Adaptive, caste-specific changes to recombination rates in a thelytokous honeybee population

Abstract

The ability to clone oneself has clear benefits—no need for mate hunting or dilution of one's genome in offspring. It is therefore unsurprising that some populations of haplo-diploid social insects have evolved thelytokous parthenogenesis—the virgin birth of a female. But thelytokous parthenogenesis has a downside: the loss of heterozygosity (LoH) as a consequence of genetic recombination. LoH in haplo-diploid insects can be highly deleterious because female sex determination often relies on heterozygosity at sex-determining loci. The two female castes of the Cape honeybee, Apis mellifera capensis, differ in their mode of reproduction. While workers always reproduce thelytokously, queens always mate and reproduce sexually. For workers, it is important to reduce the frequency of recombination so as to not produce offspring that are homozygous. Here, we ask whether recombination rates differ between Cape workers and Cape queens that we experimentally manipulated to reproduce thelytokously. We tested our hypothesis that Cape workers have evolved mechanisms that restrain genetic recombination, whereas queens have no need for such mechanisms because they reproduce sexually. Using a combination of microsatellite genotyping and whole-genome sequencing we find that a reduction in recombination is confined to workers only.

1. Introduction

Thelytokous parthenogenesis is the asexual reproduction of a female [1,2]. That is, a female lays an egg that results in a daughter without the involvement of a sperm. In the haplo-diploid Hymenoptera (ants, wasps and bees), thelytoky is typically automictic (meiosis-based) [2] with central fusion (figure 1). Meiosis I begins within the oocyte as it develops within the mother's ovary. After the egg is laid, the two diploid products of meiosis I divide again in meiosis II, resulting in four haploid pronuclei (pronuclei because no cell or nuclear membrane forms around each of the four sets of chromosomes). The two A pronuclei are derived from one product of meiosis I and the two B pronuclei are derived from the other. Shortly after the egg is laid the pronuclei align as A₁A₂B₁B₂ (figure 1) [4]. Remarkably, an A and a B pronucleus then fuse [4,5], restoring diploidy (figure 1), and allowing female development. Without the restoration of diploidy, the haploid egg would develop as a haploid male [6].

Figure 1.

Thelytokous parthenogenesis in A. m. capensis. (a) Central fusion without recombination results in retention of all heterozygosity present in the mother in offspring. (b) With frequent recombination, alleles are randomized between chromatids. This means that when any one chromatid is chosen at random, one of the other three is identical. This leads to the prediction that at any one locus there will be a 1/3 LoH between mother and daughter [3]. (Online version in colour.)

With this kind of fusion and in the absence of genetic recombination, all heterozygosity present in the mother is retained in offspring (figure 1). However, if there is genetic recombination, heterozygosity present in the mother can be lost in offspring. Extensive recombination randomizes the two maternal alleles among the four chromatids at meiosis I (figure 1). This randomization leads to the expectation of an average of 1/3 loss of heterozygosity between mother and daughter [3,7]. The 1/3 arises because once one of the four alleles present at a locus is chosen at random, one of the remaining three alleles is the same as the first.

An important model organism for the study of thelytoky in Hymenoptera is the Cape honeybee, Apis mellifera capensis, hereafter Capensis, which is endemic in the southern-most part of South Africa. In contrast with all other honeybee (sub-)species, Capensis workers typically produce thelytokous eggs that result in daughters [8,9]. This extremely unusual ability allows Capensis workers to be genetically reincarnated as a queen. Indeed, up to half of all Capensis queens are descended from workers, and many of the worker-mothers of queens are social parasites from other colonies [10,11].

Another remarkable manifestation of social parasitism in Capensis is ‘the Clone’. The Clone is a lineage of thelytokous clonal parasites. The current population is derived from a single Capensis worker that lived in 1990 [12,13]. The Clone infests the commercial beekeeping industry in the northern part of South Africa, which is based on another subspecies, A. m. scutellata (hereafter Scutellata). Clone workers enter Scutellata colonies, where they lay eggs that develop thelytokously into more Clones [14,15]. Eventually, the host colony fades away, and the Clones disperse to new host colonies.

Despite the obvious benefits of thelytoky to Capensis and other social insect workers, thelytoky comes at a potential cost. The extremely high rates of genetic recombination that are observed in honeybees [16] and other social insects [17] mean that the thelytokous daughters of workers should lose 1/3 of their heterozygosity relative to their mothers [7,18]. Homozygosity is a major problem for hymenopterans, particularly if they have a single-locus sex determination such as csd in the honeybee. csd must be heterozygous for expression of the female phenotype; individuals that are homozygous at csd are non-viable [6]. Therefore, because of the huge genetic load imposed by the need for heterozygosity at csd [19], one might predict that lower rates of recombination might evolve in workers relative to queens. Indeed, in Capensis, there is evidence that recombination rates are lower than predicted in the Clone [20,21], and the Clone population retains a high level of heterozygosity [12,22]. In the little fire ant, Wasmannia auropunctata, there is also a dramatic reduction in the recombination frequency and loss of heterozygosity (LoH) in the progeny of both thelytokous workers and queens [23].

In contrast with Capensis workers, mature, mated Capensis queens never reproduce thelytokously, even when they themselves are the clonal daughter of a worker [24,25]. Queens mate with many unrelated males, meaning that their female offspring are heterozygous and genetically diverse. Therefore, unlike Capensis workers, Capensis queens do not pay the cost of homozygosity in their offspring due to thelytoky and are under no selection to reduce the frequency of recombination.

Although Capensis queens always mate and reproduce sexually [24], virgin Capensis queens can be experimentally manipulated to reproduce thelytokously [26,27]. This provides an experimental device whereby the recombination rates in a thelytokous queen can be directly compared with those seen in thelytokous workers. We can thus directly test the hypothesis that Capensis workers have evolved mechanisms to curtail genetic recombination and thereby reduce LoH.

Here, we describe an investigation into the LoH in the progeny of virgin thelytokously laying Capensis queens, and thelytokous Capensis workers via a combination of microsatellite genotyping and whole-genome sequencing of the worker progeny. We predicted that in thelytokous workers, genetic recombination, as evidenced by LoH between worker-mother and worker-daughter, would be curtailed because LoH reduces worker fitness [20,28]. By contrast, we predicted that recombination rates in a thelytokous queen, which do not occur in nature, would be similar to those seen in sexual queens. If so, this would indicate that recombination rates are adaptively tuned in Capensis in a caste-specific manner.

2. Material and methods

(a) . Biological material

Fieldwork was performed at the Plant Protection Research Institute in Stellenbosch South Africa. To obtain genetically uniform workers, we instrumentally inseminated [29] a Capensis queen with the semen of a single Capensis male, retaining the cadaver of the male for later sequencing. We then introduced a brood comb holding several hundred eggs laid by this queen into an A. m. scutellata (Scutellata) colony to be reared. The Scutellata colony had been transported from Douglas, 500 km to the north of Stellenbosch. Scutellata workers over-feed Capensis larvae due to inappropriate signalling between larva and nurse worker [30,31]. This over-feeding generates the large workers that are typical of the daughters of Capensis social parasites that are reared in parasitized Scutellata colonies [30,31]. After 17 days, we placed the comb, now with Capensis pupae ready to emerge, in an incubator overnight at 34.5°C. We individually marked workers that were less than 18 h old and added 1–2 individuals to each of four micro-colonies comprising about 500 queenless Scutellata workers. To reduce the possibility of reproductive parasitism by Capensis workers from other colonies, we confined the micro-colonies to a screen cage that was 10 m × 10 m × 2 m and fed them sugar candy and pollen ad libitum. After 7 days, we culled any extra workers so that there was one marked Capensis worker present in each colony. However, in one colony, we missed the second marked worker, but the progeny of the two workers were easily distinguished genetically. After 14 days, the Capensis workers commenced laying and pheromonally suppressed ovary activation in their host workers. We successfully harvested four laying Capensis workers and a total of 63 of their larval progeny into ethanol for later genotyping.

To obtain the progeny of a thelytokous queen, we reared 10 Capensis queen pupae by standard methods [29]. When the pupae were about to eclose we introduced them to individual nucleus colonies containing roughly 3000 Capensis workers. We ensured that every hive was bee tight by sealing the lid to the box with duct tape and blocking the entrance with a queen excluder grid to prevent the virgins from taking mating flights.

Virgin honeybee queens can be induced to lay eggs by double narcosis with carbon dioxide [32]. Virgin Capensis queens so treated produce a high proportion of thelytokous progeny [25–27,33]. However, CO₂ narcosis has a number of physiological [34] and gene expression [35] effects on insects in general and honeybees in particular. We therefore chose to induce oviposition by sham mating, which induces oviposition in a small proportion of queens [36].

When the queens were 5 days post eclosion we constrained them in an artificial insemination apparatus [37] without narcosis. We then glued a 5 mm piece of surgical tape (Micropore, 3M, Minnesota) over the sting chamber using nail varnish (figure 2) [36]. We fitted each nucleus colony's entrance with a tunnel with a transparent top and a removable queen excluder grid [38]. This allowed us to confirm the departure and return of queens as they went on mating flights, and whether their ‘chastity belts’ (CB) were intact after each flight (figure 2).

Figure 2.

Sham mating in A. m. capensis queens. (a) A Capensis queen being fitted with a chastity belt. (b) A queen marked on the thorax with pink paint returning from her mating flight with chastity belt in place. (Online version in colour.)

We monitored the tunnels each afternoon, the natural mating flight time, until each virgin queen was observed trying to leave her colony. Once a queen was observed at the entrance with her CB in place, we allowed her to fly. If the CB had been dislodged, we did not allow the virgin to fly and replaced her CB the next morning. If a queen was observed to fly for more than 10 min, we removed the CB the next morning and clipped her wing so that she could not fly again. We monitored the queens closely for the next two weeks, to determine if and when oviposition had commenced. We collected larvae as soon as they appeared into ethanol.

Not all queens flew, not all returned from mating flights, and not all laid. In the end, we were able to harvest one queen and 25 of her larval progeny into ethanol.

(b) . Measuring loss of heterozygosity

For our microsatellite analysis, we extracted DNA from each individual using Chelex [39]. We genotyped the sham-mated virgin queen and her 25 larvae at nine unlinked microsatellite loci (primers in electronic supplementary material, table S1). We ensured that each larva was the thelytokous offspring of the resident queen by confirming that none carried alleles not found in the queen. We also checked that larvae were heterozygous at at least one locus, thereby ensuring that no haploid males were included in our survey.

To quantify LoH per larva, we compared the offspring genotype to the queen's genotype at each locus and calculated $H_M-H_O/H_M$, where H_O was the number of loci that were heterozygous in the offspring and H_M the number that were heterozygous in the mother. We then averaged LoH over progeny.

We followed the same procedures for the progeny of workers. Two of the 63 larvae harvested were discarded because genotyping revealed that they were the progeny of host workers (electronic supplementary material, table S2). We then calculated LoH in the remaining 61 larvae, the progeny of our four workers at the same nine unlinked microsatellite loci used for the queens.

To determine whether thelytokous virgin Capensis queens differ in their recombination rates depending on whether they are subjected to CO₂ narcosis, we recalculated LoH in 42 worker progeny of a thelytokous virgin queen treated with CO₂ from a previous study [26] as above. The loci used here were all on chromosome 1, but were more than 50 centimorgans apart and unlinked to the sex locus or centromere. This means that each locus provided an independent test of LoH.

Microsatellite data provide an estimate of LoH for a tiny fraction of the genome. To provide a broader picture of the frequency and localization of LoH over the entire genome, we used whole-genome sequencing to assess LoH from SNP data [22]. We selected thelytokous worker A, 14 of her larval progeny, and her father for whole-genome sequencing. Genomic DNA of the 16 samples was extracted using a standard phenol/chloroform protocol. Genomic libraries were prepared with the Illumina Nextera DNA Flex Library Prep protocol and sequenced on one lane of Illumina NovaSeq 6000, S1 300 Cycle (150 bp paired end) at the Australian Genome Research Facility in Melbourne.

We identified heterozygous SNPs in the mother-worker and SNPs that were no longer heterozygous in her thelytokous offspring using the bioinformatic pipeline [22] described in electronic supplementary material, Methods. Briefly, we estimated the LoH in each offspring larvae as the number of homozygous SNPs in the larva that had been heterozygous in the mother-worker, divided by the total number of SNPs that had been heterozygous in the mother and overlapped with that particular offspring.

We then used BEDtools to create 10 kb windows along the entire genome [22]. We then retrieved, for each individual, the number of heterozygous and homozygous SNPs in each window, only considering windows that had at least 10 SNPs. We calculated the level of heterozygosity per window as the number of heterozygous SNPs divided by the total number of SNPs in each window. Windows with a heterozygosity level of zero were deemed homozygous. All other windows were deemed heterozygous. We also identified any 10 kb windows that had consistently retained heterozygosity in the larval progeny.

3. Results

Six of nine microsatellite loci were heterozygous in the thelytokous virgin queen that had sham-mated. Of the loci that were heterozygous in this queen, an average of 5.4 (range 5–6) were scorable in the individual larva (electronic supplementary material, table S2). The average LoH in larval progeny relative to their thelytokous queen mother was 45.87% (range 0–100%, s.e. ± 6.3%) (electronic supplementary material, table S2). We attribute the range to the high variability among individuals as a consequence of the stochasticity of recombination events. Although they cannot be compared directly because different microsatellite loci were used, this high frequency of LoH is qualitatively similar to those seen in virgin queens driven to thelytoky by CO₂ narcosis (electronic supplementary material, table S3) [26]. Here, the average LoH was 32.0% ± 3.8%, range 0–100%.

Worker A produced 28 progeny. The average number of scoreable loci that were heterozygous in A and scoreable in her larvae was 6.96 (range 6–7 per larva). Worker B produced two progeny with seven informative loci for both larvae. Worker C produced 20 larvae with nine informative loci per larva. Worker D produced 11 larvae with seven informative loci per larva. The average LoH in larvae compared to their four mothers was 0.47% ± 0.32% (electronic supplementary material, table S2) an average of a 100 times less than that observed in the sham-mated queen (contingency table test of the proportions of lost and retained heterozygosity in the progeny of workers and the queen, ${\chi }_1^2=228.04$, p < 0.001).

We obtained an average sequencing depth of 36.4 ± 0.9 (mean ± s.e) across the 15 individuals chosen for whole-genome sequencing (electronic supplementary material, table S4). The mother-worker had 993 778 heterozygous SNPs. A slightly larger number were observed across the 14 offspring larvae: 1 113 965.2 ± 7110.5 heterozygous SNPs on average, a consequence of slightly lower coverage in the parent. The number of SNPs that were heterozygous in the mother-worker and homozygous in individual offspring larvae was 5076.6 ± 1955.0 SNPs on average, which corresponds to a genome-wide LoH of 0.55 ± 0.21% (table 1), similar to the rate estimated from microsatellites (0.47%). The average number of 10 kb windows that were heterozygous in the mother-worker but homozygous in offspring larva was 162.7 ± 64.0. Thus, a window-level analysis provides a similar estimate of LoH (0.99 ± 0.39% on average; electronic supplementary material, figure S1; figure 3; table 1) to individual SNPs.

Table 1.

Minimal LoH in the daughters of thelytokous workers based on SNP genotyping. LoH can only be computed for SNPs with sufficient coverage in both mother and daughter. This is why the reported counts for the mother are slightly different for each daughter, and the total number of SNPs observed in the mother exceeds those reported here.

larva	individual SNPs				10 kb windows
	heterozygous in mother-worker	heterozygous in daughter	homozygous in daughter	loss of heterozygosity (%)	heterozygous in mother-worker	heterozygous in daughter	homozygous in daughter	loss of heterozygosity (%)
1	944 538	944 148	2053	0.04	16 979	16 970	8	0.05
2	946 510	946 129	1930	0.04	16 983	16 974	10	0.05
3	932 857	932 513	2023	0.04	16 976	16 967	10	0.05
4	958 810	958 311	1996	0.05	16 985	16 978	8	0.05
5	943 488	935 767	10 463	0.82	16 774	16 561	200	1.27
6	934 453	926 883	10 619	0.81	16 773	16 561	202	1.26
7	953 237	952 779	1936	0.05	16 978	16 968	9	0.06
8	949 739	949 246	2029	0.05	16 977	16 965	14	0.07
9	964 046	963 510	1999	0.06	16 985	16 976	10	0.05
10	922 676	901 403	26 044	2.31	16 330	15 671	641	4.04
11	952 987	949 172	5874	0.40	16 822	16 652	166	1.01
12	909 825	903 299	9645	0.72	16 744	16 505	232	1.43
13	924 394	924 021	1985	0.04	16 977	16 968	10	0.05
14	910 838	890 144	25 590	2.27	16 267	15 556	690	4.37

Figure 3.

Level of heterozygosity per 10 kb window in 14 thelytokous daughters of one Capensis worker. Heterozygosity in each window ranges from 0 (complete homozygosity, blue) to 1 (complete heterozygosity, red). Each chromosome in the outer circle is represented with a different colour. Scale represents chromosome size (Mb). The first concentric circle is the mother-worker. Each of the 14 inner concentric circles (below the dashed line) is a different progeny of the mother. Grey areas correspond to centromeric regions. There is minimal loss of heterozygous SNPs from the mother to daughter. (Online version in colour.)

The proportion of homozygous SNPs that fell within homozygous windows was 2.97 ± 0.71%. This is not surprising because more than 98% of windows were heterozygous. Within heterozygous windows, homozygous SNPs were mostly contiguous: 69.36 ± 0.92% were adjacent to at least one other homozygous SNP and only 30.64 ± 0.92% were singletons. Thus, the rare SNPs that lost heterozygosity seemed to do so in unison with adjacent SNPs. This is probably a signature of rare gene conversion rather than recombination events [40].

In addition to the presumptive gene conversion events, we also observed what appears to be evidence of terminal recombination events. Electronic supplementary material, figure S1 shows regions where heterozygosity had been lost at the ends of chromosomes 1, 2, 4, 10 and 14 in seven individual larvae. The probability of recombination increases significantly with distance from the centromere because LoH in any particular window was significantly associated with its distance from the centromere (GLMM: estimate = 1.05, s.e. = 0.04, Z = 28.10, p < 0.00001). By contrast, there is no evidence of double recombination events, which result in blocks of homozygosity within chromosomes [22].

There were 17 106 windows overlapping across the mother-worker and the 14 offspring larvae. Of these, 15 180 (88.74%) were consistently heterozygous in all individuals (electronic supplementary material, figure S2). On average, the proportion of windows heterozygous in the mother-worker and also her offspring larva was 98.04 ± 0.37%. The proportion of genes intersecting with heterozygous windows in both the mother-worker and each offspring larva was 97.54 ± 1.46%. Out of 9573 genes found in all individuals, 8405 (87.80%) intersected with heterozygous windows across all individuals. Thus, the vast majority of the genome, including coding regions, retained heterozygosity across all offspring.

4. Discussion

The striking conclusion of this study is that thelytokous Capensis queens have high levels of genetic recombination, as evidenced by LoH in progeny, and similar levels to those seen in queens that are reproducing sexually [41–43]. This recombination leads to approximately 1/3 LoH between queen and daughter. By contrast, thelytokous workers have a 100-fold reduction in LoH relative to queens and therefore retain most of the heterozygosity present in their mother. This supports our hypothesis that Capensis workers have evolved caste-specific mechanisms to curtail recombination during thelytokous parthenogenesis as an adaptation to social parasitism. CO₂ exposure is not the cause of high recombination rates in virgin Capensis queens; our virgin queen that was not exposed to CO₂ had a LoH similar to that of the virgin exposed to CO₂. The death of recombinant progeny is not the cause of the low LoH seen in the progeny of workers [22,26,44]. If the death of recombinants were the cause of the observed low LoH in the progeny of thelytokous workers, then a similar reduction should operate in the progeny of thelytokous queens.

Why should workers be so different from queens? Without a reduction in recombination, frequent LoH leads to reduced viability in offspring [3,21,28,44,45]. Since parasitic workers must compete with hundreds of others for access to brood cells, any reduction in offspring viability will be subject to strong negative selection. By contrast, Capensis queens benefit from recombination and sexual reproduction because it diversifies their offspring. Since Capensis queens always mate, we assume that there is no selection pressure against a high recombination frequency in queens. It therefore seems that Capensis has evolved a caste-specific regulation of recombination rate with low recombination in workers and high recombination in queens.

Inspection of electronic supplementary material, figure S1 shows that in all 14 larvae across all 16 chromosomes there were only seven large blocks of homozygosity (indicated in orange in electronic supplementary material, figure S1). In each case the loss of homozygosity was telomeric; no blocks of homozygosity appear within a chromosome. This suggests that in the seven chromosomes where a recombination event occurred there was only one recombination event on that chromosome and it was teleomeric. We know this because if two recombination events occurred the block of homozygosity is expected internal to the ends of the chromosome (figure 1 for the expectation). This indicates an average frequency of 7 crossovers/(16 chromosomes × 14 larvae) or 0.031 crossovers per chromosome.

Interestingly, physical cross overs are generally reckoned to be essential, at least in mammals and yeast, for proper chromosome alignment and disjunction [46,47]. Therefore, the extremely low frequency of crossovers in Capensis workers represents yet another intriguing mystery of their genetics. The pattern of LoH within chromosomes is more consistent with gene conversion of very short genomic sequences, rather than chromosomal breakage and exchange [40]. This pattern is not unexpected because gene conversion that is not associated with physical crossing over is common in honeybees [43].

How does the difference in recombination frequency between workers and queens come about? The gene GB45239 (hereafter Thelytoky) on chromosome 11 is the major determining factor of the thelytokous phenotype in Capensis workers [48]. Workers that are homozygous recessive (th/th) express the thelytokous phenotype. th is close to fixation in the Capensis population, but almost absent in all other honeybee populations [48]. Capensis workers have lower expression of Thelytoky than other bees [48]. However, it is as yet unclear how Capensis queens, which are also th/th, avoid thelytokous reproduction once mated [25]. Since thelytokous Capensis queens and workers carry the same mutations at the thelytoky locus [48], but differ in their recombination rate (this study), thelytoky itself is probably not responsible for the low recombination rate in Capensis workers. We also note that a gene that influences cytokinesis in second division meiosis is unlikely to influence the frequency of recombination in first division meiosis. Rather, it seems that a genetic module that is unrelated to thelytoky is switched on in workers only, and that this module specifically reduces the frequency of recombination [21].

What could this module be? Wallberg et al. [49] identified 11 genomic regions, in addition to GB45239, that show significant differentiation between the Capensis population and the neighbouring subspecies A. m. scutellata. Any of these 11 regions may be involved in the caste-specific regulation of recombination frequency in Capensis. We predict that like GB45239, the genetic element that regulates recombination will show strong caste-specific expression.

We have previously argued that it is selection against homozygous recombinants, and not reduced recombination rates, that maintain heterozygosity in thelytokous social insect populations [12,20,22,44]. In particular, in the Clone, a break-away from the queen-sexual/worker-asexual Capensis population, the degree of homozygosity decreases as eggs develop into larvae, and larvae develop into pupae, providing strong evidence of selection against homozygosity (i.e. homozygous individuals die as they mature [44]). Clones have recombination rates that are intermediate [44] between the extremely low recombination rates seen in Capensis workers from the queen-sexual population and those seen in queens (this study). Given the rates of recombination seen in the Clone population, selection against homozygotes must contribute to the maintenance of heterozygosity within the lineage, or heterozygosity would have been completely lost many generations ago [20–22]. Nonetheless, the death of homozygotes cannot explain the retention of heterozygosity in workers of the normal Capensis population studied here. If the selection was the sole determinant of retention of heterozygosity, then we would expect that regions that are not subject to over-dominant selection would have lost heterozygosity [22]. No regions where heterozygosity was frequently lost were observed in our 14 worker sequences.

Why does the Clone population have recombination rates that are intermediate between Capensis workers from the queen-sexual population and those seen in queens? It is possible that moderate recombination frequencies have reduced the virulence of the Clone and contributed to its persistence [28,50]. If so, it is interesting to speculate as to how this may have come about. Potentially there is a partial loss-of-function mutation in the module that increases recombination in the Clone, that has not been removed by selection. A comparison between Clone, Capensis and Scutellata genomes may help identify the region(s) that control(s) the frequency of recombination. We predict that this region will show high pairwise F_ST between the sexual Capensis, the Clone and Scutellata populations and functional mutations between the Clone and sexual Capensis populations. If a single gene is involved, it would be differentially expressed between thelytokous workers and queens.

Finally, we note parallels between Capensis and another thelytokous social insect, the little fire ant (Wasmannia auropunctata). W. auropunctata populations come in two forms. In invasive populations queens produce offspring queens asexually via thelytoky with central fusion. By contrast, in native-range populations in French Guiana, nearly half of queens are sexual [51,52]. Genetic recombination is severely restricted in thelytokous populations, whereas it is frequent in obligately sexual populations [23]. This again suggests that when isolated populations of social insects evolve thelytokous parthenogenesis, be they queens or workers, a parallel reduction in recombination frequency is also necessary to prevent accumulation of homozygosity and inbreeding.

Data accessibility

Microsatellite genotype data are given in the supplementary material (tables S2 and S3) [53]. WGS data for all samples have been deposited to the NCBI Sequence Read Archive under accession number PRJNA592273.

Authors' contributions

B.P.O.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing-original draft, writing-review and editing; MJ.H.: data curation and investigation; G.B.: data curation and investigation; B.Y.: conceptualization, data curation, formal analysis, software, validation, visualization, writing-original draft, writing-review and editing; M.H.A.: conceptualization, investigation, methodology, resources, writing-review and editing; A.Z.: conceptualization, funding acquisition, writing-review and editing; M.B.: conceptualization, investigation, writing-review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This study was supported by the Australian Research Council grant no. DP180101696 to B.P.O. and A.Z.