Causes and evolutionary consequences of primordial germ-cell specification mode in metazoans

Numerous hypotheses that might explain the putative evolutionary advantage of specifying germ cells early in development have been proposed over the years (discussed in ref. 19), but most of these hypotheses have not been tested empirically or specifically investigated regarding different potential fitness advantages of inheritance vs. induction. To our knowledge, the only such hypothesis that has been subject to explicit testing is what we refer to herein as the “PGC-specification hypothesis” (presented in ref. 3 and 20). This hypothesis asserts that the selective advantage of inheritance is that the early disentanglement of germ-line specification from somatic influences results in the “liberation” of selective pressures on somatic gene regulatory networks (GRNs), thus enhancing species evolvability. In turn, this liberation accelerates an organism’s protein sequence evolution under germ plasm, particularly of genes expressed during early stages of development, and leads to species radiations.

The hypothesis’ predictions pertaining to molecular evolution of genes has been purported to be empirically supported by findings of faster evolution of protein sequences in clades with inheritance than in clades with induction (faster rates were inferred for up to 32% of the genes in a genome under inheritance) based on analysis of four pairs of vertebrates, with one member of the pair exhibiting inheritance and the other induction [anurans vs. urodeles, birds vs. crocodiles/turtles, snakes vs. lizards, and one clade of ray-finned fishes (Teleostei) vs. another (Acipenseriformes) (20)]. As with any hypothesis, testing the predictions by using different methods and systems would be helpful to assessing its generalizability, particularly because that pioneering assessment had some limitations. For instance, because of extreme divergence of taxa being compared and saturation of substitutions, only nonsynonymous coding-DNA substitutions (dN) were studied, excluding synonymous site changes (dS) and dN/dS, which would have been needed to assess or make conclusions about variation (or liberation) in selective pressures under preformation (21, 22). Importantly, the study comprised numerous overlapping nonphylogenetically independent contrasts, which is pseudoreplication (23, 24), an approach known to cause spurious correlations (23, 24). Small gene sample sizes for some taxa, the use of tissue-specific transcriptomes, and exclusion of invertebrates, which comprise the vast majority of animal life, also may have limited those findings (25) (see further details on these limitations in SI Appendix, section 2). Accordingly, further investigations using alternate approaches and organisms would be useful to assess whether the purported liberation of constraint and increase in rates of protein-coding sequence evolution under germ plasm (20) is robust to different types of methods and animal systems.

Our recent comparative large-scale molecular evolutionary analysis using a different study approach and different datasets, and spanning both vertebrates and invertebrates, did not support the predictions of the PGC-specification hypothesis (25). The assessment was based on genome-wide dN/dS of phylogenetically independent species pairs from 12 genera with different PGC-specification modes [the invertebrate genera Drosophila, Nasonia, Schistosoma, Anopheles, and Pristionchus (inheritance) and Tribolium, Echinococcus, and Apis (induction); and the vertebrate genera Falco and Xenopus (inheritance), and Alligator and Pan (induction)]. The results of these analyses supported the null hypothesis that PGC-specification mode has no consistent effect on protein sequence evolution, including on the evolution of the sequences of early developmental genes (25). Thus, these recent findings suggest that, at a minimum, the PGC-specification hypothesis does not hold in some animal genera, and that there is a good possibility that germ plasm is not strongly, or potentially even marginally, linked to rapid protein sequence divergence. Although further testing in even more taxa will be helpful as more genomic data become available, these results already comprise significant examples countering the notion that germ plasm generally and broadly releases selective constraint on genes and somatic gene networks in animals, suggesting that alternate causes of the evolution of germ plasm should also be explored.

Because previous and detailed arguments have already been made for why germ plasm should accelerate evolution (3, 20), here we consider the other possible scenario—that is, that the null hypothesis may hold for many or most animals, and that germ plasm does not give rise to rapid animal evolution, including a broad effect on gene sequence evolution (see SI Appendix, section 4 for discussion of a possible smaller-scale effect on the evolution of specific genes). To explain the phylogenetic distribution of PGC-specification modes across animals, and extended to the PGC-specification hypothesis, it was proposed that “the distribution of epigenesis and preformation must result from the influence each mode of germ cell specification has on the development of the soma” (3). The proposition that inheritance (preformation) must evolve convergently because of its effect on somatic tissues was suggested to be supported by certain characteristics of the genetic mechanisms regulating somatic development. For example, it was noted that frogs and teleosts, which specify germ cells by inheritance and exhibit a complex GRN governing embryonic mesoderm formation, possess multiple copies of the mesoderm inducers Nodal and Mix. In contrast, Axolotls (A. mexicanum, Mexican salamanders), which specify germ cells by induction, contain just one copy of each of Nodal and Mix. This finding was interpreted as resulting from greater constraint in the induction species and the evolution of novelty in the frog GRN (and presumably also in teleost GRNs) due to the liberation of constraint under germ plasm (ref. 3; see also ref. 32). Although this scenario indeed comprises one feasible possibility, it is also worth considering that the differences in mesoderm GRNs between these taxa for this sample of genes might result from various other factors. For instance, it is feasible that there is a greater tendency for tandem duplications of chromosomal regions, neofunctionalization and subfunctionalization, and/or fewer (duplicate) gene losses in frogs and teleosts (36–38), factors potentially unconnected to PGC-specification. Another contributing factor may be whole-genome duplications, which have been linked to rapid evolution of duplicated genes and are believed to have occurred within some frog (e.g., X. laevis; ref. 39) and teleost lineages (36, 40, 41), but likely not in the salamander A. mexicanum (42). In this regard, although fewer gene copies under induction could reflect greater constraint on these gene pathways (3), further studies should aim to further disentangle such an effect from other plausible factors unrelated to PGC-specification mode. For example, it would be useful to evaluate patterns of duplications of developmental patterning genes, their evolution, and their distribution among inheritance vs. induction species across a wide range of animal taxa.

The distribution of inheritance and induction modes across metazoans (Fig. 1) suggests that induction is ancestral to Bilateria and prevalent throughout both protostomes and deuterostomes, whereas inheritance has been derived in multiple lineages (1, 19). Although there are numerous examples suggesting that the inheritance mode arises from an ancestral induction mechanism (1), reports of clades that might exemplify a transition from induction to inheritance remain rare (1, 66), suggesting that the shift to germ plasm is typically irreversible. Germ plasm thus appears to follow “Dollo’s Law”—that is, structures or processes lost in evolution are unlikely to be regained by descendants in the same form as the ancestors (67, 68). In this aspect, the inheritance mode resembles other typically irreversible transitions, such as the transition from outcrossing to selfing and from hermaphroditism to dioecy (69). In plants, for example, the shift from blue to red flowers in Andean Iochroma (Solanaceae) is irreversible; molecular studies suggest that loss of a gene encoding an enzyme in the anthocyanin pathway is needed for the transition, which limits the opportunity for later reversal to the ancestral state (69, 70). Accordingly, the transition from induction to inheritance could, in principle, be made irreversible by a single gene mutation or gene loss in the induction pathway. Although it is known that germ plasm has a distinct molecular genetic mechanism from that of induction, as observed in the contrast between the mechanisms used by the models D. melanogaster and M. musculus (2), this difference alone should not be sufficient to prevent a reversal to induction, unless the induction pathway has been impaired by within-gene mutations, gene losses, or gene silencing. Indeed, although key genes needed for induction in mice and crickets, such as BMP signaling pathway members and specific downstream transcription factors (2, 71), are highly pleiotropic, they show no signs of being used in early PGC specification in Drosophila (but see ref. 72 for evidence of a role for BMP signaling in PGC fate maintenance), or in other less-well-studied taxa with germ plasm (2). Furthermore, the ablation of PGCs or their precursors in animal models with inheritance (e.g., D. melanogaster, C. elegans, D. rerio, and X. laevis) is generally not corrected by inductive signaling and de novo establishment of PGCs, suggesting that a putative ancestral inductive PGC-specification mechanism has been lost in those taxa (see, for example, ref. 73). In this regard, gene loss, and possibly mutations or silencing in upstream regulators under induction, could contribute to an irreversible transition to the inheritance mode of PGC specification in metazoans. Nonetheless, during or shortly after the transition from induction to inheritance, one might expect gradual loss of induction mechanisms (19), and thus in this period a taxon could exhibit some reversal capabilities, as has been implied for the solitary ascidian Ciona intestinalis (74). Further studies will be needed to ascertain whether unambiguous examples of species exhibiting both PGC-specification modes can be identified, in support of this hypothesized transition period, or whether a complete irreversible transition to germ plasm, including the loss of induction functionality, is prevalent and likely occurs rapidly.

Under a presumption that the inheritance mode comprises an adaptation, one must question whether this adaptation occurs via divergent or similar genetic mechanisms. Unlike other derived traits—such as the transition to wings in insects, birds, and bats (75) or from a primitive photoreceptor to the camera-eye found in octopus and vertebrates (76, 77), both of which have apparently unambiguous advantages in terms of adapting to environmental conditions—an adaptive advantage of germ plasm appears to be less obvious. Convergent phenotypes can arise from independent genetic pathways, as is believed to occur for wing formation (75, 78), or can result from similar genetic mutations that arise in independent lineages. The latter may include convergent phenotypes arising from a mutation in orthologs or orthologous pathways between lineages (parallel evolution), a shared allele that was polymorphic in ancestral populations, or from shared introgressions (collateral evolution) (75). In the case of germ plasm, the data to date suggest that this convergent phenotype results from distinct genetic pathways in different lineages. For example, the oskar gene has been shown to be sufficient for germ-plasm assembly in D. melanogaster (79–81). This gene has been identified in a number of insect lineages (66, 82), but lacks known orthologs in noninsect metazoan lineages, including those with germ plasm (e.g., birds, fish, and frogs) (82). The novelty of oskar to insects is one indication that germ plasm arose independently via different genetic mechanisms across the metazoans, at least between the insects and noninsect animal systems. Consistent with the notion of independent genetic regulators of germ plasm across animals, the bucky ball gene in the zebrafish D. rerio has been shown to be critical for germ-plasm assembly in that taxon (83–85), whereas this process is believed to be modulated by the nematode-specific MEG (maternal-effect germ-cell defective) and PGL (P-granule abnormality) genes in the model nematode C. elegans (86, 87). It is worth noting that oskar (Drosophila), bucky ball (Dario), and MEG/PGL (Caenorhabditis) differ in their pleiotropic roles. For example, although all genes are involved in germ-plasm function or assembly, only the former two genes have been shown to also play a role in embryo or egg/oocyte axial polarity (53, 85, 88–90), findings consistent with differences in their evolutionary dynamics among taxa. In summary, it appears feasible that at least in some lineages, inheritance arose rapidly, potentially due to rare highly beneficial mutations, novel gene evolution, and/or introgressions across populations, leading to fixation of germ plasm based on only a few genetic changes (75, 91).

Given that no specific example of shared mutations or introgressions have yet been identified that could explain every instance of the evolution of germ plasm across metazoans, the most plausible scenario is that this phenotype arose repeatedly via different genetic mechanisms. Furthermore, the biological properties of germ plasm differ in some respects across taxa: Germ plasm can originate as a molecular accumulation in the oocytes before fertilization (D. melanogaster, D. rerio, Gallus gallus, and X. laevis) or clustering and concentration of P granules in the early embryos (C. elegans), and each of these mechanisms will ultimately drive formation of PGCs in early embryogenesis (92). In addition, a number of marked differences have been noted between Xenopus (frogs) and Danio (zebrafish) species in terms of the embryological location and specific molecular components of germ plasm (3). Thus, germ plasm appears to be functionally (or analogously) convergent, but the underlying mechanisms are not necessarily perfectly developmentally or genetically convergent. Together, these findings concur with the hypothesis of convergent evolution of inheritance as a PGC-specification mechanism. An argument is thus available for the existence of an innate selective advantage to germ-line determination via localized maternal germ plasm, causing it to arise independently, with highly similar functionality, but with different genetic mechanisms and biological properties, across diverse systems.

A fitness advantage of germ plasm compared with induction is not currently known. Speculatively, however, it can be considered that germ plasm might be particularly beneficial under a change to an unfavorable or dynamic environment (referred to hereafter as “stressful”) that reduces reproductive output or germ-line and embryo survival. In principle, when the maternal determinants are synthesized within the female sex organs, it may make the biochemical cost of development in the early embryo substantially lower, by reducing costs of RNA and protein synthesis and/or cellular transport to asymmetrically localize axial or regional determinants, thus increasing germ-line establishment and embryo survival rates under stress. For oviparous organisms (egg-bearing reproduction), germ plasm may be particularly beneficial, because the molecules would not need to be synthesized in a nonmaternally supported, independent embryo under a shift in environmental conditions, as they would under the induction mode. In viviparous (and ovoviviparous; live-bearing reproduction) organisms, where the embryo remains supported by the female throughout embryogenesis and PGC formation, a reduced cost in establishment of the germ line and embryo proper could also, in principle, be advantageous to offspring survival. Nonetheless, assuming germ plasm is beneficial to survival under a dynamic changing environment, one may speculate that any putative benefit of germ plasm might be particularly elevated for oviparous organisms that lack the ongoing support/investment from maternal tissues, which can reduce an offspring’s probability of survival (93). Thus, as further discussed in SI Appendix, section 5, reproductive lifestyles might be a significant factor influencing the evolution of germ plasm in animals.

Given that the PGCs and germ lines are the source of all heritable genetic mutations, and thus genetic variation available in biological systems, they play a central role in evolution. Mutations are thought to arise primarily from DNA-replication errors, but can also result from faulty DNA repair, transcription-mediated mutation, and environmental and physiological agents (47, 94–98). The mutation rate in coding DNA per generation underlies a diverse set of evolutionary phenomena, including the evolution of sex, aging, recombination, mating systems, species extinctions, reproductive isolation, and speciation (30, 99). Evidence to date has shown that the mutation rate in coding DNA varies among animals (95, 100, 101), and the rate itself can be subject to selective pressures (98, 101, 102). Given the central role of germ-line mutations to evolutionary biology, it is worth considering whether and how these mutations—and particularly mutation rates in the germ lines—could be related to the evolution of the germ-line soma divide and to PGC-specification mode in animals. We highlight putative differences in the mutation rates of the germ line and soma in SI Appendix, section 6, and describe how PGC-specification mode may affect germ-line mutation rates below.

Here, we have described existing hypotheses from the literature and set forth additional hypotheses and proposals for further consideration, with respect to the causes and consequences of PGC specification mechanisms in metazoans (SI Appendix, Table S1). Together, the data to date suggest that the transition to germ plasm in metazoans occurred convergently via different genetic and developmental mechanisms, which may have involved adaptive processes, or, alternatively, may have arisen as a spandrel effect. Furthermore, PGC specification may be connected to life history parameters such as oviparity and viviparity. We argue that, because PGC specification mode is indispensable for germ-line formation, it is apt to affect the germ-line genomic mutation rate, which is one of the most crucial parameters in evolutionary biology. Expanding research in this area will thus be essential to gaining an understanding of the nature of that relationship.

We thank Eric H. Davidson, participants of the Sackler Colloquium on “Gene Regulatory Networks and Network Models in Development and Evolution” for discussion of these issues, and two anonymous reviewers for valuable comments on an earlier version the manuscript. This work was partially supported by funds from Harvard University and NIH Grant 1 R01 HD073499 (NICHD) (to C.G.E.).