Evidence against a germ plasm in the milkweed bug Oncopeltus fasciatus, a hemimetabolous insect

In sexually reproducing animals, only germ cells contribute genetic information to future generations. The germ line/soma separation is a cell fate decision shared across Metazoa (Buss, 1987). Despite the fundamental commonality of germ cell function in animals, the molecular mechanisms underlying germ cell specification are remarkably diverse across different taxa (Extavour and Akam, 2003; Extavour, 2007; Ewen-Campen et al., 2010; Juliano et al., 2010).

Primordial germ cells (PGCs) can be specified via different developmental mechanisms; here we call these “cytoplasmic inheritance” and “zygotic induction.” (We and others have previously referred to these mechanisms as “preformation” and “epigenesis” respectively (Nieuwkoop and Sutasurya, 1981; Extavour and Akam, 2003; Extavour, 2007). However, these terms can hold different meanings in other contexts of the history and philosophy of biology (e.g. Callebaut, 2008). We therefore avoid them here in favour of more mechanistically descriptive terms.) Cytoplasmic inheritance is characterised by the asymmetric formation of a specialised cytoplasmic region within the oocyte or early embryo, termed “germ plasm.” Germ plasm contains maternally provided mRNAs and proteins that are individually necessary and collectively sufficient for PGC formation. Cells that inherit germ plasm during embryogenesis acquire germ line fate. The best understood example of cytoplasmic inheritance occurs in Drosophila melanogaster, where germ plasm is maternally synthesised, localised to the posterior of the oocyte during oogenesis, and subsequently incorporated into PGCs (pole cells) during cellularisation. Removing pole cells after their formation, or compromising the molecular components of germ plasm, leads to loss of PGCs and sterility in adulthood (reviewed by Mahowald, 2001). In contrast, zygotic induction of PGCs takes place later in development and requires signalling from neighbouring somatic cells to induce germ line fate. This mode of PGC development is exemplified by Mus musculus, wherein PGCs develop from a subset of presumptive mesodermal cells after the segregation of embryonic and extraembryonic tissues in response to local signalling (reviewed by Magnúsdóttir et al., 2012).

Across Insecta, germ plasm has been almost exclusively reported in taxa nested within Holometabola (“higher” insects, which undergo complete metamorphosis) including D. melanogaster (reviewed by Kumé and Dan, 1968; Anderson, 1973; Nieuwkoop and Sutasurya, 1981), and in only three species belonging to the sister assemblage to the Holometabola (see below). Thus, although the vast majority of our knowledge of insect germ cell development comes from studies of germ plasm in D. melanogaster, this mode of germ cell specification is likely a derived feature of Holometabolous insects and their close sister taxa.

Our present knowledge of PGC specification in basally branching insects (Hemimetabola) is based almost entirely on classical histological studies of insect development conducted over the past 150 years. Nearly all of these report that PGCs arise late in embryogenesis, raising the possibility that they may be specified through inductive mechanisms (Wheeler, 1893; Heymons, 1895; Hegner, 1914; Nelsen, 1934; Roonwal, 1937). Experimental approaches to discovering germ plasm in Hemimetabola are limited, but a study involving destruction of the germ rudiment via irradiation in the cricket Gryllus domesticus (Schwalm, 1965) showed that no specific region of early embryos in this species contains a germ line determinant. Functional tests of genes that may specify germ cells in Hemimetabola have been performed in only one insect, the cricket Gryllus bimaculatus. In this cricket, the conserved germ line markers vasa and piwi are dispensable maternally and zygotically for PGC formation (Ewen-Campen et al., 2013). Most evidence available for the Hemimetabola therefore suggests the absence of germ plasm and the operation of zygotic PGC specification mechanisms.

Exceptions have been reported, however, in some members of the Paraneoptera, an assemblage of insect orders (including Hemiptera [true bugs], Psocoptera [book lice], and Thysanoptera [thrips]) that collectively form the sister group to Holometabola (Yeates et al., 2012). Cytological studies of three paraneopteran species, a book louse (Psocoptera (Goss, 1952)), a thrip (Thysanoptera (Heming, 1979)) and an aphid (Hemiptera (Chang et al., 2009)) suggested the presence of germ plasm in oocytes or early embryos, as did expression studies of vasa, piwi and nanos expression during asexual development of the pea aphid Acyrthosiphon pisum (Chang et al., 2006; Chang et al., 2007; Chang et al., 2009; Lu et al., 2011). However, A. pisum embryogenesis is highly modified relative to that of other hemimetabolous insects and even relative to other members of the same order (Miura et al., 2003). Studies of embryogenesis in most other hemipterans describe absence of germ plasm and PGC origin after cellularisation from the blastopore region at gastrulation stages (Metschnikoff, 1866; Witlaczil, 1884; Will, 1888; Seidel, 1924; Mellanby, 1935; Butt, 1949; Sander, 1956; Kelly and Huebner, 1989; Heming and Huebner, 1994). We therefore wished to examine the expression and function of germ line genes in a hemipteran displaying embryological characteristics more representative of the order.

Here we characterise germ cell formation and migration in the milkweed bug Oncopeltus fasciatus (Hemiptera). We examine the expression of 19 molecular markers including vasa, nanos, and piwi, and test the germ cell function of three of these using RNA interference. We show that in striking contrast to Drosophila, transcripts of none of these genes localise asymmetrically within Oncopeltus oocytes or early embryos. We identify PGCs using multiple criteria, and show that neither vasa nor tudor are required for PGC specification or oogenesis in this species, but that vasa is required for spermatogenesis in adult males. These data show that the PGC specification role of vasa has diverged between Oncopeltus and the Holometabola, and suggest that Oncopeltus PGCs may form in the absence of maternally supplied germ plasm.

In contrast to D. melanogaster, classical studies of Oncopeltus fasciatus embryogenesis have not revealed a germ plasm in oocytes or early embryos, and instead first identify cells with cytological characteristics of PGCs at the posterior of the embryo at the end of the cellular blastoderm stage (Butt, 1949). We used semi-thin plastic sectioning and fluorescence microscopy to confirm these observations, and traced the development of these putative PGCs throughout gastrulation and germ band elongation (supplementary material Fig. S1). Our observations of putative PGC formation in Oncopeltus were consistent with historical studies (Butt, 1949), showing that these cells first arise at the blastoderm posterior immediately prior to gastrulation (supplementary material Fig. S1). Unlike pole cells in D. melanogaster, presumptive PGCs in Oncopeltus arise on the basal side of the blastoderm surface, adjacent to the yolk (supplementary material Fig. S1G–H). In order to obtain further evidence that these cells were PGCs and test for the presence of a maternally supplied germ plasm, we examined the expression of conserved germ line markers.

We cloned fragments of vasa, nanos, and piwi (Ewen-Campen et al., 2010) and confirmed that each was the best reciprocal BLAST hit to its respective orthologue in D. melanogaster. vasa was cloned using degenerate primers (supplementary material Table S1). nanos and a single piwi gene were recovered from the Oncopeltus transcriptome, in addition to a single AGO-3 orthologue (an additional PIWI family protein belonging to a separate sub-family; not shown). We believe it is unlikely that Oncopeltus possesses additional orthologues of these genes because (1) the Oncopeltus ovarian and embryonic transcriptome, which is nearly saturated for gene discovery and has an average coverage of 23× (Ewen-Campen et al., 2011), contained only one orthologue of each gene; and (2) degenerate PCR for vasa using primers flanking the conserved DEAD box helicase domain (Rocak and Linder, 2004) recovered only a single vasa orthologue.

Phylogenetic reconstruction confirmed that Oncopeltus vasa is nested within other insect vasa genes (supplementary material Fig. S2A), and that Oncopeltus piwi belongs to the PIWI sub-family containing the Drosophila genes piwi and aubergine (which are Drosophila-specific duplications) (supplementary material Fig. S2B). The portion of animal Nanos proteins with conservation sufficient for confident alignment (48 amino acids) is too short to yield significant phylogenetic signal (supplementary material Fig. S2C, note low support values), but Oncopeltus Nanos does contain the diagnostic 2×(CCHC) zinc finger domain found in all Nanos orthologs (supplementary material Fig. S2D).

Our analysis of the Oncopeltus nanos sequence produced an unexpected result: we found that a stop codon is present 771 bp upstream of the first CCHC zinc finger domain, although no methionine is found anywhere in this region. This is unlikely to be a sequencing error, as it was identified with high coverage (22 reads/bp at this position) in the transcriptome (Ewen-Campen et al., 2011) and confirmed using Sanger sequencing of independent clones generated from a different cDNA pool than that used to generate the transcriptome. Furthermore, repeated attempts at 5′ RACE using a third independent DNA pool failed to amplify a start codon. Several lines of evidence confirm that this Oncopeltus nanos sequence represents a highly expressed mRNA and is therefore unlikely to be a pseudogene: it was recovered from a transcriptome made solely from poly(A)-RNA, and is detected via both RT-PCR (Ewen-Campen et al., 2011) and in situ hybridisation (see below). We hypothesise that a large, unspliced intron downstream of the start codon may have been present in our mRNA preparations. Alternatively, given that the length of the predicted translated region upstream of the first CCHC zinc finger domain (266 amino acids) is within the range of known arthropod Nanos orthologues (95 to 332 amino acids) (Wang and Lehmann, 1991; Curtis et al., 1995; Calvo et al., 2005; Lynch and Desplan, 2010), it may be that the Oncopeltus Nanos N terminus has a non-methionine start codon. Although rare, eukaryotic non-AUG translation initiation can occur in nuclear-encoded genes, including developmentally relevant genes (Hellen and Sarnow, 2001), and can be recognized by insect ribosomes (Sasaki and Nakashima, 2000; Jan et al., 2001). In the absence of a complete genome sequence we cannot distinguish between these hypotheses. Despite this uncertainty, we report nanos transcript expression here for the sake of completeness.

The distinct cytoplasm inherited by early-specified PGCs in multiple organisms, including D. melanogaster, Caenorhabditis elegans, Xenopus laevis and Danio rerio, contains transcripts of the highly conserved piwi, vasa and nanos gene families. Of these, only nanos mRNA is asymmetrically localised to D. melanogaster germ plasm, while piwi and vasa transcripts are ubiquitous throughout the fly oocyte and embryo. However, in several other organisms vasa orthologue transcripts are asymmetrically localised germ plasm components (reviewed by Ewen-Campen et al., 2010).

To test whether any of these transcripts were asymmetrically localised to putative germ plasm in Oncopeltus oocytes, we conducted in situ hybridisation on adult ovaries. The structure of Oncopeltus ovaries is typical of Hemiptera and several other insect orders but differs remarkably from that of Drosophila (Fig. 1A) (Büning, 1994). Rather than each oocyte developing together with its own complement of 15 nurse cells as in Drosophila, all oocytes in Oncopeltus ovarioles share a common pool of syncytial nurse cells located at the anterior of each ovariole in a region termed the “tropharium” (Fig. 1A1). The nurse cell syncytium connects to all oocytes via elongated, microtubule-rich tubes called “nutritive tubes” (Hyams and Stebbings, 1979; Harrison et al., 1991) through which maternal factors, including mRNA, proteins and mitochondria, are transported to developing oocytes (Fig. 1A2,A3) (Stebbings et al., 1985; Stebbings and Hunt, 1987; Anastas et al., 1991; Hurst et al., 1999; Stephen et al., 1999).

vasa, nanos, and piwi were expressed at high levels in Oncopeltus nurse cells and oocytes of all stages, but at no stage of oogenesis did any of these three transcripts localise asymmetrically within oocytes (Fig. 1B–D). Expression was detected in nurse cells, resting oocytes, nutritive tubes, and developing oocytes, suggesting that these transcripts are synthesised in the nurse cells and subsequently transported to oocytes via nutritive tubes (Fig. 1B–D). nanos and piwi were expressed throughout the tropharium (Fig. 1C,D), in contrast to vasa, whose expression was primarily in nurse cells of the posterior tropharium, resting and developing oocytes (Fig. 1B). In late stage oocytes, expression remained ubiquitous (not shown), similar to the expression in just-laid eggs (see below).

The expression of piwi, vasa and nanos suggests that a maternally localised germ plasm containing transcripts of these genes is not present in Oncopeltus oocytes. However, a functional germ plasm that contains gene products other than those encoded by these three genes could be present in oocytes or early embryos. To explore this possibility, we examined the expression of 14 additional genes whose transcripts are enriched in the germ plasm and germ cells of Drosophila (supplementary material Table S2) (Tomancak et al., 2002; Lécuyer et al., 2007; Tomancak et al., 2007) that were also recovered from the Oncopeltus ovarian and embryonic transcriptome (Ewen-Campen et al., 2011) based on best reciprocal BLAST hit analysis with the Drosophila proteome (Zeng and Extavour, 2012). Although several of these genes do not have documented mutant phenotypes for germ cell formation in Drosophila (supplementary material Table S2), all are expressed at high levels in germ plasm and/or pole cells and are therefore molecular markers for germ plasm in Drosophila. We reasoned that if Oncopeltus possessed germ plasm it would likely be revealed by at least one of these genes.

In addition, we examined the expression of boule and tudor, which have widely conserved functions in germ cells across Metazoa (Eberhart et al., 1996; Ewen-Campen et al., 2010; Shah et al., 2010). tudor is one of 23 Tudor domain-containing proteins in Drosophila (Ying and Chen, 2012), but there is no evidence that loss of function of other Tudor domain-containing genes have grandchildless phenotypes in Drosophila (Handler et al., 2011; Pek et al., 2012). We therefore focus only on the expression and function of the orthologue of Drosophila tudor (CG9450). We examined boule and tudor transcript expression throughout oogenesis and embryogenesis through mid-germ band stages.

None of these 16 transcripts localised asymmetrically in ovaries (Fig. 1E,F; supplementary material Fig. S3). Instead, like vasa, piwi and nanos (Fig. 1B–D), all of these genes were expressed ubiquitously throughout oogenesis. Half of the genes examined (sra, CycB, Bsg25D, Uev1A, CG16817, Unr, mael and tud) were expressed, like vasa (Fig. 1B), in nurse cells adjacent to resting oocytes, as well as in the resting and early oocytes themselves (Fig. 2E; supplementary material Fig. S3B–H). Five genes (Gap1, eIF5, bel, orb and bol) were, like piwi and nanos (Fig. 1C,D), strongly expressed in all nurse cells of the tropharium (Fig. 2F; supplementary material Fig. S3I–L). Two genes (cta and Tao) were expressed in resting and early oocytes but barely at all in the tropharium (supplementary material Fig. S3M,N), suggesting that these genes may be transcribed by resting oocyte nuclei rather than by nurse cells. Finally, aret (aka bruno), which is a translational regulator of Oskar in Drosophila (Kim-Ha et al., 1995; Webster et al., 1997), was expressed in nurse cells of the posterior tropharium and in early stages of oogenesis but excluded from resting oocytes (supplementary material Fig. S3O), suggesting that it is transcribed by oocyte nuclei after the onset of oogenesis. In summary, although transcripts of most of these genes are likely to be supplied maternally to oocytes, they are not asymmetrically localised within oocytes of any stage.

Although none of the genes examined showed asymmetric localisation during oogenesis or early embryogenesis, at late blastoderm stages many of the genes appeared enriched at the posterior pit, where PGCs had previously been identified based on cytological criteria (supplementary material Fig. S1) (Butt, 1949). However, because at this stage of development gastrulation begins at the posterior, this region of the blastoderm is multilayered. Upon close examination, we found that the apparent transcript enrichment was an artifact of tissue thickness for all genes except vasa, tudor and boule, whose transcripts appeared truly enriched in putative PGCs at late blastoderm/early gastrulation stages (Fig. 2D–H″, Fig. 4N,S).

Strikingly, we found that vasa, tudor and boule marked PGCs from the time of their formation at cellular blastoderm stages, but that none of these genes' transcripts were asymmetrically localised prior to PGC formation. Immediately after egg laying, vasa transcripts were not localised asymmetrically but rather were ubiquitously distributed throughout the embryo (Fig. 2A). As energid nuclei reached the embryonic surface (Fig. 2B), cytoplasmic islands enriched with these transcripts were distributed evenly across the embryonic surface, remaining there as these energids divided to form the uniform blastoderm (Fig. 2C). Prior to posterior pit formation, vasa expression became restricted to putative PGCs at the embryonic posterior (Fig. 2D).

To visualise vasa expression in the developing PGCs in greater detail, we collected staged embryos in two-hour intervals over the period during which PGCs arise (19 to 27 hours after egg laying (AEL)), performed in situ hybridisation for vasa (Fig. 2E–H), and sectioned the embryos in plastic resin (Fig. 2E″–H″). During this eight-hour period, the blastoderm nuclei undergo two concurrent, dynamic processes: continuing cell divisions increase the nuclear density throughout the blastoderm, and the blastoderm nuclei move towards the posterior pole and ultimately into the yolk (Butt, 1949; Liu and Kaufman, 2004) (Fig. 2E′–H′). From 19–21 hours AEL, the ubiquitous vasa expression seen in early embryos remained unchanged (Fig. 2E–E″). However, from 21–23 hours AEL vasa expression became enriched in a subset of cells at the blastoderm posterior (Fig. 2F–F″). From 23–25 hours AEL, vasa-positive cells increased in density at the blastoderm posterior and began to move into the yolk (Fig. 2G–G″). This movement appeared passive, due to the formation of the posterior pit by invagination of the germ rudiment. However, in the absence of time-lapse data we cannot rule out the possibility of active PGC movement out of the blastoderm epithelium and towards the yolk. From 25–27 hours AEL, as the germ rudiment began its invagination into the yolk, vasa-positive cells formed a distinct mesenchymal clump within the yolk at the posterior of the embryo (Fig. 2H–H″). During this and all following stages, in addition to the marked enrichment in PGCs, vasa transcripts were additionally observed ubiquitously at low levels throughout somatic tissue (Figs 2, 3).

Throughout all subsequent stages of germ band elongation and patterning, vasa continued to mark PGCs (Fig. 3). During early stages of germ band elongation prior to limb bud formation (∼28–32 hours AEL) vasa-positive PGCs remained at the embryonic posterior on the dorsal surface of the newly forming mesoderm (Fig. 3A,B,B′). The PGC cluster then became pear-shaped from 32–42 hours AEL, as the anteriormost PGCs began to move towards the anterior of the embryo (Fig. 3A,C,C′). As the head lobes enlarged (36–40 h AEL), PGCs began to migrate anteriorly on the dorsal surface of the embryo and continued their migration during limb bud stages (40–44 h AEL) (Fig. 3A). During appendage elongation stages (44–48 h AEL) PGCs split into distinct clusters spanning the midline in abdominal segments A4–A6, one cluster per segment. As appendage segmentation became morphologically distinct (48–52 h AEL), the segmental clusters split along the ventral midline into bilateral clusters in A4–A6.

tudor and boule were also expressed in PGCs at all stages in a pattern indistinguishable from that of vasa (Fig. 4), providing further evidence that the vasa-positive cells are Oncopeltus PGCs. None of the other genes we examined (supplementary material Fig. S3), including nanos and piwi (Fig. 4), were enriched in PGCs at any stage of embryogenesis.

Our gene expression analysis demonstrates that vasa, tudor, and boule are specifically expressed in PGCs beginning at the putative time of their specification at the embryonic posterior just prior to gastrulation. To determine whether these genes were required for PGC formation or development in Oncopeltus, we performed maternal RNAi (mRNAi) for each gene. We confirmed that mRNAi effectively reduced zygotic transcript levels in our experiments using RT-PCR (Fig. 5E). PGC presence or absence was determined with in situ hybridisation against PGC markers at ∼40–54 hours AEL, when germ cells are visible on the dorsal mesoderm.

RNAi knockdown of vasa or tudor did not disrupt embryonic patterning or germ band development (supplementary material Table S3), despite the widespread expression of these genes at early blastoderm stages (Figs 2, 4), and their persistent low levels of expression in somatic cells even after PGC formation (Figs 3, 4). Strikingly, germ cells were clearly present in both vasa (93.8%, n = 16) and tudor (100%, n = 20) knockdowns, suggesting that neither of these genes is required for PGC specification (Fig. 5A–C′). It is formally possible that residual vasa or tudor transcripts that may have escaped destruction by mRNAi could be sufficient to play an instructive role in PGC formation. However, we note that transcript levels of both genes in the progeny of injected mothers were barely detectable in the case of vasa, and undetectable in the case of tudor, when assessed with RT-PCR even as late as 4 days AEL (Fig. 5E). Moreover, even hypomorphic alleles of tudor (Schüpbach and Wieschaus, 1986) and vasa (Lasko and Ashburner, 1990; Schüpbach and Wieschaus, 1991; Liang et al., 1994) lead to loss of PGCs in Drosophila. We therefore hypothesise that in Oncopeltus, vasa and tudor are required neither maternally nor zygotically for germ cell specification, although they are expressed in the cells specified as PGCs.

To address the possibility of redundancy between these two genes, we performed double knockdown of vasa and tudor, which reduced transcripts of both genes to undetectable levels (Fig. 5E). Eggs laid by vasa + tudor double RNAi females had an increased rate of embryonic lethality relative to controls (supplementary material Table S3; 47.4%, n = 19 vs 10.3%, n = 39), which may mean that these genes work together to play roles in somatic development. However, embryos that escaped this lethality still had PGCs (100%, n = 10) (Fig. 5D,D′).

None of the knockdowns caused any qualitative or quantitative change in egg laying by injected females compared to controls, and ovaries of injected females showed neither morphological abnormalities nor signs of disrupted oogenesis (not shown). This indicates that, in contrast to Drosophila (Schüpbach and Wieschaus, 1991; Styhler et al., 1998; Tomancak et al., 1998; Johnstone and Lasko, 2004), vasa is not required individually or redundantly with tudor for Oncopeltus oogenesis or egg laying.

boule mRNAi caused a complete cessation of egg laying by injected females after four to five clutches (one clutch is laid every one to two days). In contrast, vasa, tudor and control mRNAi females continued to lay up to 12 clutches. Ovaries of boule dsRNA-injected females possessed only a few oocytes at early stages of oogenesis, and few or no mature oocytes (not shown), indicating a requirement for boule in the progression of oogenesis. Eggs laid by boule RNAi females displayed nearly complete embryonic lethality (81.8%, n = 22) in all but the first clutch laid. (The first clutch of Oncopeltus eggs laid following mRNAi typically displays no abnormalities, as these eggs have developed their chorion by the time of injection and are therefore impervious to dsRNA (Liu and Kaufman, 2004).) This was a striking increase in embryonic lethality compared to DsRed controls (26.8%, n = 190), vasa knockdowns (23.2%, n = 198) and tudor knockdowns (5.6%, n = 54). The oogenesis requirement for boule and resulting embryonic lethality thus prevented us from determining whether boule is required for germ cell specification in Oncopeltus, and we do not further report on the role of boule on oogenesis in the present study.

Given that in contrast to Drosophila, vasa is not required for germ cell specification or oogenesis in Oncopeltus, we wished to test for other possible functions of this gene. In mice, despite its expression in the embryonic PGCs of both sexes once they reach the genital ridge (Fujiwara et al., 1994; Diez-Roux et al., 2011). vasa is required not for PGC specification, but rather for gametogenesis in males (Tanaka et al., 2000). Similarly, we recently showed that vasa plays a role in spermatogenesis in the cricket G. bimaculatus (Ewen-Campen et al., 2013). We therefore asked whether vasa also functions during spermatogenesis in Oncopeltus.

The testes of Oncopeltus show an organisation typical of insect testes (Dumser, 1980), with stages of spermatogenesis located in an anterior–posterior progression (supplementary material Fig. S4). Unlike Drosophila, which has a single sperm tubule (testiole) per testis (Hardy et al., 1979), each Oncopeltus testis comprises seven testioles (Bonhag and Wick, 1953). In situ hybridisation for vasa showed that it is strongly expressed in secondary spermatogonia of each testiole, and at lower levels in early primary spermatocytes and post-spermatocyte stages, but not in primary spermatogonia or somatic cells (Fig. 6A). Adult males injected with dsRNA against vasa displayed multiple abnormalities in spermatogenesis. Testioles of vasa RNAi males lacked clearly defined cysts and contained large numbers of small, dense nuclei in the anterior region (Fig. 6H,I,I′), which in controls contained only spermatocytes with large, pale nuclei (Fig. 6C,D,D′; supplementary material Fig. S4E). The primary spermatogonial region of vasa RNAi testioles contained cysts of irregular size (Fig. 6I, arrowheads) with poorly defined cytoplasmic bridges (Fig. 6I′, arrows). In the spermatocyte region vasa RNAi testioles contained large, poorly defined clusters of several hundred cells (Fig. 6J, arrowheads) at varying stages of spermatogenesis (Fig. 6J′). The nuclear morphology of cells in these cysts corresponded to spermatocyte (Fig. 6E; supplementary material Fig. S4E) or early spermatid (Fig. 6F) stages, as well as shell stage-like nuclei (Fig. 6K,K′) typical of the mid-stage spermatids of controls (Fig. 6F′). Cysts of wild type shell stage spermatids are no longer syncytial as the actin-rich cytoplasmic bridges disappear during spermatocyte stages (supplementary material Fig. S4G). In contrast, the anterior shell stage-like nuclei in vasa RNAi testioles remained connected by cytoplasmic bridges (Fig. 6K′, red arrows), consistent with precocious spermatid differentiation. Moreover, although they displayed clear shell stage nuclear morphology (Fig. 6K,K′, red arrowheads), they were larger than wild type shell stage nuclei (Fig. 6F′, red arrowheads), suggesting that they had begun spermatid differentiation as syncytial diploid cells without first proceeding through meiosis as in wild type. Finally, the posteriormost region of vasa RNAi testioles contained irregular groups of cells at mixed stages of late spermatid and spermatozoon differentiation (Fig. 6L), rather than the perfectly synchronised cysts of late spermiogenic stages seen in controls (Fig. 6G). These defects were observed in testes examined 28–29 days following injection of adult males, but are not artefacts of age, as testes of 10 week old wild type adult males showed normal progression through all stages of spermatogenesis (supplementary material Fig. S4B).

Taken together, these data suggest that vasa is required for the maintenance of synchrony within cysts at multiple stages of spermatogenesis. In addition, vasa may be required for secondary spermatogonia to enter correct meiotic progression as spermatocytes, in the absence of which germ cells are nonetheless able to continue with subsequent stages of spermatogenesis.

In several cases, analyses of molecular markers such as vasa mRNA have revealed the presence of a cryptic germ plasm that had eluded prior histological studies (Yoon et al., 1997; Tsunekawa et al., 2000; Wu et al., 2011). In Oncopeltus, we have shown that none of the transcripts of an extensive suite of conserved germ cell markers localise asymmetrically within oocytes or in early embryos (Figs 1, 2, 4; supplementary material Fig. S3), including transcripts of genes that localise to and are required for the function of germ plasm in Drosophila. Gene products of at least one of these conserved germ line markers have been found in the germ plasm of every species where a germ plasm is known to exist (Ewen-Campen et al., 2010), although we note the important caveat that in Drosophila, several of these genes (vasa, piwi, and tudor) are localized as proteins rather than mRNAs. Thus, the lack of localisation of transcripts of any of these 19 genes during oogenesis or early embryogenesis suggests that Oncopeltus lacks germ plasm. Instead, our data support the hypothesis that Oncopeltus germ cells form in the absence of germ plasm, and are not present prior to the onset of posterior invagination at the end of the cellular blastoderm stage. We cannot, however, rule out the possibility that untested molecular markers, including protein products of the genes examined here, could be asymmetrically localised to a putative germ plasm in Oncopeltus.

While we provide multiple markers of PGCs, further experiments could be useful to confirm the identity of these cells. However, demonstration that these cells are functional PGCs via ablation experiments is complicated by the fact that they arise at the inner face of the blastoderm at the gastrulation center, so that their physical disruption would likely also compromise mesoderm formation and subsequent embryogenesis. Moreover, we note that while pole cell removal experiments in Drosophila result in sterility, pole cell removal in another insect with germ plasm, the wasp Pimpla turionellae, yields fertile adults despite the fact that these pole cells are bona fide PGGs in wild type embryos (Bronskill, 1959; Achtelig and Krause, 1971; Fleischmann, 1975). Further, we are currently unable to genetically ablate these cells and determine their effect on fertility, as our vasa, tudor and vasa + tudor RNAi double RNAi experiments do not disrupt their formation (Fig. 5). Lineage tracing techniques that would permit tracking of the putative PGCs over the six-week period between PGC formation and sexual maturity are not currently available for Oncopeltus. These caveats notwithstanding, the molecular and morphological evidence that the cells we identify in this report are bona fide Oncopeltus PGCs is comparable to that available for PGC identification in most studied animal species: (1) three conserved germ line genes, vasa, tudor, and boule, are specific germ cell markers in Oncopeltus (Figs 1–4); (2) transcripts of these genes first become enriched in germ cells specifically at the time that these cells were previously reported to arise based on morphological and cytological criteria (Figs 2–4) (Butt, 1949); and (3) cells with these molecular markers undergo migration and primordial gonad occupation (supplementary material Fig. S1; Figs 3, 4) consistent with the well-documented behavior of PGCs in many other hemipterans (Seidel, 1924; Mellanby, 1935; Butt, 1949; Sander, 1956; Kelly and Huebner, 1989; Heming and Huebner, 1994).

Although the posterior location of germ cells at the time of their specification is superficially similar to that of pole cells in Drosophila and other Diptera, PGC specification and development in Oncopeltus differs in several important ways. First, while Drosophila pole cells form on the exterior of the posterior syncytial blastoderm before somatic cellularisation, Oncopeltus germ cells appear on the yolk side of the cellular blastoderm. Second, while Drosophila pole cells are the first cells in the embryo to cellularise (Huettner, 1923), Oncopeltus germ cells arise after blastoderm cellularisation is complete (Butt, 1949). Third, because Oncopeltus is an intermediate-germ insect, only the gnathal and thoracic segments have been specified at the time that germ cells arise (Liu and Kaufman, 2004), whereas in the long-germ insect Drosophila, pole cells form posterior to the abdominal embryonic segments. Lastly, Oncopeltus germ cells form on the dorsal surface of the embryo, and remain on the yolk-facing surface of the mesoderm during their migration to the gonad primordium in anterior abdominal segments (Fig. 3). As a result, they do not undergo a transepithelial migration through the hindgut epithelium as in Drosophila (reviewed by Richardson and Lehmann, 2010).

Our functional analysis led to the surprising discovery that neither vasa nor tudor play instructive roles in germ cell specification in Oncopeltus. Both of these genes are required for germ cell specification in Drosophila (Boswell and Mahowald, 1985; Schüpbach and Wieschaus, 1986) and other species (Sunanaga et al., 2007; Spike et al., 2008). However, vasa has widely divergent roles across Metazoa (reviewed by Yajima and Wessel, 2011), and in many cases is dispensable for PGC specification (Tanaka et al., 2000; Braat et al., 2001; Li et al., 2009; Özhan-Kizil et al., 2009). In several organisms it plays a role in adult gametogenesis (Tanaka et al., 2000; Ohashi et al., 2007; Salinas et al., 2007; Fabioux et al., 2009; Salinas et al., 2012; Ewen-Campen et al., 2013).

Intriguingly, we find that similar to the mouse and the cricket, vasa is required for spermatogenesis in adult Oncopeltus (Fig. 6), but not for oogenesis. This sex-specific function may relate to a putative role in stem cell function. As in other hemimetabolous insects (Büning, 1994), in Oncopeltus germ line stem cells are likely active in the apex of the testes (Schmidt and Dorn, 2004) but are not thought to be present in adult ovaries. One caveat to this hypothesis is that vasa transcript was not detected by in situ hybridisation in the primary spermatogonia (Fig. 6A), although it may be present at very low levels in those stem cells. Alternatively, given its strong expression in secondary spermatocytes and the defects in cyst integrity and synchrony caused by vasa RNAi (Fig. 6), Oncopeltus vasa may play a male-specific role in the onset or synchrony of meiosis. Consistent with a conserved role for vasa in bilaterian meiosis, male germ cells in vasa knockout mice arrest just prior to meiosis onset (Tanaka et al., 2000), and in human stem cell-derived germ cells vasa overexpression enhances meiotic progression (Medrano et al., 2012). Oncopeltus vasa RNAi leads to premature spermatid differentiation by some diploid secondary spermatocytes within a cyst, resulting in cyst asynchrony. In Drosophila, mutations are known that disrupt meiosis but do not prevent sperm formation (Davis, 1971), consistent with the hypothesis that spermiogenesis can be decoupled from meiotic status.

Together with recent molecular and classical histological data on germ cell specification in other insects, our results are consistent with the hypothesis that germ plasm is a derived mode of germ cell specification that arose in the ancestor to holometabolous insects (Fig. 7) (Lynch et al., 2011; Ewen-Campen et al., 2012). The only other functional genetic analysis of germ line specification in a hemimetabolous insect to date (Ewen-Campen et al., 2013) has also provided evidence that maternally supplied posterior germ plasm is absent, and that vasa is dispensable maternally and zygotically for germ cell formation. Our data thus support the notion that germ plasm-driven germ cell specification mechanisms operative in Drosophila melanogaster and Nasonia vitripennis are derived relative to the Hemimetabola (Fig. 7).

The ubiquitous distribution of germ cell markers in early Oncopeltus embryos and their subsequent enrichment in presumptive germ cells at the blastoderm posterior is reminiscent of vasa expression in the beetle Tribolium (Fig. 7) (Schröder, 2006; C. von Levetzow, Konservierte und divergente Aspekte der twist-, snail- und concertina-Funktion im Käfer Tribolium castaneum, PhD thesis, Universität zu Köln, 2008). Further taxonomic sampling, and functional studies in Tribolium, will be needed to determine whether the PGC specification mechanisms in these two species may be the result of common ancestry (Fig. 7).

A large number of transcripts that localise to germ plasm in Drosophila are expressed ubiquitously in Oncopeltus oocytes and early embryos. This suggests that the evolution of germ plasm in Holometabolous insects involved a large-scale change in the localisation of many transcripts in the oocyte. We propose that this likely resulted from a change in the localisation of an upstream component capable of recruiting many downstream transcripts, rather than via the sequential evolution of distinct localisation mechanisms for individual transcripts. Studies on the genetic mechanism of evolutionary redeployments of multiple downstream genes have largely focused on transcription factors, as individual transcription factors are capable of regulating large numbers of target genes (Hoekstra and Coyne, 2007; Moczek, 2008; Stern and Orgogozo, 2008; Craig, 2009). Interestingly, in the case of germ plasm, transcription factors are unlikely to have been key players in the mechanisms of evolutionary change for a number of reasons. First, regulation of germ line determinants is largely post-transcriptional (Arkov and Ramos, 2010; Richter and Lasko, 2011; Sengupta and Boag, 2012; Nousch and Eckmann, 2013). Second, germ plasm transcript function relies on their subcellular localisation (often mediated via signals in their 3′UTRs) rather than their presence or absence (Rangan et al., 2009). Finally, unlike key transcription factors identified as largely sufficient to induce specific somatic cell fates (e.g. Akam, 1998; Kozmik, 2005; Baena-Lopez and García-Bellido, 2006), there is no single conserved gene that is sufficient to confer germ cell fate across metazoans. The evolution of germ plasm may therefore serve as an example of how a novelty (asymmetrically localized germ plasm in the oocyte) arose via changes in RNA localisation rather than transcriptional regulation.

Oncopeltus fasciatus were cultured at 28°C as previously described (Ewen-Campen et al., 2011). Timing of embryonic events reported here may differ from that reported in other studies using lower rearing temperatures (e.g. Liu and Kaufman, 2004).

Total RNA was extracted from mixed-stage embryos and ovaries using TRIzol (Invitrogen) and used for first strand cDNA synthesis with qScript cDNA SuperMix (Quanta BioSciences). An Oncopeltus vasa fragment was cloned using degenerate primers (supplementary material Table S1). nanos and piwi fragments were obtained from the Oncopeltus transcriptome (Ewen-Campen et al., 2011). Fragments were extended using RACE PCR (SMART RACE cDNA kit, Clontech), and used as templates for DIG-labeled in situ probes and dsRNA fragments following sequence verification (supplementary material Table S1). Genes for the in situ hybridisation screen (supplementary material Tables S1, S2) were obtained from the Oncopeltus transcriptome (Ewen-Campen et al., 2011; Zeng and Extavour, 2012) and amplified using primers containing linker sequence (5′-CCCGGGGC-3′) enabling direct addition of a T7 site to the 3′ end in a subsequent PCR reaction. Extended sequences are available from ASGARD (http://asgard.rc.fas.harvard.edu) (Zeng and Extavour, 2012). All coding sequences reported in this study have been submitted to GenBank [accession numbers KC261571–KC261587] except for orb and Uev1A, for which we obtained only 3′ UTR sequence.

Maximum-likelihood phylogenetic analysis was performed for vasa, piwi, and nanos as previously described (Ewen-Campen et al., 2012).

Embryos were fixed and stained as previously described (Liu and Kaufman, 2004; Erezyilmaz et al., 2009; Kainz et al., 2011). Adult gonads were dissected in 1× PBS and fixed in 4% formaldehyde in 1× PBS for at least one hour. Antibodies used were mouse anti-alpha tubulin DM1A (Sigma) 1:50 and goat anti-mouse Alexa Fluor 568 (Invitrogen) 1:500–1:1000, and counterstains were FITC-phalloidin (Invitrogen) 0.5–1 µl and Hoechst 33342 (Sigma) 0.1–0.5 µg/ml.

In situ hybridisation and/or Sytox Green (Invitrogen) staining were performed prior to embedding embryos in Durcupan ACM Fluka (Sigma), mixed at a ratio of 32:27:1:0.6  =  components A:B:C:D. Embryos were dehydrated through 10-minute washes in each of 50%, 70%, 90%, 2× 100% ethanol and 100% acetone, transferred to a 1:1 mixture of acetone: catalysed Durcupan, and left uncovered in a fume hood overnight. Embryos were individually transferred to fresh Durcupan in silicone molds (Electron Microscope Sciences NO. 70903) and oriented following a 30-minute initial hardening at 65°C. Resin blocks were baked for 24 hours at 65°C.

Block fronts were trimmed with a razor blade and sectioned at 5–6 µm on a Leica RM2255 microtome with a high-profile knife holder using High-Profile disposable “diamond-edge” steel knives (C.L. Sturkey NO. D554D50). Sections were collected on water droplets on charged slides, dehydrated on a heat block, and mounted in Permount (Fisher Scientific).

dsRNA for all genes (supplementary material Table S1) was prepared as previously described (Kainz et al., 2011) and resuspended in injection buffer (5 mM KCl, 0.1 mM NaH₂PO₄) to a concentration of 2 µg/uL. Male and female adults were injected three days after final molt with 5 µL of 2 µg/uL dsRNA using a Hamilton syringe and size 26 needles. Testes were collected from injected males 27–29 days after injection.

Half of each clutch laid by injected females was fixed for in situ hybridisation, and the other half was homogenised in TRIzol (Invitrogen) and stored at −80°C before isolation of total RNA. RNA was isolated separately from late blastoderm (24–29 hours AEL), early germ band (24–48 hours AEL) and late germ band (72–96 hours AEL) embryos laid by injected mothers. Genomic DNA was treated with Turbo DNase (Ambion) at 37°C for 30 minutes, followed by DNase heat-inactivation and phenol/chloroform extraction. cDNA was synthesised from 120 ng of each RNA sample using Superscript III Supermix (Invitrogen). PCR was performed using Advantage 2 DNA Polymerase from 1 µL of cDNA template and primers indicated in supplementary material Table S1 at 60°C annealing temperature with 35 PCR cycles. RT-PCR results for samples of all three embryonic ages tested yielded indistinguishable results, indicating that maternal RNAi was effective at reducing zygotic transcripts in embryos at least up to four days AEL.

We thank Omar Delannoy Bruno for conducting additional in situ hybridisations, Ian Dunn for assistance collecting embryos, and Jeremy Lynch for providing the T7-linker PCR probe synthesis protocol. This work was partially supported by National Science Foundation (NSF) awards [grant numbers IOS-0817678 and IOS-1257217] to C.G.E., an NSF Predoctoral Fellowship to B.E.-C., and a Frank Knox Memorial Fellowship to T.E.M.J.