Transcriptome profiling of chemosensory appendages in the malaria vector Anopheles gambiae reveals tissue- and sex-specific signatures of odor coding

Mosquito rearing

An. gambiae sensu stricto, which originated from Suakoko, Liberia [38], was reared as described [39]. Briefly, mosquitoes were reared in an isolated chamber at 27°C and 75% relative humidity with a photoperiod of 12:12 (L:D). Larvae were reared at low densities to ensure large adult size. Pupae were hand collected and allowed to eclose in small cages. Almost all pupae eclosed on the day after collection. The few pupae that failed to eclose were removed, such that adults in any single cage were the same age. Adult females and males were kept together in the same cages for 4-6 days and were sugar-fed with 10% sucrose ad libitum until the time of tissue collection. Females were not bloodfed prior to tissue collection, nor were they selected based upon any specific response to external stimuli. As a consequence of the rearing protocol, mosquitoes were intermittently exposed to the odor of their human caretakers. The vast majority of females were assumed to be mated based on numerous previous studies of cage-reared An. gambiae[40, 41] and our own experience. Moreover, nearly all females reared as described above in our laboratory will bloodfeed when presented with an anesthetized mouse, indicating that they are physiologically competent to host seek.

RNA isolation and sequencing

Approximately 1500 antennae or maxillary palps were hand dissected from randomly selected, age-matched cohorts of 4-6 d.o. adults (ZT10-12). Additionally, approximately 20 whole bodies of 4-6 d.o. adults were collected of each sex (ZT10-12). All collected tissues were immediately placed in RNA Later Ice (Ambion Corp.; Austin, TX) on ice prior to RNA extraction. Total RNA was isolated from each sample using RNeasy columns (Qiagen Inc.; Carlsbad, CA) according to the manufacturer's protocol. mRNA isolation and cDNA library preparation were carried out using the Illumina mRNA sequencing kit (Illumina Inc.; San Diego, CA). Libraries were sequenced using an Illumina Genome Analyzer II or HiSeq2000. A single biological replicate, representing a large sample size was used in the subsequent analysis.

AgOr and AgObp reannotations

Data processing and expression profiling

Individual Illumina read files were mapped to the recently updated (Dec. 2010) version of the assembled An. gambiae genome, to the mitochondrial genome, and to the annotated An. gambiae transcripts (http://www.VectorBase.org). For mapping purposes, all transcript isoforms for a given gene were condensed under that gene's AGAP designation. Prior to mapping, individual reads were quality checked and uniformly trimmed by 4 and 12 nucleotides on their 5' and 3' ends, respectively, to account for spurious adapter incorporation (5'end) and for sequencing reaction degeneration (3'-end). Mapping was carried out using seqmap software, configured to allow for a maximum of three mismatches per read. Processed mapping data was then consolidated based upon AGAP number and the results summarized by rseq software. Expression level output by rseq was reported in terms of unique reads, total weighted reads, and transcript length. Total weighted reads and AGAP transcript lengths were used to calculate a normalized transcript abundance level in units of Reads Per Kilobase per Million reads mapped (RPKMs), for every AGAP in every tissue type [32].

PfamA categorization

Peptide sequences from AgamP3.6 conceptual peptides (n = 12669) were compared to the PfamA dataset [46], using the default e-value threshold of 1.0.

Comparison of tissue expression profiles

Statistical significance was assigned to each pairwise tissue comparison (antennae:bodies, palps:bodies, bodies:bodies) by setting up a Fisher's Exact test, comparing the number of weighted, mapped reads for each gene to the total number of mapped reads for that tissue sample. The Agam3.6 transcript annotation contains 13319 unique, annotated transcripts and the statistical significance of the Fisher's Test was evaluated against a Bonferroni corrected p-value of 3.9 × 10^-6.

RNA Sequencing and Gene Mapping

As a means of inferring gene expression in chemosensory appendages we have employed single-end short read (43bp) RNA-seq technology to characterize the relative abundances of poly-adenylated RNAs in antennae, maxillary palps and whole bodies of female and male adult mosquitoes. We established tissue-specific gene expression profiles for each of our six samples by mapping the read sequence files against the annotated An. gambiae transcriptome, using an approach that quantitated transcript abundance per gene and which accounted for all annotated transcripts per gene (see Materials and Methods). As our reference transcriptome, we used the AgamP3.6 version of the An. gambiae gene annotation, which contains 12669 protein-coding genes and 650 non-coding RNAs (http://www.VectorBase.org). For each of the tissue types assayed, we obtained an average of 30.5 million sequence reads per tissue type and mapped them to the An. gambiae transcriptome, nuclear and mitochondrial genomes (Table 1). Additional file 1 contains the complete RNA-seq data set described above, including the number of reads from each tissue sample that mapped to all 13319 annotated An. gambiae genes. On average, 57.4% of the reads per sample mapped to annotated genes, 91.5% to the nuclear genome (Table 1), and 2% to the mitochondrial genome (Table 1). Of the reads that mapped only to the genome, many of them are likely to represent unannotated 5'or 3' untranslated regions (UTRs). Moreover, there likely remain regions of the genome, most notably the Y-chromosome, where novel exons and transcripts remain [47].

On a whole-genome level, comparison of the mapping density of reads sequenced from the female body along all chromosomes showed a high degree of correspondence between the number of reads mapped to the nuclear genome and the number of reads mapped to the transcriptome (Figure 1). There are a few areas of asymmetry where a higher degree of mapping to either the transcriptome or to the genome was observed, most noticeably in the gene-rich autosomal telomeres and in several regions of the X chromosome (Figure 1). Greater mapping frequency to the genome represents actively transcribed regions that are unannotated as distinct genes. Greater mapping frequency to the transcriptome can generally be explained as reads that map to exon-exon junctions, which by their nature would not map to the genome. For example, the observed asymmetry in the 2R telomeric region is due to the high number of exon junction reads that mapped to two rhodopsin-family genes (Figure 1).

To quantify relative differences in gene expression levels within each tissue, we calculated a Reads Per Kilobase per Million (RPKM) reads mapped value for each gene within a sample [32]. Mean and median RPKM values for each tissue type in this study were very similar across samples, as were the number of genes showing basal or greater levels of transcription (Table 1). RPKM values spanned more than 6 orders of magnitude for each of the tissue types examined (see Additional file 1).

We assessed fold-differences in transcript abundance by independently comparing ratios of RPKM values between pairs of tissues within each sex: antennae to bodies and maxillary palps to bodies. For each of these pairwise comparisons we performed a Fisher's Exact Test on counts of mapped reads and assigned statistical significance using a Bonferroni-corrected p-value (p < 3.9 × 10^-6; see Materials and Methods). Furthermore, we have used the term "enhanced" to describe any gene that displayed at least a 2-fold, significant difference in transcript abundance between the samples being compared (Figure 2). These conservative criteria were applied to avoid false positives stemming from variations within the samples themselves, as well as to reduce the numbers of genes that were used for subsequent analyses [48, 49].

Gene Expression Profiling in Chemosensory Tissues

To examine global gene expression patterns, we have compared RPKM values pairwise for whole bodies versus either antennae or maxillary palps in both sexes. One such comparison is shown in Figure 2 where 4587 genes displayed enhancement in the female antenna to body comparison (Figure 2, black dots). Of those, 2277 were enhanced in the antenna (Figure 2, right half). Similarly, we found that 1906 genes were enhanced in female palps, 3037 genes were enhanced in male antennae, and 2284 genes were enhanced in male palps. These 4 gene sets formed the basis of our subsequent analyses where we compared enhanced gene profiles between chemosensory tissues and across sexes (Figure 3).

Comparing the enhanced gene sets between the female antennae and palps revealed significant overlap, with 1158 genes (61% of palp set) enhanced in both tissues (see Additional file 2). Similarly, male antennae and palps showed significant overlap with 1208 genes enhanced in both tissues (53% of palp set; see Additional file 2). Interestingly, the most well-represented gene families in both of these overlapping sets were 7-transmembrane receptors (PF00001), protein kinases (PF00069), cytochrome P450s (PF00067), trypsins (PF00089), carboxylesterases (PF00135), and potential transcription factors (PFs 00046 and 00096; see Additional files 2 and 3, bottom tables). However, we also observed several differentially enhanced gene sets between the antennae and palps (see Additional files 1 and 2). The An. gambiae Ors (AgOrs; [12]; PfamA family PF02949) were the most prevalent class in female antennae (see Additional file 2, left table) and second-most in the male antennae (see Additional file 3, left table). Other chemosensory gene families, such as ligand-gated ion channels, which include the recently identified ionotropic receptors (AgIrs; [13]; PF00060), and odorant binding proteins (AgObps; [42]; PF01395), were highly represented in the antennae (see Additional files 2 and 3). It is clear from these antennae-to-palp analyses that both extensive overlap and significant distinctions in gene expression profiles exist. The consistent identification of the same PfamA families in all enhanced gene sets implicates functional groups that can be studied in greater detail to elucidate their potential roles in mosquito chemosensation.

To evaluate gene expression patterns between sexes, we have compared female and male antennal-enhanced gene sets and palp-enhanced gene sets. In antennae, 2277 female, and 3037 male antennal-enhanced genes shared a common set of 1381 genes (Figure 3 and Additional file 4). Once again, this set included AgOrs, AgIrs, and AgObps (see Additional file 4, bottom table). Despite many commonalities in gene expression, there were also 896 female antennae-specific enhanced genes and, surprisingly, nearly 1700 male antennae-specific enhanced genes (Figure 3 and Additional file 4). In the maxillary palps, as in the antennae, considerable overlap was found in gene expression profile between the sexes. In the palp, 778 genes were common between the 1906 female palp-enhanced gene set and the 2284 male palp-enhanced gene set (Figure 3 and Additional file 5). Interestingly, the fraction of enhanced gene overlap was much lower in the palps than in the antennae; 61% of the total female antennal-enhanced set that was shared with males (see Additional file 4) while only 41% of the total female palp-enhanced set was shared with males (see Additional file 5). This result may indicate the presence of cryptic sex-specific specializations in the maxillary palps.

Given the obvious sexual dimorphisms of An. gambiae antennae and maxillary palps (Figure 4) comparisons of their gene expression profiles is not necessarily straightforward. Chemosensory sensilla, and AgOr- containing neurons in particular, are distributed over the entire length of the female antenna, whereas male antennae house ~3-fold fewer chemosensilla that are restricted to distal segments 12 and 13 [4, 50–52]. Furthermore, while female antennae are predominantly chemosensory, male antennae are also highly specialized for hearing [53, 54]. Accordingly, the An. gambiae orthologs of the D. melanogaster trpV channels Nanchung and inactive, which are required for hearing in the fruit fly, were enhanced in antennae of both An. gambiae sexes (AGAPs 012241 and 000413, respectively; Table 2), but their expression levels were much higher in male antennae (RPKMs of 183.92 and 104.49 in males and 20.54 and 7.66 respectively, in females) [55, 56]. This elevated expression of auditory-associated genes in the male antenna is consistent with male An. gambiae mating behavior where an acute sense of hearing facilitates the recognition of female wing beats within the context of a male swarm [40, 53, 54]. Given that wild females are likely to mate just once, while males swarm daily in search of a mate [20, 40], the specialization shift away from olfaction and toward audition in the principle male sensory organ is reasonable, presumably as a mechanism to increase male mating success.

These comparisons also revealed multiple classes of genes beyond the known chemosensory gene families that displayed enhanced tissue expression. A detailed examination of the expression patterns of a subset of other gene families is provided in Table 2, many of which are represented in figures found in Additional files 2, 3, 4, and 5. Nearly half of the members of the large superfamily of 7-transmembrane (7tm) receptors (114 of the 241 recognized by PfamA) were enhanced in at least one of the chemosensory tissues examined (Table 2). This may indicate unrecognized roles in sensory reception or regulation of chemoreceptor neuron or accessory cell function. Importantly, efferent projections from serotonergic, or tachykinin neuroendocrine cells have been identified in mosquito chemosensory appendages [57–59]. Thus the expression of serotonin (AGAPs 002232, 002679, 004222, 004223, 007136, and 011481), and tachykinin (AGAPs 001592 and 012824) receptor homologs in An. gambiae antennae and maxillary palps (see Additional file 1) is consistent with a neuromodulatory role for these compounds.

Other gene families with multiple members that displayed chemosensory enhancement include the CD36 family, some members of which function in insect olfaction [60, 61] and ion channels and transporters, including the recently identified chemosensory ionotropic receptors [13, 62, 63]. Enhanced levels of such biotransformation enzymes as carboxylesterases and cytochrome P450s could hint at a potential role in odor degradation [64–66]. Carbonic anhydrases involved in carbon dioxide detection in mammals [67] and transcription factors, including the An. gambiae homologs of acj6 and pdm3, D. melanogaster pou-type transcription factors involved in DmOr gene regulation and ORN axon targeting [68–74] were also enhanced in chemosensory tissues (Table 2).

We also identified a number of genes encoding small, soluble proteins with enhanced expression in chemosensory tissues of both sexes (Table 2). Transcripts encoding the CRAL-TRIO (PF00650), cysteine-rich secretory protein (PF00188), and haemolymph juvenile hormone binding proteins (JHBP, PF06585) were highly enhanced. To our knowledge, the first two gene families have not been linked to chemosensation, but the members of the JHBP family have been identified in screens of highly expressed genes in mosquito antennae [75, 76]. Moreover the JHBP gene, takeout, links the circadian clock and feeding behavior in D. melanogaster[77], and modulates aggregation behavior in Locusta migratoria[78]. The extremely high expression levels of some members of these 3 gene families suggest potential chemosensory functions analogous to other soluble lipophilic carriers such as the Obps.

Chemosensory Gene Families

In light of the existing literature on the molecular mechanisms underlying the processes of peripheral chemosensation in vector mosquitoes, we examined in detail the expression patterns of AgOrs, AgIrs, AgObps and gustatory receptors (AgGrs). As expected, the vast majority of AgOrs were highly enhanced in antennae. Of the 76 AgOrs, 58 showed enhanced expression in female antennae as compared to only 36 in male antennae (Figure 5). The entire set of male-enhanced AgOr s was contained within the female enhanced set. None of the larval-specific AgOrs: 37, 40, 52, or 58, was enhanced in adult antennae or palps, supporting previous observations [79]. In the palps, only AgOrs8 and 28 and AgOrco (formerly AgOr7, recently renamed to reflect its capacity as an obligate Or co-receptor) were enhanced in female maxillary palps (Figure 5), a result consistent with our previous study on odor coding in the An. gambiae maxillary palps [11]. The same 3 AgOrs were enhanced in male palps (Figure 5).

Several members of the recently described AgIr gene family [13, 62], displayed significant enhancement in antennae of both sexes (Figure 6), further supporting their potential roles as chemosensory receptors in An. gambiae. A high degree of overlap was observed between the sexes, where 21 AgIrs were enhanced in both. Similar to the AgOrs, there were many fewer AgIrs enhanced in the palps compared to the antennae, with 7 and 6 enhanced in female and male palps, respectively. Furthermore, the degree of overlap (3 genes) between the sexes was much less pronounced in the palp (Figure 6).

The enhanced AgGrs were the only class that did not overlap in the antennae between the sexes, with very few showing enhancement in either females or males (Figure 7). Only AgGr1 was enhanced in female antennae, while AgGrs, 33, 48, 49, and 50 were enhanced in male antennae. Notably, one member of this large gene family, AgGr33 was strongly enhanced in male antennae (Figure 7), perhaps indicating a specialized function in males. Interestingly, AgGr33 shares significant homology with D. melanogaster Gr28[12], some splice forms of which are expressed in non-chemosensory tissues, including the Johnston's organ [80]. In contrast to the acute sexual dimorphism displayed in the antennae, both sexes showed enhanced expression of AgGrs 22, 23, and 24, in their maxillary palps (Figure 6). These three AgGrs are homologs of the D. melanogaster carbon dioxide receptors [81–83], and are expressed in capitate peg sensilla on the maxillary palps where they have been directly implicated in An. gambiae CO₂ sensing [11].

Enhanced chemosensory expression of members of the large AgObp family was evident across all tissues and sexes (see Additional file 1). Sixteen classical and 3 Plus-C AgObp s were significantly enhanced in the female antennae (Figure 8). Of these, 17 (see Additional file 1) were also significantly enhanced in the male antennae (Figure 8) including the D. melanogaster LUSH homolog, AgObp4[84]. AgObp19 was the only one that demonstrated significantly enhanced expression in the female antennae and in no other tissue. In the maxillary palp, enhancement of AgObp transcripts also displayed substantial overlap between sexes, where the 4 male enhanced AgObps were all similarly elevated in females. Overall, the AgObp expression pattern was nearly identical between male and female chemosensory tissues (Figure 8). Several AgObps were also enhanced in bodies, with a greater number of them being enhanced in male bodies (Figure 8). A direct comparison of AgObp expression in female and male bodies revealed 17 AgObps that were specifically enhanced in males, while only 5 were enhanced in females. The high representation of AgObps in the male body without the coordinate expression of known chemoreceptors (Figures 5, 6, and 7) suggests uncharacterized roles for these AgObps, perhaps as general lipophilic carriers in male-specific physiology or in chemosensory processes in tarsal and wing sensilla.

In contrast, atypical AgObp s were not enhanced in any of the tissues examined, which is consistent with previous results suggesting that expression of this subfamily is limited to pre-adult stages [42]. With the exception of AgObps 47, 48, 57, which had RPKMs of >1000, expression levels of the members of the Plus-C AgObp subfamily was very low. Of these, it is noteworthy that AgObp48 was one of the most highly expressed genes (RPKM = 32311) in any tissue, with greatly elevated expression levels in both the male and female olfactory tissues (see Additional file 1). While AgObps, and insect Obps in general are among the most highly expressed gene families in chemosensory tissues their role in non-pheromone chemosensation remains largely undefined.

With regard to olfaction, it has been hypothesized that Obps act principally as molecular shuttles/chaperones, which deliver odorants to receptors and/or transiently protect specific odorants from biotransformation enzymes [43]. If individual Obps bind a subset of odorants, it is reasonable to hypothesize, that in tissues with high Or and therefore odor-coding complexity such as the antennae, the Obp landscape would need to be similarly complex in order to bind the required range of odorants. The converse would also be expected for tissues with reduced odor coding complexity such as the maxillary palp.

Female antennae showed enhanced expression of 58 conventional AgOrs, while only 3 AgOrs were enhanced in the female palp (Figure 5). Furthermore, the odorant response profiles of the palp-expressed AgOrs 8 and 28 were also vastly different from the deorphanized antennal AgOrs[11, 17, 18]. These differences in AgOr coding capacity and their expression profiles strongly suggest that the ability of the female antennae to sense odors is much greater than the maxillary palp.

In An. gambiae females both the antennae and maxillary palps expressed 21 AgObp family members with an RPKM >10, of which 19 were found in both (see Additional file 1). While not all of these AgObps' transcript levels met our criteria for enhancement, they were nevertheless expressed in these tissues. Although the AgObp complexity was almost identical in these two appendages, they displayed vastly different AgOr complexity and thus odor coding capacities. This analysis confounds standing theories about Obp function; if a large number of Obps are required in the antennae for signaling, then their presence in the palp, with its more limited odor coding capacity, would appear superfluous. Given the broad expression of AgObps and a demonstrated lack of functional overlap between the antennae and palps, our analysis suggests that in at least some instances, Obps act as low-pass filters for environmental odorants rather than as specific odorant-carrier agents. Therefore, Obps may act to solubilize odors in some cases, but as molecular sinks in others, adding yet another dimension to peripheral odor coding. In addition, the near ubiquitous expression in both sexes of some AgObps suggests that they are playing completely uncharacterized roles outside of chemosensory processes.

Diverse Roles for Chemosensory Tissues

To explore the effect of morphology on observed AgOr expression, we have attempted to normalize sex-specific differences in transcript abundance by scaling up male AgOrs in proportion to the number of female chemosensilla. AgOrs are expressed in the trichoid sensilla, the predominant sensillar type, and not in grooved peg (GP) sensilla [14]. Sensilla counts have indicated that female antennae house an average of 630 trichoid sensilla while male antennae house an average of 225 trichoid sensilla [4, 51, 85]. Multiplying the male AgOr RPKM levels by a factor of 2.8 (630/225), resulted in a sex-normalized, AgOr expression profile that was qualitatively very similar in both sexes (Figure 9, top panel), yet male AgOr RPKM values remained lower than those in females. Alternatively, normalization of male AgOr expression by a factor equal to the fold-difference in AgOrco RPKM levels between female and male antennae (916/186 or 4.9) was also performed. In the case of AgOrco normalization, AgOr expression profiles appeared to be very similar between sexes (Figure 9, middle panel). Based on the same logic, we also normalized AgIr expression in male antennae (Figure 9, bottom panel). We postulated that AgIrs are localized in GP sensilla neurons, as they are in D. melanogaster[13, 61], we used a GP normalization factor of 4.2, which is the fold difference in GP numbers between female and male An. gambiae antennae [4]. As with AgOrs, the AgIr gene expression patterns were qualitatively similar in both sexes after normalization (Figure 9). Taken together these results suggest that male antennae express AgOr and AgIr chemoreceptor repertoires that are highly similar to those expressed in female antennae.

The AgOr and AgGr expression profiles in the maxillary palps support a similar conclusion. Although AgOrco, AgOrs 8, and 28, and AgGrs 22, 23 and 24 were enhanced in both sexes, their expression levels were lower in males than in females (Figures 5 and 7). As is the case for An. gambiae antennae, the maxillary palps are sexually dimorphic with males having about 4-fold fewer chemosensilla [4, 11]. This could account for the apparent lower chemosensory gene transcript abundances in males. Normalizing male palp AgOrs and AgGrs by this factor elevated their RPKM values closer to those of females, but did not affect the qualitative observation that the identical chemoreceptors were enhanced there (data not shown). The same could be said for AgObps in the antennae and maxillary palps (Figures 6 and 8), which were generally more enhanced in females than in males. Assuming the expression profiles seen here are meaningful at the functional level, both sexes would potentially be receptive to a qualitatively similar odor space, with females perhaps having a lower threshold response to odors and thus greater chemoreceptive power.

There is some precedent for this hypothesis. Based on electrophysiological recordings, female An. gambiae antennae responded to volatiles contained in larval breeding site water at much lower thresholds than antennae of An. gambiae males [86]. Furthermore, at least 2 species of Mansonia females were attracted to mammalian hosts at greater distances than conspecific males [87]. The basis of these differences is thought to be a function of the reduced number of antennal and palpal chemosensilla in Mansonia males relative to conspecific females [88]. In either case, the aforementioned differences in gene expression profiles could also be functionally relevant and serve as the basis for distinguishing qualitatively and quantitatively female and male chemosensory abilities. Additionally, higher order processing could contribute significantly to sexually dimorphic behaviors, thus the perception of the same odors may elicit very different responses in the sexes. These competing hypotheses are directly testable using a combination of electrophysiological recording and behavioral response assays. Moreover, the requirement in chemoreception for any of the differentially expressed genes could potentially be explored by gene silencing.