Male spiders control offspring sex ratio through greater production of female-determining sperm

Abstract

Sex allocation theory predicts that when sons and daughters have different reproductive values, parents should adjust offspring sex ratio towards the sex with the higher fitness return. Haplo-diploid species directly control offspring sex ratio, but species with chromosomal sex determination (CSD) were presumed to be constrained by Mendelian segregation. There is now increasing evidence that CSD species can adjust sex ratio strategically, but the underlying mechanism is not well understood. One hypothesis states that adaptive control is more likely to evolve in the heterogametic sex through a bias in gamete production. We investigated this hypothesis in males as the heterogametic sex in two social spider species that consistently show adaptive female-biased sex ratio and in one subsocial species that is characterized by equal sex ratio. We quantified the production of male (0) and female (X) determining sperm cells using flow cytometry, and show that males of social species produce significantly more X-carrying sperm than 0-sperm, on average 70%. This is consistent with the production of more daughters. Males of the subsocial species produced a significantly lower bias of 54% X-carrying sperm. We also investigated whether inter-genomic conflict between hosts and their endosymbionts may explain female bias. Next generation sequencing showed that five common genera of bacterial endosymbionts known to affect sex ratio are largely absent, ruling out that endosymbiont bacteria bias sex ratio in social spiders. Our study provides evidence for paternal control over sex allocation through biased gamete production as a mechanism by which the heterogametic sex in CSD species adaptively adjust offspring sex ratio.

1. Introduction

Equal production of male and female offspring is generally considered an evolutionarily stable strategy [1,2], however, when sons and daughters have different reproductive values, sex allocation theory predicts parents to adjust offspring sex ratio towards the sex with the higher fitness return [3–6]. While adaptive sex ratio bias is well understood in haplo-diploid systems, species with chromosomal sex determination (CSD) have been presumed to be constrained by Mendelian segregation in their ability to strategically adjust sex ratio [3]. There is accumulating evidence against this idea [7,8], but we still have limited understanding of possible sex biasing mechanisms that are mainly limited to theoretical hypotheses (such as steroid and glucose levels [9–14]), and almost no empirical evidence for mechanisms through which CSD species with adaptive sex ratio bias exert control over offspring sex. One hypothesis proposes that control over offspring sex is more likely to evolve in the heterogametic sex through a bias in the production of gametes [15,16].

A striking example of CSD species with highly female-biased sex ratio is found in social spiders, which are cooperative breeders that live in permanent groups with an obligatory inbreeding mating system [17–20]. Social spider groups are usually composed of around 85% females [17], and female bias is considered adaptive by reducing competition for fertilization between brothers (local mate competition [5]) and/or increasing the number of females as the helping sex (local resource enhancement [2,4,6,21,22]). However, the mechanism of sex ratio bias is unknown. By contrast, closely related subsocial species that cooperate in the juvenile stage but disperse prior to reproduction show equal sex ratios [17]. In most spiders, sex is determined by an X0 sex chromosome system where males are heterogametic with one copy of the X chromosome (X0) and females are homogametic (XX) [23]. Males produce two types of sperm cells, female-determining with X chromosomes (X-sperm) and male-determining without X chromosomes (0-sperm). These sperm types are expected to be produced in equal numbers after Mendelian segregation during meiosis. However, a bias towards X-sperm in social spiders would produce more female offspring and thereby function as a mechanism for adjusting offspring sex ratio. To test this hypothesis, we used flow cytometry to quantify the proportion of X- and 0-sperm cells from two social spider species of the genus Stegodyphus; Stegodyphus dumicola and Stegodyphus mimosarum with female-biased sex ratio [17]. We also included one subsocial species Stegodyphus africanus with equal sex ratio [24] as control.

Female-biased sex ratio in arthropods, including spiders, can also occur through inter-genomic conflict caused by infection with certain endosymbiotic bacteria ([25–32], but see [33]). The effect of a cytoplasmic distorter on offspring sex ratio can vary with environmental (i.e. temperature) and genetic (i.e. host suppression) effects that influence the bacteria's abundance [34]. To control for potential confounding effects caused by endosymbiont bacteria, we screened the microbiome of the two social species (S. dumicola, S. mimosarum) for the five endosymbiont genera known to be able to cause female-biased sex ratio, i.e. Wolbachia, Rickettsia, Candidatus Cardinium, Spiroplasma and Arsenophonus.

2. Methods

(a). Biased X-sperm proportion

(i). Flow cytometry

We quantified the proportion of male- (0-sperm) and female-determining sperm (X-sperm) using flow cytometry. Using a DNA stain, flow cytometry allows the automated determination of the DNA content of thousands of sperm nuclei, based on fluorescence intensity [35,36]. Males from two social species (S. mimosarum and S. dumicola) and one subsocial species (S. africanus, sister species to S. mimosarum [37]) were used. Spiders load sperm into reproductive organs called pedipalps, which are used for external transfer of sperm to the female sperm storage organ [38] (pedipalps are indicated by arrowheads in figure 1). Sperm present in the pedipalp, therefore, represents the male ejaculate. Both live males and males that were stored at −80°C were included (see the electronic supplementary material, table S1 for sample details). Previous analysis revealed that, next to haploid sperm cells, substantial amounts of diploid cells are also present in the pedipalp [39]. In order to correctly identify sperm cells we, therefore, included a leg sample where diploid somatic cells are present but haploid sperm cells are absent (electronic supplementary material, figure S1). DNA of the isolated nuclei was stained with propidium iodide (PI) using the protocol described in Vanthournout et al. [39] (adapted from [40] and [41]). Preparations were stored at 4°C for up to 2 h and protected from light using tin foil. DNA-content analysis of prepared nuclei was performed on a BD Biosciences FACSaria flow cytometer and Fortessa (Argon laser emitting at 488 nm).

Figure 1.

Dot plot of propidium iodide-stained sperm nuclei (PI-A, corresponding to DNA content) and forward scatter of the nucleus (FSC-A, corresponding to particle size) + PI-A histogram, isolated from one pedipalp of one social S. dumicola (a), subsocial S. africanus (b) and social S. mimosarum (c) male. For S. dumicola, a graphical representation of a male spider, with pedipalps indicated by arrowheads, is added. Circles indicate populations of X-sperm (top circle) and 0-sperm (bottom circle). Circles are used for illustrative purposes only and are not used in estimating density curves. Estimated proportion of X-sperm for the three individual males is 0.77, 0.51 and 0.63, respectively. Sperm sample sizes consist of 1620 (S. dumicola), 5049 (S. africanus) and 959 (S. mimosarum) sperm nuclei. (Online version in colour.)

(ii). Data analysis

We visualized the PI intensities (representing DNA amount) as a function of the forward scatter (FSC) (representing cell size) using FCS Express 6 (DeNovosoftware). We manually selected populations representing sperm nuclei (see the electronic supplementary material), and exported the data into a text file. We observed a positive correlation between FSC and PI intensity, and corrected the PI intensity values according to the strength of the correlation (for sperm scatter plots, see data accessibility section). We estimated the proportion of X- and 0-sperm using the normalmixEM function (with the following specifications: k = 2, mean.constr. = NULL, sd.constr = NULL) in R. This function allows us to fit two normal distributions (representing 0-sperm and X-sperm) to the PI intensity [42] and hence provides estimates of sperm proportions. This ensures an objective approach by taking into account potential overlap in X- and 0-sperm populations. We bootstrapped the PI intensities from each sample 100 times, and ran the normalmixEM on the bootstrapped datasets to obtain confidence intervals (CIs). Because this approach produces point estimates of sperm proportions, these are not expected to follow a binomial distribution. Indeed, normality tests (PROC UNIVARIATE, SAS v. 9.4, SAS Institute Inc. (2002–2012)) indicated that the distribution of the point estimates did not deviate significantly from a normal distribution (see the electronic supplementary material). For this reason, we performed a one-sample t-test (assuming normal distribution, PROC TTEST, SAS v. 9.4, SAS Institute Inc. (2002–2012)) on the point estimates to test whether sperm ratios within a species are significantly different from 0.5. A generalized linear model (assuming normal distribution, PROC GLM, SAS v. 9.4, SAS Institute Inc. (2002–2012)) was used to analyse the differences in X-sperm production between species. If two pedipalps were analysed of the same male, the average proportion was used.

(b). Microbiome

(i). Small dataset: Ion Torrent

16S rRNA amplicon libraries were prepared for three adult females from each of the social spider species S. mimosarum and S. dumicola, along with one individual of the solitary dwarf spider Oedothorax retusus originating from a laboratory-reared line that is confirmed to be infected with Wolbachia, Rickettsia and Ca. Cardinium [26]. DNA was extracted from whole spiders using DNeasy Blood and Tissue Kit (Qiagen) according to the manufacturer's protocol. 16S rRNA gene amplicon libraries of variable region V4 were prepared according to Life Technologies protocol (Ion Plus Fragment Library Kit), using primers Univ519F and Univ802R [43,44], and sequenced on an Ion Torrent PGM (Life Technologies) (average read length = 285 bp). Raw sequence reads were quality-screened using fastQC software v. 0.96 (Babraham Bioinformatics), trimmed to 260 bp and quality filtered and clustered with usearch v. 8.1.1861 and the UPARSE pipeline [45]. Operational taxonomic units (OTUs) were generated based on a high-quality subset of the data (maxee = 1.0), and the remaining reads were mapped onto the obtained OTUs with a 97% similarity cut-off. OTUs were classified to genus level using mothur v. 1.36.1 [46] with the Silva SSU Ref NR release 123 database as reference [47].

(ii). Large dataset: MiSeq

To increase the small sample size of microbiome data, a larger-scale sampling of the social spiders was also screened for sex ratio distorting endosymbionts. Adult female S. dumicola (n = 64) and S. mimosarum (n = 61) were collected from six populations across South Africa and Madagascar in April–June of 2012. Several nests were sampled from each population, and several individuals were sampled from each nest (see the electronic supplementary material for sample details). The spiders were placed in animal tissue lysis buffer (Qiagen) and frozen in the field. Whole spider DNA extraction was performed as described above. 16S rRNA gene amplicon libraries of variable regions V3 and V4 were prepared according to Illumina's 16S Metagenomic Sequencing Library Preparation guide, using Bac 341F and Bac 805R primers [48] and sequenced on a MiSeq desktop sequencer (Illumina). Sequence analysis, OTU clustering and taxonomic classification was done using mothur v. 1.39.0 [46], with Silva SSU NR release 128 as a reference [47]. The analysis protocol is available at https://github.com/ianpgm/AU_microbio_16S_protocol.

Further data analysis of both datasets was done in R v. 3.3.1 [49] using in-house scripts. The data were filtered to exclude non-bacteria and OTUs classified as common laboratory contaminants [50]. The OTU lists were screened for five endosymbiont sex ratio distorters: Wolbachia, Rickettsia, Ca. Cardinium, Spiroplasma and Arsenophonus. Results of the screenings were plotted using the R package ggplot2 [51].

3. Results and discussion

(a). Biased X-sperm proportion

We screened the sperm of males of two social species S. dumicola (n = 8) and S. mimosarum (n = 9, with female bias), and one subsocial species S. africanus (n = 7). In the dot plot of PI-stained sperm nuclei (corresponding to DNA content) and FSC of the nucleus (corresponding to particle size), originating from a single pedipalp, two populations of sperm nuclei are visible based on DNA content (figure 1). The lower population consists of 0-sperm nuclei (lower amount of DNA), while the upper population consists of X-sperm nuclei (higher amount of DNA). This corresponds to the left (0-sperm) and right (X-sperm) peaks of the PI-histograms in figure 1. Males of the two social species both produced significantly higher numbers of X-sperm nuclei compared to 0-sperm: S. dumicola: 0.70 ± 0.07; (t₇ = 7.85, p < 0.0001) and S. mimosarum: 0.68 ± 0.12; (t₈ = 4.47, p = 0.001). For males of the subsocial S. africanus a lower, yet significant, bias was found: 0.54 ± 0.03 (t₆ = 4.01, p = 0.004; figure 2). Stegodyphus africanus is an outcrossing species with pre-mating dispersal, and we, therefore, expected unbiased sperm cell production. This bias was significantly lower compared to the social species (generalized linear model: F₂ = 7.57, p = 0.003). Our results indicate substantial variation in sperm ratio between males, even originating from the same colony (figure 2).

Figure 2.

Proportion of X-sperm in the pedipalps of social S. dumicola and S. mimosarum males and subsocial S. africanus (average ± 95% CI). Striped lines depict the average proportion for each species and the solid line depicts an equal sperm ratio. Data points indicated with the same number originate from two pedipalps (left and right) of the same male. Bars indicate males originating from the same nest.

Our result of an average bias in X-sperm of S. dumicola males of 70% corresponds with the primary sex ratio bias found by karyotyping eggs from a single nest, where female embryos ranged from 68 to 92%, with a median of 83% [52]. While our data provide evidence for direct paternal influence over sex ratio, we cannot rule out the possibility of additional maternal influence on offspring sex ratio, as spider sperm is transferred in an inactive state, stored and subsequently activated in the female tract. This leaves ample opportunity for female control [53] and makes it unlikely that other factors such as difference in motility between X- and 0-sperm play an important role.

In mammals, several studies found bias in the proportion of X- and Y-sperm produced by a male ([54,55] and 19 studies reviewed in [16]), while to our knowledge, only three studies exist that link a bias in X/Y sperm proportion (either directly through fluorescence in situ hybridization of sperm cells [56] or indirectly using sperm nucleus area as a proxy [57]) or differences in male fertility [58] to offspring sex ratio. However, the potential adaptive function of sex ratio bias is not well understood. By contrast, female-biased sex ratio in social spiders applies across populations and species and is considered adaptive [17,18]. Social spiders have an obligatory inbreeding mating system and as a consequence, extraordinary high homozygosity and almost no genetic variation [20]. This social spider system can, therefore, be viewed as an extended version of the local mate competition model that predicts increased allocation to daughters, to reduce competition between highly related males within the nest over access to females over multiple generations (haystack model) [2,19,59]. Similarly, the local resource enhancement model [6] predicts allocation to daughters as the helping sex in social spiders, where females cooperate in tasks such as web building, prey capture and brood care [21,22]. Both of these explanations imply that females and males have aligned interests over the production of daughters. It is, therefore, unlikely that sexual conflict over offspring sex ratio exists in this system, as is frequently the case in other species [60].

Biased sperm production in arthropods is often associated with sex chromosome meiotic drive, in which males produce a biased proportion of sperm types because the driving sex chromosome is over-represented [61–63] (several Drosophila species, stalk eyed flies, mosquitos, etc.). Driving loci are considered as examples of selfish genetic elements rather than a mechanism of adaptive sex ratio adjustment. Our results are highly consistent with the actions of X chromosome meiotic drive. Although there is no sex chromosome towards which the driving X chromosome can be directed (X0 males in spiders), meiotic drive has been recorded in a nematode with a similar X0 sex chromosome system and was suggested as a way to control offspring sex ratio [64]. Furthermore, meiotic drive is hypothesized to play a role in the facultative adjustment of sex ratio in birds, where females are the heterogametic sex [65] and in the adult sex ratio in tetrapods [66]. The evolution of meiotic drive could lead to species extinction and should be strongly selected against [63]. However, it is possible that an incomplete suppressor of the meiotic drive action has evolved in social spiders. Further investigations into the cytological characteristics of spermatogenesis, identification of genomic regions involved in meiotic drive and the construction of a suitable genetic model are needed to establish the presence of a meiotic drive system. If confirmed, this would be a striking example of how a selfish genetic element can be recruited to function as a mechanism to adjust sex ratio adaptively [64].

(b). Microbiome

(i). Small dataset: Ion Torrent

The screening of six social spiders and one positive control revealed 509 bacterial OTUs. The number of reads classified as known sex ratio distorting endosymbionts are presented in table 1, along with total read numbers for each sample. No OTUs were classified as Spiroplasma or Arsenophonus, indicating that these bacteria were not present. Rickettsia, Wolbachia and Ca. Cardinium were each represented as a single OTU, and only showed high abundance (25–48% of total reads) in the positive control sample of Oedothorax retusus, a dwarf spider previously shown to be infected with these bacteria, causing a female bias [26]. In the social Stegodyphus species, these endosymbionts were present in a very low number of reads and never exceeded 0.02% of the total reads; this minimal detection might result from contamination by ingesting infected prey or by misassignments during de-multiplexing [67].

Table 1.

Counts of OTUs classified as known sex ratio distorters in positive control Oedothorax retusus and social species Stegodyphus dumicola and Stegodyphus mimosarum.

	control	social
	O. retusus	S. dumicola	S. dumicola	S. dumicola	S. mimosarum	S. mimosarum	S. mimosarum
Rickettsia	3641	0	3	0	6	0	1
Wolbachia	6985	3	4	0	3	0	1
Candidatus Cardinium	3721	0	1	0	2	0	0
Spiroplasma	0	0	0	0	0	0	0
Arsenophonus	0	0	0	0	0	0	0
total reads	14 495	17 788	15 154	51 631	27 799	11 983	7969

(ii). Large dataset: MiSeq

This larger screening confirmed the absence of OTU's classified as Ca. Cardinium or Arsenophonus in this dataset. More than half of all social spiders sampled were completely free of sex ratio distorting endosymbionts (S. mimosarum 33 out of 61, S. dumicola: 45 out of 64; figure 3). In infected individuals, endosymbionts are generally found at very low abundances, only comprising more than 1% of total reads in eight individuals (figure 3). Of those eight, three S. dumicola and two S. mimosarum stand out, having clear infections of either Rickettsia or Spiroplasma. Detailed data are available in the electronic supplementary material, table S2.

Figure 3.

Percentage of total 16S rRNA gene amplicon reads assigned to Rickettsia, Spiroplasma and Wolbachia endosymbionts in individuals of social S. dumicola and S. mimosarum. Localities are indicated with MAH (Madagascar, Mahavanana), SAK (Madagascar, Isalo National Park), TANA (Madagascar, Antananarivo), PON (South Africa, Pongola Game Reserve), WEE (South Africa, Weenen Game Farm), KRU (South Africa, Kruger National Park), ADDO (South Africa, Addo), PAA (South Africa, Paarl). First number indicates colony number while second number is the number of the individual spider, belonging to the respective colony.

Because of differences in library preparation methods and sequencing platforms, the data from the two sequencing experiments are not directly comparable. However, both datasets point to the same conclusion: sex ratio distorting endosymbionts are not responsible for sex ratio bias in social Stegodyphus.

4. Conclusion

Our results show the production of more female- than male-determining sperm cells as the likely mechanism underlying female-biased sex ratio in two social Stegodyphus spider species. This supports the hypothesis of sex ratio adjustment through a bias in gamete production in the heterogametic sex. We also show that five common genera of bacterial endosymbionts known to affect sex ratio are largely absent, ruling out that endosymbiont bacteria influence sex ratio bias in social spiders. In CSD species, offspring sex ratio adjustment is often expected to occur in the female reproductive tract at fertilization, suggesting female control [11]. Our study instead lends support for the heterogametic sex—here males—to evolve the ability to strategically bias sex ratio in a low cost manner by skewing gamete production [15,16].

Ethics

Research on the species in this study does not require ethical licences.

Data accessibility

Flow cytometry data have been uploaded onto FlowRepository (http://flowrepository.org/id/RvFr4wSgP6tERF9mxpOW30F941OaQDY73ZVArWWoQzG3eCocpmyKWPmytoSiUZT8). Excel files containing the FSC-A and PI-A data for the gated sperm nuclei populations per individual spider can be found in the Dryad repository: http://dx.doi.org/10.5061/dryad.r3206 [68]. Source data for figure 2 can be found in the electronic supplementary material, table S1. Microbiome sequence data are available at NCBI, Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) accession numbers SRP130747, SRP130740 and SRP130742.

Authors' contributions

All authors contributed to the design of the study and to writing of the manuscript. B.V., J.B., F.H. and T.B. performed the flow cytometry experiment and analysed the data, microbiome screening and data analysis was performed by M.M.B. and A.S.

Competing interests

We declare we have no competing interests.

Funding

This research has been funded by the European Research Council (ERC StG-2011-282163 to T.B.).