Abstract
Evolution under asexuality is predicted to impact genomes in numerous ways, but empirical evidence remains unclear. Case studies of individual asexual animals have reported peculiar genomic features which have been linked to asexuality, including high heterozygosity, a high abundance of horizontally acquired genes, a low transposable element load, and the presence of palindromes. However, it is unclear whether these features are lineage-specific or general consequences of asexuality. We reanalyzed published genomes of 24 asexual animals and found that not a single genome feature is systematically replicated across a majority of these species, suggesting that there is no genomic feature characteristic of asexuality. We found that only asexuals of hybrid origin were characterized by high heterozygosity levels. Asexuals that were not of hybrid origin appeared to be largely homozygous, independently of the cellular mechanism underlying asexuality. Overall, despite the importance of recombination rate variation for understanding the evolution of sexual animal genomes, the genome-wide absence of recombination does not appear to have the dramatic effects which are expected from classical theoretical models. The reasons for this are probably a combination of lineage-specific patterns, impact of the origin of asexuality, and a survivor bias of asexual lineages.
- Abbreviations
- TE
- transposable element;
- HGT
- horizontal gene transfer
Introduction
Sex: What is it good for? The reasons for why most eukaryotes take a complicated detour to reproduction, when straightforward options are available, remains a central and largely unanswered question in evolutionary biology 1,2. The species that use asexual reproduction as their sole form of replication typically occur at the tips of phylogenies and only few of them have succeeded like their sexually reproducing counterparts 3. In other words, most asexual lineages may eventually be destined for extinction. These incipient evolutionary failures are however invaluable because by understanding the evolutionary fate of asexual species, something may be learned about the adaptive value of sex.
An accumulating number of studies have sequenced the genomes of asexually reproducing animals, often with the aim of identifying features that would distinguish them from sexual species (Figure 1). In asexual animals, females produce daughters from unfertilized eggs via so-called thelytokous parthenogenesis (hereafter asexuality) 4. Asexuality is predicted to have many consequences for genome evolution, since gamete production via meiosis and the restoration of somatic ploidy levels via fertilization no longer take place. Predicted consequences include for example the accumulation of deleterious mutations 5–7, as well as changes in heterozygosity levels 8,9 and transposable element (TE) dynamics 10. Some predictions have been tested without genomic data, using a handful of housekeeping genes. However, conclusions based on such small and non-random subsets of genomes can lead to erroneous conclusions 11. With the advent of high-throughput sequencing it is possible to evaluate classical predictions of asexuality at the genome scale, and furthermore to test new predictions, such as the accumulation of palindromes (see below), which could not be studied with single gene approaches. In the present study, we compare the published genomes of 24 asexual animal species (Figure 1) to assess whether we can identify any key features characteristic of asexual animals. The 24 species comprise four species of bdelloid rotifers, a group that likely persisted and diversified in the absence of canonical sex for over 40 million years 12. Bdelloids have thus far overcome the predicted dead-end fate of asexuality, which raises the question of what mechanisms protect them from extinction, and whether these mechanisms are visible in specific characteristics of their genomes.
Because the predicted consequences of asexuality are strongly affected by how asexuality evolved from the sexual ancestor (Box 1) as well as by the cellular mechanisms underlying asexuality (Box 2), we include biological differences among asexual species in our comparisons. For example, some asexual species have evolved via hybridization (Box 1), which generates high heterozygosity and can result in increased activity of transposable elements 13–15. In such instances, it can be difficult to disentangle consequences of hybridization from consequences of asexuality. Similarly, some cellular mechanisms underlying asexuality involve meiotic divisions, with a secondary restoration of somatic ploidy levels, while others do not. In the former case, heterozygosity in the asexual species is expected to decay rapidly, while in the latter case, it could be maintained or even increase over time (Box 2). Finally, because genome studies differed in their focus and in the methods used, we reanalyzed published genomes with standardized approaches. Whenever possible, we conducted quantitative comparisons between groups of asexual species. However, for interpretation, it is important to consider that the available genomes are not a random nor representative sample of asexual animals, and that not all of these genomes reflect evolutionarily independent events.
Box 1: Transitions to asexuality
Meiotic sex evolved once ~1.5 billion years ago, and since then remained the predominant mode of reproduction in eukaryotes 3,39. Current asexual animals therefore derive from a sexual ancestor, but how transitions from sexual to asexual reproduction occur can vary. While the molecular changes underlying different types of transitions are unknown, the expected genomic consequences of asexuality vary extensively among them 14.
Hybrid origin
Hybridization between sexual species can generate hybrid females that reproduce asexually 14,40. Asexuality caused by hybridization generates high levels of heterozygosity, corresponding to the divergence between the parental sexual species prior to hybridization. Hybridization can also result in a burst of transposable element activity 13.
Endosymbiont infection
Infection with intracellular endosymbionts (such as Wolbachia, Cardinium or Rickettsia) can cause asexuality, a pattern that frequent in species with haplodiploid sex determination 41. This type of transition often (but not always) results in fully homozygous lineages because asexuality induction frequently occurs via gamete duplication (see Box 2).
Spontaneous mutations/Contagious asexuality
Spontaneous mutations can also underlie transitions from sexual to asexual reproduction. In addition, asexual females of some species produce males that mate with females of sexual lineages, and thereby generate new asexual strains (contagious asexuality). In both cases, the genomes of incipient asexual lineages are expected to be very similar to those of their sexual relatives and subsequent changes should be largely driven by the cellular mechanism underlying asexuality (Box 2).
Box 2: Cellular mechanisms of asexuality
In sexual species offspring is generated through the fusion of male and female gametes. In asexuals, females generate diploid (or polyploid) offspring from unfertilized oocytes via different cellular mechanisms. The cellular mechanism used is predicted to affect genome evolution and especially heterozygosity levels. For details see 4,42.
Mitotic asexuality (Apomixis)
Under mitotic asexuality, no ploidy reduction occurs and offspring are clones of their mother.
Meiotic asexuality (Automixis)
Under meiotic asexuality, meiotic divisions occur partially or completely, but somatic ploidy levels are maintained via different mechanisms. Some of these mechanisms have similar genomic consequences as mitotic asexuality, even though meiosis is involved (for example, endoduplication in hybrid asexuals results in offspring that are clones of their mother. Such mechanisms are often referred to as “functionally mitotic” (or functionally apomictic), especially when the cellular mechanisms are not known in detail but genotyping data suggest that offspring are clones of their mother.
Endoduplication
A duplication of the entire chromosome set occurs before normal meiosis, during which ploidy is reduced again. If recombination occurs between identical chromosome copies rather than between chromosome homologs, endoduplication produces offspring that are clones of their mother.
Inverted meiosis with terminal fusion (gonoid thelytoky)
During the first meiotic division, sister chromatids separate instead of homologous chromosomes. The homologues are separated in the second meiotic division. In the absence of recombination, inverted meiosis with terminal fusion generates offspring that are clones of their mother (and though mechanistically different, is conceptually equivalent to central fusion without recombination). Holocentric chromosomes seem to be a prerequisite for this type of mechanism 43,44.
Central fusion and terminal fusion
Under these two mechanisms, somatic ploidy levels are restored through the fusion of two of the four meiotic products (products separated during the first meiotic division merge under central fusion, products separated during the second division merge under terminal fusion). In the absence of recombination, central fusion generates offspring that are clones of their mother.
Gamete duplication
After a full meiosis, a haploid meiotic product undergoes duplication. This results in a diploid, but fully homozygous offspring.
We uncovered a number of unusual features in the genomes of asexual animals, including extreme loads of transposable elements and highly asymmetric divergence among haplotypes in polyploid species of hybrid origin. However, none of these were systematically replicated across even a majority of analyzed species, let alone all of them, suggesting that there is no universal genomic feature specific to asexual species. We found that a hybrid origin of asexuality was the most important factor affecting heterozygosity, with potential effects of asexuality being masked by effects of hybrid ancestry. Unexpectedly, asexuals that are not of hybrid origin are largely homozygous, independently of the cellular mechanism underlying asexuality.
Results
Overview of species and genomes studied
We reanalyzed the published genomes of 24 asexual animal species with the aim of identifying general genomic signatures of asexualilty. The 24 species correspond to at least 16 independent transitions to asexuality and cover a broad taxonomic range, including chordates, rotifers, arthropods, nematodes and tardigrades. In addition to covering this taxonomic range, the asexual species vary in the cellular mechanisms underlying asexuality, in the mechanisms that caused the transition to asexuality, as well as in other biological aspects (Figure 1, Supplementary Tables 1 & 2). This variation allows us to assess whether asexuality generates universal genomic signatures independently of species-specific traits.
The cellular mechanisms underlying asexuality were studied in 20 of the 24 species. Eight of them use mitotic asexuality, while the 12 remaining species use different types of meiotic asexuality (Figure 1). All but one of the eight species with mitotic asexuality are polyploid, the amazon molly P. formosa being the only diploid studied. Conversely, all but one species with meiotic asexuality are diploid. This is expected given that polyploidy can generate problems during meiosis (reviewed in 45). Nevertheless, the nematode Panagrolaimus sp. is characterized by both meiotic asexuality and triploidy 30 (see Supplementary Table 1 for details).
Information on how asexuality evolved is available for 15 of the 24 sequenced species (Figure 1). A hybrid origin has been suggested for ten of these. Endosymbionts are the most likely cause of asexuality in four species (the springtail, both wasps and the thrips), and spontaneous mutation in one (the ant). Across the 24 species, hybrid origin is correlated with polyploidy. Six of the 11 polyploids in our sample are of hybrid origin, while for the five others a hybrid origin has thus far not been suggested, but is supported by our results (see below). It is important to note however that there are many polyploid asexual animals that are not of hybrid origin, including several well studied asexual species such as the New Zealand mudsnail Potamopyrgus antipodarum, the bush cricket Saga pedo, or the bagworm moth Dahlica triquetrella. None of these has a published genome yet which precludes their inclusion in our study.
Most if not all predicted consequences of asexuality are expected to accumulate over time, meaning that their effect size as well as the power to detect them increases in old asexual lineages. However, estimating the age of asexual lineages is difficult and always associated with large uncertainties 46,47. We therefore did not include quantitative comparisons among asexuals with respect to their age. However, because our set of species comprises asexuals believed to be ‘ancient’ (i.e., several million years old, see Supplementary Table 1), we discuss, where appropriate, potential age effects in a qualitative manner.
Mutation accumulation and positive selection
One of the classical predictions linked to asexuality is that it reduces the efficacy of selection 5–7,48–50. This reduction occurs because linkage among loci in asexual species prevents selection from acting individually on each locus. This can allow for deleterious mutations to accumulate over time, because they are linked to other sites under selection. It can also reduce the rate of adaptation, because beneficial mutations cannot reach fixation in a population as easily as under sexual reproduction.
The prediction that deleterious mutations accumulate more rapidly in asexual than sexual lineages has been tested in over twenty groups of different asexual species (reviewed in 51 and with three additional studies published since 16,21,52), with results generally supporting the prediction. However, in only eight studies were the tests conducted genome wide, while tests in the remaining studies were based on one or a few genes only. Note that four 11,52–54 of these studies were based on transcriptomes and are therefore not included in our systematic reanalysis. Among the genome wide tests, results are much more mixed than among the ‘single or few genes’ studies, raising the question whether the latter are representative of the genome as a whole. Specifically, only two of the eight genome-wide studies found support for deleterious mutation accumulation in asexuals 52,53. However, two studies found that sexual taxa experienced more deleterious mutation accumulation than asexual taxa 11,19 while the four remaining ones found no differences between sexual and asexual taxa 16,21,24,54. In the case of the water flea D. pulex, the study specifically reported that earlier inferences of deleterious mutation accumulation under asexuality were incorrect, as the detected deleterious mutations in asexual strains were inherited from the sexual ancestor and did not accumulate after the transition to asexuality 24.
In summary, results from genome-wide studies addressing the prediction of deleterious mutation accumulation in asexual species are equivocal. More studies are therefore needed. A major constraint for studying deleterious mutation accumulation, and the reason why it was not studied in most genome studies of asexuals species (Figure 1), is that it requires sexual references for comparison. Such references are either unknown or not included in most published genome studies of asexuals.
The same constraints likely explain why no study has thus far addressed adaptive evolution in the genome of an asexual species. The question of adaptive evolution was addressed indirectly in the amazon molly, by studying the amount of segregating variation at immune genes (where variation is known to be beneficial). The authors found very high diversities at immune genes 16. However, these were difficult to interpret because standing variation was not compared to sexual relatives, and because the amazon molly is a hybrid species. Hence the high diversity could be a consequence of the hybrid origin rather than of asexuality.
Heterozygosity
Expected heterozygosity levels in asexual organisms are influenced by three major factors: (1) the mechanism of transition to asexuality (which determines the initial level of heterozygosity) (Box 1), (2) the cellular mechanism underlying asexuality (which determines whether heterozygosity should increase or decrease over time) (Box 2), and (3) for how long a species has been reproducing asexually (because effects of asexuality accumulate over time).
As expected, all of the asexual species with a known hybrid origin display high heterozygosity levels (1.73% - 8.5%, Figure 2), while the species with an intraspecific origin of asexuality show low heterozygosity levels (0.03% - 0.53%, Figure 2). However, it is important to note that hybrid origin is correlated with polyploidy in our dataset, and that heterozygosity does not have a clear definition in polyploids (Box 3). Our measures of heterozygosity are based on the proportion of sites with more than one allele present among all copies, where the total number of copies includes all homologous genome regions (Box 3).
Box 3: Quantification of heterozygosity for different ploidy levels
The classical definition of heterozygosity describes heterozygosity as a measure of allelic divergence, where alleles are defined through chromosome pairing 56. This definition poses a problem for genomes where chromosome pairing is not known (e.g., in polyploid genomes), as well as for genomes of mitotic asexuals where chromosome pairing may not occur at all. In these cases, concepts such as divergence between alleles (i.e., heterozygosity) vs. divergence of paralogs, that are clearly distinct in diploids, become blurred. We therefore quantify heterozygosity as the proportion of nucleotides that differ in at least one of the homologous chromosomes. Three examples of genomes with similar heterozygosity but different ploidy levels are shown on the scheme for illustration. Grey bars highlight specific loci, coloured bars represent alternative alleles at a given locus.
The heterozygosity levels present at the inception of asexuality should decay over time for most forms of meiotic asexuality 42,55. Under mitotic asexuality, heterozygosity is expected to increase over time as haplotypes can accumulate mutations independently of each other (generating the so-called ‘Meselson effect’)8. However, gene conversion can strongly reduce haplotype divergence and, if high enough, can even result in a net loss of heterozygosity over time, even under mitotic asexuality 8,17. In spite of the prediction that the cellular mechanism of asexuality should affect heterozygosity, the cellular mechanism of asexuality appears to have little or no effect on heterozygosity levels once we control for the effect of hybrid origins (Figure 2). However, we have very little power to detect such effects, especially because our dataset does not include any asexual species that uses mitotic asexuality but is not of hybrid origin. Nevertheless, it is interesting to note that species with different forms of meiotic asexuality (including gamete duplication and central fusion) feature similarly low heterozygosity levels. This suggests that although the rate of heterozygosity loss is expected to vary according to mechanisms of asexuality, this variation is only relevant very recently after transitions to asexuality, and no longer affects heterozygosity among established asexual species. Alternatively, variation in heterozygosity caused by different forms of meiotic asexuality may be too small to be picked up with our methods.
Heterozygosity structure in polyploids
Heterozygosity in polyploids is estimated via the proportion of sites that differ in at least one of the homologous regions (see Box 3). This means that the estimated genome-wide heterozygosity could be generated by a single haplotype that is highly divergent while others are similar, or by homogeneous divergence across all copies present, or a combination of these. We therefore decomposed genome-wide heterozygosity for each polyploid genome into portions with different divergence structures (Figure 3).
In polyploid species of hybrid origin (sexual or asexual), heterozygosity is generally driven by divergence between haplotypes originating from different species (hereafter homoeologs, following the terminology of Glover et al 57). In our dataset, the polyploid species with confirmed hybrid origins are nematodes in the genera Meloidogyne (five species) and Panagrolaimus (one). As expected, heterozygosity in these species is largely dominated by divergence between homoeologs. In the triploid species, divergence is between a single homoeolog and two similar homologs (yellow portions in Figure 3), consistent with previous findings 30,34. In the tetraploid species, genome-wide heterozygosity is generated by a combination of genome portions comprised of one homoeolog and three similar homologs and of other portions comprised of pairs of homoeologs (pink portions in Figure 3).
Given that in asexual polyploids of hybrid origin we expect and observe highly heterogeneous divergences among haplotypes, while polyploidy of intra-specific origin is predicted to generate homogeneous divergences, haplotype divergences can be used to infer the origin of asexuality in polyploid species. Notably, the highly asymmetric divergence levels between haplotypes in the four bdelloid rotifers (Figure 3) are best explained by a hybrid origin of bdelloids. When tetraploidy was first discovered in bdelloids, it was proposed that tetraploidy stemmed from either a whole genome duplication or a hybridization event in their ancestor 58. However, studies of bdelloid rotifers traditionally refer to the divergent haplotypes as “ohnologs” (e.g., 17,18), which, following the unified vocabulary of Glover et al 57 would imply that the diverged haplotypes are products of a whole genome duplication. However, given their likely hybrid origin, referring to them as homoeologs appears more appropriate.
Whether the crayfish is also of hybrid origin remains as an open question. Its genome features two nearly identical haplotypes and one that is substantially divergent (1.8%, Figure 3), which is suggestive of a hybrid origin. However, the divergence of the latter is within the range of heterozygosity commonly observed in sexual species, and therefore we cannot clearly distinguish between an intra-specific or a hybrid origin.
Our analyses also reveal that the divergence of homologs varies extensively among bdelloid rotifer genera. Divergence is very low in Rotaria (0% in R. magnacalcarata and 0.25% R. macrura) and low in A. ricciae (0.5%) but relatively high in A. vaga (2.4%). The mechanisms causing these differences remain unknown. In A. vaga it has been suggested that gene conversion reduces divergence between homologs in some genome regions 17. It is possible that rates of gene conversion are higher in Rotaria than Adineta, for unknown reasons. Independently of the mechanisms causing the differences between bdelloid genera and species, it is important to note that with such low levels of divergence between homologs, there can be no genome-wide ‘Meselson effect’ in bdelloid rotifers (see also 17). It remains possible that the subset of genomic regions with divergence between homologs in Adineta feature allele phylogenies as expected under the ‘Meselson effect’. This is the case in the asexual unicellular eukaryote Trypanosoma brucei gambiense: some genome regions feature high heterozygosity and allele phylogenies as expected under the ‘Meselson effect’, while others are largely homozygous 59. Again, it remains unknown why there is such extensive heterogeneity in divergence across the genome in this species. A possible explanation is that the heterozygous genome regions are the consequence of ancient introgression, and that gene conversion rates are low in such regions with very high heterozygosity (see Discussion).
Palindromes and gene conversion
Palindromes are duplicated regions on a single chromosome in reverse orientation. Because of their orientation, palindromes can align and form hairpins, which allows for gene conversion within duplicated regions (Supplementary Figure 2). Palindrome-mediated gene conversion was shown to play a major role in limiting the accumulation of deleterious mutations for non-recombining human and chimpanzee Y chromosomes 60–62. Indeed, approximately one third of coding genes on these Y chromosomes occur in palindromes, and the highly concerted evolution of palindromic regions indicates that the rates of gene conversion are at least two orders of magnitude higher in the palindromes than between homologous chromosomes. The reports of palindromes in the genomes of the bdelloid rotifer Adineta vaga and the springtail Folsomia candida led to the hypothesis that palindromes could play a similar role in asexual organisms – reducing deleterious mutation accumulation in the absence of recombination. However, the potential benefit of palindrome-mediated gene conversion depends on the portion of genes in palindromic regions 61. In addition to identifying palindromes, it is therefore important to also quantify the number of genes affected by palindrome-mediated gene conversion.
Methods for palindrome identification depend on genome assemblies (contrary to the other genome features we re-analysed in our study). Palindromes are less likely to be detected in highly fragmented assemblies and artificial palindromes can be generated by erroneous scaffolding (see also 18). Our analyses and interpretations assume that there are no systematic scaffolding errors in the published assemblies. Palindrome identification methods rely on genome annotations, which are available for 22 of the 24 asexual species (all except D. pulex and A. rufus). We screened these 22 genomes for the presence of palindromic arrangements (See Methods and Supplementary Text S2 for details). We identified 19 palindromes in A. vaga, 16 in F. candida, and up to four palindromes in seven additional genomes (Table 1). Not a single palindrome was detected in the remaining 13 species. The frequency of palindromes had no phylogenetic signal; for example, although we found 19 palindromes in A. vaga, we found no palindromes in the three other bdelloid rotifers (in agreement with 18). There is also no indication for major rearrangements being present solely in very old asexuals; among the very old asexuals, the non-A. vaga rotifers along with the Diploscapter nematodes have either no or only a single palindrome.
Adineta vaga and F. candida are the only two species with more than 100 genes potentially affected by palindrome-mediated gene conversion, but even for these two species, the overall fraction of genes in palindromes is very small (1.23% and 0.53% respectively). The fraction of genes in the other seven species ranges between 0.01% and 0.16%, suggesting that palindromes do not play a major role in the genome evolution of any of the asexual lineages analyzed. Our findings substantiate the conclusion of a previous study 18 that major genomic rearrangements and the breaking of gene syntenies do not occur at high rates in asexual organisms. They appear to occur at rates similar to those known in recombining genome portions of sexual species 63,64.
Mitotic gene conversion can also occur outside of palindromic regions, for example when double-stranded DNA breaks are repaired using the homologous chromosome as a template 65,66. It can, in theory, contribute to the loss of heterozygosity under all forms of asexuality, but mitotic gene conversion rates have only rarely been studied in asexual species – or sexual ones for that matter. Gene conversion rates are estimated differently in different studies and are therefore difficult to compare: in the water flea D. pulex, they were estimated to amount to approximately 10−6 locus−1 generation−1 24,25,67, in the amazon molly P. formosa to 10−8 16. Up to 11 % of the genome of the nematode D. pachys is suggested to be homozygous as a consequence of gene conversion, and studies also argued for an important role of gene conversion for genome evolution in root knot nematodes 34 and rotifers 17,18, although no quantitative estimates are available for these species groups.
Transposable elements
Transposable elements (TEs) are DNA sequences that can autonomously change positions in a genome via various ‘cut-and-paste’ and ‘copy-and-paste’ mechanisms 68,69. TEs can invade genomes even though they generally provide no adaptive advantage to the individual carrying them 70–72. To the contrary, new TE insertions in coding or regulatory sequences disrupt gene functions and cause deleterious effects in the host; only very rarely can specific insertions be co-opted to acquire novel, adaptive, functions for the host 72. In sexual organisms, TEs can spread through panmictic populations because of their ability to rapidly colonize new genomes 10,73. At the same time, sexual reproduction facilitates the purging of deleterious TE insertions, because recombination, segregation and genetic exchange among individuals improve the efficacy of selection 74,75. In the absence of sex, TEs could therefore accumulate indefinitely, which led to the prediction that TEs could frequently drive the extinction of asexual lineages. Only asexual lineages without active TEs, or with efficient TE suppression mechanisms, would be able to persist over evolutionary times 75. Consistent with this view, a study in bdelloid rotifers reported extremely low TE loads 76. This prompted the authors to suggest that bdelloid rotifers could have been able to persist in the absence of sex for over 40 million years thanks to their largely TE-free genomes.
Our analysis of asexual animal genomes does not support the view that bdelloid rotifers have unusually low TE contents. The TE content of bdelloid rotifers (0.8% to 9.1%) is comparable to other asexual animal taxa (Figure 4), all of which are considerably younger than the bdelloids. Across the 24 genomes, there was large variation in total TE content, overall ranging from 6.6% to 17.9%, but with one species, the marbled crayfish, reaching 34.7%. Nevertheless, the abundance of TEs in asexual animal genomes appears to be generally lower than in sexual species, which range typically from 8.5-37.6% (median: 24.3%) 77. Whether this difference is indeed driven by asexuality remains an open question as TE loads are known to be highly lineage-specific 20,78. Furthermore, we annotated TEs in each genome via homology searches in general databases (see methods). This can result in an underestimation of TE loads relative to annotations based on species-specific TE libraries. However, this is unlikely to be of major concern in our study since the methods we used allowed us to identify more TEs than most of the individual genome studies (Supplementary Table 3). Specifically, most studies estimate TE loads from genome assemblies, which underestimates TE loads because regions with high repetitive contents are generally not assembled.
In addition to other lineage-specific characteristics, the cellular mechanisms underlying asexuality could also affect TE loads. For example, most forms of meiotic asexuality can allow for the purging of heterozygous TE-insertions, given the loss of heterozygosity between generations (Box 2). Barring potential gene conversion events, this form of purging cannot occur under mitotic asexuality. However, in the genomes analyzed here, we did not find any effect of cellular mechanisms on TE loads (Supplementary Figure 3), likely because the expected effect of the cellular mechanisms is very small relative to lineage-specific mechanisms. Moreover, host TE suppression mechanisms can contribute to the inactivation and subsequent degeneration of TE copies over time, independently of the cellular mechanism of asexuality 72,79.
Two asexual animals clearly stand out (Figure 4), one for very low TE contents (the rotifer A. ricciae; <1% of the genome) and one for very high contents (the marbled crayfish P. virginalis >34%). There is currently no known mechanism that could help explain why A. ricciae differs so extensively from other bdelloid rotifers. In the case of the marbled crayfish, it is unknown whether its extreme repetitive content is a heritage from its sexual ancestor or a consequence of a possible hybrid origin with a subsequent burst of TE activity. In the absence of information on TE loads in the sexual relative P. fallax, these possibilities cannot be evaluated. More generally, in most studies quantifying TE contents in asexual species, no comparisons to related sexual species are made. In the cases where this was done, no differences could be detected 16,20,21,26,35,80.
Independently of the question of whether asexuality affects genome-level TE loads, our dataset allows us to study whether hybrid species have higher TE loads than non-hybrid species. Indeed, TE activity in hybrids is expected to be high because of mismatches between species-specific TEs and silencing machineries 13,15,81,82. However, we do not find any difference in TE content according to hybrid vs intraspecific origin of asexuals (Supplementary Figure 2).
Gene loss
Asexual animals are predicted to lose genes underlying sexual reproduction traits, including male-specific traits and functions (e.g. male-specific organs, spermatogenesis), as well as female traits involved in sexual reproduction (e.g., pheromone production, sperm storage organs) 83. In the absence of pleiotropic effects, gene loss is expected due to mutation accumulation in the absence of purifying selection maintaining sexual traits, as well as to directional selection to reduce costly sexual traits 84. Some gene loss consistent with these predictions is documented. For example, the sex determination genes xol-1 and tra-2 are missing in the nematode D. coronatus 28. Furthermore, genes believed to be involved in male functions harbour an excess of deleterious mutations in the wasp Leptopilina clavipes 19, which could represent the first step towards the loss of these genes. However, a similar excess of deleterious mutations in genes with (presumed) male-specific functions was not detected in the amazon molly P. formosa 16.
Species reproducing via mitotic asexuality are further predicted to lose genes specific to meiotic processes 85. The genes involved in meiosis have been studied in three of eight mitotic parthenogens, as well as in Rotaria rotifers and Diploscapter nematodes, whose cellular mechanisms of asexuality are unknown. Most meiotic genes have been found in the four bdelloid rotifers 17,18 and in both species of Diploscapter nematodes 28,29 There was also no apparent loss of meiosis genes in the amazon molly P. formosa 16 As much as the idea is appealing, there does not seem to be any support for the predicted loss of meiotic genes in mitotic asexuals. We note that the lack of our understanding of meiosis on the molecular level outside of few model organisms (particularly yeast and C. elegans) makes the interpretation of gene loss (or absence thereof) difficult. This is best illustrated by the fact that losses of meiosis genes have also been reported in different sexual species, where meiosis is clearly functional 86.
In summary, some gene loss consistent with the loss of different sexual functions has been reported in several asexual species. However, a clear interpretation of gene loss in asexual species is problematic because the function of the vast majority of genes is unknown in these non-model organisms.
Horizontal gene transfer
Asexual species could harbour many genes acquired via horizontal gene transfer (HGT) as a consequence of relaxed selection on pairing of homologous chromosomes. It has also been proposed that HGTs represented an adaptive benefit which allows for the long term maintenance of asexuality 87. Indeed, bdelloid rotifers have been reported to carry an unusually large amount (6.2% - 9.1%) of horizontally acquired genes compared to sexual lophotrochozoan genomes (0.08% - 0.7%) 18,88. Many of these have contributed to adaptive divergence between bdelloid rotifer species 89. However, there are no other ancient asexuals sequenced and evaluating the role of HGTs in the long-term persistence of asexuality is therefore not possible. In more recent asexuals, levels of HGT appear mostly low, e.g. in Panagolaimus (0.63% - 0.66%) and in two tardigrade species (0.8% - 0.97%)18,30,37. The only genome with a high reported fraction of HGT (2.8%) outside of the rotifers is the springtail F. candida 23. This is a meiotic asexual, hence a relaxed constraint on chromosome pairing did not contribute to the high retention of horizontally acquired genes. Nevertheless, the presence of a gene for lignocellulose degradation in the springtail and in the root-knot nematode M. incognita, which was likely acquired via HGT in both species, supports an adaptive role of HGT in these asexuals 23,33 However, such isolated events of adaptive HGTs are not specifically linked to asexuality, since they are reported in sexual species as well 90. The potential relation of HGT and asexuality will remain unclear until we are able to reliably identify HGTs in more genomes of asexual as well as sexual species. Indeed, current reports of HGT are often unreliable because of the difficulty of distinguishing HGT from contamination 37.
Gene family expansions
Most genome papers, including those focussing on asexual animals, scan for expansions of specific gene families. Such expansions are then discussed in the light of the focal species’ biology. The expansion of specific gene families per se is thus generally a species-specific trait 91 that is not related to asexuality. For example, expansions of stress response genes in M. incognita 33, Panagrolaimus spp. 30, and R. varieornatus 38 were suggested to be associated with the evolution of cryptobiosis in these species. To our knowledge, the only example of a gene family expansion that could be directly associated with asexuality is the diversification of the RNA silencing machinery of TEs in bdelloid rotifers 17. TEs are expected to evolve reduced activity rates in asexual hosts (see section Transposable elements), and an improved RNA silencing machinery could be the mechanism underlying such reduced activity rates.
However, mitotic asexuality might allow for extensive variation in gene copy numbers between homologous chromosomes as a consequence of relaxed constraints on chromosome pairing (see also section on Horizontal gene transfer). Gene family expansions (and contractions) could therefore be more extensive and be retained more frequently in asexual than sexual species. To test this hypothesis, an overall comparison of gene family expansions in sexual and asexual species is needed (see Supplementary Text S3). Four studies have surveyed gene family expansions in asexual species as well as in (sometimes distantly related) sexual counterparts, but these studies found no differences between reproductive modes l6,21,23,31. However, only two of the four studies are based on asexuals with mitotic asexuality (i.e., where chromosome pairing is not required), and additional studies are therefore needed to address the question of whether asexuality affects gene family expansions.
Discussion
We re-analyzed 24 published genomes of asexual animals to investigate whether we can detect genomic features that are characteristic of asexual animals in general. Many of the original genome studies highlighted one or a few specific features in their focal asexual species, and suggested that it might be linked to asexuality. However, our analyses and review of published studies show that none of these genome features appear to be a general consequence of asexuality given that none them was systematically replicated across even a majority of analyzed species.
The variation among genomes of asexual species is at least in part due to species- or lineage-specific traits. But variation among the features detected in the published single-genome studies is also generated by differences in the methods used. Such differences are often less obvious, and maybe less interesting to discuss, yet they can be critical in our assessment of genome diversity among animals. In this work we thus re-analyzed several key genome features with consistent methods. To minimize the effect of differences in genome quality, we have used in priority robust methods, e.g. based on sequencing reads rather than from assemblies. For example, re-estimating heterozygosity levels directly from reads of each species allowed to show a strong effect of hybrid origin, but not of cellular mechanism of asexuality (Figure 2). Another advantage of using the same methods over all species is that it diminishes the “researcher degrees of freedom” 92–94. For example, the analysis of polyploid genomes requires choosing methods to call heterozygosity and ploidy. By providing a common framework among species, we have shown that homoeolog divergence is very diverse among polyploid asexuals.
We have identified hybrid origin as the major factor affecting heterozygosity levels across all asexual animal species with available genomic data. This is consistent with the conclusions of two studies that focussed on individual asexual lineages: hybridization between diverse strains explains heterozygosity in Meloidogyne root knot nematodes and Lineus ribbon worms 34,54. This rule applies more generally to all the species analysed with known transitions to asexuality, but it is important to highlight that all the non-hybrid species in our dataset are hexapods. Thus in principle the low heterozygosity could be a hexapod specific pattern, for example due to high gene conversion rates in hexapods. The taxonomic range of the sequenced species is wide but we are missing several clades rich in asexual species, such as mites or annelids 95,96. These clades would be useful foci for future genomic studies of asexual species. Independently of the findings of such future studies, our results suggest that mitotic gene conversion (that acts independently of palindromes) plays a significant and highly underappreciated role in the evolution of asexual species of intraspecific origin. For example, it has been argued that one of the main benefits of sex could be the masking of recessive deleterious mutations (referred to as “complementation”) which would be exposed under many forms of meiotic asexuality 97,98. If gene conversion is indeed pervasive, these arguments would extend to functionally mitotic forms of asexuality. Conversely, high rates of gene conversion could also allow for the purging of deleterious mutations while in the heterozygous state, as in highly selfing species (eg. 99,100). Such purging could help explain why most of the genome scale studies did not find support for the theoretical expectation that asexual reproduction should result in increased rates of deleterious mutation accumulation (see section Mutation accumulation and positive selection). More generally, given the major differences in genome evolution for asexuals of intra-specific vs. hybrid origin, our study calls for future theoretical approaches on the maintenance of sex that explicitly consider the loss vs. the maintenance of heterozygosity in asexuals.
In our evaluation of the general consequences of asexuality, we were not able to take two key aspects into account: survivor bias of asexual lineages, and characteristics of sexual ancestors. How often new asexual lineages emerge from sexual ancestors is completely unknown, but it has been speculated that in some taxa asexual lineages might emerge frequently, and then go extinct rapidly because of negative consequences of asexuality. In other words, asexuals that would exhibit the strongest consequences of asexuality, as predicted by theoretical models, are expected to go extinct the fastest. Such transient asexuals remain undetected in natural populations, because research focuses on asexual species or populations, and not on rare asexual females in sexual populations. Indeed, most of the species included in our study have persisted as asexuals for hundreds of thousands to millions of years. They might thus be mostly representative of the subset of lineages that suffer weaker consequences of asexuality. Finally, the key constraint for identifying consequences of asexuality is that almost none of the published genome studies of asexual animals included comparisons to close sexual relatives. This prevents the detection of specific effects of asexuality, controlling for the variation among sexual species - which is extensive for all of the genome features we analyzed and discussed in our study. Overall, despite the importance of recombination rate variation for understanding the evolution of sexual animal genomes (e.g., 101,102), the genome-wide absence of recombination does not appear to have the dramatic effects which are expected from classical theoretical models. The reasons for this are probably a combination of lineage-specific patterns, differences according to the origin of asexuality, and survivor bias of asexual lineages.
Methods
We combined different methods into a complete pipeline that collects published assemblies, sequencing reads, and genome annotation data from online databases, and automatically computes the genome features discussed here. The methods for the different steps in the pipeline are detailed below. The pipeline is available at https://github.com/KamilSJaron/genomic-features-of-asexual-animals. We used this pipeline to gather and analyze the data for 29 sequenced individuals from 24 asexual species. For some species, additional genomes to the ones we used were available, but we did not include them because of low data quality and/or unavailable illumina reads (this was the case for one sample of M. incognita, M. floridensis and multiple samples of D. pulex 24,33,36). Overall, the genome features computed were: ploidy, genome size, heterozygosity, heterozygosity, haplotype divergence structure, transposable elements/ repeat content, conserved gene content (see Supplementary Text S3), and palindrome abundance.
Core genome features (ploidy, haploid genome size, heterozygosity, repetitive fraction of the genome, and characterisation of TE content) were estimated directly from sequencing reads to avoid potential assembly biases in reference genome-based approaches. The raw reads were publicly available for 27 samples and for three more samples shared by authors on request. We cleaned the raw reads by removing adaptors and low quality bases using Skewer (parameters “-z -m pe -n -q 26 -l 21”) 103.
We used smudgeplot v0.1.3 (available at https://github.com/tbenavi1/smudgeplot) to estimate ploidy levels. This method extracts from the read set unique kmer pairs that differ by one SNP from each other. These kmer pairs are inferred to derive from heterozygous genome regions. The sum of coverages of the kmer pairs is then compared against their coverage ratio. This comparison separates different haplotype structures (Supplementary Figure 1b). The most prevalent structure is then indicative of the overall ploidy of the genome. We used this ploidy estimate in all species, except A. vaga. The most prevalent structure suggested that this species is diploid. A. vaga is well characterized as tetraploid 58, but we were unable to detect tetraploidy because homoeologs are too diverged to be identified as such by the kmer-based smudgeplot method.
Using the inferred ploidy levels, we then estimated genome size and heterozygosity using an extended version of GenomeScope 104. GenomeScope estimates genome wide heterozygosity via kmer spectra analysis, that is, by fitting a mixture model of evenly spaced negative binomial distributions, where the number of fitted distributions is decided given the input ploidy. Estimated distributions correspond to kmers that occur once, twice, etc., in the genome. Fits are then used to estimate heterozygosity, the fraction of repeats in the genome, as well as the 1n sequencing coverage. The latter is subsequently used for estimation of genome size. The definition of heterozygosity for polyploids is not well established (see Box 3), but GenomeScope distinguishes different types of heterozygous loci in polyploids (as shown in Figure 3).
Kmer spectra analysis is affected by the choice of kmer length. Longer kmers require higher sequencing coverage, but lead to more informative kmer spectra. We have chosen the default kmer size 21 nt for all species except the marbled crayfish, where we chose kmer length 17 nt due to low sequencing coverage.
We quantified transposable elements using DnaPipeTE v1.2 105. The method uses haploid genome size (parameter -genome_size) to subsample sequencing reads to 0.5x coverage (parameter -genome_coverage). Subsampled reads are then assembled using an assembler that can deal with uneven coverages, and annotated using the database of known TEs. This process is repeated three times (parameter -sample_number), and the union of results represents the repeat library. Additionally, repeats are annotated as TEs if their sequence matches known TEs by homology (for details see 105). Our reported values of TE loads include only repeats that were annotated as TEs, i.e., we did not include ‘unknown’ repeats which consist of tandem repeats (satellite repeats), duplications or very divergent/unknown TEs.
The palindrome analysis was based on genome assemblies and their published annotations, from 27 samples of 22 species (annotations were not available for D. pulex and A. rufus). We performed collinearity analysis using MCScanX (untagged version released 28.3.2013) 106, allowing even a single gene to form a collinear bloc (parameter -s) if there were fewer than 100 genes in between (parameter -m). The output was then filtered to contain only blocs on the same scaffold in a reverse order. Furthermore we filtered all the homologous gene pairs that have appeared on the same strand. All the remaining blocks are palindromes, blocs built of reverse complementary genes on the same scaffold. See Supplementary Text S2 for more details.
Funding
This study was supported by funding from Swiss SNF grants PP00P3_170627 and CRSII3_160723.
Author contributions
K.S.J. and T.S. conceived the study, K.S.J. and J.B. collected the data; K.S.J. and T.R.R.B. performed the analyses; K.S.J., J.B. and T.S. reviewed the literature; K.S.J., T.S., J.B. and M.R.R. wrote the manuscript; All authors were involved in discussions about results and interpretations.
Competing interests
None declared.
Acknowledgments
We would like to thank Philipp H. Schiffer, Jean-Francois Flot, Reuben W. Nowell, Yuko Ulrich, Dave Lunt, Julian Gutekunst and Mark Blaxter for comments and discussions. We would also like to thank all the authors who deposited their data in public databases with correct metadata and all the authors who were helpful when we asked for data or clarifications.
Footnotes
Data Availability: The code of the pipeline for gathering data, calculating and plotting results is available at https://github.com/KamilSJaron/genomic-features-of-asexual-animals; the majority of sequencing reads are available in public databases under the accessions listed in Supplementary Table 1. The data without publicly available sequencing reads were obtained via personal communication with the corresponding authors.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.↵
- 107.↵
- 108.↵
- 109.↵