), 1 µM UTP, ADP, or UDP, or 10 µM ADP +/− 0.16 nm EGF. To understand how nucleotides regulate cell migration, we asked whether integrins mediated the response. Integrins are known to play an important role in cell migration, and the integrin binding tri-peptide, RGD, that is found in the first extracellular loop of the P2Y2 receptor, has been shown to interact with integrins [20]. In addition, expression of α6β4 in corneal epithelial cells is up-regulated by EGFR activation [18]. To evaluate the role of integrin receptors in nucleotide mediated migration, RGD peptide was added to the cell suspension prior to its addition to the Transwell chamber. RGE peptide was used to control for non-specific binding. Cells were migrated to binding buffer or 1 µM ATP. When either the RGD or RGE peptide was added to the cells migrating to binding buffer, there was no detectable difference over control (Figure 9). Cells stimulated with ATP migrated 362 (±92) per field. Interestingly, cell migration was reduced almost fourfold when RGD was present (96 ± 43 cells/1.48 mm2) (Figure 9). The control peptide did not decrease migration significantly from control (284 ± 59). Since RGD peptide can inhibit migration, it is likely integrins play a role in ATP induced cell migration. The nucleotide receptors, EGFR, and integrins mediate calcium signaling and cell migration. Thus, interaction between these protein families may be critical in regulating the cellular signals after injury. Figure 9ATP and integrins mediate cell migration. Transwell migrations were performed for 8 h at 37 °C with the indicated nucleotides. HCE-T cells were stained with propidium iodide, counted in six randomly chosen fields (1.46 mm2) and averaged. RGD or RGE peptide, 1 µM, was added to the cell suspension. Cells were stimulated with binding buffer (control) or 1 µM ATP. (t-test, *P < 0.0005) Experiments were performed in triplicate. Discussion The results of these studies demonstrate that P2 receptors play a role in the initial cellular responses that occur after injury and the long-term signals that are necessary for wound closure in corneal epithelial cells. We have demonstrated that nucleotides are the active component released with injury and are responsible for the intercellular Ca2+ wave that is propagated in neighboring cells [5, 13]. We have shown that ATP, UTP, ADP, UDP and BzATP promote an intracellular Ca2+ increase, β,γMe-ATP does not. Our goal was to demonstrate that nucleotides play a prominent role in the cellular response to injury, in migration, and in the cross-talk with EGF. Recently involvement of P2 receptors in wound healing has begun to be demonstrated [2]. Desensitization experiments demonstrate the role of nucleotides in the injury response. We have shown pretreatment of cells with ATP and UTP inhibits the response to a subsequent wound (Figure 3). Several explanations are plausible: 1) Specific receptors desensitize via receptor internalization or signal transduction down regulation [25] or 2) When cells are exposed to a solution of a tri-nucleotide, ATP and/or UTP receptors have bound agonist thus saturating the receptors. We predict that the amount of nucleotide used to desensitize the wound approximates the concentration needed for the maximal Ca2+ response (Figure 1). In either case, the result is that nucleotides that are released would not be able to induce any further effect upon the receptors. In addition, injury, ATP and UTP, induce a decrease in the EGF induced Ca2+ response (Figure 6). Since nucleotides are released with injury [13], we hypothesize that desensitization of the EGF receptor occurs via nucleotides. In contrast, while ADP and UDP cause homologous desensitization of their own Ca2+ signal [7], ADP, UDP, and BzATP do not affect either the EGF or wound response even when the receptors are saturated (Figures 4 and 6). Therefore, we hypothesize that P2Y1, P2Y6, and P2X7 receptors do not play a role in the Ca2+ wave that occurs after injury. We have demonstrated that the source of the calcium for the injury induced wave is intracellular ([5] and Figure 5). Pretreatment with either thapsigargin or BAPTA inhibits the Ca2+ wave but not oscillations of cells at the wound margin. In addition, pretreatment with either ATP or UTP inhibits only the wave but not the Ca2+ increase in the cells adjacent to the wound. Cells adjacent to the wound edge utilize extracellular calcium (Figure 5), but the method of extracellular calcium entry is unknown. One possibility is that the plasma membranes of the injured cells are damaged, and Ca2+ freely enters. The response could also be mediated via stress receptors or P2X receptors, both of which are ion channels that permit entry of extracellular calcium [26]. While these cells are not involved in the propagation of the wave, we cannot rule out the possibility that they play a role in wound healing, and that is an area of current exploration. Activation of intracellular signaling pathways by P2 receptors is emerging as intricate patterns of cell specific processes. Involvement of αVβ3 integrins has been shown to be essential for both ATP induced intracellular Ca2+ increase and ERK1/2 activation in 1321N1 P2Y2 transfected cells [20]. We have shown in our system that integrins play a role in the response to nucleotides (Figure 9). As integrins are known to play a role in cell migration and wound healing, their coordinated role in nucleotide signaling is of great interest. Purinergic signaling is also known to activate downstream signaling events using a number of receptors. We have shown that ATP, UTP, ADP, and UDP promote activation of ERK1/2, but α,β-MeATP and β,γ-MeATP do not [13]. P2X2 and P2Y1 receptors have been shown to play a role in activating ERK1/2 in stretch induced injury of astrocytes in Ca2+ dependent manner [27]. In contrast in embryonic kidney cells, P2X7 activates ERK1/2 in a Ca2+ independent manner [28]. Additionally PKC and PKD both have been shown to play role in P2X7 induced ERK1/2 activation [29]. In pro-monocytic cells, activation of ERK1/2 by P2Y2 has been shown to involve c-Src but not PKC [30]. Recently, the P2YRs have been hypothesized to play a role in cross-talk with the EGFR [7, 13, 21, 31, 32]. There are a number of potential pathways, several of which may occur. One current hypothesis is cross-talk occurs via the triple membrane pass system where GPCR activated matrix metalloproteases (MMPs) cleave pro-heparin binding (HB)-EGF from the membrane and HB-EGF can then activate the EGFR [33, 34]. Another potential method for cross-talk is via Src. While Src is known to play a role in the phosphorylation of EGFR, its involvement in corneal epithelial injury has not been determined. GPCR stimulation has been found to lead to phosphorylation of the EGFR in a Src and/or Pyk2 dependent manner in many cell types [21, 31, 35, 36]. Future experiments will be performed to test the hypothesis that Src and Pyk2 play an active role in nucleotide-mediated activation of the EGFR in corneal epithelial cells. Inhibition of the EGFR activation with tyrphostin has been shown to inhibit ATP induced cell migration in epithelial cells [7]. These indicate that cross talk between GPCR and the EGFR occur over the entire time course of wound response and repair. ATP and UTP are the major players in Ca2+ waves after injury and the cross talk with the EGFR (Figures 3 and 6). However, this does not preclude that other nucleotides may mediate other facets of the wound response. While ADP and BzATP do not play a role in the injury induced Ca2+ wave, they can stimulate migration, albeit at a higher concentration than ATP or UTP (Figure 7). Though UDP does not stimulate migration, both ADP and UDP have the ability to activate downstream signals. We have shown that ERK1/2 and paxillin become phosphorylated upon stimulation with both tri- and di-nucleotides [7, 13]. Additionally, BzATP has been shown to activate downstream signals such as ERK1/2 in other cell systems [28, 29]. The signals propagated by di-nucleotides may be artifacts of saturating the system with larger than physiological concentrations [37]. Another possibility is that the signal is real and that after injury the receptors are available at a density that is unable to elicit a detectable downstream cascade. This may explain why ADP can induce migration at 10 µM. However, this does not take into account that high concentrations of ADP are still not able to inhibit the injury induced Ca2+ wave (data not shown). Proper signaling relies on optimal concentrations of nucleotides. At high concentrations where maximal Ca2+ release upon nucleotide exposure occurs (>100 µM), epithelial cells do not migrate toward the stimuli. The desensitization of the calcium response that occurs at 100 µM may explain why cell migration is minimal. In contrast, at lower concentrations where the maximal migration occurs, little desensitization occurs. This bell shape curve is not atypical and has been demonstrated to occur with nucleotide induced migration in epithelial and dendritic cells [7, 10]. It is likely that different signals are needed for the initial communication of an injury and the long-term need for wound healing, such as cell migration and changes in protein expression. The receptor specificity and the numerous signaling pathways provide an intricate network allowing for differential regulation of short versus long term signals. Our study provides evidence that P2Y2 and possibly P2Y4 play a role in the initial signals involved in the injury response and these receptors and other P2 receptors play a role in long term cell migration. Future studies will involve understanding the critical interactions between signaling pathways that play a role in wound healing.

Anal_Bioanal_Chem-3-1-1914286

Improvement of the liquid-chromatographic analysis of protein tryptic digests by the use of long-capillary monolithic columns with UV and MS detection

Optimisation of peak capacity is an important strategy in gradient liquid chromatography (LC). This can be achieved by using either long columns or columns packed with small particles. Monolithic columns allow the use of long columns at relatively low back-pressure. The gain in peak capacity using long columns was evaluated by the separation of a tryptic bovine serum albumin digest with an LC–UV–mass spectrometry (MS) system and monolithic columns of different length (150 and 750 mm). Peak capacities were determined from UV chromatograms and MS/MS data were used for Mascot database searching. Analyses with a similar gradient slope for the two columns produced ratios of the peak capacities that were close to the expected value of the square root of the column length ratio. Peak capacities of the short column were 12.6 and 25.0 with 3 and 15 min gradients, respectively, and 29.7 and 41.0 for the long column with 15 and 75 min gradients, respectively. Protein identification scores were also higher for the long column, 641 and 750 for the 3- and 15-min gradients with the short column and 1,376 and 993 for the 15- and 75-min gradients with the long column. Thus, the use of long monolithic columns provides improved peptide separation and increased reliability of protein identification. Introduction Identification of unknown proteins is a key step in proteome analysis. The standard method of protein identification consists of enzymatic digestion of the protein(s), usually by using trypsin, followed by mass spectrometry (MS) analysis of the resulting digest. When analysing a single protein, e.g. from an excised 2D gel spot, a peptide separation is usually unnecessary. When analysing a digest of a mixture of proteins, the resolution of only a mass spectrometer is usually not sufficient. In those cases, a peptide separation is applied before MS detection. In shotgun proteomics, where a whole proteome is digested without prior protein fractionation, 2D liquid chromatography (LC) is the method of choice owing to the high complexity of the peptide mixtures. A drawback of multidimensional methods is the time frame, as typical analysis times for 2D-LC analysis of peptide mixtures range from several hours to more than 1 day [1–4]. 1D-LC methods can be used for separation of less complex samples (digests of fewer than 100 proteins), but this requires very efficient separations in combination with the additional separation power of MS detection. Such high-efficiency separations have been reported for analyses with long, packed columns, using (ultra-) high pressure (about 70 MPa) systems [5, 6]. Peak capacity is the primary parameter for evaluation of efficiency in gradient chromatography [7]. Peak capacity was first defined by Giddings [8] as the maximum number of peaks that can be separated by a phase system. This theory was later adapted for gradient chromatography by Horváth and Lipsky [9]. There are generally two approaches for optimisation of the peak capacity. The first is increase of the gradient length; gradients of up to 10 h have been reported on single columns [10]. However, according to theory, peak capacity increases to a maximum and then decreases as the gradients become longer [11, 12]. The second approach is the use of longer columns, as the peak capacity increases linearly with the square root of the plate number and thus with the square root of the column length. Wang et al. [13] illustrated this by connecting several columns packed with a pellicular stationary phase. They showed that the ratio of the peak capacity and the square root of the column length was constant for columns ranging in length from 7.5 to 60 cm. The main limitation to simply increasing column length is the higher back-pressure. A possible solution for this problem is the use of monolithic columns. Monolithic columns have a higher permeability compared with packed columns, facilitating fast separations or the use of long columns [10, 14–16]. Wang et al. [13] reported a back-pressure of 28.5 MPa for a 60 cm × 2.1 mm packed column at a linear flow of 0.94 mm/s. In contrast, other groups have obtained 3 times higher linear flow rates for monolithic columns of similar length [10, 17]. The higher permeability of monolithic columns makes it possible to operate long columns at higher flow rates while using conventional LC equipment. Luo et al. [10] separated a bacterial protein digest at a linear flow rate of 2.4 mm/s with a 70-cm monolithic column at 34.5 MPa (5,000 psi). Tolstikov et al. [17] analysed plant metabolomic extracts with a flow rate of 2.6 mm/s for a 60-cm column. The quality of the analysis of a protein digest can be expressed in different ways. One way is to describe chromatographic efficiency in terms of parameters like peak width, peak capacity and resolution. Another way is to use the reliability of protein identification, like SEQUEST [18] or Mascot (http://www.matrixscience.com) [19] scores. Such an approach is also useful, as the mass spectrometer adds a second dimension to the separation, which is overlooked when only chromatographic efficiency is measured. In this paper the evaluation of capillary monolithic silica columns of different lengths for the LC-UV-MS analysis of a bovine serum albumin (BSA) tryptic digest is described. Columns of 150- and 750-mm length were investigated using gradient times varying from 3 to 75 min. Chromatographic peak capacities, based on UV detection, and protein identification, based on Mascot scoring data of the MS detection, were determined as a measure for the efficiency of peptide separation. Theoretical aspects The peak capacity is defined as the maximum number of bands that will fit within a chromatogram with a resolution of Rs = 1.0 [20]. In gradient LC, where peak width is about constant throughout the separation, the theoretical peak capacity is given by Eq. 1: where tg is the gradient time and wav is the average baseline peak width. For large values of PC, this approaches tg/wav. Because sample peaks often do not occupy the entire length of the gradient, the sample peak capacity can then be defined as where ta and tz are the retention times of the first and last eluting peaks, respectively. Wang et al. [13] investigated the effect of column length on peak capacity in packed columns and found that the peak capacity is proportional to if the gradient slope is proportional to L. The gradient slope can be defined as where Δϕ is the change in organic modifier fraction during the gradient (0≤ϕ≤1) and t0 is the column dead time. If the linear flow rate is constant, t0 is only dependent on column length; therefore, tg should be varied proportionally to column length in order to keep the gradient slope constant. Experimental Materials and reagents BSA, trypsin (porcine, type IX-S, EC 3.4.21.4) and 1,4-dithiothreitol (DTT) were purchased from Sigma (St. Louis, MO, USA), and iodoacetamide (IAA) and NH4HCO3 from Fluka (Buchs, Switzerland). Acetonitrile was high-performance LC gradient grade (Biosolve, Valkenswaard, the Netherlands), and spectroscopy-grade trifluoroacetic acid (TFA) was obtained from Merck (Darmstadt, Germany). All solutions were prepared using water from a Milli-Q water-purifying system (Millipore, Bedford, MA, USA). All reagents for the digestion of BSA were prepared in 200 mM NH4HCO3 buffer, pH 8. The tryptic digest was prepared as follows. A 100-μl aliquot of BSA stock solution (3.5 μg/μl in water) was set to pH 8 by adding 25 μl of 1 M NH4HCO3 buffer (pH 8). After addition of 25 μl of 10 mM DTT solution, the sample was incubated at 50 °C for 30 min to reduce disulfide bonds. After cooling to room temperature, 25 μl of 30 mM IAA solution was added and the sample was incubated in the dark for 60 min to alkylate the free thiols. Trypsin was dissolved in 10 μl of buffer to reach a trypsin-to-protein mass ratio of 1:50 in the final solution; the trypsin solution was added to the sample, which was incubated at 37 °C overnight for 15 h. The digestion was stopped by addition of 15 μl of 10% TFA. The sample was diluted to 200 ng/μl (3 μM) with LC mobile phase A (water plus 0.05% TFA) and injected without further purification. Apparatus and LC columns All analyses were performed with an Agilent 1100 nanoLC system (Agilent Technologies, Waldbronn, Germany), consisting of a vacuum degasser, a binary Nano-Pump, a μ-well plate sampler and a column switching module with a trapping column in the 1–4 position of the six-port column-switching valve. The trapping pump was a Gynkotek model 480 (Gynkotek, Germering, Germany). Detection was performed by UV and MS detection, with the detectors connected in series. The UV detector was an MU 701 UV–VIS detector (ATAS GL International, Veldhoven, the Netherlands), equipped with an external optical-fibre flow cell (6 nl, 3-mm light path); peptides were detected at 215 nm. The mass spectrometer was an Agilent LC/MSD Trap XCT (Agilent Technologies, Waldbronn, Germany) ion-trap mass spectrometer, equipped with an orthogonal electrospray ionisation (ESI) interface. The external flow cell of the UV detector allows minimal time delay and band-broadening between UV and MS detection. The monolithic columns were provided by GL Sciences (Tokyo, Japan). The columns were a 150 mm × 0.1 mm MonoCap for nano-flow C18-silica monolithic column and a 750 mm × 0.2 mm MonoCap high resolution C18-silica monolithic column. For trapping of the digest a 5 mm × 0.3 mm column packed with 5 μm Zorbax 300 SB-C18 (Agilent Technologies, Waldbronn, Germany) was used. Method and data analysis LC solvent A was water plus 0.05% TFA (v/v); solvent B was acetonitrile plus 0.04% TFA (v/v). The trapping solvent was a mixture of 5% (v/v) solvent B in solvent A. After injection of the digest (0.25 μl for the 0.1-mm inner diameter column and 1.0 μl for the 0.2-mm inner diameter column), the sample was trapped on the trapping column at a flow rate of 5 μl/min. After 5 min, the trapping column was switched on-line with the separation column and the gradient was started. All separations were performed at room temperature using a gradient of 5–50% solvent B with gradient times varying from 3 to 75 min. MS spectra were acquired in the positive ion mode over the 400–2,000 m/z range, after which the two most intense ions (with a preference for doubly charged ions) were selected for fragmentation. MS/MS fragmentation spectra were acquired over the 100–2200 m/z range. An ESI spray voltage of -3 kV was used for all experiments. The effect of separation efficiency on protein identification was evaluated using the Mascot search engine [19]. LC-MS/MS data were converted to the Mascot generic format (.mgf file) using the data-analysis software, and the .mgf files were searched against the MSDB database using Mascot’s MS/MS ion search module. The database was searched for tryptic peptides from all entries in the database, allowing one missed cleavage per peptide and containing carbamidomethyl cysteine as a variable modification. Mass tolerances were set to default values: peptide mass tolerance ±2.0 Da, MS/MS tolerance ±0.8 Da. Results and discussion Liquid chromatography–UV analysis Because of the difference in diameter, the 150 mm × 0.1 mm and the 750 mm × 0.2 mm columns were used with different flow rates. For the 150- and the 750-mm columns, the flow rates were 0.5 and 2.0 μl/min, respectively, resulting in a linear flow rate of 1.06 mm/s. Injection volumes were also proportional to the square of the column diameter, 0.25 μl of the digest for the 0.1-mm column and 1.0 μl onto the 0.2-mm column. During the gradient, the maximum back-pressure of the 750-mm column was below 20 Mpa, which is well below the manufacturers limit of 30 Mpa. Figure 1 shows the LC-UV chromatograms of 3- and 15-min gradients run on the 150-mm column and 15- and 75-min gradients run on the 750-mm column. When the chromatograms of the analyses with similar gradient slopes are compared (Fig. 1a,b, and Fig. 1c,d), it is clear that an increase in column length improves the peptide separation. In order to quantify the efficiency of the separation, the sample peak capacity was calculated for all analyses. Because of the incomplete resolution of the digest, the peak capacity was estimated by using the average peak width of a selected number of peaks that appeared to contain only a single peptide. Using this method, we calculated peak capacities for all analyses and the results are summarised in Table 1. The peak capacities found for the short column are comparable to those found in the literature for similar columns [21, 22]. As expected, the peak capacities of the long column are higher than those of the short column, but they are relatively low compared with the values reported in [10]. However, when gradient time is taken into consideration, the difference is significantly less: PC**/tg is 0.55 peaks per minute with the present system (75-min gradient on the 750-mm column) and 1.62 peaks per minute for the 260-min gradient reported in [10]. A possible negative effect on the resolution of our system is the use of a trapping column, filled with a different stationary phase, which probably has a selectivity that differs from that of the analysis column. Fig. 1Liquid chromatography (LC)–UV chromatograms of a tryptic bovine serum albumin (BSA) digest, separated by monolithic silica capillary columns of 150 mm × 0.1 mm (a, c) and 750 mm × 0.2 mm (b, d) using a gradient of 5–50% acetonitrile (0.04% v/v trifluoroacetic acid, TFA) in water (0.05% v/v TFA). a A 3-min gradient (15%/min); b, c 15-min gradient (3%/min); d 75-min gradient (0.6%/min). The run times include 5 min of trapping and 2.5 and 12.5 min of gradient delay time for the 150-mm and 750-mm columns, respectivelyTable 1Chromatographic parameters from liquid chromatography–UV analysis 150 mm × 0.1 mm750 mm × 0.2 mmtg (min)wavΔtPC**wavΔtPC30.243.0512.60.283.1611.2150.4411.025.00.3911.629.7751.2737.629.60.7731.741.0tg gradient time, wav average width of seven to 12 peaks which were found to contain only a single peptide (based on mass spectrometry data), Δt elution window between the first and last eluting peptides, PC** sample peak capacity (Δt/wav) If chromatograms with the same gradient slope are compared (3- and 15-min gradients for the 150-mm column and 15- and 75-min gradients for the 750-mm column, respectively), the ratio of the peak capacities should be close to the square root of the column length ratio (2.24). For the 3-/15-min gradient pair this ratio is 29.7/12.6 = 2.37; for the 15-/75-min gradient pair the ratio is 41.0/25.0 = 1.64. For the short column, no increase in peak capacity is observed with a gradient of more than 15 min, but for the long column the peak capacity increases up to a 75-min gradient. A possible explanation can be found in the study of Stadalius et al. [11, 12], who demonstrated that peak capacity will increase with gradient time, until it reaches a maximum, after which it will even decrease. The gradient time at which this maximum is obtained is greater for longer columns. Liquid chromatography–mass spectrometry analysis To assess the influence of peptide separation on protein identification, MS/MS data were investigated by database-searching using the MS/MS ion search module from the Mascot search engine. The results are expressed as a Mascot score, the number of unique identified peptides and the percentage of the BSA amino acid sequence covered by these peptides (Table 2). These results were compared with the scores obtained for direct infusion of the BSA digest at the same flow rates as for the LC separations. The protein identification parameters for the infusion experiments were about similar for both flow rates. All database searches gave bovine albumin as the top protein match and the only other significant matches were albumins from other species. Table 2Mascot® database search results 150 mm × 0.1 mm750 mm × 0.2 mmtgMascot scorePeptidesaCoverageb (%)Mascot scorePeptidesaCoverageb (%)Infusionc31791737610183 min6411428787132515 min75016291,376243875 min69316299931929aNumber of unique peptides identified by database search.bPercentage of amino acid sequence covered by identified peptides.cDirect infusion of the digest into the mass spectrometer. Figure 2 shows base peak chromatograms of separations with the same gradient slope, a 15-min gradient for the 150-mm column and a 75-min gradient for the 750-mm column. The Mascot scores for the 750-mm column are higher than those for the 150-mm column. The average score per identified peptide is between 52 and 60 for the long column and between 42 and 51 for the short column. Combined with the larger number of identified peptides on the long column, this adds up to a higher Mascot score (Table 2). Fig. 2LC–mass spectrometry (MS) base peak chromatograms and mass spectra of tryptic BSA digest, separated by monolithic silica columns of 150 mm × 0.1 mm (a) and 750 mm × 0.2 mm (b), using a gradient of 5–50% acetonitrile (0.04% v/v TFA) in water (0.05% v/v TFA). a A 15-min gradient (3%/min); b 75-min gradient (0.6%/min). The run times include 5 min of trapping and 2.5 and 12.5 min of gradient delay time for the 150-mm and 750-mm columns, respectively Despite providing only a low resolution, even the shortest gradients cause a considerable increase in the reliability of protein identification. Increase in gradient length leads to cleaner mass spectra (Fig. 3) and a higher number of identified peptides and sequence coverage, compared with direct infusion. These numbers, however, decrease beyond a certain gradient time. This could possibly be attributed to the lower peak heights for longer gradient times. This result indicates that there is a gradient slope where an increase in chromatographic separation no longer improves protein identification. Fig. 3Averaged mass spectra of peptide YICDNQDTISSK (m/z 722.32, M2H2+) as identified from extracted ion chromatograms in the LC-MS analysis of a tryptic BSA digest. a A 150-mm × 0.1-mm silica monolithic column, 15-min gradient of 5–50% acetonitrile (0.04% v/v TFA) in water (0.05% v/v TFA); b 750-mm × 0.2-mm column, 75-min gradient Conclusions The use of long silica-based capillary monolithic columns provides a clear advantage over use of shorter columns, i.e. an increase of chromatographic efficiency and reliability of protein identification. As expected from chromatography theory, a factor 5 longer column gives a 1.6–2.4 times increase in peak capacity for separations with similar gradient slope. The use of longer gradients also leads to an initial improvement of the protein identification score, but the score seems to have a maximum at longer gradient times. While the use of longer columns for the separation of peptides has a clear advantage because of the gain in chromatographic efficiency, this also gives a longer analysis time. As maximum protein identification scores for rather simple digests are reached at relatively short gradient times, it is important to find a compromise between chromatographic efficiency and analysis time. However, if the sample is more complex, the use of longer columns is more attractive as longer gradients are necessary to achieve sufficient separation. In the near future, short and long columns of the same diameter (0.1-mm inner diameter) will be compared. Further improvement of the separation might be obtained by optimisation of the combination of the trap column and the analysis column. Moreover, the potential of longer monolithic capillary columns will be demonstrated by the analysis of more complex and real samples.

J_Mol_Biol-2-1-2267215

Structure of the Tandem Fibronectin Type 3 Domains of Neural Cell Adhesion Molecule

Activation of the fibroblast growth factor receptor (FGFR) by neural cell adhesion molecule (NCAM) is essential for NCAM-mediated neurite outgrowth. Previous peptide studies have identified two regions in the fibronectin type 3 (FN3)-like domains of NCAM as being important for these activities. Here we report the crystal structure of the NCAM FN3 domain tandem, which reveals an acutely bent domain arrangement. Mutation of a non-conserved surface residue (M610R) led to a second crystal form showing a substantially different conformation. Thus, the FN3 domain linker is highly flexible, suggesting that it corresponds to the hinge seen in electron micrographs of NCAM. The two putative FGFR1-binding segments, one in each NCAM FN3 domain, are situated close to the domain interface. They form a contiguous patch in the more severely bent conformation but become separated upon straightening of the FN3 tandem, suggesting that conformational changes within NCAM may modulate FGFR1 activation. Surface plasmon resonance experiments demonstrated only a very weak interaction between the NCAM FN3 tandem and soluble FGFR1 proteins expressed in mammalian cells (dissociation constant > 100 μM). Thus, the NCAM–FGFR1 interaction at the cell surface is likely to depend upon avidity effects due to receptor clustering. Introduction The neural cell adhesion molecule (NCAM) is the prototype and founding member of the immunoglobulin (Ig) superfamily cell adhesion molecules (CAMs).1–4 NCAM is present on the cell surface of neurons, astrocytes and oligodendrocytes, where it mediates homophilic and heterophilic cell adhesion. NCAM is involved in neuronal migration, axon growth and guidance, as well as in synaptic plasticity associated with learning and memory.5–7 Alternative splicing of the NCAM1 gene results in isoforms of three size classes that differ in their membrane attachment and cytosolic regions but have in common an extracellular domain consisting of five Igs and two fibronectin type 3 (FN3) domains.8 The two larger isoforms have a transmembrane helix and cytosolic domains of different sizes, while the smallest isoform has a glycophosphatidylinositol membrane anchor. Variable use of alternative exons in the extracellular domain results in small insertions into Ig4 or between the FN3 domains.9–12 NCAM function is further regulated by an unusual posttranslational modification, namely, the addition of polysialic acid to Ig5.13 The molecular basis of homophilic adhesion by NCAM has been a subject of intense study, and the results from biochemical and biophysical studies have not always been consistent.14,15 A crystal structure of NCAM Ig1–Ig3 has led to a zipper model of adhesion, which postulates both cis and trans interactions by the Ig1–Ig3 region of NCAM.16 While a conclusive picture has yet to emerge, it appears that there may be multiple modes of homophilic interaction.17 NCAM is also engaged in heterophilic interactions. There is now convincing evidence that NCAM-mediated neurite outgrowth, as well as tumour development and progression, critically involves the activation of fibroblast growth factor receptor 1 (FGFR1), through a cis interaction of NCAM and FGFR1.18–22 The four FGFRs and their 23 growth factor ligands control a variety of cellular processes, including development, angiogenesis, hematopoiesis and tumourigenesis.23–26 Alternative splicing of the four FGFR genes results in at least 48 receptor isoforms that vary in their ligand binding profiles and kinase domains. The longest FGFR1 ectodomain variant consists of three Ig domains, D1–D3, with a stretch of acidic amino acids (the “acid box”) inserted into the D1–D2 linker; shorter forms lack D1 and the acid box. The NCAM–FGFR1 interaction was originally proposed based on indirect biological evidence,27 but biochemical evidence has been obtained to suggest that the interaction is mediated by direct binding of the FN3 domains of NCAM to FGFRs (Fig. 1).20,28,29 We and others identified NCAM-derived peptides capable of stimulating FGFR1 signalling and inducing neurite outgrowth. One peptide (FRM peptide) is derived from the first FN3 domain of NCAM (1FN3),30 while another (FGL peptide) is derived from the second FN3 domain (2FN3).28 The structures of the 1FN3 and 2FN3 domains in isolation have been determined,28,31 but how the two domains cooperate in FGFR1 activation is unclear. In this study, we determined the crystal structure of the NCAM FN3 tandem (1FN3–2FN3) in two crystal forms. We report that 1FN3–2FN3 assumes a bent conformation in both forms, with evidence of substantial flexibility of the domain linker. In direct binding experiments with fully glycosylated proteins of mammalian origin, we observed only a very weak interaction (dissociation constant > 100 μM) of NCAM 1FN3–2FN3 with two FGFR1 ectodomain constructs. We conclude that the NCAM–FGFR1 interaction at the cell surface may be transient or stabilised by avidity effects resulting from receptor clustering and that conformational changes within NCAM may have a profound role in FGFR1 activation. Results Crystal structure of the NCAM FN3 tandem To obtain insight into the relative orientation of the two FN3 domains of NCAM, we determined the crystal structure of 1FN3–2FN3 at 2.3-Å resolution (Table 1). Both 1FN3 and 2FN3 adopt the typical β-sandwich fold of all FN3 domains consisting of seven strands arranged in two antiparallel sheets (ABE and GFCD) (Fig. 2a and b). Preceding strand A in both domains are short proline-rich segments that are integrated into the FN3 fold, with the proline tetrahydropyrrole rings pointing into the hydrophobic core (Pro500 and Pro503 in 1FN3; Pro601 and Pro604 in 2FN3). A similar feature has been observed in other FN3 domains (e.g., in gp13033 and titin34). 1FN3 contains an unusual α-helix situated between strands D and E, as reported previously.31 1FN3 in our FN3 tandem structure matches the crystal structure of 1FN3 in isolation,31 with an r.m.s.d. of 0.50 Å for 100 Cα atoms. 2FN3 in our FN3 tandem matches the solution structure of 2FN3 in isolation,28 with an r.m.s.d. of 1.3 Å for 92 Cα atoms (the main differences are concentrated in the B–C and C–D loops). The relative orientation of the two FN3 domains in the 1FN3–2FN3 tandem is characterised by an unusually bent conformation with an interdomain angle of ∼ 80° (calculated between the long axes of the two FN3 domains). The interface between 1FN3 and 2FN3 buries 630 Å2 of solvent-accessible surface (calculated with the CCP4 program AREAIMOL), which is in the typical range for rigid FN3 (and Ig) domain interfaces. The interface is dominated by polar interactions, between the A–B loop of 1FN3 on the one hand and the domain linker and B–C loop of 2FN3 on the other (Fig. 2c) The key interface residues (Tyr511, Ser513, Thr514, Pro601 and Asp625) are strictly conserved in all vertebrate NCAM sequences (Fig. 2d), suggesting that the bent conformation may be physiologically relevant. However, residues 511–514 have also been implicated in FGFR1 binding,30 and the linker region could assume a very different structure when NCAM is bound to FGFR1 (see below). The conformation of multidomain proteins often is influenced by the crystal lattice. In this regard, we noted that the asymmetric unit of our crystals contains two 1FN3–2FN3 molecules arranged in a tightly interlocked dimer (Fig. 2b). There are two main contacts responsible for dimer formation: the α-helix of 1FN3 packs against the GFCD sheet of 2FN3 of the other molecule, and the 2FN3 domains of the two molecules interact via their A and G strands. Altogether, these contacts bury as much as 3040Å2 of solvent-accessible surface. We do not think that the dimer observed in our crystals is physiologically relevant, as dimer interface residues are only poorly conserved (Fig. 2d) and NCAM dimers were never observed in electron microscopy studies.35,36 Because we were concerned that the tight association of 1FN3–2FN3 molecules in the dimer may have forced the unusually bent interdomain conformation, we sought to disrupt the dimer by mutagenesis and crystallise a mutant 1FN3–2FN3 protein in a different crystal form. Structure of the M610R mutant We expressed three point mutants of NCAM 1FN3–2FN3 (M610R, Y672E and R690E), all of which should be incompatible with the dimer structure seen in crystals of the wild-type protein. Importantly, all three mutations target surface residues and are unlikely to have an effect on NCAM folding. Whether the mutations are functionally neutral could not be determined due to a lack of suitable assays. When examined by size-exclusion chromatography, all three mutants eluted as a mixture of monomers and dimers, similar to the wild-type protein (data not shown). We obtained crystals of the M610R mutant and determined its structure at 2.7-Å resolution by molecular replacement (Fig. 3a). The asymmetric unit of the crystals contains a hexamer of NCAM 1FN3–2FN3 M610R. The hexamer can be regarded as a trimer of dimers, with the dimers having a completely different mode of association compared with the wild-type structure (the individual domains are very similar, as expected, with r.m.s.d. values of 0.93 and 1.1 Å for 1FN3 and 2FN3, respectively). The relative orientation of the two FN3 domains in the M610R mutant is less severely bent than in the wild-type structure but still far from fully extended (interdomain angle ∼ 120°; Fig. 3b). The FN3 pair opens up in the M610R mutant, and 1FN3 is additionally twisted about its long axis. The combined hinge opening and 1FN3 twisting amounts to a pure rotation of 73° for 1FN3 when the structures are superimposed on their 2FN3 domains. There is no domain interface to speak of in the mutant (240 Å2 buried), and the conformation appears to be stabilised entirely by the crystal lattice. Thus, crystal lattice forces can have a profound influence on the conformation of the NCAM FN3 tandem, suggesting that the 1FN3–2FN3 linker may act as a flexible hinge in native NCAM. We compared the NCAM FN3 tandem with other FN3 tandems of known structure. Tandems from the extracellular matrix proteins fibronectin and tenascin generally assume an extended conformation.38–40 In contrast, many cytokine and hormone receptors (e.g., gp130)33 feature bent FN3 pairs that are superficially similar to the NCAM structures reported here (data not shown). Of particular interest is the structure of the FN3 pair of neuroglian, a Drosophila CAM.37 The neuroglian tandem, which has an extensive domain interface that incorporates a bound sodium ion, adopts a conformation that is intermediate between the two conformations we report for the NCAM tandem. Thus, a similar bend in the FN3 linker(s) may be a general feature of other animal CAMs containing FN3 domains in their membrane-proximal region. Location of putative FGFR1 binding site Previous studies have implicated two NCAM 1FN3–2FN3 regions in FGFR1 binding. Kiselyov et al. identified a bioactive peptide from the F–G loop of 2FN3 (FGL peptide),28 and we identified a bioactive peptide from the A–B loop of 1FN3 (FRM peptide).30 In agreement with our earlier prediction,30 the FRM and FGL loops are indeed located in close proximity on the same face of the wild-type NCAM molecule (Fig. 4a). However, in the M610R mutant, the opening up of the two domains combined with rotation of 1FN3 places the FRM and FGL loops much farther apart. Thus, conformational changes at the NCAM 1FN3–2FN3 hinge may modulate the interaction of NCAM with FGFR1. Surface plasmon resonance analysis of the NCAM–FGFR1 interaction We wanted to map the FGFR1 binding site on the NCAM FN3 tandem by structure-based mutagenesis and first sought to establish a suitable binding assay. A solid-phase assay with immobilised NCAM and Fc-tagged FGFR1 proteins did not show any appreciable interaction (data not shown). We therefore used surface plasmon resonance (SPR) to analyse the binding of NCAM 1FN3–2FN3 to two FGFR1 ectodomain constructs. The FGFR1 D1–D3 construct used (residues 22–364) spans essentially the full ectodomain and contains the acid box situated between domains D1 and D2. The FGFR1 D2–D3 construct used (residues 151–364) lacks D1 and the acid box but retains the binding site for FGFs; this construct is similar to the construct previously used by Kiselyov et al. in SPR studies.28 Both soluble FGFR1 proteins were produced by the 293-EBNA cells in good yields. Due to the presence of multiple N-linked glycosylation sites in FGFR1 (see below), the purified recombinant proteins migrate as diffuse bands of higher-than-calculated molecular mass on SDS-PAGE (Fig. 5). In a first set of experiments, the two FGFR1 constructs were immobilised on a CM5 sensor chip [8000 resonance units (RU) of D1–D3 and 3850 RU of D2–D3]. Recombinant FGF1 injected at a concentration of 100 nM produced sensorgrams characteristic of a high-affinity interaction, confirming that the immobilised proteins are functional (Fig. 6a and b). In contrast, wild-type NCAM 1FN3–2FN3 up to a concentration of 70 μM did not produce a signal on the FGFR1 D1–D3 surface and showed only very weak binding to FGFR D2–D3 (Fig. 6c and d). In a second set of experiments, the order of proteins was reversed. NCAM 1FN3–2FN3 proteins were immobilised on a CM4 sensor chip (1800 RU of wild-type protein and 1900 RU of M610R mutant), and the two soluble FGFR1 constructs were used as analytes up to a concentration of 100 μM. We again observed only weak interactions for all pairings (Fig. 6e–h). Wild-type and M610R NCAM 1FN3–2FN3 behaved almost identically in these experiments, and, as before, it appeared that the affinity of NCAM for FGFR1 D2–D3 was higher than that for FGFR1 D1–D3. The fast association and dissociation steps in the sensorgrams prevented the fitting of kinetic constants. We used the plateau values at equilibrium to estimate a dissociation constant of > 100 μM for the interaction of NCAM 1FN3–2FN3 with FGFR1 D2–D3 (not shown), but we emphasise that this value is very approximate given the weak resonance signals obtained. In view of the weakness of the NCAM–FGFR1 interaction in our assay, we were unable to pursue our initial plans of mapping the binding site(s) by mutagenesis. Discussion NCAM ectodomain structure The current view of the NCAM ectodomain structure is based on early studies by rotary shadowing electron microscopy of tissue-derived NCAM.35 Electron microscopy visualised the NCAM ectodomain as ∼ 28-nm rods bent at a flexible hinge located ∼ 10 nm from the C-terminus; the angle between the two arms varied from 50° to 140° (average = 100°). The hinge was attributed to the proline-rich linker between Ig5 and the first FN3 domain, and the long and short arms were described as rigid domain tandems (Ig1–Ig5 and 1FN3–2FN3, respectively) in extended conformations.35 A crystal structure of NCAM Ig1–Ig3 indeed showed a largely extended structure.16 In sharp contrast, the present crystal structure analysis of the NCAM 1FN3–2FN3 domain pair has revealed a prominent bend between the two domains in two independent crystal forms (Fig. 3b). This finding is difficult to reconcile with the uniformly straight appearance of the short arm in the electron micrographs and suggests that the hinge point may actually lie between 1FN3 and 2FN3. In this respect, we note that the Ig5–1FN3 linker is actually rather short and may well be rigid (in the Ig5–1FN3 linker sequence, ILVQADTPSSP, the isoleucine and valine residues are predicted to contribute to the hydrophobic core of Ig5 and the first proline is already part of the 1FN3 fold). If the hinge is instead situated in the 1FN3–2FN3 linker, the juxtamembrane domain of NCAM must have contributed to the short arm seen in electron micrographs, as a single FN3 domain would only account for half of the short arm length. The serine/threonine-rich juxtamembrane domain of NCAM (sequence TSAQPTAIPANGSPTSGLSTGA) is predicted to be extensively modified by O-linked glycosylation (NetOGlyc 3.1 server†) and could easily assume the extended and rigid conformation required to span the remaining ∼ 5 nm.41 Further structural analysis, in particular, of the Ig5–1FN3 pair, is required to conclusively pinpoint the site of articulation within the NCAM ectodomain. NCAM splice variants Interestingly, the flexible hinge linking the two FN3 domains of NCAM 1FN3–2FN3 is known to be modified by alternative splicing. Our structure is of the shortest isoform (linker sequence TQPVREPSAP), whereas the underlined arginine residue is replaced by QG, HSPPPQG or even longer sequences in other variants.9,10,12 The HSPPPQG insertion has been suggested as a potential hinge region.1 The biological relevance of these splicing events is not well understood, but in vitro experiments have demonstrated that the NCAM isoforms differ in their capacity to support cell adhesion and spreading,42,43 as well as myoblast fusion.44 We prepared NCAM 1FN3–2FN3 proteins with QG or HSPPPQG inserts in the domain linker, but, unfortunately, these proteins were very prone to aggregation in physiological buffers and could not be used for SPR or structural analysis (data not shown). Interaction of NCAM with FGF Hinge bending and alternative splicing at the 1FN3–2FN3 junction could affect NCAM function by modulating either its homophilic binding properties or its heterophilic interactions with other proteins. There is currently no evidence for the former scenario. In contrast, the functional interaction of NCAM with FGFR1 in cis (i.e., at the same cell membrane) is well established1,22,30 and would appear to be an attractive candidate for regulation by alternative splicing. Peptides from two regions of the NCAM 1FN3–2FN3 tandem, one in each domain, have been shown to modulate FGFR1-dependent neurite outgrowth.28,30 Intriguingly, the corresponding loop regions are in close proximity in the acutely bent conformation observed in crystals of the wild-type protein, suggesting that they are part of a larger FGFR1 binding site extending over the domain junction. In crystals of the M610R mutant, the two putative FGFR1-binding loops are farther apart and no longer on the same face of the 1FN3–2FN3 structure (Fig. 4). Thus, changes in the 1FN3–2FN3 conformation, either by biomechanical forces resulting from cell–cell contact or by alternative splicing, could have a profound effect on the NCAM–FGFR1 interaction. Other NCAM activities that might be affected by alternative splicing of the 1FN3–2FN3 linker include interactions with polysialyltransferases31 and prion protein,45 which both bind to the FN3 domains. Previous biochemical studies reported a dissociation constant of ∼ 10 μM for the interaction between NCAM 1FN3–2FN3 and FGFR1 D2–D3.28,29 Another study, using cell-based assays, concluded that the acid box in the FGFR1 D1–D2 linker was essential for the NCAM–FGFR1 interaction.20 We wanted to use SPR binding experiments to identify NCAM residues involved in FGFR1 binding and test the effect of splice inserts in the 1FN3–2FN3 linker. Unfortunately, using our recombinant proteins expressed in mammalian cells (Fig. 5), we were unable to detect substantial binding between soluble FGFR1 and NCAM proteins, regardless of whether FGFR1 or NCAM was immobilised on the sensor chip. A very weak interaction (estimated dissociation constant > 100 μM) was evident between FGFR1 D2–D3 and NCAM 1FN3–2FN3, but we observed no NCAM binding to the FGFR1 D1–D3 construct containing the acid box (Fig. 6). The most likely explanation for the discrepancy between our findings and those of Christensen et al.29 is the difference in glycosylation of the FGFR1 proteins used. Christensen et al. expressed FGFR1 D2–D3 in insect cells, which produce N-linked oligosaccharides of the high-mannose type, whereas our expression system (human embryonic kidney cells) produces complex-type oligosaccharides, which more closely resemble the glycan present on mammalian FGFRs. Human and rodent FGFR1 proteins are highly glycosylated, and the glycan is known to influence ligand binding.46 We think that there may be electrostatic repulsion between the acidic NCAM 1FN3–2FN3 protein (isoelectric point 5.0) and the terminal sialic acids present on FGFR1 expressed in human cells. The very weak interaction we observed between soluble NCAM and FGFR1 proteins does not preclude a critical role of this interaction at the cell surface, where avidity effects due to receptor clustering may be substantial. An attractive hypothesis is that the functional state of NCAM (i.e., whether it is engaged in a homophilic contact or not) is linked to FGFR1 binding and activation. It is tempting to speculate that the flexible 1FN3–2FN3 linkage revealed by our structural analysis could provide the molecular means for such a regulatory mechanism. Experimental Procedures Expression vectors NCAM constructs were made by PCR amplification from a bacterial expression vector coding for the FN3 pair of human NCAM. Our NCAM numbering scheme corresponds to SwissProt entry P13591 up to residue 598 but differs by − 1 from P13591 for all subsequent residues due to the replacement of Gln599–Gly600 by Arg, a naturally occurring splice variant in the brain and muscle.12 The M610R mutation in NCAM was introduced by strand-overlap-extension PCR. FGFR1 constructs were made by PCR amplification from a complete cDNA clone of human FGFR1 (IIIc isoform; our numbering scheme corresponds to SwissProt entry P11362). The PCR products were cloned into a modified pCEP-Pu vector47 coding for proteins with a C-terminal His6 tag. The insert sequences of all expression vectors were verified by DNA sequencing. The domain boundaries of the constructs are as follows: NCAM 1FN3–2FN3, QADTP…VFRTS (496–692); FGFR1 D1–D3, RPSPT…EALEE (22–364); and FGFR1 D2–D3, VAPYW…EALEE (151–364). Vector-derived APLA and AAAHHHHHH sequences are additionally present at the N- and C-terminus, respectively. Protein expression and purification All proteins were purified from the conditioned medium of episomally transfected 293-EBNA cells. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Invitrogen), transfected using Fugene reagent (Roche Applied Science) and selected with 1 μg/ml of puromycin (Sigma). Proteins were purified by a combination of affinity and size-exclusion chromatography performed on an Äkta platform (GE Healthcare). Typically, 1.5 l of conditioned serum-free medium was loaded onto a 5-ml HisTrap column (GE Healthcare) equilibrated in phosphate-buffered saline (PBS) buffer, pH 7.45 (140 mM NaCl, 10 mM Na2PO4 and 3 mM KCl), and eluted with 500 mM imidazole in PBS. The eluate was concentrated using Vivaspin centrifugal devices (Sartorius AG) and further purified on a 24-ml Superdex 200 size-exclusion chromatography column (GE Healthcare) with Tris-buffered saline (TBS) buffer, pH 7.4, as the running buffer. Purified proteins were analysed by SDS-PAGE, quantified by measuring their absorption at 280 nm, concentrated to the final desired concentrations and flash-frozen in liquid nitrogen for storage at − 80 °C. Final yields were 10–20 mg of pure protein per litre of cell culture medium. Crystallisation and structure determination NCAM 1FN3–2FN3 was concentrated to 13 mg/ml in TBS, and crystals were obtained by hanging drop vapour diffusion at room temperature using 2.2 M ammonium sulfate, 0.1 M sodium citrate, pH 5.2, 0.2 M potassium/sodium tartrate and 3–5% ethanol as precipitant. Crystals grew within 2 days and belong to space group P212121 with unit cell dimensions a = 52.77 Å, b = 71.35 Å and c = 98.22 Å. There are two 1FN3–2FN3 molecules in the asymmetric unit, resulting in a solvent content of ∼ 38%. Crystals were flash-frozen in liquid nitrogen after brief soaking in mother liquor supplemented with 20% glycerol. A crystal was soaked in mother liquor supplemented with 300 mM potassium iodide for 30 s before freezing to obtain a heavy atom derivative. Diffraction data from native and KI derivative crystals were collected at 100 K on station 14.1 at the Synchrotron Radiation Source (SRS) Daresbury and on station ID29 at the European Synchrotron Radiation Facility Grenoble, respectively. The NCAM 1FN3–2FN3 M610R mutant was concentrated to 14 mg/ml in TBS, and crystals were obtained by sitting drop vapour diffusion at room temperature using 2 M ammonium sulfate and 0.1 M sodium acetate, pH 4.6, as precipitant. Crystals grew within 3–4 days and belong to space group P212121 with unit cell dimensions a = 92.74 Å, b = 107.49 Å and c = 161.18 Å. There are six copies of mutant 1FN3–2FN3 in the asymmetric unit, resulting in a solvent content of ∼ 42%. Crystals were flash-frozen in liquid nitrogen after brief soaking in mother liquor supplemented with 20% glycerol, and diffraction data were collected at 100 K on station 10.1 at the SRS Daresbury. The diffraction data were processed with MOSFLM‡ and programs of the CCP4 suite.48 The structure of NCAM 1FN3–2FN3 was solved by single-wavelength anomalous dispersion phasing of a KI-soaked crystal using SHARP (Globalphasing Ltd., Cambridge) in full automatic mode. The structure was rebuilt with O49 and refined with Crystallography & NMR System.50 The structure of the NCAM M610R mutant was solved with some difficulty by molecular replacement with PHASER,51,52 using the isolated FN3 domains of the NCAM 1FN3–2FN3 structure as search models. Data collection, phasing and refinement statistics are summarised in Table 1. The figures were made with PyMOL§. SPR experiments Binding experiments were performed on a Biacore 3000 instrument (GE Healthcare) at 25 °C. Proteins were immobilised on activated CM4 or CM5 chips using standard amine coupling procedures following the manufacturer's instructions. Briefly, flow cells were activated with 20 μl of a mixture of 0.2 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and 0.05 M N-hydroxy-sulfosuccinimide at a flow rate of 5 μl/min. The proteins to be immobilised (50–100 μg/ml in 10 mM sodium acetate, pH 4.5–5.5) were allowed to pass over activated flow cells to reach ∼ 2000–8000 RU, after which unreacted groups were blocked with 20 μl of 1 M ethanolamine, pH 8.5. Reference flow cells without protein were treated identically. The chips were equilibrated in 10 mM Hepes, pH 7.4, 150 mM NaCl, 50 μM ethylenediaminetetraacetic acid and 0.005% surfactant P20 (HBS-EP buffer), and serial dilutions of analyte proteins in PBS (our recombinant proteins or FGF1 from PeproTech) were injected at 20 μl/min for 300 s, followed by 200 s of pure buffer to monitor dissociation. Chips were regenerated using HBS-EP with 1 M glycine–HCl, pH 3.5. The sensorgrams were analysed with the BiaEvaluation 4.1 software. Protein Data Bank accession codes Coordinates and structure factors for wild-type and M610R NCAM 1FN3–2FN3 have been deposited in the Protein Data Bank with codes 2vkw and 2vkx, respectively.

Int_Arch_Occup_Environ_Health-4-1-2413125

The influence of psychosocial factors at work and life style on health and work ability among professional workers

Objectives The purpose of this article is to explore the associations of psychosocial factors at work, life style, and stressful life events on health and work ability among white-collar workers. Introduction Many western countries face the challenge of an aging population, which also affects the workforce. From the biological perspective, aging means a progressive deterioration in various physiological systems, which is accompanied by changes in physical and mental capacities of workers (Ilmarinen, 1997). Aging of the workforce will result in an increased prevalence of work-relevant symptoms and diseases. Therefore, the role of (functional) health in working life is of interest, especially since modern welfare states are prolonging working life by increasing the statutory retirement age. A recent study on the relation between health and working life showed that a perceived poor health predicts staying or becoming unemployed (Schuring et al. 2007). This calls for better adjustments of the working life demands with the individual’s health as a crucial element for a longer career at work. Within this framework, the concept of work ability has been developed as an important tool to identify workers at risk for imbalance between health, capabilities and demands at work. The work ability concept is based on the assumption that work ability is determined by an individual’s perception of the demands at work and the ability to cope with them. The work ability index (WAI) is a well-accepted instrument to conceptualize work ability. Several studies have shown that a low score on the index is highly predictive of work disability during follow-up (Liira et al. 2000; Tuomi et al. 1991). Previous research, predominantly in physically demanding jobs, showed that the WAI is negatively influenced by older age, high physical work demands, high psychosocial work demands (e.g. lack of possibilities to control one’s own work), unhealthy lifestyle (lack of physical activity), and a poor physical fitness (Alavinia et al. 2007; Ilmarinen et al. 1997; Pohjonen, 2001a; Tuomi et al. 2001). Few studies have addressed determinants of work ability in occupational populations with predominantly mental demands at work. Among office workers Sjögren-Rönkä (2002) showed that low stress at work and a better self-confidence were directly related to higher work ability. Seniority in the job and job satisfaction were also associated with a better work ability among office workers (Martinez and Latorre 2006). However, the knowledge of determinants of work ability in mentally demanding occupations is scarce and hence, it remains unclear whether in these jobs the relative importance of personal and work-related factors is similar to their well-known contribution in physically demanding jobs. The purpose of this study was to explore the associations of psychosocial factors at work, stressful life events, and life style on health and work ability among white-collar workers. Methods Subjects In the period between 2003 and 2007 a total of 2,666 white-collar workers from six companies in commercial services were invited for a health examination. Twenty percent of the subjects were employed at three consultancy firms, 62% at two insurance companies and 18% at an information technology company. The health examination consisted of two parts, i.e. a questionnaire and a physical examination. Both parts were offered independently to workers and their participation was entirely voluntary. The response to the questionnaire was 69.4% (n = 1,850). The response to the physical examination was 67.8% (n = 1808). Selection of subjects with both a filled out questionnaire and a physical examination comprised the study population of 1141 (42.8%) subjects. Work ability Work ability was measured with the work ability index (WAI). The WAI consists of an assessment of the physical and mental demands on an individual in relation to his work, previously diagnosed diseases, limitations in work due to disease, sick leave, work ability prognosis, and psychological resources. The WAI consists of seven dimensions and the index is derived as the sum of the ratings on these dimensions. The range of the summative index is 7–49, which is classified into a poor (7–27), moderate (28–36), good (37–43), or excellent (44–49) work ability (Tuomi et al. 1998). Functional health status Functional health status was assessed using the Short-Form Health Survey (SF-12) version 2, the shortened alternative for the 36-item health survey. This measure provides two weighted summary scores assessing physical function (physical health component summary, PCS) and mental well-being (mental health component summary, MCS) (Ware et al., 2002). The mental health summary score ranges from 8 to 74, whereas the physical health summary score ranges from 4 to 73, with a higher score indicating a better health state. Psychosocial factors at work Psychosocial factors at work were measured by the stress monitor (Petri et al. 2001). The original monitor consists of four dimensions, whereas three dimensions (teamwork, stress handling, and self-development) were used in the current study. The three dimensions consist of 27 items on a five-point scale varying from “totally disagree” to “totally agree”. The dimension teamwork (Cronbach’s alpha = 0.85) reflects social support and work spirit and consists of 12 items, e.g. “I can rely on my colleagues and trust them” and “We are not a team at work”. The stress-handling dimension (Cronbach’s alpha = 0.77) reflects active coping and self-efficacy and consists of seven items, such as “In difficult situations I do not wait and see, but take action” and “I can cope well with the demands of my job”. The dimension self-development (Cronbach’s alpha = 0.82) reflects possibilities for self-fulfilment and consists of eight items. Examples are: “My abilities are full employed” and “I need a new challenge”. The scores on items within each dimension were transformed to a 0–100 scale with a higher score indicating good teamwork, better stress handling, and more opportunities for self-development in work. The sum scores for the variables teamwork, stress handling, and self-development were not normally distributed. Tertiles were calculated to assign subjects into low, intermediate and high levels per dimension. Stressful life events The occurrence of stressful life events in the past 12 months was measured using a shortened social readjustment rating questionnaire (SRRQ) (Holmes and Rahe 1967). The original SRRQ consists of 43 life events (e.g., divorce, job change, death of family members and so forth), listed by rank order based on their mean life change values. Life change values classify the impact of the events and were obtained by scaling the life events based on the amount of coping required to deal with the event. The total score counts the life change values of all events in the past 12 months. In the current study the 25 events most appropriate for the population under study were selected. In theory, the total score can range from none of these events (0) up to all events (1077). Life style factors Life style factors were measured with the Dutch version of the Stanford wellness inventory (Sallis et al. 1985). Lifestyle factors of interest concerned moderate physical activity, vigorous activity, smoking, and alcohol use. The questionnaire has single questions on regular participation in moderate activities for 30 minutes or more and participation in vigorous activities for 20 min or more, both on a five-point scale ranging from “never” to “5 days or more per week”. Those who reported moderate physical activity on at least 5 days per week were considered in agreement with the recommendation on moderate-intensity physical activity, and subjects with vigorous exercises at least 3 times per week were considered in agreement with the recommendation on vigorous-intensity physical activity (Pate et al. 1995). Current smoking was assessed with the question “Do you smoke?”. A five-point-response scale was used to assess alcohol drinking by average number of alcohol drinks per week (1–7, 8–14, 15–21, 22–28, more than 28). Problematic drinkers were defined as those who consumed more than 14 units of alcohol per week for women and more than 21 units for men (Health Council Netherlands 2006). Physical examination Physical examinations were performed using MicroFit equipment in accordance with the protocol of the American College of Sports Medicine (ACSM, 1975). During the physical examination biometry was recorded, including weight, height, biceps strength, and cardio respiratory fitness. The body mass index (BMI) was used to define subjects as normal (BMI ≤ 25), overweight (BMI 25–30), or obese (BMI ≥ 30). Maximal isometric muscular strength of the biceps was measured after one practice trial with a calibrated dynamometer with the subjects in standing position with 90° flexion in the elbows for 3 s. The isometric biceps strength was calculated as the average of several hundred readings over the 3 s period. Cardio respiratory fitness was assessed by a 12 min sub-maximal bicycle ergometer test, supervised by instructors. Subjects pedaled at 60 rev. min−1 for 12 min on the cycle ergometer, at an exercise intensity designed to produce a heart rate between 120 and 170 beats min−1, in order to reach a level of 80% of the theoretical maximal heart rate of the participant for 3 min, after a warming up period of minimal 3 min. This level was sustained for 3 min and the heart rate was measured at the end of each minute. The VO2 max (mL min−1 kg−1) was calculated by the work intensity (watts) and heart rates at the end of all the stages at exercise level. Statistics The effects of individual characteristics (age and sex), life style, psychosocial factors at work, stressful life events, life style, and physical condition on the outcome variables work ability, and mental and physical health were investigated with linear regression analysis. Probability plots and Kolmogorov–Smirnov tests showed that none of the determinants measured at continuous level were normally distributed. However, the evaluation of the distributions of residuals in the regression analyses showed that for those variables measured at ratio scale (i.e. age, VO2 max, and biceps strength), the assumption of linearity was not violated. These variables were included in the linear regression analyses as continuous variables. Due to considerable ceiling effects for the psychosocial variables and skewed distribution for life stress events, these variables were treated as categorical variables, defined by cut-off values based on tertiles. The analysis started with univariate regression models to determine the single effects of all determinants of interest. A backward regression technique was used to determine the multivariate model with the best overall fit. In this analysis, independent variables with a P-value of 0.05 or less were retained in the final model. The results of the regression analyses are presented by the regression coefficients and associated standard errors. A regression coefficient is an expression of the change in the work ability score due to a change in one unit of measurement of the independent variable of interest. For categorical variables, this reflects the effect on the work ability score of the presence of this determinant. The regression analysis on determinants of work ability was stratified for three age groups. All significant determinants in the multivariate model for one age group were included in the models for other age groups as well in order to provide an appropriate comparison. All analyses were carried out with the statistical package for social sciences version 11.0 for Windows (SPSS, 1999). Results The study population included 769 men (67%) and 372 (33%) women in a variety of jobs (Table 1). The median for age was 35.7 years (18–63). The distribution of excellent, good, moderate, and poor work ability was 42.8, 45.4, 9.7, and 2.1%, respectively. Subjects scored almost equal on mental health as on physical health, whereas the Pearson correlation coefficient between both measures of health was −0.20. The Pearson correlation coefficients between WAI and mental and physical health were 0.49 and 0.35, respectively. The three psychosocial factors at work were strongly interrelated with Pearson correlation coefficients varying from 0.45 to 0.57. Table 1Characteristics of 1,141 commercial workers who participated in a voluntary medical examinationCharacteristicsCasesMedian (min–max)Frequency (%)Individual characteristics Age (year)1,14135.7 (18–63)67.4 Male769Work ability Excellent (44–49)48842.1 (9–49)42.8 Good (37–43)51845.4 Moderate (28–36)1119.7 Poor (7–27)242.1Health Mental health component summary (MCS) (8–74)1,14154.2 (10.9–67.9) Physical health component summary (PCS) (4–73)1,14153.4 (18.2–70.6)Psychosocial factors at work Teamwork (0–100)1,13681.0 (27–100) Stress-handling (0–100)1,13668.0 (11–100) Self-development (0–100)1,13678.0 (9–100)Stressful life events (0–100)1,1365.5 (0−38.4)Life style Lack of moderate physical activity (<5 days week−1)79870.2 Lack of vigorous physical activity (<3times week−1 )88678.0 Current smoker14512.8 Problematic alcohol use424.5Physical examination Overweight (BMI 25–30 kg m−2)37135.9 (11.4–61.7)34.6 Obesity (BMI ≥ 30 kg m−2)5737.0 (8.0–94.0)5.3 VO2max (ml kg −1 min−1)1,117 Biceps strength (kg)1,134 Table 2 shows mental health was statistically significant influenced by psychosocial factors at work, stressful life events, and, life style factors, whereas physical health was influenced by lifestyle factors and physical condition in the univariate model. The multivariate model explained 22% of the variance in mental health. An increase in age with one year increased the mental health score with 0.1 point, and decreased the physical health score with 0.1 point. In the multivariate analysis most determinants remained statistically significant, albeit with a lower regression coefficient, especially for teamwork and self-development. The multivariate model explained only 5% of the variance in physical health. It is of interest to note that neither problematic alcohol use nor overweight or obesity was associated with physical health. Table 2Results of backward regression analysis: effects of psychosocial factors at work, stressful life events, lifestyle and physical condition on mental health and physical health among workers in commercial services (n = 1141)Mental Health (MCS) Physical health (PCS)Univariate modelMultivariate modelUnivariate modelMultivariate modelβSEβSEβSEβSEIndividual characteristics Age (year)0.09*0.030.07*0.02–0.07*0.02–0.09*0.02 Male2.41*0.511.52*0.471.66*0.381.91*0.38Psychosocial factors at work Low versus high teamwork –5.90*0.56–2.71*0.66–0.92*0.43n.s. Intermediate high versus teamwork –2.39*0.57–0.700.58–0.680.45n.s. Low versus high stress-handling –6.94*0.56–4.71*0.62–0.260.45n.s. Intermediate versus high stress-handling –2.39*0.58–1.42*0.590.010.46n.s. Low versus high self-development –5.44*0.60–2.12*0.65–0.730.46n.s. Intermediate versus high self-development –2.19*0.58–0.590.57–0.810.45n.s.Stressful life events High versus low stressful life events–3.13*0.59–3.13*0.54–0.620.44n.s. Intermediate versus low stressful life events–1.91*0.59–1.98*0.53–0.050.44n.s.Life style Lack of moderate physical activity–0.140.53n.s.–0.010.39n.s. Lack of vigorous physical activity–1.27*0.58–1.37*0.52–1.79*0.43–1.71*0.42 Current smoker–2.02*0.72–1.96*0.65–1.14*0.54n.s. Problematic alcohol use–1.761.27n.s.–1.610.92n.s.Physical examination Obesity (BMI ≥ 30) versus normal (BMI < 25)–0.841.14n.s–1.540.83n.s. Overweight (BMI 25–30) versus normal 0.340.54n.s–0.300.39n.s. VO2 max (ml/kg/min)–0.010.03n.s0.05*0.02n.s. Biceps strength (kg)0.030.02n.s0.04*0.02n.s.n.s not significant, P > 0.05 Table 3 shows work ability was statistically significant influenced by psychosocial factor at work, stressful life events, lack of vigorous physical activity, and obesity in the univariate model. The multivariate model explained 29% of the variance in work ability. Again, in the multivariate model most determinants remained statistically significant, although with lower regression coefficients. The influence of stressful life events increased in the multivariate model. Table 3Results of backward regression analysis: effects of psychosocial factors at work, stressful life events, lifestyle and physical condition on work ability among workers in commercial services (n = 1141)Work abilityWork abilityUnivariate modelMultivariate modelβSEβSEIndividual characteristics Age (year)–0.07*0.02–0.09*0.01 Male2.13*0.312.08*0.28Psychosocial factors at work Low versus high teamwork–4.02*0.32–1.32*0.40 Intermediate versus high  teamwork–1.52*0.34–0.200.35 Lowversus versus high stress-handling–4.39*0.34–2.75*0.35 Intermediate versus high  stress-handling–1.41*0.35–0.79*0.35 Low versus high  self-development–4.11*0.35–2.20*0.39 Intermediate versus high  self-development –1.67*0.34–0.91*0.34Stressful life events High versus low stressful life  events–1.36*0.36–2.01*0.32 Intermediate versus low stressful  life events–0.97*0.36–1.14*0.32Life style Lack of moderate physical  activity0.490.32n.s0.31 Lack of vigorous physical  activity–0.71*0.35–0.71* Current smoker–0.680.44n.s. Problematic alcohol use–0.520.74n.s.Physical examination Obesity (BMI ≥ 30) versus  normal (BMI < 25)–2.02*0.68–1.21*0.59 Overweight (BMI 25–30) versus  normal (BMI < 25)–0.490.32–0.320.28 VO2 max (ml −1kg −1min)0.030.02n.s. Biceps strength (kg)0.030.01n.s.n.s  not significant, P > 0.05 No significant interaction was observed for age, sex, and psychosocial factors at work. Table 4 shows that in each age group sex, stress handling, and self-development were associated with the work ability index. Lifestyle factors were associated with work ability only in the oldest age group of workers, over 45 years. Obesity no longer was statistically significant. Table 4Results on backward regression analysis per age-group: effects of psychosocial factors at work, stressful life events, lifestyle and physical condition on work ability among workers in commercial services (n = 1141)Work abilityMultivariate modelAge ≤ 32 year (n = 335)Age 32–45 year (n = 366)Age > 45 year (n = 200)βSEβSEβSEIndividual characteristics Male1.43*0.421.37*0.493.19*0.83Psychosocial factors at work Low versus high teamwork–0.440.61–1.40*0.63–1.070.98 ntermediate versus high  teamwork–0.490.50–0.390.60–0.580.89 Low versus high  stress-handling–2.85*0.55–3.22*0.63–2.44*0.95 Intermediate versus high  stress-handling–0.960.52–1.080.61–0.830.86 Low versus high  self-development–2.59*0.64–1.64*0.59–3.57*1.01 Intermediate versus high  self-development –1.27*0.51–0.920.57–1.630.84Stressful life events High versus low stressful life  events–1.25*0.49–1.67*0.51–2.60*0.84 Intermediate versus low  stressful life events–0.640.53–1.78*0.52–0.340.72Life style Lack of moderate  physical activity–0.280.470.020.511.45*0.70 Lack of vigorous physical  activity–0.580.49–0.560.53–1.62*0.79 Problematic alcohol use1.700.93–0.851.19–2.62*1.33P > 0.05 Discussion This study showed that work ability of white-collar workers in commercial services industry was strongly associated with psychosocial factors at work, such as teamwork, stress handling, self-development, and, to a lesser extent, with stressful life events, lack of physical activity, and obesity. Work ability was strongly associated with mental and physical health. Determinants of mental health were very similar to those of work ability, whereas physical health was influenced primarily by lack of life physical activity. Some limitations must be taken into account in this study. First, the cross-sectional design does not permit exploration of causal relationships between the determinants and work ability. Therefore, it remains unknown whether, for example, poor stress handling will decrease work ability or decreased work ability will cause a poorer stress handling. Nevertheless, the results are still of interest as they give a first insight into important factors for interventions among white-collar workers. Second, data were drawn from voluntary participants. Information on non-response for both measures showed that age and sex did not bias response. Non-response differences between questionnaire and physical examination did not show any bias; none of the questionnaire variables were associated with not participating in the physical examination; and also none of the physical examination variables were associated with not participating in the questionnaire. Third, the reliability of the physical examination highly depends on the professional skills of the instructor and the standardization of the examination. The maximum oxygen uptake was indirectly calculated using the heart rate, which can be easily increased by minor distractions, such as room temperature, and talking during the test. In this study among white-collar workers in commercial services industry, the proportion of workers with poor work ability was 2.1% and the mean WAI was 41.1 (SD = 5.1). These results are slightly higher than the Finnish reference data in mentally demanding work (mean 39) (Tuomi et al. 1998). Work ability in this study population was influenced by sex, age, psychosocial factors at work, stressful life events, and life style factors. These factors together explained 29% of the total variance in work ability in this study population. Male sex increased work ability with two points, which means 4% of the maximum score. An increase in age of 40 years decreases the WAI score with four points, which is 7.3% of the maximum score, which indicates a rather modest influence of age on work ability. Psychosocial factors each had an effect on WAI comparable to sex, whereas the combined effect of the psychosocial factors is approximately 1.5-fold the effect of 40 years of aging. Lack of vigorous physical activity decreases the WAI score with only 0.7 points, which is no more than 1.5% of the maximum score. Obesity (5% of the population) compared to normal weight decreases the WAI score with 1.2 points, which is 2.4% of the maximum score. Each psychosocial factor at work was negatively associated with work ability. Univariate results showed comparable strength in associations, while the multivariate model showed lower regression coefficients, especially for teamwork. It seems that the association between teamwork and work ability was more influenced by other determinants included in the multivariate model, than the associations between work ability and stress handling and self-development. In previous research, inconsistent results were found regarding the influence of psychosocial factors at work on work ability. For example, in the metal industry an increase in teamwork and increase in opportunities for development was not predictive of an increase in work ability during a 2-year follow-up (Tuomi et al. 2004). Negative associations between mental stress and work ability have been found among office workers (β = −0.17), but this association was minimized when including age in the regression model (Sjögren-Rönkä, 2002). Among bus drivers, significant associations were observed for high control by superiors and lack of responsibility at work with lower WAI scores (Kloimüller et al. 2000). The negative association of stressful life events with work ability in the current study is in agreement with earlier findings by Pohjonen (2001b), who found an increased risk for poor work ability [OR = 3.62 (2.2–5.9)] for a hard life situation outside of work. The results showed that a lack of vigorous physical activity was associated with decreased work ability, whereas associations between work ability and biceps strength and maximum oxygen uptake were not found in the multivariate model. The lack of significant results for maximum oxygen uptake and biceps strength is in line with findings of Eskelinen et al. (1991), Nygard et al. (1991), and Pohjonen (2001a). It may be hypothesized that in mentally demanding jobs a good physical condition is not required to meet the work demands and, thus, will have no influence on work ability. Stratification by age showed the importance of lifestyle in the oldest age group, but not among younger workers. This effect may be explained by the fact that health problems due to an unhealthy lifestyle, most notably diabetes mellitus and cardiovascular disease, occur primarily at older age. In the total study population, obesity was significantly associated with lower work ability, whereas no significant associations were found in the stratified analyses. This is partly due to lack of statistical power in these strata with smaller numbers of workers, since the magnitude of the regression coefficients were comparable but the standard errors increased substantially. The Pearson correlation coefficient of mental and physical health was −0.20, which was in line with results of van Duijn et al. (2004). In a univariate analysis both mental health and physical health were associated with work ability. However, determinants of work ability were similar to determinants of mental health. This finding can be explained by the fact that the work setting of the white-collar workers in the current study is characterized by high mental demands. An exception to the similarity in factors influencing both mental health and work ability was, smoking. Smoking was related to mental health, but not to work ability. The results of the current study outline the importance of work-related factors in white-collar workers, with regard to work ability. The combined impact of psychosocial factors is much stronger than is for individual factors, and is amenable to change, in contrast to individual factors as age, and sex. In conclusion, among white-collar workers in commercial services industry psychosocial factors at work, stressful life events, lack of vigorous physical activity, and obesity were significant with regard to work ability. The strong associations between psychosocial factors at work and mental health and work ability suggest that in this study population health promotion should address working conditions rather than individual life style factors, although the importance of life style factors seems to increase with aging of the worker.

Ann_Biomed_Eng-2-2-1705482

Evaluation of Important Treatment Parameters in Supraphysiological Thermal Therapy of Human Liver Cancer HepG2 Cells

This study was aimed at simulating the effect of various treatment parameters like heating rate (HR), peak temperature (PT) and hold/total treatment time on the viability of human liver cancer HepG2 cells subjected to different thermal therapy conditions. The problem was approached by investigating the injury kinetics obtained using experimentally measured viability of the cells, heated to temperatures of 50–70°C for 0–9 min at HRs of 100, 200, 300 and 525°C min−1. An empirical expression obtained between the activation energy (E) and HR was extended to obtain the E values over a broad range of HRs from 5 to 600°C min−1 that mimic the actual conditions encountered in a typical thermal therapy protocol. Further, the effect of the HR (5–600°C min−1) and PT (50–85°C) on the cell survival was studied over a range of hold times. A significant drop in survival from 90% to 0% with the simultaneous increase in HR and PT was observed as the hold time increased from 0 to 5 min. For complete cell death, the hold time increased with the increase in the HR for a given PT, while the total time showed presence of minima for 60, 65 and 70°C at HRs of 50, 100 and 200°C min−1, respectively. Introduction Primary liver cancer is the fifth most common malignancy in the world, with a global annual incidence of about one million new patients.6 In 2004, the American Cancer Society estimated 18,920 new cases of Hepatocellular Carcinoma (HCC) and the estimated deaths were 14,720.1 Surgical resection, the gold standard for the treatment of primary liver cancer, shows low success rate (20–37%) and a high recurrence (50–60%) rate due to many surgical complications.10,25 Liver transplantation is limited because of fewer donors.18 Combination treatment methods, using chemotherapy, embolization and chemoembolization have a limited effect even with the newer drugs available and their beneficial effects on the patient survival remain controversial in randomized studies.18,24 Use of radiation therapy is limited to alleviating the symptoms such as pain or just to shrink the tissue rather than destroying it. Ablation therapies that use heat to destroy the tumor are gaining increasing attention as an alternative because the treatment procedure is faster, simpler, less painful and cheaper.20 There are two known approaches for the application of these ablative procedures: hyperthermia, where the treatment modalities use temperatures ranging from 42 to 50°C for periods of 30 min to few hours. Another approach is thermal therapy, where the destruction of the tumor takes place by the application of heat at temperatures higher than 50°C within very short time periods of few minutes to seconds. The tumor is heated using any one of radiofrequency, microwave, laser or high intensity focused ultrasound energy sources. Several clinical trials have been performed to test the efficacy of these thermal ablative therapies in treatment of liver cancer using the above-mentioned energy sources. The results have shown survivals of 83–94% after 1 year, 50% after 2 years, 33% after 3 years and 33–40% after 5 years for hyperthermic radiofrequency ablation in studies conducted on 29–123 patients.18 Survival of 86% after 1 year in 55 patients has been reported for hyperthermic interstitial laser therapy;12 survivals of 73–86% after 3 years, 18 50% after 4 years and 48.6% after 5 years17 have been reported for microwave coagulation therapy. In spite of all the above-mentioned clinical trials, no data exists showing the effect of thermal therapy on the survival of liver cancer cells. Effect of hyperthermic temperatures on human liver cancer cell lines has also been investigated in few in vitro studies. Callari et al.7 used HTC hepatoma cells to study the action of retinol on viable cell recovery after in vitro hyperthermia at temperatures between 42°C and 44°C for 1 h. Hasumura et al.13 studied the effect of TNF along with hyperthermia on JHH-4, JHH-5 and JHH-7 human hepatoma cells between temperatures of 41.4°C and 42.5°C. However, none of the studies demonstrated the effect of temperatures above 50°C any liver cancer cell lines. Since in thermal therapy, the injury accumulates at very high temperatures within a short period of time, knowledge of the injury kinetics is very useful for predicting the cell/tumor damage15 for these thermal histories. Several thermal therapy studies have been performed to obtain the injury kinetics in different cell lines. Landry et al.19 heated HeLa cells up to 55°C using water bath, with thermal equilibrium time of 1 min for 55°C. To reduce this equilibrium time Borrelli et al.5 heated cells to 57°C on 0.025-mm thick mylar pieces. Cell injury kinetics was obtained by measuring the cell survival using clonogenic assay in both the studies. In addition, studies on T24 human bladder carcinoma,22 skeletal muscle,11 SN12 renal cell carcinoma15 and Dunning AT-1 prostate cancer cells4 in the temperature range of 40–70°C obtained cell injury kinetics using dye uptake assays and thus showed that dye uptake assay is a reasonable conservative estimate of the cell survival. Also, two recent in vitro studies in human benign prostatic hyperplasia tissue and rodent prostate cancer tissue showed that the cell injury kinetics measured by membrane integrity vital dye assay and histology assays are very similar.2,3 Therefore, a key observation from the above studies is that vital dye uptake assay can be used as an alternate and reliable conservative maker as it provides rapidity, automation and better control.15 A key difference of high temperature–short time thermal therapy protocols from the traditional hyperthermia protocols is the significant injury accumulation during the non-isothermal portion of the thermal history (heating up/cooling down period). Therefore, injury accumulation tends to be a complex function of hold time as well ramp up and cooling time. Hence, accurate prediction of the cell injury kinetics requires the knowledge of heating/cooling rates in addition to peak temperatures (PTs) and hold/total time. This model also better represents the thermo-clinical applications because the rate at which the PTs are achieved at different locations inside the tumor vary with the applicator location and plays a paramount role in determining the amount of injury accumulated. The knowledge of the injury accumulated in reaching a PT or the PT required to obtain a desired injury under different heating rates (HRs) are some of the very important parameters in designing better and optimal clinical protocols. This study was used to investigate experimentally and through simulations the parameters affecting thermal therapy of primary human liver cancer HepG2 cells between 50°C and 70°C using Ethd-1 uptake cell membrane integrity assay. The cells were thermally challenged on a programmable heating stage which can control and record the entire thermal history profile, allowing us to recreate typical thermal therapy protocols. Cell membrane integrity dye uptake (Ethd-1) was used as an irreversible injury marker. The first-order Arrhenius rate model was used to extract the kinetic parameters activation energy, E (kJ mol−1) and frequency factor A (s−1).16 The survival data was fitted into the Arrhenius model for various HRs and the E and A values were predicted by minimizing the least square between the model predictions and the experimental data for suspended and attached cells. Based on the predicted E values, a relationship between the E and HR was obtained experimentally and was further used to simulate the survival map for different hold periods by varying the target PT and the HR. Finally, the hold time and the total injury time necessary for complete cell death (Survival ≤ 0.001%) was predicted for target PTs of 60–85°C and HRs of 5–600°C min−1 representing the typical transient thermo-clinical protocol. Materials and methods HEPG2 Cell Culture and Sample Preparation Human liver cancer Hepg2 cells (ATCC, VA) were propagated in MEM cell culture media with 10% Fetal Bovine Serum (Hyclone, UT) and 1% penicillin/streptomycin (Invitrogen, CA). They were then incubated in 10% CO2 and 95% humidified air at 37°C in 75-cm2 T flasks. The cells were trypsinized with Trypsin-EDTA (ATCC, VA) for 5–10 min, centrifuged at 1200 rpm for 5 min and re-suspended in the media to an appropriate concentration. Suspended cell samples were prepared by placing 3 μl of cell suspension onto the center of the 12-mm diameter coverglass, covering it with another coverglass to prevent evaporation during heating. For attached cell studies, cell suspension was scattered over about 18–20, 12-mm cover glasses lying in a 110-mm diameter Petri dish. Overnight attachment of the cells on top of the coverglass was allowed in the incubator and then the cell sample was prepared in the same way as the suspended sample. Heating Stage Attached and suspended cell samples were heated on a programmable heating stage, at different HRs. Figure 1 shows the setup diagram of the heating stage and the feedback control used. Cement T-type thermocouple (Omega Engg Inc., CT) was used for temperature detection, which is interfaced with a data acquisition board (DAB) (Keithly Inc., Cleveland, OH) connected to the computer port. The interface between the user and the DAB is through a Visual Basic code that calculates the output voltage based on the set point and instantaneous temperature and controls it based on proportional and derivative constants used in the feedback loop. Table 1 shows the combinations of proportionality (kp) and derivative (kd) constants used in the feedback loop to achieve different HRs (100, 200, 300 and 525°C min−1). An OPAMP (Analog Devices Inc, MA) circuit is used to amplify the feedback voltage signal from the DAB and heat the stage. The stage was calibrated using fixed temperature tempi labels and sticks (MSC Industrial Supply Co., NY) to an accuracy of ±0.3°C for all PTs at each HR used in the study. Figure 1.Block diagram of the amplifying OPAMP circuit and feedback control of the heating stage. The OPAMP circuit amplifies the voltage signal from the data acquisition board to heat the stage, while the feedback control system, through the VB code, controls the amount of voltage based on the instantaneous temperature of the stage.Table 1.Values of proportionality (kp) and derivative (kd) constants for different HRs to obtain desired PTs.HR (°C min-1)PT (°C)kpkd100600.01550.0925650.01550.0925700.01550.0925200600.0350.108650.0350.108700.0350.121300600.0850.255650.0850.255700.09030.275 Heating Studies and Ethd-1 Dye Uptake Assay Suspended and attached HepG2 cell samples were heated to different time–temperature histories. Isothermal heating was carried out for suspended and attached cells at temperatures between 50°C and 70°C for 0.5–9 min. The cell sample was placed on the center of the heating stage after the PT was reached. The time for each cell sample to equilibrate to any given PT was experimentally calculated to be ∼5 s and was added to the hold time during heating. HRs of 100, 200 and 300°C min−1 were employed for heating attached cells non-isothermally between temperatures of 60°C and 70°C for a hold time of 0–3 min using the entire temperature–time history, which includes the rise, hold and the cooling period. The HRs were assumed to be constant for the entire thermal history and were obtained by taking a tangent to the exponential curve of the temperature rise. Therefore, the cell sample was placed on the center of the heating stage before starting the heating cycle. The cooling time for each sample from a given PT was calculated based on the experimentally measured cooling rate of 100°C min−1. Cellular injury post heating for isothermal and non-isothermal studies was quantified using Ethd-1 (Sigma-Aldrich, MO) vital dye assay by counting the number of cells stained with Ethd-1 dye (dead only) and the total number of the cells stained with Hoechst (Sigma-Aldrich). Thermally treated cell samples were placed in a 50 μl drop of 2.5 μM Ethd-1 and 10 μM Hoechst dye solution in a 35-mm Petri dish and incubated for 3 h. The incubation time of 3 h was based on the experiments conducted after 1, 2 and 3 h of incubation (data not included), which confirmed that membrane damage equilibrates within this period and a significant amount of media is pulled between the cover glasses to stain the cells. Control samples underwent the same procedure without heating. After 3 h, the dead and the total number of cells were counted with a fluorescent microscope (Nikon Eclipse TS 100, Tokyo) using a 20× objective. Multiple fields with at least 150–300 cells were counted for each run. Normalized Cell Survival For all the cell samples, the normalized cell survival (Se) was calculated as follows: NH and NE are the number of total and dead cells stained with Hoechst and Ethd-1 respectively; subscripts ‘t’ and ‘c’ represent thermally treated and control cells, respectively. The calculated cell survival was based on three separate experiments, with two runs for every data point in each experiment. The Cell Injury Model Thermally induced cellular injury was considered as a first-order irreversible process as shown below and as used in many previous heating studies; where V represents the viable state of a cell, I represents the injured state of a cell and k is the cell injury rate. The predicted cell viability (Sc) was calculated using the following equation: δ is the total treatment time (s). The Arrhenius model16 was adopted in this study as it has been extensively used in previous studies to obtain the cell injury kinetics14,21 as shown below; A is frequency factor (s−1), E is activation energy (J mol−1) and R is the universal gas constant (8.314 J mol−1 K−1). A and E are related by an empirical formula, obtained by compiling the injury kinetics of various proteins, cells and tissues.15 Determination of Arrhenius Model Parameters The kinetic parameters E and A were determined by fitting Eqs. (4) and (5) in Eq. (1) as obtained by He and Bischof15 They were further refined using a simple optimization subroutine to minimize the root mean square error between the experimental data and model prediction. N is the total number of data points, vector ‘X’ contains variables E and A. Se,i and Sc,i represent the ith measured data point and model fit, respectively. The goodness of the fit (R2) is calculated as follows: Sav is the measured average survival from the Etdh-1 studies. Results The results were generated by counting the number of total and dead cells from the micrographs of the cell samples as shown in Fig. 2 and using Eq. (1) to calculate the normalized cell survival for each sample. Figure 2a, b represent the control sample, where the number of dead cells, stained with Ethd-1 (red) dye, is very less compared to the total number of cells, stained with Hoechst (blue) dye and the cell survival is higher than 0.96. The cell survival decreases slightly as the number of dead cells compared to the total number of cells increases for the cell sample heated at 50°C for 9 min, as shown in Fig. 2c, d. The survival drops dramatically with a sharp increase in the number of dead cells compared to total cells at 70°C as seen in Fig. 2e, f. Figure 2.Micrographs showing cell samples stained with Hoechst (Blue-all cells) and Ethd-1 (Red-dead cells only). (a and b) – control samples without heating; (c and d) – samples heated at 50°C for 9 min; (e and f) – samples heated at 70°C for 1 min. Isothermal Heating Studies The isothermal heating studies were used to understand the effect of different temperature–time histories on the viability of attached and suspended HepG2 cells and to quantify the cellular injury associated with these histories. All the cell samples reached the PTs in ∼5 s at an approximate HR of 525°C min−1 resulting in a negligible difference between the hold time and the total time of the treatment. Thus heating at the rate of 525°C min−1 was referred as isothermal heating. The data in Fig. 3a, b represents the mean and the standard error for measured cell viability of attached and suspended HepG2 cells respectively using Ethd-1 dye uptake assay. Both the figures show that for a given PT, the survival drops with the increase in the hold time. At 50°C, the survival drops from 0.84 to 0.43 for suspended cells and from 0.89 to 0.62 for attached cells, as the hold time increases from 2 to 9 min. Similar trend was observed for heating at all the PTs. The figures also show that for a given hold time, the cell survival drops with the increase in the PT for suspended cells. For a hold time of 2 min, the cell survival drops from 0.84 to 0.52 and from 0.52 to 0.29 as the PT increases from 50°C to 55°C and from 55°C to 60°C, respectively. For a hold time of 1 min, the cell survival drops from 0.45 to 0.1 and from 0.1 to 0.003 as the PT increases for 50°C to 55°C and from 55°C to 60°C, respectively. The same trend is also seen for the attached cells but with slightly higher survival. Figure 3.(a) Plot of experimentally measured and predicted survival of attached cells heated isothermally versus hold time in minutes for single cell heating at PTs of 50, 55, 60, 65 and 70°C. Cross (×), filled circles (•), filled diamonds (♦), filled squares (■) and filled triangles (▲) represents heating of samples at 50, 55, 60, 65 and 70°C, respectively. The lines represent the exponential fit through all the data points based on the predicted E and A values. (b) Plot of experimentally measured and predicted survival of suspended cells heated isothermally versus hold time in minutes for single cell heating at PTs of 50, 55, 60, 65 and 70°C. Cross (×), filled circles (•), filled diamonds (♦), filled squares (■) and filled triangles (▲) represents heating of samples at 50, 55, 60, 65 and 70°C respectively. The lines represent the exponential fit through all the data points based on the predicted E and A values. (c) Direct comparison between suspended and attached HePG2 cell viability, using student t-test assuming unequal variance. Each data point is an average of three separate runs using two different cell samples for each run. The error bar in the figure represents the standard error of the mean. Data points marked with asterisks indicate the data points where viability of suspended and attached cells is significantly different. The figures also show the effect of PTs on the slope of the survival curve. The drop in the survival with the increase in the hold time is slow at low PTs. The survival curve gets steeper as the PT increases. When heated at 50°C for up to 9 min, the cell survival is still higher than 0.43 for both suspended and attached cells. This trend changes for 55°C, where the cell survival drops to 0.03 after only 6 min of heating. This trend magnifies sharply as the PTs increases to 60, 65 and 70°C. For suspended cells, the survival at 60°C varies from 0.58 to almost zero survival as the hold time increases from 0.5 to 3 min. The survival dropped from 0.2 to 0.003 for an increase in the hold time from 15 s to a min at 70°C. The attached cells also showed a similar trend with slightly higher survival than the suspended cells for all the data points. The student t-test assuming unequal variance was used to observe the statistical significance between the experimentally measured cell viability of suspended and attached HepG2 cells as shown in Fig. 3c. With p = 0.05 as the criterion, no significant difference in the cell viability was observed for most of the data points. The data points that were significantly different (p < 0.05) are 50°C – 5 min, 50°C – 9 min, 60°C – 3 min and 70°C – 1 min. This difference for data points of 60°C – 3 min and 70°C – 1 min can be neglected as the experimentally calculated cell viability in suspended and attached HepG2 for both these thermal insults is of the order 10−3. The p-value of 0.088 for 60°C – 2 min data point suggests the difference in the cell viability is not significant, but is still different. As no significant difference between the survival of suspended and attached HepG2 cells was observed for most of the time–temperature histories and since attached cells better describe the arrangement of the cells in a tissue, only attached cells were used for further non-isothermal studies. The parameters of cell injury, E and A, were obtained by fitting the Arrhenius rate model into the survival data. This was done by minimizing the function in Eq. (6) using Eqs. (3) and (4). The predicted E and A values for suspended and attached cells for isothermal heating are shown in Table 2. The E and A values obtained for suspended cells are 229.46 (kJ mol−1) and 3.495 × 1031 (s−1) and those for attached cells are 248.64 (kJ mol−1) and 5.396 × 1036 (s−1), respectively. The predicted survival based on the predicted E and A values for attached and suspended cells is shown by the solid lines in Fig. 3a, b, respectively. The goodness of the fit between the predicted and experimentally measured values, obtained using Eq. (7) for suspended and attached cells, is 0.86 and 0.84, respectively. Table 2.Activation energy E (kJ mol−1) and frequency factor A (s−1) using all the data point for different HRs.HR (°C min−1)Activation energy E (kJ mol−1)Frequency factor A (s-1)100272.47.757 × 1040200262.021.502 × 1039300257.382.576 × 1038525 (Isothermal – Attached)248.645.396 × 1036525 (Isothermal – Suspended)229.463.495 × 1031 Non-Isothermal Heating Studies The study focused next on the investigation of the effect of different HRs on the cell viability of attached HepG2 cells and extraction of kinetic parameters of the cellular injury associated with each HR. The HRs of 100, 200 and 300°C min−1 were employed in this study, as they all showed a considerable difference in the rise-time at all the PTs. The total injury time (rise, hold and cooling time) was taken into account to quantify the thermal injury for all the HRs. The data in Fig. 4a–c represents the mean and the standard error of the measured cell viability using Ethd-1 dye uptake assay for attached HepG2 cells at HRs of 100, 200 and 300°C min−1, respectively. All the figures show that, for a given HR, the cell survival drops with the increase in the hold time and the PT, similar to that observed in isothermal heating. At the HR of 100 C min−1, the survival drops from 0.82 for no hold time to 0.18 for a hold time of 2 min for a PT of 60°C. This trend remains the same for the PTs of 65°C and 70°C and for all the HRs. The survival drops from 0.82 at 60°C to 0.69 at 65°C and further to 0.11 at 70°C, for no hold time at HR of 100°C min−1. This trend continues for all the HRs and at all the hold times. The figures also illustrates that for any given PT and for a given hold time, the cell survival increases with the increase in the HR. At 60°C for no hold time, the survival increases from 0.82 at 100°C min−1 to 0.85 at 200°C min−1 and further to 0.88 at 300°C min−1. At 65°C for a hold time of 0.5 min, the survival increases from 0.15 at 100°C min−1 to 0.19 at 200°C min−1 and then to 0.22 at 300°C min−1. This trend also continues for 70°C for any given hold time. Figure 4.Plot of experimentally measured and predicted survival of attached cells heated at various heating rates versus time in minutes for single cell heating at temperature 60, 65 & 70°C. (a) 100°C min−1. (b) 200°C min−1. (c) 300°C min−1. Filled diamonds (♦), filled squares (■) and filled triangles (▲) represents heating of samples at 60, 65 and 70°C respectively. The lines represent the exponential fit through all the data points based on the predicted E and A values. A trend was observed in the Arrhenius parameters obtained from the non-isothermal heating studies as the E, and the corresponding A value, decreased with the increase in the HR. Table 2 shows the predicted E and A values for HRs of 100, 200 and 300°C min−1. The activation energy decreases from 272.40 kJ mol−1 at 100°C min−1 to 262.02 kJ mol−1 at 200°C min−1 and then to 257.38 kJ mol−1 at 300°C min−1. The solid lines in the Fig. 5a–c represent the predicted cell survival based on above values respectively. The goodness of fit between the predicted and measured cell survival for all data points obtained using Eq. (7) is 0.97 for all the HRs used in this study. Figure 5.Plot of the normalized survival of attached cells for various hold times at different peak temperatures and heating rates based on the predicted E and A values. (a) 0 min. (b) 0.25 min. (c) 1 min. (d) 5 min. Relation Between Activation Energy and HRs The following relation was obtained by fitting a polynomial trend line through the data in Table 2 that includes the isothermal heating at 525°C min−1; E represents the activation energy and HR represents the HR. The above equation was then used to obtain the values of activation energy over a range of HRs from 5 to 600°C min−1 in order to simulate the effects of various parameters on the cell survival. The values of the activation energy shows a small decrease from 282.09 to 248.29 kJ mol−1 over a large increase in the HR increased from 5 to 600°C min−1. Effects of PTs, HRs and Hold Time on the Cell Survival Based on the E and A values obtained at HRs of 5–600°C min−1 using the E–HR relation, the cell survival was predicted at different hold times for PTs between 50°C and 85°C. The combined effect of the increase in the hold time and PT on the cell survival is shown in the Fig. 5a–d for all the HRs mentioned above. For a given hold time and HR, the cell survival drops consistently with the increase in the PT. For a HR of 20°C min−1, the cell survival drops from 0.98 at 50°C to 0.091 at 65°C and then to no cell survival at temperature above 75°C, when the hold time is 0 min. The trend remains the same for any HR and hold time combination. The drop in the cell survival for any hold time is lower when the PTs are 55°C and less. The survival drops slightly from 0.96 at 50°C to 0.84 at 55°C for heating at 200°C min−1 for a hold time of 1 min. This drop is higher, from 0.84 to 0.48, when the temperature increases to 60°C. This trend is magnified with the increase in the PT to 65°C where the survival drops to 0.05 and to almost complete cell destruction for a further 5°C rise in the PT. At any given hold time and PT, the cell survival increases with the increase in the HR. The survival increased from 0.006 to 0.56 at 60°C for a hold time of 1 min and from 0 to 0.002 for a hold time of 3 min at 65°C as the HR increased from 5 to 600°C min−1. For a given PT and a HR, the cell survival drops with the increase in the hold time. At 60°C and 50°C min−1 the cell survival drops from 0.76 to 0.008 when the hold time increases from 0 to 5 min. For the same increase in the hold time, the survival drops from 0.12 to zero at 70°C and 400°C min−1. For any hold time, almost complete cell destruction takes place at all the HRs when the PTs are 75°C and above. For HRs of 50°C min−1 and lower, the PT to acquire almost complete cell death decreases with the increase in hold time. For zero hold time, 70°C shows a survival of 0.004 and less for HRs of 50°C min−1 and less. When the hold time is 1 min, complete cell death is observed at 65°C for all the HRs below 50°C min−1 and at 60°C for HR of 5°C min−1. The survival at 55°C decreases considerably with hold time at these low HRs but is still significantly high (>10%), while the survival at 50°C shows a very less drop. For a HR of 10°C min−1, the survival at 55°C drops from 0.79 to 0.28 with the increase in the hold time from 0 to 5 min. At 50°C, this drop is from 0.96 to 0.79. This trend is depicted in Fig. 5a–d as the drop in the slope of the vertical portion with the increase in the hold time. Hold and Total Injury Time for Complete Cell Destruction The relationship between activation energies and HRs was further used to predict the hold time and total injury time required to attain complete cell destruction at different target PTs and HRs that represents different scenarios of a clinical thermal therapy protocol. Figure 6a, b shows the calculated hold time and total time in minutes resulting in complete cell death at target PTs between 60°C and 85°C and HRs between 5 and 600°C min−1, which represents the transient clinical problem strictly under thermal therapy conditions. Figure 6.(a) Plot of hold time in minutes to attain complete destruction of attached HepG2 cells at various PTs between 50°C and 85°C and at HRs from 5 to 600°C min−1. (b) Plot of total time (min) to attain complete destruction of attached HepG2 cells at various PTs between 50°C and 85°C and at HRs from 5 to 600°C min−1 . Figure 6a shows that, for a given target PT, the hold time increases with the increase in the HRs. It increases from 0.45 to 8.4 min, from 0 to 2.2 min, from 0 to 0.61 min, from 0 to 0.17 min and from 0 to 0.05 min at a target PT of 60, 65, 70, 75 and 80°C, respectively. The increase in the hold is negligible at 85°C. Further, the hold time decreases with the increase in target PT, for any given HR. At 300°C min−1 the hold time for complete cell destruction decreases from 7.2 min at 60°C to 0.4 min at 80°C. For very low HRs of 5 and 10°C min−1, complete cell destruction is achieved for target PTs of 65°C and above without any hold time. Similarly, no hold time is required when the target PTs are 70°C and above at 20°C min−1 and 75°C and above at 50°C min−1. Heating, targeted at PTs of 80°C and 85°C, kills all the cells without any hold period for any of the above used HR. These results demonstrate that significant cell death takes place during the ramp up and ramp down period. Figure 6b shows the variation in the total injury time (rise, hold and cooling time) to achieve complete cell destruction with the change in the target PTs and HRs. For 60, 65 and 70°C, the total injury time showed a minima with the decrease in the HR. For example, the total time at the target PT of 60°C, decreases from 8.8 for 600°C min−1 to 5.9 for 50°C min−1 and then again increases to 7.8 min for 5°C min−1. At 65°C, the total time decreases from 2.7 min to 2 min and then increases to 3.2 min for a decrease of HR from 600 to 100°C min−1 and again to 20°C min−1, respectively. The trend remains the same at 70°C, the transition of hold time being at a higher rate of 200°C min−1. For target PTs of 75, 80 and 85°C, this trend reverses as the total injury time consistently increases with the decrease in the HR, unlike for all other target PTs. The total time increased almost twice from 0.76, 0.69 and 0.7 at 75, 80 and 85°C respectively, as the HR dropped from 600 to 50°C min−1. For the target PTs (85°C to 65°C) where no hold time is required to attain complete cell death, total injury time was obtained as 1, 1.35, 2.4, 3.96 and 6.93 min at HRs of 100, 50, 20, 10 and 5°C min−1, respectively. The trend in the change of the total injury time with change in target PT, for any given HR remains the same as that for the hold time. Discussion Injury Kinetics of Attached and Suspended Cells The comparison of the survival data for the suspended and the attached cells (Fig. 3a, b) indicate that the suspended cells are a slightly more susceptible to heat than the attached ones. He and Bischof15 also obtained similar results for SN12 human renal cell carcinoma within the same temperature range, where the reason for this difference between suspended and attached cells was assumed to be the difference in their protein synthesis and gene expression. It may be assumed that some of the above differences also exist between attached and suspended HepG2 cells used in this study. The results shows a clear difference between the frequency factors for both the cell types, but the activation energies does not show much variation. But since the difference in the survival of suspended and attached cells is small either of the cell type can be used in studying the cell injury kinetics in HepG2 cells. Comparison with Previous Hyperthermic Studies Hyperthermic studies are mainly characterized by the presence of a shoulder region followed a slope in the survival curve. In these studies, the shoulder region is typically associated with considerably higher (∼30–40%) cell survival within a long heating period of few hours and has been observed in the temperature range of 42–45°C in many studies. Chinese hamster ovary (CHO) cells showed a small (15%) drop in the survival followed by reduction in the slope of the survival curve when heated at 42°C for 3 h.9 Human melanoma HTB-66 cells, when heated between 42°C and 45°C, also showed a considerably high survival (∼30%) followed by reduction in the slope at 42.5°C even after 5 h of heating.23 In this study, heating at 45°C for 15 min (data not included) and at 50°C for 9 min also shows a significantly higher (>42%) survival for attached and suspended HepG2 cells indicating the possibility of a shoulder region for these short time protocols. However, whether they biologically behave in a similar fashion to the hyperthermic shoulders by demonstrating thermotolerance has not been studied and further investigation is clearly warranted to verify these facts. Another important aspect of the single cell heating studies is the presence of a break point, after which the slope of the Arrhenius plot drops significantly with the increasing temperature. The break point is believed to be the indication of an achieved thermotolerance and change in the mechanism/target of the thermally induced cellular injury beyond it.9 Such break points have been observed at 43.5°C for human cell lines and 43°C for rodent cell lines. We have observed no significant break point for our studies between temperatures of 55–70°C. However, one interesting observation in our study was that heating at 60°C for 3 min for attached and suspended cells, shows a sudden change in the slope of the survival curve. The survival drops from ∼0.25 at 2 min to 0.003 after 3 min of heating, suggesting the possibility of a distinct change in the response to heating at 60°C for HepG2 cells. Comparison of Arrhenius Parameters This study shows a linear relationship between activation energies and HRs for a temperature range of 50–70°C. The activation energies between 282.09 kJ mol−1 and 248.29 kJ mol−1 were obtained for HRs of 5–600°C min−1, which are within the range of the activation energies obtained in most of the dye uptake assays. Thermal injury in Dunning AT-1 attached prostate cancer cells from 40°C to 60°C gave activation energy of 245 kJ mol−1.4 Activation energy values obtained in renal cell carcinoma studies are 290–360 kJ mol−1.15 and in skeletal muscle tissue is 230 kJ mol−1.11 The activation energies for clonogenic assays for various studies like AT-1 prostate cancer cells (40–70°C),4 T24 human bladder carcinoma cells (48–65°C) and CHO (48–65°C)22 were found to be 500, 600 and 680 kJ mol−1, respectively. These values are higher than the ones obtained in this study which is not surprising since clonogenic assay yields higher activation energies than dye uptake assays. Since the activation energies obtained in this study are in the same range of the values obtained for other dye uptake assays, we suggest that the mechanism of thermal injury and its measurement is consistent with other studies. Clinical Relevance Clinical thermal therapy is a transient conduction problem as the tumor gets heated by deposition of energy from the source and by diffusion of the heat from the applicator. Based on the size and heat absorption pattern of the tumor, different radial locations are heated to different PTs at different rates. The regions closer to the applicator gets heated to higher PTs at higher HRs than the ones that are further away resulting in different survival patterns at various locations inside the tumor. Thus, accurate prediction of the hold time of the applicator inside the tumor is very important to ensure desired cellular injury accumulation at every location. As in these high temperature–short time thermal therapy protocols a considerable (>10%) injury accumulates within the non-isothermal portion of the thermal history, the determination of the total treatment time (rise, hold and cooling) is also necessary to obtain desired injury accumulation. Since the maximum size of a liver tumor that can be treated using various energy sources is limited to 4–6 cm diameter,8 the HRs between 600°C min−1 and 5°C min−1 are representative of all the locations within the tumor for any selected energy source, while the temperatures between 85°C and 50°C are descriptive of the actual temperature distribution inside the tumor during the thermal application. Temperatures higher than 85°C might result in carbonization near the applicator as well as overheating of the healthy tissue beyond the edge of the tumor, while temperatures lower than 50°C may result in insufficient injury accumulation within the short treatment time. Therefore, design of clinical thermal therapy protocols requires knowledge of the survival patterns for different hold times and precise prediction of hold and total treatment times to obtain desired cellular injury within the above mentioned range of HRs and PTs. Such an attempt has been made in a previous study carried out by He and Bischof15 on SN12 renal cell carcinoma. The cell injury kinetics, based on the non-isothermal heating at an average rate of 100°C min−1, were used to predict the PTs necessary to obtain a considerable (>10%) amount of cellular injury in cells during the rise period. This prediction was extended to various HR–activation energy combinations just by selecting HRs between 2°C min−1 and 200°C min−1 and activation energies up to 1000 kJ mol−1. However, experimental calculations to determine E as a function of HRs was not included in above study. The unique feature of this study is that for all PTs, prediction of survival and treatment time at any given HR is based on a specific experimentally calculated E–A combination for that HR, which results in a close approximation of the actual cellular injury for these high temperatures-short time thermal treatments. An important assumption made in this study is that the HR used to attain a target PT remains constant. The temperature rise on the heating stage, based on the constants kp and kd in the feedback loop, was found to be exponential. The HR was approximated by taking a tangent to the temperature–time curve, resulting in a constant value for the entire temperature rise. However, inside a tumor, the bioheat equation is required to model the transient temperature rise which is rarely linear. But as an initial step in understanding the effect of important parameters governing thermal therapy, this assumption may be a good approximation to predict desired cellular injury and Figs. 5 and 6 may provide important insights in this direction. The predicted survival at different hold times in Fig. 5a–d shows that for any HR, the PT required to attain a considerable amount of injury accumulation (>90%) decreases as the hold time increases. A smaller flat portion representing complete cell death and the steeper vertical portion representing considerable (>10%) survival for zero hold time in Fig. 5a intersects at 70°C and complete cell death is observed for HRs of 50°C min−1 and lower. This suggests that for HRs of 50°C min−1 and lower, typically at the boundary of the tumor, PTs of only 70°C and above are capable of accumulating significant amount of injury. This information is helpful in developing treatment protocols for small sized tumors, where the use of high target PTs will not significantly damage the healthy tissue beyond the range of the tumor, even though the temperatures at those locations are higher than 60°C. As the hold time increases the intersection shifts toward lower PTs increasing the flat portion of complete cell destruction. The decrease in steepness of the vertical portion also implies that sufficient cell damage can be obtained at tumor boundaries at these low temperatures. For example, when the hold time is 5 min, more than 80% survival is observed even at 55°C at the boundary of the tumor where the HR is as low as 5°C min−1. This shows that thermal protocols with higher hold times are beneficial for cell injury in large size tumors. Since it is necessary that the intersection should coincide with the tumor boundary to maximize the damage within the tumor as well as minimize the damage of the surrounding healthy tissue, the contours obtained in Fig. 5a–d are very important in the selection of the hold times based on the size of the tumor. Since the goal of every clinical application is complete tumor destruction, it is very important to know the hold time required to destroy all the tumor cells for different PT–HR combinations (that may represent different regions within a tumor depending upon its geometry and the energy source used). Figure 6a shows the hold time necessary to acquire complete damage in attached HepG2 cells at PTs between 60°C and 85°C and HRs between 5°C min−1 and 600°C min−1 and suggests that the hold time increases with the simultaneous decrease in the PT and the HR. Therefore, as the distance from the applicator increases, decreasing the PT and the HR, the hold time for tumor destruction increases. For example, Fig. 6a shows that the hold time increases from almost zero at 85°C – 500°C min−1 to 4.84 min at 60°C – 50°C min−1. Significant increase in the hold time for PTs of 60°C and 65°C suggests that in the regions away from the applicator (HRs of 100°C min−1 and lower), the injury accumulation strongly depends on the hold time. However, the hold time is very less for PTs of 70°C and above even at high HRs and decreases to almost zero as the HR decreases, suggesting that the majority of the cell damage at these temperatures occurs just within the rise and cooling period. At 70°C and above for HRs of 20°C min−1 and lower, the hold time completely vanishes and the thermally induced cellular injury is just a function of the non-isothermal portion of the thermal history. Therefore it is necessary to investigate the variation in the total treatment time (rise, hold and cooling) to attain complete cellular destruction at various PT–HR combinations. Figure 6b shows the variation in the total treatment time based on the temperature distribution within the tumor, depending upon its geometry and the energy source used. For PTs of 75°C and above, unlike hold time, the total treatment time increases with the decrease in the HR. This is due to the fact that for HRs above 50°C min−1, the hold time is very small and negligible compared to the rise time to reach these high PTs and the time to cool down back to the room temperature. It was also observed that for a given target PT (>70°C)–HR (<50°C min−1) combination, where the hold time is zero, the total treatment time is lower than the calculated time. This suggests that if damage within a tumor is desired at very slower rate at PTs higher than 70°C, the tumor actually gets destroyed before even reaching the target PT at any location. But the interesting observation in Fig. 6b is the presence of a minimal total treatment time at PTs of 70, 65 and 60°C for HRs of 200, 100 and 50°C min−1, respectively. This implies that based on the size of the tumor and the energy source selected, if the temperature distribution within the tumor results in any of the above PT–HR combinations at the boundary, complete cell death can be achieved in the entire tumor within a minimal total treatment time. For example, if the temperature distribution is such that the boundary of the tumor reaches 60°C at 50°C min−1, the entire tumor can be destroyed in 5.9 min. Thus based on the size of the tumor, Fig. 6a, b may help in selecting hold and total treatment time in order to design an optimal clinical protocol. Since this study is based on experimental single cell data to obtain the activation energies at different HR, it can serve as an accurate model for designing better thermo-surgical protocols using appropriate energy sources of heat delivery. Also, for a clinician practicing the technique, it can serve as a database, from where an optimal thermal therapy protocol can be selected for the treatment of liver cancer. By selecting appropriate blood flow and tissue properties and by using the bioheat equation, this in vitro protocol can be extended to in vivo clinical protocols. Summary The investigation of injury kinetics in human liver cancer HepG2 cells between temperatures of 50°C and 70°C at HRs of 100, 200, 300 and 525°C min−1 (isothermal) was used to extract Arrhenius parameters, E and A, that showed a minor decay with the increase in the HR. The extrapolation of the kinetic parameters over a range of heating conditions showed that the E value decreased from 282.09 to 248.29 kJ mol−1 as the HR increases from 5 to 600°C min−1. The measured and the predicted survival at all the HRs dropped as the PT and hold time increased. A sharp drop in the survival was observed after 2 min of heating at 60°C for all HRs. The measured and predicted cell survival increased with the increase in the HR at any PTs and hold time. This increase in the survival reduced significantly for target PTs of 65°C and above and for hold times of a minute and higher. The hold time for complete cell damage decreases with the increase in the target PT and increases with the increase in the HR. No hold time is required for target PTs of 75°C and above with the decrease in the HR below 100°C min−1. The cellular damage is a dominant function of hold time at target PTs of 60°C and less and for HRs of 50°C min−1 and higher. Total time (rise, hold and cooling time) plays a major role in accumulating the cellular injury when the temperatures are 70°C and above and when the HRs are between 0°C min−1 and 50°C min−1. Minimal total treatment time for complete cellular damage was observed at target PTs of 60, 65 and 70°C and HRs of 50, 50 and 20°C min−1, respectively.

Virchows_Arch-3-1-1888719

TNM staging of foregut (neuro)endocrine tumors: a consensus proposal including a grading system

The need for standards in the management of patients with endocrine tumors of the digestive system prompted the European Neuroendocrine Tumor Society (ENETS) to organize a first Consensus Conference, which was held in Frascati (Rome) and was based on the recently published ENETS guidelines on the diagnosis and treatment of digestive neuroendocrine tumors (NET). Here, we report the tumor–node–metastasis proposal for foregut NETs of the stomach, duodenum, and pancreas that was designed, discussed, and consensually approved at this conference. In addition, we report the proposal for a working formulation for the grading of digestive NETs based on mitotic count and Ki-67 index. This proposal, which needs to be validated, is meant to help clinicians in the stratification, treatment, and follow-up of patients. Background It has been known for a long time, and was finally defined within the World Health Organization (WHO) classification of endocrine and digestive tumors, that neuroendocrine tumors (NET) arising at different anatomical sites of the digestive system represent tumor entities that differ in their biology [5, 7, 9, 31]. Several recent publications focused on the application of the “new” WHO classification and proved its effectiveness, supporting the concept that the different endocrine tumor types also differ in their clinical behavior [3, 4, 19, 20, 27, 32]. Malignant gastroenteropancreatic NETs may be fatal, though at a significantly slower pace than their exocrine counterparts. A number of retrospective papers and epidemiological data solidly support such statements [8, 9, 12–14, 21, 24, 25, 32]. This peculiar clinical feature attracted the interest of pathologists very early and was the reason for the special designation of such tumors as “carcinoid” by Oberndorfer [15]. As gastroenteropancreatic NETs are rare [9, 13, 14], it is tempting to lump them together and equate all digestive “carcinoids” with the appendiceal “carcinoid,” probably the best known NET with the most benign behavior [28]. However, in recent years it has become clear that gastroenteropancreatic NETs, especially foregut NETs, are heterogeneous in their morphological and biological features. In the last two decades efforts were therefore made by the WHO to define NET features that discriminate true benign behavior (low risk) from low-grade malignant well-differentiated NETs in the different parts of the digestive system. Although the new WHO classification is an important step toward defining the diverse tumor biology of NETs, further efforts are necessary to improve the prognostic assessment of the individual NET. The demand for standards in the stratification and treatment of patients with gastroenteropancreatic NETs prompted the recently established European Neuroendocrine Tumor Society (ENETS) to define guidelines [22, 33]. Such guidelines underwent scrutiny for consensus in the first of two meetings entitled “Consensus Conference on the ENETS Guidelines for the Diagnosis and Treatment of Neuroendocrine Gastrointestinal Tumors, Part 1: Foregut Tumors” held in Frascati (Rome, Italy) from November 2–5, 2005. During this meeting the clinical need for a tumor–node–metastasis (TNM) classification of gastroenteropancreatic NETs was felt. Here we report the TNM staging classification proposal for foregut NETs that was approved at this consensus conference. In addition, we suggest a simple grading system with some pointers that may help to standardize the prognostic assessment of gastroenteropancreatic NETs. Materials and methods Sixty-two experts in the field of digestive endocrine tumors from 20 different countries attended the Consensus Conference. The attendees represented all medical branches involved in managing patients with gastroenteropancreatic NETs. They formed four working groups according to their specific clinical expertise: (1) pathology and genetics (11 participants, all listed as authors and G. Klöppel), (2) surgery (10 participants, including the coauthors H. Alhman and M. Falconi), (3) imaging and radiology (10 participants), (4) medicine and clinical pathology (31 participants, including the coauthors M. Caplin, W.W. de Herder, B. Erikssson, and B. Wiedenmann). The Conference was divided sequentially into eight sessions devoted to specific topics on an anatomical basis (gastric NET sessions 1–2, duodenal NET, pancreatic NET sessions 1–4, and poorly differentiated endocrine carcinomas). A working booklet with the ENETS guidelines and specific queries had been prepared in advance by the Organizing Committee. The work was organized such that, after a short case presentation in a plenary session, each working group gathered separately to discuss group-specific questions. Once agreement was reached within each group, consensus statements were discussed and approved or rejected by all participants gathered in the plenary session. This procedure was followed for all eight sessions. The TNM staging proposal was made by the Pathology and Genetics working group and amended and approved by the plenary session of the consensus conference. The grading system was discussed and defined by the Pathology and Genetics working group only. Results and discussion The consensus guidelines are reported elsewhere. Here, we report the TNM staging proposal for gastroenteropancreatic NETs of the foregut together with a grading system that may be relevant for the prognostic assessment by the pathologist. The foregut NETs were separated into gastric, duodenal (including ampulla and proximal jejunum), and pancreatic NETs, but were not distinguished according to specific functional activity, main tumor cell type, and specific genetic background. TNM staging proposal (see Tables 1, 2 and 3) The currently published TNM format was adopted as working template [29]. Tumor The proposed definition of tumor in situ applies to the stomach only and adheres to the literature [30]. No definition is given for the duodenum and pancreas because none has been agreed upon in spite of recent working proposals [1, 2]. The size limits indicated for T1 are those defined by the WHO for tumors with “benign behavior” according to site-specific clinicopathological correlations [5, 7, 31]. Similarly, for T2 of the stomach and duodenum, the sizes are those indicated for tumors of “uncertain behavior.” In the pancreas the size limit given for T2 needs to be validated [5]. Deeply invasive tumors are included under the T3 and T4 definitions, taking into account site-specific features. Nodes N1 indicates the presence of any single or multiple metastases in regional lymph nodes, according to TNM rules. Although the presence of regional lymph-node metastases is, per se, a negative prognostic factor in gastroenteropancreatic NETs [11], the prognostic significance of the number of metastatic nodes is not known. In light of this, stage 3B of Tables 1, 2 and 3 is proposed to mark the N1 status for future validation. Table 1Proposal for a TNM classification and disease staging for gastric endocrine tumorsTNMT—primary tumor TXPrimary tumor cannot be assessed T0No evidence of primary tumor TisIn situ tumor/dysplasia (<0.5 mm) T1Tumor invades lamina propria or submucosa and ≤1 cm T2Tumor invades muscularis propria or subserosa or >1 cm T3Tumor penetrates serosa T4Tumor invades adjacent structuresFor any T, add (m) for multiple tumorsN—regional lymph nodes NXRegional lymph nodes cannot be assessed N0No regional lymph node metastasis N1Regional lymph node metastasisM—distant metastasis MXDistant metastasis cannot be assessed M0No distant metastases M1aDistant metastasisStage Disease stages  Stage 0TisN0M0  Stage IT1N0M0  Stage IIaT2N0M0   IIbT3N0M0  Stage IIIaT4N0M0   IIIbAny TN1M0  Stage IVAny TAny NM1aM1 specific sites defined according to Sobin and Wittekind [29]Table 2Proposal for a TNM classification and disease staging for endocrine tumors of the duodenum/ampulla/proximal jejunumTNMT—primary tumor TXPrimary tumor cannot be assessed T0No evidence of primary tumor T1Tumor invades lamina propria or submucosa and size ≤1 cma T2Tumor invades muscularis propria or size >1 cm T3Tumor invades pancreas or retroperitoneum T4Tumor invades peritoneum or other organsFor any T, add (m) for multiple tumorsN—regional lymph nodes NXRegional lymph nodes cannot be assessed N0No regional lymph node metastasis N1Regional lymph node metastasisM—distant metastases MXDistant metastasis cannot be assessed M0No distant metastases M1bDistant metastasisStage Disease stages  Stage IT1N0M0  Stage IIaT2N0M0    IIbT3N0M0  Stage IIIaT4N0M0    IIIbAny TN1M0  Stage IVAny TAny NM1aTumor limited to ampulla of Vater for ampullary gangliocytic paragangliomabM1 specific sites defined according to Sobin and Wittekind [29]Table 3Proposal for a TNM classification and disease staging for endocrine tumors of the pancreasTNMT—primary tumor TXPrimary tumor cannot be assessed T0No evidence of primary tumor T1Tumor limited to the pancreas and size <2 cm T2Tumor limited to the pancreas and size 2–4 cm T3Tumor limited to the pancreas and size >4 cm or invading duodenum or bile duct T4Tumor invading adjacent organs (stomach, spleen, colon, adrenal gland) or the wall of large vessels (celiac axis or superior mesenteric artery)For any T, add (m) for multiple tumorsN—regional lymph nodes NXRegional lymph node cannot be assessed N0No regional lymph node metastasis N1Regional lymph node metastasisM—distant metastases MXDistant metastasis cannot be assessed M0No distant metastases M1aDistant metastasisStage Disease stages  Stage IT1N0M0  Stage IIaT2N0M0   IIbT3N0M0  Stage IIIaT4N0M0   IIIbAny TN1M0  Stage IVAny TAny NM1aM1 specific sites defined according to Sobin and Wittekind [29] Distant metastasis M1 indicates the presence of any single or multiple metastases at any distant anatomical site (including nonregional nodes). Because there is evidence that extrahepatic bone metastases are a particularly ominous sign [6, 20], it is recommended to specify the anatomical site of the metastasis according to the TNM classification rules (PUL, pulmonary; HEP, hepatic; OSS, osseous; etc.) [29]. Staging The proposed staging system lists stage 0 only for the stomach because this is the only anatomical site where Tis is defined. Stage I encompasses the T1 NETs with limited growth. Stage II identifies tumors that are larger in size or more invasive, either T2 or T3, though always in the absence of metastasis. At stage III the increased malignancy refers either to invasion into surrounding structures (stage IIIa) or to the presence of regional node metastasis (stage IIIb). Stage IV always implies the presence of distant metastasis. Grading proposal (see Table 4) Grading It has been widely discussed and generally accepted that no histological grading system effectively predicts the behavior of well-differentiated endocrine tumors. The major obstacle to developing a practically effective grading system is the fact that severe cytological atypia, as, for instance, in pheochromocytomas, has no impact on the clinical behavior and malignancy of such tumors. However, recent studies in well differentiated NETs of the foregut, including the pancreas, and of the midgut have shown the usefulness of a grading system [10, 25, 32]. Thus, well-differentiated endocrine tumors with a more solid appearance and distinct proliferative activity, which also lead to difficulties in the differential diagnosis vs poorly differentiated endocrine carcinomas, seem to have a worse prognosis than NETs without these features [16–18, 22, 23]. It was therefore decided to introduce a grading system that could be of help in distinguishing the well-differentiated NETs into G1 and G2 categories. As a working suggestion, we propose to apply to foregut NETs a grading system modified from that adopted by the WHO for endocrine tumors of the lung, though exclusively referring to the proliferation status. In brief (see Table 4), three tumor categories are identified: G1, <2 mitosis per 2 mm2 [10 high power fields (HPF) 40× magnification] and/or Ki-67 index ≤2%; G2, 2–20 mitosis per 2 mm2 and/or Ki-67 index between 3 and 20%; G3 with 21 or more mitosis per 2 mm2 and Ki-67 index >20%. Table 4Grading proposal for foregut (neuro)endocrine tumorsGradeMitotic count (10 HPF)aKi-67 index (%)bG1<2≤2G22–203–20G3>20>20a10 HPF: high power field=2 mm2, at least 40 fields (at 40× magnification) evaluated in areas of highest mitotic densitybMIB1 antibody; % of 2,000 tumor cells in areas of highest nuclear labeling In general, G1 and G2 should refer to well-differentiated NETs displaying diffuse and intense expression of the two general immunohistochemical neuroendocrine markers, chromogranin A and synaptophysin [26]. Punctate necrosis is, per se, indicative of a more aggressive tumor, pointing to a G2 status, which, however, has to be confirmed by the mitotic count. G3 indicates a poorly differentiated neuroendocrine carcinoma. It has high mitotic counts/Ki-67 index, is often associated with fields of necrosis, and shows significantly reduced chromogranin A expression, while maintaining intense staining for synaptophysin. It is relevant to remind here that the diagnosis of G3 carcinoma is based on a specific histologic pattern according to the current WHO criteria [5, 7, 31]. In addition, the clinical behavior of G3 poorly differentiated neuroendocrine carcinomas of the gastroenteropancreatic tract does not necessarily correspond to that of small cell cancers of the lung or of any other sites. Mitotic count and Ki-67 index We propose that mitoses should be counted on hematoxylin and eosin-stained slides in at least 40 HPFs, where possible. The mitoses should be assessed in areas where they are most frequent after a general slide survey. For Ki-67 assessment, the MIB1 antibody is recommended at the conditions that have been established at the laboratory in question. The Ki-67 index should be assessed in 2,000 tumor cells in areas where the highest nuclear labeling is observed (often but not exclusively at the tumor periphery). Concluding remarks Requests for standardization in the management of patients with gastroenteropancreatic NETs recently resulted in the development of several guidelines, including those proposed by ENETS [16, 17, 22, 23]. However, it was never attempted to reach consensus on specific practical issues. The TNM staging system we propose here was developed to meet a clinical need, is based on the current WHO classifications of endocrine and digestive tumors, and is the result of a consensus conference held by specialists involved in the management of digestive endocrine tumor patients. Along the same line, the grading system described here attempts to close the gap between the advances of the most recent WHO classifications and the need for a better prognostic assessment of NETs. It is obvious, of course, that all our proposals have to be validated by future clinicopathological work.

Diabetologia-3-1-2039867

Anthropometry, carbohydrate and lipid metabolism in the East Flanders Prospective Twin Survey: heritabilities

Aims/hypothesis We determined the genetic contribution of 18 anthropometric and metabolic risk factors of type 2 diabetes using a young healthy twin population. Introduction Type 2 diabetes is a heterogeneous disease that involves both genetic and environmental factors. Its incidence is rising rapidly worldwide and consequently much research is focused on the genetic components of the disorder, in order to get a better understanding of the pathogenesis and eventually to achieve better, more personalised diagnostics, treatment and prevention [1]. The starting point in the search for genes is to estimate the degree of heritability of intermediate traits leading to the disease in the studied population. Heritability is the proportion of phenotypic variation of a trait that can be attributed to genetic variation [2]. The degree of heritability is an important determinant of the power to detect and localise disease-related genes [3]. Although heritabilities can in principle be estimated from all kinds of related individuals, twin studies allow the variation to be split up into genetic, shared environmental and unique environmental components, thus offering one of the most valid estimations [2]. A variety of studies estimating heritabilities of risk factors of type 2 diabetes in adult twins have been carried out. A summary of the larger studies (>200 twin pairs) on the risk factors obesity, glucose intolerance/insulin resistance and dyslipidaemia is provided in Electronic Supplementary Material (ESM) Table 1. The heritability of BMI, the most frequently used measure of obesity, has been estimated extensively and ranges between 40 and 90% [4–21]. For all the other traits, fewer studies have been performed; the total variation of fasting glucose and fasting insulin is explained by genetic factors for 12 to 50% [22, 23] and 14 to 54% [7, 18, 22–24], respectively; for total cholesterol, LDL-cholesterol, HDL-cholesterol and triacylglycerol heritability estimates range between 0 and 98% [7, 10, 13, 17, 21, 25–34]. In addition to sample numbers and statistical methodologies used, the large variation in heritability estimates may also be caused by the genetic background of the studied populations and the environmental exposures experienced (see ESM Table 1). For the majority of intermediate traits related to type 2 diabetes, the number of larger studies performed is quite small and studies in which heritabilities of several risk factors of type 2 diabetes were estimated in the same population are even scarcer (ESM Table 1). To determine the genetic contribution to type 2 diabetes, we used variance components modelling to estimate the heritabilities of 18 anthropometric and metabolic risk factors of the disease, including parameters quantifying obesity, glucose intolerance/insulin resistance and dyslipidaemia. In addition, we also determined the heritabilities of the hormones insulin-like growth factor binding protein-1 (IGFBP-1), which has been shown to be related to several cardiovascular risk factors, and leptin, which plays a role as satiety signal regulating body composition and energy expenditure [35]. The characteristics were measured in 756 healthy twins, divided into 240 monozygotic (MZ) and 138 dizygotic (DZ) twin pairs in the age range of 18 to 34 years, recruited from the East Flanders Prospective Twin Survey (EFPTS). Methods Participants The EFPTS is a population-based twin register that started in 1964 and has recorded all multiple births in the Belgian Province of East Flanders until the present day. A detailed description of the EFPTS has been published [36]. Zygosity was determined using sequential analysis based on sex, fetal membranes, umbilical cord blood groups, placental alkaline phosphatase and DNA marker analysis. Between July 1964 and May 1982, the Twin Survey had registered 2,141 twin pairs who met the World Health Organization criteria for live born infants (birthweight  ≥ 500 g or gestational age  ≥ 22 weeks if birthweight unknown). Pairs of whom one or both members were stillborn, died in neonatal or later life or suffered from major congenital malformations were excluded. We randomly contacted 803 pairs using an envelope system. To assure equally distributed groups, we stratified for birth year and zygosity. Since the twin survey was originally set up to determine chorionicity, the number of MZ twin pairs was over sampled. Eventually, 424 twin pairs (52.7%) volunteered to participate in the study. Sex, gestational age and birthweight did not differ between participants and non-participants (i.e. those who where eligible, but refused to participate or had not been contacted) (p > 0.05), although participants were slightly older than the non-participants and the proportion of monochorionic (MC) twins was higher in the group of participants (p < 0.05). A detailed description of the methods used to measure the anthropometric and metabolic characteristics of the twins is provided in the Electronic Supplementary Material. For the present analysis participants suffering from type 1 diabetes were excluded and only twin pairs of whom both members participated were incorporated. Consequently, phenotypic data were available for 378 complete pairs, consisting of 113 monozygotic men (MZM), 127 monozygotic women (MZF), 46 DZ men (DZM), 49 DZ women (DZF) and 43 DZ opposite sex (DZOS) pairs. Additionally, for the analysis of lipid and carbohydrate parameters, participants taking drugs with potential effects on lipid or carbohydrate metabolism were excluded (n = 11). Twins were randomly assigned to be the first or the second member of a pair. The Ethics Committee of the Faculty of Medicine of the Katholieke Universiteit Leuven approved the project and all participants gave informed consent. Descriptive statistical analysis Anthropometric and metabolic characteristics are expressed as mean ± SD according to chorion type, for men and women separately. BMI, sum of four skinfold thicknesses (S4SF), fat mass, IGFBP-1, fasting insulin, insulin resistance, beta cell function, leptin, total cholesterol:HDL-cholesterol ratio and triacylglycerol values had a skewed distribution. After transforming these data into natural logarithms a normal distribution was obtained and the transformed data were used when performing statistical tests. Differences in means between MZ MC, MZ dichorionic (DC) and DZ twins were calculated using the PROC MIXED method implemented in SAS version 9.1 (SAS Institute, Cary, NC, USA). A random intercept model was used, where the intercept of each twin pair was modelled as a function of the population intercept plus a unique contribution of the twin pair. In addition, we allowed the variance–covariance structure of the random intercept to differ between MZ MC, MZ DC and DZ pairs. Differences in means were considered significant if the 2 dfF test indicated p < 0.05. For all traits, effects of potential covariates were also checked using the random intercept model of PROC MIXED, where the variance–covariance structure of the random intercept was allowed to differ between MZ and DZ pairs. The variables BMI, WHR and S4SF were checked for the effect of potential confounding by sex and age; body mass, fat mass and lean body mass were checked for the effect of sex, age and height; the blood parameters were checked for the effect of sex, age and fat (BMI, WHR or S4SF). Covariates were considered significant if p < 0.05. Intra-pair correlation coefficients were calculated for MZ MC and MZ DC twin pairs and for each of the five sex by zygosity groups (MZM, MZF, DZM, DZF and DZOS) before and after adjustment for significant covariates using the Mx software package [37]. In addition, we used a linear regression analysis to test whether twin correlations differed between MZ MC and MZ DC twin pairs before and after adjusting for covariates. Twin model fitting Twin methodology makes use of the fact that MZ twins are genetically identical, whereas DZ twins share 50% of their segregating genes. Assuming that MZ and DZ twins share their common environment to the same extent, a higher concordance rate in MZ twins than in DZ twins reflects genetic influences. To estimate the genetic and environmental components of variance of the traits, twin model fitting of raw data was implemented using the statistical package Mx [37]. Scripts were downloaded from the GenomEUtwin Mx-script library (http://www.psy.vu.nl/mxbib/). Univariate twin analyses were performed, where the phenotypic variance can be decomposed into additive genetic (A, additive effects of genes on multiple loci), non-additive genetic (D, interactions between alleles at the same locus [dominance] or on different loci [epistasis]), common environmental (C, environmental effects shared by twins reared in the same family) and unique environmental effects (E, environmental effects unique to the individual). MZ twins are assumed to share the same A and D genetic variance; DZ pairs are assumed to share one-half of the A variance and one-quarter of the D variance. The C variance is assumed to be the same for both MZ and DZ twin pairs. The broad sense heritability (H2), which estimates the extent to which variation of a trait in a population can be explained by genetic variation, is defined as the proportion of genetic variance to total phenotypic variance. As non-additive genetic (D) and shared environmental effects (C) cannot be identified simultaneously in data from twins reared together, ACE and ADE models were fitted separately. The significance of variance components A, C or D in the model was tested by dropping these parameters and comparing the fit of the models [5]. To test whether genetic and environmental factors influence a trait to the same degree in men and women, we compared a quantitative heterogeneity model (variance components free to differ across sexes) with a homogeneity model (variance components equal for both sexes). In addition, we verified whether the distribution of a trait differed among men and women by testing a scalar model, which assumes that the female variance components are common multiples of the male variance components. In this model the variance components were constrained to be equal for both sexes, but total variances were allowed to differ between men and women (Fig. 1) [5]. Fig. 1Path diagram for a univariate quantitative heterogeneity (a), scalar (b) and homogeneity (c) model only presenting opposite sex pairs. Observed phenotypes (PM and PW) for the male and female twins are shown in rectangles, while latent factors (A, C and E) and latent phenotypes (LPM and LPW) are shown in circles. Path coefficients of observed variables on the different latent factors are shown in lower case: a, additive genetic effect; c, common environmental effect; e, unique environmental effect; k, scalar factor. The genetic correlation is represented by rg (1 for MZ and 0.5 for DZ twins) and the common environmental correlation is represented by rc (1 for MZ and DZ twins) If for a certain trait the correlation of the opposite-sex pairs is smaller than the correlations for the like-sexed DZ pairs, it is suggested that the correlation between additive genetic factors in opposite-sex pairs is smaller than 0.5. This indicates that different genetic factors influence this trait in men and women. To test this, a full heterogeneity model (variance components free to differ across sexes, plus the correlation between the additive genetic factors in opposite-sex pairs free to be estimated between 0 and 0.5) was compared with a quantitative heterogeneity model. All variance components were estimated both unadjusted and adjusted for the covariates. In the models, phenotypic means were adjusted for significant covariates by modelling them as definition variables in the means model. The difference in fit between the nested models was evaluated with the likelihood ratio χ2 test, which uses the difference between the −2log likelihood of the full and the restricted model. This difference is distributed as χ2. The df of the test were calculated as the difference in df between the models. When the χ2 was not statistically significant (p < 0.05), the most parsimonious model was selected, i.e. the model with the best fit given the number of df. When comparing the fit of non-nested models (e.g. ACE with ADE), the model with the lowest Akaike’s information criterion was preferred. Results Descriptive statistical analysis The mean values of the anthropometric and metabolic characteristics of the twins are presented in Table 1 according to chorion type for men and women separately. In men, fasting glucose levels were lower in DZ twins than in MZ MC and MZ DC twins. In women, MZ DC twins were younger and had lower total cholesterol and LDL-cholesterol levels than MZ MC and DZ twins (Table 1). Table 1Anthropometric and metabolic characteristics of the twin population according to chorion type for men and women separately in the EFPTSCharacteristicMenWomenMZ MCMZ DCDZp valueMZ MCMZ DCDZp valuen13492135142112141Age (years)a25.3 ± 4.525.0 ± 4.925.7 ± 4.70.3325.9 ± 4.424.0 ± 4.825.7 ± 4.60.02Body height (cm)178.0 ± 5.7 178.3 ± 7.2178.4 ± 6.50.92165.0 ± 6.3165.6 ± 6.1166.2 ± 6.40.59Body mass (kg)70.7 ± 10.569.6 ± 8.470.9 ± 10.20.6760.5 ± 9.760.5 ± 10.660.8 ± 10.20.98BMI (kg/m2)22.1 ± 1.121.8 ± 1.122.1 ± 1.20.7322.0 ± 1.221.8 ± 1.221.8 ± 1.20.86WHR (%)83.4 ± 5.483.4 ± 5.882.8 ± 5.50.8473.4 ± 4.573.3 ± 4.672.5 ± 4.30.30S4SF (mm)b35.9 ± 1.534.6 ± 1.534.8 ± 1.50.8957.1 ± 1.455.6 ± 1.553.6 ± 1.40.50Fat mass (kg)b11.9 ± 1.611.2 ± 1.512.2 ± 1.50.3316.9 ± 1.316.7 ± 1.416.8 ± 1.30.95Lean body mass (kg)57.7 ± 6.957.6 ± 6.057.7 ± 6.60.9942.9 ± 5.143.1 ± 5.643.4 ± 6.00.89IGFBP-1 (ng/ml)b11.0 ± 1.910.3 ± 1.811.5 ± 1.90.4616.8 ± 2.0 16.6 ± 2.218.4 ± 2.30.58Fasting insulin (pmol/l)b33.5 ± 1.5 32.7 ± 1.632.7 ± 1.60.9438.6 ± 1.535.8 ± 1.640.1 ± 1.50.24Insulin resistance (HOMA)b1.2 ± 1.61.2 ± 1.61.2 ± 1.60.691.3 ± 1.51.2 ± 1.71.4 ± 1.50.26Fasting glucose (mmol/l)5.0 ± 0.54.9 ± 0.44.8 ± 0.40.014.5 ± 0.34.6 ± 0.44.6 ± 0.40.23Beta cell function (HOMA)b77.2 ± 1.578.2 ± 1.686.5 ± 1.70.10135.2 ± 1.7115.7 ± 1.6126.7 ± 1.60.17Leptin (ng/ml)b1.7 ± 3.01.4 ± 2.91.7 ± 3.00.4211.5 ± 2.111.2 ± 2.211.5 ± 1.90.92Total cholesterol (mmol/l)4.8 ± 0.94.7 ± 1.04.9 ± 1.10.315.3 ± 0.94.9 ± 0.85.3 ± 1.00.006LDL-cholesterol (mmol/l)3.0 ± 0.92.9 ± 0.93.1 ± 1.00.353.1 ± 0.82.7 ± 0.82.9 ± 0.80.012HDL-cholesterol (mmol/l)1.3 ± 0.31.3 ± 0.31.4 ± 0.40.941.8 ± 0.41.8 ± 0.41.9 ± 0.40.39Total cholesterol:HDL-cholesterol ratiob3.6 ± 1.43.5 ± 1.33.7 ± 1.40.673.0 ± 1.32.8 ± 1.32.9 ± 1.30.34Triacylglycerol (mmol/l)b0.9 ± 1.60.9 ± 1.50.9 ± 1.50.780.9 ± 1.50.9 ± 1.50.9 ± 1.50.83NEFA (mmol/l)0.5 ± 0.20.5 ± 0.20.5 ± 0.20.050.7 ± 0.20.7 ± 0.30.7 ± 0.20.81Data are expressed as mean ± SDHOMA, homeostasis model assessmentap value calculated using standard linear regression, because convergence criteria could not be met using a random intercept modelbGeometric mean ± SDThe covariates adjusted for, the intra-pair correlations of MZ MC and MZ DC pairs, and the intra-pair correlations of each sex by zygosity group before and after adjustment are summarised in Table 2. Intra-pair correlations of the anthropometric and metabolic characteristics did not differ between MZ MC and MZ DC pairs (p > 0.05). Adjusting for covariates strongly reduced the correlations of the leptin concentrations. The correlations of the other traits were only minimally affected by adjustment. Twin correlations for total cholesterol and LDL-cholesterol were high in both MZ and DZ twin pairs, indicating a common environmental effect. The DZ correlations of IGFBP-1, fasting insulin and insulin resistance were less then one-half of the MZ correlations, suggesting that non-additive genetic effects might be important. Correlations of the other traits were in agreement with a model containing additive genetic and unique environmental influences (Table 2). Table 2Intra-pair correlations of MZ MC and MZ DC pairs, and of each sex by zygosity group before and after adjusting for covariates in the EFPTSCharacteristicMonozygoticMonozygoticDizygoticCovariatesMCDCMenWomenMenWomenOSn (of pairs)138102113127464943Body mass 0.85/0.790.84/0.760.86/0.820.76/0.730.38/0.280.58/0.570.26/0.35Sex, age, heightBMI0.80/0.780.81/0.770.86/0.830.77/0.740.46/0.310.53/0.560.47/0.46AgeWHR0.87/0.690.88/0.710.79/0.740.70/0.660.39/0.280.44/0.480.31/0.15Sex, ageS4SF0.82/0.730.84/0.750.81/0.790.72/0.680.46/0.370.64/0.630.36/0.31Sex, ageFat mass0.85/0.780.85/0.750.85/0.820.73/0.690.46/0.360.43/0.460.42/0.35Sex, ageLean body mass0.93/0.810.93/0.790.86/0.820.79/0.780.43/0.390.65/0.580.25/0.39Sex, age, heightIGFBP-10.49/0.390.60/0.530.55/0.510.45/0.430.31/0.210.08/0.060.12/−0.05Sex, age, BMIFasting insulin0.57/0.480.52/0.500.49/0.450.58/0.520.07/0.130.18/0.190.07/−0.01Age, S4SFInsulin resistance0.54/0.470.53/0.500.49/0.460.57/0.510.03/0.080.14/0.170.04/−0.05Sex, age, S4SFFasting glucose0.74/0.660.73/0.670.65/0.650.70/0.690.28/0.240.57/0.600.31/0.32Sex, BMIBeta cell function0.71/0.580.66/0.600.52/0.500.68/0.660.32/0.400.47/0.460.37/0.33Sex, age, S4SFLeptin0.85/0.530.85/0.570.70/0.580.64/0.520.35/0.020.66/0.310.38/0.37Sex, age, S4SFTotal cholesterol0.76/0.740.77/0.720.78/0.740.73/0.730.52/0.510.51/0.440.63/0.51Age, S4SFLDL-cholesterol0.78/0.790.81/0.740.81/0.790.77/0.750.52/0.510.59/0.520.68/0.58Sex, age, S4SFHDL-cholesterol0.78/0.710.84/0.770.75/0.760.74/0.740.31/0.300.44/0.440.52/0.52Sex, S4SFTotal cholesterol:HDL-cholesterol ratio0.81/0.780.86/0.810.84/0.820.78/0.760.50/0.410.50/0.490.65/0.54Age, WHRTriacylglycerol0.58/0.560.67/0.600.59/0.540.63/0.600.29/0.410.34/0.340.16/0.17Sex, S4SFNEFA0.49/0.350.43/0.390.39/0.380.34/0.370.10/0.140.20/0.180.25/0.26Sex, S4SFValues are unadjusted intra-pair correlation/adjusted intra-pair correlation.OS, opposite sex Twin model fitting The variance components and 95% CIs of the best fitting models before and after adjustment for covariates are presented in Table 3. The best fitting model for lean body mass and the obesity parameters body mass, BMI, WHR, S4SF and fat mass was an AE model containing a major genetic component. For total cholesterol and LDL-cholesterol, the ACE model was the best fitting model. However, after adjusting for covariates, the AE model became the best fitting model. For IGFBP-1, fasting insulin and insulin resistance, a DE model containing a non-additive genetic and unique environmental component had the best fit. The variation of the remaining blood parameters, including fasting glucose, beta cell function, leptin, HDL-cholesterol, total cholesterol:HDL-cholesterol ratio, triacylglycerol and NEFA levels were best explained by an AE model (Table 3). Table 3Variance components estimates and 95% CIs of best-fitting models expressed in percentagesCharacteristicUnadjustedAdjusted for covariatesModelSexa2 (H2)c2d2 (H2)e2SModelSexa2 (H2)c2d2 (H2)e2SBody mass AEM90 (86–93)––10 (7–14)AEM84 (78–88)––16 (12–22)AEW80 (73–85)––20 (15–27)AEW74 (66–80)––26 (20–34)BMI AEM87 (82–90)––13 (10–18)AEM85 (79–89)––15 (11–21)AEW76 (68–82)––24 (18–32)AEW75 (67–81)––25 (19–33)WHRAEM94 (92–96)––6 (4–8)AE70 (63–75)––30 (25–37)M>WAEW70 (59–77)––30 (23–41)S4SFAEM88 (83–91)––12 (9–17)AE74 (68–79)––26 (21–32)M>WAEW75 (67–81)––25 (19–33)Fat mass AEM88 (84–92)––12 (8–16)AEM81 (74–86)––19 (14–26)AEW74 (65–80)––26 (20–35)AEW70 (60–77)––30 (23–40)Lean body mass AEM97 (96–98)––3 (2–4)AE81 (76–84)––19 (16–24)M>WAEW82 (76–87)––18 (13–24)IGFBP-1 DE––56 (47–64)44 (36–53)W>MDE––47 (36–56)53 (44–64)W>MFasting insulin DE––54 (45–62)46 (38–55)DE––49 (39–58)51 (42–61)Insulin resistanceDE––52 (42–60)48 (40–58)DE––48 (38–57)52 (43–62)Fasting glucose AE72 (66–77)––28 (23–34)M>WAE67 (60–73)––33 (27–40)M>WBeta cell function AE69 (63–75)––31 (25–37)AE62 (54–69)––38 (31–46)Leptin AEM92 (88–94)––8 (6–12)AE53 (44–61)––47 (39–56)M>WAE W63 (52–72)––37 (28–48)Total cholesterol ACE49 (28–74)29 (4–48)–22 (18–28)AE75 (69–79)––25 (21–31)LDL-cholesterol ACE43 (24–67)37 (14–54)–20 (16–25)M>WAE78 (73–82)––22 (18–27)HDL-cholesterol AE 82 (78–86)––18 (14–22)W>MAE76 (70–81)––24 (19–30)W>MTotal cholesterol:HDL-cholesterol ratioAEM88 (83–91)––12 (9–17)AE79 (74–83)––21 (17–26)M>WAEW79 (72–85)––21 (15–28)TriacylglycerolAE61 (52–68)––39 (32–48)AE58 (49–66)––42 (34–51)NEFAAE46 (36–54)––54 (46–64)W>MAE37 (25–47)––63 (53–75)W>MValues are variance components estimates (95% CI)M, men; W, women; a2, additive genetic variance; c2, common environmental variance; d2, non-additive genetic variance; e2, unique environmental variance; S, scalar effect; H2, broad heritabilityQuantitative sex differences were present in body mass, BMI, WHR, S4SF, fat mass, lean body mass, leptin and total cholesterol:HDL-cholesterol ratio, because variance components estimates were significantly different between men and women (Table 3). The influences of additive genetic factors were larger in men than in women. For some traits, scalar sex differences were observed, implying that although variance components are equal across sexes, the total variances differ. As a result, total variance of IGFBP-1, HDL-cholesterol and NEFA levels in women was larger than in men, but smaller for fasting glucose and HDL-cholesterol levels (Table 3). After adjusting for covariates, quantitative sex differences remained significant only for body mass, BMI and fat mass (Table 3). In addition, scalar sex differences were significant for WHR, S4SF, lean body mass, IGFBP-1, fasting glucose, leptin, HDL-cholesterol, total cholesterol:HDL-cholesterol ratio and NEFA levels. Total variance of IGFBP-1, HDL-cholesterol and NEFA levels was larger in women than in men, but smaller for WHR, S4SF, lean body mass, fasting glucose, leptin and total cholesterol:HDL-cholesterol ratio (Table 3). The adjusted correlations of WHR, fasting insulin and triacylglycerol levels for opposite-sex pairs were smaller than the correlations for like-sexed DZ pairs, suggesting that different genetic factors influence this trait in men and women (Table 2). However, the correlations between the additive genetic factors in opposite-sex pairs were not significantly smaller than 0.5 (p > 0.05). Broad-sense heritability estimates (encompassing both additive and non-additive genetic effects) were slightly lower after adjusting for covariates, with the exception of total cholesterol and LDL-cholesterol, which had a higher heritability after adjustment (Table 3). In summary, heritability estimates of body mass, BMI and fat mass were 84, 85 and 81% for men and 74, 75 and 70% for women, respectively. WHR, S4SF and lean body mass had heritability estimates of 70, 74 and 81%, correspondingly. For fasting glucose, fasting insulin, insulin resistance, beta cell function and IGFBP-1 levels, genetic factors explained 67, 49, 48, 62 and 47% of the total variation, respectively. Finally, heritability estimates of total cholesterol, triacylglycerol, HDL-cholesterol, LDL-cholesterol, total cholesterol:HDL-cholesterol ratio, NEFA and leptin levels were 75, 58, 76, 78, 79, 37 and 53%, respectively (Table 3). Discussion To determine the genetic contribution to type 2 diabetes, we estimated the heritabilities of 18 anthropometric and metabolic characteristics related to this disease, including parameters quantifying obesity, glucose intolerance/insulin resistance and dyslipidaemia. The traits were measured in 138 DZ and 240 MZ young healthy twin pairs recruited from the population-based EFPTS [36]. Obesity, defined as excess accumulation of adipose tissue, is a complicated trait to measure and numerous methods exist for its determination [38]. In the present study several parameters defining obesity, including body mass, BMI, S4SF, WHR and fat mass were used and heritability estimates ranged from 70 to 85%. According to ESM Table 1, the heritabilities identified by us for BMI and body mass are in agreement with those reported in other twin studies, but the heritability identified by us for WHR is somewhat higher [7, 15, 18, 39]. Heritability estimates for S4SF and fat mass are not often reported and therefore we were not able to make a valid comparison. Despite the variety of measurements used to quantify obesity, the heritabilities of the different measurements were consistently in a high range, indicating that the contribution of genetic factors to total phenotypic variation in obesity is of great importance in the studied twin sample. Quantitative sex differences were significant for body mass, BMI and fat mass, with heritability estimates of 84, 85 and 81% in men and 74, 75 and 70% in women, respectively. Our results confirm those from other twin studies reporting quantitative sex differences in the heritability of body mass and BMI [4–6, 9, 20, 40]. However, in other twin studies the highest heritability estimates were not always observed in men (ESM Table 1), indicating that the influence of environmental factors in the total variation of these traits according to sex may vary in different populations. In the present study, 49% of the total variation of fasting insulin was explained by genetic factors, which is in agreement with other twin studies (ESM Table 1) [41, 42]. The heritability of fasting glucose (H2 = 67%) was higher than those observed in other twin studies, where it ranged between 12 and 58% (ESM Table 1) [42, 43]. The heritabilities of beta cell function (H2 = 62%) and insulin resistance (H2 = 48%) was not higher than those of fasting glucose (H2 = 67%) and fasting insulin levels (H2 = 49%). This indicates that no additional power was obtained by calculating these alternative phenotypic markers with the homeostasis model assessment. The variation of total cholesterol, triacylglycerol, HDL-cholesterol, LDL-cholesterol and total cholesterol:HDL-cholesterol ratio was primarily explained by additive genetic factors, accounting for 75, 58, 76, 78 and 79% of the total variance, respectively. No evidence was observed for quantitative sex differences in heritability estimates for these traits. The percentages are slightly higher than the heritability estimates reported in the literature, but are in the same high range (ESM Table 1). To the best of our knowledge, ours is the first large twin study reporting the heritability of NEFA levels, which was 37%. This is somewhat surprising, since excessive circulating fatty acids represent a major contributor to the development of insulin resistance [44]. The heritability determined for NEFA levels in the current study was low compared with those reported for the other blood lipid parameters, indicating that environmental factors, e.g. nutrition, may be more important in determining the variation of NEFA levels than genetic factors. Additive genetic variance explained 53% of the variation of total leptin concentration. In other western adult twin studies, heritabilities of total leptin concentrations ranged from 34 to 55% [45, 46], which is in agreement with our findings. Compared with other traits, the correlations and heritability estimates of leptin were strongly reduced after adjusting for covariates, which is probably caused by the adjustment for fat (using S4SF). This is not very surprising, since leptin is synthesised in and secreted from adipose tissue and plasma leptin levels are increased in obese humans in direct proportion to body fat mass [35]. The heritability of IGFBP-1 (H2 = 47%) levels determined in our study was also relatively high, since the heritability in two other twin studies ranged from 0 to 36% [41, 47]. Interestingly, the heritabilities of IGFBP-1 levels in the latter studies were determined in older twins (mean age 63 years). This might suggest that the heritability of IGFBP-1 decreases with advancing age. Longitudinal twin studies need to be carried out to verify this observation. We are the first to report that a model containing non-additive instead of additive genetic factors is the best-fitting model for insulin resistance, fasting insulin and IGFBP-1 levels [7, 18, 22, 24, 41–43]. In addition to the twin model fitting procedure, twin correlations can also give an impression of whether non-additive genetic effects might be involved, i.e. a DZ correlation of less than one-half of the MZ correlation suggests a contribution of non-additive genetic effects [18, 41–43]. The smaller sample size of some studies could be responsible for their not observing a DE model. Furthermore, the presence of non-additive genetic effects has not always been tested. However, if non-additive genetic effects are observed, this implies that a part of the variance is explained by interactions between alleles at the same locus or on different loci. Although this observation is likely to be the result of lack of power, we cannot exclude real non-additive genetic effects in the absence of additive genetic effects, as these three traits are closely related to or are a direct gene product. MZ MC twins have a more adverse intra-uterine environment than MZ DC and DZ twins, resulting in a significantly lower weight at birth and higher perinatal mortality and morbidity [36]. Hence, it has been hypothesised that for disorders like type 2 diabetes, in the development of which prenatal environment plays a role, the classical twin study might be an unreliable method of estimating heritabilities [48, 49]. In the current twin sample, intra-pair correlations did not differ between MZ MC and MZ DC twins, suggesting that the chorion type of MZ twins did not bias the heritability estimates of the studied traits. With the exception of NEFA levels, the genetic contribution to the traits studied in our twin sample was high. When comparing the heritability values found by us with those reported in other twin studies (Electronic Supplementary Material Table 1), our heritability estimates are in a slightly higher range. An important aspect affecting heritability estimates is the accuracy of the zygosity determination used in the present study via sequential analysis based on sex, fetal membranes, umbilical cord blood groups, placental alkaline phosphatase and DNA marker analysis. In other twin studies, zygosity was often determined using questionnaires [4], which sometimes leads to misclassification and consequently to underestimation of heritability [50]. Our high heritability estimates might also be a consequence of the homogeneous composition of the studied sample. All twins included were born in the Belgian province of East Flanders, which is a relatively small but rather densely populated area [36]. Another factor possibly contributing to the high heritability estimates is age, as shown by the fact that several large twin studies have reported a decrease in the heritability of obesity (BMI) in adults with increasing age [5, 12, 16]. Our twin sample is young (mean = 25 years) and the age range is small (18–34 years). As the power to detect quantitative trait loci is positively correlated with the size of the genetic effect [3], our study suggests that the search for candidate genes might be more efficient in a young homogeneous population. In conclusion, this study is the first large study to give a comprehensive overview of the heritabilities of multiple risk factors related to type 2 diabetes within the same twin sample. The variation of traits related to obesity, glucose intolerance/insulin resistance and dyslipidaemia in this Belgian twin sample is in general highly explained by genetic factors. These high heritabilities are very promising for follow-up to this research, in which, in order to explain some of the genetic variance observed, sib-pair linkage analyses will be performed using microsatellite markers located in the close vicinity of selected candidate genes. Electronic supplementary material Below is the link to the electronic supplementary material. ESM 1 (PDF 64 kb)

J_Med_Internet_Res-5-4-1550572

Adolescents Searching for Health Information on the Internet: An Observational Study

Background Adolescents' access to health information on the Internet is partly a function of their ability to search for and find answers to their health-related questions. Adolescents may have unique health and computer literacy needs. Although many surveys, interviews, and focus groups have been utilized to understand the information-seeking and information-retrieval behavior of adolescents looking for health information online, we were unable to locate observations of individual adolescents that have been conducted in this context. Introduction The Internet has become an important tool for many people with health concerns [1,2], especially for adolescents [3,4]. Concerns about confidentiality, coupled with the fact that many teenagers find accessing care through traditional providers difficult [5], make access to information via the Internet particularly important. Given rapidly-expanding Internet access among young people, it is not surprising, then, that more than 70% of 15 to 17 year-olds say they have used the Internet to look up health information (written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA). This percentage is likely to increase if Internet access from home continues to rise as it has in recent years [6]. Because of the enormous amount of unstructured online content, it is crucial to understand how youth navigate through the Web to find health information. Prior research, primarily from library and information science literature and education literature, has highlighted several search characteristics that are either unique or more pronounced in adolescents. For example, adolescents take more time to complete online tasks than college students [7], search less systematically [7- 10], have difficulty formulating search queries due to misspelling and problems with the level of specificity [8- 11], utilize less-advanced search syntax [7], and rarely consider the source of Web pages [8,9]. While informative, this literature is based primarily on adolescents searching for answers to homework questions rather than health information. Searching for online health information involves distinctive challenges including unfamiliar terminology [12]; encounters with pornography-blocking software (written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA), [13]; and the importance and difficulty of determining health information quality [14]. However, despite the need for research that details the online search behavior of health consumers, the authors were only able to locate a few articles in which health science researchers actually observed, recorded, and analyzed consumers of any age searching for health information [14- 16]. Instead, surveys (eg, written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA; and [1,2,4]) have been the predominant method used to understand health consumers' online searching behavior, despite problems with participant recall and the inability of surveys to capture specific search tactics. In addition, the authors found a handful of studies in the medical informatics literature that have also looked at log data from particular medical Web sites, but these studies are also limited in scope since they do not observe the actual searcher or see the broader context in which the searcher is acting [17] (see also [18] for similar studies performed on search engine data). The value of directly observing users was demonstrated in the Eysenbach study, which revealed that adults said they paid attention to the source of health sites during interviews, although this behavior was not found during the actual observations [15]. Observational research specific to the adolescent age group and online search behavior for health information is also sparse. There have been some good surveys that answer many useful questions concerning why adolescents go to the Internet, what they search for, if they find it, and what they do with it (written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA;and [4,19]). The only observational study we were able to locate conducted 27 focus groups where groups of adolescents searched online for health information as they discussed their own experiences [14]. Many of the findings concerning adolescent search behavior found in the library and information science literature were confirmed and additional issues were raised, including concerns about low health literacy and trouble judging the quality of information, that may be more pronounced in adolescents than adults. However, that study only begins to paint a picture of adolescent search behavior for online health information, because the searches were performed in a group setting and the success, failure, and specific search tactics used were not coded or analyzed. The study reported here provides a more in-depth understanding of how adolescents search for health information using the Internet and what implications this may have on access to health information. To capture enough detail, the study recorded specific actions taken by adolescents which were later coded and analyzed. Participants were encouraged to share their thought process out loud as they searched for answers to a list of predetermined health questions. The result was a rich set of both quantitative and qualitative data that was thoroughly analyzed for common themes and events. Specific questions of interest include, but are not limited to: What are the various search strategies used? What factors contribute to finding correct and useful answers? When using a search engine, how many results pages are viewed and utilized? What types of search strings are entered into search engines? Answers to these and related questions should be of interest to a number of parties including educators (eg, health educators, librarians, teachers), Web site and search engine designers, health care practitioners, and researchers (eg, to create a sample of URLs by simulating online searching behavior [20]). Methods Sample Twelve students from 1 middle school (N= 4) and 2 high schools (N = 4 and N= 4) in southeast Michigan were recruited for this study. Staff at each school were asked to select 4 students who were (a) comfortable using computers, (b) comfortable searching for information on the Internet, and (c) strong students who could afford to miss one class period. Students received a University of Michigan T-shirt, valued at roughly $8, in return for their participation. The parent or guardian of every student signed an informed consent document that described the purpose and procedure of the study. Students also signed separate assent forms with similar information. The University of Michigan Behavioral Science Institutional Review Board approved this study and the consent and assent documents. Data Collection Three methods of data collection were used. First, one of the two members of the research team present during each of the observations coded searching behavior in real time while the second member of the research team interacted with the student. Second, TechSmith Camtasia 3.0.1 commercial tracking software [21] was installed on the computer. This software captured the students' voices and took pictures of the screen (screen captures) twice per second during the entire session. Finally, a video camera was positioned to capture the screen and the students' voices, but not the students' images. Observations coded in real time were used to develop a more detailed and systematic coding system for use when reviewing the tracking software records. It is data from the tracking software coding that is reported here. All observations of adolescents were conducted during January 2002. Each school provided a room in which to conduct the observations. Students were brought to the observation room one at a time. Two researchers were present at every observation. For each student, one of the researchers first reviewed the assent form to introduce the project and obtain the student's permission to participate. The students were then asked 14 questions about demographics (age, race/ethnicity, and gender) and their prior computer use (eg, how often they use computers or the Internet, what health topics they have searched, which search engines they used, and whether they have a computer and access to the Internet at home). Once the brief interview had been completed, the observed searches began. To help the students understand the procedure and to reinforce the importance of thinking out loud while doing their searches, each student was first asked to do an easy non-health-related search looking for the next day's local weather forecast. As with the subsequent health-related searches, the local-weather question was first read to the student by a researcher and then a card with the question on it was set next to the computer in case the student needed to read it. As part of the think-aloud protocol, the experimenter asked the student to talk out loud about what they were doing, so that researchers could better understand the reasons behind the searching behavior. If a student stopped talking during the search, he or she was reminded by the observers to "keep talking," but the experimenters did not ask students to elaborate on any specific thing they said. Concurrent verbal reports more accurately reflect a subject's mental state at the time of observed behaviors than do retrospective reflections, and this minimal think-aloud protocol has been shown to slow subjects down, but not to qualitatively change their problem solving behavior [22]. After the students completed the practice local-weather search, they were given a sequence of up to 6 predetermined health information questions (see Table 1), 1 at a time. Questions were framed in a way that took into consideration the broader information concern that the question attempted to resolve. To eliminate confounding by learning effects between searches, we used a 6 x 6 Latin square to determine the order in which the questions were presented to the participants. The computer that students used was provided by the researchers, but connected to the school's network so that the students were protected from controversial or pornographic material by the same blocking or filtering software used by the school. The 3 different schools used 3 different filtering systems. Each observation session lasted one class period. No time limit was given for each question, but when the class period ended, any ongoing search was terminated and any remaining questions were skipped. Table 1 Health-related questions Your aunt was just told she has diabetes. She isn't sure what kinds of food she can or can't eat. Using the Internet, find some information for your aunt about what foods she should or should not eat. A friend recently started taking a drug called Paxil for depression. He seems to be tired all the time, and even falls asleep in class. Use the Internet to find out if the drug might be making him sleepy. Your older brother has a problem with drinking too much alcohol. He wants to go to a local Alcoholics Anonymous meeting. Use the Internet to help him find a local meeting. You want to get an HIV test, but you don't want anyone to know. You also don't have any money to pay for it. Use the Internet to find a place to get a free and confidential HIV test. For class, you need to learn about medicine that can help people stop smoking. Using the Internet, find the names of these medicines. You are about to get a tattoo, but a friend warned you that some places spread infections like HIV and hepatitis. Use the Internet to find out if this is true. Topics for the health-related questions were chosen based upon responses to a survey of adolescents conducted by the Kaiser Family Foundation (written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA). Certain topics including homosexuality, teen pregnancy, and abortion were purposefully avoided so as not to expose participants to overly-controversial information. Data Analysis After all the observations were completed, 3 researchers including a physician, health educator, and human-computer interface specialist met as a group to review the real-time coding results and to clarify or augment the coding scheme before the definitive final coding of the tracking-software records. The final coding scheme was designed to record data on the person searching, the question being asked, the time it took to find an answer, the search strategy utilized (eg, utilize search engine or directly type in URL); search strings used; number of search engine results pages reviewed; number of pages viewed within a particular site; and the use of menus, advertisements, and directories. One of the 3 coders was assigned as a primary reviewer for each of the observation sessions. The assigned primary reviewer was responsible for a detailed coding of the observation session and any coding problems were resolved in a second group discussion. The reviewers classified each of the answers found by the students as correct or incorrect, complete or incomplete, and, for location questions only, useful or not useful. To avoid being overly narrow in our classification of correct for the more open-ended questions such as the question on healthy foods for a person with diabetes, we used the following general rule for classification: to be considered correct, the content of the answer had to be the kind of information that might be discussed in a medical school or school of public health. This classification system was validated in previously-published work by the research team and resulted in a high inter-rater reliability (κ = 0.84) [13]. The more-specific questions such as the question asking about a location for an Alcoholics Anonymous meeting were considered correct if the student found a Web page listing a meeting location and time or contact phone number. Answers were complete if the students were able to answer all parts of the question. For example, if the student found a discussion about HIV transmission by tattoo parlors, but did not find an answer about hepatitis it was classified as incomplete. Useful answers pertained to location questions. An Alcoholics Anonymous meeting in another state was not useful. A summary measure classifying each search as successful, partially successful, or unsuccessful was computed using the correct, complete, and useful ratings. To obtain a rating of successful, the answer had to be complete, correct, and useful. If the student gave up before finding an answer, the search was classified as unsuccessful. Results Twelve middle school students and high school students in southeast Michigan participated. Students ranged in age from 12 to 17 years old, with a mean of 14 years. Half of the students were female. Of the 12 students, 7 were white, 2 were African American, 1 was Indian American, 1 was Hispanic, and 1 was Asian American. Of the 12 students, only the 6 oldest students had searched for health information on the Internet before. The variation by age is consistent with other findings that youth age 15 to 17 years are significantly more likely to have looked up health information (32%) than youth age 12 to 14 years (18%) [23]. All of the students, however, had computers and access to the Internet at home. Students reported using a computer from 1 hour per week to 3 hours per day, with a mean of 12.3 hours per week. Eleven students attempted all 6 searches, while the remaining student attempted 3, for a total of 69 searches. One search was not included since the Internet connection was not working properly, making a total of 68 searches that were analyzed. Searches took an average of 5 minutes and 41 seconds, ranging from just under a minute to nearly 24 minutes. This time frame is essentially the same as Eysenbach recorded for adults [15]. Although direct comparison is inappropriate since different questions were asked, the similar order of magnitude is suggestive. Overall Search Strategy As students thought aloud, the researchers got a sense of what students were looking at on each page. Students seemed to skip around a lot, and didn't skim results pages or specific Web sites in any methodical or thorough ways, sometimes missing links or text that contained the answer to questions. This is also consistent with findings from non-health-related searching behavior as summarized in Hsieh-Yee [24]. Table 2 Distribution of pages viewed per site Pages Viewed Per Site Sites n % Cumulative % 1 143 70.4 70.4 2 27 13.3 83.7 3 11 5.4 89.2 4 8 3.9 93.1 5 8 3.9 97.0 6 2 1.0 98.0 8 1 0.5 98.5 9 1 0.5 99.0 15 2 1.0 100.0 Total 203 100 Students used multiple methods to locate Web sites that they believed contained answers to the 68 questions. In 60 cases, the student started looking for an answer by visiting a search engine and entering in a search term or phrase. In 2 cases, the student started by selecting from directory menus (eg, choosing the topic health). In 6 cases, the student started by entering a URL (other than a search engine) directly into the browser address bar. In total, there were 215 attempts to access non-search-engine or directory Web sites. Nearly all of these attempts were made by following a link from a search engine either after a search or through the use of a directory. Of the 215 attempted site visits, 4 were broken links, 3 were blocked by the filters utilized at certain schools, and 5 were PDF files (read by Acrobat Reader) which students either could not download or chose not to download because downloading was too slow. This left 203 sites that were viewed with an average of 1.8 pages viewed per site. The distribution of pages visited per site is shown in Table 2. Note that the distribution is roughly consistent with a power law as observed in previous studies [25]. At a reviewer's request, this data was looked at on an individual student level. Students varied a great deal in the total number of visited sites. Eleven of the 12 students went only 1 page deep on the majority of visited sites. Although the individual-level data is not large enough to analyze more rigorously, the power law seems to operate on an individual level as well as the aggregate level. Even when students found a Web site that contained the answer to a question, they did not always find the answer. One example is the Alcoholics Anonymous site [26] where 8 of the 11 students ended up while searching for a local meeting. Although there was a link to a site that contained local information, only 3 of the 8 students were able to find the link, 1 of whom only found it on the second visit to the Alcoholics Anonymous site, after viewing a total of 16 pages within the site. Similarly, 6 of the 11 students who searched for whether or not Paxil causes drowsiness visited the official Paxil site [27]. Only 3 of the 6 students were able to successfully answer the question based upon the information they found at the site. Two of them failed to find the list of side effects and 1 of them found the list but did not understand it enough (or read it carefully enough) to answer the question correctly. Search Engine Tactics Seven search engines were used, including 2 meta-search engines (Dogpile and Locate.com). The meta-search engine Locate.com offers the user a number of search engines to choose from. Searches performed from the Locate.com Web site that utilized another search engine (eg, Yahoo!) are reported as if the search occurred on the destination search engine (eg, Yahoo!). Table 3 summarizes the number of times that a particular search engine was used. If a search engine was used multiple times while searching for an answer to the same question, it is only counted once. Because students occasionally switched search engines while trying to answer the same question, there are more searches using a search engine (79) than there are attempts to answer questions (68). In total, 6 of the 12 students used only Google, 1 used only Yahoo!, and the remaining 5 changed search engines at some point. Table 3 Search engine usage Search Engine Times Used n % Google 38 48.1 Yahoo! 13 16.5 Ask 12 15.2 MSN 7 8.9 Hotbot 6 7.6 Dogpile 2 2.5 AltaVista 1 1.3 A total of 132 search phrases were entered into the various search engines. Only 104 of those search phrases were unique. The most-frequent 2 phrases used were "diabetes" and "Paxil," each of which had 5 occurrences. There was an average of 3.6 words typed in per search phrase and 80% of the time there were 4 or fewer words per search phrase. Table 4 Distribution of search-result links viewed Bands of Search-Result Links Viewed Chosen Links n % Cumulative % Results 1-10 137 82.5 82.5 Results 11-20 8 4.8 87.3 Results 21-30 11 6.6 94.0 Results 31-40 4 2.4 96.4 Results 41-50 4 2.4 98.8 Results 51-60 1 0.6 99.4 Results 61 or more 1 0.6 100.0 Of the 132 search phrases, 30 contained at least 1 word that was misspelled (eg, "tatoo," "Alchoholics," or "smokeing"), despite the fact that students could read the correctly-spelled word on the index card containing the question. Some search engines (eg, Google) offer a feature that recommends an alternate search string with the correct spelling of a word. For example, if a student typed "alchoholics anonymous," the first page of results began with, "Do you mean 'alcoholics anonymous?'" Students were offered a new search string with correct spelling on 15 separate occasions, but only noticed and used it 6 times. The remainder of the times they used the results that were offered for the incorrect spelling. Of the 7 students who were offered corrected spelling suggestions, only 2 ever used them. Once a search string was entered into a search engine, students varied in the number of results pages that were viewed. Students viewed only the first results page 78% of the time and 4 pages or less of results 93% of the time. Because search engines report a different number of links per page of search results, Table 4 reports how often links were selected from the first 10 results, the second 10, and so on. Only 3 blocked links were encountered during all of the searches, suggesting that blocking software did not have a significant impact on these results. Successful Searching Characteristics Of the 68 questions that students attempted to answer, 7 searches were abandoned after the student gave up or, in 2 cases, when the class period ended. Of the remaining 61 searches, 47 were successful in finding a complete, correct, and useful answer to the health question and the remaining 14 were unsuccessful. Six of the unsuccessful answers were completely incorrect and not useful, 4 were useful but only partially correct, and 4 were fully correct but not useful. Several factors contributed to the success of finding a correct, complete, and useful answer. One important factor was the individual who was performing the search. Although every student answered at least 1 question correctly there was wide variation in the number of correct answers. Two students successfully answered 6 out of 6 questions, 3 students successfully answered 5 questions, 4 students successfully answered 4 questions, and the remaining 3 students only successfully answered 1 or 2 questions. While our sample of students was too small to draw conclusions from, no distinct patterns were observed that would indicate that race, gender, Internet experience, or health searching experience were significant determinants of success. However, the older adolescents (16-17 year olds) were successful 87% of the time (26 of 30) as compared to 68% (21 of 31) for the younger adolescents. Another important factor was the difficulty level of the questions themselves. Table 5 shows the failure rate for each question. The 4 partially-correct answers were split evenly between the Alcoholics Anonymous and tattoo questions. All 4 of the correct but not useful answers resulted from the HIV test question. Table 5 Unsuccessful searches by search topic Search Topic Unsuccessful Searches n % HIV test 8 38.1 Paxil 4 19.0 Alcoholics Anonymous 3 14.3 tattoo 3 14.3 smoking 2 9.5 diabetes 1 4.8 Total 21 100.0 Certain search actions led to sites that contained the answer more often than others. Overall, students found answers on 22% of the sites they accessed (47 of 215). They accessed sites in 5 ways. Although not often taken, the action with the highest probability of success (47%; 7 of 15) was following a link from 1 non-search-engine site (eg, www.aa-intergroup.org) to another site (eg, www.alcoholics-anonymous.org). In most of these cases, the student accessed the first site directly from a search engine. Clicking on search engine results led to a site where students found an answer 21% of the time (35 of 166). Success rates were similar for following a recommended link from a list or menu provided by the search engine (18%; 4 of 22). Directly typing in a URL, bypassing search engines entirely, was successful only 9% of the time (1 of 11). A sponsored link from a search engine was followed only once, and the student found an incorrect answer on that site. Another contributing factor related to success was misspelling of search terms. Of the 14 completed but unsuccessful searches, 29% (4 searches) had at least 1 misspelling compared to only 15% (7 searches) of the 47 successful searches. Perhaps even more telling, both successful and unsuccessful searches with misspellings took students 1.5 minutes longer on average than searches without misspellings. Observations confirmed that some students were unable to find an answer until they discovered and corrected their misspelling, resulting in higher quality and more-relevant results. Other search characteristics did not have statistically significant impacts on whether searches were successful, although this may have been due to small sample sizes. For example, the search engines were not significantly different in their percentages of successful searches. Similarly, the average number of words per search string was not significantly related to search success rate. (Data not shown.) Qualitative Analysis Certain common behaviors of the adolescent searchers were observed which were not apparent from the quantitative analysis. First, the students were very comfortable and confident while searching online for health information. Most students knew where they wanted to start the search and navigated using quick mouse clicks and shortcut keys. However, this characteristic was likely over-represented in our population due to their strong academic performance and Internet proficiency. Second, several searchers did not take much time in formulating a search strategy or (when applicable) choosing search terms. Instead, these searchers seemed to type in the first search string that came to mind. If the results were not what were anticipated, another search string was typed in, sometimes without even clicking on any results from the first search string. The overall approach was a trial-and-error method with frequent backtracking. The most-common problem with search strings was that they were not specific enough. For example, 2 different students typed in the search string "hiv" when looking for a place that administers free and confidential HIV tests. Third, most students quickly scanned pages, jumping from place to place within a page, rarely reading an entire paragraph. In some cases the answer to a question was contained on a page, but the student left before finding it. In other cases a link that would have led to the answer was missed. This finding supports prior research on adolescent search behavior related to nonhealth topics [7- 10]. Fourth, students mentioned that they purposefully avoided sponsored links and advertisements, despite the fact that many of the search engines present these results first. The qualitative data confirmed this practice, as only 1 sponsored link was ever selected. Finally, little to no attention was paid to the source of the answer. In the vast majority of cases, once an answer was located, it was simply assumed to be correct. Discussion When compared with prior research, the findings of this study show many similarities and a few key differences between the behaviors of adolescents and adults while searching for health information. This study found that adolescents searching for health information utilized search engines nearly every time. This finding was similar to that for adults as described in the Eysenbach study [15]. These observational studies also suggest that after-the-fact survey questions concerning the use of search engines may underestimate this behavior. For example, 2 nationally-representative surveys reported that 58% of youth (written communication, 2001 Dec; Generation RX.com Survey printouts; V. Rideout, Henry J. Kaiser Foundation, Menlo Park, CA) and 81% of adults [1,2] started seeking health information at search engines. Our study found that adolescents relied upon links from only the first few results pages, and rarely explored far within any site. These results also were similar to adult searching behaviors [15], although youth seem to be more likely to search beyond the first 10 search results. Adolescents often chose search strings that were too general and/or contained misspellings, so that they did not always find useful sites that were available. Eysenbach also reported search strings by adults that were too general [15], however, spelling seems to be more of a problem with youth. Adolescents were unsystematic in their reading of Web sites and some sites were poorly organized so that they did not always find the information they were looking for, even when it was present in a site they examined. Future research is needed to better understand if adolescents do not understand information provided on these sites, whether they simply have less patience, or some other explanation. In summary, many of the specific search tactics are similar for adults and adolescents, but a few issues related to spelling, browsing of Web sites, and understanding of content are notably different. Simulation of Searches The results from this study have implications for anyone who simulates adolescent health searches, for providers of health information, and for educators. There are many reasons to simulate adolescent health searches. For example, an educator preparing a lesson plan may want to informally simulate searches in order to anticipate what students are likely to find if given certain particular search tasks. A researcher may want to simulate adolescent searches more systematically to evaluate the availability and accessibility of information on particular topics, to evaluate which search engines should be recommended to adolescents, or to evaluate whether the installation of filtering software will have a detrimental impact on accessibility of health information [13]. Because many of the search behaviors modeled by these simulations are similar for both adolescents and adults, results from studies that simulated one or the other group likely apply to both groups. The results of this study suggest that such simulations can focus on the use of search engines, but that very-broad search terms and, especially for adolescents, common spelling errors should be considered. Ads and other nonresult links can be ignored. Since more than 80% of the links that were followed appeared in the top 10 results, and more than 95% were among the top 40, a search simulation need not consider result links beyond these. Providers of Internet Health Content Given the patterns of adolescent searching behavior found in this study, providers of health content can do several things to increase the probability that adolescents will find their sites. Since adolescents rely primarily on the first few results from search engines and do not tend to look at ads, it is important to ensure that health sites appear near the top of the results for searches on health terms. Choices of keywords in the domain name, page title, meta tags, and the first few sentences, as well as links from other sites, can all affect placement in search results. It may also be useful to include some common misspellings in meta-tag keywords and in the body of the text in order to make a site appear in the results page of searches using those misspellings of related search terms. Because most major English-language search engines no longer use the keyword feature of meta tags, site designers are left with the difficult task of working misspelled words (eg, misspelt) into the text without coming across as poor spellers themselves. It is also important that the site descriptions displayed in search engines be attractive to adolescent searchers: while our study did not analyze the various reasons that adolescents chose to follow one link over another, we did observe that they made choices based upon the link descriptions and did not simply select the first link offered. Books and articles, software, and consulting services are all widely available to improve search engine placement and to influence the short summary text that search engines extract for display in search results [28,29]. Organizations that invest large amounts of money in developing sophisticated health-information sites would do well to spend a little bit more to ensure these sites are easily found. Another area that Internet content providers should focus on is within-site navigation. Because students tend to skip around from place to place within a page and read little in sequence, it is important that sites with a significant adolescent audience are well organized, concise, and understandable. Long paragraphs, too many links, and difficult vocabulary all decrease the likelihood of adolescents finding health information they are seeking, even if it is contained within a site. Internet content producers should attempt to understand the needs of the site visitors and build hierarchal structures that reflect those needs. For example, if one of the primary needs of individuals visiting the Alcoholics Anonymous site is to find a local meeting, the first page of the site should include an obvious link (eg, "Find an AA Meeting Near You") that leads to another page that returns the nearest meetings after entering in a zip code or city name. While ease of within-site navigation is important for all visitors to health information sites, some information providers may want to develop sites targeted specifically to adolescents. While they might like the targeted information once they found it, we observed that adolescents tend to rely on general-purpose search engines. Thus, developing special youth-targeted versions of information sites may be of somewhat limited utility, unless also accompanied by advertising or education campaigns that make adolescents more likely to find such sites. Rather than changing Web sites or their presentation in search engines, it may also be useful to undertake education campaigns to improve the search strategies and tactics that adolescents use when seeking health information. It may be helpful to guide them towards youth-oriented directories or search engines, rather than general-purpose search engines. For example, both Yahoo! and Google offer directories with subcategories of sites designed for teens that cover various health topics. This approach may be facilitated by including links to such resources on the Web browser's starting page in schools and libraries. Alternatively, adolescents might be taught techniques for formulating and refining search terms at general-purpose search engines, adding or dropping more-specific words based on the kinds of results returned. They might also be taught to notice potential search term misspellings based on surprising search results. Finally, adolescents might also be taught techniques for systematically exploring within a Web site to find the kind of information they are looking for. Limitations and Future Research There are several important limitations to the interpretation of these results. First, this was not a representative or random sample of adolescents. It was a small convenience sample with a selection bias toward adolescents with strong Internet searching skills. While the results cannot be generalized to all adolescents and do not capture the full range of adolescent searching experience, we can assume that the average adolescent would have had even more trouble than our study participants in finding health information on the Internet. Second, the health-related search questions were deliberately constructed to avoid controversial topics such as safe sex, abortion, and homosexuality. Given that adolescents are often faced with health problems related to sexuality, their actual search behavior and success at finding health information related to sexuality may not be reflected in our results. Another concern is that participants may have changed their search behavior because of the presence of observers and because they were aware that their search behaviors were being recorded. For example, students who had trouble finding an answer may have persisted in their search longer than they would have in a nonresearch setting. Alternatively, because students knew they had several search questions to answer during a single class period, they may not have been as persistent as they might have been with a more personally-relevant question and less-restricted search time. Thus, the data here reflect a rough estimate of persistence for an adolescent looking for health-related information. Also, searching was conducted individually, while in practice many searches both at home and at school are conducted with friends, teachers, or family close by. While it is difficult to know how this would affect searching behavior without future research, it is possible that students would act differently (eg, receive help with spelling). Finally, while components of our classification scheme for successful versus unsuccessful searching have been previously validated, the overall scheme was modified to more accurately code the search results as correct, complete, and useful. A more-systematic validation of coding schemes for health information search results is an important area for future research. More research is needed to validate the results presented in this article, as well as determine if results vary for different populations (eg, age, race, and experience with health searching) and different health questions (eg, finding a practitioner versus finding the answer to a question). Additionally, instead of focusing on how adolescents currently search for health information, future studies may also want to explore interventions aimed at improving their searches. For example, should health portal sites designed for adolescents or online directories be used? Or would the current practice of using common search engines, but with adolescents learning improved search tactics be more effective? Also, which search strategies lead to sites that are the most likely to be accurate and influence adolescents to change their behavior? Conclusions This study provides a useful snapshot of current adolescent searching patterns. The results have implications for constructing realistic simulations of search behavior, and for both information providers and educators. Analyzing search behavior through actual observation should be a cornerstone in any effort to improve adolescents' access to health information.

Cardiovasc_Intervent_Radiol-3-1-1914265

Subintimal Double-Barrel Restenting of an Occluded Primary Stented Superficial Femoral Artery

In-stent re-stenosis is a frequent complication of endovascular stents, especially in the superficial femoral artery (SFA). Endovascular re-intervention of in- or peri-stent occlusive disease consists of recanilization through the occluded stent. In our case report, we describe the endovascular treatment of a previously placed stent in the SFA. We unintentionally passed the affected stent subintimally, in a double barrel fashion next to the first stent. The procedure was without any complications and with a successfull angiographic result. At one year follow-up the patient still has no complaints and the stent is still patent. In-stent restenosis is a frequent complication, especially in the superficial femoral artery (SFA). If in-stent restenosis occurs there are several therapeutic options available such as conservative therapy and endovascular or surgical intervention. Endovascular reintervention of in- or peri-stent occlusive disease is one of the less invasive options. Endovascular therapy means passage through the occluded stent (recanalization) by additional means and redilatation or restenting of the lesion. In some cases passage of the occluded stent is not possible. In this case report we describe the unsuccessful attempt to recanalize an occlusion in a stent in the SFA followed by the successful bypass of the stent using a subintimal technique. The concept of utilizing this secondary stent was to create enough radial force to keep open the neo-lumen next to the occluded stent. Although the per-interventional extraluminal recanalization (PIER) technique is usually performed for longer lesions of the SFA, in this case a shorter segment was successfully managed [1, 2]. An internet search for articles describing equivalent cases revealed no results. No publications on subintimal passage of a reoccluded stent and double-barrel stenting were found in the literature. Case Report A 75-year-old man underwent several percutaneous transluminal angioplasty (PTA) procedures in both upper leg vessels as well as stenting of a right SFA lesion in the adductor canal after an insufficient PTA result in an other facility (Figs. 1 and 2). Fig. 1. See text. Fig. 2. See text. The patient presented in 2004 with progressive complaints of the right leg. The patient’s maximum walking distance was 800 m but he smoked 5–10 cigarettes a day. As consistent medication the patient was on antihypertensive drugs, 40 mg simvastatin (10 mg Zocor, MSD) as well as 100 mg carbasalaatcalcium (Ascal Cardio, Viatris). There was no cardiopulmonary comorbidity. Right pedal pulses were absent with a normal pulse at the right groin. Blood pressure was 160/95 mmHg. The maximum walking distance was 160 m with a drop in the ankle/brachial index (ABI) after excercise of 0.62 to 0.23 on the right side and 0.98 to 0.55 on the left side. In conclusion, this patient had peripheral arterial occlusive disease (PaOD) staged as Fontaine class 2a. A color Doppler ultrasound (CCDS) examination of the right femoropopliteal arteries revealed in- and peri-stent restenosis with a peak systolic volume (PSV) ratio of 5.7 and distal in-stent occlusion noted over 4 cm. At that time the patient could not be persuaded to undergo any intervention and was managed conservatively. In March 2005 the patient presented again with unchanged symptoms. On physical examination the walking distance was now 240 m with an ABI drop after excercise of 0.44 to 0.22 on the right side and 0.91 to 0.52 on the left side. The patient and his attending physician discussed endovascular treatment such as recanalization of the affected vessel with PTA and/or additional stenting as a proper next option for his PaOD Fontaine class 2a lifestyle-limiting symptoms. In April 2005, an endovascular percutaneous intervention of the right SFA was performed at the Department of Radiology in our hospital. Because the patient was overweight an antegrade puncture was not a good option and the right SFA was approached by a crossover technique from the contralateral side. The right leg vessels were visualized angiographically demonstrating a stenosis of approximately 70% at the origin on the right SFA. More distally in the SFA an occlusion was noticed reaching from 4 cm proximal of the occluded stent to the upper border of the patella (The total length of this lesion was approximately 11 cm.) The 5 Fr Brite Tip sheath (Cordis, New York, USA) in the right femoral artery was changed for a long 6 Fr Balkin Up & Over flexorsheath (Cook, Bloomington, IN, USA). The sheath tip was placed in the left external iliac artery, just above the hip joint. First a PTA was performed at the origin of the SFA with a 6.0/40.0 mm balloon (Pheron, Biotronik, Switzerland) after an intravenous (IV) injection of 2,500 IU heparin, with a satisfactory angiographic result. However, it proved impossible, after several catheter-guidewire manipulation efforts, to recanalize the occluded stent with a 0.035-inch J-shaped Terumo (Terumo, USA) guidewire. Next the occluded stent was bypassed subintimally PIER-wise and the native lumen successfully re-entered just distal to the occlusion. The neo-lumen thus created was predilated with an unsatisfactory result due to recoil and consequently stented with a self-expandable 6.0/60.0 mm and a 6.0/80.0 mm stent (Absolute, Guidant, Indianapolis, IN, USA). Post-dilatation was performed with a 6.0 mm Pheron balloon. Angiographic verification of the treated arterial segment revealed an acceptable angiographic result and good runoff to patent vessels below the knee. The 2,500 IU heparin was antagonized by an IV injection of 2,500 IU protamine. The angiographic sheath was withdrawn and the puncture site managed by manual compression for 10 min without any problems. The procedure was carried out without any minor or major events or complications. One day after the procedure the patient underwent CCDS proving patency of the treated proximal SFA and newly inserted stents with no residual stenotic segments. The ABI on the right side was 1.0 and on the left side was 1.1. The patient reported a slightly painful sensation on the upper-medial side of his lower (treated) limb. On his planned check-up at the department of surgery in May 2005 the patient reported himself very satisfied with the result. The pain sensations had disappeared and there were no more limitations to his walking distance.In August 2005 his ABI on the right side was 0.85 with a drop, after exercise, to 0.25. On the left side his ABI was 0.88 with a drop, after exercise, to 0.62. On color Doppler ultrasound investigation a patent secondary stent segment and no new stenotic lesions were observed and a normal flow pattern was described. Discussion Peri- and in-stent restenosis is a common and well-known complication in the SFA region. There are several therapeutic options to solve this problem including endovascular treatment and bypass surgery. Because of our patient’s physical condition and his lifestyle-hampering symptomatology we chose was endovascular treatment in an attempt to recanalize the affected stent. During the procedure it proved impossible after several attempts to recanalize the occluded stent in the SFA, probably because of already organized and fibrosed thrombus material in the stent. The crossover was not the cause of this failure, since the pushability and steerability of the guidewire-catheter assembly was sufficient. No angioplasty could be performed in the occluded stent. After several attempts we finally managed to pass the occluded stent subintimally, successfully re-entering the native lumen distally and creating a neo-lumen side-by-side with the primary occluded stent. The neo-lumen was stented to create enough radial force to prevent its elastic recoil. Two self-expandable stents were positioned in a double-barrel fashion alongside the previously occluded stent with proper clinical and angiographic result. A retrograde puncture was preferred because of the lower complication rate compared with an antegrade approach in this obese male. Moreover a crossover introducer sheath assembly creates more possibilities for easier catheter-guidewire manipulations. Conclusion If a patient presents with claudication symptoms caused by in-stent restenosis or occlusion, one can try to recanalize the occluded stent. This is often not possible, especially if the occlusion is of an older date and consists of fibrosed tissue, mainly because of reorganizing thrombus. Mechanical debridement is an option, but this is more successful in (sub)acute (re)stenosis [3]. Subintimal passage of a occluded stent using the PIER technique could be an option in such cases. The PIER technique is intended as a primary therapy, usually for longer lesions of the SFA and for patients with severe (critical) limb ischemia, and has a 1-year patency rate of 22% and a success rate of 62% [1, 2]. As shown in our case, PIER appears to be good not only for recanalizing occlusions but also for providing options for the treatment of nonpenetrable in-stent occlusions by trespassing PIER-wise and subintimal stenting of the neo-lumen.

Cancer_Immunol_Immunother-4-1-2253649

Clinical impact of HLA class I expression in rectal cancer

Purpose To determine the clinical impact of human leukocyte antigen (HLA) class I expression in irradiated and non-irradiated rectal carcinomas. Introduction The immune system is thought to have an important function in controlling tumor growth and eliminating metastasizing tumor cells. The expression of human leukocyte antigen (HLA) class I, presenting tumor-associated antigens on the tumor cell surface, is considered as a prerequisite for an effective T cell immune response [34]. As a consequence, tumor cells with down-regulated HLA class I expression might escape this immune response, resulting in a selective outgrowth of these tumor cells. Many studies described HLA class I expression in cancer [10, 17, 22, 33]. Only limited studies have reported on the clinical impact of HLA class I expression in colorectal cancer with contrasting results. Some studies found no significant correlation between staining intensity and survival [1, 19, 20], while others found a prognostic correlation between HLA expression and survival [18, 35]. The latter two studies had in common that total absence of HLA class I resulted in a favorable prognosis as compared to patients with down-regulated expression of HLA class I of tumor cells. The discrepancy between these two studies is, whereas the one described high expression of HLA class I in tumor cells that resulted in a better prognosis as compared to the partial down-regulation of HLA class I [35], the other found the opposite [18]. These studies both analyzed a mixed population of colon and rectal cancer patients. For rectal cancer patients, the clinical impact of HLA class I expression is still unknown. Since HLA class I expression is often absent in microsatellite instable (MSI) tumors [9, 16] and MSI is more frequently observed in right-sided colon tumors than in rectal tumors [30], therefore the results obtained from a mixed population of colon and rectal cancer patients might not hold true for rectal cancer patients. The purpose of this study was to analyze the clinical relevance of HLA class I expression for rectal cancer patients. In addition to determining the impact of MSI on HLA class I expression, the tumors most at risk for MSI i.e. HLA-negative tumors were examined for MSI by determining the expression of the mismatch repair proteins, mismatch mutL homolog 1 (MLH1) and postmeiotic segregation increased 2 (PMS2), that are most absent in sporadic MSI tumors [7, 36]. Radiotherapy has been described recently to increase cell surface expression of major histocompatibility complex (MHC) class I molecules in a murine colon adenocarcinoma cell line [26]. Therefore, our study also evaluated the effect of radiotherapy on HLA class I expression in rectal cancer patients. For these purposes, HLA class I expression was evaluated in a set of 1,135 formalin-fixed paraffin-embedded rectal cancer specimens. The tumors studied were obtained at the time of surgery from patients of a prospective multicenter trial, who were randomized between standardized preoperative radiotherapy treatment followed by surgery or surgery alone [15]. Materials and methods Study population Patients were obtained from the Dutch TME trial, a multicenter trial that evaluated total mesorectal excision (TME) surgery with or without preoperative radiotherapy (5 × 5 Gray) [15]. Radiotherapeutical, surgical and pathological procedures were standardized and quality-controlled [15, 21]. Tumor staging was determined using the tumor node metastasis (TNM) classification [29]. Patients with the hereditary Lynch syndrome also known as hereditary non-polyposis colorectal cancer (HNPCC) were excluded from the TME trial. Sufficient formalin-fixed paraffin-embedded tumor material was available for 1,135 Dutch patients. Three 2 mm cores of each tumor sample were arrayed into tissue microarrays (TMA) as previously described [5]. Immunohistochemistry and microscopic analysis TMAs [5] were immunohistochemically stained for HLA class I using the mAb antibodies HCA2 and HC10 and the rabbit anti-β2 m polyclonal Ab (A 072; DAKO Cytomation, Glostrup, Denmark). The HCA2 and HC10 antibodies were applied in immunohistochemistry as hybridoma culture supernatant, kindly provided by Prof. J.J. Neefjes from the Netherlands Cancer Institute (Amsterdam, The Netherlands). The reactivity spectrum of HCA2 includes HLA-A (except HLA-A24), HLA-B73 and HLA-C molecules as well as HLA-E, HLA-F and HLA-G antigens [27, 28, 31]. HC10 reacts with HLA-B and HLA-C molecules and HLA-A10, -A28, -A29, -A30, -A31, -A32 and -A33 heavy chains [13, 23, 31, 32]. The immunohistochemical procedures are described in detail elsewhere [18]. All tumor specimens were stained simultaneously to avoid intra-assay variation. Microscopic analysis was assessed by two independent observers (M.M. v. B. and M. v. V.) in a blinded manner. HCA2, HC10 and β2 m stainings were scored in six categories. Essentially, the scoring was divided into quartiles but for tumors with less than 25% stained cells, there was a distinction made between those with 6–25% positively stained tumor cells, those with approximately 1–5% positively stained cells and those with absolute no positively stained tumor cells [3, 11]. Where discrepancies arose between the staining of cores from the same tumor, an average of the scores was taken, with confirmation by two observers using a double-headed microscope with a consensus decision taken in all cases. Tissue stromal cells, normal epithelium or lymph follicles served as positive internal controls to ascertain the quality of the staining. Patients were excluded if stromal cells of tumor were not stained for HCA2 or HC10. Twenty-five tumors with negative staining of the stromal cells for HCA2 were excluded. HC10 showed in all tumors staining of the stromal cells. Also TNM stage 0 patients, tumors lost due to technical failure and ineligible patients were excluded, leaving 1,092 tumors in which HC10, and 1,035 in which HCA2 could be evaluated. Combining the results for HCA2 and HC10 staining resulted in 1,008 eligible stage I–IV rectal cancer patients for analyses of clinical impact of HLA class I expression. Tumors negatively stained for HCA2 and/or HC10 were stained for mismatch repair proteins MLH1 and PMS2. MLH1 and PMS2 are deficient in sporadic MSI tumors. Therefore, the expression of these proteins was used to differentiate MSI and MSS rectal cancers. Tissue stromal cells, normal epithelium or lymph follicles served as positive internal controls when analyzing MLH1, PMS2 expression. The expression of MLH1 and PMS2 was scored positive if tumor cells showed expression, and negative if tumor cells showed no expression of either MLH1 or PMS2, provided that and tissue stromal cells did show expression, indicating microsatellite stable (MSS) and microsatellite instable (MSI) tumors, respectively [7]. Statistical analyses All analyses were performed with SPSS statistical software (version 12.0 for Windows, SPSS Inc, Chicago, USA). Mann–Whitney U, t test and χ2-tests were used to compare variables. Kaplan–Meier analyses were performed to analyze patient survival. The entry date for the survival analyses was the time of surgery of the primary tumor. Events for time to local recurrence, distant recurrence, cancer-specific survival, disease-free and overall survival were defined as follows: from time of surgery to time of local disease relapse (for local recurrence), time of distant disease relapse (for distant recurrence), time of disease relapse or death by disease (for cancer specific survival), time of disease relapse or death (for disease free survival) and time of death, respectively (for overall survival). Non-irradiated and irradiated patients were first separately analyzed in univariate analysis and second, variables with a P value of <0.10 in the univariate analyses were subjected to a multivariate analysis. Multivariate analysis was performed on the whole group of irradiated and non-irradiated patients with the following variables: HLA class I, randomization for radiotherapy, TNM and circumferential margin. Cox’ regression analyses were used to calculate hazard ratios (HR) with 95% confidence intervals (CI). Results Scoring methods Several methods are described to analyze HLA class I expression in cancer. The standard is defined by the International HLA and Immunogenetics Workshop (IHIW) [3, 11]. A recent paper describing HLA class I expression in colorectal cancer used an adjusted form of this scoring method [35]. Our scoring was primarily adapted from IWIH, i.e. division into quartiles, but for tumors with less than 25% stained cells a distinction was made between those with 6–25% positive tumor cells, those with approximately 1–5% positive tumor cells and those with absolute no HLA class I positive-stained tumor cells. After scoring and analyzing this method we found that patients in the groups with absolute numbers, 1–5%, 6–25% and 26–50% HLA class I expression of tumor cells did not differ in prognosis but had a worse prognosis as compared to patients with HLA class I expression in groups with 50–75% and >75% of tumor cells expressing HLA class I. Therefore, we distinguished two categories. These two categories were (1) 0–50%; and (2) >50–100% of tumor cells expressing HLA class I. HCA-2 and HC10 staining in rectal cancer Immunohistochemical staining with HCA2 and HC10 antibodies demonstrated strong positive membrane staining of stromal cells and tumor-infiltrating inflammatory cells, indicating the success of the staining. A total of 1,035 and 1,092 tumors were evaluated with HCA2 and HC10; 324 (65%) irradiated tumors and 312 (58%) non-irradiated tumors showed at least 50% of all tumor cells positive for HCA2. Staining with HC10 resulted in 403 (76%) irradiated tumors and 436 (77%) non-irradiated tumors that showed more than 50% positive tumor cells. The complete results are shown in Table 1. Representative examples of the immunohistochemical stainings of tumors are displayed in Fig. 1a–f. These results show that about 35% of irradiated and 42% of non-irradiated patients showed less than 50% of the tumor cells expressing HCA2. HC10 is expressed in less than 50% of the tumor cells in about 25% of both irradiated and non-irradiated rectal cancer patients. Table 1Most rectal tumors have high numbers of tumor cells positive for HCA2 or HC10Irradiated patients N (%)Non-irradiated patients N (%)HCA2 High324 (65%)312 (58%) Low142 (28%)174 (32%) Absence31 (6.2%)52 (9.7%)HC10 High403 (76%)436 (77%) Low117 (22%)116 (21%) Absence8 (1.5%)12 (2.1%)Numbers (N) of patients are indicated with percentages shown in parentheses, showing: expression of HCA2 and HC10 in more than 50% of the tumor cells (high), expression in less than 50% of the tumor cells (low) and total absence (absence)Fig. 1Examples of HCA2 and HC10 immunohistochemical staining of rectal tumors; a–c HCA2, d–f HC10 expression. a, d Expression of HLA class I in >50% tumor cells; b, e expression of HLA class I in <50% tumor cells; c, f epithelial cells show total absence for HCA2 or HC10 and only stromal and infiltrative cells show positive staining for HCA2 or HC10; Original magnification ×20 Analysis of HLA class I expression in rectal tumors Together, the results obtained with HCA2 and HC10 are expected to reflect HLA class I expression in rectal cancer. In a group of 64 tumors it was studied whether an additional staining for β2 m would better define HLA class I expression. The results of the addition of β2 m to HCA2 and HC10 were comparable to those obtained with HCA2 and HC10, i.e. only 1 of 64 tumors was differently classified. Therefore, β2 m was not scored in the whole cohort and HLA class I expression was assessed by combining HCA2 and HC10. A total of 406 (85%) irradiated and 445 (84%) non-irradiated tumors exhibited expression of at least one of the two markers showing >50% positive staining of all tumor cells (further referred to as ‘the HLA class I high expression group’). A total of 70 (15%) irradiated and 87 (16%) non-irradiated tumors showed reduced numbers (≤50%) of HLA class I positive tumor cells. Only 3 (0.6%) irradiated tumors and 8 (1.5%) non-irradiated tumors showed total loss of HLA class I (negative for both HCA2 and HC10). Survival results of patients with total absence of HLA class I on tumor cells did not show significant difference from patients with reduced numbers of HLA class I positive tumor cells. Therefore, these groups were combined and will be further referred to as ‘the HLA class I low-expression group’. The complete results are shown in Table 2. The number of patients in the group of the HLA class I high-expression group and the HLA class I low-expression group was equally distributed between irradiated and non-irradiated tumors (χ2 = 0.519, P = 0.471), indicating that irradiation had no effect on HLA class I expression in these patients. Table 2Expression of HLA class I in rectal cancer using HCA2 and HC10 antibodiesHCA2HC10HLA class IHigh (N)Low + absence (N)N (%)IrradiatedHigh (N)27037406 (85%)Low + absence (N)997070 (15%)Non-irradiatedHigh (N)27732445 (84%)Low + absence (N)1368787 (16%)Expression of results of HCA2 and HC10 staining in a cross table for numbers (N) of irradiated and non-irradiated patients; expression of HCA2 and HC10 in more than 50% of the tumor cells (high) versus expression in less than 50% of the tumor cells (low) is shown. A significant correlation was noted between HCA2 and HC10 staining for both irradiated (χ2 = 53.947, P < 0.001) and non-irradiated patients (χ2 = 61.257, P < 0.001). The right side of the table displays HLA class I expression estimated on HCA2 and HC10 expression. A total of 406 (85%) irradiated and 445 (84%) non-irradiated tumors exhibited expression of at least one of the two markers showing >50% positive staining of all tumor cells. A total of 70 (15%) irradiated and 88 (16%) non-irradiated tumors showed reduced numbers (≤50%) of HLA class I positive tumor cells. The number of patients in the group of the HLA class I high expression group and the HLA class I low expression group was equally distributed between irradiated and non-irradiated tumors (χ2 = 0.519, P = 0.471) HLA class I negative cells and microsatellite instability It has been described that a majority of MSI colorectal tumors do not express HLA class I, while only a minority of MSS tumors do not express HLA class I [8, 16]. Therefore, HLA class I negative rectal tumors are the most at risk to be MSI tumors. To evaluate the numbers of sporadic MSI tumors in our study, HCA2 or HC10 negative tumors were analyzed for the expression of PMS2 and MLH1. PMS2 and MLH1 are mismatch repairs proteins that are most frequently absent in MSI sporadic tumors [6]. Of the HLA class I negative tumors only 1 out of 11 tumors did not express PMS2 and MLH1. In the tumor group negative for only 1 of the 2 HLA class I markers, 2 of 81 tumors displayed no PMS2 and MLH1 staining. These results indicate that HLA class I down-regulation is not associated with MSI in rectal cancer and are in accordance with the previous findings that only a very small minority of rectal tumors are MSI [4, 25]. HLA class I expression and clinicopathological parameters The relationship between HLA class I expression and patient/tumor characteristics was assessed (Table 3). The HLA class I expression levels were distributed equally in non-irradiated and irradiated patients with regard to most clinical and pathological parameters. Three significant differences were observed. For the non-irradiated patients, significantly more men appeared in the HLA class I low-expression group (P = 0.03). The group of irradiated tumors with HLA class I low expression contained significantly more stage III and IV tumors (P = 0.01) and also more patients with a tumor-positive circumferential resection margin (P = 0.02) when compared with tumors with high HLA class I expression. Table 3Clinicopathological characteristics of irradiated and non-irradiated patients with high or low numbers of HLA class I positive tumor cells Non-irradiated patientsIrradiated patientsHigh N = 445Low N = 87P valueHigh N = 406Low N = 70P valueGender Male (%)63750.0365660.90Age Median years65680.3265650.99TNM stage (%) I31240.5233240.01 II27303021 III36383240 IV58514Circumferential margin Negative (%)83770.2886740.02Distant from anal verge (%) ≥10 cm28330.1727320.30 5–10 cm41314636 < 5 cm31362732Operation type (%) Low anterior resection66610.7765660.89 Abdomino-perineal resection29332930 Hartmann5664Number (N) of patients with expression of total HLA class I expression in more than 50% of the tumor cells (high) and expression in less than 50% of the tumor cells (low)* Statistical significant P values are in bold Expression of HLA class I and clinical prognosis Because radiotherapy might influence local tumor recurrences [15], irradiated and non-irradiated tumors were analyzed separately in order to evaluate the impact of HLA class I expression on tumor recurrence and patient survival. The HLA class I expression was not related with distant or local recurrence rates. The patients with low expression of HLA class I had a worse overall survival and disease-free survival when compared to patients with HLA class I high expression, irrespective of treatment (Fig. 2; overall survival: P = 0.008 and P = 0.01; disease free survival: P = 0.01, P = 0.006 in irradiated and non-irradiated patients, respectively). Irradiated patients with low HLA class I expression also had a worse cancer-specific survival (P = 0.003). For non-irradiated patients, HLA class I expression had no significant effect on cancer-specific survival (Fig. 2). All results of univariate analysis are shown in Table 4. Univariate analysis showed a better outcome for overall survival and disease-free survival in patients with high HLA class I expression. Fig. 2Examples of Kaplan–Meier curves showing overall survival and cancer specific survival for irradiated and non-irradiated patients (a–d). Kaplan–Meier curves for overall survival (a, b) and cancer specific survival (c, d); curves show prognosis for non-irradiated (a, c) and irradiated patients (b, d) for HLA class I expression in more than 50% of the tumor cells (high) versus expression in less than 50% of the tumor cells (low). P value is based on univariate log rank analysesTable 4Both irradiated and non-irradiated patients with high expression of HLA class I have a better overall, and disease free survivalNon-irradiated patientsIrradiated patientsHigh (%)Low (%)P valueHigh (%)Low (%)P valueOverall survival65.558.50.01267.551.30.008Disease free survival62.253.50.01562.248.30.006Cancer specific survival74.371.40.4180.161.80.003Local recurrence8.913.70.224.73.20.72Distant recurrence26.728.70.8824.729.30.34Survival and recurrence rates indicated in percentages after 5-years of follow-up for non-irradiated and irradiated patients for HLA class I expression in more than 50% of the tumor cells (high) versus expression in less than 50% of the tumor cells (low). P value is based on univariate log rank analyses for overall survival, disease free survival, cancer specific survival, local recurrence and distant recurrence* Statistical significant P values are in bold Multivariate analysis Multivariate analysis was performed to identify factors with independent prognostic significance and to calculate hazard ratios (HR). Analyses included TNM, circumferential margin, randomization for preoperative radiotherapy and HLA class I expression (low vs. high HLA class I positive tumor cells) (Table 5). Advanced pathological (TNM) stage and tumor-positive circumferential resection margins retained their strength as independent prognostic factors in these survival analyses. HLA class I expression showed independent prognostic value for overall survival and disease-free survival (HR: 1.3, P = 0.042 and HR: 1.4, P = 0.006, respectively), but not for cancer-specific survival. Table 5Multivariate analysis confirms independent better overall, and disease free survival for rectal cancer patients with high expression of HLA class IOverall survivalDisease free survivalCancer specific survivalHR (95% CI) P valueHR (95% CI) P valueHR (95% CI) P valueHLA High10.04210.00610.653 Low1.3 (1.0–1.6)1.4 (1.1–1.8)1.1 (0.8–1.5)Randomization TME10.63210.21410.282 TME + RT1 (0.8–1.2)0.9 (0.7–1.1)1.1 (0.9–1.5)TNM I111 II2.2 (1.7–3.0)<0.0012.1 (1.6–2.8)<0.0013.5 (2.0–6.1)<0.001 III3.1 (2.4–4.1)<0.0013.1 (2.3–4.0)<0.0019.0 (5.4–14.9)<0.001 IV11.8 (8.1–17.1)<0.001––50.3 (28.5–89.1)<0.001CRM Negative1<0.0011<0.0011<0.001 Positive1.3 (1.1–1.5)1.8 (1.4–2.2)1.3 (1.1–1.5)Multivariate analysis for cancer specific, overall and disease free survival was performed to identify factors with independent prognostic significance and to calculate hazard ratios (HR) with 95% confidence intervals (CI) shown in parentheses. HLA class I expression in more than 50% of the tumor cells (high) versus expression in less than 50% of the tumor cells (low), total mesorectal excision (TME), Radiotherapy (RT); circumferential margin (CRM); P value is based on Cox’ regression analyses* Statistical significant P values and HR are in bold Discussion We showed that rectal cancer patients from the HLA class I low expression group had an independent worse overall and disease-free survival when compared to patients from the HLA class I high-expression group. These data imply that the expression of HLA class I in tumor cells predicts survival for rectal cancer patients. Although significant better cancer-specific survival for irradiated patients with high HLA class I was found in univariate analysis, the predictive value was lost in multivariate analysis. This observation can be explained by the fact that the group with low HLA class I significantly included more stage III/IV and more patients with a positive circumferential margin as compared to the group of patients with high expression of HLA class I. Also no predictive value of HLA class I expression was found with regard to recurrence-free survival of these patients. Therefore, we have no indications that support the notion that better survival of high HLA class I expression is due to the better antigen presenting function of these tumor cells, as has been suggested [18, 35]. In our study, no difference was found between irradiated and non-irradiated patients for HLA class I expression in tumor cells. It has been described that γ-irradiation induces enhanced peptide production and surface expression of MHC class I in a colorectal mouse tumor cell line [26]. The fact that we could not find more HLA class I expression in irradiated tumors than in non-irradiated tumors indicates that radiotherapy does not induce HLA class I expression in vivo. Immunohistochemistry, however, is less suitable to measure subtle expression changes. Therefore, additional research is required to determine the impact of radiotherapy on expression levels of HLA class I in human tumors. In our study, more tumors showed HLA class I down-regulation after immunohistochemical staining using HCA2 than using HC10. This difference might be due to differences in reactivity spectrum of both antibodies (see Materials and methods) or to the fact that HLA alleles are differently affected in colorectal cancer. If the latter is the case, our results suggest that HLA A alleles preferentially show down-regulation in rectal cancer. Previous reports evaluated HLA class I expression in mixed patient populations of colon and rectal cancer patients [18, 35]. Watson et al. also found in a large group of colorectal cancer patients that patients with low expression of HLA class I had a poor prognosis [35]. However, in contrast to our results, both studies described a substantial population of patients with tumors showing absence of HLA class I. In addition, they described that absence of HLA class I was associated with better prognosis as compared to tumors expressing reduced numbers of HLA class I positive tumor cells. A relatively low number (1.1%) of HLA class I negative tumors was observed in our cohort of rectal cancer patients only. These patients showed no survival advantage when compared to patients with reduced numbers of HLA class I positive tumor cells. There are several explanations for the discrepancy in the number of HLA class I negative tumors between the study of Watson et al. and ours, like different definition of HLA class I expression, differences in staining techniques, different patient cohort and number of MSI tumors. We showed that tumors that do not stain HC10 can stain positive for HCA2 and thus are still able to present antigens. Therefore, an explanation for the differences with the results of Watson et al. is that we used strict criteria to classify tumors as HLA class I absent (defined as both HCA2 and HC10 negative) as compared to Watson et al. (defined as negative for HC10 or negative for β2 m instead of negative for both). Another important explanation is that we examined HLA class I expression in a relative more homogeneous population of patients with a rectal tumor, while the other cohorts are more heterogeneous, consisting of both colon and rectal cancer. Although combining results from colon and rectum is generally accepted when predicting prognosis, this might influence results [14]. In colon cancer patients, approximately 50% of all proximal colon tumors show MSI, whereas almost all distal colon and rectal cancers are MSS tumors [24, 30]. Loss of HLA class I has been described in at least 60% of all sporadic right-sided MSI colorectal tumors but in only 17% of MSS right-sided colon tumors loss of HLA class I is found [8, 16]. In our cohort, only 1 out of 11 HLA negative tumors and 2 out of 81 tumors negative for HCA2 or HC10 did not express MLH1 and PMS2 and were thus likely MSI tumors. This indicates that rectal cancers are mainly MSS tumors, as has previously been described [4, 14, 25]. Of the multiple mechanisms that have been shown to underlie defects in HLA class I expression in colorectal cancer (mutations in the individual HLA class I genes, mutations in β2 m [16], and defects in components of the HLA class I-associated antigen processing machinery (APM) [2, 16]), only the first will result in allele-specific aberrancies while the other affect total HLA class I expression and may result in total absence in a tumor cell. These observations imply that a population of colorectal tumors with total absence of HLA class I probably contains a disproportionate large number of MSI tumors when compared to colorectal tumors expressing HLA class I. In addition, MSI colorectal tumors have a better prognosis when compared to MSS colorectal tumors [12, 24]. Therefore, HLA class I negative tumors are more likely to be MSI tumors with a different clinical behavior as compared to MSS colorectal tumors. It is likely that MSI influences prognostic results when considering HLA class I expression in colorectal tumors. Our results show that HLA class I expression in rectal cancer affects the patient’s prognosis. We hypothesize that both oncogenic pathway and HLA class I expression dictate clinical tumor progression. We suggest that in future prognostic studies, analyzing expression of HLA class I or other biomarkers in colorectal cancer, the impact of MSI should be considered.

Invest_New_Drugs-3-1-1915607

Human mass balance study of the novel anticancer agent ixabepilone using accelerator mass spectrometry

Summary Introduction The taxanes paclitaxel and docetaxel are important anti-cancer drugs [1]. However, their use is complicated by low oral bioavailability and the development of resistance due to MDR1 over-expression or β-tubulin mutation [2, 3]. A search for microtubule stabilizing agents with improved characteristics compared to taxanes led to the discovery of epothilone A and B in extracts of the myxobacterium Sorangium cellulosum [3], which were subsequently shown to bind to tubulin at the same binding site as paclitaxel [4]. In vitro, epothilone B was more active than in vivo (mouse models) [5], and in vivo hydrolysis of the lactone ring was suspected of producing inactive metabolites. To improve the metabolic stability of the naturally occurring epothilones, these were modified by replacing the ester bond with an amide bond [6,7]. In general, lactam analogues displayed tubulin-polymerizing and cytotoxic potencies inferior to their naturally occurring counterparts. The exception to this rule was the lactam analogue of epothilone B, ixabepilone (BMS-247550, NSC 71028, EpoB-lactam, aza-EpoB, 15-desoxy-15-aza-epothilone B), see Fig. 1 for structure [6]. Ixabepilone has undergone a number of phase II clinical trials with varying schedules at doses from 6 to 40 mg/m2 administered as an intravenous infusion over 1 to 3 h [8–11]. Response rates in metastatic breast (12–44%), non-small cell lung (14%), hormone-refractory metastatic prostate cancer (33%), and metastatic renal cell carcinoma (14%) were modest to promising and warrant further investigation [8]. The pharmacokinetics of ixabepilone have been described [9,12,13] and the contribution of two known chemical degradation products was shown to be negligible [14], however, other than that it is a CYP3A4/5 substrate [15], little is known about the excretion and metabolic fate of ixabepilone. Fig. 1Chemical structures of epothilone B and its lactam analogue ixabepilone The present investigation was aimed at determining the pharmacokinetics and excretory pathways of [14C]ixabepilone. To this end, we initiated a mass balance study of ixabepilone in humans. A mass balance is an elaborate pharmacokinetic investigation employing a radioactive tracer. It investigates the plasma pharmacokinetics and excretion of both the unchanged drug and the total radioactivity (drug and metabolites), and allows elucidation of the metabolic fate of a drug. The main objective is the maximum recovery of the radioactive dose in urine and faeces [16]. Due to autoradiolysis, [14C]ixabepilone at a typical specific activity of 1.5 μCi/mg was unstable. However, [14C]ixabepilone at a specific activity of less than 10 nCi/mg was found to be stable. Therefore a study was conducted with a low radioactive dose of 80 nCi of 14C-ixabepilone, as opposed to a more typical human radio-tracer dose of circa 100 μCi dosed in study. For detection of these very low levels of radioactivity we used an ultra-sensitive detection method, Accelerator Mass Spectrometry (AMS) [17–19], which measures the amount of 14C as if it were a stable isotope label instead of measuring its decay. AMS separates elemental isotopes through differences in mass, charge, and energy, resulting in a [14C]/[12C] isotope ratio. Subtracting the natural [14C] background level and taking into account the total carbon content of the sample results in a total [14C] amount that is converted to a conventional unit for radioactivity (DPM, Bq, or Ci). Given the very low dose of radioactivity administered (80 nCi compared to 100 μCi in a typical radio-tracer study), administration would result in negligible radiation exposure. Materials and methods Study design This was an open-label, single-dose study in which eight patients with histologically confirmed solid tumours received an intravenous fixed dose of 70 mg, 80 nCi [14C]ixabepilone over 3 h. The protocol and informed consent form were reviewed and approved by the Institutional Review Board of the Antoni van Leeuwenhoek Hospital/The Netherlands Cancer Institute, and each patient gave written informed consent before receiving the study medication. Eligibility criteria most relevant to the mass balance study were: availability for in-house admittance during the first 8 days, adequate hepatic and renal function, a performance status (PS) of 0–2, and a life expectancy of at least 12 weeks. Patients Eight patients (three male and five female) with a median age of 54.5 years and a WHO performance status of 0–2 were enrolled in the study. All patients had advanced cancer for which no standard anti-cancer therapy was available. The primary tumours were: squamous cell carcinoma, colon cancer (n = 2), gastric carcinoma, ovarian cancer, non-small cell lung cancer, sigmoid adenocarcinoma, and pancreas carcinoma. Study medication Individual vials of radiolabeled ixabepilone contained 20 mg, 22.8 nCi [14C]ixabepilone (specific activity 1.14 nCi/mg). The [14C]ixabepilone was synthesized by Bristol-Myers Squibb by growing the myxobacterium Sorangium cellulosum in 14C containing medium. This resulted in labeling throughout the ixabepilone molecule. Radiochemical and chemical purity was confirmed at >99%. Vials were reconstituted using 10 ml Cremophor®EL:ethanol (1:1, v/v) each. A 35 ml volume of the reconstituted [14C]ixabepilone was added to 140 ml Lactated Ringer’s Infusion (LRI). After mixing, a 2 ml sample was taken for quantitation of radioactivity by liquid scintillation counting (LSC). LSC radioactivity values were recalculated to amounts of [14C]ixabepilone for use in calculating recovery from feces and urine. After the 3-h intravenous administration, infusion lines were flushed with 500 ml LRI. Gravimetric measurements of the infusion flask prior to, and after administration enabled calculation of the amount of radioactivity actually administered. To prevent Cremophor®EL-related hypersensitivity reactions, patients were pre-medicated with oral H1 and H2 blockers. Sample collection Blood was collected at 0 (pre-dose), 1.5, 3 (prior to end of infusion), 3.25, 3.5, 3.75, 4, 5, 6, 8, 12, 24, 48, 72, 120, and 168 h after start of infusion. An indwelling catheter contra-lateral to the administration site was used for serial blood sampling. Separate, but simultaneous, blood samples were taken for the quantitation of total radioactivity and unchanged ixabepilone, respectively. Plasma was obtained by immediate centrifugation of the blood (10 min, 1000 × g, 4°C). The plasma layer was aspirated and stored below −20°C until analysis. We collected complete urinary and faecal output up to 7 days after administration. Urine was collected over 24-h intervals in refrigerated collection jugs. At the end of each collection interval, the respective urine samples were mixed and total volume was recorded. Aliquots were stored below −20°C until analysis. Faecal samples were collected and stored at −20°C per portion and combined per 24-h interval. Faeces were homogenized after addition of water (2:1, w/w). Aliquots were stored below −20°C until analysis. AMS analysis of total radioactivity (TRA) in plasma, urine and faeces TRA in plasma, urine and faeces was determined using AMS (Exceleron, York, UK) [20, 21]. Briefly, sample pre-treatment consisted of the conversion of the carbon within the samples to graphite via a two-step process of oxidation and reduction [22]. Oxidation (combustion) was performed by heating the dried sample with copper oxide under vacuum at 900°C for 2 h in a heat-sealed tube. Reduction (graphitisation) was performed by heating the carbon dioxide formed from the oxidised sample with zinc, titanium hydride and cobalt at 500°C for 4 h, followed by a further 6 h at 550°C. After the graphitisation process, the graphite was pressed into a cathode. In the AMS-procedure, this cathode was bombarded with high-energy caesium (Cs+) ions. The resulting negative carbon ion beam was introduced into the AMS instrument, accelerated by an electric field of 4.5 MV, and the carbon atoms were stripped of their valency electrons. This resulted in separation of the carbon isotopical elements as C4+ ions. The carbon isotopes 12C and 13C were quantitated using Faraday cups, while 14C was quantitated by a gas ionization detector. Because AMS provides an isotope ratio that is [14C]/[12C] and not an absolute value, it is necessary to know the carbon content of the sample plus any added carrier. This was determined using a carbon, hydrogen, nitrogen analyser (Elemental Microanalysis Ltd., Okehampton, UK). The natural background [14C]/[12C] level per matrix was subtracted to obtain the amount of ixabepilone-related 14C. The total carbon content and the specific activity of [14C]ixabepilone was used to convert the [14C]/[12C] ratio to ixabepilone equivalents (ng/ml sample or % of dose). Using this sophisticated technique, it was possible to quantitate <4 DPM/ml plasma, 0.01 DPM/ml urine and 0.1 DPM/g faeces (1 DPM corresponds to 395 ng [14C]ixabepilone). Analysis of ixabepilone in plasma and urine Concentrations of ixabepilone parent drug in plasma and urine were determined with a validated liquid chromatography assay equipped with tandem mass spectrometric detection (Bristol-Myers Squibb, Princeton, NJ). Briefly, after addition of internal standard (BMS-212188) to 0.2 ml of each sample, calibration standard and quality control sample, the samples were precipitated with acetone. The supernatant was further extracted with 1-chlorobutane. The organic layer was removed and evaporated to dryness. The residue was reconstituted and injected into the LC-MS/MS system. Chromatographic separation was achieved, isocratically, on a YMC ODS-AQ column (4.6 × 50 mm internal diameter) at a flow rate of 0.3 ml/min with detection by electrospray tandem mass spectrometry. The mobile phase contained acetonitrile—0.01 M ammonium acetate (pH 5.0) (65:35, v/v). The standard curve, which ranged from 2 to 500 ng/ml, was constructed with a 1/x weighted quadratic regression model. The within-run precision for ixabepilone in plasma and urine was within 15 and 9%, respectively. The between-run precision for ixabepilone in plasma and urine was within 13 and 9%, respectively. The accuracy was within 11 and 6% of the nominal values in plasma and urine, respectively. Pharmacokinetic analyses The pharmacokinetic analyses of the ixabepilone plasma concentration versus time data were performed by non-compartmental methods using Kinetica Version 4.2 (InnaPhase Corporation, Philadelphia, PA). The peak plasma concentration (Cmax) and the time to reach maximal concentration (Tmax) represent observed values. The area under the plasma concentration versus time curve (AUC) was calculated using a combination of linear and log trapezoidal summations. The AUC was extrapolated to infinity (AUCinf) by dividing the last measurable concentration by the terminal rate constant, k. The absolute value of k was used to estimate the apparent terminal half-life, t1/2 = ln 2/ k. The total body clearance (Cl) was determined by dividing dose by AUCinf and the volume of distribution at steady-state (Vss) was calculated as dose × area under the moment curve (AUMCinf)/(AUCinf)2 with appropriate corrections for infusion duration. Renal clearance was calculated by dividing amount excreted by the AUC. Similarly, pseudopharmacokinetic parameters were calculated for TRA. Outliers were statistically assessed using Dixon’s Q test. Results To evaluate the pharmacokinetics and excretory pathways of [14C]ixabepilone, we performed a mass balance study in eight patients. Figure 2 shows the mean plasma concentrations over time of both TRA and ixabepilone as determined by AMS and LC-MS/MS, respectively. Fig. 2Mean (±SD) plasma concentration-time curves of [14C]ixabepilone derived radioactivity (filled square) and unchanged [14C]ixabepilone (open square) for eight patients Both TRA and ixabepilone in plasma exhibited an initial distribution phase of approximately 10 h followed by a much slower terminal elimination phase with half-lives of 73.1 and 50.3 h, respectively. The time to maximum plasma concentration was typically around the end of infusion. Plasma ixabepilone concentrations were quantifiable through 168 h in all patients and plasma TRA concentrations were quantifiable through 168 h in all patients except in patient 1 in whom plasma TRA concentrations were only quantifiable for 48 h. Pharmacokinetic parameters are summarized in Table 1. The ratio of the ixabepilone AUC to the total radioactivity AUC, or AUC-ratio, is a measure of exposure to [14C]ixabepilone relative to total [14C]ixabepilone related compounds. The AUC ratio of patient 1 was larger than unity. This was caused by the underestimated TRA AUC, which was based on measurements until only 48 h, after which plasma concentrations fell below the lower limit of quantitation. Besides the TRA AUC, the t1/2 of patient 1 was also inaccurately estimated at 8.3 h. Therefore, patient 1 was omitted from calculation of mean plasma TRA parameters. Figure 3 shows the plasma concentrations of ixabepilone relative to TRA for patients 1–7 and patient 8. Patient 8 experienced a blocked bile duct (reflected by a maximum plasma bilirubin concentration of 47 μmol/l (3 × upper limit of normal) and alkaline phosphatase of 500 U/l (4 × upper limit of normal) at day 7) resulting in a higher plasma TRA, Cmax and AUC. The higher TRA plasma concentrations in patient 8 resulted in an AUC ratio of plasma ixabepilone to TRA that was lower than that of the other patients. In contrast, unchanged ixabepilone plasma pharmacokinetics were not affected in this patient. Table 1Plasma pharmacokinetic parameters of ixabepilone and total radioactivity in cancer patients after a 3 h intravenous infusion of 70 mg, 80 nCi [14C]ixabepiloneNrIxabepiloneRadioactivity (ixabepilone equiv.)AUClast ratioaCmax (ng/ml)AUClast (μg*h/ml)AUCinf (μg*h/ml)t1/2(h)Cl (l/h)Vss (l)Cmax (ng/ml)AUClast (μg*h/ml)AUCinf (μg*h/ml)t1/2 (h)12773.063.3551.720.912753221.891.968.31.62b22872.752.9753.023.6117449711.011.848.70.2534332.832.9951.723.492968110.911.336.50.2644863.824.3267.416.2102794020.121.646.90.1952662.422.5223.527.875650320.044.3185.20.1262732.172.2936.730.591257311.916.399.80.1872012.252.4971.528.115204498.539.5961.20.2681871.932.2247.231.61635150232.534.434.40.06Mean3012.652.8950.325.3115468316.4b21.3b73.1b0.19bSD1050.600.6915.45.23083788.44b13.2b54.0b0.08baAUClast ratio is calculated by dividing AUClast ixabepilone by AUClast TRA and expresses the relative contribution of ixabepilone to TRA exposure in plasma.bPatient 1 was not included in the calculation because the radioactivity AUC and t1/2 could not be determined accurately (see text). This also resulted in the deviating AUClast ratio (outlier).Fig. 3Mean (+SD) [14C]ixabepilone to total radioactivity ratio in plasma over time for patients 1–7 (filled square) and patient 8 (open square) Figure 4 displays the mean excretion of TRA in urine and faeces and the total, in time, and Fig. 5 displays the mean cumulative urinary excretion of [14C]ixabepilone derived radioactivity and unchanged [14C]ixabepilone. Table 2 shows individual 168 h excretions of [14C]ixabepilone and TRA in urine and faeces. The U/F ratio is the ratio of urinary recovery relative to faecal recovery. Patient 3 displayed a rather low recovery of radioactivity from both urine (14.8%) and faeces (38.3%). Plasma pharmacokinetics were not different compared to the other patients. A low total recovery of 37.8% was also obtained for patient 4, attributable to a very low faecal recovery (12.7%), while urinary recovery was normal (25.1%). This patient only defecated at 24 h and at 168 h (a few grams), which almost certainly explains the observed low recovery. Although the plasma pharmacokinetics and urinary excretion did not deviate extremely from the other patients, patient 4 displayed a somewhat elevated plasma exposure to both radioactivity and ixabepilone, and the lowest clearance for unchanged ixabepilone. If we now look at the U/F ratios (a measure of urinary versus faecal elimination), patient 4, who had the low fecal output, is an outlier (Q = 0.86, critical value p = 0.05: 0.57) compared to patients 1, 2, 3, 5, 6 and 7. Patient 8, who experienced the blocked bile duct stent, is also an outlier (Q = 0.91, critical value p = 0.05: 0.52) compared to the same patients. The urinary recovery of patient 8 was highest at 47.7% causing the large standard deviation of urinary recovery. However, the urinary recovery of unchanged ixabepilone of 4.0% was within the normal range. The faecal recovery of this patient was the lowest (excluding patient 4) at 15.6%. Apparently, there was no accumulation of the active drug despite a decreased biliary excretion. Fig. 4Mean urinary (open square, −SD), faecal (open circle, +SD) and total (filled square, +SD) cumulative excretion (168 h) of [14C]ixabepilone derived radioactivity for eight patientsFig. 5Mean (±SD) cumulative urinary excretion of [14C]ixabepilone derived radioactivity (filled square) and unchanged [14C]ixabepilone (open square) for eight patientsTable 2Urinary and faecal recovery (0–168 h) of ixabepilone and total radioactivity in cancer patients after a 3 h intravenous infusion of 70 mg, 80 nCi [14C]ixabepiloneNraRecovery (% of dose)Radioactivity (U/F ratioc)Ixabepilone (Clrenal (l/h))IxabepiloneTotal radioactivityUrineUrineFaecesTotal15.0714.5659.6974.250.241.1624.6623.2874.7898.060.311.1937.7614.7738.2953.060.391.9243.3225.0812.7137.791.970.6155.4619.6665.0284.680.301.5866.1230.7864.5995.370.481.9876.0325.0586.72111.770.291.8784.0247.6515.6263.273.051.46Mean5.3025.1052.1877.280.33b1.47SD1.3810.6327.1724.940.08b0.47aPatients 1, 2, and 8 were male.bPatient 4 was not included because of the extremely low production of faeces and patient 8 was not included in the calculation because his bile duct obstruction (relevance illustrated by increasing plasma bilirubin concentrations) may have reduced faecal excretion.cThe U/F ratio is the ratio of urinary recovery relative to faecal recovery. On average, more than 77% of the drug was excreted over the 7-day period with the majority of 52% being excreted in faeces and 25% in urine. Excretion in faeces occurred at a relatively constant rate over the entire collection period, with only 10 and 12% of dose recovered in the 0–24 and 24–48 h collection intervals, respectively. In contrast, urinary excretion displayed two distinct phases with 65% of total urinary radioactivity (corresponding to 17% of dose) and 76% of total urinary ixabepilone being excreted in the first 24 h. From 24 to 168 h, urinary excretion occurred at a relatively slow constant rate. This may reflect the biphasic plasma curve, with high concentrations in the distribution phase and lower concentrations in the terminal elimination phase. At first urinary excretion predominates and after 24 h, faecal excretion becomes the dominant route of excretion. Discussion The present investigation was aimed at determining the pharmacokinetics and excretory pathways of [14C]ixabepilone by performing a mass balance study. In conventional human mass balance studies, approximately 100 μCi of [14C] is typically used to trace the fate of a drug [16]. Due to autoradiolysis of ixabepilone, this level of labeling was not possible. The level of radioactive labeling we used was more than 1,000-fold lower at 80 nCi. The most commonly used method for detection of radiotracers in mass balance studies is liquid scintillation counting (LSC). This method however, relies on the detection of β-emissions and therefore can only exploit a very small proportion of the isotope. 14C has a half-life of 5740 years and therefore only about 0.012% of the 14C atoms in a sample decay over the course of a year. AMS however, is a nuclear physics technique that detects the individual 14C atoms, as if they were stable isotopes, and is therefore much more sensitive than LSC [19]. Originally, AMS technology was developed in the 1970s for radiocarbon dating, but has recently found novel uses in biomedical research. It has been used to study DNA adducts of carcinogens, and it can replace decay counting as a method of analysis in biological samples [17, 23–25]. A first human mass balance using this ultra-sensitive technique has already been reported and is expected to be followed by many others [26]. The ixabepilone plasma pharmacokinetic parameters observed in the present investigation are similar to the values previously reported for a standard phase I dose escalation study in which 11 patients were administered 40 mg/m2 of ixabepilone in an expansion of the recommended phase II dose [9]. However, in our study, the coefficient of variation for AUCinf and Cl of ixabepilone was 24 and 21% as compared to 46 and 38% in the phase I dose escalation study. The lower inter-patient variability may result from the greater rigor toward sample collection and processing that is possible in a mass balance study during which the patient is hospitalized. It should also be noted that the administration of ixabepilone as a flat 70 mg dose in the present study did not result in any increase in inter-patient variability compared with the dose, based on body surface area, which was administered during the phase I study. Our results show that ixabepilone is extensively metabolised. This is indicated by the plasma AUC ratio of 0.19 and confirmed by the recovery of ixabepilone relative to TRA in urine of approximately 0.21. Patient 8 experienced a blocked bile duct resulting in a higher plasma TRA, Cmax and AUC, but interestingly, unchanged ixabepilone plasma pharmacokinetics were not affected and excretion of ixabepilone metabolites was re-routed to urine. The calculated half-life of TRA was longer than that of ixabepilone (t1/2 of 73.1 and 50.3 h respectively), however, we do not conclude that TRA was eliminated more slowly than ixabepilone. Comparison of ixabepilone and TRA half-lives per patient revealed that in 5 cases (patient 2, 3, 4, 7, and 8) the TRA half-life was shorter than the ixabepilone half-life. The mean TRA half-life was disproportionally influenced by the long half-life observed in patient 5 (185 h). The very long half-life of TRA in this patient is not likely to be clinically relevant because this patient excreted 85% of the dose within 7 days. Based on the half-lives for plasma radioactivity observed in the other patients, in the once every 3 weeks administration schedule, accumulation of metabolites should not occur. The seemingly long ixabepilone half-life of over 50 h is not unusual when compared to other tubulin-interacting drugs. Half-lives of e.g. vinca-alkaloids range from 12–42 h for vindesine and up to a half-life of several days for vincristine. With an average recovery of 77% of the drug excreted over the 7-day period, the recovery of radioactivity of this mass balance study is reasonably good [16]. The incomplete recovery is explained by the still ongoing excretion at the end of the 7 day collection period, and the long plasma half-life of TRA. Immediate assessment of the still appreciable amounts of radioactivity excreted on day 7 to possible extend the period of collecting excreta, as suggested in the literature [16], was not possible. The shipping and lag-time in quantitating radioactivity by AMS precluded this. Another contributing route of elimination not monitored in the current study is loss of radioactivity through expiration of 14C O2 [16]. At 52% of the dose versus 25%, faecal excretion is a quantitatively more important route of elimination than urinary excretion. Biliary excretion is governed by a molecular weight limit. In humans, polar compounds with a molecular weight above approximately 500 g/mol are predominantly excreted via bile [27, 28]. Ixabepilone metabolites will probably be more polar (phase I metabolism, oxidation, hydrolysis etc.) and have higher masses (conjugates or phase II metabolites) than ixabepilone (Mwt 506.7 g/mol, and rather hydrophobic). Whether this is the case for ixabepilone remains to be investigated by metabolic profiling of the excreta. Given the normal ixabepilone plasma concentrations of patient 8 while experiencing a blocked bile duct, biliary excretion of unchanged ixabepilone into faeces may not be a very important elimination route, and the results suggest that ixabepilone was equally well metabolised but that the biliary excretion of these metabolites was hindered by the blocked bile duct stent, resulting in a shift to urinary excretion of total radioactivity. The importance of metabolism in the elimination of ixabepilone has been confirmed by this study as apparent from the low contribution of unchanged ixabepilone to TRA in plasma and urine. Ixabepilone related radioactivity is predominantly excreted in the faeces. Future investigations must be aimed at elucidating the metabolic fate of ixabepilone, and determining the activity of the metabolites. Subsequent identification of drug metabolizing enzymes involved in ixabepilone metabolism may result in explaining pharmacokinetic variability of ixabepilone in individual patients.

Virchows_Arch-3-1-2039779

Increased epithelial cell proliferation in the ileal pouch mucosa of patients with familial adenomatous polyposis

To eliminate the risk of colorectal cancer in patients with familial adenomatous polyposis (FAP), reconstructive proctocolectomy is performed. Although most colonic mucosa is resected during the ileal pouch anal anastomosis, adenomas and carcinomas may develop in the pouch. This may be caused by altered cell kinetics due to intraluminal changes in the pouch. In 32 patients with FAP, biopsy specimens from the mucosa of the pouch and also of the afferent ileal loop were taken. Tissue sections were immunohistochemically processed with the monoclonal antibodies M30 and MIB-1 to assess apoptotic and proliferative indices, respectively. Cell proliferation was also assessed by a modified sign test. There were no significant differences in apoptotic rates between the mucosa of the pouch and the mucosa of the afferent ileal loop. However, cell proliferation was significantly higher in the mucosa of the pouch vs afferent ileal loop, both by using the quantitative (68.3% vs 61.6%, p = 0.001) and semiquantitative methods (p < 0.05). Our newly developed semiquantitative approach outperformed previously described methods. The higher cell proliferation in the pouch as compared to the afferent ileal loop may contribute to the increased risk for adenomas and carcinomas in the pouch of patients with FAP and emphasizes the need for regular endoscopic surveillance. Introduction Familial adenomatous polyposis (FAP) is an autosomal-dominant disease with an estimated prevalence of 1 in 5,000 to 1 in 7,500 [5]. It is characterized by the development of hundreds of adenomas in the large intestine. Without surgical intervention, virtually all patients will develop colorectal cancer at relatively young age. To eliminate the risk of colorectal cancer, a restorative proctocolectomy with ileal pouch anal anastomosis (IPAA) is accepted as one of the surgical treatments of choice in these patients. In the IPAA procedure, almost all colonic mucosa is resected, thus reducing cancer risk significantly. However in patients with FAP, adenomas are also present in the upper gastrointestinal tract; approximately 5% of the patients having gastric, 60 to 90% duodenal [4], 50% jejunal [14], and 9 to 25% ileal adenomas [14, 23, 34]. Occurrence of ileal adenomas is of special interest because the terminal ileum, which is used to construct the pouch, may already contain adenomas. There is accumulating evidence that adenomas develop in the pouch. The reported incidence after a follow-up of at least 5 years is 8 to 60%, increasing up to 75% in certain subgroups [14, 26, 33, 38], which is much higher as compared to the reported incidence of 9–25% in the preoperative ileum of patients with FAP [14, 23, 34]. Therefore, it seems that in the ileal pouch of patients with FAP, the development of adenomas is accelerated. At present, the malignant potential of the pouch adenomas is unclear. However, 13 patients with a carcinoma in the pouch have been recently described in the literature [2, 3, 7, 9, 13, 24, 25, 29, 35–37]. Stasis in the pouch causes a change in the luminal content, which is in close contact with the ileal mucosa. There is an increase in the concentration of short-chain fatty acids to colonic levels [10], an increase in anaerobic bacterial counts resulting in a more colonic type of flora, and an increased deconjugation and dehydroxylation of bile acids by anaerobic bacteria [20, 21]. The intestinal flora could play a role in initiation and promotion of colon cancer by activation of various classes of carcinogens, such as nitrosamines, polycyclic aromatic hydrocarbons, and glycosides [17]. A direct link between intestinal bacterial flora and tumor promotion in the APCMin/+ mouse model for FAP has already been established [22]. Therefore, we hypothesize that the intraluminal ecological changes in the pouch are responsible for enhanced tumor formation in the mucosa of the pouch. Alteration of cell kinetics in the mucosa, including a lower rate of apoptosis and a higher rate of cell proliferation, is associated with tumor initiation and promotion. In colorectal cancer, the balance between apoptosis, cell proliferation, and cell shedding is disturbed, and such changes may favor the development of adenomas [30]. Up to now, no data are available on changes in proliferation and apoptosis rates of the ileal mucosa of the pouch after IPAA in patients with FAP. The aim of this study is to investigate changes in apoptosis and cell proliferation rates, occurring in the mucosa of the pouch of patients with FAP, in comparison with the ileum of the afferent loop. The results may contribute to a better understanding of the enhanced adenoma formation in the pouch compared to normal ileum. Materials and methods Patients and tissues The study was approved by the regional medical ethical commission, and informed consent was obtained from all patients. Patients with FAP and an IPAA, who were under surveillance in the Radboud University Nijmegen Medical Centre or regional affiliated hospitals, were invited to participate in this study. Thirty-two patients with FAP were included. The diagnosis FAP was based on either a clinical presentation of at least 100 colonic adenomas or a mutation in the APC gene. Data concerning the surgical procedures were obtained from medical records. From each patient, mucosal biopsy specimens of both the pouch and the afferent ileal loop were obtained during a regular surveillance endoscopy, in the period January 2002 until April 2004. Patients were fasted overnight. On the day of examination, patients were encouraged to drink liberally. No laxatives or cathartic enemas were given. To clear the pouch of fecal ruminants, two 250-ml water enemas were given before the endoscopy. The endoscopy was performed with an Olympus GIF-1T140 video endoscope. From January 2002 until August 2003, a 2.8-mm diameter biopsy forceps (FB 13K-1 Olympus, Tokyo, Japan) was used, and from September 2003 until April 2004, a 3.0-mm diameter biopsy forceps (B102-C1-30.160 MedWork/Treier Endoscopie GA, Beromünster, Switzerland) was used. The afferent loop was introduced up to 20 cm proximal of the pouch. The mucosa was sprayed with 1% indigo carmine dye (Laboratoires SERB, Paris) at 1:1 dilution with water, where after photographs were taken to evaluate number and size of adenomas present. For pathological examination, at least four biopsies were taken at random locations from the afferent ileal loop (10 to 20 cm proximal from the pouch), four biopsies from the pouch mucosa (at least 5 cm proximal from the anal verge), and four biopsies from adenomas if present. The biopsies from adenomas were used only for pathological evaluation, i.e., to exclude serious dysplasia in pouch adenomas, and were not used for research purposes. The biopsies were stretched on filter paper to maintain correct orientation of crypts, fixated in formalin, and embedded in paraffin. Immunohistochemistry Cell proliferation and apoptosis were measured in the stretched formalin-fixed nonadenomatous tissue. The crypt cell proliferation activity was assessed after staining with the monoclonal antibody MIB-1 (Dako A/S, Glostrup, Denmark), which recognizes the Ki-67 nuclear antigen of dividing cells in formalin-fixed paraffin-embedded tissue [8]. Apoptosis was assessed by staining with the monoclonal antibody M30 (Roche Diagnostics, Mannheim, Germany), which recognizes cleaved cytokeratin 18. Immunoreactivity of M30 is confined to the cytoplasm of apoptotic epithelial cells and is expressed during early apoptosis [18]. Tissue sections of 4-μm thickness were cut from paraffin blocks, mounted on electrostatic slides (Super Frost Plus, Menzel-Gläser, Germany), and dried overnight, followed by drying in a stove at 50°C for 15 min. Sections were put in xylol for 10 min and taken from xylol through 100% alcohol to water. After deparaffinization, endogenous peroxidase was blocked by treatment with 3% hydrogen peroxide in phosphate-buffered saline (PBS) for 30 min. Pretreatment was performed by heating the tissue sections in citrate buffer (10 mmol/l, pH 6.0) at 180-W power in a microwave oven for 10 min. After cooling at room temperature for 1.5 h, sections were rinsed with PBS. Then, 20% normal horse serum (Vector Laboratories, Burlingame, CA, USA) was applied for 10 min. The sections were then incubated overnight at 4°C with either the mouse monoclonal antibodies MIB-1 at 1:1,000 dilution or M30 at 1:100 dilution. Thereafter, the sections were rinsed in PBS for 10 min and incubated with biotinylated horse anti-mouse IgG (Vector) for 30 min. After rinsing for 10 min in PBS, incubation with ‘avidin and biotinylated horseradish peroxidase macromolecular complex’ (Vector) was performed for 45 min. Finally, 5 mg/10 ml diaminobenzidine was used as chromogen, and the sections were put in 0.5% CuSO4/0.9% NaCl for 5 min to intensify the staining. Mayer hematoxylin counterstaining was applied. Sections were dehydrated through 100% alcohol, cleared in xylol, and coverslipped with permount (Fisher Scientific, Fiar Lawn, NJ, USA). Tissue sections of rectal carcinoma were used as positive controls. Evaluation of immunostaining results Investigators were blinded for the origin of the tissue sections, regarding patient and biopsy location. For evaluation of M30 staining, tissue sections were examined by light microscopy. M30 positivity was identified as brown cytoplasmic staining. M30-positive cells were marked by a first investigator (BvH) and reevaluated by an expert pathologist (IN). In all cases, the complete section was evaluated, and all M30-positive epithelial cells were counted. The apoptotic index was expressed as the number of M30-positive cells per tissue area in square millimeters. Tissue area was assessed by using a Zeiss KS400 computer-aided system. In each MIB-1-immunostained tissue section, crypts whose entire length could be visualized were photographed under ×400 magnification using a Zeiss KS400 computer-aided system. Crypts were excluded when they did not reach the muscularis mucosae or had multilayered bases. MIB-1 positivity was identified as brown nuclear staining. The number of MIB-1-positive epithelial cells and the total number of epithelial cells in up to five crypts per tissue section were counted from screen. The labeling index for each crypt was given by the ratio of MIB-1-positive cells and the total number of crypt epithelial cells and is expressed as percentage of total. For each patient, the labeling indices of pouch and afferent ileal loop were expressed as means of three to five counted crypts. If less than three crypts could be photographed for either pouch or ileal loop, the patient was excluded from analysis. The photographed crypts of five randomly selected patients were counted twice by one investigator (BvH) to determine intraobserver variability. Although great effort was made to obtain well-orientated mucosal crypts when using the quantitative method, however, not all biopsies reached the criteria mentioned above, and therefore, could not be examined. This problem mainly occurred in the biopsies taken from the pouch and might be due to friability of the pouch mucosa. We therefore developed a new semiquantitative scoring system. A representative part of the biopsies showing several complete crypt/villous axes was photographed under ×100 magnification. The photographs were judged pair-wise (pouch vs afferent loop) during which the investigators had to choose from four possible outcomes; one of the two locations showed most MIB-1 positivity. MIB-1 positivity did not differ or no judgement could be made. Judgement was based on relative length of the area of positive cells and the relative size of the stem cell compartment. Five investigators, two pathologists (IN, HvK), two gastroenterologists (PF, FN), and one junior investigator (BvH) independently compared the paired photographs of biopsies of pouch and afferent ileal loop of all patients. When three or more observers agreed in their judgement, this judgement was denoted as consensus judgement. If this criterion was not met, no consensus was reached. If the quality of the tissue sections was poor, no judgment was made. One investigator (BvH) judged the whole series twice for evaluation of intraobserver reliability. Statistical analyses Values for apoptosis and cell proliferation in the quantitative study were not expected to be normally distributed; therefore, they were presented as median and range. The Wilcoxon matched-pairs signed-ranks test was used to compare the paired observations in the apoptosis staining and the paired mean labeling indices in the cell proliferation study. Consensus judgements on the semiquantitative assessment of cell proliferation, favoring either pouch or ileal afferent loop, were compared with a Sign test. To evaluate the reliability of this semiquantitative method to assess cell proliferation, Cohen’s kappa was calculated for the first and second series of judgements by the prime investigator to determine intraobserver reliability. Also, for each pair of investigators, a Cohen’s kappa was calculated. The mean Cohen’s kappa was taken as value for interobserver reliability. Consensus judgments were compared to the difference in mean labeling index between pouch and ileal afferent loop for each evaluable patient. A p value less than 0.05 was considered as significant (SPSS for Windows 11.0.1, 2001). Results Patient characteristics Patient characteristics are given in Table 1. The median age of the 32 patients (19 men, 13 women) included in the study was 32 (range 16–72) years. Table 1Patient characteristics ApoptosisCell proliferationNumber of patients studied3220Male/Female19/1312/8Median age in years (range)32 (16–72)29 (16–62)Median age at surgery in years (range)24 (10–55)20 (10–52)Median age pouch in months (range)96 (9–216)105 (9–216)IPAA: hand-sewn/double-stapled/ unknown9/21/25/14/1Carcinoma at surgery40Patients with adenomas at biopsy:yes/no24/815/5For quantitative cell proliferation analysis, 12 patients had to be excluded from analysis (right column) due to absence of sufficient evaluable crypts (see under “Materials and methods”). Twenty-three patients were operated in the Radboud University Nijmegen Medical Centre. The median age at the time of reconstructive colectomy was 24 (range 10–55) years. A mucosectomy with hand-sewn IPAA was performed in 9 patients, and a double-stapled IPAA was performed in 21 patients. For two patients, the information about the performed technical procedure could not be retrieved. At the moment of colectomy, four patients had a colorectal adenocarcinoma localized in the rectum, sigmoid, hepatic flexure, or appendix, respectively. At the time of endoscopy, the median age of the pouch was 96 (range 9–216) months. The medication used was loperamide by 18 patients, psyllium fibres by 3 patients, and iron, metoclopramide, colestyramine, omeprazole, sulindac, tramadol, tamoxifen, gosereline, nifedipine, metoprolol, furosemide, or losartan each by 1 patient. Thirteen patients were not on medication 3 months before endoscopy. One patient used sulindac, a nonspecific cyclooxygenase inhibitor, which is thought to influence cell proliferation and especially apoptosis [15, 27]. Exclusion of this patient from analyses had no effect on the results. Histological examination revealed pouch adenomas in 24 patients (75%). Apoptosis In both pouch and afferent ileal loop, M30-positive cells were predominantly detected in the mucosal villi (Fig. 1). Fig. 1Apoptotic epithelial cells (encircled) using M30 immunohistochemistry. Original magnification ×200 (left) and ×400 (right) In the 32 pairs of biopsies from pouch and afferent ileal tissue investigated, the median absolute number of apoptotic cells in the pouch mucosa was two per tissue section (range 0–9), which was identical to the values in the afferent ileal loop (2, range 0–19). The median apoptotic index (expressed as number of M30-positive cells per mm2 tissue section area) did not differ between pouch (median 0.4/mm2, range 0–2.9) and afferent ileal tissue (median 0.3, range 0–2.7; see Fig. 2). Fig. 2Box-Whisker plots of apoptosis expressed as number of M30-positive epithelial cells per mm2 tissue section area of pouch and afferent ileum of 32 patients with FAP. Difference between groups was not significant. Values are given as median (fat line), lower to upper quartile (green box), and minimum and maximum values (error bars) Cell proliferation Quantitative comparison The intraobserver reliability for counting of the photographed crypts (Fig. 3) was rs = 0.855, p = 0.002. Fig. 3Detection of proliferating crypt cells with MIB-1 immunohistochemistry. Brown stained nuclei are positive. Nonproliferative crypt stem cells are visualized at the base of the crypt. Original magnification ×400 In 12 pouch tissue sections, less than three crypts were available for counting, and these samples were therefore excluded from analysis. For the same reason, two afferent ileal tissue sections were excluded. In the 20 pairs of pouch and afferent ileum tissue sections left for comparison, median labeling index (expressed as percentage of MIB-1-positive epithelial crypt cells) was significantly higher in the pouch as compared to the afferent ileum (median 68.3%, range 52.9–79.6% vs median 61.6%, range 38.0–73.9%; Wilcoxon signed-ranks test, p = 0.001; Fig. 4). Fig. 4Box-Whisker plots of cell proliferation expressed as median labeling index (MIB-1-positive crypt cells per total number of crypt cells) of 3–5 crypts in tissue sections of pouch and afferent ileum of 20 patients with FAP. Difference between groups was significant (p = 0.001). Values are given as median (fat line), lower to upper quartile (red box), and minimum and maximum values (error bars) Semiquantitative comparison The results of the semiquantitative comparison of the photographed tissue sections from the pouch and afferent ileum mucosae (Fig. 5) are visualized in Fig. 6. In 15 patients (47%), the proliferation was higher in the pouch; in 7 patients (22%), there was no difference; in 5 patients (16%), proliferation was higher in the afferent ileal loop; in 2 patients (6%), no consensus could be achieved; and in 3 patients (9%), no judgement was possible due to poor quality of the tissue sections. These differences were significant (p < 0.05). Fig. 5Photographs used for semiquantitative comparison of cell proliferation in MIB-1 immunohistochemically stained tissue sections of pouch (left) and afferent ileal mucosae (right; original magnification ×100)Fig. 6Histogram of semiquantitative comparison; 0 ileal mucosa shows more cell proliferation than pouch mucosa, 1 no difference in cell proliferation between pouch and ileal mucosae, 2 pouch mucosa shows more cell proliferation than ileal mucosa, 3 no judgement possible, 4 no consensus could be reached Intraobserver reliability was κ = 0.87. Cohen’s kappa for each pair of observers ranged from 0.26 to 0.59 with a mean Cohen’s kappa of 0.38, which corresponds with “fair agreement” following Byrt’s guidelines [6]. Thus, interobserver reliability was acceptable. In cases in which a labeling index could be assigned to both pouch and afferent ileal loop, consensus judgment was available in 18 cases (Fig. 7). In all cases in which the semiquantitative analyses showed more proliferation in the pouch (n = 11), this was confirmed by the quantitative method; the median value of the difference in proliferation between pouch and ileum was 0.107 (range 0.032–0.175). In the five cases where no difference was observed with the semiquantitative approach, the median value of the difference in proliferation between pouch and ileum was 0.023 (range −0.096 to 0.120). There were only three cases in which the semiquantitative method showed more proliferation in the afferent loop, and the median value of the difference in proliferation between pouch and ileum in these cases was 0.098 (range −0.021 to 0.262). The Spearman’s correlation coefficient between the semiquantitative and the quantitative methods was 0.273 (p = 0.244). Fig. 7Semiquantitative comparison vs difference in mean labeling index of pouch minus afferent ileal mucosa; 0 ileal mucosa shows more cell proliferation compared to the pouch mucosa, 1 no difference in cell proliferation between pouch and ileal mucosae, 2 pouch mucosa shows more cell proliferation compared to the ileal mucosa Discussion Although both the pouch and its afferent loop contain the same preexisting ileal mucosa, adenomas occur more frequently in the pouch of patients with FAP than in the afferent ileal loop, suggesting an accelerated adenoma formation in the pouch [14, 26, 33, 38]. To investigate the role of cell kinetics as a possible explanation for this observation, apoptosis and cell proliferation rates in the mucosa of the pouch were compared with those of the afferent ileal loop from the same patient to eliminate bias caused by interindividual differences. Cell proliferation was significantly higher in the pouch mucosa in comparison to mucosa of the afferent ileal loop. No significant difference in apoptosis was found in the mucosa of the pouch and afferent ileal loop. A low amount of apoptotic cells in the pouch as well as in the ileal mucosa was found. This might be caused by the APC mutation-induced apoptotic resistance [16]. Only a trend but no significant difference in apoptosis was found. Taking into account the broad range of apoptotic rates as found in our study samples, a significant difference might be found when a larger group of patients could be studied. There are no clear guidelines for estimation of ileal proliferation. For estimation of colonic proliferation, an accepted method is to count proliferating cells in five colonic crypts. Although most biopsy specimens were stretched and orientated directly after endoscopy, only a lower number of crypts could be counted completely along the longitudinal axis in most patients. For this reason, we accepted three full crypts as the minimal number to assess proliferation. Using this criterion, tissue sections from 12 patients still could not be used to determine the cell proliferation and had to be excluded from the study. Especially in the tissue sections of the pouch, this problem was evident and is possibly caused by a higher fragility of this tissue. In the remaining 20 pairs of tissue sections left for comparison, we found a significantly higher median labeling index in the pouch compared to the ileal afferent loop. To overcome the problem of the relative low number of assessable crypts, another quicker but less quantitative method was applied in which five investigators compared photographs of tissue sections of pouch and ileal mucosae. In accordance with the results of the first method, we found significantly higher cell proliferation in the mucosa of the pouch compared to that of the ileal mucosa. Although no significant correlation could be found in a case-by-case comparison between both methods, in all cases in which the semiquantitative analyses showed more proliferation in the pouch, this was confirmed by the quantitative analyses of crypts. Furthermore, the inter- and intraobserver variability was good. In addition, the semiquantitative method is far less time-consuming and can therefore give a relatively fast and easy impression of eventual differences in cell proliferation. Evaluation of this method in a larger study may further demonstrate its value. Data on cell proliferation in ileal mucosa in patients with FAP are limited. Previously, de Silva et al. [11] reported a labeling index of 19.8% in afferent ileal loop mucosa of patients with FAP or ulcerative colitis (UC), far lower than the 61.6% that was found here. Their labeling index for cell proliferation in ileal pouch mucosa (33.6%) was also much lower than in the present study. Only the labeling index of 51.7% as found by them in pouches with pouchitis nears the values we obtained. However, the majority of patients included in the study of de Silva et al. were patients with UC (23 of the 26 patients), so direct comparison with our results therefore seems inappropriate. In addition, Goldberg et al. [12] reported a median labeling index of 34.9% in the ileal pouch of patients with FAP, which is much lower than the 68.3% we found. However, their labeling indices were based on a minimum of three counted crypts, and if this number could not be reached, halves of crypts were included in the analyses. Moreover, their study group consisted of only 5 patients with FAP, whereas in the present study, 3 to 5 whole crypts of 20 patients were counted. Several studies indicated that cell proliferation of normal-appearing colorectal mucosa of patients with adenomas or carcinomas was 19 to 86% higher compared to colorectal mucosa of healthy persons [1, 28, 30, 32]. These findings strongly suggest an association between the presence of adenomas or carcinomas and an increased mucosal cell proliferation. However, the difference in cell proliferation between pouch and afferent ileal mucosae of 6.7% as found in this study is less pronounced in comparison to the above referred findings. A possible explanation for this relatively small difference in cell proliferation between pouch and ileal mucosae in our study may be that intestinal epithelial cell proliferation is already very high in patients with FAP [19, 31], and these high cell proliferation rates make a further increase less pronounced. The higher proliferation found in the pouch mucosa in comparison to mucosa of the afferent ileal loop can only be explained by intraluminal changes that occur after construction of the pouch. Whether changes in bacterial flora, bile acid composition, short-chain fatty acids, or other compounds are responsible for this finding remains unclear, but a better understanding of this process is necessary to find a possible treatment for this group of patients. In conclusion, the increased cell proliferation in the ileal pouch mucosa compared to the mucosa of the afferent ileal loop may contribute to the enhanced risk for adenomas and carcinomas in the pouch of patients with FAP and emphasizes the need for regular endoscopical surveillance of the pouch in these patients. In addition, cell proliferation can be used as an early endpoint marker in chemopreventive studies in these patients. The applied new method for semiquantitative evaluation of cell proliferation in immunohistochemically stained tissue sections seems promising, as it offers a relatively fast and easy means of assessment.