Chemoreceptor Evolution in Hymenoptera and Its Implications for the Evolution of Eusociality

1Chemosensation, or the perception of chemical cues (e.g. smell and taste) from biotic and abiotic sources, is fundamental to many aspects of insect lifecycles such as host-seeking (e.g. for blood-feeding mosquitoes), mating choice, and searching for oviposition sites. In social insects, where multiple individuals live as a group and cooperate on tasks like brood care and colony defense, chemosensation serves a particularly important role in mediating the recognition and communication between members of the same society. For instance, ants and other eusocial insects form sophisticated societies organized according to specialized behavioral castes, notably the reproductive caste (queen) and sterile caste (worker). The intricate interactions within and between castes that maintain the organization of such societies are coordinated via various types of signals such as chemical (e.g. queen pheromones) as well as acoustical and visual (e.g. the dance language in honeybee) cues. Chemical communication is perhaps the most universal and important one of such mechanisms, and largely involves the accurate discrimination of a diverse set of cuticular hydrocarbons (CHCs). Therefore, it is reasonable to hypothesize that the sophisticated chemosensory systems in social insects, and their underlying molecular components, represent adaptations that have facilitated the evolution of sociality.

2A key step in insect chemosensation is the detection of chemicals by receptor proteins present on peripheral sensory neuron membrane, which initiate downstream signaling ultimately leading to behavioral responses. Three classes of insect chemosensory receptors are currently known. Odorant receptors (Ors) and gustatory receptors (Grs) belong to the same superfamily and both have seven transmembrane domains. However they differ in their general functions: Ors are generally, but not exclusively, expressed in antenna and other chemosensory appendages where they detect various volatile compounds; in contrast, Grs display diverse expression patterns in gustatory organs and other tissues, and the few Grs with known function include receptors for sweet and bitter tastants, as well as for carbon dioxide. The third chemoreceptor class is comprised of ionotropic receptors (Irs), which are members of the ancient ionotropic glutamate receptor family. Their roles in chemosensation were only recently discovered and known ligands include amines and acids.

3The genes encoding these chemoreceptors have experienced highly dynamic evolution in insects; they are among the most rapidly evolving genes in the genome, and their copy numbers can vary considerably among insect lineages. Intriguingly, some of the largest chemoreceptor repertoires are found in eusocial insects. Each of the four ant genomes with available chemoreceptor annotations contains more than 300 Ors as well as 17–97 Grs and 23–32 Irs. The honeybee Apis mellifera also has 163 Ors, which exceeds most solitary insects. We recently compared the chemoreceptor repertoires of selected hymenopteran species and demonstrated that they are largely shaped by rapid gain-and-loss throughout the evolution of these lineages. In particular, the Or family appears to have experienced a dramatic expansion in the common ancestor of ants.

4Although these findings are consistent with the adaptive role of chemoreceptors in the evolution of social insects, contradictory observations have also been made. For example, the solitary parasitoid jewel wasp Nasonia vitripennis contains ∼60 more Ors and ∼40 more Grs than A. mellifera, seemingly challenging the notion that social insects require a greater variety of chemoreceptors than solitary insects. Additionally, a recent molecular evolutionary comparison of Ors found lower levels of positive selection in ants relative to N. vitripennis. However, N. vitripennis is the only solitary hymenopteran species with high-quality chemoreceptor annotations. Therefore, it is difficult to determine whether N. vitripennis represents an exception in chemoreceptor evolution.

5Recently, several additional hymenopteran genomes have become available, including: two solitary wasps, Ceratosolen solmsi, and Microplitis demolitor; one socially polymorphic halictid bee, Lasioglossum albipes, which represents another independent origin of eusociality in addition to honeybee and ants; and several additional ant species, notably the dorylomorph Cerapachys biroi which represents the second most early-branching ant species following the ponerine Harpegnathos saltator. These new genomes provide powerful comparative genomics resources to study the evolutionary patterns of chemoreceptor genes in solitary and social insects. Here, we report characterization of the Or and Gr repertoires in eight hymenopteran genomes where careful chemoreceptor annotations were previously lacking. We focused on Or and Gr gene families because they showed more dramatic copy number variations among previously analyzed hymenopteran genomes. By comparing >4,000 Ors and Grs from 13 solitary and social Hymenoptera species, we further investigated the evolution of hymenopteran chemoreceptor genes at levels of copy number dynamics, sequence evolution, and expression divergence.

6Genome assemblies of the eight hymenopteran species listed in table 1 were downloaded from their respective sources (supplementary table S8, Supplementary Material online). Or and Gr genes were annotated in these species as described previously. In brief, protein sequences of previously reported insect Or and Gr genes were used as queries to perform TBLASTN search (e value cutoff of 1e-5) against each of the eight hymenopteran genome. Putative Or and Gr coding regions were identified from the search results after filtering out low-scoring and short hits. For each putative coding region, the most similar query sequence was used as reference for homology-based gene prediction using GeneWise v2.2.0. All predicted gene models were checked for the presence of the characteristic domains of Ors (IPR004117) or Grs (IPR009318 or IPR013604) in their coded protein sequences using InterProScan v5. Previously annotated chemoreceptor genes of N. vitripennis, A. mellifera, and five other ants were used for this study. Genes encoding proteins no shorter than 350 amino acids were included in subsequent phylogenetic and selection analyses. It should be noted that C. obscurior and M. pharaonis have their genomes sequenced most recently during the preparation of this manuscript. The chemoreceptors of C. obscurior and M. pharaonis were annotated and provided as resources, but the phylogenetic and selection analyses were not rerun to include these two species because ants are already well represented in this study by eight other ant species.

7We first reconstructed the phylogenies of the whole Or and Gr families, containing 3,545 and 806 sequences respectively, as following: 1) translated protein sequences were aligned using MAFFT v7.215 with the high-accuracy option “E-INS-i”; 2) the alignment was filtered using trimAl v1.4 with the “automated1” option; 3) a maximum-likelihood (ML) tree was estimated using RAxML v8.1.16 with the “PROTCATJTTF” model and 100 bootstrap (BS) replicates; 4) the protein sequences were aligned again using MAFFT (“E-INS-i”) with the ML tree as guide; 5) the alignment was evaluated using GUIDANCE v1.5 with the 100 BS trees from step 3) and unreliably aligned residues (confidence score <0.8) were masked; 6) gap-rich columns were removed using trimAl with the “gappyout” option; and 7) the final ML tree was generated as in step 3).

8Subfamilies were delineated from the whole family phylogenies and analyzed individually as following: 1) protein sequences were aligned using MAFFT (“E-INS-i”); 2) the alignment was evaluated using GUIDANCE with 100 BS replicates and residues with score <0.8 were masked; 3) gap-rich columns were removed using trimAl (“gappyout”); 4) a ML tree was estimated using RAxML with the “PROTGAMMAAUTO” model and 100 BS replicates; and 5) steps 1) to 4) were repeated, only that the alignment was performed using PRANK v140603 with the “-F” option and the ML tree as guide, and the 100 BS trees were used for the GUIDANCE evaluation. The final protein sequence alignments were reverse-translated into codon alignments for selection analysis.

9To estimate the number of gene gain and loss events, the phylogenetic trees of subfamilies were first corrected using TreeFix v1.1.10 with the “long” search settings given the reference organismal phylogeny shown in figure 1; the corrected trees were then reconciled with the same organismal phylogeny using NOTUNG v2.8. All selection analyses were based on trees corrected using TreeFix as well. The results of Or and Gr subfamilies were then summed, respectively, to reflect the copy number dynamics of the whole families. The global ω ratio of each subfamily was estimated using HyPhy v2.2.4. The recently developed adaptive branch-site random effects likelihood (aBSREL) approach implemented in HyPhy was used to detect evidence of episodic positive selection on all branches in all subfamilies. Some very large subfamilies, including Or subfamilies 9-exon, E, L, V, and Gr subfamily F, were condensed by removing selected ant and wasp species (supplementary table S5 and S6, Supplementary Material online) to make the analyses feasible. The FDR method was applied to the P values of all 5,439 tests of Or and Gr branches to correct for multiple testing.

10The following antennal transcriptome data were obtained from their respective sources (supplementary table S9, Supplementary Material online): A. mellifera nurses and foragers; C. biroi workers and males; C. floridanus major workers and males; as well as H. saltator workers and males. Reads were trimmed for low-quality positions and then aligned to respective genome assemblies using STAR v2.4.0 with gene annotations as guidance (supplementary table S9, Supplementary Material online). Uniquely mapped reads were counted using HTSeq v0.6.1p1 and reads per kilobase per million mapped reads (RPKM) values were calculated for all genes as a measure of expression level. Percentile ranks of genes’ RPKM values in their respective transcriptomes were determined, and the ranks of Ors were compared between A. mellifera and each ant using the Wilcoxon rank-sum test. GFOLD v1.1.2 was used to determine the fold changes in transcript abundances of genes between workers and males in ants. To account for potential bias introduced by using only uniquely mapped reads, Cuffdiff v2.2.1 (which incorporates nonuniquely mapped reads as well) was used as an alternative approach to calculate the transcript abundances and fold changes and the results were highly similar (supplementary fig. S3 and table S10, Supplementary Material online).

11This project has been deposited at Figshare under the accession 10.6084/m9.figshare.1485695. The deposited data include all the sequences, raw and filtered alignments, phylogenetic trees of Or and Gr families and subfamilies, results of gene gain/loss and selection analyses, and commands for software tools used in this study.

12We annotated Or and Gr genes in two solitary wasps (C. solmsi and M. demolitor), one halictid bee (L. albipes), and five ants (Acromyrmex echinatior, Atta cephalotes, Cardiocondyla obscurior, Monomorium pharaonis, and Solenopsis invicta). As a result, a large number of chemoreceptor genes were identified in these species (table 1; supplementary table S1, Supplementary Material online). We then compared the size of all currently characterized chemoreceptor repertoires in Hymenoptera. To streamline downstream analyses, we focused in this study on nearly intact genes encoding proteins of 350 or more amino acids.

13The numbers of Ors are largely comparable among ants where most species have 300 or more copies (fig. 1A and table 1). The tramp ant C. obscurior has notably fewer Ors than other ants. The genome of C. obscurior is the smallest among sequenced ants and contains genomic regions with high densities of transposable elements (TEs) in which Ors were enriched, suggesting that C. obscurior Ors are likely to be affected by genome streamlining and random copy number changes caused by TE activity. There is a reduced number of Ors in the pharaoh ant M. pharaonis as well, however it is likely biased by the current, highly fragmented genome assembly. The two bees each has ∼160 Ors, highly consistent with the recent report of 159 Ors in bumblebee (Sadd et al. 2015). On the other hand, while both N. vitripennis and M. demolitor have ∼200 Ors, we found only 56 Ors in the fig wasp C. solmsi which is an obligate specialist on figs and may have likely experienced a reduction in its chemosensory gene repertoire. More wasp genomes are needed to better characterize the variability of Or family size in this lineage.

14In the Gr family, however, our results further strengthen the previous observation that Gr copy numbers vary considerably even between related species (fig. 1B and table1). For instance, three of the five ant genomes we annotated in this study have more than 100 Grs, whereas most previously analyzed ants have 50 or less copies. Strikingly, we found 219 Grs in the highly invasive fire ant S. invicta, on a par with the highest number of Grs previously reported in insects (Tribolium castaneum has 220 intact Grs). Interestingly, M. pharaonis which is also an invasive ant has a greatly expanded Gr family as well, despite its fragmented genome assembly. Among wasps, while C. solmsi has only five Grs, which is consistent with its specialist lifestyle, M. demolitor contains almost twice as many Grs as N. vitripennis (79 vs. 47). On the contrary, both A. mellifera and L. albipes have relatively small Gr repertoires, potentially reflecting the much reduced need for gustatory reception due to the mutualistic relationships between bees and plants.

15The vast majority of the chemoreceptor genes, we annotated here are tandemly arrayed in genomes (supplementary fig. S1, Supplementary Material online). Tandem duplication is known to have important contributions to the chemoreceptor repertories in most insect genomes, particularly so in Hymenoptera. Indeed, we found the largest known chemoreceptor gene clusters in the leafcutter ant A. cephalotes (a cluster of 68 Ors) and the fire ant (a cluster of 66 Grs). The formation of tandem clusters may accelerate the turnover of chemoreceptors as one duplication (or loss) event can involve more than one gene simultaneously (see one such example in the bumblebee genome).

16Our phylogenetic analyses of the Or and Gr genes in 13 solitary and social hymenopterans (fig. 1) found 30 and 11 well-supported subfamilies (supplemental table S1, Supplementary Material online) respectively, each representing potentially one or a few genes in the common ancestor of the suborder Apocrita within Hymenoptera (wasps, bees, and ants). These include all previously reported subfamilies and also several additional subfamilies that were previously orphan lineages. We then reconstructed the histories of gain and loss for Or and Gr genes to examine the dynamics of chemoreceptor gene numbers in solitary and social hymenopterans (fig. 1A and B; supplementary table S2 and S3, Supplementary Material online). Consistent with our previous study, the results showed that the respective ancestors of Apocrita, Aculeata (bees and ants), the three solitary wasps, and the two bees all had relatively small number of chemoreceptors, suggesting that the many chemoreceptors found in wasps and bees were mostly derived from lineage-specific expansions. Similarly, the ants with higher numbers of Grs all showed clear patterns of recent duplications (fig. 1B). Our current analysis including additional early-branching ant species also confirmed the dramatic expansion of Ors inferred in the common ancestor of ants. Subsequent duplication/loss events occurred at high rates in all ant lineages despite their overall similar Or numbers (fig. 1A).

17The vast majority of gene gains and losses can be attributed to a few large subfamilies which also displayed divergent evolutionary patterns between species (fig. 1C and D; supplementary fig. S2 and S3, Supplementary Material online). For instance, the 9-exon subfamily is the largest Or subfamily in ants, halictid bee, and jewel wasp, containing at least 25% of the Ors in each of these species (fig. 1C). However, the honeybee has many more Ors in the subfamily L, and the other wasp M. demolitor has 50% less 9-exon Ors compared with N. vitripennis. Even between ants, the number of 9-exon Ors can differ dramatically (e.g. 104 in A. cephalotes vs. 164 in C. biroi). Additionally, ants, bees, and wasps each showed lineage-specific expansions in the subfamilies U, J, and F, respectively (fig. 1C). Similarly, while the majority of ant Gr duplications was harbored in the subfamilies F and Gr8/9, expansions of wasp Grs occurred almost exclusively in the subfamily Gr4 (fig. 1D). Particularly, a S. invicta-specific expansion included 79 Grs in the subfamily E where all other species including ants have either none or only one copy (fig. 1D). In sum, our results revealed highly diverse patterns of gene gain and loss among subfamilies of chemoreceptors and among species.

18To investigate the sequence level evolution of chemoreceptors, we first examined the global selective pressure (the ratio of nonsynonymous/synonymous substitution rates, or ω) for all subfamilies and revealed a diverse range of ω values (supplementary table S4, Supplementary Material online). In both Or and Gr families, the strongest levels of purifying selection were found in subfamilies that are strictly single-copy in all species, namely Orco (ω = 0.044), which encodes the coreceptor required for Or function, and Gr1/Gr2 (ω = 0.105 and 0.155), which likely encode sugar receptors. Conversely, the highest ω values were displayed by the most expanded subfamilies in each family, including the Or subfamily 9-exon (ω = 0.384), as well as the Gr subfamilies E, F, and Gr8/9 (ω = 0.484, 0.500, and 0.474, respectively), suggesting increased levels of functional divergence in these subfamilies due to much relaxed selection.

19We then performed branch-site tests (see Materials and Methods) to detect episodic positive selection during chemoreceptor gene evolution in Hymenoptera. We quantified the respective proportions of branches under positive selection in wasps, halictid bee, honeybee, and ants, representing one solitary lineage and three independent origins of eusociality (table 2; supplementary tables S5 and S6, Supplementary Material online). Notably, halictid bee branches showed evidence of positive selection much more frequently in both Or and Gr families. The evolution of sociality in the halictid bee family is highly dynamic; there are multiple recent origins of eusociality, and social polymorphisms are often displayed by closely related species or even populations of the same species (e.g. in L. albipes). As such, halictid bees can be viewed to be at a relatively early stage of eusocial evolution. Our results suggest that chemoreceptors may have played an adaptive role in the establishment of eusociality in this lineage. Interestingly, one of the six genes under accelerated evolution between social and solitary L. albipes individuals is an Or.

20However, overall, ant Ors showed lower level of positive selection relative to solitary wasps (table 2), consistent with previous findings comparing the jewel wasp and two of the ants. In light of the predicted dramatic expansion of Ors in the ant common ancestor and the aforementioned results in L. albipes, we further characterized the pattern of adaptive evolution on branches representing the ant common ancestor (i.e. branches in gene phylogenies that correspond to the branch leading to the node “n22” in the organismal phylogeny shown in supplementary table S2, Supplementary Material online). Indeed, we found elevated level of positive selection on Ors during early ant evolution compared with the overall levels in ants or wasps (table 2; although not significant at 10% FDR level in the latter case). On the other hand, ant Grs showed positive selection mostly among recent duplicates, likely reflecting a greater role for Grs in species-specific adaptations.

21Honeybee chemoreceptors similarly displayed lower overall levels of positive selection than wasps (table 2). It is possible that the use of other communication mechanisms such as the dance language in honeybee reduced the adaptive pressure on chemoreceptors. Alternatively, chemoreceptors may have experienced greater levels of positive selection at particular stages in the evolution of honeybees, however the detection of such episodic events is difficult given that A. mellifera is the only species sampled in the family Apidae. Indeed, a recent population genomic study comparing four genetically distinct A. mellifera populations found Ors to be enriched among genes under positive selection during recent evolution of honeybees. At larger scale, the origin of eusociality and transition from simple to complex eusociality occurred multiple times in Apidae. Future analyses including more solitary and social members of Apidae will be necessary to better understand the evolution of chemoreceptors in this family.

22The recent availability of antennal transcriptome data from honeybee and several ants allows for a comparison of Or transcript abundance profiles between social lineages. We used the percentile rank of each Or’s transcript abundance for the cross-species comparison to account for differences in sample preparation, sequencing methodologies, and species. Interestingly, we found much higher overall Or abundances in the workers of all three ants relative to both young and old A. mellifera workers (median percentile rank: 59.0–68.1% vs. 40.7–41.3%; Wilcoxon rank-sum test P value < 1e-15 for all pairwise comparisons between ant and honeybee workers), while the two neuronal markers (Elav and Brp) displayed similar or lower abundances in ants compared with A. mellifera (fig. 2A; supplementary table S7, Supplementary Material online). The elevated Or abundances in worker antennae of ants, together with their substantially larger chemoreceptor repertoires, suggest a higher chemosensory ability of ant workers which is concordant with the more complex ant chemical communication systems. Additionally, in contrast to ants and other insects such as mosquitoes, the obligate Orco coreceptor is not the most abundant Or in A. mellifera worker antenna (fig. 2A), further indicating divergent Or expression between honeybee and ants.

23We further compared the sex-specific patterns of honeybee and ant Ors. A previous study based on microarray and quantitative-PCR has identified several worker-/male-biased A. mellifera Ors (AmOrs), and examination of ant Ors in the same subfamilies revealed both conserved and divergent patterns (fig. 2B and C; supplementary table S7, Supplementary Material online). Many ant Ors in the subfamilies H and P showed worker-enrichment, consistent with AmOr150/Or151 and AmOr63 in the respective clades (fig. 2B). Importantly, several AmOrs in the subfamily L were more abundant in males, including AmOr11 which encodes a receptor for (E)-9-oxo-2-decenoic acid (9-ODA), a major queen substance that functions in both attracting males and regulating worker reproduction. Although the use of 9-ODA seems highly specific to A. mellifera, a number of H. saltator and C. biroi Ors in this subfamily also showed strong enrichment in males (fig. 2C), suggesting their potential function as sex pheromone receptors as well. The lack of male-enriched subfamily L Ors in Camponotus floridanus, on the other hand, may indicate an alteration in its use and/or perception of sex pheromone (fig. 2C). Interestingly, unlike AmOr170, ant Ors in the subfamily A were mostly enriched in workers (fig. 2C), indicating a functional divergence between the two lineages.

24In this study, we have characterized a large number of Or and Gr genes in eight hymenopteran genomes; the annotations themselves are likely to be valuable resources for future comparative and functional studies. More importantly, the analyses of these chemoreceptors provide insights into their evolution and their roles in the evolution of social Hymenoptera. Our phylogenetic comparison of multiple solitary and social lineages indicates that an expanded chemoreceptor repertoire is likely a general feature of Apocrita without any exclusive association with sociality. At the same time, all social insects sequenced so far have complex chemical communication systems and moreover contain relatively large chemoreceptor repertoires (note that the damp-wood termite harbors a greatly expanded family of Irs instead of Ors and Grs). The net gain of more than 200 Ors in the ant common ancestor also represents one of the most striking expansions of insect chemoreceptors. Furthermore, we found elevated positive selection on chemoreceptors during the transition to eusociality in two independent lineages (halictid bees and ants), consistent with an adaptive role of chemoreceptors in social evolution.

25These findings raise the possibility that the frequent expansion of chemoreceptors, while allowing for rapid adaptation to changing chemical environments, also independently increase the probability of developing sophisticated chemical communication mechanisms compatible with the emergence of social life. For example, nestmate and caste recognition in social insects are largely mediated by the CHC profiles whose composition can be highly complex and vary qualitatively and quantitatively between species, colonies, and castes. Functional divergence between duplicated chemoreceptors may allow them each to be tuned for a subset of CHCs, thus collectively achieving a high specificity and accuracy in communication. Furthermore, as illustrated by the divergent patterns of gene expansion and expression revealed in our analyses, independently evolved social lineages often explore different subsets of chemoreceptors for their communication. This observation again highlights the possibility that the numerous chemoreceptor duplications provide the potential for complex social communications but the details of how these duplications have contributed to eusociality are likely to have varied in each case. Recent genome-wide comparisons among multiple ants and bees revealed a similar trend that independent origins of eusociality differ considerably in their underlying molecular details.

26Finally, our results suggest several future directions. First, functional information about chemoreceptors is critical to linking the evolution and expression patterns of these genes with the ecology of organisms. Such data are extremely scarce in Hymenoptera and there is an urgent need for relevant functional characterizations. Second, given that honeybee chemoreceptors displayed the lowest levels of episodic positive selection, it would be interesting to further investigate the patterns of selection in the very recently sequenced genomes of ten bees that encapsulate a range of social complexity. Lastly, as our results suggest a potential, adaptive role of chemoreceptors in the transition to eusociality, future studies comparing closely related species (or populations of the same species) that are socially polymorphic may provide further insights. Similarly, the comparison of populations exhibiting alternative social organizations (e.g. monogyne and polygyne forms of S. invicta) may help to elucidate the role of chemoreceptors in the further development of eusociality.

27Supplementary figures S1–S3 and tables S1–S10 are available at Genome Biology and Evolution online (http://www.gbe.oxfordjournals.org/).