{"PMCID": "PMC176545", "title": "The Transcriptome of the Intraerythrocytic Developmental Cycle of Plasmodium falciparum", "abstract": "Plasmodium falciparum is the causative agent of the most burdensome form of human malaria, affecting 200\u2013300 million individuals per year worldwide. The recently sequenced genome of P. falciparum revealed over 5,400 genes, of which 60% encode proteins of unknown function. Insights into the biochemical function and regulation of these genes will provide the foundation for future drug and vaccine development efforts toward eradication of this disease. By analyzing the complete asexual intraerythrocytic developmental cycle (IDC) transcriptome of the HB3 strain of P. falciparum , we demonstrate that at least 60% of the genome is transcriptionally active during this stage. Our data demonstrate that this parasite has evolved an extremely specialized mode of transcriptional regulation that produces a continuous cascade of gene expression, beginning with genes corresponding to general cellular processes, such as protein synthesis, and ending with Plasmodium -specific functionalities, such as genes involved in erythrocyte invasion. The data reveal that genes contiguous along the chromosomes are rarely coregulated, while transcription from the plastid genome is highly coregulated and likely polycistronic. Comparative genomic hybridization between HB3 and the reference genome strain (3D7) was used to distinguish between genes not expressed during the IDC and genes not detected because of possible sequence variations. Genomic differences between these strains were found almost exclusively in the highly antigenic subtelomeric regions of chromosomes. The simple cascade of gene regulation that directs the asexual development of P. falciparum is unprecedented in eukaryotic biology. The transcriptome of the IDC resembles a \u201cjust-in-time\u201d manufacturing process whereby induction of any given gene occurs once per cycle and only at a time when it is required. These data provide to our knowledge the first comprehensive view of the timing of transcription throughout the intraerythrocytic development of P. falciparum and provide a resource for the identification of new chemotherapeutic and vaccine candidates.", "fulltext": "Introduction Human malaria is caused by four species of the parasitic protozoan genus Plasmodium . Of these four species, Plasmodium falciparum is responsible for the vast majority of the 300\u2013500 million episodes of malaria worldwide and accounts for 0.7\u20132.7 million annual deaths. In many endemic countries, malaria is responsible for economic stagnation, lowering the annual economic growth in some regions by up to 1.5% ( Sachs and Malaney 2002 ). While isolated efforts to curb malaria with combinations of vector control, education, and drugs have proven successful, a global solution has not been reached. Currently, there are few antimalarial chemotherapeutics available that serve as both prophylaxis and treatment. Compounding this paucity of drugs is a worldwide increase in P. falciparum strains resistant to the mainstays of antimalarial treatment ( Ridley 2002 ). In addition, the search for a malaria vaccine has thus far been unsuccessful. Given the genetic flexibility and the immunogenic complexity of P. falciparum , a comprehensive understanding of Plasmodium molecular biology will be essential for the development of new chemotherapeutic and vaccine strategies. The 22.8 Mb genome of P. falciparum is comprised of 14 linear chromosomes, a circular plastid-like genome, and a linear mitochondrial genome. The malaria genome sequencing consortium estimates that more than 60% of the 5,409 predicted open reading frames (ORFs) lack sequence similarity to genes from any other known organism ( Gardner et al. 2002 ). Although ascribing putative roles for these ORFs in the absence of sequence similarity remains challenging, their unique nature may be key to identifying Plasmodium -specific pathways as candidates for antimalarial strategies. The complete P. falciparum lifecycle encompasses three major developmental stages: the mosquito, liver, and blood stages. It has long been a goal to understand the regulation of gene expression throughout each developmental stage. Previous attempts to apply functional genomics methods to address these questions used various approaches, including DNA microarrays ( Hayward et al. 2000 ; Ben Mamoun et al. 2001 ; Le Roch et al. 2002 ), serial analysis of gene expression ( Patankar et al. 2001 ), and mass spectrometry ( Florens et al. 2002 ; Lasonder et al. 2002 ) on a limited number of samples from different developmental stages. While all of these approaches have provided insight into the biology of this organism, there have been no comprehensive analyses of any single developmental stage. Here we present an examination of the full transcriptome of one of these stages, the asexual intraerythrocytic developmental cycle (IDC), at a 1-h timescale resolution. The 48-h P. falciparum IDC ( Figure 1 A) initiates with merozoite invasion of red blood cells (RBCs) and is followed by the formation of the parasitophorous vacuole (PV) during the ring stage. The parasite then enters a highly metabolic maturation phase, the trophozoite stage, prior to parasite replication. In the schizont stage, the cell prepares for reinvasion of new RBCs by replicating and dividing to form up to 32 new merozoites. The IDC represents all of the stages in the development of P. falciparum responsible for the symptoms of malaria and is also the target for the vast majority of antimalarial drugs and vaccine strategies. Figure 1 Parasite Culturing and Data Characteristics of the P. falciparum IDC Transcriptome Analysis (A) Giemsa stains of the major morphological stages throughout the IDC are shown with the percent representation of ring-, trophozoite-, or schizont-stage parasites at every timepoint. The 2-h invasion window during the initiation of the bioreactor culture is indicated (gray area). (B\u2013D) Example expression profiles for three genes, encoding EBA175, DHFR-TS, and ASL, are shown with a loess fit of the data (red line). (E) MAL6P1.147, the largest predicted ORF in the Plasmodium genome, is represented by 14 unique DNA oligonucleotide elements. The location of each of the oligonucleotide elements within the predicted ORF and the corresponding individual expression profiles are indicated (oligo 1\u201314). A red/green colorimetric representation of the gene expression ratios for each oligonucleotide is shown below the graph. The pairwise Pearson correlation for these expression profiles is 0.98 \u00b1 0.02. (F) The percentage of the power in the maximum frequency of the FFT power spectrum was used as an indicator of periodicity. A histogram of these values reveals a strong bias toward single-frequency expression profiles, indicating that the majority of P. falciparum genes are regulated in a simple periodic manner. This bias is eliminated when the percent power was recalculated using random permutations of the same dataset (inset). For reference, the locations of EBA175 (peak B), DHFR-TS (peak C), and ASL (peak D) are shown. Our laboratory has developed a P. falciparum\u2013 specific DNA microarray using long (70 nt) oligonucleotides as representative elements for predicted ORFs in the sequenced genome (strain 3D7) ( Bozdech et al. 2003 ). Using this DNA microarray, we have examined expression profiles across 48 individual 1-h timepoints from the IDC of P. falciparum . Our data suggest that not only does P. falciparum express the vast majority of its genes during this lifecycle stage, but also that greater than 75% of these genes are activated only once during the IDC. For genes of known function, we note that this activation correlates well with the timing for the respective protein's biological function, thus illustrating an intimate relationship between transcriptional regulation and the developmental progression of this highly specialized parasitic organism. We also demonstrate the potential of this analysis to elucidate the function of the many unknown gene products as well as for further understanding the general biology of this parasitic organism. Results Expression Profiling of the IDC The genome-wide transcriptome of the P. falciparum IDC was generated by measuring relative mRNA abundance levels in samples collected from a highly synchronized in vitro culture of parasites. The strain used was the well-characterized Honduran chloroquine-sensitive HB3 strain, which was used in the only two experimental crosses carried out thus far with P. falciparum ( Walliker et al. 1987 ; Wellems et al. 1990 ). To obtain sufficient quantities of parasitized RBCs and to ensure the homogeneity of the samples, a large-scale culturing technique was developed using a 4.5 l bioreactor (see Materials and Methods ). Samples were collected for a 48-h period beginning 1 h postinvasion (hpi). Culture synchronization was monitored every hour by Giemsa staining. We observed only the asexual form of the parasite in these stains. The culture was synchronous, with greater than 80% of the parasites invading fresh RBCs within 2 h prior to the harvesting of the first timepoint. Maintenance of synchrony throughout the IDC was demonstrated by sharp transitions between the ring-to-trophozoite and trophozoite-to-schizont stages at the 17- and 29-h timepoints, respectively ( Figure 1 A). The DNA microarray used in this study consists of 7,462 individual 70mer oligonucleotides representing 4,488 of the 5,409 ORFs manually annotated by the malaria genome sequencing consortium ( Bozdech et al. 2003 ). Of the 4,488 ORFs, 990 are represented by more than one oligonucleotide. Since our oligonucleotide design was based on partially assembled sequences periodically released by the sequencing consortium over the past several years, our set includes additional features representing 1,315 putative ORFs not part of the manually annotated collection. In this group, 394 oligonucleotides are no longer represented in the current assembled sequence. These latter ORFs likely fall into the gaps present in the published assembly available through the Plasmodium genome resource PlasmoDB.org ( Gardner et al. 2002 ; Kissinger et al. 2002 ; Bahl et al. 2003 ). To measure the relative abundance of mRNAs throughout the IDC, total RNA from each timepoint was compared to an arbitrary reference pool of total RNA from all timepoints in a standard two-color competitive hybridization ( Eisen and Brown 1999 ). The transcriptional profile of each ORF is represented by the mean-centered series of ratio measurements for the corresponding oligonucleotide(s) ( Figure 1 B\u20131E). Inspection of the entire dataset revealed a striking nonstochastic periodicity in the majority of expression profiles. The relative abundance of these mRNAs continuously varies throughout the IDC and is marked by a single maximum and a single minimum, as observed for the representative schizont-specific gene, erythrocyte-binding antigen 175 ( eba175 ), and the trophozoite-specific gene, dihydrofolate reductase\u2013thymidylate synthetase ( dhfr-ts ) ( Figure 1 B and 1C). However, there is diversity in both the absolute magnitude of relative expression and in the timing of maximal expression (phase). In addition, a minority of genes, such as adenylosuccinate lyase ( asl ) ( Figure 1 D), displayed a relatively constant expression profile. The accuracy of measurements from individual oligonucleotides was further verified by the ORFs that are represented by more than one oligonucleotide feature on the microarray. The calculated average pairwise Pearson correlation ( r ) is greater than 0.90 for 68% (0.75 for 86%) of the transcripts represented by multiple oligonucleotides with detectable expression during the IDC ( Table S1 ). Cases in which data from multiple oligonucleotides representing a single putative ORF disagree may represent incorrect annotation. The internal consistency of expression profile measurements for ORFs represented by more than one oligonucleotide sequence is graphically shown in Figure 1 E for the hypothetical protein MAL6P1.147, the largest predicted ORF in the genome (31 kb), which is represented by 14 oligonucleotide elements spanning the entire length of the coding sequence. The average pairwise correlation ( r ) for these features is 0.98\u00b10.02. Periodicity in genome-wide gene expression datasets has been used to identify cell-cycle-regulated genes in both yeast and human cells ( Spellman et al. 1998 ; Whitfield et al. 2002 ). Owing to the cyclical nature of the P. falciparum IDC dataset, a similar computational approach was taken. We performed simple Fourier analysis, which allowed us to calculate both the apparent phase and frequency of expression for each gene during the IDC (see Materials and Methods ). The fast Fourier transform (FFT) maps a function in a time domain (the expression profile) into a frequency domain such that when the mapped function is plotted (the power spectra), sharp peaks appear at frequencies where there is intrinsic periodicity. The calculated power spectra for each expression profile confirmed the observation that the data are highly periodic. The majority of profiles exhibited an overall expression period of 0.75\u20131.5 cycles per 48 h. We have used the FFT data for the purpose of filtering the expression profiles that are inherently noisy (i.e., that have low signal) or that lack differential expression throughout the IDC. Since the majority of the profiles display a single low-frequency peak in the power spectrum, we have taken advantage of this feature to classify profiles, similar to the application of a low-pass filter in signal processing. By measuring the power present in the peak frequency window (the main component plus two adjacent peaks) relative to the power present at all frequencies of the power spectrum, we were able to define a score (percent power) that we have used to stratify the dataset. The resulting distribution of expression profiles, scored in this way, is shown in Figure 1 F for all oligonucleotides. For reference, the positions of profiles corresponding to eba175 (peak B), dhfr-ts (peak C), and asl (peak D) are indicated. It is striking that 79.5% of the expression profiles have a very high score (greater than 70%). For comparison, we applied our FFT analysis to the Saccharomyces cerevisiae cell cycle data, yielding only 194 profiles (3.8%) above a 70% score ( Figure S1 ). In addition, we randomly permuted the columns of the complete dataset 1,000 times, each time recalculating the FFT, for a total of 5 million profiles (see inset in Figure 1 F). The randomized set exhibits essentially no periodicity: the probability of any random profile scoring above 70% is 1.3 \u00d7 10 \u22125 . P. falciparum Transcriptome Overview To provide an overview of the IDC transcriptome, we selected all 3,719 microarray elements whose profiles exhibited greater than 70% of the power in the maximum frequency window and that were also in the top 75% of the maximum frequency magnitudes. Although hierarchical clustering is extremely useful for comparing any set of expression data, regardless of the experimental variables, we sought to specifically address temporal order within the dataset. To accomplish this, the FFT phase was used to order the expression profiles to create a phaseogram of the IDC transcriptome of P. falciparum ( Figure 2 A). The overview set represents 2,714 unique ORFs (3,395 oligonucleotides). An additional 324 oligonucleotides represent ORFs that are not currently part of the manually annotated collection. Figure 2 Overview of the P. falciparum IDC Transcriptome (A) A phaseogram of the IDC transcriptome was created by ordering the transcriptional profiles for 2,712 genes by phase of expression along the y-axis. The characteristic stages of intraerythrocytic parasite morphology are shown on the left, aligned with the corresponding phase of peak gene expression. (B\u2013M) The temporal ordering of biochemical processes and functions is shown on the right. Each graph corresponds to the average expression profile for the genes in each set and the mean peak-to-trough amplitude is shown in parentheses. The IDC phaseogram depicts a cascade of continuous expression lacking clear boundaries or sharp transitions. During the first half of the IDC, a large number of genes involved in general eukaryotic cellular functions are induced with broad expression profiles. This gradual continuum includes the transition from the ring to the early trophozoite stage and the trophozoite to the early schizont stage, encompassing approximately 950 and 1,050 genes, respectively. Next, the mid- and late-schizont stages are marked by a rapid, large amplitude induction of approximately 550 genes, many of which appear to be continually expressed into the early-ring stage. However, owing to the level of synchrony in the culture, the ring-stage signal may be partially attributed to cross-contamination from residual schizonts. In the final hours of the IDC, approximately 300 genes corresponding to the early-ring stage are induced, indicating that reinvasion occurs without obvious interruptions to initiate the next cycle. The expression profiles for developmentally regulated genes in the P. falciparum IDC transcriptome reveal an orderly timing of key cellular functions. As indicated in Figure 2 B\u20132M, groups of functionally related genes share common expression profiles and demonstrate a programmed cascade of cellular processes that ensure the completion of the P. falciparum IDC. Ring and Early-Trophozoite Stage In the following text, we have grouped the genes according to temporal expression phases based on their association with the common P. falciparum cytological stages. Following invasion, approximately 950 ORFs are induced during the ring and early trophozoite stage, including genes associated with the cytoplasmic transcriptional and translational machinery, glycolysis and ribonucleotide biosynthesis ( Figure 2 B\u20132E). Represented in this group are 23 ORFs involved in transcription, including the four subunits of RNA polymerase I, nine subunits of RNA polymerase II, three subunits of RNA polymerase III, and four transcription factors. The average expression profile for this group is shown in Figure 2 B. (See Table S2 for all functional group details.) Also in this set are three previously identified P. falciparum RNA polymerase genes: the large subunits of P. falciparum RNA polymerase I ( Fox et al. 1993 ) and RNA polymerase II ( Li et al. 1989 ) and RNA polymerase III ( Li et al. 1991 ). The cytoplasmic translation gene group ( Figure 2 C) consists of 135 ORFs including homologues for 34 small and 40 large ribosomal subunits, 15 translation initiation factors, five translation elongation factors, 18 aminoacyl-tRNA synthetases, and 23 RNA helicases. In addition to the manually annotated ORFs, the translation gene group contains three ORFs predicted only by automated annotation including two ribosomal proteins (chr5.glm_215, chr5.glm_185) and a homologue of eIF-1A (chr11.glm_489) ( PlasmoDB.org ). In one case, chr5.glm_185 overlaps with the manually annotated ORF PFE0850w, which is found on the opposite strand. Oligonucleotide elements for both of these ORFs are present on the array. The oligonucleotide corresponding to the automated prediction yielded a robust FFT score and a phase consistent with the translation machinery, yet no PFE0850w expression was detected. These results suggest that the automated prediction for chr5.glm_185 most likely represents the correct gene model for this genomic locus and illustrates the use of the IDC expression data for further verification of the P. falciparum genome annotations. Another set of 33 ORFs with homology to components of the translational machinery displayed an entirely distinct expression pattern, being induced during the late-trophozoite and early-schizont stage. This group includes 11 homologues of chloroplast ribosomal proteins, four mitochondrial/chloroplast elongation factors, and six amino acid tRNA synthetases ( Table S2 ). These ORFs also share a common pattern of expression, suggesting that these factors are components of the mitochondrial and/or the plastid translation machinery. This observation is supported by the presence of predicted apicoplast-targeting signals in 18 of these proteins ( PlasmoDB.org ). In addition, one of these factors, ribosomal protein S9, has been experimentally immunolocalized within the plastid ( Waller et al. 1998 ). These data suggest that the peak of expression for the cytoplasmic translation machinery occurs in the first half of the IDC, whereas plastid and mitochondrial protein synthesis is synchronized with the maturation of these organelles during the second half of the IDC. In addition to transcription and translation, genes involved in several basic metabolic pathways were also induced during the ring and early-trophozoite stage, including glycolysis and ribonucleotide biosynthesis ( Figure 2 D and 2E). Unlike the majority of P. falciparum biochemical processes, most of the enzymes involved in nucleotide metabolism and glycolysis have been identified ( Reyes et al. 1982 ; Sherman 1998 ). The glycolysis group ( Figure 2 D) is tightly coregulated throughout the IDC and contains all of the 12 known enzymes. Expression initiates after reinvasion and continues to increase toward maximal expression during the trophozoite stage, when metabolism is at its peak. The glycolytic pathway is very well preserved in P. falciparum and exemplifies how data from this study can complement the homology-based interpretation of the genome. First, the genome contains two putative copies of pyruvate kinase on chromosomes 6 and 10, MAL6P1.160 and PF10_0363, respectively ( Gardner et al. 2002 ). However, only one of these genes, MAL6P1.160, has a similar expression profile to the other known glycolytic enzymes, suggesting that this enzyme is the main factor of this step in the glycolytic pathway. Interestingly, PF10_0363 contains a putative apicoplast-targeting signal ( PlasmoDB.org ). In another case, the malaria genome sequencing consortium has predicted two homologues of triose phosphate isomerase, PF14_0378 and PFC0381w. The latter is not detected by our analysis, suggesting that this gene is utilized in another developmental stage or may be a nonfunctional, redundant homologue. P. falciparum parasites generate pyrimidines through a de novo synthesis pathway while purines must be acquired by the organism through a salvage pathway ( Gero and O'Sullivan 1990 ). The mRNA levels of 16 enzymes corresponding to members of the pyrimidine ribonucleotide synthesis pathway, beginning with carbamoyl phosphate synthetase and ending with CTP synthetase, were uniformly induced immediately after invasion ( Figure 2 E). The relative abundance of these transcripts peaked at approximately 18\u201322 hpi and then rapidly declined. Similar expression characteristics were detected for the enzymes of the purine salvage pathway, including the nucleoside conversion enzymes, hypoxanthine\u2013guanine\u2013xanthine phosphoribosyltransferase, and both guanylate and adenylate kinases ( Figure 2 E; Table S2 ). Trophozoite and Early-Schizont Stage The mRNA expression data indicate that ribonucleotide and deoxyribonucleotide production is clearly bifurcated into two distinct temporal classes. While ribonucleotide synthesis is required in the early stages of the IDC, deoxyribonucleotide metabolism is a trophozoite/early-schizont function. mRNA transcripts for enzymes that convert ribonucleotides into deoxyribonucleotides, including DHFR-TS and both subunits of ribonucleotide reductase, were induced approximately at 10 hpi, peaking at approximately 32 hpi ( Figure 2 F). This represents a temporal shift from the induction of ribonucleotide synthesis of approximately 8\u201310 h. The expression of the deoxyribonucleotide biosynthesis is concomitant with the induction of DNA replication machinery transcripts, reflecting a tight relationship between DNA synthesis and production of precursors for this process. Thirty-two ORFs with homologies to various eukaryotic DNA replication machinery components are transcribed during the late-trophozoite and early-schizont stage. The timing of their transcription presages cell division. This functional gene group ( Figure 2 G), with peak expression around 32 hpi, contains the previously characterized P. falciparum DNA Pol\u03b1, DNA Pol\u03b4, and proliferating cell nuclear antigen, as well as the vast majority of the DNA replication components predicted by the malaria genome sequencing consortium ( Gardner et al. 2002 ). These additional components include eight predicted DNA polymerase subunits, two putative origin recognition complex subunits, six minichromosome maintenance proteins, seven endo- and exonucleases, seven replication factor subunits, and two topoiosomerases. Interestingly, a number of proteins typically required for eukaryotic DNA replication, including the majority of the subunits of the origin recognition complex, have not yet been identified by conventional sequence similarity searches of the P. falciparum genome. All genes necessary for the completion of the tricarboxylic acid (TCA) cycle were detected in the Plasmodium genome ( Gardner et al. 2002 ), although earlier studies indicate an unconventional function for this metabolic cycle. These studies suggest that the TCA cycle does not play a major role in the oxidation of glycolytic products. Instead, it is essential for the production of several metabolic intermediates, such as succinyl-CoA, a precursor of porphyrin biosynthesis ( Sherman 1998 ). The peak of expression for all TCA factors was detected during the late-trophozoite and early-schizont stage ( Figure 2 H). Consistent with the model suggesting a disconnection of the TCA cycle from glycolysis during the IDC, no expression was detected for the subunits of the pyruvate dehydrogenase complex, including the \u03b1 and \u03b2 chains of pyruvate dehydrogenase and dihydrolipoamide S-acetyl transferase, the typical links between glycolysis and the TCA cycle. On the other hand, expression of TCA cycle genes is well synchronized with the expression of a large number of mitochondrial genes, including the three ORFs of the mitochondrial genome ( Feagin et al. 1991 ), and several factors of electron transport ( Table S2 ). Although some of the TCA cycle proteins have been localized to the cytoplasm ( Lang-Unnasch 1992 ), the expression data suggest an association of this biochemical process with mitochondrial development and possibly with the abbreviated electron transport pathway detected in this organelle. Schizont Stage A transition from early to mid-schizont is marked by the maximal induction of 29 ORFs predicted to encode various subunits of the proteasome ( Figure 2 I). Seven \u03b1 and six \u03b2 subunits of the 20S particle and 16 ORFs of the 19S regulatory particle were identified in this gene group. The common expression profile for the subunits of both of the 26S particle complexes suggests the involvement of ubiquitin-dependent protein degradation in the developmental progression of the parasite. The peak of proteasome expression coincides with a transition in the IDC transcriptome from metabolic and generic cellular machinery to specialized parasitic functions in the mid-schizont stage. This suggests an association between transcriptional regulation and protein turnover during this and possibly other transitions during the progression of the P. falciparum IDC. In the schizont stage, one of the first specialized processes induced was expression from the plastid genome ( Figure 2 J). The essential extrachromosomal plastid (or apicoplast) genome contains 60 potentially expressed sequences, including ribosomal proteins, RNA polymerase subunits, ribosomal RNAs, tRNAs, and nine putative ORFs, including a ClpC homologue ( Wilson et al. 1996 ). Very little is known about the regulation of gene expression in the plastid, but it is thought to be polycistronic ( Wilson et al. 1996 ). In support of this observation, we find that 27 of the 41 plastid-specific elements present on our microarray displayed an identical expression pattern ( Figure 3 C). The remaining elements correspond mainly to tRNAs and failed to detect appreciable signal. The highly coordinated expression of the plastid genome, whose gene products are maximally expressed in the late-schizont stage, is concomitant with the replicative stage of the plastid ( Williamson et al. 2002 ). Note that not all plastid ORFs are represented on the microarray used in this study, and thus it is a formal possibility that the expression of the missing genes may differ from those shown in Figure 3 C. Figure 3 Coregulation of Gene Expression along the Chromosomes of P. falciparum Is Rare, While Plastid Gene Expression Is Highly Coordinated Expression profiles for oligonucleotides are shown as a function of location for Chromosome 2 ([A], Oligo Map). With the exception of the SERA locus (B), coregulated clusters of adjacent ORFs are seldom observed, indicating that expression phase is largely independent of chromosomal position. (C) In contrast to the nuclear chromosomes, the polycistronic expression of the circular plastid genome is reflected in the tight coregulation of gene expression. This is an expanded view of the plastid-encoded genes from Figure 2 J. Genomic differences between strain 3D7, from which the complete genome was sequenced, and strain HB3 were measured by CGH. The relative hybridization between the gDNA derived from these two strains is shown as a percent reduction of the signal intensity for 3D7 ([A], CGH Data). Differences between the two strains are predominately located in the subtelomeric regions that contain the highly polymorphic var, rifin, and stevor gene families. Intrachromosomal variations, as observed for the msp2 gene, were rare. Offset from the plastid by approximately 6 h, a set of approximately 500 ORFs exhibited peak expression during the late-schizont stage. Merozoite invasion of a new host cell is a complex process during which the parasite must recognize and dock onto the surface of the target erythrocyte, reorient with its apical tip toward the host cell, and internalize itself through invagination of the erythrocytic plasma membrane. The entire sequence of invasion events is facilitated by multiple receptor\u2013ligand interactions with highly specialized plasmodial antigens ( Cowman et al. 2000 ). The merozoite invasion group contains 58 ORFs, including 26 ORFs encoding antigens previously demonstrated to be important for the invasion process (see Figure 2 K). These include integral membrane proteins delivered to the merozoite surface from the micronemes (AMA1 and EBA175), GPI-anchored proteins of the merozoite membrane (MSP1, MSP4, and MSP5), proteins extrinsically associated with the merozoite surface during their maturation in the PV (MSP3 and MSP6), and soluble proteins secreted to the parasite\u2013host cell interface (RAP1, RAP2, and RAP3). In addition, late-schizont-specific expression was observed for several antigens whose functions are not completely understood, but which have been associated with the invasion process. These ORFs include the merozoite-capping protein (MCP1), erythrocyte-binding-like protein 1 (EBL1), reticulocyte-binding proteins (RBP1 and RBP2), acid basic repeat antigen (ABRA), MSP7, and a homologue of the Plasmodium yoelii merozoite antigen 1. As expected, peak expression of these antigens coincides with the maturation of merozoites and development of several apical organelles, including rhoptries, micronemes, and dense granules. Many of these proteins have been considered as vaccine candidates since antibodies against these antigens were readily detected in the immune sera of both convalescent patients as well as individuals with naturally acquired immunity ( Preiser et al. 2000 ). The sensitivity of invasion to protease and kinase inhibitors indicates an essential role for these activities in merozoite release as well as in the reinvasion process ( Dluzewski and Garcia 1996 ; Blackman 2000 ; Greenbaum et al. 2002 ). The merozoite invasion gene group contains three serine proteases, including PfSUB1, PfSUB2, and an additional homologue to plasmodial subtilases (PFE0355c), and two aspartyl proteases, plasmepsin (PM) IX and X. Peak expression during the mid-schizont stage was also observed for seven members of the serine repeat antigen (SERA) family, all of which contain putative cysteine protease domains. In addition to the proteases, expression of 12 serine/threonine protein kinases and three phophorylases was tightly synchronized with the genes of the invasion pathway, including six homologues of protein kinase C, three Ca + -dependent and two cAMP-dependent kinases, phosphatases 2A and 2B, and protein phosphatase J. Another functionally related gene group whose expression is sharply induced during the late-schizont stage includes components of actin\u2013myosin motors (see Figure 2 L) ( Pinder et al. 2000 ). As in other apicomplexa, actin and myosin have been implicated in host cell invasion ( Opitz and Soldati 2002 ). Schizont-specific expression was observed for three previously described class XIV myosin genes, one associated light chain, two actin homologues, and three additional actin cytoskeletal proteins, including actin-depolymerizing factor/cofilin (two isoforms) and coronin (one isoform). Although the molecular details of plasmodial actin\u2013myosin invasion are not completely understood, the tight transcriptional coregulation of the identified factors indicates that the examination of schizont-specific expression may help to identify additional, possibly unique elements of this pathway. Early-Ring Stage The expression data are continuous throughout the invasion process, with no observable abrupt change in the expression program upon successful reinvasion. However, a set of approximately 300 ORFs whose expression is initiated in the late-schizont stage persists throughout the invasion process and peaks during the early-ring stages (see Figure 2 M). It was previously determined that immediately after invasion, a second round of exocytosis is triggered, ensuring successful establishment of the parasite within the host cell ( Foley et al. 1991 ). One of the main P. falciparum virulence factors associated with this process is ring-infected surface antigen 1 (RESA1). RESA1 is secreted into the host cell cytoplasm at the final stages of the invasion process, where it binds to erythrocytic spectrin, possibly via its DnaJ-like chaperone domain ( Foley et al. 1991 ). The early stages of the IDC contain a variety of putative molecular chaperones in addition to RESA1, including RESA2 and RESAH3, plus five additional proteins carrying DnaJ-like domains. However, the functional roles of these chaperones remain unclear. Despite the cytoplasmic role of RESA1, abundant antibodies specific for RESA1 are present in individuals infected with P. falciparum , indicating that RESA1 is also presented to the host immune system ( Troye-Blomberg et al. 1989 ). Several genes encoding additional antigenic factors are found among the early ring gene group, including frequently interspersed repeat antigen (FIRA), octapeptide antigen, MSP8, and sporozoite threonine- and asparagine-rich protein (STARP). Like RESA1, antibodies against these antigens are also found in the sera of infected individuals, suggesting that the final stages of invasion might be a target of the immune response. Overall, the genes expressed during the mid- to late-schizont and early-ring stage encode proteins predominantly involved in highly parasite-specific functions facilitating various steps of host cell invasion. The expression profiles of these genes are unique in the IDC because of the large amplitudes and narrow peak widths observed. The sharp induction of a number of parasite-specific functions implies that they are crucial for parasite survival in the mammalian host and hence should serve as excellent targets for both chemotherapeutic and vaccine-based antimalarial strategies. IDC Transcriptional Regulation and Chromosomal Structure Transcriptional regulation of chromosomal gene expression in P. falciparum is thought to be monocistronic, with transcriptional control of gene expression occurring through regulatory sequence elements upstream and downstream of the coding sequence ( Horrocks et al. 1998 ). This is in contrast to several other parasites, such as Leishmania sp. , in which polycistronic mRNA is synthesized from large arrays of coding sequences positioned unidirectionally along the arms of relatively short chromosomes ( Myler et al. 2001 ). Recent proteomic analyses failed to detect any continuous chromosomal regions with common stage-specific gene expression in several stages of the P. falciparum lifecycle ( Florens et al. 2002 ). However, transcriptional domains have previously been suggested for Chromosome 2 ( Le Roch et al. 2002 ). The availability of the complete P. falciparum genome coupled with the IDC transcriptome allows us to investigate the possibility of chromosomal clustering of gene expression (see Figure 3 A). To systematically explore the possibility of coregulated expression as a function of chromosomal location, we applied a Pearson correlation to identify similarities in expression profiles among adjacent ORFs. The pairwise Pearson correlation was calculated for every ORF pair within each chromosome ( Figure S2 ). Gene groups in which the correlation of 70% of the possible pairs was greater than r = 0.75 were classified as putative transcriptionally coregulated groups. Using these criteria, we identified only 14 coregulation groups consisting of greater than three genes, with the total number of genes being 60 (1.4% of all represented genes) ( Table S3 ). In eight of the 14 groups, the coregulation of a pair of genes may be explained by the fact that they are divergently transcribed from the same promoter. A set of 1,000 randomized permutations of the dataset yielded 2.25 gene groups. Contrary to the nuclear chromosomes, there was a high correlation of gene expression along the plastid DNA element, consistent with polycistronic transcription (see Figure 3 C). The average pairwise Pearson correlation for a sliding window of seven ORFs along the plastid genome is 0.92\u00b10.03. The largest group demonstrating coregulation on the nuclear chromosomes corresponds to seven genes of the SERA family found on Chromosome 2 (see Figure 3 B) ( Miller et al. 2002 ). Besides the SERA gene cluster and a group containing three ribosomal protein genes, no additional functional relationship was found among the other chromosomally adjacent, transcriptionally coregulated gene groups. The limited grouping of regional chromosomal expression was independent of strand specificity and, with the exception of the SERA group, did not overlap with the groups of \u201crecently duplicated genes\u201d proposed by the malaria genome sequencing consortium ( Gardner et al. 2002 ). Three major surface antigens, the var , rifin , and stevor families, have a high degree of genomic variability and are highly polymorphic between strains and even within a single strain ( Cheng et al. 1998 ; Afonso Nogueira et al. 2002 ; Gardner et al. 2002 ). Expression profiles for only a small subset of these genes were detected in the IDC transcriptome and were typically characterized by low-amplitude profiles. This could be due to two nonmutually exclusive possibilities: first, the HB3 DNA sequence for these genes may be substantially rearranged or completely deleted relative to the reference strain, 3D7; second, only a few of these genes may be selectively expressed, as has been proposed ( Deitsch et al. 2001 ). To identify regions of genomic variability between 3D7 and HB3, we performed microarray-based comparative genomic hybridization (CGH) analysis. Array-based CGH has been performed with human cDNA and bacterial artificial chromosome-based microarrays to characterize DNA copy-number changes associated with tumorigenesis ( Gray and Collins 2000 ; Pollack et al. 2002 ). Using a similar protocol, CGH analysis revealed that the majority of genetic variation between HB3 and 3D7 is confined to the subtelomeric chromosomal regions containing the aforementioned gene families ( Figure 3 A; Figure S3 ). Only 28.3% of rifin , 47.1% of var , and 51.0% of stevor genes predicted for the 3D7 strain were detected for the HB3 genomic DNA (gDNA) when hybridized to the 3D7-based microarray. Thus, the underrepresentation of these gene families in the HB3 IDC transcriptome is likely due to the high degree of sequence variation present in these genes. Excluding the three surface antigen families in the subtelomeric regions, 97% of the remaining oligonucleotide microarray elements exhibit an equivalent signal in the CGH analysis. However, 144 of the differences detected by CGH reside in internal chromosomal regions and include several previously identified plasmodial antigens: MSP1, MSP2 ( Figure 3 A), S antigen, EBL1, cytoadherence-linked asexual gene 3.1 (CLAG3.1), glutamine-rich protein (GLURP), erythrocyte membrane protein 3 (PfEMP3), knob-associated histidine-rich protein (KAHRP), and gametocyte-specific antigen Pfg377 ( Table S4 ). These results demonstrate a high degree of genetic variation within the genes considered to be crucial for antigenic variation between these two commonly used laboratory strains of P. falciparum . Implications for Drug Discovery The majority of the nuclear-encoded proteins targeted to the plastid are of prokaryotic origin, making them excellent drug targets ( McFadden and Roos 1999 ). Moreover, inhibitors of plastid-associated isoprenoid biosynthesis, DNA replication, and translation have been shown to kill the P. falciparum parasite, demonstrating that the plastid is an essential organelle ( Fichera and Roos 1997 ; Jomaa et al. 1999 ). The plastid has been implicated in various metabolic functions, including fatty acid metabolism, heme biosynthesis, isoprenoid biosynthesis, and iron\u2013sulfur cluster formation ( Wilson 2002 ). It is clear that, within the plastid, functional ribosomes are assembled to express the ORFs encoded by the plastid genome ( Roy et al. 1999 ). However, nuclear-encoded components are required to complete the translational machinery as well as for all other plastid metabolic functions. A bipartite signal sequence is required for efficient transport of these nuclear proteins from the cytoplasm to the plastid via the endoplasmic reticulum ( Waller et al. 2000 ). Computational predictions suggest that the P. falciparum genome may contain over 550 nuclear-encoded proteins with putative transit peptides ( Zuegge et al. 2001 ; Foth et al. 2003 ). Given that over 10% of the ORFs in the P. falciparum genome are predicted to contain an apicoplast-targeting sequence, we sought to use the IDC transcriptome as a means to narrow the search space for candidate apicoplast-targeted genes. As mentioned above, the expression profiles for genes encoded on the plastid genome are tightly coordinated (see Figure 3 C). We reasoned that genes targeted to the plastid would be expressed slightly before or coincidentally with the plastid genome. Therefore, we utilized the FFT phase information to identify ORFs in phase with expression of the plastid genome (see Materials and Methods ) ( Table S5 ). Because the genes of the plastid genome are maximally expressed between 33 and 36 hpi, we searched for all genes in the dataset with an FFT phase in this time window and then cross-referenced the list of predicted apicoplast-targeted sequences ( PlasmoDB.org ), resulting in a list of 124 in-phase apicoplast genes ( Figure 4 A). Within this list are two ORFs that have been directly visualized in the apicoplast, acyl carrier protein and the ribosomal subunit S9 ( Waller et al. 1998 ), as well as many ORFs associated with the putative plastid ribosomal machinery, enzymes involved in the nonmevalonate pathway, additional caseineolytic proteases (Clps), the reductant ferredoxin, and replication/transcriptional machinery components. However, this list contains only 14 of the 43 proteins categorized in the Gene Ontology (GO) assignments at PlasmoDB.org as apicoplast proteins by inference from direct assay (IDA). In addition, 30% of the nuclear-encoded translational genes that are not coexpressed with the known cytoplasmic machinery are found within this small group of genes. More importantly, 76 ORFs (62%) are of unknown function, with little or no homology to other genes. This limited subgroup of putative plastid-targeted ORFs are likely excellent candidates for further studies in the ongoing search for malaria-specific functions as putative drug targets. Figure 4 Temporal Distribution of the Apicoplast-Targeted Proteins and P. falciparum Proteases, Potential Antimalarial Drug Candidates (A) The expression profiles of all putative plastid-targeted genes represented on our microarray are shown. The yellow box encompasses a highly synchronized group of genes, which are in-phase with plastid genome expression. The average expression profile for this in-phase group of genes is shown and includes most of the known apicoplast-targeted genes as well as many hypothetical genes. For reference, the average expression profile for the plastid genome is shown (dashed gray line). (B) Proteases represent an attractive target for chemotherapeutic development. The broad range of temporal expression for various classes of proteases and their putative functions are displayed. Abbreviations: HAP, histo-aspartyl protease (PM III); Clp, caseineolytic protease; sub1, 2, subtilisin-like protease 1 and 2. Similarly, P. falciparum proteases have received much attention, since they are candidates as drug targets and have been shown to play important roles in regulation as well as metabolism throughout the IDC ( Rosenthal 2002 ). A temporal ordering of expression profiles for several well-characterized P. falciparum proteases is shown in Figure 4 B, demonstrating the broad significance of these enzymes throughout the IDC. One of the principal proteolytic functions is considered to be the degradation of host cell hemoglobin in the food vacuole (FV) to produce amino acids essential for protein synthesis. This elaborate process is carried out by a series of aspartyl proteases, cysteine proteases, metalloproteases, and aminopeptidases ( Francis et al. 1997 ). A family of ten aspartyl proteases, the plasmepsins (PMs), has been identified in the P. falciparum genome, four of which have been characterized as bona fide hemoglobinases: PM I, II, III (a histo-aspartic protease [HAP]), and IV ( Coombs et al. 2001 ). Our data reveal that the PMs are expressed at different times throughout the lifecycle, suggesting that they are involved in different processes throughout the IDC. PM I, II, HAP, and PM IV are adjacent to one another on Chromosome 14 and have been localized to the FV. While HAP and PM II are expressed in the mid-trophozoite stage, during peak hemoglobin catabolism, PMI and IV are maximally expressed in the ring stage along with the cysteine protease falcipain-1 (FP-1). FP-1 has recently been implicated in merozoite invasion and has been localized to the interior of the PV ( Greenbaum et al. 2002 ). The coincident expression of these proteases implies that the development of the PV and the FV occurs during the very early-ring stage. This observation is corroborated by similar expression profiles for the PV-associated protein RESA1 and the FV protein PGH1. Subsequently, a second group of hemoglobinases, including the m1-family aminopeptidase, FP-2, and falcilysin, is expressed simultaneously with HAP and PM II during the trophozoite stage of the IDC. The expression of PM V and the newly identified FP- 2 homologue during this stage suggests they are also important in the trophozoite stage. The other known falcipain, FP-3, does not show a marked induction in expression throughout the IDC. We fail to detect any transcripts for PM VI, VII, and VIII during the IDC. These genes may have roles in any of the other sexual, liver, or mosquito stages of development. In addition to the hemoglobinases, P. falciparum contains a variety of proteases involved in cellular processing, including a group of Clps and signal peptidases that are all expressed maximally at the late-trophozoite stage ( Figure 4 B). The timing of these genes may play a key role in protein maturation during trafficking to various compartments, including the plastid. The three Clps contain putative leader peptides and may actually function within the plastid. Finally, a group of proteases are expressed in the schizont stage and include the P. falciparum subtilisin-like proteases PfSUB1 and PfSUB2 as well as PMs IX and X. PfSUB1 and PfSUB2 are believed to be involved in merozoite invasion and have been localized apically in the dense granules. Interestingly, there are two PfSUB1 protease homologues (PFE0355c and PFE0370c); PM X parallels the expression of PfSUB1 (PFE0370c), suggesting that aspartyl proteases may also be involved in merozoite invasion. In addition, the phase of the PfSUB1 homologue suggests a concomitant role, with PM IX slightly preceding merozoite invasion. In total, we have detected gene expression for over 80 putative proteases throughout the entire IDC ( Table S6 ). This set includes over 65 proteases from a group of recently predicted proteases ( Wu et al. 2003 ). The differing temporal expression of these proteases may allow for a multifaceted approach toward identifying protease inhibitors with efficacy at all stages of the IDC. Implications for New Vaccine Therapies Merozoite invasion is one of the most promising target areas for antimalarial vaccine development ( Good 2001 ). Many vaccine efforts thus far have focused primarily on a set of plasmodial antigens that facilitate receptor\u2013ligand interaction between the parasite and the host cell during the invasion process ( Preiser et al. 2000 ) (see Figure 2 K and 2M). Merozoite invasion antigens are contributing factors to naturally acquired immunity, triggering both humoral and antibody-independent cell-mediated responses ( Good and Doolan 1999 ). Antibodies against these antigens have been demonstrated to effectively block the merozoite invasion process in vitro and in animal models ( Ramasamy et al. 2001 ). Owing to the highly unique character of merozoite surface antigens, homology-based searches have yielded only a limited set of additional invasion factors. We utilized the IDC transcriptome to predict a set of likely invasion proteins by identifying expression profiles with characteristics similar to previously studied merozoite invasion proteins. The expression profiles for all known invasion factors undergo a sharp induction during the mid- to late-schizont stage and are characterized by large expression amplitudes (see Figure 2 A). Among these proteins are seven of the best-known malaria vaccine candidates, including AMA1, MSP1, MSP3, MSP5, EBA175, RAP1, and RESA1. To identify ORFs with a possible involvement in the merozoite invasion process, we have calculated the similarity, by Euclidian distance, between the expression profiles of these seven vaccine candidates and the rest of the IDC transcriptome. A histogram of the distance values reveals a bimodal distribution with 262 ORFs in the first peak of the distribution ( Figure S4 ). This represents the top 5% of expression profiles when ranked by increasing Euclidian distance ( Table S7 ). In addition to the seven vaccine candidate genes used for the search, essentially all predicted P. falciparum merozoite-associated antigens were identified in this gene set ( Figure 5 ). These include the GPI-anchored MSP4; several integral merozoite membrane proteins, such as EBA140 and EBL1; three RBPs (RBP1, RBP2a, RBP2b); and a previously unknown RBP homologue. In addition, components of two proteins secreted from the rhoptries to the host cell membranes, RhopH1 and RhopH3, or to the PVs RAP1, RAP2, and RAP3 were found in the selected set. Surprisingly, CLAG2 and CLAG9 were also classified into the merozoite invasion group. Although the biological function of these genes is believed to be associated with cytoadherence of the infected erythrocyte to the vascular endothelium, a highly related homologue, CLAG3.1 (RhopH1), was recently detected in the rhoptries, suggesting a possible secondary role for these genes in merozoites ( Kaneko et al. 2001 ). Figure 5 Phaseogram of Putative Vaccine Targets The similarity of all expression profiles to seven known vaccine candidates (boxed) was calculated. The top 5% of similar profiles correspond to 262 ORFs, 28 of which have been previously associated with plasmodial antigenicity and the process of merozoite invasion. A number of antigens are presently in various stages of clinical trials and are yielding encouraging results ( Good et al. 1998 ). However, many single-antigen vaccine studies indicate that the most promising approach will require a combination of antigenic determinants from multiple stages of the complex plasmodial lifecycle ( Kumar et al. 2002 ). Searches for new target antigens in the P. falciparum genome are thus vital to the development of future vaccines, since no fully protective vaccine has been assembled thus far. Of the 262 ORFs whose expression profiles were closest to the profiles of the seven major vaccine candidates, 189 are of unknown function. These ORFs represent a candidate list for new vaccine targets. Discussion The transcriptome of the IDC of P. falciparum constitutes an essential tool and baseline foundation for the analysis of all future gene expression studies in this organism, including response to drugs, growth conditions, environmental perturbations, and genetic alterations. Essentially all experiments involving asexual intraerythrocytic-stage parasites must be interpreted within the context of the ongoing cascade of IDC-regulated genes. In our global analysis of the P. falciparum transcriptome, over 80% of the ORFs revealed changes in transcript abundance during the maturation of the parasite within RBCs. The P. falciparum IDC significantly differs from the cell cycles of the yeast S. cerevisiae ( Spellman et al. 1998 ) and human HeLa ( Whitfield et al. 2002 ) cells, during which only 15% of the total genome is periodically regulated. Instead, the P. falciparum IDC resembles the transcriptome of the early stages of Drosophila melanogaster development, which incorporates the expression of over 80% of its genome as well ( Arbeitman et al. 2002 ). Unlike the development of multicellular eukaryotes, there is no terminal differentiation and, with the exception of gametocytogenesis, the parasite is locked into a repeating cycle. In this respect, the P. falciparum IDC mirrors a viral-like lifecycle, in which a relatively rigid program of transcriptional regulation governs the progress of the course of infection. The lack of continuous chromosomal domains with common expression characteristics suggests that the genes are regulated individually, presumably via distinct sets of cis - and trans -acting elements. However, the extent and the simple mechanical character of transcriptional control observed in the IDC suggest a fundamentally different mode of regulation than what has been observed in other eukaryotes. It is plausible that a comparatively small number of transcription factors with overlapping binding site specificities could account for the entire cascade. While further experiments are ongoing, it may be the case that P. falciparum gene regulation is streamlined to the extent that it has lost the degree of dynamic flexibility observed in other unicellular organisms, from Escherichia coli to yeast. This observation also implies that disruption of a key transcriptional regulator, as opposed to a metabolic process, may have profound inhibitory properties. While a few putative transcription factors have been identified in the P. falciparum genome, no specific regulatory elements have been defined in basepair-level detail. A further analysis of the upstream regions of genes with similar phases should facilitate the elucidation of regulatory regions and their corresponding regulatory proteins. In general, the timing of mRNA expression for a given gene during the IDC correlates well with the function of the resultant protein. For example, replication of the genome occurs in the early-schizont stage and correlates well with the peak expression of all factors of DNA replication and DNA synthesis. Also, organellar biogenesis of several intracellular compartments such as mitochondria, the plastid, or the apical invasion organelles is concomitant with the maximal induction of mRNAs encoding proteins specific to these organelles. In addition, our data are generally in good agreement with proteomic analyses that have detected intraerythrocytic-stage proteins from the merozoite, trophozoite, and schizont stages. More than 85% of the 1,588 proteins detected in these studies were also expressed in our analysis ( Florens et al. 2002 ; Lasonder et al. 2002 ). However, a more detailed proteomic analysis at different stages of the IDC will be needed to ascertain the temporal changes of these proteins. We initially expected that a high percentage of the genome would be specialized for each lifecycle stage (mosquito, liver, blood), yet this was not observed; the mRNA transcripts for 75% of proteins determined to be gamete-, gametocyte-, or sporozoite-specific by mass spectrometry are also transcribed in the plasmodial IDC. These findings confirm previous studies demonstrating that not only genes used for generic cellular processes are present in multiple developmental stages, but also factors of highly specialized Plasmodium functions ( Gruner et al. 2001 ). This may indicate that only a small portion of the genome may actually be truly specific to a particular developmental stage and that the majority of the genome is utilized throughout the full lifecycle of this parasite. It is also feasible to speculate that a multilayer regulatory network is employed in the progression of the entire P. falciparum lifecycle. In this model, the same cis - and trans -acting regulatory elements driving the actual mRNA production in IDC are utilized in other developmental stages. These elements are then controlled by an alternate subset of factors determining the status of the lifecycle progression. These findings also outline two contrasting properties of the P. falciparum genome. The Plasmodium parasite devotes 3.9% of its genome to a complex system of antigenic determinants essential for host immune evasion during a single developmental stage ( Gardner et al. 2002 ). On the other hand, large portions of the genome encode proteins used in multiple stages of the entire lifecycle. Such broad-scope proteins might be excellent targets for both vaccine and chemotherapeutic antimalarial strategies, since they would target several developmental stages simultaneously. While there are certainly proteins specific to these nonerythrocytic stages, a complementary analysis of both proteomic and genomic datasets will facilitate the search. With malaria continuing to be a major worldwide disease, advances toward understanding the basic biology of P. falciparum remain essential. Our analysis of the IDC transcriptome provides a first step toward a comprehensive functional analysis of the genome of P. falciparum . The genome-wide transcriptome will be useful not only for the further annotation of many uncharacterized genes, but also for defining the biological processes utilized by this highly specialized parasitic organism. Importantly, candidate groups of genes can be identified that are both functionally and transcriptionally related and thus provide focused starting points for the further elucidation of genetic and mechanistic aspects of P. falciparum . Such biological characterizations are presently a major objective in the search for novel antimalarial strategies. The public availability of the dataset presented in this study is intended to provide a resource for the entire research community to extend the exploration of P. falciparum beyond the scope of this publication. All data will be freely accessible at two sites: http://plasmodb.org and http://malaria.ucsf.edu . Materials and Methods Cell culture. A large-scale culture of P. falciparum (HB3 strain) was grown in a standard 4.5 l microbial bioreactor (Aplikon, Brauwweg, Netherlands) equipped with a Bio Controller unit ADI 1030 (Aplikon, Brauwweg, Netherlands). Cells were initially grown in a 2% suspension of purified human RBCs and RPMI 1640 media supplemented with 0.25% Albumax II (GIBCO, Life Technologies, San Diego, California, United States), 2 g/l sodium bicarbonate, 0.1 mM hypoxanthine, 25 mM HEPES (pH 7.4), and 50 \u03bcg/l gentamycin, at 37\u00b0C, 5% O 2 , and 6% CO 2 . Cells were synchronized by two consecutive sorbitol treatments for three generations, for a total of six treatments. Large-scale cultures contained 32.5 mM HEPES (pH 7.4). The bioreactor culture was initiated by mixing 25.0 ml of parasitized RBCs (20% late schizonts, approximately 45 hpi) with an additional 115.0 ml of purified RBC in a total of 1.0 l of media (14% hematocrit). Invasion of fresh RBCs occurred during the next 2 h, raising the total parasitemia from an initial 5% to 16%. After this period, the volume of the culture was adjusted to 4.5 l, bringing the final RBC concentration to approximately 3.3% to reduce the invasion of remaining cells. Immediately after the invasion period, greater than 80% of the parasites were in the ring stage. Temperature and gas conditions were managed by the Bio Controller unit. Over the course of 48 h, 3\u20134 ml of parasitized RBCs was collected every hour, washed with prewarmed PBS, and flash-frozen in liquid nitrogen. RNA preparation and reference pool. P. falciparum RNA sample isolation, cDNA synthesis, labeling, and DNA microarray hybridizations were performed as described by Bozdech et al. (2003 ). Samples for individual timepoints (coupled to Cy5) were hybridized against a reference pool (coupled to Cy3). The reference pool was comprised of RNA samples representing all developmental stages of the parasite. From this pool, sufficient cDNA synthesis reactions, using 12 \u03bcg of pooled reference RNA, were performed for all hybridizations. After completing cDNA synthesis, all reference pool cDNAs were combined into one large pool and then split into individual aliquots for subsequent labeling and hybridization. Microarray hybridizations were incubated for 14\u201318 h. DNA microarray hybridizations and quality control. In total, 55 DNA microarray hybridizations covering 46 timepoints were performed. Timepoints 1, 7, 11, 14, 18, 20, 27, and 31 were represented by more than one array hybridization. Data were acquired and analyzed by GenePix Pro 3 (Axon Instruments, Union City, California, United States). Array data were stored and normalized using the NOMAD microarray database system ( http://ucsf-nomad.sourceforge.net/ ). In brief, a scalar normalization factor was calculated for each array using unflagged features with median intensities greater than zero for each channel and a pixel regression correlation coefficient greater than or equal to 0.75. Quality spots were retained based on the following criteria. The log 2 (Cy5/Cy3) ratio for array features that were unflagged and had a sum of median intensities greater than the local background plus two times the standard deviation of the background were extracted from the database for further analysis. Subsequently, expression profiles consisting of 43 of 46 timepoints (approximately 95%) were selected. For those timepoints that were represented by multiple arrays, the ratio values were averaged. FFT analysis of the expression profiles. Fourier analysis was performed on each profile in the quality-controlled set (5,081 oligonucleotides). Profiles were smoothed with missing values imputed using a locally weighted regression algorithm with local weighting restricted to 12% using R ( http://www.R-project.org ). Fourier analysis was performed on each profile using the fft() function of R, padded with zeros to 64 measurements. The power spectrum was calculated using the spectrum() function of R. The power at each frequency ( Power() ), the total power (P tot ), and the frequency of maximum power (F max ) were determined. The periodicity score was defined as Power [(F max\u22121 ) + (F max ) + (F max+1 )]/P tot . The most frequent value of F max across all profiles was deemed the major frequency (m) and used in determining phase information. The phase of each profile was calculated as atan2\\[\u2212(I (m)],R (m)\\, where atan2 is R's arctangent function and I and R are the imaginary and real parts of the FFT. Profiles were then ordered in increasing phase from \u2212\u03c0 to \u03c0. The loess smooth profiles were drawn through the raw expression data using the loess() function found in the modern regression library of R (version 1.5.1). The default parameters were used, with the exception that local weighting was reduced to 30%. For the averaged profiles of the functional groups (see Figure 2 B\u20132M), the loess smooth profiles were calculated for each expression profile individually and subsequently averaged to create the representative profile. These same methods were applied to both the randomized set (see the inset to Figure 1 F) and the yeast cell cycle dataset (see Figure S1 ). The raw results files ( Dataset S1 ), the fully assembled raw dataset ( Dataset S2 , the overview dataset ( Dataset S3 , and the quality control dataset ( Dataset S4 ) are available as downloads. Evaluation of coexpression along chromosomes. The evaluation of coexpression of genes along chromosomes was carried out as follows. The Pearson correlation coefficient was calculated for each pair of profiles. For ORFs with multiple oligonucleotides, the average profile was calculated. The neighborhood of each ORF profile was defined as a window of between one and ten adjacent ORF profiles. If any window in an ORF profile's neighborhood displayed more than 70% pairwise correlation of greater than 0.75, it was flagged as enriched. The length of the window was then recorded as a region of coexpression. This process was repeated without strand separation of ORFs and with randomly permuted datasets. Comparative genomic hybridization. P. falciparum strains 3D7 and HB3 were cultured as previously described at a concentration of 10% parasitaemia. Genomic DNA (gDNA) was isolated from a minimum of 500 ml of total culture for each P. falciparum strain, as previously described ( Wang et al. 2002 ). Isolated gDNA from each strain was sheared by sonication to an average fragment size of approximately 1\u20131.5 kb and then was purified and concentrated using a DNA Clean and Concentrator kit (Zymo Research, Orange, California, United States). Amino-allyl-dUTP first was incorporated into the gDNA fragments with a Klenow reaction at 37\u00b0C for 6\u20138 h with random nonamer primers and 3 \u03bcg of sheared gDNA. After purification and concentration of the DNA from the Klenow reaction, CyScribe Cy3 and Cy5 dyes (Amersham Biosciences, Buckinghamshire, United Kingdom) were coupled to HB3 DNA and 3D7 DNA, respectively, as previously described ( Pollack et al. 1999 ). Uncoupled fluorescent dye was removed using a DNA Clean and Concentrator kit. Labeled DNA fragments were hybridized to the oligonucleotide-based DNA microarrays. Fluorescence was detected and analyzed using an Axon Instruments scanner and GenePix Pro 3.0 software. Only features that had median intensities greater than the local background plus two times the standard deviation of the background in each channel were considered for further analysis. For each feature, the percent of the total intensity was determined using the signal in the 3D7 channel as the total amount of intensity for each oligonucleotide; intensity differences less than 50% were considered to be significant for subsequence analysis. Calculation for in-phase plastid-targeted genes. The range of FFT-based phases for the expression profiles of the plastid genome is between 0.32 and 1.05 (or roughly \u03c0/9 \u2212\u03c0/3). Using the list of 551 apicoplast-targeted genes available at PlasmoDB.org , we first ordered these genes by phase and then grouped all genes with a phase range between 0.00 and 1.40 (0\u20134\u03c0/9), resulting in 124 genes represented by 128 oligonucleotides on the microarray. This select group represents the in-phase plastid targeted genes (see Table S6 ). Calculation for vaccine targets. To select the expression profiles most related to the AMA1, MSP1, MSP3, MSP5, EBA175, RAP1, and RESA1 vaccine candidates, we calculated the similarity of all expression profiles in the dataset to those of these antigens by Euclidian distance. The minimum Euclidian distance calculated for every profile was then binned into 60 bins and plotted as a histogram. A natural break in the histogram was seen that included the set of 262 ORFs (see Figure S2 ). Supporting Information Dataset S1 Raw GenePix Results (29.5 MB ZIP). Click here for additional data file. Dataset S2 Complete Dataset (3.7 MB TXT). Click here for additional data file. Dataset S3 Overview Dataset (2.4 MB TXT). Click here for additional data file. Dataset S4 Quality Control Set (3.1 MB TXT). Click here for additional data file. Figure S1 Histogram of the Percent Power at Peak Frequencies for the Yeast Cell Cycle Data The percent of power in the maximum frequency of the FFT power spectrum was used to determine periodicity of the yeast cell cycle data from Spellman et al. (1998 ). The histogram reveals periodic regulation of gene expression for only a small subset of genes (% power >70%). (223 KB EPS). Click here for additional data file. Figure S2 Pearson Correlation Maps for the P. falciparum Chromosomes A matrix of the pairwise Pearson correlations was calculated for every expression profile along the chromosomes. The analysis included all annotated ORFs. The gray areas correspond to a Pearson correlation d(x, y) = 0 and indicate ORFs with no detectable IDC expression or ORFs not represented on the microarray. The starting point (left) and the end point (right) of the chromosomes and the ORF order along the chromosomes are identical to the order in PlasmoDB.org . (30.9 MB EPS). Click here for additional data file. Figure S3 CGH of 3D7 versus HB3 for All Chromosomes Genomic differences between strain 3D7 and strain HB3 were measured by CGH. The relative hybridization between the gDNA derived from these two strains is shown as a percent reduction of the signal intensity for 3D7 along individual chromosomes. (1.7 MB ZIP). Click here for additional data file. (A) (216 KB EPS) Click here for additional data file. (B) (232 KB EPS) Click here for additional data file. (C) (237 KB EPS) Click here for additional data file. (D) (240 KB EPS) Click here for additional data file. (E) (252 KB EPS) Click here for additional data file. (F) (232 KB EPS) Click here for additional data file. (G) (235 KB EPS) Click here for additional data file. (H) (235 KB EPS) Click here for additional data file. (I) (249 KB EPS) Click here for additional data file. (J) (265 KB EPS) Click here for additional data file. (K) (283 KB EPS) Click here for additional data file. (L) (270 KB EPS) Click here for additional data file. (M) (305 KB EPS) Click here for additional data file. (N) (332 KB EPS) Click here for additional data file. Figure S4 Distribution of Euclidian Distances between Expression Profiles of the IDC Genes and Seven Vaccine Candidates The similarity between each IDC expression profile and the profiles of the seven selected vaccine candidate genes was evaluated by Euclidian distance calculations, d(x,y) = \u03a3(x i \u2212 y i ) 2 . The Euclidian distance value to the closest vaccine homologue was selected for each IDC profile and used to generate this plot. Genes with d(x,y) < 20 were selected for the phaseogram of putative vaccine targets (see Figure 5 ). (494.02 KB EPS). Click here for additional data file. Table S1 Pearson Correlation for ORFs Represented by Multiple Oligonucleotides This table contains all of the ORFs in the analyzed dataset that are represented by multiple oligonucleotides on the DNA microarray. The average Pearson correlation value has been calculated for the expression profiles of all oligonucleotides for each given ORF. (44 KB TXT). Click here for additional data file. Table S2 P. falciparum Functional Gene Groups This table contains all of the P. falciparum groups discussed. The groups include the following: transcription machinery, cytoplasmic translation machinery, the glycolytic pathway, ribonucleotide synthesis, deoxyribonucleotide synthesis, DNA replication machinery, the TCA cycle, the proteaseome, the plastid genome, merozoite invasion, actin\u2013myosin motility, early-ring transcripts, mitochondrial genes, and the organellar translational group. (291 KB TXT). Click here for additional data file. Table S3 Coregulation along the Chromosomes of P. falciparum This table contains the regions of coregulation found in the chromosomes of P. falciparum determined by calculating the Pearson correlation between expression profiles for contiguous ORFs. The cutoff was 70% pairwise correlation of greater than 0.75 for each group. Only groups of two ORFs or more are listed. (6 KB TXT). Click here for additional data file. Table S4 3D7 versus HB3 CGH Data This table contains all of the intensity data from CGH of gDNA derived from the 3D7 and HB3 strains of P. falciparum . The averaged intensities from three microarray hybridization experiments are listed. (414 KB TXT). Click here for additional data file. Table S5 Putative Apicoplast-Targeted Genes and Expression Profiles This table contains all of the predicted apicoplast-targeted ORFs from PlasmoDB.org . The presence of each ORF on the DNA microarray is tabulated, as well as whether each ORF is present in the overview set. Finally, the plastid ORFs in-phase with plastid genome expression are listed, as well as the corresponding oligonucleotide identifiers. (147 KB TXT). Click here for additional data file. Table S6 Putative P. falciparum Proteases and Their Expression Data The table was constructed by searching the database for any putative protease annotations and contains all of the 92 proteases identified by Wu et al. (2003 ). (59 KB TXT). Click here for additional data file. Table S7 Vaccine Candidate Correlation Table The similarity of all expression profiles to seven known vaccine candidates was evaluated by a Euclidian distance calculation to all expression profiles measured. These 262 ORFs constitute the top 5% of genes in the IDC with minimum distance to these seven ORFs. The seven candidates used are AMA1, MSP1, MSP3, MSP5, EBA175, RAP1, and RESA1. (204 KB TXT). Click here for additional data file."}
{"PMCID": "PMC1043859", "title": "Perceptual and Neural Olfactory Similarity in Honeybees", "abstract": "The question of whether or not neural activity patterns recorded in the olfactory centres of the brain correspond to olfactory perceptual measures remains unanswered. To address this question, we studied olfaction in honeybees Apis mellifera using the olfactory conditioning of the proboscis extension response. We conditioned bees to odours and tested generalisation responses to different odours. Sixteen odours were used, which varied both in their functional group (primary and secondary alcohols, aldehydes and ketones) and in their carbon-chain length (from six to nine carbons).The results obtained by presentation of a total of 16 \u00d7 16 odour pairs show that (i) all odorants presented could be learned, although acquisition was lower for short-chain ketones; (ii) generalisation varied depending both on the functional group and the carbon-chain length of odours trained; higher generalisation was found between long-chain than between short-chain molecules and between groups such as primary and secondary alcohols; (iii) for some odour pairs, cross-generalisation between odorants was asymmetric; (iv) a putative olfactory space could be defined for the honeybee with functional group and carbon-chain length as inner dimensions; (v) perceptual distances in such a space correlate well with physiological distances determined from optophysiological recordings of antennal lobe activity. We conclude that functional group and carbon-chain length are inner dimensions of the honeybee olfactory space and that neural activity in the antennal lobe reflects the perceptual quality of odours.", "fulltext": "Introduction Stimulus discrimination and generalisation constitute two major abilities exhibited by most living animals. Discrimination allows treating different signals as distinct, while generalisation allows treating different but similar stimuli as equivalents [ 1 , 2 , 3 ]. Similarity along one or several perceptual dimensions determines the degree of generalisation between stimuli [ 2 ]. Determining such dimensions is fundamental for defining an animal's perceptual space. This objective remains, however, elusive in the case of the olfactory modality in which the dimensions along which odours are evaluated are not well known. Characteristics such as the functional chemical group or the carbon-chain length of a chemical substance may influence olfactory perception. It is known that at least some features of odorant molecules influence olfactory perception. For instance, some enantiomers can be discriminated by humans and nonhuman primates [ 4 ]. If and how chemical group and carbon-chain length are integrated as inner dimensions into an olfactory perceptual space remains unknown. Vertebrate and invertebrate nervous systems show important functional as well as anatomical similarities in the way in which olfactory signals are detected and processed in their brains, particularly at the level of their first olfactory centres, the olfactory bulb in the case of vertebrates and the antennal lobe (AL) in the case of insects [ 5 , 6 , 7 ]. Insects are useful models for studying olfaction, as their behaviour heavily relies on the use of olfactory cues. The honeybee Apis mellifera is one such model in which behavioural and neurobiological studies have been performed to unravel the basis of olfaction [ 8 , 9 , 10 , 11 ]. Honeybee foragers are \u2018flower constant' and learn and memorise a given floral species that they exploit at a time as long as it is profitable. Floral cues, among which odours play a prominent role, are then associated with nectar or pollen reward [ 12 , 13 ]. However, under natural conditions, the blends of volatiles emitted by floral sources vary widely in quantity and quality both in time and in space [ 14 , 15 ]. To cope with such changes in an efficient way, a \u2018flower constant' forager should be able to generalise its choice to the same kind of floral sources despite fluctuations in their volatile emissions. In a pioneering investigation, von Frisch [ 16 ] trained freely flying bees to visit an artificial feeder presenting several essential oils (odour mixtures). Using a set of 32 odour mixtures, von Frisch observed that after learning that a blend was associated with sucrose solution, bees tended to prefer this odour blend, but they sometimes visited other blends that were similar (to the human nose) to the rewarded one. Olfactory generalisation in honeybees was mainly studied on restrained honeybees using the conditioning of the proboscis extension reflex (PER) [ 17 , 18 ]. In this paradigm, harnessed honeybees are conditioned to odours associated with a sucrose reward. When the antennae of a hungry bee are touched with sucrose solution, the animal reflexively extends its proboscis to reach out towards and to lick the sucrose. Odours presented to the antennae do not usually release such a reflex in naive animals. If an odour is presented immediately before sucrose solution (forward pairing), an association is formed and the odour will subsequently trigger the PER in a subsequent unrewarded test. This effect is clearly associative and involves classical conditioning [ 18 ]. Thus, the odour can be viewed as the conditioned stimulus (CS), and sucrose solution as an appetitive unconditioned stimulus (US). Bees conditioned to individual odours or to olfactory mixtures can generalise PER to a wide range of different olfactory stimuli. Using the PER paradigm, Vareschi [ 19 ] showed that bees generalise most often between odours with similar carbon-chain lengths and between odours belonging to the same functional group. However, Vareschi conditioned odours in a differential way, with two rewarded and many unrewarded odours, so that several generalisation gradients (excitatory and inhibitory) may have interacted in an unknown way to determine the generalisation responses exhibited by the bees [ 19 ]. Using a similar approach and a restricted (6 \u00d7 6) set of odour combinations, Smith and Menzel [ 20 ] confirmed that bees generalise among odours with the same functional group, but their analysis did not detail the results obtained with individual odour combinations, thus rendering impossible the analysis of generalisation between odours with similar carbon-chain lengths. Free-flying bees trained in a differential way to a rewarded odour presented simultaneously with multiple unrewarded odours also generalise between odours with similar functional groups [ 21 ]. As for Vareschi's study [ 19 ], such an experimental design makes it difficult to interpret the generalisation responses due to unknown interactions between excitatory and inhibitory generalisation gradients. Recently, optical imaging studies facilitated our understanding of how olfactory stimuli are detected and processed in the bee brain [ 22 , 23 , 24 , 25 , 26 ]. The first relay of the bee's olfactory system involves the ALs, which receive sensory input from the olfactory receptor neurons of the antennae within a number of 160 functional units, the glomeruli [ 27 , 28 , 29 ]. Within each glomerulus, synaptic contacts are formed with local interneurons and projection neurons (PNs). PNs send processed information from the ALs to higher brain centres such as the mushroom bodies and the lateral protocerebrum [ 30 ]. Stimulation with an odour leads to a specific spatiotemporal pattern of activated glomeruli, as shown, using in vivo calcium imaging techniques that employ fluorescent dyes to measure intracellular calcium in active neurons [ 22 , 24 , 31 ]. The odour-evoked activity patterns are conserved between individuals and constitute therefore a code [ 23 , 24 ]. Odours with similar chemical structures tend to present similar glomerular activity patterns [ 23 ]. Furthermore, it is believed that the neural code of odour-evoked glomerular patterns measured in the bee brain actually represent the perceptual code, although this idea was never tested directly. In the present work, we studied behavioural olfactory generalisation, using the PER conditioning paradigm, with 16 odorants varying in two chemical features, functional group and chain length. The odours belonged to four chemical categories: alcohols with the functional group on the first or second carbon of the carbon chain (henceforth primary and secondary alcohols, respectively), aldehydes, and ketones. They possessed therefore three functional groups (alcohol, aldehyde, ketone). Their chain length ranged from six to nine carbon atoms (C6, C7, C8, and C9). The pairwise combination of 16 odours defined a 16 \u00d7 16 matrix. These odours are well discriminated by free-flying bees [ 21 ] and give consistent odour-evoked signals in optical imaging studies [ 23 ]. Using a behavioural approach, we measured similarity between odours and calculated their perceptual distances in a putative olfactory space. These perceptual distances were correlated with physiological distances measured in optical imaging experiments [ 23 ]. The correlation between both datasets was highly significant, thus indicating that odours that are encoded as physiologically similar are also perceived as similar by honeybees. Although other studies have addressed the issue of perceptual correlates of neural representations [ 32 , 33 ], we show for the first time that neural olfactory activity corresponds to olfactory perception defined on the basis of specific dimensions in a putative olfactory space, a finding that is of central importance in the study of the neurobiology of perception. Results We trained 2,048 honeybees along three trials in which one of the 16 odours used in our experiments was paired with a reward of sucrose solution (conditioned odour). Afterwards, each bee was tested with four odours that could include or not include the trained odour. Acquisition Phase The level of PER in the first conditioning trial was very low (between 0% and 8.60%) for all odours ( Figure 1 ). All the 16 odours were learnt but not with the same efficiency. An overall (trial \u00d7 odour) analysis of variance (ANOVA) showed a significant increase in responses along trials ( F 2, 4064 = 2215.50, p < 0.001) and a significant heterogeneity among odours ( F 15, 2032 = 8.80, p < 0.001). Responses to the CS in the last conditioning trial reached a level of approximately 70% for primary and secondary alcohols, 80% for aldehydes, and 61% for ketones. Figure 1 Acquisition Curves for Primary Alcohols, Secondary Alcohols, Aldehydes, and Ketones The ordinate represents the percentage of proboscis extensions to the training odour (CS). The abscissa indicates the conditioning trials (C1, C2, C3) and the test with the CS (T). The curves correspond to molecules with 6 (white triangles), 7 (white diamonds), 8 (black circles) and 9 carbons (black squares); ( n = 128 bees for each curve). As not all 128 bees were tested with the odour used as CS, the sample size in the tests was smaller ( n = 32). Different letters (a, b, c) indicate significant differences either between acquisition curves for different chain-length molecules (in the case of the ketones) or between test responses (post hoc Scheff\u00e9 tests). In the case of aldehydes and primary and secondary alcohols, no significant chain-length effect within functional groups was found over the whole conditioning procedure (chain length \u00d7 trial ANOVA; chain-length effect for primary alcohols: F 3, 508 = 0.18, p > 0.05; secondary alcohols: F 3, 508 = 1.47, p > 0.05; and aldehydes: F 3, 508 = 1.26, p > 0.05). In contrast, bees conditioned to ketones showed a significant chain-length effect in the acquisition (chain length \u00d7 trial ANOVA; chain-length effect: F 3, 508 = 20.00, p < 0.005). Scheff\u00e9 post hoc comparisons showed that acquisition was significantly better for nonanone (81.25% responses in the last conditioning trial) than for all other ketones. Octanone (68.75% responses in the last conditioning trial) was also better learned than hexanone and heptanone (45.31% and 48.44% responses in the last conditioning trial, respectively) ( Figure 1 , bottom right). The effect over trials was significant in all cases ( p < 0.05) as bees learned all odours. The analysis of acquisition for each chain length separately revealed that it varied significantly depending on the functional group (functional group \u00d7 trial ANOVA; C6: F 3, 508 = 18.89; p < 0.005; C7: F 3, 508 =10.78; p < 0.005; C8: F 3, 508 = 3.84; p < 0.01; C9: F 3, 508 = 2.73, p < 0.05). Scheff\u00e9 post hoc comparisons generally showed that this effect was mainly due to ketones being less well learned than aldehydes and alcohols. Generally, the longer the carbon chain, the lower the heterogeneity in acquisition between functional groups. Thus, apart from short-chain ketones, all odours were learned similarly (reaching a level of acquisition between 60% and 80% in the last conditioning trial). Test Phase When the conditioned odour was presented in a test ( Figure 1 , grey panels), the level of PER recorded corresponded mainly to that found in the last acquisition trial (McNemar tests [2 \u00d7 2 Table]: in all cases p > 0.05). To compare generalisation after conditioning, and because acquisition levels were heterogeneous between odours, we built a generalisation matrix in which only bees responding to the CS at the end of training (3rd conditioning trial) were considered ( Figure 2 ). The number of individuals included in the statistical analysis varied within each \u2018training odour/test odour' pair. The number of bees completing the tests varied between 17 and 28 for primary alcohols, between 13 and 29 for secondary alcohols, between 23 and 30 for aldehydes, and between 11 and 31 for ketones. The responses to the CS in the tests ranged between 70% and 100% in the generalisation matrix. All further analyses were carried out on this matrix. In the following sections, we will use the matrix data to analyse generalisation within and between functional groups, within and between chain lengths, and the asymmetries in olfactory generalisation. Figure 2 Olfactory Generalisation Matrix The generalisation matrix represents the percentage of PER in the tests performed by bees that actually learned the CS, that is, bees that responded to the CS at the third conditioning trial ( n = 1,457). Upper part: percentages recorded. Lower part: colour-coded graphic display grouping the level of responses in ten 10% response categories. Red, maximal response; light blue, minimal response. Generalisation within Functional Groups Figure 3 A shows the percentage of PER to odours having different (white quadrants) or the same (grey quadrants) functional group as the conditioned odour. High levels of PER to odours different from the trained one correspond to high generalisation. In order to better visualise generalisation as depending on functional groups, we pooled all the observed responses within each quadrant of Figure 3 A (i.e., not considering chain length) and calculated the resulting percentage of PER ( Figure 3 B). Grey bars correspond to generalisation to the same functional group; white bars correspond to generalisation to different functional groups. Generalisation mainly occurred within a given functional group (grey bars). This pattern was clearest for aldehydes ( Figure 3 B, 3 rd row) because bees conditioned to aldehydes responded with a high probability to other aldehydes but showed lower responses to any other odour (see also the clear aldehyde \u201cresponse block\u201d in Figure 2 ). Figure 3 Generalisation Depending on Functional Groups (A) Data of the generalisation matrix (see Figure 2 ) represented as two-dimensional graphs for each conditioned odour. The right ordinate represents the CSs categorised in four functional groups, primary alcohols, secondary alcohols, aldehydes, and ketones (from top to bottom). The abscissa represents the test odours aligned in the same order as the conditioned odours (from left to right). The left ordinate represents the percentage of proboscis extensions to the test odours after being trained to a given odour. Each quadrant in the figure represents generalisation responses to one functional group after training for the same (grey quadrants) or to a different functional group (white quadrants). (B) Same data as in (A), but the observed responses within each quadrant were pooled and the resulting percentage of responses per quadrant was calculated. The abscissa and the right ordinate represent the four functional groups. The left ordinate represents the percentage of proboscis extensions to each of these groups after being trained to a given group. Grey bars correspond to grey quadrants in (A) and represent generalisation to the same functional group as the conditioned one. White bars correspond to white quadrants in (A) and represent generalisation to a functional group different from the conditioned one: 1-ol, 2-ol, al, and one mean primary alcohol, secondary alcohol, aldehyde, and ketone, respectively. Asterisks indicate significant differences along a row or a column ( p < 0.001) (C) Within-functional group generalisation, depending on chain length. The abscissa represents the functional groups tested. The ordinate represents the percentage of proboscis extensions to the functional groups tested after being trained to a given chain-length (lines). Thus, for instance, the first point to the left for C9 molecules (black circles) represents generalisation to 1-hexanol, 1-heptanol, and 1-octanol after conditioning to 1-nonanol. A significant heterogeneity was found in within-functional group generalisation for C8 and C9 but not for C6 and C7 molecules. (D) Generalisation within-functional groups. The figure shows results from pooling the data of (C) corresponding to each functional group. Each point shows the percentage of proboscis extensions to odours of the same functional group as the conditioned odour. Within-group generalisation was significantly heterogeneous (asterisks, p < 0.001). Pairwise comparisons showed that generalisation within aldehydes was significantly higher than within primary alcohols or ketones and marginally higher than within secondary alcohols (different letters indicate significant differences). We analysed within-functional group generalisation as depending on chain length (see Figure 3 C). To this end we represented generalisation from C6, C7, C8, and C9 molecules having a given functional group to the other compounds having the same functional group (e.g., Figure 3 C, black circle curve, first data point: generalisation to 1-hexanol, 1-heptanol, and 1-octanol after conditioning to 1-nonanol). A significant heterogeneity appeared for C8 and C9 molecules (\u03c7 2 = 12.60 and 14.30, respectively, p < 0.01 in both cases, n = 67\u201385) but not for C6 and C7 molecules ( p > 0.05). In the case of C8 and C9 molecules, generalisation was significantly higher within aldehydes ( p < 0.05). When comparing within-group generalisation over all four functional groups ( Figure 3 D), a significant heterogeneity appeared (\u03c7 2 = 14.40, df = 3, p < 0.01, n = 276\u2013316). Pairwise comparisons (using a corrected threshold for multiple comparisons: \u03b1\u2032 = 0.017) showed that generalisation within aldehydes was significantly higher than within primary alcohols (\u03c7 2 = 11.80, df = 1, p < 0.0006) and ketones (\u03c7 2 = 9.90, df = 1, p < 0.005) and close to significance in favour of aldehydes when compared to secondary alcohols (\u03c7 2 = 4.40, df = 1, 0.017 < p < 0.05). Generalisation within Chain Lengths Figure 4 A shows the generalisation responses of bees to odours having different (white quadrants) or the same (grey quadrants) chain length as the conditioned odour. In order to better visualise generalisation as depending on chain length, we pooled all the observed responses within each quadrant of Figure 4 A and calculated the resulting percentage of PER ( Figure 4 B). Grey bars correspond to generalisation to the same chain length; white bars correspond to generalisation to different chain lengths. Generalisation was highest in the case of odours with the same or similar chain length. Figure 4 Generalisation Depending on Chain Length (A) Data of the generalisation matrix (see Figure 2 ) represented as two-dimensional graphs for each conditioned odour. The right ordinate represents the CSs categorised in four chain lengths, C6, C7, C8, and C9 molecules (from top to bottom). The abscissa represents the test odours aligned in the same order as the conditioned odours (from left to right). The left ordinate represents the percentage of proboscis extensions to the test odours after being trained for a given odour. Each quadrant in the figure represents generalisation responses to one chain length after training for the same (grey quadrants) or to a different chain length (white quadrants). (B) Same data as in (A), but the observed responses within each quadrant were pooled and the resulting percentage of responses per quadrant was calculated. The abscissa and the right ordinate represent the four chain-length categories. The left ordinate represents the percentage of proboscis extensions to each of these categories after being trained for a given chain-length category. Grey bars correspond to grey quadrants in (A) and represent generalisation to the same chain length as the conditioned one. White bars correspond to white quadrants in (A) and represent generalisation to a chain length different from the conditioned one: C6, C7, C8, and C9 mean chain length of 6, 7, 8, and 9 carbons, respectively. Asterisks indicate significant differences along a row or a column ( p < 0.001). (C) Within chain-length generalisation as depending on functional group. The abscissa represents the chain lengths tested. The ordinate represents the percentage of proboscis extensions to the same chain length after being trained to a given functional group (lines). Thus, the first point to the left for ketones (red circles) represents generalisation to 1-hexanol, 2-hexanol, and hexanal after conditioning to 2-hexanone; the second point represents generalisation to 1-heptanol, 2-heptanol, and heptanal after conditioning to 2-heptanone. A significant heterogeneity was found in within-chain-length generalisation for aldehydes and ketones. (D) Generalisation within-chain lengths. The figure results from pooling the data of (C) corresponding to each chain length. Each point shows the percentage of proboscis extensions to odours of the same chain length as the conditioned odour. Within-chain-length generalisation was significantly heterogeneous (asterisks, p < 0.001). Pairwise comparisons showed that generalisation within C9 molecules was significantly higher than within C7 and C6 molecules and marginally higher than within C8 molecules (different letters indicate significant differences). We analysed within-chain length generalisation as depending on functional group ( Figure 4 C). To this end we represented generalisation from primary alcohols, secondary alcohols, aldehydes, or ketones of a given chain length to the other compounds having the same chain length (e.g., Figure 4 C, red circle curve, first data point: generalisation to 1-hexanol, 2-hexanol, and hexanal after conditioning to 2-hexanone). Generalisation within-chain length was generally higher for longer than for shorter chain lengths. This effect was significant for aldehydes (\u03c7 2 = 28.70, df = 3, p < 0.01, n = 75\u201380) but not for primary and secondary alcohols (\u03c7 2 = 5.20 and 3.4, df = 3, p > 0.05, n = 67\u201373 and n = 61\u201366, respectively). For ketones, a significant heterogeneity was found (\u03c7 2 = 10.00, df = 3, p < 0.05, n = 40\u201379), but generalisation was more important between C8 than between C7 molecules. The generalisation corresponding to other chain lengths fell in between. When comparing within-chain length generalisation over all four chain-length groups ( Figure 4 D, i.e., not considering functional group), a significant heterogeneity appeared \u03c7 2 = 23.2, df = 3, p < 0.001, n = 247\u2013293). Pairwise comparisons (using a corrected threshold for multiple comparisons: \u03b1\u2032 = 0.017) showed that within-chain length generalisation was significantly higher within C9 than within C6 (\u03c7 2 = 18.50, df = 1, p < 0.0001) and C7 molecules (\u03c7 2 = 15.00, df = 1, p < 0.0001). Generalisation within C8 molecules was close to significance when compared to generalisation within C9 molecules (\u03c7 2 = 5.00, df = 1, 0.017 < p < 0.05), and it was significantly higher than generalisation within C6 molecules (\u03c7 2 = 4.3, df = 1, 0.017 < p < 0.05). Generalisation between Functional Groups To analyse generalisation between groups, we took into account the responses to functional groups different from the conditioned one (see white bars in Figure 3 B). Bees showed heterogeneous patterns of generalisation (all vertical and horizontal comparisons in Figure 3 B were significant: \u03c7 2 > 37.70, df = 3, p < 0.001, in all eight cases). We found high between-group generalisation for primary and secondary alcohols: bees conditioned to secondary alcohols responded preferentially to primary alcohols, somewhat less to aldehydes, and even less to ketones (see Figures 3 A and 3 B, second row). A similar but less obvious response gradation was found for bees conditioned to primary alcohols Figures 3 A and 3 B, first row). In fact, the overall generalisation patterns were very similar for primary and secondary alcohols sharing the same chain length (see, for instance, the very close relationship between the two sets of blue [primary alcohol] and green curves [secondary alcohols] in Figure 4 A). As indicated before, bees conditioned to aldehydes generalised very little to odours belonging to other functional groups (see Figure 3 B, third row). Contrarily, bees conditioned to other functional groups highly generalised to aldehydes (see third column \u2018al' in Figure 3 B). This shows that generalisation between aldehydes and odours belonging to other functional groups was asymmetrical. The topic of asymmetric generalisation will be considered below in more detail. Generalisation between Chain Lengths To analyse generalisation between chain lengths, we took into account the responses to chain lengths that differed from the conditioned one (see white bars in Figure 4 B). In general, responses to molecules with different chain lengths followed a clear decreasing gradient, depending on the difference in the number of carbon atoms between the molecules considered (see Figure 4 B; all horizontal and vertical comparisons were significant, \u03c7 2 > 16.3, df = 3, p < 0.001 in all eight cases). For instance, when conditioned to a C9 molecule (see Figure 4 B, fourth row), bees responded in 53%, 31%, and 23% of the cases to C8, C7, and C6 molecules, respectively, while they responded to C9 molecules in 67% of the cases. This gradient was also evident when generalisation took place between functional groups: for instance, after training with 2-nonanol (see Figure 3 A, second row), the response of bees to odours of different functional groups (solid lines in white boxes) always followed a similar decreasing tendency with the same (C9) or similar (C8) chain length on top. Asymmetry in Olfactory Generalisation As previously mentioned, some groups like aldehydes induced asymmetrical cross-generalisation (i.e., bees responded less to other functional groups after training for aldehydes than to aldehydes after training for other functional groups). We analysed this asymmetrical generalisation and built an asymmetry matrix ( Figure 5 A). To this end, we calculated for each odour pair (A and B) the difference (in percentage) between generalisation from A to B and generalisation from B to A. Such differences were ranked in 10% categories from \u221255% to 55%. White boxes indicate no asymmetries. Blue shades in Figure 5 A indicate that cross-generalisation was biased towards odour A (i.e., conditioning to A resulted in lower generalisation to B while conditioning to B resulted in higher generalisation to A); red shades indicate that cross-generalisation was biased towards odour B (i.e., conditioning to A resulted in higher generalisation to B while conditioning to B resulted in lower generalisation to A). This representation showed that some odours induced generalisation while other odours diminished it. For instance, hexanal was well learnt but induced low generalisation to other odours, except to other aldehydes. On the other hand, bees conditioned to other odours very often generalised to hexanal. Thus, a clear blue row (or a red column) corresponds to hexanal in the asymmetry matrix. Conversely, 2-hexanone induced high generalisation to other odours but received few responses as a test odour. Thus a red row (or a blue column) corresponds to 2-hexanone in the asymmetry matrix. Most odours, however, showed little or no asymmetry. Figure 5 B presents the mean asymmetry found for each training odour. In six cases, the mean asymmetry deviated significantly from zero, which represents a theoretically perfect symmetry ( t -test). Two odours (red bars) significantly induced generalisation (2-hexanone and 2-hexanol, t -test, df = 14, p < 0.001 and p < 0.01, respectively), while four odours (blue bars) diminished it significantly (hexanal, heptanal, and octanal, and 2-nonanone, t -test, df = 14, p < 0.001 for the former and p < 0.01 for the three latter odours). Figure 5 Asymmetric Generalisation between Odours (A) The asymmetry matrix depicts asymmetric cross-generalisation between odours. For each odour pair (A and B), the difference (percentage) between generalisation from A to B and generalisation from B to A was calculated. Such differences were ranked in 10% categories varying from blue (\u221255%) to red (55%). Blue shades indicate that cross-generalisation was biased towards odour A (i.e., conditioning to A resulted in lower generalisation to B, while conditioning to B resulted in higher generalisation to A); red shades indicate that cross-generalisation was biased towards odour B (i.e., conditioning to A resulted in higher generalisation to B, while conditioning to B resulted in lower generalisation to A). For this reason, each odour pair (A and B) appears twice in the matrix, once in the upper-left of the black diagonal line, and once in the lower-right of the black diagonal line, with opposite values. See, for example, the two cells outlined in green for the pair 2-hexanone/2-octanol. (B) Mean generalisation induced or diminished by each odour A in (A). Each bar represents the mean asymmetry of the respective horizontal line in the asymmetry matrix. Red bars show that an odour induced more generalisation than it received, while blue bars show the opposite. Significant generalisation asymmetries were found in six out of 16 cases (**, p < 0.01; ***, p < 0.001). Olfactory Space In order to define a putative olfactory space for the honeybee, we performed a principal component analysis (PCA) on our data to represent in a limited number of dimensions the relative relationships between odorants in a 16-dimension perceptual space ( Figure 6 A). The first three factors represented 31%, 29%, and 15% of overall variance in the data (total of the first three factors: 75%). The analysis showed a clear organisation of odours depending on their chemical characteristics. First, chain length was very clearly represented by the first factor (see upper-right graph in Figure 6 A), from C6 to C9 molecules from the right to the left. On the other hand, the chemical group was mostly represented by factors 2 and 3. Whereas factor 2 separated mostly aldehydes from alcohols, with ketones falling between them, factor 3 segregated ketones from all other odours (lower-right graph, Figure 6 A). None of these factors separated primary and secondary alcohols. This analysis indicates that the chemical features of molecules (chain length and functional group), which are sometimes thought of as artificial perceptual (psychophysical) dimensions determined by experimenters [ 34 ] can be considered as true inner dimensions of the bees' perceptual space. Cluster analyses performed on the data segregated odours mostly according to their chain length. In the first group ( Figure 6 B, upper part), we found two subgroups, short-chain alcohols (C6 and C7, primary and secondary alcohols) and short-chain ketones (C6 to C8). On the other hand ( Figure 6 B, lower part), three clear subgroups were formed: short-chain aldehydes (C6 and C7), long-chain alcohols (C8 and C9, primary and secondary alcohols), and a last group with long-chain aldehydes (C8 and C9) and 2-nonanone. Very similar results were obtained using Euclidian or city-block metrics. Figure 6 A Putative Honeybee Olfactory Space (A) Left: The olfactory space is defined on the basis of the three principal factors that accounted for 76% of overall data variance after a PCA performed to represent the relative relationships between odorants. Primary alcohols are indicated in blue, secondary alcohols in green, aldehydes in black, and ketones in red. Different chain-lengths are indicated as C6, C7, C8, and C9, which corresponds to their number of carbon atoms. For each functional group, arrows follow the increasing order of carbon-chain lengths. Right: Chain length was very clearly represented by factor 1. C6 to C9 molecules are ordered from right to left. The chemical group was mostly represented by factors 2 and 3. Whereas factor 2 separated mostly aldehydes from alcohols, with ketones falling between them, factor 3 separated ketones from all other odours. None of these three factors separated primary and secondary alcohols. (B) Euclidean cluster analysis. The analysis separated odours mostly according to their chain length. Linkage distance is correlated to odour distances in the whole 16-dimension space. The farther to the right two odours/odour groups are connected, the higher the perceptual distance between them (odour colour codes are the same as in [A]). Correlation between Optophysiological and Behavioural Measures of Odour Similarity We asked whether optophysiological measures of odour similarity, obtained using calcium imaging techniques at the level of the honeybee AL [ 22 , 23 , 24 , 35 ], correspond to perceptual odour similarity measures as defined in our putative honeybee olfactory space. We thus calculated the Euclidian distance between odour representations in our 16-dimension \u201cbehavioural\u201d space for all odour pairs (120 pairs). We then calculated distances between odours in optical imaging experiments, using the odour maps by Sachse et al. [ 23 ]. A correlation analysis was performed between both datasets. This analysis was possible because both the study by Sachse et al.[ 23 ] and our study used the same set of odours delivered under the same conditions. Figure 7 A presents the correlation obtained, including all 120 odour pairs. Both sets of data were highly significantly correlated ( r = 0.54, t 118 = 7.43, p < 2.10 \u201310 ), a result that shows that odours, which were found to be physiologically similar in the optical imaging study, were also evaluated as similar in behavioural terms. Note, however, that data points cluster quite broadly around the main trend line, showing that many exceptions were found. In order to use a more exact measure of physiological odour similarity, we used the correlation results between primary and secondary alcohol maps provided by Sachse et al. [ 23 ]. By correlating this more exact value of physiological similarity with our behavioural data, we also found a highly significant relationship between physiological and behavioural data ( Figure 7 B; r = 0.82, t 26 = 7.83, p < 7.10 \u20138 ). The correlation coefficient achieved with this second method was significantly higher than that achieved with the first method ( Z = 2.52, p < 0.05). A better fit between the two datasets was thus found, although outliers were still present in the data. These two analyses show that optophysiological and behavioural measures of odour similarity correlate well using the methods described here. Thus, in the case of the honeybee, olfactory neural activity corresponds to olfactory perception. Figure 7 Correspondence between Perceptual and Physiological Odour Similarity (A) Correlation between optophysiological measures of odour similarity (carried out using calcium imaging recordings [ 23 ]) and our behavioural measures of odour similarity. Euclidian distance between odour representations in our 16-dimension \u201cbehavioural\u201d space for all odour pairs (120 pairs, x axes) and distances between odours in optical imaging experiments, using the odour category maps displayed by Sachse et al. [ 23 ] (also 120 pairs, y axes) were calculated. This correlation, including all 120 odour pairs, was highly significant ( r = 0.54, p < 0.001). Odours found to be similar in the optical imaging study were also similar in the behaviour. Data points cluster quite broadly around the main trend line, showing that many exceptions were found. (B) Correlation between measures of optophysiological similarity carried out using the optical imaging technique [ 23 ] and our behavioural measure of odour similarity. Using the exact data given for primary and secondary alcohols [ 23 ], a much better correlation between the two datasets was achieved than in (A) ( r = 0.82, p < 0.001), although outliers were still found in the data. Discussion In the present work, we have studied perceptual similarity among odorants in the honeybee, using an appetitive-conditioning paradigm, the olfactory conditioning of the PER [ 17 , 18 ]. We showed that all odorants presented could be learned, although acquisition was lower for short-chain ketones. Generalisation varied, depending both on the functional group and on the carbon-chain length of odours trained. Generalisation was very high among primary and secondary alcohols, being high from ketones to alcohols and aldehydes and low from aldehydes to all other tested odours; thus, in some cases, cross-generalisation between odorants was asymmetric. Some odours, like short-chain ketones or aldehydes, induced more asymmetries than other odours. Higher generalisation was found between long-chain than between short-chain molecules. Functional group and carbon-chain length constitute orthogonal inner dimensions of a putative olfactory space of honeybees. Perceptual distances in such a space correlate well with physiological distances determined from optophysiological recordings performed at the level of the primary olfactory centre, the AL [ 23 ] such that olfactory neural activity corresponds to olfactory perception. Previous studies have attempted to describe olfactory generalisation in honeybees and to study structure\u2013activity relationships [ 19 , 20 , 36 , 37 , 38 ]. These studies generally supported the view that generalisation mainly happens when odours belong to the same chemical group. Moreover, they also suggested that the rules underlying olfactory learning and perception of different chemical classes [ 20 ] or of particular odorants (e.g., citral [ 20 , 37 ]) may vary. However, these studies used differential training, thus inducing several generalisation gradients (excitatory and inhibitory) that make the interpretation of generalisation responses difficult [ 21 , 36 ]. Furthermore, these studies were carried out on a rather discrete number of odour pairs [ 37 ], did not detail the results obtained with individual odour combinations [ 20 ], or used a very reduced number of bees per conditioned odour ([ 21 ]; two bees per odorant).Thus, the present study is the first one to provide (i) generalisation data based on absolute conditioning (i.e., only one odour conditioned at a time), (ii) a systematical test of all odour combinations, (iii) robust sample sizes for each experimental situation, and (iv) important generalisation gradients. These are in our view crucial prerequisites to describe odour perception and similarity in a precise way. Chemical Group and Chain Length Several studies in other species have shown the importance of functional group and carbon-chain length of the odour molecules for behavioural responses to odours. Differences in the response between molecules of diverse aliphatic and aromatic homologue odour classes (i.e., differing in functional group, chain length, and overall molecule form) were investigated in moths [ 39 , 40 ], cockroaches [ 41 ], rats [ 42 ], squirrel monkeys [ 4 , 43 ] and humans [ 38 , 44 , 45 ]. These studies show that both functional group and chain length affect the perceived quality of an odorant. Concerning chain length, the greater the difference in the number of carbons between odours, the easier the discrimination and the lower the generalisation ([ 21 , 40 , 42 , 44 ] and present study). In our study, both chemical group and chain length of odour molecules determined the bees' generalisation responses. Bees mostly generalised to other odours when these shared the same functional group. This effect was observed for all functional groups (see Figure 3 B) but was strongest for aldehydes. Other studies have found that aldehydes induced high within-group generalisation [ 20 , 21 , 36 ]. Thus, aldehydes may represent a behaviourally relevant chemical class for honeybees. Between-functional group generalisation depended on the functional group considered. It was high between primary and secondary alcohols, which appear therefore perceptually similar to the bees, and low between other chemical groups. Bees clearly generalised between odours that shared the same chain length. Increasing chain length promoted generalisation. Moreover, generalisation to other chain lengths decreased if the difference in the number of carbons between odours increased. This suggests a perceptual continuum between different chain lengths (but see below). Thus, the chemical structure of the odorants is critical for determining the amount of generalisation. A Putative Olfactory Space for the Honeybee We found that the two controlled physical characteristics of odour molecules used in this study, functional group and chain length, correspond to internal dimensions in the bees' olfactory perceptual space such as the three most important factors extracted in our PCA analysis, one mainly represented chain length and the other two were mostly influenced by functional group. Cluster analyses allowed separating odours in clusters according to their functional groups and their chain length. Interestingly, C6 and C7 molecules and C8 and C9 molecules were mainly grouped together, so that, for instance, all short-chain primary and secondary alcohols were grouped on one side, and all long-chain alcohols on the other side. The same happened for aldehydes, and in a different way for ketones (C9 separated from the rest). This discrepancy suggests that, although chain length appears mostly as a perceptual continuum in the PCA analysis, there may be a perceptual \u201cjump\u201d between short-chain and long-chain molecules. Neural Bases of Odour Perception Both in vertebrates and in invertebrates, studies quantifying the neural responses to structurally similar odours in the first relay of the olfactory pathway have been performed (olfactory bulb: e.g., [ 46 , 47 , 48 , 49 ]; AL: [ 23 , 50 ]). These studies show that activity patterns are more similar when the difference in the number of carbons between molecules is small. It was hypothesised that such a physiological similarity is the basis for olfactory discrimination and generalisation as measured behaviourally. This has indeed been reported for mucosal activity in mice [ 51 ], electrical mitral cell activity [ 42 ], and/or radiolabelled 2-deoxyglucose uptake in the rat olfactory bulb [ 32 ]. Also, in Manduca sexta, qualitative similarities were observed between the degree of behavioural generalisation according to chain length [ 40 ] and the degree of overlap between electrophysiological temporal patterns of activity across AL neurons [ 50 ]. Several correspondences, but also discrepancies, can be found between our behavioural results and the physiological results obtained at the level of the bee AL [ 23 ]. First, within the regions of the AL accessible to optical imaging (about 25% of the glomeruli), patterns of glomerular activity for different odours are highly dependent on chain length, but much less so on chemical group. Thus, most active glomeruli respond to several functional groups as long as the chain length corresponds, but respond differentially to different chain lengths. Glomeruli T1\u201328 and T1\u201352 are specialised in short-chain molecules (respectively C5\u2013C7 and C6\u2013C7), whilst glomeruli T1\u201333 and T1\u201317 are specialised in long-chain molecules (respectively C7\u2013C9 and C8\u2013C9). These glomeruli also respond to most functional groups but in a graded way. For instance, glomerulus T1\u201317 responds more to alcohols in the intermediate range than to aldehydes or ketones, whereas T1\u201352 generally responds more to ketones in the short range, more to aldehydes in the long range, and overall little to alcohols. No individual glomerulus was found that responds specifically to a chemical group. However, it should be kept in mind that some regions of the ALs are not yet accessible to calcium imaging techniques (about 75% of the lobe; see below). Thus, a possible explanation is that glomeruli responding to specific chemical groups (or with responses more dependent on chemical groups than on chain length) were not imaged. Second, primary and secondary alcohols induce extremely similar activation patterns in the AL, but subtle differences could be found, so that for a given chain length, the representation of a secondary alcohol was between that of the primary alcohol of the same chain length and that with one less carbon atom (see Figure 6 B in Sachse et al. [ 23 ]). We found a similar arrangement of alcohol representations, with primary and secondary alcohols alternating on a common axis (see Figure 6 A). Third, optical imaging data showed that higher chain lengths support more similarity between patterns (see Figure 6 C in Sachse et al. [ 23 ]). Our finding that longer chain lengths induce more generalisation agrees with the imaging data. These last two points suggest that the general rules governing odour similarity at the neural and the behavioural level are similar. The Correspondence between Perceptual and Physiological Odour Similarity We aimed at comparing behavioural and physiological data in a more precise way, using correlation analyses between our behavioural similarity matrix, in which distances between two odour points represent psychological distances between stimuli, and a physiological similarity matrix obtained from optophysiological recordings of glomerular activation patterns [ 23 ]. Comparing distances between odours in these two matrixes resulted in a good correlation. This means that glomerular activity patterns recorded in the brain could predict behavioural responses and vice versa. The optophysiological dataset of Sachse et al. [ 23 ] has nevertheless some limitations with respect to the objectives of our work: (i) bath application measurements of AL activity using calcium green as a dye [ 23 ] record the combined activity of several neuronal populations of the AL, among which primary-afferent activity seems to have the most important contribution [ 52 ]; (ii) such measurements survey only the dorsal part of the AL, which constitutes 25% of the neuropile studied; and (iii) learning alters odour representations in the AL [ 35 , 53 , 54 ] such that there could be a mismatch between our data collected after olfactory conditioning and the dataset of Sachse et al. [ 23 ], which was obtained from naive bees. With respect to the first point, it could be argued that the AL circuitry transforms the primary-afferent representations of odours [ 25 ] such that recordings where primary-afferent receptor activity is predominant are not very useful for evaluating optophysiological similarity. However the very fact that we found a significant correlation between our behavioural data and the imaging data by Sachse et al. [ 23 ], strongly suggests that the perceptual quality of odorants mostly appears at the peripheral level. Clearly, this correlation was not perfect, and odour quality is most probably refined by further processing within the AL, and/or at higher stages of the olfactory pathway, such as in the mushroom bodies or the lateral protocerebrum. In honeybees, new methods have been developed, which allow recording selectively the activity of the efferent PNs [ 25 ]. However, the two studies published using this method [ 25 , 26 ] do not provide an extensive odorant matrix as that provided by Sachse et al. [ 23 ]. In this sense the study on which we based our correlation analysis is certainly the only one of its kind published to date. However, in the future, a careful comparison of our behavioural data with both bath-applied imaging data emphasising receptor neuron input (as done here) and selective imaging of PNs would be extremely helpful in understanding to what extent AL processing shapes odour perceptual quality. With respect to the second point, calcium imaging recordings of AL activity are certainly limited to the dorsal part of the AL, which is the region accessible when the head capsule is opened in order to expose the brain for recordings. This is an inherent limitation of the method that the use of two-photon microscopy during calcium imaging measurements will soon allow us to overcome, as shown already by recordings obtained in the fruit fly Drosophila melanogaster [ 55 ]. Finally, with respect to the third point, it is known that learning alters odour representations in the AL, when bees are trained in a differential conditioning procedure, with one odour rewarded and another odour unrewarded [ 53 ]. This is not the conditioning procedure used in our work, which was absolute (only one odour rewarded at a time). In the bee, changes in the olfactory code due to absolute conditioning seem to be difficult to detect (C. G. Galizia, personal communication), such that this point may not be so critical for our correlation analysis. In any case, if there are changes in odour representations due to conditioning, recording glomerular activity patterns after conditioning would only improve our correlation analyses. Generalisation Asymmetries between Odours We have found a number of asymmetries in olfactory cross-generalisation, with bees responding more to odour B after learning odour A than in the reverse situation. Previous studies have observed such a phenomenon, but it was mostly related to olfactory compounds with pheromonal value (aggregation pheromone citral [ 20 , 37 ] and alarm pheromones 2-heptanone and isoamyl acetate [ 56 ]). In the present study, we found that six out of the 16 odours used induced significant generalisation asymmetries over the whole matrix; none of these six odours was related to any known pheromone (see Table 1 ). Generalisation asymmetries seem to be a general feature of honeybee olfaction. Table 1 Chemical and Biological Characteristics of the Odours Used The odours were listed by functional groups (primary alcohols, secondary alcohols, aldehydes, and ketones) and purity. Odour vapour pressure values (VP), pheromone characteristics and occurrence in floral scents (after Knudsen et al. [ 66 ]) are also given a Notation: *1, releases altering at hive entrance and stinging, repels clustering bees, inhibits scenting, repels foragers (sting chamber); *2, releases altering at hive entrance, inhibits foraging activity, repels foragers (sting chamber); *3, repels at hive entrance, releases stinging, encourages foraging activity (sting chamber); *4, releases stinging, inhibits foraging activity, repels foragers (mandibular glands) Odour concentration can affect stimulus salience. In our work, generalisation asymmetries could not be directly explained by differences in odour concentration (through differences in vapour pressure), because, for instance, the two odours with the highest vapour pressure in our sample (2-hexanone and hexanal) produced totally opposite results: 2-hexanone induced important generalisation, while hexanal strongly reduced generalisation. Also, although we used 16 different odours with a range of different vapour pressures, we found that acquisition was very similar for most odours, except for the short-chain ketones, which were less easily learned. This suggests that almost all odours used had a good salience for bees. Wright and Smith [ 57 ] studied the effect of odour concentration in generalisation in honeybees. They found that discrimination increased with concentration for structurally dissimilar odours but not for similar odours. Further experiments using odorants at different concentrations should be carried out to determine the effect of odour concentration on generalisation asymmetries. Generalisation asymmetries could be due to innate or experience-dependent differences in the salience of odours for honeybees, such that more salient odours would induce higher generalisation than less salient odours. This interpretation implies that most aldehydes (hexanal, heptanal, and octanal) are highly salient odours for honeybees, because aldehydes showed a clear \u201cfunctional group\u201d effect, which could reveal a certain bias of the olfactory system towards these odours. Ketones, on the other hand, showed a heterogeneous effect, as 2-hexanone seemed to have a low salience (it was not well learnt) and induced a high generalisation to other odours, while 2-nonanone consistently reduced generalisation to other odours. In the group of alcohols, only 2-hexanol induced generalisation to other odours. Therefore, only aldehydes showed a clear group effect on generalisation asymmetry. This effect could be due to innate odour preferences [ 58 , 59 ] or to previous odour exposure within the hive [ 60 , 61 ]. Innate odour preferences could be related to natural, floral odours that were more consistently associated with food resources [ 20 , 62 ]. It is thus important to investigate whether or not such ecological trends exist in the natural flora associated with the honeybee and whether or not other bee species also present such clear biases, in particular towards aldehydes. Conversely, asymmetries could be the result of the conditioning procedure. This would be the case if conditioning modifies odour representation in an asymmetric way. Indeed, experience-induced modifications of odour representations have been found at the level of the honeybee AL. Thus, odour-evoked calcium signals in the AL can be modified by elemental [ 53 ] and nonelemental olfactory learning paradigms [ 35 ] such that the representations of odours that have to be discriminated become more distinct and uncorrelated as a result of learning. In the fruit fly D. melanogaster , new glomeruli become active after olfactory learning [ 54 ], while in the moth M. sexta new neuronal units in the AL are recruited after olfactory learning [ 63 ]. These elements suggest that modifications of odour representation after learning two different odours could indeed be asymmetrical: if, for instance, the neuronal representation of A after conditioning becomes A\u2032, which is slightly farther away from B than A in the bee's olfactory space, and if the perceptual representation of B becomes B\u2032 after conditioning, which is closer to A than B, then bees would show less generalisation in behavioural tests from A to B than from B to A. On the level of the AL network, glomeruli are connected via lateral inhibitory interneurons [ 25 , 64 , 65 ]. Due to this, glomerular activation by an odour A will transiently inactivate parts of the network and possibly parts encoding a subsequent odour B. Optical imaging experiments have shown that inhibition between glomeruli may be asymmetric [ 25 ]. In our case, glomeruli activated by odour A may inhibit glomeruli coding for odour B, while glomeruli coding for odour B may not inhibit those coding for odour A. In this hypothesis, asymmetric cross-generalisation could reflect a sensory phenomenon. Nevertheless, we believe that inhibitions at the level of the AL are rather short-lived such that a purely sensory priming effect seems improbable. If, however, the strength of lateral inhibitions between glomeruli can be modified by learning as proposed by Linster and Smith [ 65 ], then asymmetrical generalisation would come from the fact that inhibitory lateral connections are modified. In order to determine the physiological mechanisms underlying asymmetrical cross-generalisation and the possible role of AL networks in it, future work will aim at visualising the evolution of glomerular activity patterns during and after olfactory conditioning with odours that showed asymmetries in our study. Conclusion We have shown that the two odorant physical dimensions that varied in our study, functional group and chain-length, correspond to internal dimensions of the bees' olfactory space. Generalisation was mainly due to these two characteristics with generalisation within functional group being more important. Such generalisation was particularly high for aldehydes, a fact that suggests that these odours may have an intrinsic value for bees. Generalisation between functional groups was mostly found between primary and secondary alcohols. Furthermore, a gradient in generalisation was found with respect to chain length. Asymmetric cross-generalisation was found in the case of certain odorants. Such asymmetries were neither strictly linked to chain length nor to functional group, but depended on particular odorants. The 16 odours used in our work represent a small part of the odorants that bees may encounter in nature (see Knudsen et al. [ 66 ]). For a complete description of the bees' olfactory perceptual space, more odours having other molecular features have to be studied. New dimensions in the bees' perceptual space could then be found. Finally, and most important, the perceptual distance between odours can be predicted on the basis of the differences in the patterns of glomerular activation in the first relay of the olfactory pathway: the AL, and vice versa. This emphasises the relevance of studying activity patterns in the brain in imaging studies and trying to relate them to perceptual tasks. Our work shows that this objective, which is at the core of cognitive neurosciences, can be achieved using an invertebrate model such as the honeybee. Materials and Methods Insects Every experimental day, honeybees were captured at the entrance of an outdoor hive and were cooled on ice for 5 min until they stopped moving. Then they were harnessed in small metal tubes in such a way that only the head protruded. The mouthparts and the antennae could move freely. Harnessed bees were left for 3 h in a resting room without disturbance. Fifteen minutes before starting the experiments, each subject was checked for intact PER by lightly touching one antenna with a toothpick imbibed with 50% (w/w) sucrose solution without subsequent feeding. Extension of the proboscis beyond the virtual line between the open mandibles was counted as PER. Animals that did not show the reflex were not used in the experiments. Stimulation apparatus The odours were delivered by an odour cannon, which allowed the presentation of up to seven different odours, and a clean airstream [ 67 ]. Each odour was applied to a filter paper placed within a syringe (see below) that was connected to the cannon. An airstream was produced by an air pump (Rena Air 400, Annecy, France) and directed to the relevant syringes with electronic valves (Lee Company, Voisins-le-Bretonneux, France) controlled by the experimenter via a computer. In the absence of odour stimulation, the airstream passed through a syringe containing a clean filter paper piece (clean airstream). During odour stimulation, the airstream was directed to a syringe containing a filter paper loaded with odour. After a 4-s stimulation, the airstream was redirected to the odourless syringe until the next stimulation. Stimuli Sixteen odours (Sigma Aldrich, Deisenhofen, Germany) were used in our work as CS and test stimuli (see Table 1 ). Racemic mixtures were used in the case of molecules that had chiral carbons. These odours are present in flowers and some in pheromones (see Table 1 ). Pure odorants (4 \u03bcl) were applied to 1-cm 2 filter paper pieces, which were transferred to 1-ml syringes, cut to 0.7 ml to make them fit into the odour cannon. Fifty percent sugar solution was used throughout as US. Experimental design Our work was designed to obtain a generalisation matrix with 16 different odours. Ideally, after conditioning each of the 16 odours as CS, the response to each odour (including the CS) should be measured (i.e., 16 \u00d7 16 = 256 cells). However, testing 16 odours implies presenting them without reward, a situation that may result in extinction of the learned response due to the repeated unrewarded odour presentations. Preliminary experiments were performed in which four groups of 180 bees were trained along three trials to 1-hexanol, 2-octanol, linalool, and limonene, respectively. Training was followed by tests with the four different odours, including the conditioned one. These experiments showed that after three conditioning trials, the response of the bees to the CS in the four tests remained at the same level, independently of the order of occurrence of the CS such that it was not influenced by extinction. We thus kept this protocol for the 16 \u00d7 16 matrix. Each of the 2,048 bees used in this study was thus subjected to three conditioning trials with their respective CS, and to four test trials, each with a different odour chosen among the 16 possible odours. Intertrial intervals of 10 min were used throughout. A randomisation schedule (detailed below) was developed for the test phase to reduce any possible day- and odour-combination effects. Conditioning trials One bee at a time was placed into the conditioning setup. The total duration of each trial was 37.5 s After 15 s of familiarisation to the experimental context, the CS was presented to the bee for 4 s. Three sec after onset of the CS, the antennae were stimulated with the US, leading to a proboscis extension. The bee was allowed to feed for 3 s. Stimulus overlap was 1 s (interstimulus interval, 3 s). The bee was left in the conditioning place for 17.5 s and then removed. Test trials The procedure was similar to that for conditioning trials but no US was given after odour delivery. After the four test trials, PER to the US was checked once again. Animals unable to show PER at this point were not considered for the analyses. Overall, less than 2% of the bees died during the experiment, and less than 1% of the survivors showed no US reaction at the end of the tests. Randomisation schedule On each day, two to three experimenters worked in parallel, each training 16 bees at a time. In the training phase, the 16 bees were divided into four groups of four bees, and each group was trained to one of the 16 different odours. In the test phase, four out of 16 odours were presented to each of the 16 bees. The combination of four odours tested together changed in each experiment, so that any effect of having particular odours in the same test combination was suppressed. The whole experiment was planned in such a way that in any of our experimental groups, two given odours appeared at least once, but a maximum of three times together in a test sequence. This was possible by carefully picking out eight of the 16! (2.1 \u00d7 10 13 ) possible experimental plans. Additionally, within each group, the testing order for the four test odours was determined randomly. Data analysis and statistics During the experiments, we recorded the response to the presented odour, that is, whether bees extended their proboscis after the onset of the odour and before the presentation of the sucrose solution in the case of reinforced trials, such that the anticipatory response recorded was due to the odour and not to the US. Multiple responses during a CS were counted as a single PER. The percentages of PER recorded during acquisition were used to plot acquisition curves (see Figure 1 ). To test whether bees learnt the different odours in a similar way, ANOVAs for repeated measurements were used both for between-group and for within-group comparisons. Monte Carlo studies have shown that it is permissible to use ANOVA on dichotomous data only under controlled conditions [ 68 ], which are met by the experiments reported in this study: equal cell frequencies and at least 40 df of the error term. The \u03b1 level was set to 0.05 (two-tailed). To ensure that we analysed a true generalisation response in the tests, and hence built a true generalisation matrix, we kept only those bees which had actually learnt the CS (71% of the bees used in this work). We therefore performed new analyses that only included those bees that responded to the CS before the presentation of the US in the third conditioning trial. A lack of response to an odour in the tests could be due either to the fact that the bees had not made any association between CS and US or because their motivational level was low. For all odours tested, we observed that responses to the CS in the third conditioning trial were equivalent to responses to the CS in the tests (McNemar test; see Results). We represented the responses of the selected bees to the test odours (see Figure 2 ). As the numbers of bees were now heterogeneous in the different groups, we could not use ANOVAs to analyse the responses in the tests (see above). We thus used \u03c7 2 tests for all further between-group comparisons. In the case of multiple two-by-two comparisons, the significance threshold was corrected using the Dunn\u2013Sidak correction [\u03b1\u2032 = 1 \u2212 (1 \u2212 \u03b1) 1/k where k is the number of two-by-two comparisons in which each dataset is used] in order to reduce the type I errors. Alpha values between \u03b1\u2032 and 0.05 were considered as near significant. Olfactory space To observe the relationships between odours in a reduced number of dimensions, we performed a PCA, which identified orthogonal axes (factors) of maximum variance in the data, and thus projected the data into a lower-dimensionality space formed of a subset of the highest-variance components. We calculated the three factors, which accounted for most of the observed variance. Calculating distances between odours in the resulting putative olfactory space allowed the evaluation of their perceptual similarity, not only based on direct generalisation between these odours (i.e., generalisation from odour A to odour B and vice versa), but also including responses to these odours after conditioning to other odours (e.g., C, D, E, etc.). We performed cluster analyses to group odours, according to their respective distance in the olfactory space, using both Euclidian and city-block metrics, with Ward's classification method. Both metrics gave very similar results, so we later used only Euclidian metrics. Euclidian (i.e., direct) distances in the 16-dimensional space are defined as with i and j indicating odours, p the number of dimensions\u2014that is, conditioning groups\u2014and X ik the response of bees to odour i after conditioning to odour k. These distances were used in correlation analyses with optical imaging data (see below). Correlation analysis between perceptual and optophysiological similarity measures We studied whether or not physiological similarity between odours as determined by optical imaging studies of AL activity [ 22 , 23 , 35 ] actually reflects perceptual odour similarity for the bees. To this end, we performed correlation analyses between published optical imaging data that were obtained using the same set of odours as in our work [ 23 ] and our behavioural data. We used two sets of physiological data. First, to perform such a correlation on the whole dataset (including all 16 odours), we transcribed the activation maps presented by Sachse et al. [ 23 ] (see Figure 7 ) into activation levels for each glomerulus from zero to three, according to the following signal scale: dark blue (0%\u201320%) and light blue (>20%\u201340% activity), zero; green (>40%\u201360% activity), one; yellow (>60%\u201380% activity), two; and red (>80% activity), three. As the activity under 40% was less accurately separated from noise, activation levels between 0% and 40% were ranked as 0. Scaling the physiological data in this way instead of using the original imaging activation data, gave a good overview of physiological similarity between odours for imaging data ( see Results ). To provide a more precise correlation analysis between behavioural and imaging data, albeit on a more limited odour dataset (eight odours), we used exact correlation data ([ 23 ], Table 1 ). Each correlation value C, as calculated by Sachse et al. [ 23 ] between activity patterns for all pairs of primary and secondary alcohols, was converted into physiological distances by the operation 100 \u2212 C. All linear correlations were assessed by calculating Pearson's r, and using Student's t -test. Comparison between correlation coefficients obtained with the two methods was carried out statistically using a Z test as in [ 69 ]."}
{"PMCID": "PMC1044832", "title": "Recent Origin and Cultural Reversion of a Hunter\u2013Gatherer Group", "abstract": "Contemporary hunter\u2013gatherer groups are often thought to serve as models of an ancient lifestyle that was typical of human populations prior to the development of agriculture. Patterns of genetic variation in hunter\u2013gatherer groups such as the !Kung and African Pygmies are consistent with this view, as they exhibit low genetic diversity coupled with high frequencies of divergent mtDNA types not found in surrounding agricultural groups, suggesting long-term isolation and small population sizes. We report here genetic evidence concerning the origins of the Mlabri, an enigmatic hunter\u2013gatherer group from northern Thailand. The Mlabri have no mtDNA diversity, and the genetic diversity at Y-chromosome and autosomal loci are also extraordinarily reduced in the Mlabri. Genetic, linguistic, and cultural data all suggest that the Mlabri were recently founded, 500\u2013800 y ago, from a very small number of individuals. Moreover, the Mlabri appear to have originated from an agricultural group and then adopted a hunting\u2013gathering subsistence mode. This example of cultural reversion from agriculture to a hunting\u2013gathering lifestyle indicates that contemporary hunter\u2013gatherer groups do not necessarily reflect a pre-agricultural lifestyle.", "fulltext": "Introduction The Mlabri are an enigmatic group of about 300 people who nowadays range across the Nan, Phrae, and Phayao provinces of north and northeastern Thailand and the Sayaburi province of western Laos [ 1 , 2 ]. Their traditional lifestyle is to move frequently through the dense forests of the high mountains, building temporary structures of bamboo sticks thatched with banana leaves, which they occupy for a few days, until the leaves turn yellow (thus accounting for their traditional Thai name, Phi Tong Luang, which means \u201cspirit of the yellow leaves\u201d). First contacted by Europeans in 1936 [ 3 ], they are unique among the hill tribes of northern Thailand in that, until recently, they subsisted by hunting and gathering combined with occasional barter trade with villagers. The origins of the Mlabri are controversial. Some investigators have assumed that there is a direct connection between the Mlabri and the ancient Hoabinhian hunting\u2013gathering culture of Southeast Asia [ 1 ]. However, a limited investigation of blood group variation [ 4 ] raised the possibility that the Mlabri originated via a founder event from an agricultural group, and preliminary linguistic analyses support this idea. The Mlabri language seems lexically most closely related to Khmu and Tin, two languages of the Khmuic branch of the Mon-Khmer sub-family of Austro-Asiatic languages, both of which are spoken in agricultural highland villages [ 5 ]. The cluster of dialects jointly referred to as Tin, or Mal/Prai, [ 6 ] is spoken in the Thailand\u2013Laos border region that the Mlabri also occupy, whereas Khmu is spoken over a much wider area [ 7 ]. The grammar of Mlabri additionally has features that deviate markedly from typical Mon-Khmer, suggesting that Mlabri developed as a result of contact between speakers of a Khmuic language and speakers of a quite different language of unknown affiliation [ 2 , 8 ]. We report here the results of an investigation of genetic diversity in the Mlabri, to see whether patterns of genetic variation might provide further insights into the question of an agricultural versus hunting\u2013gathering origin for the Mlabri. The rationale for using genetic analyses to investigate this question is that previous work has shown that hunter\u2013gatherer groups typically differ from their agricultural neighbors in having reduced genetic diversity and high frequencies of unique mtDNA types [ 9 , 10 , 11 , 12 , 13 , 14 ], so we might expect a similar pattern if the Mlabri have always been hunter\u2013gatherers. The genetic results, combined with linguistic and cultural evidence, suggest that the most probable explanation for the origin of the Mlabri is an extreme founder event from an agricultural group, followed by adoption of a hunting\u2013gathering lifestyle. Results/Discussion Genetic Analyses: mtDNA Diversity We analyzed 360 bp of the first hypervariable segment (HV1) of the mtDNA control region in 58 Mlabri; surprisingly, all of the sequences were identical, with the following differences from the reference sequence [ 15 ]: 16140C, 16189C, and 16266A, as well as the common Asian 9-bp deletion in the intergenic region between the cytochrome oxidase subunit II and lysine tRNA genes [ 16 ]. No other human population has been found to lack mtDNA HV1 variation, and mtDNA HV1 variation in six other hill tribes (all agricultural groups) from the same region of Thailand was significantly higher ( Figure 1 ; Table 1 ). Figure 1 Genetic Diversity in the Mlabri and Other Hill Tribes Genetic diversity based on mtDNA HV1 sequences, Y-STR haplotypes, and autosomal STR (A-STR) genotypes in the Mlabri, compared to the average genetic diversity for six other hill tribes. The haplotype diversity is indicated for the mtDNA and Y-STR data, while the average heterozygosity is indicated for the autosomal STR loci. Table 1 Genetic Diversity Parameters Based on mtDNA HV1 Sequences, Y-STR Haplotypes, and Autosomal STR Genotypes for the Mlabri and the Six Other Hill Tribes Diversity in the Mlabri is significantly lower than the average for the other groups for all three genetic systems, based on t -tests (not shown) a Probability of the observed heterozygosity excess under the stepwise mutation model, Wilcoxon one-tailed test Y-Chromosome Diversity We analyzed nine short tandem repeat (STR) loci on the Y chromosome in 54 Mlabri, and again found significantly reduced variation in the Mlabri compared to the other six hill tribes ( Figure 1 ; Table 1 ). The Mlabri had just four Y-chromosome STR (Y-STR) haplotypes, two of which differed by a single repeat at a single locus from one each of the other two haplotypes ( Table 2 ). The Y-STR haplotype diversity in the Mlabri is again lower than that reported for any other human population [ 17 , 18 ]; the Akha, one of the six other hill tribes, also exhibited very low Y-STR diversity ( Table 1 ). The average variance in the allele size distribution at the nine Y-STR loci shows an even greater contrast between the Mlabri and the other hill tribes: the average variance was 0.11 for the Mlabri, versus an average of 1.45 for the other six hill tribes. Table 2 Y-STR Haplotypes in the Mlabri The number of repeats for the allele at each locus in the four haplotypes is given Autosomal DNA Diversity We analyzed nine autosomal STR loci in the Mlabri and the other six hill tribes, and again found significantly reduced variation in the Mlabri ( Figure 1 ; Table 1 ). The genotype frequencies did not deviate significantly from Hardy\u2013Weinberg expectations for any locus in the Mlabri; however, even though these nine STR loci are on different chromosomes and hence unlinked, eight pairs of loci exhibited significant linkage disequilibrium (LD) ( p < 0.05; Figure 2 ), as measured by a likelihood ratio test [ 19 ]. This is significantly more ( p < 0.01) than the 1.8 pairs expected by chance (out of 36 pairwise comparisons) to exhibit this level of LD. For each of the six agricultural hill tribes, the number of pairs of loci exhibiting significant LD was within expectations ( Figure 2 ). Moreover, the p -value of the likelihood ratio test is a measure of the strength of the association between two loci [ 19 ]; the average p -value was 0.20 for the Mlabri, versus 0.31\u20130.55 for the other six hill tribes, indicating that overall associations between these unlinked loci were stronger in the Mlabri than in the other hill tribes. However, the sample size for the Mlabri for the autosomal STR analyses was larger than the sample size for the other hill tribes ( n = 35 for the Mlabri, versus n = 29\u201330 for the others), so it is possible that the lower average p -value for the Mlabri reflects more statistical power due to a larger sample size and not more LD. To test this, we sampled 30 Mlabri at random and redid the LD analysis; the conclusions did not change, indicating that the lower average p- value for the Mlabri does reflect more LD in the Mlabri. Figure 2 Associations amongst Unlinked Autosomal STR Loci in the Mlabri and the Other Hill Tribes Probability values of the likelihood ratio test of association versus no association for nine unlinked autosomal STR loci in the Mlabri and six other hill tribes (average probability over the 36 pairs of loci in parentheses). One explanation for the reduced diversity at mtDNA, Y-STR loci, and autosomal STR loci, and the significant number of pairs of unlinked autosomal STR loci in LD, is a severe reduction in population size in the Mlabri. Following such an event, the number of alleles is reduced more than the heterozygosity, leading to an excess of observed heterozygosity compared to that expected for the observed number of alleles under mutation\u2013drift equilibrium [ 20 ]. We therefore compared the observed and expected heterozygosity (at mutation\u2013drift equilibrium, conditioned on the observed number of alleles) for the autosomal STR loci in the Mlabri and the six other hill tribes, under a stepwise mutation model. Only the Mlabri exhibited a significant excess of observed heterozygosity ( Table 1 ). Although more complicated scenarios are possible, the simplest explanation is that the Mlabri (but not the other hill tribes) have undergone a severe reduction in population size, as also indicated by the mtDNA and Y-STR haplotype data, and as also suggested by a previous study of blood group variation [ 4 ]. Population Size Reduction in the Mlabri Assuming that there was a reduction in population size in the Mlabri that set the mtDNA and Y-chromosome diversity near or equal to zero, the coalescence times for the Mlabri mtDNA and Y-STR haplotypes provide an upper estimate as to when the population reduction occurred. We therefore applied Bayesian-based coalescence analysis [ 21 ] to the mtDNA sequences and the Y-STR haplotypes from the Mlabri and the other six hill tribes. For the six agricultural hill tribes, the resulting estimates of coalescence time are broadly distributed ( Figure 3 ), indicating little information in the data (except for the Akha, who do show a pronounced peak in the posterior probability distribution for the Y-STR data, in accordance with their lower Y-STR haplotype diversity). By contrast, the estimates of coalescence time for the Mlabri show a sharp peak ( Figure 3 ), with a median time of 770 y (approximate 95% credible interval 250\u20134,270 y) for the mtDNA sequences and 490 y (approximate 95% credible interval 170\u20131,290 y) for the Y-STR haplotypes. Figure 3 Time to the Most Recent Common Ancestor for mtDNA and Y-STR Types for the Mlabri and the Other Hill Tribes Posterior probability distribution of the time back to the most recent common ancestor for the mtDNA (A) and Y-STR haplotype (B) data for the Mlabri and six other hill tribes. Both the mtDNA and the Y-STR data therefore indicate that the Mlabri underwent a substantial reduction in population size about 500\u2013800 y ago (and not more than about 1,300 y ago, if the mtDNA and Y-chromosome data reflect the same event). There are two possible scenarios: (1) a bottleneck, in which the Mlabri were reduced from a formerly large population to a much smaller population size, which then increased to the current level of about 300 individuals; or (2) a founder event, in which the Mlabri began as a very small number of individuals, became isolated, and then increased over time to their present size. Similar reductions in genetic diversity are predicted under either scenario, so the genetic data cannot distinguish between these. But some information can be obtained by considering the magnitude of the reduction in population size needed to completely eliminate mtDNA diversity in the Mlabri. The amount of population size reduction needed to completely eliminate mtDNA diversity in the Mlabri depends on how much mtDNA diversity was present prior to the size reduction. We assumed that the ancestral Mlabri population would have the same mtDNA diversity as one of the other hill tribes and then estimated the amount of population size reduction needed to completely eliminate mtDNA diversity by resampling with replacement various numbers of mtDNA types from the ancestral (pre-bottleneck) population. For example, we started with an ancestral population with the same distribution of mtDNA types as the Akha. We then sampled two mtDNA types (with replacement) from this ancestral population, repeated this procedure 1,000 times, and found that 243 out of the 1,000 resamples of size two had no mtDNA diversity; thus, the probability is 0.243 that a reduction to just two individuals would eliminate mtDNA diversity in an ancestral population that started with the same mtDNA diversity as the Akha. We then repeated this procedure, sampling three mtDNA types (with replacement), and obtained a probability of 0.007 that there would be no mtDNA diversity following a reduction to three individuals. Therefore, if the Mlabri were derived from a population with the same mtDNA diversity as the Akha, the population had to be reduced to not more than two unrelated females, in order to completely eliminate mtDNA diversity. This resampling analysis was carried out six times, with the putative ancestral mtDNA diversity corresponding to each of the six hill tribes. The results of this analysis were that for five of the ancestral populations, resampling three (or more) individuals gave a probability of no mtDNA diversity of less than 0.05; for the remaining ancestral population (which had the same starting mtDNA diversity as the Red Karen), resampling four (or more) individuals gave a probability of no mtDNA diversity of less than 0.05. We also carried out a similar analysis for the Y-STR types in the Mlabri. Here we again assumed an ancestral population with the same Y-STR haplotype diversity as one of the other hill tribes, then determined the maximum number of individuals that could be sampled at random that would have not more than two Y-STR types (since the four Y-STR types in the Mlabri consist of two pairs that differ by a single-step mutation at a single locus). The results of this analysis were that at most 3\u20136 individuals (depending on which hill tribe the ancestral population resembled most in terms of Y-STR diversity) could have been present after the size reduction, otherwise, with greater than 95% probability, more than two Y-STR types would have been retained. A critical assumption is the amount of genetic diversity present in the ancestral Mlabri population prior to the size reduction. The estimates used in the above analysis are based on agricultural populations, which in general have more mtDNA diversity than hunter\u2013gatherer populations. We therefore also constructed putative ancestral populations with frequency distributions of mtDNA types identical to those found in the !Kung and in African Pygmies [ 22 ]; the results of the resampling analysis were the same. Another assumption of this analysis is that the event that led to the population size reduction completely eliminated the mtDNA diversity. Alternatively, some mtDNA diversity may have been present after the population size reduction, but was subsequently lost because of drift. Loss of mtDNA diversity due to subsequent drift is not likely if there was a single event reducing the Mlabri population size that was followed by population growth, since mtDNA diversity is retained in growing populations [ 23 ]. However, if the reduction in size occurred over several generations, then it may not have been as dramatic a bottleneck as the resampling analysis implies. To investigate this further, we employed a Bayesian approach, following the procedure previously used to estimate the number of founders for the Maoris [ 24 ] but allowing for new mutations, to estimate the number of founders for the Mlabri, assuming various time periods since the founding event. The results ( Figure 4 ) indicate that the most probable number of founders is one over all time periods; however, for longer time periods since the founding event, there is decreasing information on the number of founders from the observation of no mtDNA diversity in the Mlabri. As expected, the longer the time since the founding event (i.e., the slower the population growth rate), the greater the influence of drift in eliminating diversity that might have been present in the founding population. Nevertheless, given that the coalescent analyses indicate an upper date for the origin of the Mlabri of about 1,000 y ago, the lack of mtDNA diversity in the Mlabri is most consistent with a very small founding population size, perhaps even only one female lineage. Figure 4 Number of Founding Individuals in the Mlabri, Given No mtDNA Diversity Posterior probability distribution for the number of founding individuals (k), conditioned on the observation of no diversity in a sample of 58 mtDNA sequences and the time since the founding event. The prior probability is indicated by the dashed black line. Origin of the Mlabri The group that gave rise to the founder event that established the Mlabri could have been either a hunter\u2013gatherer group, in which case the Mlabri maintained their hunting\u2013gathering lifestyle from before, or an agricultural group, in which case the Mlabri subsequently adopted their current hunting\u2013gathering lifestyle. While the genetic data cannot unequivocally distinguish between these two possibilities, they do suggest the latter. Other hunter\u2013gatherer groups typically share few, if any, mtDNA types with neighboring agricultural groups, consistent with long-term isolation of the hunter\u2013gatherer groups. For example, !Kung, African Pygmies, Andamanese Islanders, and south Indian hunter\u2013gatherer groups can readily be distinguished from nearby agricultural groups on the basis of their mtDNA sequences [ 9 , 13 , 14 , 25 ]. By contrast, the Mlabri mtDNA sequence has been reported in other, agricultural hill tribes [ 26 , 27 ], and identical or closely related sequences have also been reported from Southeast Asia and China [ 9 , 28 , 29 ]. Similarly, the Mlabri Y-STR haplotypes are identical or closely related (differing by a single-step mutation at one locus) to Y-STR haplotypes found in Southeast Asia and Oceania [ 30 , 31 ]. Also, the Mlabri do not exhibit any alleles at the nine autosomal STR loci that are not found in the agricultural hill tribes. The widespread sharing of mtDNA, Y-STR, and autosomal STR alleles between the Mlabri and agricultural groups in Southeast Asia are not expected if the Mlabri have always been hunter\u2013gatherers. Instead, the genetic data suggest that the Mlabri are derived from an agricultural group. Moreover, the Mlabri vocabulary and folklore also give some evidence of ancient familiarity with agriculture coexisting with hunting and gathering (J. Rischel, personal communication). While preliminary in nature, the available linguistic evidence suggests that the present-day Mlabri language arose after some speakers of a Khmuic language, most likely Tin, became isolated and subsequently experienced intensive contact with speakers of some other, presently unknown language [ 2 , 8 ]. Just how long ago the Mlabri and Tin languages diverged cannot be determined, but it has been suggested that Tin branched from Khmu about 600 y ago, and that Tin then branched into two varieties (Mal and Prai) some 200\u2013300 y ago [ 6 , 32 ]. These time estimates are based on a calibration of the chronology of sound changes in Tin against reasonably secure datings of sound changes in neighboring languages; the actual time depth may be underestimated, but most likely by not more than a few centuries. Thus, the linguistic evidence would date the origin of the Mlabri at less than 1,000 y ago, in excellent agreement with the genetic evidence. Other data that may shed light on the origins of the Mlabri, such as historical information, are scarce, since the Mlabri do not have a written language and the first recorded contact was only in 1936. However, the Tin Prai have an oral tradition concerning the origin of the Mlabri (J. Rischel, personal communication), in which several hundred years ago, Tin Prai villagers expelled two children and sent them downriver on a raft. They survived and escaped into the forest, turning to a foraging lifestyle and thus becoming the Mlabri. Although it is difficult to know how to evaluate such oral traditions, this story nevertheless intriguingly parallels the genetic and linguistic evidence concerning the origins of the Mlabri. In sum, genetic, linguistic, and cultural data all suggest a founding event in the Mlabri, involving a single maternal lineage and 1\u20134 paternal lineages some 500\u20131,000 y ago, from an ancestral agricultural population. The Mlabri then subsequently adopted their present hunting and gathering lifestyle, possibly because the group size at the time of founding was too small to support an agricultural lifestyle. Other examples of such cultural reversion are rare; probably the best known involves Polynesian hunter\u2013gatherers on the Chatham Islands and the South Island of New Zealand [ 33 ], who abandoned agriculture and adopted a maritime-based foraging subsistence because of the rich marine resources and the inability of these islands to support cultivation of tropical crops. Other hypothesized examples of cultural reversion, such as the Punan of Borneo [ 34 ], the Guaj\u00e1 and other lowland Amazonian groups [ 35 ], and the Sirion\u00f3 of Bolivia [ 36 ], are controversial, as it is not clear whether these groups are descended directly from earlier hunter\u2013gatherer groups or whether they indeed have undergone cultural reversion. Detailed genetic analyses, as carried out here for the Mlabri, may shed further light in these cases. In any event, our conclusion that the Mlabri, a present-day group of hunters and gatherers, was founded recently and in all probability from an agricultural group further supports the contention that contemporary hunter\u2013gatherer groups cannot be automatically assumed to represent the pre-agricultural lifestyle of human populations, descended unchanged from the Stone Age [ 37 ]. Indeed, even if they have not reverted from an agricultural lifestyle, most (if not all) contemporary hunter\u2013gatherer groups interact with, and have evolved and changed along with, agricultural groups [ 38 ]. The Mlabri provide a unique opportunity to investigate the circumstances and consequences of a reversion from an agricultural to a hunting\u2013gathering lifestyle that apparently was not dictated by purely ecological reasons (as in the case of Polynesian hunter\u2013gatherers). Materials and Methods Samples There are three linguistically distinct subgroups of Mlabri [ 39 ], designated A, B, and C. Subgroup A (also known erroneously as \u201cMrabri\u201d) is the only group that has been studied in detail [ 1 ]; subgroup B (minor Mlabri) is practically extinct [ 2 ], and subgroup C (formerly \u201cYumbri\u201d) comprises less than 30 people [ 39 ]. Blood samples and genealogies of 91 Mlabri (all from subgroup A) were obtained with informed consent in 1999, and cell lines were prepared and DNA was extracted from the cell lines. The genealogical data were used to identify and exclude known relatives from the genetic analyses. Data on mtDNA and Y-STR variation from six agricultural hill tribes in the same geographic region (Akha, Lahu, White Karen, Red Karen, CR Lisu [from near Chiang Rai], and MHS Lisu [from near Mae Hong Son]), all of whom speak Sino-Tibetan languages, were published previously [ 26 ]. Genetic analyses The first hypervariable segment (HV1) of the mtDNA control region (nucleotide positions 16,024\u201316,385) was amplified and sequenced directly, as described previously [ 29 ], from 58 Mlabri. PCR analysis of the intergenic region between the cytochrome oxidase subunit II and lysine tRNA genes, which harbors an informative 9-bp deletion, was carried out as described previously [ 28 ]. Nine Y-STR loci (DYS385a, DYS385b, DYS389I, DYS389II, DYS390, DYS391, DYS392, DYS393, and DYS394) were amplified and genotypes determined, using previously described methods [ 30 ], for 54 Mlabri. Nine autosomal STR loci (D3S1358, vWA, FGA, D8S1179, D21S11, D18S51, D5S818, D13S317, and D7S820) plus the amelogenin locus were amplified with the AmpFLSTR Profiler Plus PCR Amplification Kit (Applied Biosystems, Foster City, California, United States), using 2\u20134 ng of DNA in a 15-\u03bcl reaction volume. Genotypes were determined by fragment analysis on an ABI377 (Applied Biosystems) for 35 Mlabri, 29 Lahu, and 30 individuals from each of the other hill tribes. Statistical analyses Genetic diversity, heterozygosity, and tests for goodness of fit to Hardy\u2013Weinberg expectations were calculated with Arlequin 2.000 [ 40 ]. LD was estimated as the probability of the likelihood of the data assuming linkage equilibrium versus the likelihood of the data assuming association [ 19 ]; Arlequin 2.000 was used to obtain maximum-likelihood estimates of the haplotype frequencies for each pair of loci with the EM algorithm [ 41 ], and the null distribution of the p -value of the likelihood ratio test was generated by 10,000 random permutations. The program Bottleneck ( http://www.ensam.inra.fr/URLB/bottleneck/bottleneck.html ) was used to compare the observed heterozygosity at each autosomal STR locus to that expected at mutation\u2013drift equilibrium for the observed number of alleles, assuming a stepwise mutation model. Bayesian-based coalescence analyses of Y-STR haplotypes [ 42 ] were performed using the software Batwing ( http://www.maths.abdn.ac.uk/~ijw/downloads/batwing/batguide/node6.html ) and previously described prior distributions for the initial effective population size, population growth rate, and Y-STR mutation rates [ 30 ]. The coalescence time for mtDNA HV1 sequences was also estimated by a Bayesian procedure [ 21 ] as described previously for Xq13.3 sequences [ 43 ], using the same initial effective population size and population growth rate priors as for the Y-STR analysis, and a \u03b3-distribution with parameters \u03b1 = 14.74 and \u03b2 = 0.0005 (corresponding mean = 0.00737) as a prior for the mutation rate [ 44 ]. Resampling of mtDNA and Y-STR types, in order to estimate the magnitude of population size reduction needed to eliminate mtDNA and reduce Y-STR diversity, was performed with the software Resample ( http://www.pbs.port.ac.uk/~woodm/resample.htm ). Bayesian analysis of the number of founders for the Mlabri was performed by pooling the mtDNA types in the other hill tribes to obtain a starting population, from which a certain number of founding mtDNA types were selected at random, assuming a uniform prior distribution between one and 20 founders. The sample was then allowed to grow from the number of founders to size 300 (the current size of the Mlabri population) over various time intervals, such that the shorter the time interval, the faster the growth rate. Simulations were performed both under the assumption of no new mutations, and with a mutation rate of one mutation/sequence/10,000 y. For each combination of parameters, 1,000,000 simulations were carried out. The simulation results were converted via Bayes's theorem into a posterior probability for the number of founding individuals, conditioned on the observation of no diversity in a random sample of size 58 (the sample size in this study). In practice, the posterior probability distributions were independent of the mutation rate (analyses not shown)."}