Acquisition of Germ Plasm Accelerates Vertebrate Evolution

See allHide authors and affiliations

Science  11 Apr 2014:
Vol. 344, Issue 6180, pp. 200-203
DOI: 10.1126/science.1249325

Tangling Evolutionary Trees

Evolutionary rates tend to vary among taxa and may result in phylogenetic trees that do not reflect the true relationships among taxa, depending on the sequences input into the analysis. Examining vertebrate trees, Evans et al. (p. 200) demonstrate that differences in evolutionary rates, leading to phylogenetic distortions, are correlated with the mechanisms underlying germ cell formation. Evolutionary rate is faster in cases where germ cells are established by maternal molecules (“preformed”) relative to those that are induced during embryogenesis (“epigenesis”) in slowly evolving and, presumably, ancestral lineages. For example, frogs evolve more rapidly than salamanders, and teleosts more rapidly than ascipenseriform fishes. Thus, epigenesis constrains the ability of gene regulatory networks to change, with the repeated and convergent evolution of preformation eliminating this constraint.


Primordial germ cell (PGC) specification occurs either by induction from pluripotent cells (epigenesis) or by a cell-autonomous mechanism mediated by germ plasm (preformation). Among vertebrates, epigenesis is basal, whereas germ plasm has evolved convergently across lineages and is associated with greater speciation. We compared protein-coding sequences of vertebrate species that employ preformation with their sister taxa that use epigenesis and demonstrate that genes evolve more rapidly in species containing germ plasm. Furthermore, differences in rates of evolution appear to cause phylogenetic incongruence in protein-coding sequence comparisons between vertebrate taxa. Our results support the hypothesis that germ plasm liberates constraints on somatic development and that enhanced evolvability drives the evolution of germ plasm.

The germ line of metazoans is established early in development with the specification of primordial germ cells (PGCs). Among vertebrates, the conserved mechanism for PGC specification involves their induction from pluripotent cells by extracellular signals, a process referred to as epigenesis (1, 2). However, in several lineages of vertebrates, an alternative mechanism evolved, termed preformation. Here, PGCs are determined by inheritance of germ plasm. Preformation evolved by convergence, which suggests that it may confer a selective advantage. Accordingly, the evolution of germ plasm is associated with morphological innovations and enhanced numbers of species within individual clades (1, 3, 4). Why this derived mode of PGC specification evolved repeatedly in vertebrates is unknown.

The best-studied contrast of epigenesis and preformation is within amphibians. The PGCs of urodele amphibians (salamanders) are specified by epigenesis, whereas in its sister lineage, anurans (frogs), PGCs contain germ plasm (5). Using the axolotl (Ambystoma mexicanum) as a model urodele, the ancestral gene regulatory networks (GRNs) for pluripotency and mesoderm specification in vertebrates were identified (6, 7). These GRNs were conserved through the evolution of mammals (6, 7), which also employ epigenesis (8). In contrast, in frogs the master regulators of pluripotency as employed in mammals have been deleted (6, 9, 10), and the GRN for mesoderm underwent expansions of key regulatory molecules (7, 11). Similar genetic innovations evolved in the GRNs for zebrafish development (12), which also uses preformation (13). The correlation of germ plasm with genetic change has been proposed to result from the relaxation of constraints on somatic development imposed by maintaining the PGC induction pathway (1, 3, 4). To investigate this possibility, we compiled available expressed sequence tag, mRNA and cDNA sequences from vertebrates (fig. S1A and table S1) identifying ortholog pairs shared between sister taxa with different modes of PGC specification and an appropriate mammal and outgroup sequence (14) (fig. S1B). To increase sequence numbers from organisms using epigenesis, we generated transcriptomes from the axolotl and an Acipenseriforme, Acipenser ruthenus (the sterlet) (14), identifying 82,954 sequence clusters across all vertebrates. All analyses were performed with protein coding DNA sequence, excluding the saturated third position (14) (figs. S2 and S3).

Of the 56 published gene trees involving an anuran and a urodele, 29 do not recapitulate the known species phylogeny (table S2). The majority of the incongruent gene trees group urodele and mammal species together, both of which undergo epigenesis, to the exclusion of anurans. We generated unrooted four-taxon trees to investigate the extent of this incongruity (14), presenting trees rooted on the known outgroup (a Teleostei sequence) (Fig. 1A). Within these trees, 54.1% (4355 of 8045) of amphibian sequences show the expected species phylogeny (>70% bootstrap), grouping anuran with urodele. The majority of the remainder [32.1% (2584 of 8045)] incongruently group urodeles with mammals (Fig. 1A). The Shimodaira-Hasegawa (SH) test reduces the number of significant trees overall (P < 0.05), increasing the proportion of trees reflecting the species phylogeny (14). Orthology groups that do not reflect the species phylogeny (28%) are three times as likely to place the urodele sequence with the mammal (Fig. 1A, fig. S4A, and table S3). We next considered each amphibian species in turn, grouping them by mode of PGC specification (Fig. 1, B and C). We show that when a tree is incongruent, any given anuran sequence is less likely than its orthologous urodele sequence to group with mammals (Fig. 1, D and E). These results do not depend on the inclusion of the urodele transcriptome data (fig. S4).

Fig. 1 Amphibian four-taxon tree topologies.

(A) Number of significant trees by bootstrapping (>70%) and SH test (P < 0.05) for each topology rooted with a Teleostei sequence. (B and C) The proportions of species phylogeny (black), mammal-urodele (gray), and mammal-anuran (white) topologies per species. (D and E) The likelihood of each species grouping with mammals when the tree is incongruent; species using preformation are shown in red, those using epigenesis in blue. Dashed lines indicate equal probability of species grouping with mammal or outgroup. [(B) to (E)] Only species with >20 significant trees are shown. The results excluding the transcriptome are shown in fig. S4.

Within Actinopterygii (ray-finned fishes), Teleostei (teleosts) use preformation, whereas Acipenseriformes (sturgeons and paddlefish), which maintain primitive embryological and adult traits, most likely have retained epigenesis (1, 4, 13). We identified 19,394 trees with >70% bootstrap support, of which 68.2% (13,233) reflect the species phylogeny. The majority of the remainder [24.5% (4757)] incongruently group Acipenseriformes with mammals (Fig. 2A). The SH test reduces the total, but still Acipenseriforme sequences are 5 times as likely to group with mammals when the species phylogeny is not obtained (Fig. 2A). Subdividing the data by species reveals a clear distinction between Teleostei and Acipenseriformes; in incongruent trees, Acipenseriformes are more likely to group with a mammal (Fig. 2, B to E). This is true for all 59 Teleostei analyzed with bootstrap-supported trees and 22 of 23 Teleostei supported by the SH test (fig. S5, A and B, and table S4). These results remain true even if the transcriptome data are excluded (fig. S5, C and D)

Fig. 2 Actinopterygian four-taxon tree topologies.

(A) Number of significant trees by bootstrapping (>70%) and SH test (P < 0.05) for each topology, rooted with an amphioxus sequence. (B and C) The proportion of species phylogeny (black), Mammal-Acipenseriforme (gray) and Mammal-Teleostei (white) topologies per species. (D and E) The likelihood of each species grouping with mammals when the tree is incongruent; species using preformation are shown in red, those using epigenesis in blue. Dashed lines indicate equal probability of species grouping with mammal or outgroup. [(B) to (E)] Only Acipenseriformes with >20 significant trees and eight Teleostei species are shown; the results for all species are in fig. S5, A and B. The results excluding the transcriptome are shown in fig. S5, C and D.

We next investigated the sauropsids (reptiles and birds), determining four-taxon tree topologies. In sauropsids, preformation evolved independently in lepidosaurs (lizards and snakes) and in archosaurs (crocodiles and birds) (1518). The lepidosaurs, which experienced a change in the rate of evolution (19), and archosaurs, separated ~280 million years ago (20). The turtle lineage (testudines), using epigenesis, is closer to the archosaurs than lepidosaurs (21). Thus, we analyzed the sauropsids in two subdivisions—the archosaurs and testudines, and the lepidosaurs. Within these groups, we compared birds (preformation) with crocodiles and testudines (epigenesis) and similarly snakes (preformation) with Gekkota and Iguanidae lizards (epigenesis) (1518). Within the sauropsids, almost all the four-taxon trees support the expected species phylogeny and do not subdivide by the mode of germ cell specification in this analysis (fig. S6 and table S5), although the total number of sequence comparisons was low (fig. S1A).

Nonetheless, among the amphibians and actinopterygians, but not the sauropsids, when in an incongruent tree, species using epigenesis are more likely to group with mammals (Figs. 1,2). Such incongruent phylogenies may be driven by differences in the rate of sequence evolution (19, 22), and organisms that have acquired preformation are typically more speciose than those using epigenesis (3, 4). We therefore used three-taxon multiple alignments to determine how the relative rate of sequence evolution differs between sister taxa. Among amphibian species, 32.3% of sequences are evolving at significantly different rates (P < 0.05), of which 87% show urodele sequences evolving slower than anurans (Fig. 3A and fig. S7B). Within the actinopterygians, ~50% of sequences evolve at significantly different rates (P < 0.05), with almost all showing that Acipenseriforme sequences are slower than Teleostei (Fig. 3B and fig. S7, A and B). Furthermore, in the sauropsids, ~20 to 25% of sequences are evolving at significantly different rates, with the majority of slow-evolving sequences in organisms using epigenesis (P < 0.05) (Fig. 3, C and D). Thus, sauropsid sequences exhibit differences in the rate of sequence evolution that correlate with the mode of PGC specification.

Fig. 3 Relative-rate test results.

(A to D) Proportion of sequences evolving at significantly slower (filled) or faster (clear) rates in each species (P < 0.05; preformation shown in red, epigenesis in blue). (A) Amphibians. (B) Actinopterygians, including eight Teleostei species (all Teleostei shown in fig. S7G). (C) Archosaurs and Testudines. (D) Lepidosaurs. Only species with >20 significant sequences are shown. (E) Summary of relative-rate data across all vertebrates grouping species by epigenesis or preformation. [(A) (B), and (E)] Excluding transcriptomes are in figs. S7, C to E, respectively.

Combining these data across classes, 56% (69,165 of 121,373) of all analyses show no significant difference in the rate of sequence evolution (Fig. 3E and table S6). Only 2319 of 121,373 relative-rate tests (<2%) showed a sequence derived from an organism with epigenesis evolving faster than its ortholog. The remaining 41.1% of comparisons (49,898 of 121,373) suggest that sequences from organisms using epigenesis are evolving more slowly. Ranking each species by the proportion of slower-evolving sequences separates organisms using epigenesis from those using preformation, regardless of taxonomic class (fig. S7F).

To investigate functional properties of sequences showing accelerated rates of evolution and incongruent phylogenies, we mapped our results to the mouse and zebrafish genomes (see supplementary text). The proportion of sequences showing evidence of accelerated evolution is significantly higher among genes expressed early in development (chi-square test, P < 0.05) and decreases in genes expressed at later stages (fig. S8). Previous reports demonstrate that early genes are under the highest levels of developmental constraint (23), and our data suggest that early genes are the most likely to evolve at faster rates in species employing preformation. Together, this supports the hypothesis that the evolution of germ plasm liberates constraints on early development (1, 4).

We next considered the correlation of results where individual sequences had been tested for both rate and incongruent tree topologies (table S11). All four-taxon trees were rooted on species using preformation, yet our preceding analyses suggest that these outgroups also have accelerated rates of evolution compared with their sister taxa (Fig. 4A). Because tree reconstruction may fail as a consequence of two long branches clustering (long-branch attraction)(Fig. 4B) (24), we asked whether the incongruent trees were driven by the differences in rates observed in the outgroup sister taxa.

Fig. 4 Phylogenetic incongruence driven by rate of evolution.

(A) Summary of relative-rate test results. (B) Tree diagrams illustrating long-branch attraction driven by outgroup choice in four-taxon trees. (C) Number of amphibian four-taxon tree topologies grouped by relative-rate differences between anurans and urodeles, and Teleostei and Acipenseriformes. n.s, not significant; N/A, not available; Anu, Anuran; Uro, Urodele; Tel, Telesotei; Aci, Acipenseriforme. (D and E) For the four common relative-rate test results between Amphibians and Actinopterygii, the proportions of Amphibian four-taxon tree topologies are shown when using Teleostei or Acipenseriforme outgroups. (D) Bootstrap trees. (E) SH test trees.

For the majority of amphibian four-taxon trees where the outgroup sister taxa differ in rate, the Teleostei sequence is evolving significantly faster than its Acipenseriforme ortholog [P < 0.05, (fig. S9A)]. We therefore rebuilt trees using Acipenseriformes as the outgroup. This increased the proportion of trees congruent with the species phylogeny at the expense of trees grouping urodele sequences with mammals (fig. S9B and table S12). Grouping sequences by both relative rate and tree topology revealed that the highest proportion of incongruent trees occur when the relative rate differs within amphibians (Fig. 4C). If rate differences drive incongruence, changing outgroup should only affect those trees where the outgroup rates differ. Where the actinopterygian sequences do not significantly differ in rate, proportions of incongruent trees remain similar as the outgroup species changes (Fig. 4, D and E, and table S13). Where actinopterygian sequences significantly differ in rate, the proportion of incongruent trees is reduced using an Acipenseriforme rather than a Telesotei outgroup (P < 0.05). The most dramatic change occurs when both amphibian and actinopterygian sequences significantly differ in rate (P < 0.05), suggesting that the faster rate of evolution in both anuran and Teleostei sequences drives the observed incongruence.

The natural history of vertebrates is punctuated with the repeated evolution of germ plasm associated with embryological innovations, gross morphological changes in adults, and enhanced speciation (14). Germ plasm functions to segregate PGCs from somatic cells at the inception of development, and we propose that this relaxes genetic constraints on the mechanisms that govern early somatic development (4). Our results identify a consistent bias in changes in the rate of sequence evolution in species using preformation compared with their sister taxa that use epigenesis. No other biological property correlates as well with the observed changes in rate (fig. S10). Sequences expressed during early development are under high levels of developmental constraint (23), and we show that these sequences exhibit a release of constraint in species using preformation. Taken together, these data suggest that the acquisition of germ plasm liberates developmental constraints, leading to increased rates of sequence evolution and enhanced speciation. They support the hypothesis that enhanced evolvability is responsible for the repeated evolution of germ plasm (14).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S10

Tables S1 to S13

References (2578)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: All data are available from the authors’ Web site at and are deposited at (DOI:10.5061/dryad.rd70f). The authors thank B. Crother, G. Morgan, and M. Blythe for helpful discussion. Sequencing was carried out at the Genome Centre, Queen Mary, University of London by L. Bhaw-Rosun and M. Struegbig. This work was supported by the Biotechnology and Biological Sciences Research Council, Medical Research Council, and University of Nottingham.
View Abstract

Navigate This Article