Genome architecture and evolution of a unichromosomal asexual nematode

SUMMARY

Asexual reproduction in animals, though rare, is the main or exclusive mode of reproduction in some long-lived lineages. The longevity of asexual clades may be correlated with the maintenance of heterozygosity by mechanisms that rearrange genomes and reduce recombination. Asexual species thus provide an opportunity to gain insight into the relationship between molecular changes, genome architecture and cellular processes. Here, we report the genome sequence of the parthenogenetic nematode Diploscapter pachys with only one chromosome pair. We show that this unichromosomal architecture is shared by a long-lived clade of asexual nematodes closely related to the genetic model organism Caenorhabditis elegans. Analysis of the genome assembly reveals that the unitary chromosome arose through fusion of six ancestral chromosomes, with extensive rearrangement among neighboring regions. Typical nematode telomeres and telomeric protection-encoding genes are lacking. Most regions show significant heterozygosity; homozygosity is largely concentrated to one region and attributed to gene conversion. Cell-biological and molecular evidence are consistent with the absence of key features of Meiosis I, including synapsis and recombination. We propose that D. pachys preserves heterozygosity and produces diploid embryos without fertilizationthrough a truncated meiosis. As a prelude to functional studies, we demonstrate that D. pachys is amenable to experimental manipulation by RNA interference.

eTOC Blurb

By genome sequencing, Fradin et al. discover that the single chromosome in an asexual group of nematodes resulted from a fusion of six ancestral chromosomal domains. Due to the lack of recombination between alleles at most loci, high heterozygosity has evolved, providing one explanation for the unexpected longevity of this asexual lineage.

INTRODUCTION

Sex is generally considered to provide an adaptive advantage over asexual reproduction for the long-term survival of species because the reshuffling of genetic material by meiotic recombination allows new beneficial allelic combinations and the removal of deleterious alleles [1, 2]. Shorter term, heterozygosity resulting from sexual reproduction may be advantageous by masking recessive deleterious mutations [3]. Sexual reproduction is also the most prevalent reproductive mode among eukaryotes and is considered ancestral, yet asexual species have evolved from sexual ancestors [4]. Because asexual eukaryotes are scarce but occur in many kingdoms, debate about the benefits of asexual reproduction versus sexual reproduction is ongoing [5–8]. Consistent with their rarity, most eukaryotic asexual lineages are short-lived [8]. However, some exclusively asexual lineages—such as bdelloid rotifers and darwinulid ostracods—appear to be surprisingly long-lived [9, 10].

Several mechanisms that maintain heterozygosity with little or no recombination may contribute to the longevity of asexual clades [3, 11]. Successful asexual lineages commonly generate heterozygous progeny either by skipping Meiosis I or by fusing two genetically distinct meiotic products (“central fusion”). Chromosomal rearrangements can further promote genetic diversity and their preservation may be facilitated by suppression or loss of homologous chromosome pairing and meiotic recombination. For example, the genome of the bdelloid rotifer Adineta vaga [12] is incompatible with regular meiosis: extensive rearrangements preclude homologous pairing and segregation, with some allelic regions even residing on the same chromosome rather than on homologues. This organization arose from a genome duplication (tetraploidy) followed by multiple translocations and the subsequent divergence of ancestral pairing regions [12]. Mechanisms such as polyploidization by hybridization may disrupt meiosis and lead to an unconventional genome organization. For example, triploids such as the nematode Meloidogyne incognita cannot segregate chromosomes evenly during meiosis and thus can only survive as asexual species [13]. In a counter-example, obligately asexual Daphnia pulex lineages are extremely short-lived despite meiotic suppression, likely due to loss of heterozygosity by interallelic gene conversions and accumulation of deletions [14]. Aside from these examples, we know little about the genome architecture of evolutionarily long-lived asexual species.

Here we explore the reproductive biology and genome sequence of Diploscapter pachys, Steiner 1942, an asexual member of the Protorhabditis group of nematodes [15] and close relative of the sexually reproducing model organism Caenorhabditis elegans. Caenorhabditis is a clade of dioecious, obligately crossing species; C. elegans is a recently evolved androdioecious hermaphrodite that can cross or self [16, 17]. We confirm the existence of a long-lived asexual clade within Protorhabditis. Because species in the Protorhabditis clade are morphologically, taxonomically, and geographically diverse, they are a good model for elucidating how long-lived asexual lineages can evolve. Whereas other species in the same family, e.g. C. elegans, have six chromosomes, rarely five or seven [18–20], D. pachys, like its recently sequenced relative Diploscapter coronatus, has only one chromosome pair [21, 22]. This unusual karyotype has been described in only two other eukaryotes, both sexual species: the nematode Parascaris univalens [23] and the ant Myrmecia croslandi [24]. Genome comparisons of D. pachys and sexual relatives revealed that fusion of all ancestral chromosomes and extensive rearrangements among neighboring regions occurred in the asexual lineage. We find no evidence of typical nematode telomeres in the D. pachys genome, suggesting a possible fusion mechanism. Most regions show a high degree of heterozygosity, as in D. coronatus [22]. We provide cell biological and molecular evidence for a model in which D. pachys oocytes undergo meiosis without synapsis or recombination and auto-activate, resulting in diploid embryos without fertilization.

RESULTS

Asexual reproduction and a unichromosomal karyotype evolved jointly in the Protorhabditis group

Based on molecular phylogenetic analyses [25–28], Diploscapter belongs to the Protorhabditis group, the sister group of Caenorhabditis that includes the genera Protorhabditis, Prodontorhabditis and Sclerorhabditis (Figure 1). Reports of asexual reproduction in Diploscapter species are based on the absence of observed males, sperm and/or mating [32, 33]. To confirm that asexual reproduction is common in this clade, we looked at early embryonic development in seven species of the Protorhabditis group. We observed a single nucleus in presumed asexually reproducing species, but two pronuclei that fuse before the first embryonic division in Protorhabditis species with equal numbers of males and females (Figure 1A, Figure S1A–C). We detected sperm in only one species lacking males, Protorhabditis sp. 4 (n=13, Figure S1D–E); however, the single nucleus in one-celled embryos suggested no genetic contribution from sperm. Sperm could still be required for egg-activation (pseudogamy), as reported for another nematode Rhabditis aberrans [34]. The rare males observed in D. lycostoma appear to make no genetic contribution: we detected no sperm in female gonads, no mating when males were present, and only a single nucleus in one-celled embryos.

Figure 1. Phylogenetic relationships, reproductive modes, and karyotypes of the Protorhabditis group and related nematodes.

(A) Molecular phylogeny (see STAR Methods) and key aspects of reproductive biology for sexually reproducing lineages (orange) and asexually reproducing lineages (purple, gray box) of the Protorhabditis group. C. = Caenorhabditis, D. = Diploscapter, H. = Haemonchus, Het. = Heterorhabditis, P. = Protorhabditis, Pan. = Panagrellus, Poik. = Poikilolaimus, Pr. = Prodontorhabditis, Prist. = Pristionchus. Sources of karyotypic data: a [29]; b [30]; c, wiki.wormbase.org; d [19]; e [20]; f [31]; all others from this study. (B–F) Condensed chromosomes in C. elegans (B) and members of the Protorhabditis group: (C and D) sexually reproducing species in diakinesis form bivalents (tetrads); and (E and F) univalents (separate dyads) at equivalent stage in asexually reproducing species. Homologous chromosome pairs are arbitrarily enumerated. Fixed oocytes are stained with DAPI. All micrographs are maximum intensity z-stack projections. Scale bar, 5 μm. See also Figure S1, Table S1 and STAR Methods).

A molecular phylogeny shows that all parthenogenetic species form a clade derived from a sexually reproducing stem species (Figure 1A; STAR Methods). The potentially pseudogamous Protorhabditis sp. 4 occupies a basal position in the asexual clade, consistent with the view that pseudogamy can be an intermediate step in the transition to asexual reproduction [35]. By comparing divergence rates with those of Caenorhabditis [16], we estimate that the asexual clade originated ca. 18 Myr ago (Table S1; STAR Methods), indicating that this asexual lineage is one of the oldest described so far [9].

To investigate the origin of the unichromosomal karyotype, we counted chromosomes in oocytes and early embryos from seven Protorhabditis-group species (Figure 1). Rhabditid nematodes generally have six chromosome pairs (Figure 1A–D). Whereas sexually reproducing Protorhabditis-group species had six or seven pairs (Figure 1A–D), all asexually reproducing species had only one (Figure 1A,E,F); we detected no intermediate karyotypes. The most parsimonious evolutionary reconstruction thus involves a single transition from multiple chromosomes to a single chromosome pair in the stem species of the asexual clade. The drastic drop in the number of chromosomes coincides with the transition from sexual to asexual reproduction, drawing a potential connection between these two evolutionary changes.

While surveying early embryogenesis in this clade, we identified D. pachys as particularly amenable to cultivation in the laboratory [36, 37]. Moreover, providing potential for functional studies, at least the soma responds to RNAi feeding: 86% (n = 81) of D. pachys nematodes died upon RNAi targeting all five C. elegans actin genes, vs. 0% (n = 64) of controls; see STAR Methods and [37]). The Protorhabditis group also holds interest for functional studies into the evolution of early embryogenesis; for example, these species show a linear arrangement of blastomeres at the four-cell stage rather than the more typical rhomboid pattern seen in other rhabditids and which in C. elegans is important for proper specification of the body plan [36, 37].

The single chromosome resulted from fusions

The singular chromosome pair within the Protorhabditis group could result from chromosome loss, fusions, or both. To investigate further the structure and origin of the single chromosome characteristic of the asexual lineage, we sequenced the D. pachys genome (STAR Methods).

The D. pachys genome was assembled de novo from Illumina paired-end libraries, and scaffolding was improved using PacBio reads (Figure S2). We obtained a genome assembly of 158 megabases (Mb) with an N50 of 124 kilobases (Kb), comparable to other recently published nematode genome assemblies (Table S2). Of the assembly, 97% is contained in 2,734 scaffolds larger than 3 Kb, with an average gene length of 3,292 base pairs (bp); most genes should thus be entirely contained within a single scaffold. Using independent RNA-seq data, we estimated that coverage of the assembled D. pachys genome is ≥94%. To annotate the genome, we predicted gene models in silico, informed by both the genome assembly and RNA-seq transcripts assembled de novo from Illumina paired-end libraries (Figure S2). We detected 247 out of 248 genes encoding deeply conserved eukaryotic proteins [38], signaling that close to all D. pachys genes are represented (STAR Methods).

Several pieces of evidence suggest that the single D. pachys chromosome did not result from a large-scale loss of chromosomal material. First, the haploid size of its genome is ~88 Mb (STAR Methods), similar to that of the six-chromosome C. elegans genome (100 Mb) [39]. Second, we detected no large-scale gene loss in D. pachys compared with C. elegans and other nematodes (Figure S3). Third, all C. elegans chromosomes encode a very similar proportion of proteins with homology to a D. pachys protein (between 63 and 70%; Table S3), ruling out the loss of individual chromosomes. Last, a fusion hypothesis is consistent with the larger physical size of condensed chromosomes in the unichromosomal species compared with those from sexual relatives (Figure 1B–F).

To reconstruct how this single chromosome may have arisen, we inferred the likely ancestral chromosomal origin of each D. pachys scaffold. The strategy was to use C. elegans as a reference for assigning ancestral chromosomal identity. To test if this was a valid approach, we confirmed that the chromosomal identity for most orthologs has been conserved between C. elegans and C. briggsae (for which the genome is resolved at the chromosomal level [40, 41]) and Haemonchus contortus [42]. H. contortus also has six chromosomes, but is much more distantly related to Caenorhabditis than is Caenorhabditis to Diploscapter, thus serving as an outgroup representative for these comparisons (Figures 1A, S4). Conservation of chromosomal identity for orthologs shared between C. elegans or C. briggsae and H. contortus is 77.6% and 78.6%, respectively (Figure S4) [26, 42].

We then searched for traces of macrosynteny between D. pachys scaffolds and genome sequences of C. elegans and C. briggsae. We assigned a C. elegans chromosomal identity to each D. pachys scaffold that contained a plurality of genes whose C. elegans orthologs reside on a particular chromosome (Figure S4A–B). Using this method, the majority of D. pachys scaffolds could be matched with a specific C. elegans chromosome: for the 428 scaffolds containing ten or more genes with C. elegans orthologs, on average 74.8% of homologous genes map to the same C. elegans chromosome (74.1% to the same C. briggsae chromosome). Thus, genomic regions in D. pachys have generally maintained the same chromosomal identity since diverging from Caenorhabditis, consistent with the high chromosome-level macrosynteny detected between other nematodes [42–46]; however, more rearrangements occurred in the lineage to Diploscapter since its common ancestor with Caenorhabditis than occurred during the much longer divergence between Caenorhabditis and Haemonchus.

We attempted to infer an order of fusion events between ancestral chromosomes based on syntenic comparisons and known patterns of rearrangements. The chromosome-level macrosynteny between nematodes suggests that segmental rearrangements occur more often within than between chromosomes [44]. The rate of rearrangements between chromosomal regions also correlates with their spatial proximity in three dimensions [47]. Assuming that genomic regions located closer together on a linear chromosome are also closer spatially, the likelihood of rearrangement between regions should drop as the chromosomal distance between them increases (Figure 2A). Specifically we expect that, in a relative of C. elegans in which six individual chromosomes have persisted, such as H. contortus (Figure 1A), inter-chromosomal rearrangements are rare and similar in rate between all chromosomes. In contrast, we expect that in species with a single fused chromosome such as D. pachys, previously disjoint ancestral chromosomal domains (ACDs) that are now contiguous will have greater opportunity for rearrangement overall. Furthermore, the frequency of rearrangements between ACDs should correlate with their proximity along the chromosome. In species with a single chromosome, ACDs derived from different ancestral chromosomes should thus experience a higher average rate of exchange, as well as greater variation (dispersion) in exchange rates between them, in comparison with species that have retained the ancestral number of chromosomes.

Figure 2. Evidence of chromosome fusion in D. pachys.

(A) Evolutionary scenario for rearrangements on ancestral chromosomes (colored bars) in (left) species in which individual chromosomes persist over time (e.g. C. elegans, C. briggsae or H. contortus) versus (right) species in which all ancestral chromosomes have fused into one (e.g. D. pachys). Colored arrows: intra-chromosomal rearrangements. Inter-chromosomal rearrangements (not represented) are expected to be far less frequent and unbiased with respect to chromosomes involved. Once fused, rearrangement rates should be higher between proximal ACDs than distant ACDs. (B) Estimation of gene exchange between ACDs of D. pachys. Scaffolds are binned by their assigned ACD (rectangular blocks), as inferred from comparisons with C. elegans and H. contortus (see text, STAR Methods, Figure S4). Histograms above each block display the distribution of genes with orthologs mapping to different C. elegans chromosomes (expressed as gene density across all scaffolds in each bin) (Figure S4C). Values for inferred ACDs are italicized. Gene densities for scaffolds with unusually strong matches to other C. elegans chromosomes are underlined; these ACDs show reciprocal rearrangements at higher and more heterogeneous rates than expected for ACDs on independent chromosomes. Outlined histogram bars highlight scaffolds used to infer the order of ancestral chromosomal fusions. We thus infer a spatial order of I-X-III-II (bracket) for ACDs in the fusion chromosome. ACDs IV and V are placed hypothetically, as their relative positions could not be determined. See Figures S3 and S4, Table S3.

Because our outgroup comparison with H. contortus shows that overall macrosynteny of C. elegans is much closer to the ancestral condition than is that of Diploscapter, we used macrosynteny with C. elegans to infer an ancestral chromosome identity (ACD) for each D. pachys scaffold (Figure S4A). In this way, 92% of the assembled sequences for D. pachys (1,655 scaffolds) could be assigned. For each D. pachys scaffold, we then assessed the extent of rearrangement between ACDs by calculating the proportion and (normalized) density of genes whose C. elegans orthologs are located on chromosomes other than the assigned ACD (Figure S4C). For D. pachys we found an average of 4.48 genes per Mb mapping to one of the other five C. elegans chromosomes, versus an average of 0.98 genes per Mb for H. contortus. Overall the H. contortus genome displayed much lower gene density than both D. pachys (this study) and C. elegans [42]. D. pachys ACDs also showed greater variation in the density of genes exchanged between ACDs than H. contortus (observed s.d. is 1.70 in D. pachys vs. 0.79 expected based on the H. contortus distribution, Figure S4C). These results are consistent with more rearrangements occurring between ACDs once they fused.

We next attempted to infer the order of ancestral chromosomes in the D. pachys genome, based on the extent of exchange between scaffolds with an assigned ancestral chromosome identity. We identified specific patterns of rearrangements consistent with an order of ACD fusions (Figure 2B). For example, D. pachys scaffolds in ACDs II and III showed a higher rate of reciprocal rearrangements than expected (8.93 and 8.48 rearranged genes/Mb, respectively, vs. the average 4.48 genes/Mb over all scaffolds). Reciprocal exchange was also greater between ACDs X and I (8.21 and 5.64 genes/Mb) and X and III (6.58 and 5.24 genes/Mb), whereas data for IV and V were inconclusive. Similar patterns of rearrangements were observed when using orthology to C. briggsae (Figure S4D). The data thus allow us to propose that four ancestral chromosomes became fused in the order I-X-III-II (Figure 2B). As we were unable to conclusively place ACDs IV and V in linear order, an interesting possibility that we cannot exclude is that the D. pachys chromosome is circular, even though its condensed form appears linear in the micrographs. Such ambiguity could also be due to different ACD order in the homologous chromosomes (e.g. due to independent fusion events).

Telomere-specific repeats and proteins required for telomere maintenance are not detected

One possible mechanism for the fusion of the ancestral chromosomes involves the alteration of telomeres. Telomeres contain short repeat sequences and associated proteins that constitute a protective cap at the ends of chromosomes, which keep chromosome ends from being recognized as double-stranded breaks (DSBs) by the cell’s DNA repair machinery [48]. Defects in telomere assembly or maintenance, or the absence of telomeres, can lead to chromosomal fusions [48] and thus are plausible mechanisms for fusions in the D. pachys ancestor.

To test this idea, we attempted to detect telomeres or their remnants in D. pachys. Within the phylum Nematoda, telomeric TTAGGC repeats have been characterized and sequenced in C. elegans [49] and its distant relatives Ascaris suum and P. univalens [50]. Clusters of the TTAGGC motif are also found in the genomes of C. briggsae, Pristionchus pacificus, Brugia malayi, Ancylostoma ceylanicum, Haemonchus contortus, Trichuris suis and Romanomermis culicivorax (Figure 3), suggesting that the motif is well conserved across nematodes. It is unclear at present why Meloidogyne hapla and Trichinella spiralis are missing the standard telomere sequence (as well as alternative repeats with strings of Gs) [51].

Figure 3. Detection of telomere repeats in nematode genomes.

Number of “unique” hexanucleotide motifs detected in clusters of ≥3 consecutive repeats in the genome assemblies of 11 species across Nematoda. Totals for a “unique” motif include all circular permutations and their reverse-complements. The number of “microsatellite-specific” AAAAAT repeats is used as a positive control for repeat detection since it is found across eukaryotes [52]. Vertebrate telomere repeats are used as a negative control because they are not expected to be conserved in nematodes. The phylogeny is derived from [15, 53–55]. See also Figure S5.

In contrast, we were unable to detect the TTAGGC motif in the D. pachys genome assembly, despite the clear presence of several other pentanucleotide or hexanucleotide repeats, including the microsatellite repeat AAAAAT (Figure 3). To confirm this, we looked for the TTAGGC motif in the unfiltered raw Illumina reads from the D. pachys genome. The expectation for a species with a similar telomere structure as C. elegans is to find entire reads (~100 bp long) containing TTAGGC repeats, since each of the 12 C. elegans telomeres typically display between 60 and 300 copies [56]. No more than four successive repeats of the telomere motif were found, and in only one out of more than a billion reads. This finding suggests that telomeres are either absent from the D. pachys genome or are highly divergent from the ancestral telomeric sequence.

We also surveyed the D. pachys genome for homologs of conserved proteins required for telomere maintenance. In C. elegans, telomeres are maintained by dedicated protein machinery (Figure S5A) whose disruption can result in chromosomal fusion [57, 58]. Because completeness of the D. pachys proteome is high (Table S4), failure to detect orthologs for the C. elegans proteins is meaningful, particularly when they are clearly present in more distantly related species such as H. contortus. We found putative H. contortus homologs for all ten conserved C. elegans telomere-associated proteins, but no D. pachys homologs of the four conserved mammalian POT1 (Protection Of Telomeres 1) proteins POT-1, POT-2, POT-3 and MRT-1 (Figure S5B), suggesting that maintenance of typical nematode telomere structures has diverged in D. pachys. Notably, MRT-1 is required for telomere repeat addition and C. elegans mrt-1 mutants display chromosome fusions [59]. The combined absence of telomere sequences and maintenance proteins suggests a connection between chromosome fusion and loss of telomeres: perhaps a fusion resulted from altered telomere structure or telomere protection became obsolete after fusion, leading to secondary loss of these genes.

The genome of D. pachys displays high heterozygosity

Similar to some other asexual species, such as the nematode Meloidogyne incognita [60] and bdelloid rotifers [12, 61, 62], the D. pachys genome is highly heterozygous. While analyzing genome completeness using a set of core eukaryotic genes, we found an average of 2.12 D. pachys orthologs for them (Table S4). Our hypothesis that these might be divergent alleles was supported by two additional analyses. First, we observed two peaks in the depth of coverage distribution (Figure 4A): a main peak corresponding to the actual coverage of the diploid assembly (316 reads) and a smaller peak at twice the coverage (633 reads). The main peak corresponds to the presence of two distinct haplotypes (i.e. “heterozygous regions”), while the secondary peak is explained by regions with (nearly) identical sequences that were collapsed by the assembler (SOAP denovo v2) into one (i.e. “homozygous regions”). Second, using MCScanX [63] to detect gene collinearity, at least 51% (14,020) of D. pachys protein-coding genes (27,341) could be organized into collinear segments of 3 or more coding sequences (CDSs) that segregate into two distinct sets of scaffolds (Figure 4B, Table S5). Thus, the genome assembly of D. pachys displays precisely two distinct haplotypes, also indicating no variation among the individuals in the sequenced population. Given that we sequenced nematodes derived from a single mother, these data are consistent with a clonal type of asexual reproduction that maintains the same level of heterozygosity among all progeny.

Figure 4. Heterozygosity in the D. pachys genome.

(A) Distribution of read depth based on alignment of Illumina reads. Primary and secondary peaks represent heterozygous regions that assembled separately (316 reads) and homozygous regions (633 reads). (B) Collinearity between the five largest scaffolds (colored, right side) and thirty other scaffolds of the genome assembly (grey, left side), labeled in Kb. Ribbons: MCScanX collinear segments, showing number n of collinear proteins detected on each of the five largest scaffolds and average percent nucleotide identity between pairs of CDSs. (C) Localization of homozygous (orange) and heterozygous (grey) regions in the D. pachys genome. Outer circle: D. pachys genome organized into bins of scaffolds assigned to their putative ancestral chromosomal identity based on comparisons with C. elegans (Figure S4B). Within ACDs, scaffolds are ordered by average position of orthologs on C. elegans chromosomes. Inner disk: histogram of depth of coverage across the genome smoothened by using a 1-Kb sliding window with 100-bp step (see STAR Methods). Orange: homozygous regions with 2× depth of coverage. Regions with highest homozygosity are found on ACD I and X scaffolds (dashed ellipses). See also Figure S4 and Table S5.

Taking the presence of two haplotypes into consideration, we estimate the haploid genome size to be around 88 Mb and the haploid number of protein coding genes to be 17,130 (Table S2), about 2,000 fewer genes than in related sexual species like C. elegans. The average DNA sequence identity between pairs of collinear (allelic) CDSs is 96% (Figure 4B; based on 4,203 pairs of CDSs). Homozygous regions (2× coverage) are no longer than 220 Kb, representing 11% of assembled sequences. Notably, 84% of homozygous regions localize to ACDs I and X (Figure 4C), which are neighbors in the inferred chromosomal fusion, and constitute close to a third of sequences in these scaffolds (I, 27%; X, 31%).

D. pachys skips Meiosis I during oocyte maturation

The considerable heterozygosity in a sample that originated from a single D. pachys founder suggests that heterozygosity has been maintained over many generations of parthenogenetic reproduction. Maintenance of heterozygosity during parthenogenesis can result from several types of modified meiosis: (1) pre-meiotic genome doubling where the entire genetic material from the mother along with its heterozygosity is preserved in the first meiotic division, which thus loses its reductional property; (2) central fusion, where two genetically distinct meiotic end-products fuse, forming a diploid heterozygous embryo; or (3) skipping of the first reductional division (Meiosis I), whereby the mother’s entire genetic material and its heterozygosity are also maintained [64, 65]. Complete heterozygosity is maintained unless crossing over or gene conversion occurs (in type 2 or 3) [64, 65].

To determine how oocytes are generated in D. pachys, we first observed DNA structures in fixed D. pachys oocytes and early embryos stained with DAPI (Figure 5). Replicated chromosomes begin to condense in prophase (Figure 5A, A′). Fully condensed chromosomes take on an appearance similar to holocentric metaphase figures from other rhabditid species but do not appear to engage in meiotic synapsis (Figure 5B, B′; cf. Figure 1B–D). Oocytes show a single nuclear division (Figure 5C, C′) with no evidence for an additional meiotic division. In early embryos, one set of segregated chromosomes decondenses and moves centrally (Figure 5D, D′). The other remains compact (Figure 5D″) and localizes near the cortex, where in C. elegans and other animals the meiotic divisions and polar body extrusion take place [66]. This compact structure persists in subsequent divisions, and we interpret it as a residual polar body that has not been extruded. Because we see a polar body—the hallmark of meiosis—we infer that D. pachys oocytes undergo partial meiosis.

Figure 5. Oocyte maturation in D. pachys.

(A–D) Stages of maturation: maximum intensity projections of microscopic images from DAPI-stained D. pachys oocytes (A–C) and early embryo (D). Dashed lines outline oocyte or embryo (A–D) and nuclear envelope (A). Scale bar, 10 μm. (A′–D′,D″) Magnified nuclear (A′–D′) and polar body (D″) DNA from the image above. Scale bar, 5 μm. (A) Initial condensation of homologous chromosomes in the oocyte. (B) Maximum condensation of replicated homologous chromosomes. Note that no pairing is observed (contrast with Figure 1B–D). (C) First and only division during oocyte maturation, in which identical sister chromatids have just separated. (D) One-cell embryo before the first mitotic division. Decondensed nuclear DNA (D′) and compacted polar body (PB, D″) are visible. The non-extruded PB remains detectable in later embryos. (E, F) Schematic diagrams of (E) C. elegans meiosis and (F) a model of parthenogenesis in D. pachys. The single pair of homologous chromosomes duplicates and condenses but never forms synapsed bivalents. Sister chromatids separate during a single division, generating a non-extruded diploid PB and diploid embryonic nuclear DNA. The nuclear envelope is not present at these stages (dashed circles).

Based on our cumulative observations, we propose a model of asexual reproduction in D. pachys (Figure 5F) in which oocytes skip Meiosis I and undergo a single nuclear division that separates sister chromatids and not homologous chromosomes (first division restitution without crossing over [65]). In support of this model, we detected no recombination events between eight heterozygous regions distributed along the large D. pachys chromosome (STAR Methods). In the absence of chromosome pairing and synapsis, the single nuclear division in the oocyte proceeds as in Meiosis II but produces two diploid products. This division is not followed by cytokinesis, yet retains the meiotic characteristics of a small spindle near the cortex and the formation of a residual polar body from one of the products. This model explains how diploidy and two distinct haplotypes are maintained in D. pachys in the absence of sexual reproduction.

Genes essential to Meiosis I are missing

Since D. pachys appears to lack the first meiotic division, we might expect less conservation of genes required for chromosome pairing, synapsis, and homologous recombination. We thus searched the proteome of D. pachys for homologs of 91 proteins involved in C. elegans meiosis [67] (Table S6). Forty-three of these genes can be considered “meiosis-specific” since they have no reported role outside of meiosis. The H. contortus proteome served as control for detection.

More meiotic protein-coding genes could not be detected (due to absence or divergence) in D. pachys (25/91) than in the outgroup representative H. contortus (15/91) (Figure S6). This set of genes is very similar to the set that was undetected in the D. coronatus genome [22]. Ten of the 25 undetected genes in D. pachys are conserved in C. elegans and H. contortus; seven are meiosis-specific with key roles in homologous chromosome pairing and recombination. We could not detect orthologs of the four C. elegans pairing center proteins ZIM-1, ZIM-2, ZIM-3, and HIM-8, whose genes are organized in an operon in C. elegans and H. contortus. Each of these proteins binds to sites called pairing centers, which are required for synapsis of homologous chromosomes [68–70]. Consistent with a lack of functional pairing centers, D. pachys chromosomes were never observed to form bivalents (tetrads) (Figure 5B′, n=12), which are clearly evident in sexually reproducing relatives (Figure 1B–D). Additional key meiosis-specific proteins could not be detected in D. pachys: XND-1, a chromatin factor regulating the global distribution of crossovers [71]; HIM-17, a chromatin factor required for meiotic DSB formation and recombination [72]; COM-1, a crucial regulator required for proper DSB repair during meiotic recombination [73]; and REC-8, a highly conserved meiosis-specific cohesin subunit [74]. The conserved inner kinetochore protein HCP-4/CENP-C is also not detected in D. pachys. HCP-4 provides a base for the assembly of outer kinetochore proteins and is needed for the resolution of sister centromeres in diakinesis and their proper orientation for capture by spindle microtubules [75]. Our inability to detect these proteins is consistent with the hypothesis that D. pachys skips the first meiotic division and has lost homologous pairing and crossover mechanisms. Notably, C. elegans rec-8 mutants show a similar phenotype: during Meiosis I, sister chromatids separate instead of homologous chromosomes and the embryos fail to extrude a second polar body [74]. When rec-8 mutant hermaphrodites are mated to wild-type males, a quarter of the embryos hatch and are fertile [74], suggesting that perturbing cohesion during Meiosis I is not lethal and can be achieved with a single mutation.

Despite the lack of several meiosis-specific genes, essentially all genes for key proteins in gene conversion [76] are present in the D. pachys and D. coronatus [22] genomes, including HIM-6/BLM, MRE-11/MRE11A, RAD-51/Rad51, RAD-54/Rad54, RPA-1/RPA, SPO-11/SPO11 and TOP-3/Topo-III-alpha. Along with the presence of homozygosity in particular regions of the genome (Figure 4B, C), this suggests that gene conversion is a viable mechanism in Diploscapter, at least via the SDSA (synthesis-dependent strand-annealing) pathway [76].

DISCUSSION

In asexual eukaryotes, meiosis has been modified to allow the generation of an embryo without fertilization. Consequently, the organization of genetic material in parthenogenetic species can be unconventional, as in the rotifer Adineta vaga [12]. Asexual species thus provide a unique opportunity to gain insight into the connection between molecular changes, chromosome structure, and cellular processes in eukaryotes.

Here, we show that species in an asexual clade closely related to the well-studied Caenorhabditis clade also share an extremely rare unichromosomal genome organization. Our phylogeny and comparative sequence analyses indicate that the single D. pachys chromosome arose by fusion of six (or seven) ancestral chromosomes. Recently, the Diploscapter coronatus genome was also sequenced [22]. In preliminary comparisons, we found that several homologous alleles in D. pachys and D. coronatus have diverged nearly as much as the same homologous loci have between the biological species C. briggsae and C. nigoni, supporting their designation as different Diploscapter species. Like D. pachys, D. coronatus has truncated meiosis, maintains high heterozygosity and has a single chromosome pair [22].

What factors could have favored the evolution of the unichromosomal karyotype within the Protorhabditis group? The Caenorhabditis clade and most other nematode species have holocentric chromosomes, in which kinetochores are assembled along the chromosome rather than confined to a single, discrete centromeric region. Holocentric chromosomes are more tolerant of fusion and fission because DSBs do not result in centromere-less fragments and fusions do not result in chromosomes with competing centromeres, both of which lead to genome instability in species with point centromeres [48]. Thus, holocentrism could have provided permissive conditions for evolution of the unichromosomal karyotype.

Defects in telomere maintenance can lead to chromosome fusions [77, 78]. Fusions can be elicited by mutations in telomere binding proteins or by misregulation of the DNA repair machinery or cell cycle checkpoints that monitor telomere integrity [48, 79]. Chromosomal fusion usually occurs by non-homologous end joining and is one of several mechanisms that compensate for telomere loss [48, 79]. Notably, typical telomere repeats and several conserved genes required for telomere maintenance are not detected in the D. pachys genome. Degeneration of telomers and maintenance machinery could have facilitated or resulted from chromosome fusion, especially if the chromosome were circular, lacking ends needing protection. Intriguingly, one way that fission yeast survives loss of telomerase is by circularizing its chromosomes, which also results in severely impaired meiosis [48].

Although not as unusual as a single chromosome, parthenogenesis is also an exceptional feature within nematodes. It is found mostly in phylogenetically isolated species and often combined with genome duplication or hybridization [80]. What molecular changes might have presaged the evolution of parthenogenesis in D. pachys? Parthenogenesis in D. pachys involves an alteration of meiosis in which the reductional division during Meiosis I is skipped; the two chromosomes condense but do not pair or synapse. In C. elegans, similar perturbations of meiosis can be accomplished with very few molecular changes. For example, the phenotype of C. elegans rec-8 mutants is reminiscent of the modified meiosis in D. pachys [74]. Interestingly, rec-8 is one of the conserved meiosis genes that was not detected in the D. pachys and D. coronatus genomes [22]. Thus, rec-8 is a candidate gene in which a mutation could have preceded the evolution of parthenogenesis in the Diploscapter lineage.

Several other key proteins (e.g. ZIM-1-3, HIM-8) required for homologous recombination and pairing during Meiosis I in nematodes [69, 81] have diverged or are missing in both D. pachys and D. coronatus [22]. After transition to a single chromosome in the ancestor of D. pachys, multiple pairing centers in different chromosomal domains would have become unnecessary and perhaps even deleterious, since genome integrity could be jeopardized by multiple forces pulling in different directions at once. Chromosomal fusion could thus favor the loss of chromosome pairing and the divergence or degeneration of pairing center proteins.

One feature of the Diploscapter genomes arising from the alteration of meiosis and recombination may be advantageous for maintaining parthenogenetic lineages [3]: both D. pachys and D. coronatus [22] are highly heterozygous, with ~4% average difference between alleles. This is comparable to the highest values of genetic diversity reported so far in sexually reproducing species [82–84]. Preliminary phylogenetic comparisons indicate that the two alleles for several genes have evolved independently at least since D. coronatus and D. pachys diverged if not before. At least some of the observed heterozygosity might even predate the evolution of parthenogenesis; subsequently, nonrecombining homologs would have diverged by accumulated mutation. High heterozygosity is also expected if the parthenogenetic ancestor arose by hybridization of two different Diploscapter species. The latter model was favored for D. coronatus based on the unsubstantiated assumption that the genome-wide dN/dS (the ratio of nonsynonymous to synonymous divergences between alleles) should be <1 after hybridization but >1 for independently accumulating mutations (“Meselson effect”) [22].

However, none of these scenarios explain why 84% of homozygous regions in the D. pachys genome reside on ACDs X and I (Figure 4). Here we propose one plausible scenario (Figure 6). In a typical XY or XO sex determination system, the X chromosome experiences a lower effective population size (N_e) than the autosomes because it is present in only one copy in males, whereas autosomes are present in two copies in both sexes [85]. The X chromosome consequently tends to show reduced heterozygosity and higher linkage disequilibrium relative to autosomes. In D. pachys, ACDs X and I are inferred neighbors in the fusion chromosome (Figure 2) and share the same level of homozygosity (Figure 4C). Thus, they likely shared an evolutionary history in linkage disequilibrium that was independent of the other chromosomes for some time (Figure 6). Fusion of X and I would have formed a neo-X chromosome and a neo-Y (the unfused homologue of I in males); stronger drift on this neo-X would have resulted in greater homozygosity shared by ACDs I and X. Higher similarity between alleles on ACDs X and I would also render them more susceptible to repeated gene conversion, potentially explaining why some reads were combined by the genome assembler (Figure 4B). The large allelic difference across other regions (averaging 4%) would explain why they might have escaped gene conversion [76], preserving the high heterozygosity, in contrast to Daphnia pulex, for example [14]). This scenario assumes an ancestral intermediate that reproduced sexually (Figure 6, Step 2). Although we cannot distinguish if the autosomes fused to the neo-X before, during or after the evolution of parthenogenesis, their striking difference in heterozygosity levels may favor a final fusion after the loss of recombination. Regardless of the actual sequence of events in its evolution, preservation of high heterozygosity over most (89%) of the D. pachys genome likely explains the longevity of this clade of asexual animals by circumventing loss of complementation, otherwise proposed to be a fitness cost to asexual species [3].

Figure 6. Possible 3-step scenario for the evolution of a single chromosome and parthenogenesis.

(Step 1) Sexual ancestor had 6 pairs of homologous chromosomes (colored bars) with XX females and XO males. (Step 2) In a still-sexual ancestral intermediate, the X chromosome fused with chromosome I, forming a neo-X; neo-Y resulted from an unfused chromosome I. Due to lower effective population size (N_e) of the neo-X than the autosomes, drift would have resulted in higher homozygosity on the neo-X. Note that the autosomes may have fused before, around the same time as or after the X-I fusion; alternatively (not shown), those fusions might have occurred in association with events described in Step 3. (Step 3) Parthenogenesis evolved in conjunction with loss of Meiosis I. Autosomes fused with the neo-X to produce a single large chromosome. A hermaphroditic or pseudogamous intermediate is possible but not necessary. The absence of homologous recombination during Meiosis I preserves heterozygosity across most of the genome. Especially in the former neo-X region, gene conversions maintain homozygosity. Additional details in the Discussion.

Some of the above inferences, like the fusion order of ACDs and possible circularity of the D. pachys chromosome, could be tested by improving genome sequence contiguity to the full chromosome level. For example, we cannot rule out that our inability to order all six ACDs unambiguously is due to structural differences between the two homologous chromosomes, which, like inversions and translocations, would contribute to meiotic crossover suppression [86]. Other hypotheses, like those linking divergence of the telomere maintenance machinery with chromosome fusion and the independent evolution or gene conversion of alleles, could be tested by sequencing genomes of other known sexual and asexual species related to Diploscapter. Any discovery of a new species in this group with an intermediate karyotype could also shed light on the temporal order of the chromosome fusions.

In summary, D. pachys provides a new example of a successful asexual animal lineage, adding insights to the different ways that reproductive biology can be modified during evolution without deleterious effects on fitness. As in other successful asexual lineages of animals, heterozygosity is preserved, in this case through a modified meiosis that yields diploid products. Genes required for synapsis, homologous recombination and telomere maintenance are not detected, consistent with the absence of crossing-over and the unusual unichromosomal karyotype. Our study describes a novel, experimentally accessible model asexual organism for studying how genome structure and reproductive mode evolve. Future comparative studies of the very different genomic architectures of D. pachys and closely related organisms, such as the sexual model C. elegans, hold promise to yield insights into the constraints and requirements for the maintenance of genome and chromosome integrity and function.

STAR⋆METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, David H. A. Fitch (david.fitch@nyu.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Maintenance of Protorhabditis group nematodes

Maintenance of D. pachys

D. pachys strain PF1309 nematodes were grown at 20°C on Nematode Growth Medium (NGM, USBiological) agar poured into 35mm petri plates. NGM agar was layered with an HT115 Escherichia coli bacterial lawn used as food source for the nematodes. To propagate populations, small chunks of agar containing D. pachys embryos were transferred from crowded petri plates to fresh ones. To avoid contamination and synchronize populations of nematodes, transferred chunks were then drowned on the plate in a drop of bleach. The bleach killed most contaminations as well as nematode larvae and adults, but nematode embryos survived and hatched if exposure to the bleach was short. Generating the D. pachys strain PF1309. In order to limit genetic variation, all the DNA and RNA samples used for sequencing were extracted from a population derived from a single mother from strain PS2123. For nine times, we transferred only one offspring to a new plate before propagating the population anew as described previously. We named the newly generated strain PF1309 to distinguish it from its parent strain.

Other Protorhabditis-group strains

Strains were cultured at 20°C on medium petri plates poured with 1% agar supplemented with 5 μg/ml cholesterol. Strains were maintained by transferring a small chunk of agar from a crowded plate to a fresh one. A chunk of NGM agar was added on top of the water plate, except for Pr. wirthi for which a small piece of dried brown algae was added. The following strains were used: Pr. wirthi (DF5074), Pr. prodontis (SB435), Protorhabditis sp. 2 (SB406), Protorhabditis sp. 4 (JB122), Diploscapter sp. 2 (JU359), D. lycostoma (PS2017). Strains are available from the New York University Rhabditid Collection (NYURC, david.fitch@nyu.edu) or the Caenorhabditis Genetics Center (CGC, https://cgc.umn.edu).

METHOD DETAILS

Genome sequencing

The complete genome assembly and annotation pipeline is summarized in Figure S2. Comparison with other nematode genome projects is summarized in Table S2.

Collection of D. pachys for sequencing

Collection of mixed stage D. pachys PF1309 nematode cultures was performed similarly for DNA and RNA extraction. Nematodes were maintained as described above. One-month old crowded plates were used for collection. Because D. pachys nematodes burrow into agar, only a small portion of the nematode population can be washed off the surface of a plate. To extract nematodes from the agar, we used a version of the Baermann funnel method [99]. Two Kimwipe paper leaves were laid crosswise on top of each other in a funnel connected to plastic tubing closed with a metal clamp. Large chunks of agar were transferred to the Kimwipes, covered with M9 buffer, and left overnight. With the help of gravity, active nematodes crawled to the bottom of the funnel, and were small enough to migrate through the Kimwipes. The nematodes remained suspended overnight in M9, removing some of the bacteria present in the gut. The liquid was then collected in 50 ml plastic conical tubes by releasing the metal clamp. Nematodes were separated from bacteria (mostly their single source of food HT115 Escherichia coli) through sucrose floatation [100]. Clean nematode pellets were flash frozen in liquid nitrogen and stored at −80°C.

Genomic DNA extraction

For each extraction, 600 μl of Cell Lysis Solution from Gentra® Puregene® Tissue Kit (QIAGEN®) were added to D. pachys nematode pellets in microcentrifuge tubes. Tubes were then repeatedly frozen in liquid nitrogen and thawed at 37°C for ten times to facilitate rupture of th e nematode cuticle. Protein and RNA were digested by incubating with Puregene® Proteinase K (20 mg/ml) at 55°C overnight and RNAse A at 37°C for 1h. To pu rify the DNA, phenol/chloroform extractions were done using Phase Lock Gel tubes (5 PRIME). DNA pellets were precipitated with ethanol, and re-solubilized in 100 μl DNA Hydration Solution (QIAGEN®). To ensure that the extracted DNA was of high molecular weight, the quantity and quality were assessed using standard gel electrophoresis with a 0.5–0.8% agarose gel and the Agilent 2100 Bioanalyzer.

Genomic sequencing and read filtering

The DNA was sheared in a Covaris sonicator and Illumina paired-end libraries with insert sizes of 170 bp, 210 bp, and 430 bp, were constructed using the NEXTflex DNA Sequencing kit (Bioo Scientific). Libraries were sequenced with an Illumina HiSeq 2000 system. 2 x 100 bp paired-end reads in FASTQ format were filtered using custom Python scripts to remove low quality reads (<Q20). Reads were trimmed at the first nucleotide with a quality of 2 (undetermined base), and filtered out when containing Illumina adaptor sequences using custom scripts. Quake [89] was also used to error correct the reads. Additional filtering was performed to remove bacterial sequences. Most of these sequences were E. coli sequences, which accounted for 3% of total sequences. Additional bacterial contamination from genera Pseudomonas and Sphingonomas was detected based on the coverage distribution of the genome, and filtered out (<1%).

PacBio library preparation and sequencing were performed according to the manufacturer’s instructions (Pacific Biosciences), and as described in reference [101]. Libraries were sequenced on the Pacific Biosciences (PacBio) RS system of the Icahn School of Medicine at Mount Sinai, NY, using 8 SMRT® Cells. To improve the quality of the PacBio reads, we processed them with the PacBioToCA tool that mapped the shorter high quality Illumina reads to perform error correction of the longer PacBio reads [90]. The average read length after correction was 2.5 Kb.

Genome assembly and coverage

Because of high divergence between nematode sequences even amongst Caenorhabditis species [15], we could not use C. elegans as a reference genome for the assembly of the D. pachys genome. We therefore assembled the genome de novo. We first assembled contigs from the short Illumina reads using the assembler SOAPdenovo v2 [91] run with a kmer of size k=63. PacBio reads were artificially tiled using custom scripts to produce in silico paired-end libraries with an insert size from 1 to 5 Kb. These reads were then used as would “mate-pair” libraries to improve the scaffolding step in SOAPdenovo, along with the short Illumina reads. The scaffolding did improve: the N50 assembly went from 90 Kb without the PacBio to 124 Kb with the PacBio data. We used GapCloser from SOAPdenovo to fill some of the gaps left after scaffolding. Our final assembly is 158 Mb, out of which only 0.77% are undetermined residues (gaps).

Illumina reads were realigned to the assembly to determine the depth of coverage: 270X. The PacBio reads transformed into long-insert paired reads have an artificial depth of coverage of 20X. We also estimated sequencing coverage using the D. pachys RNA-seq data that was collected and sequenced independently from the genomic DNA. The filtered RNA-seq data was assembled using the SOAP-Trans assembler. We then aligned all the assembled transcripts to the genome (BLASTN). Only best hits were retained (e-value cutoff of 1 × 10⁻¹⁰), and showed that more than 94% of the assembled contig sequences aligned to the genome assembly.

Transcriptome Sequencing

RNA extraction

RNA was extracted from frozen pellets of mixed stage Diploscapter. sp. 1 nematode cultures (20–100 μl) using the RNeasy Micro kit (QIAGEN®). Nematode pellets were kept frozen while being ground with a pestle prior to the addition of buffer RLT with β-mercaptoethanol. To homogenize the RNA, the lysate was passed through a syringe with a 20G needle at least 10 times. rRNA was depleted from total RNA using Ribo-Zero™ magnetic gold kit (human/mouse/rat) (Epicentre). The integrity of the RNA was verified using the BioAnalyzer 2100.

cDNA synthesis and Illumina libraries

To create stranded cDNA libraries for Illumina sequencing, we adopted a protocol from the Albert Einstein College of Medicine Genomics core that can be found at: http://waspsystem.einstein.yu.edu/wiki/index.php/Protocol:directional_WholeTranscript_seq.

The first strand synthesis reaction was performed using random hexamer and OligodT primers. To ensure strand-specificity, a solution of dUTP was used in place of dTTP in the second strand synthesis reaction. Illumina paired-end libraries were prepared using the NEXTflex DNA Sequencing kit (Bioo Scientific) with slight modifications: first, no size selection was performed; second, PCR amplification was performed as described in the Einstein protocol, with uridine removal through treatment with Uracil N-glycosylase (UNG) prior to amplification. Sequencing was carried out on an Illumina HiSeq 2000 system.

Transcriptome assembly and genome annotation

Illumina reads were filtered as described for the genome assembly. We assembled the transcriptome independently of the preexisting D. pachys genome assembly using the RNA-seq de novo assembler Trinity [92], which produced 872,066 initial transcripts. We then realigned these transcripts to the genome using the software PASA (Program to Assemble Spliced Alignments) [93, 95], which generated 252,494 evidence-driven consensus gene models. We used the TransDecoder tool [94] to extract a D. pachys training set consisting of 7,524 non-overlapping Open Reading Frames (ORFs) based on the PASA transcripts. We trained the gene predictor Augustus [96] with the D. pachys training set. We then ran Augustus on the D. pachys’ genome assembly, which predicted 39,821 in silico gene models. EVidenceModeler (EVM) [95] was used to combine the Augustus in silico gene models and the PASA evidence-driven gene models into consensus gene structures. Finally, we ran PASA again to adjust splice junctions, add UTR information when detectable, and include isoforms. The final gene predictions for D. pachys include 27,341 coding sequences, with a total of 38,422 isoforms.

Homology detection

We used three different methods to establish homology relationships between D. pachys proteins and proteins of other organisms:

First, to detect homology with a large number of other organisms, we used the tool Exonerate [97] to compare the final 27,341 predicted protein sequences against a database consisting of the 20 nematode proteomes available on the WormBase web site (www.wormbase.org, release WS241, date 3/17/2014), as well as the Uniref90 protein database [102]. 77% of the proteins shared some homology with at least one protein from another organism. 74% shared homology with at least one protein from another nematode.

Second, for protein conservation and synteny analyses, one-to-one reciprocal homologs between C. elegans and D. pachys, C. elegans and H. contortus, C. briggsae and D. pachys, as well as C. briggsae and H. contortus, were inferred by comparison of all proteins by BLASTp (e-value ≤ 10⁻⁵) using custom perl scripts. We paired proteins based on reciprocal best hits, and allowed for two D. pachys homologs for one C. elegans protein to account for heterozygosity in the D. pachys genome.

Third, to look at one-to-many and many-to-many homology relationships for protein conservation analyses, we computed orthology groups with OrthoMCL [98] for five species of closely related rhabditid species: D. pachys, C. elegans, C. briggsae, H. contortus and H. bacteriophora (Figure S3). We found 4,590 OrthoMCL gene clusters to be orthologous among all species, and 621 being shared between all species except D. pachys. On the other hand, 2,193 gene clusters and 6,856 singleton genes (53% of the total number of D. pachys’ OrthoMCL gene clusters and singleton genes) were found to be unique relative to the other four species. Consistent with their phylogenetic positioning, C. elegans and C. briggsae share the most gene families (12,311). More surprisingly, D. pachys shares approximately 7,000 gene families with either C. elegans or C. briggsae, and so does H. contortus, even though the Protorhabditis group diverged more recently from the Caenorhabditis clade. This suggests that the D. pachys lineage may have diverged more than expected considering its phylogenetic position.

Completeness of the genome

We assessed completeness of the final genomic assembly of D. pachys with CEGMA v2.5 [38] to find how many complete or partial matches it had to 248 highly conserved eukaryotic single-copy genes (Table S4). CEGMA detected full-length matches for 96.8% of its core eukaryotic gene set, and partial matches for 98.8%. The average frequency of matches per conserved gene is 2.12 for complete matches and 2.33 for partial matches. Finally, when we analyzed the eight proteins that CEGMA did not detect in the complete matches category, we found seven of them in our annotation (six with two copies). Therefore, the annotated assembly contains 247 out of these 248 highly conserved eukaryotic proteins, suggesting that close to all of the D. pachys genes are represented in the present genomic assembly.

Genome size and protein-coding genes

There is no available experimental estimate of the genome size of D. pachys. The total size of the assembled sequences is 158 Mb. However, high heterozygosity in the genome was supported by several pieces of evidence. We decided to preserve the heterozygosity in our assembly so we could analyze this important aspect of D. pachys biology. But to estimate the actual haploid size of the genome, we binned regions of the genome based on their Illumina coverage in two categories: once the coverage depth (heterozygous regions), or twice the coverage of depth (homozygous regions). For that, we used a sliding window analysis with a window size of 1 Kb, and a sliding step of 100 bp. Using custom R scripts, we calculated that 11% of the genome corresponds to homozygous regions, which then led me to estimate the actual haploid genome size: ~ 88 Mb.

Based on that same sliding window analysis, coverage of depth was calculated for individual gene models. Each gene model was assigned to a “heterozygous”, “homozygous” or “undetermined” category based on the distribution of coverage. Out of the 27,341 gene models from our annotation, 20,421 were “heterozygous”, 2,400 “homozygous”, and 4,520 in the “undetermined” category. Based on these categories, we estimated the haploid number of proteins in D. pachys to be approximately 17,000 (20,421/2 + 2,400 + 4,520 ≈ 17,130).

Collinearity analysis

To detect collinear blocks within the D. pachys genome, we used the tool MCScanX tool [63]. MCScanX identifies putative homologous chromosomal regions using genes as anchors. We allowed for a minimum of three genes required for a collinear block. 14,020 genes were involved in at least one collinear block (51%). In Figure 3A, the edges of the MCScanX ribbons that are pictured are defined by the positions of the first and last gene of the matching collinear segments.

Coding sequences of collinear proteins were aligned using ClustalW2 [103] and only alignments with a null E-value were included in the calculation of the average percentage identity (58% of the alignments). The percentage identity corresponds to the number of identical nucleotides divided by the total number of aligned positions. The 35 scaffolds represented on Figure 3A account for 4% of the total assembly.

Telomere repeats in 11 nematode genomes

Using a perl custom script, hexanucleotide and pentanucleotide motifs were counted as a repeat when found in a cluster of at least three consecutive motifs for hexanucleotide motifs and four for pentanucleotide motifs. The overall number of unique hexanucleotide and pentanucleotide motifs found in each genome is indicated on Figure 2 to demonstrate that the absence or near-absence of nematode telomeric repeats in D. pachys, M. hapla and T. spiralis, is not due to a poor resolution of repeat regions in the assemblies. Genome sequences used for this analysis are from WormBase release WS248 (ftp://ftp.wormbase.org/pub/wormbase/species/), except for R. culicivorax (http://www.nematodes.org/genomes/romanomermis_culicivorax/). See also Figure S5.

Microscopy

DAPI staining

For staining of early oocytes and embryos, well-fed nematodes were washed with M9 buffer and left from two hours to overnight rotating on a carrousel. Nematodes were then centrifuged and washed with M9 three times, before being concentrated in approximately 50 μl. The last wash would be performed with M9 with 0.1% Tween 20. 10 μl of the nematode suspension would then be pipetted onto a Superfrost slide (Fisher Scientific). An equal volume of 4% formaldehyde was added directly on the slide (final concentration is 2% formaldehyde) and allowed to incubate for 5 min. After covering the slide with a cover glass, excess liquid was removed with a Kimwipe on the edge of the cover slip until embryos were slightly flattened and displayed a clearer transparent aspect. Slides were then flash frozen in liquid nitrogen. After at least 10 min, slides were removed, and, using a razor blade, the cover slip was removed very quickly to crack open the nematodes cuticles and embryo eggshells. The sample was fixed with 100% methanol at −20°C for 10 min, transferred to P BS/0.1% Tween (PBST), washed three times for 10 min in PBS/1%Triton-X and stained 10 min in 0.5 μg/ml DAPI/PBST. Finally samples were de-stained in PBST for 1 h and mounted with Vectashield mounting medium (Vector Laboratories).

For gonad stainings, ten to twenty adults were transferred to a 10-μl drop of water on a Superfrost slide before being dissected with a 20G 11/2 hypodermic needle to expose the gonads. Slides were then fixed and stained as described above.

Pictures were acquired with a 100X lens on a Leica microscope using a Hamamatsu camera, or on a Zeiss AxioImager A1 microscope using a Zeiss AxioCam MRm CCD camera. Images were processed and overlaid using ImageJ [104].

DIC time-laps movies

Gravid females were placed on a coverslip in 5 to 10 μl of water. Either they were dissected to free early embryos, or paralyzed by adding 2 μl of 100 μm Levamisole. They were then transferred to a 2% agarose pad and the coverslip was sealed with melted petroleum jelly to reduce the evaporation of liquid. DIC time-lapse movies were generated by imaging every 30 seconds using Openlab or LAS X software and a Leica microscope with a 40X, 63X or 100X lens depending on the number of embryos.

RNA interference (RNAi)

To test RNAi sensitivity in the Protorhabditis group, an RNAi-expressing bacterial clone from the Ahringer RNAi library [105] was used that targets all five highly conserved C. elegans actin genes. Four species were tested by feeding (as previously described [105]) with E. coli HT115 expressing the RNAi compared to HT115 with an empty L4440 plasmid vector; two species showed RNAi sensitivity (D. pachys PF1309, D. sp. JU359) and two did not (Protorhabditis sp. 4 JB122, Prodontorhabditis wirthi DF5074). In the two Diploscapter strains, the phenotype was similar to that observed in C. elegans, although it took longer for Diploscapter to be affected. After 2–3 days, larvae began to look transparent and sick, subsequently dying without producing progeny. Living (embryos, larvae, adults) and dead nematodes were counted 14 days after 60–80 freshly hatched larvae were plated onto NGM plates with control or RNAi bacteria. Sensitivity to RNAi was quantified as the proportion of dead/total nematodes.

Recombination experiments

A D. pachys larva was singled out on an NGM medium plate. Since D. pachys nematodes can burrow into the agar, we used thin NGM agar layers—4ml instead of 8ml, to increase the number of progeny accessible on the surface of the plates. The female nematode was followed during its egg-laying stage—around two weeks—and transferred to a new plate regularly. The progeny, once adult and when retrievable from the agar surface, were picked individually into PCR tubes containing 11μl of 500 μg/ml Proteinase K (New England Biolabs) in 1X ThermoPol PCR buffer (New England Biolabs). PCR tubes were then frozen immediately at −80°C, and stored u ntil used. We designed PCR primers for eight heterozygous loci distributed across the large D. pachys chromosome (using macrosynteny information). We selected loci with alleles of different sizes, in order to be able to visualize both allelic copies on an electrophoresis gel. 78 PCR amplifications were performed on 30 stored progeny for one or more of the eight selected allelic regions with no loss of heterozygosity observed.

QUANTIFICATION AND STATISTICAL ANALYSIS

Details of the quantification and statistical analyses related to the genome sequencing, assembly, annotation and analysis are provided in the Results and Method Details sections, as well as in the Supplemental Information. Details of the phylogenetic and divergence analyses are provided below.

Phylogenetic analysis

We constructed a novel molecular phylogeny for the Protorhabditis-group species that are under study in this work. Seven nematode species from the Protorhabditis group were characterized with segments of small subunit 18S and large subunit 28S rDNA sequence and placed in molecular phylogenetic context with other nematodes from the rhabditid family. Figure 1 represents the Maximum Likelihood (ML) best SSU 18S and 28S phylogenetic tree for these species as implemented in RAxML [87] with 100 bootstrap runs. The alignment was also subjected to Bayesian inference (BI) analysis using the software MRBAYES [88]. This phylogeny confirms the paraphyly of the Protorhabditis genus. It also verifies that the group consists of two clades, which had been described previously [27]. One clade—the Xylocola group—is formed by some Protorhabditis species and the Diploscapter genus, which itself is monophyletic and derives from within these Protorhabditis species. Its sister clade—the Oxyuroides group—includes other Protorhabditis species and the Prodontorhabditis genus, also forming a monophyletic clade that derives from within the Protorhabditis genus. In addition, the phylogeny confirms the position of the Protorhabditis group as the sister clade of the Caenorhabditis clade.

Divergence comparisons

To estimate the age of the asexual Protorhabditis/Diploscapter clade, we compared species divergences within this clade to those in the sister Caenorhabditis clade, because good estimates have been obtained for dates of Caenorhabditis species divergences [16]. Orthologous sequences encoding part (1,863 nt) of the largest subunit of RNA polymerase II were used from four Caenorhabditis taxa (C. elegans, C. briggsae, C. japonica and C. drosophilae) and four Protorhabditis-group taxa (Diploscapter sp. 2 JU359, Protorhabditis sp. 4 JB122, Protorhabditis sp. DF5055 and Prodontorhabditis wirthi DF5074) [26]. This molecule was used for comparison because of the relative evenness of evolutionary rates among different lineages (i.e. lack of significant heterotachy; Kiontke et al. 2004). Divergences were calculated by summing branch lengths in phylogenetic trees, as discussed previously [106]. Phylogenetic trees were generated using maximum likelihood (as implemented in PAUP* 4.0b10) using a general time-reversible model for nucleotide changes, accounting for invariant sites and three categories of rate differences at the different codon positions (GTR+I+SS). Rate parameters and frequencies of nucleotides were estimated from the data. Relative rates at the different codon positions were: 0.383 at pos. 1 vs. 0.120 at pos. 2 vs. 2.483 at pos. 3. To determine the most recent date of the last common ancestor of the entire asexual clade, the divergence between two most divergent species in the asexual clade is used, i.e. that between any Diploscapter species (here D. sp. 2 JU359) and Protorhabditis sp. 4 JB122. This divergence is 0.268 ± 0.021 substitutions per site (Table S6). This is almost identical to the divergence between C. elegans and C. briggsae, which is 0.264 ± 0.018 substitutions per site (Table S6). Because the date of divergence for C. briggsae and C. elegans is estimated about 18 Myr ago [16], the origin of the asexual Protorhabditis/Diploscapter clade must have occurred around the same time, ca. 18 Myr ago.

DATA AND SOFTWARE AVAILABILITY

Genome submission

The genome assembly and annotation of D. pachys have been deposited as a Whole Genome Shotgun project at DDBJ/EMBL/GenBank under the accession LIAE00000000. The version described in this paper is version LIAE01000000.

Supplementary Material

1

Table S5. Collinear segments in D. pachys detected using MCScanX. Related to Figure 4. Output from MCScanX shows pairwise collinear blocks between pairs of D. pachys scaffolds. Each collinear block contains at least three pairs of similar proteins in the same order on both scaffolds. A score and an e-value are reported for each block (referred to as an Alignment in the output). Pairs of proteins that belong to the collinear block are listed with the e-value from the initial alignment form.

Table S6. Conservation of meiotic C. elegans genes in D. pachys and H. contortus. Related to Figure S6. Listed are the 91 C. elegans meiotic genes described in [67]. The homology column indicates the number of BLASTP hits (R, reciprocal; NR, non-reciprocal) between C. elegans and D. pachys/H. contortus. NONE indicates that no homology was detected. Meiosis specificity is assessed based on [67] and Wormbase [107] descriptions. A gene is considered meiosis-specific if only meiotic functions are known as 11 August 2017.

Highlights.

• Asexuality and a single chromosome evolved around the same time in Diploscapter

• The single chromosome resulted from fusions of 6 ancestral chromosomes

• Meiosis is truncated and many meiosis and telomere maintenance genes are missing

• These features have likely favored the evolution of significant heterozygosity