In eusocial Hymenoptera (ants, bees and wasps), queen and worker adult castes typically arise via environmental influences. A fundamental challenge is to understand how a single genome can thereby produce alternative phenotypes. A powerful approach is to compare the molecular basis of caste determination and differentiation along the evolutionary trajectory between primitively and advanced eusocial species, which have, respectively, relatively undifferentiated and strongly differentiated adult castes. In the advanced eusocial honeybee, Apis mellifera, studies suggest that microRNAs (miRNAs) play an important role in the molecular basis of caste determination and differentiation. To investigate how miRNAs affect caste in eusocial evolution, we used deep sequencing and Northern blots to isolate caste-associated miRNAs in the primitively eusocial bumblebee Bombus terrestris. We found that the miRNAs Bte-miR-6001-5p and -3p are more highly expressed in queen- than in worker-destined late-instar larvae. These are the first caste-associated miRNAs from outside advanced eusocial Hymenoptera, so providing evidence for caste-associated miRNAs occurring relatively early in eusocial evolution. Moreover, we found little evidence that miRNAs previously shown to be associated with caste in A. mellifera were differentially expressed across caste pathways in B. terrestris, suggesting that, in eusocial evolution, the caste-associated role of individual miRNAs is not conserved.
Eusocial insects are characterized by exhibiting castes, with the queen caste being specialized for reproduction and the worker caste being specialized for non-reproductive, colony tasks1,2,3. The occurrence of castes represents a leading example of polyphenism, where multiple adult phenotypes develop from the same genotype as a result of environmental influences during development4,5. Explaining the molecular basis of polyphenism represents a fundamental challenge in biology. The eusocial Hymenoptera (ants, bees and wasps) are the largest and most diverse group of eusocial insects1. Accordingly, to address the molecular basis of caste development, several previous studies have used genomic approaches such as high-throughput sequencing to profile the genes associated with caste determination and differentiation in a range of eusocial Hymenoptera6,7,8,9,10,11,12,13,14.
While these studies have been highly successful at isolating protein-coding genes associated with caste determination and differentiation, the role of microRNAs (miRNAs) in the same processes has been comparatively understudied. MiRNAs are a class of regulatory short RNAs (21–23 bp) that target mRNAs following transcription and prevent them from being translated into proteins15,16. They have been shown to have essential roles in development, metabolism, and other processes in animals17. Recent research has also shown that miRNAs are associated with polyphenisms in non-social insects, such as winged and wingless morphs in the pea aphid, Acyrthosiphon pisum18, and that they regulate the transition between solitary and gregarious morphs in the locust Locusta migratoria19. In eusocial Hymenoptera, the potential role of miRNAs in caste development has been studied in the honeybee, Apis mellifera, where miRNAs have been shown to be differentially expressed in queen- and worker-destined larvae and pupae20,21,22 and experimentally implicated in caste determination23.
A. mellifera is a well-studied example of the advanced eusocial Hymenoptera, which are typified by queens and workers that are morphologically strongly differentiated and highly specialized for their respective roles. In primitively eusocial Hymenoptera, morphological differentiation of the castes is weaker and many differences between queens and workers are behavioural1. As a relatively primitively eusocial species, the bumblebee Bombus terrestris occupies an intermediate position with respect to caste differences, in that queens and workers are similar in external morphology and ovariole number but differ in body mass, physiology and behaviour, with workers being unable to mate24,25. The genus Bombus shares a primitively eusocial common ancestor with Apis26. The extant clades descended from this common ancestor are either advanced eusocial (Apini, comprising Apis; and Meliponini), primitively eusocial (Bombini, comprising Bombus) or non-eusocial (Euglossini)26. Hence Bombus represents the closest extant model for the primitively eusocial common ancestor with Apis, albeit one that, since the split, will have undergone independent evolution towards its own form of eusociality. For these reasons, B. terrestris makes an excellent system in which to investigate the molecular basis of queen-worker caste differentiation and determination, including the role of miRNAs, in the evolutionary trajectory between primitive and advanced eusociality.
In B. terrestris, queens and workers develop through four larval instars, identifiable by mass and head width27. Caste fate (adult phenotype) is determined, i.e. differentiation is initiated, early in the development of female larvae24,28. Larvae are irreversibly determined from a totipotent state (i.e. in which larvae are capable of developing into either caste) towards worker development provided that, during a sensitive larval phase for 3–5 days after hatching (in the second instar), the colony queen is present and emitting a pheromonal signal27,29,30,31. It is hypothesized that, later in the colony cycle, the queen stops producing the pheromone, so allowing female larvae to develop as queens30,31,32. Two key hormone groups involved in this process are juvenile hormone (JH) and the ecdysteroids, with high larval levels of each hormone being associated with queen development27,28,33,34. Queen-destined larvae undergo JH peaks in their second and third instars that delay larval moult, so extending their developmental time and leading to the greater adult body size of queens relative to workers27,28. However, some female larvae developing in the absence of queen pheromone may still become worker-like adults if, during the fourth instar, they receive insufficient food35. By contrast, worker-destined larvae lose totipotency and cannot change to the queen pathway once they have passed the second instar without a JH peak, i.e. if they have passed through the sensitive phase in the queen’s presence28.
Overall, this system of caste determination differs from that of A. mellifera, since, in B. terrestris, the queen’s presence (via queen pheromone) and amount of food jointly influence caste, whereas in A. mellifera totipotent female larvae must be fed specialized food (royal jelly) to develop as queens36. However, the extent to which the molecular basis of caste determination differs across the two taxa remains unclear. By profiling early- and late-instar queen- and worker-destined female larvae for miRNAs in B. terrestris, we were able, for the first time, to investigate the role of miRNAs in queen-worker caste determination in a relatively primitively eusocial species, to characterize miRNA profiles in larval development before and after the change from totipotency to fixity in caste fate, and to compare the roles of miRNAs across the gradient of eusociality represented by B. terrestris and A. mellifera.
SRNA-seq reveals caste-associated miRNAs in B. terrestris
We used deep sequencing of small RNAs (sRNA-seq) from queen- and worker-destined larvae that were collected from their colonies either while still totipotent (early-instar larvae) or after their caste fate was fixed (late-instar larvae). Each phenotype was sequenced in four biological replicates, comprising paired but separate pools of early- and late-instar larvae within each caste pathway from each of eight separate colonies, yielding 16 samples in total. We previously used worker-destined larvae of both instars to characterize 101 miRNAs in B. terrestris female larvae, of which 16 were new miRNAs (i.e. not isolated in any other organism) and 85 were shared with A. mellifera and so appeared conserved across the corbiculate bees37. In the present study, sRNA-seq of queen- and worker-destined B. terrestris female larvae returned six conserved miRNAs that had high normalized read counts (≥200 reads per 4 million reads) and that were differentially expressed between the caste phenotypes. These were Bte-miR-13a, Bte-miR-87a, Bte-miR-100, Bte-miR-306, Bte-miR-6001–5p and Bte-miR-6001-3p (Fig. 1; Supplementary Information), with the last two miRNAs being, respectively, the 5′ and 3′ arms of the Bte-miR-6001 duplex. Bte-miR-13a and Bte-miR-306 were most highly expressed in worker-destined larvae in early but not late larval instars (Fig. 1). Bte-miR-87a, Bte-miR-100, Bte-miR-6001-5p and Bte-miR-6001-3p were most highly expressed in queen-destined larvae in late but not early larval instars (Fig. 1). Both Bte-miR-6001-5p and Bte-miR-6001-3p showed the highest degree of differential expression (16.5 and 15.8 fold change between caste phenotypes, respectively), while the other four miRNAs showed only 2.2–3.2 fold change between expression levels between caste phenotypes (Fig. 1).
Northern blot validation of caste-associated miRNAs
We used Northern blots to validate the six miRNAs found to be differentially expressed between caste phenotypes on the basis of the sRNA-seq data. We also used Northern blots to validate four other miRNAs selected on the basis of homology to miRNAs associated with larval or pupal caste differences in A. mellifera20,21,22,23 (Supplementary Table S1) or differential expression between larval instars within castes in B. terrestris (Supplementary Information). Each phenotype was analysed in at least two of five biological replicates drawn from paired but separate pools of early- and late-instar larvae within each caste pathway from each of ten separate colonies. The patterns of expression of Bte-miR-6001-5p and Bte-miR-6001-3p observed in the sRNA-seq were confirmed by the Northern blots. Specifically, both miRNAs were more highly expressed in late-instar queen- versus worker-destined larvae (Fig. 2). By contrast, the Northern blots failed to validate the four remaining B. terrestris miRNAs that had been identified as differentially expressed between caste phenotypes by sRNA-seq (Fig. 2), i.e., while the miRNAs were expressed at a level detectable by Northern blot, the assay showed no evidence of their being differentially expressed between the caste phenotypes (Fig. 2). In addition, the Northern blots similarly failed to validate the four miRNAs associated with larval or pupal caste differences in A. mellifera or with differences between larval instars within castes in B. terrestris (Fig. 2).
We also used Northern blots to investigate the expression patterns of Bte-miR-6001-5p and Bte-miR-6001-3p in specific tissues of late-instar queen-destined larvae (head, digestive tract and outer cuticle) and in whole bodies of queen-destined larvae and pupae. We found that, in late-instar queen-destined larvae, both miRNAs were most highly expressed in outer cuticle (Fig. 3). In addition, both miRNAs were expressed at lower levels in late queen pupae compared to late-instar queen-destined larvae and early queen pupae (Fig. 3).
Genomic context of caste-associated miRNAs
Genome scanning for the gene bte-mir-6001 against the B. terrestris genome37 showed that this miRNA and its precursor sequence comprise the entire fourth intron of very high density lipoprotein (Vhdl, RefSeq accession number: NM_001331111.1; Fig. 4), which is a gene with homology to Vitellogenin. This signifies that this miRNA is a mirtron, i.e. an intronic miRNA that replaces the Drosha cleavage with splicing to generate the pre-miRNA38,39. A BLAST search of this sequence against the A. mellifera genome40 revealed that the mirtron was conserved in the same intron of the same gene in A. mellifera (Fig. 4).
Target prediction of caste-associated miRNAs
Target prediction for Bte-miR-6001-5p and Bte-miR-6001-3p showed that many putative targets for these miRNAs are genes involved in development and reproductive differentiation in Apis and Drosophila, including genes associated with ovary and oocyte development, neurodevelopment, larval development and larval moulting (Table 1). Specifically, predicted targets of Bte-miR-6001-5p included ecdysone-induced protein 75 and ferredoxin (Table 1).
We have isolated and validated two miRNAs (Bte-miR-6001-5p and Bte-miR-6001-3p) that are more highly expressed in queen- than in worker-destined late-instar larvae in B. terrestris (Figs 1and 2). This expression pattern suggests that these miRNAs are associated with queen-worker caste determination and/or differentiation in larvae. Such a role is consistent with expression differences in these miRNAs occurring in late- but not early-instar larvae (Figs 1and 2), i.e. after the sensitive phase of caste determination in the second instar in which worker-destined larvae lose totipotency28, and with the decline in their expression in late queen pupae (Fig. 3). It is also consistent with higher expression of Bte-miR-6001-5p and Bte-miR-6001-3p in the cuticle of late-instar queen-destined larvae (Fig. 3), given that differentiation of queens in B. terrestris involves delayed moulting of later larval instars28 and that, in the B. terrestris transcriptome, genes involved in larval cuticular biogenesis show elevated expression relative to other genes41. In addition, a role in caste determination for Bte-miR-6001-5p and Bte-miR-6001-3p matches the finding that the predicted targets of Bte-miR-6001-5p include ecdysone-induced protein 75 and ferredoxin, both of which have been associated with responses to ecdysone in Drosophila42,43. Ecdysone is an ecdysteroid, a well-known family of insect hormones that mediate many processes including caste determination in both A. mellifera44 and B. terrestris33.
A role for miRNAs in caste determination in B. terrestris and other eusocial Hymenoptera is consistent with miRNAs having been found to mediate the development of alternative adult phenotypes in other insects18,19, although miRNAs may not underpin behavioural caste differences in adults in other primitively eusocial Hymenoptera45. In B. terrestris, one possibility is that one or both arms of the Bte-miR-6001 duplex targets a protein-coding gene that prevents larvae from developing along the queen-destined pathway, so that, by silencing the gene, the miRNA allows larvae to develop as queens. The detailed mechanism by which this occurs remains open for future investigation, but could involve a peak in Bte-miR-6001 responding to the JH peaks exhibited by queen-destined larvae, or to the absence of queen pheromone, or the greater amount of food received by late-instar queen-destined larvae27,28,29.
Our finding that bte-mir-6001 is a mirtron within Vhdl suggests a novel link between miRNA regulation of caste determination and the Vitellogenin family of genes, including the gene Vitellogenin (Vg), to which Vhdl has sequence homology. Vitellogenins are an important class of nutritive proteins induced by JH and linked to reproduction in numerous insects46, including eusocial Hymenoptera. For example, in bees, vitellogenins are associated with ovarian activation in adult female Bombus47,48 and Apis49 and with larval queen-worker caste differentiation in Apis9. These associations suggest that Vhdl is also a candidate for a caste-associated gene in eusocial Hymenoptera in general.
Despite validating the sequencing results for the Bte-miR-6001 duplex, the Northern blots failed to confirm expression patterns for four other B. terrestris miRNAs that were differentially expressed according to sRNA-seq (Bte-miR-13a, Bte-miR-87a, Bte-miR-100, Bte-miR-306; Fig. 2). This could have occurred for several reasons. First, the sequencing results could have been false positives, with the Northern blot results reflecting a true lack of differential expression between the caste phenotypes. Second, it is possible that differential expression was present but that the low absolute levels of expression of these miRNAs (in contrast to those of the miR-6001 duplex) meant that differential expression was not detected by the Northern blots. Third, a high level of variation between replicates in the absolute level of expression of these four miRNAs could have masked any differential expression present. Previous studies50,51, including a study on miRNAs differentially expressed between Apis behavioural phenotypes51, have also found that Northern blots do not always show the same results as sRNA-seq analysis.
Our results show that the miRNAs associated with caste differentiation in Bombus and Apis exhibit little overlap. Specifically, the Northern blots showed that the homologues of four miRNAs associated with queen-worker caste determination in A. mellifera (Ame-miR-9a, -184, -71 and -275; Supplementary Table S1) were not differentially expressed between queen- and worker-destined larvae in B. terrestris (Fig. 2). Other studies have shown that, in A. mellifera, Ame-miR-6001-5p was not differentially expressed between queen- and worker-destined larvae21 and that Ame-miR-6001-3p was weakly differentially expressed between castes in larvae21 or was differentially expressed between female and male larvae but not between castes in larvae22. In sum, four out of four caste-associated miRNAs in A. mellifera that we investigated in B. terrestris lack such a role in B. terrestris, and at least one of two caste-associated miRNAs in B. terrestris lack this role in A. mellifera. Moreover, our previous profiling of miRNAs in B. terrestris showed that there was little overlap in the total set of miRNAs in the genomes of the two species, with A. mellifera having up to 103 miRNAs in its genome that are not found in the genome of B. terrestris37. Overall, therefore, our findings suggest that, within the bees, miRNAs are associated with caste determination and differentiation relatively early in eusocial evolution but that the role of individual miRNAs is not generally conserved as eusocial evolution proceeds. Hence, as well as the emergence of novel (taxonomically restricted) genes underpinning eusocial evolution across the Hymenoptera45,52,53, novelty of function could be a feature of the molecular basis of eusocial evolution. Further investigations, as well as focusing on the mechanism by which the Bte-miR-6001 duplex might affect caste, on its putative targets and on Vhdl, would therefore benefit from considering how its role has been modified during the change from primitive to advanced eusociality in bees.
We collected female larvae from B. terrestris colonies between May 2011 and December 2013 (see Supplementary Information for full details of all methods). We removed the queen from a sample of colonies to generate queen-destined larvae and we obtained worker-destined larvae from the remaining, queenright (with a queen) colonies. Caste fate of all sampled larvae was verified by tracking the caste fate of larvae retained in the same batch and allowed to develop to adulthood8. We thereby collected larvae with four different phenotypes: early- and late-instar worker-destined larvae and early- and late-instar queen-destined larvae. Colonies were separated into three cohorts: cohort 1 (8 colonies; Supplementary Table S2) was used to produce RNA for sRNA-seq; cohort 2 (10 colonies; Supplementary Table S3) was used to produce RNA for Northern blot validation; and cohort 3 (1 colony) was used to produce larval tissues (head, cuticle, digestive tract) and pupae (early-pupae and late pupae) for Northern blot investigation of miR-6001-5p and miR-6001-3p in queen-destined developmental pathways.
We pooled larvae that had the same phenotype and that were collected from the same colony within the total set of colonies, and then extracted total RNA from each pool using Trizol. Within each caste pathway, early- and late-instar larvae were sampled from the same colony; therefore instar stages were paired within colonies and each colony pool was a biological replicate for each phenotype. In this way, we created four biological replicates for each phenotype from cohort 1 for sRNA-seq (Supplementary Tables S2, S4) and five biological replicates from cohort 2 for Northern blots (Supplementary Tables S3, S5). We used RNA from the four biological replicates of the four phenotypes from cohort 1 to generate 16 cDNA libraries (Supplementary Table S4). We prepared the libraries using the TruSeq small RNA library preparation kit v.1.5 (Epicentre Technologies, Madison, Wisconsin, USA) with HD modifications to the 3′ adapter to reduce sequencing bias54. The libraries were sequenced using Illumina sequencing on the HiSeq2000 platform.
The 16 sRNA-seq libraries produced a total of 86 million reads (2.4–6.3 million reads per library; Supplementary Table S4). After excluding reads that contained unassigned nucleotides, we trimmed the 3′ adapter sequence and the HD signature from each read. We excluded sequences shorter than 16 nucleotides from further analysis. We mapped remaining reads to the B. terrestris genome v.1.037, full length, allowing one mis-match and no gaps, using the software PatMaN55. On average, 69.4% of the redundant reads (all reads) and 54.9% of the non-redundant reads (unique reads) mapped to the B. t. terrestris genome across all 16 libraries (Supplementary Table S4). As expected, a high average proportion of these genome-matching reads was incident to miRNAs (24.7% of the redundant reads, Supplementary Table S6). We normalized the 16 libraries using the read count per total normalization method56,57,58, with the normalization total for each library set at 4 million, which was the median total read count (rounded up to the nearest million) of the accepted reads across all libraries57. After checking the efficiency of the normalization, we excluded four libraries (one from each phenotype) from further analysis (see Supplementary Information). We aligned (full length, allowing up to two mis-matches and no gaps) the sequences to the Hexapoda miRNAs listed on miRBase v.2159. Following60,61, we identified the differentially expressed miRNAs using a maximal expression interval approach, and the degree of differential expression was determined using the log offset fold change (OFC) method across all libraries for pairwise comparisons of phenotypes. To isolate miRNAs associated with caste, we compared worker- and queen-destined larvae within each instar. To isolate miRNAs associated with development, we compared early- and late-instar larvae within each caste pathway. We defined miRNAs as differentially expressed when log2(OFC) ≥ 1 between phenotypes60.
We used miRanda62 to identify a list of potential targets of the two miRNAs that had a validated pattern of differential expression (Bte-miR-6001-5p, Bte-miR-6001-3p). We identified the putative function of each target using the descriptions presented in NCBI (http://www.ncbi.nlm.nih.gov/) and FlyBase (http://flybase.org/).
To validate the expression of miRNAs that were identified as differentially expressed between castes in cohort 1, we used Northern blots to probe for miRNAs in RNA extracted from pooled samples from cohort 2 (Supplementary Tables S5, S7). For each miRNA, we produced Northern blots comparing the expression in each phenotype in at least two biological replicates. The miRNA expression was considered validated if both biological replicates showed a pattern of differential expression that was the same as the sRNA-seq. In addition, we used Northern blots of RNA extracted from larval and pupal samples in cohort 3 to investigate the tissue- and stage-specificity of the two caste-associated miRNAs that were validated by the Northern blots (Bte-miR-6001-5p and Bte-miR-6001-3p). For all Northern blots U6, a stably expressed nuclear RNA, was used as a loading control.
The raw sRNA-seq data and raw counts have been submitted to GEO NCBI (GEO accession numbers: GSE64512 and GSE77870).
How to cite this article: Collins, D. H. et al. MicroRNAs Associated with Caste Determination and Differentiation in a Primitively Eusocial Insect. Sci. Rep. 7, 45674; doi: 10.1038/srep45674 (2017).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank Carly Hayes, Jacob Holland and Karim Sorefan for practical advice and help, and Pierre Blacher, Timothy Huggins, Marjorie Labédan, Eamonn Mallon, David Prince, Matthew Ronshaugen, and Anders Wirén for comments. This work was funded by a studentship from the University of East Anglia held by M.B., a studentship from the University of East Anglia and the U.K. Biotechnology and Biological Sciences Research Council (BBSRC) held by D.C. and by BBSRC grants to V.M. and T.D. (grant reference number BB/L021269/1) and A.B. and T.D. (grant reference number BB/M001482/1).
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/