Earlier studies in plants, and animals revealed associations between genome size, extinction rates and species richness (Knight et al., 2005, Kraaijeveld, 2010, Olmo, 2006 and Vinogradov, 2004b). The association between genome size and species richness becomes especially apparent in groups with genome sizes larger than 5 pg in amniotes and 14 pg in plants, suggesting that large genomes above these sizes are maladaptive (Knight et al., 2005 and Vinogradov, 2003). Among vertebrates, salamander genome sizes are exceeded only by lungfish. Salamanders and lungfish, however, appear to be an exception to the hypothesis that large genomes increase the risk of extinction (Metcalfe and Casane, 2014).

Here, we examined genome size variation in direct developing, metamorphic and paedomorphic salamanders. Fig. 1A shows the distribution of genome sizes for each salamander family examined, indicating that each family has a restricted and characteristic range of genome sizes. Consistent with earlier reports, families comprising either obligate paedomorphic or partially metamorphic salamanders (Cryptobranchids, Sirenids, Amphiumids, Proteids) typically have larger genomes than other non-obligate neotenes and terrestrial salamanders in the Plethodontidae family. With the exception of the Proteidae, families including paedomorphic and partially metamorphic species in general display a relatively smaller variation in genome size that decreases with increasing C-value (Fig. 1A; supplementary material, Fig. 2S).

Fig. 1B reveals that the Plethodontidae, in contrast, display a large variance in genome size compared to other salamander families. Genome size in Plethodontidae varies 5X, from an average of 14 pg in Desmognathus to an average of almost 60 pg in the Hydromantes lineage. The genus Plethodon, which are among the most species-rich of salamander genera, display the largest variation in genome size among Plethodontidae. Similarly, the Bolitoglossa genus, which is also species-rich, displays a wider range of genome size than other less speciose families. The larger C-value variance in the direct developing Plethodon and Bolitoglossa supports the niche-width hypothesis, according to which genetic variance increases in lineages with greater habitat heterogeneity. The results are summarized in Table 1.

Fig. 2 shows that the average genome size across salamanders increases linearly with the phylogenetic age (evolutionary duration) of the respective family. The time of origin was obtained from Marjanović and Laurin, 2007. An earlier report found that DNA content per nucleus increased with the geological duration of neoteny in obligate paedomorphs (Martin and Gordon, 1995). This trend is reproduced here and extended to include the Plethodontidae (facultative paedomorphs, metamorphic and direct developing salamanders). Pleurodeles and Desmognathinae, for example, are of the most recent origin among the salamanders examined here, and they exhibit correspondingly smaller genome sizes.

Interestingly, Eurycea, comprising species with some of the smallest genome sizes among paedomorphs, is also of more recent evolutionary origin. Previously, it was suggested that neotene salamanders have large genomes because of their fluctuating aquatic environments, which are expected to result in frequent genetic bottlenecks and small effective population sizes (Larson, 1981, Parker and Kreitman, 1982 and Shaffer and Breden, 1989). The smaller Eurycea genomes, however, suggest that life-history, although important, is not predominantly responsible for the evolution of large genomes in paedomorphic salamanders. Preliminary evidence also suggests that the Eurycea, which include facultative paedomorphs, have higher amounts of genetic variation and significantly faster rates of substitution than do most other paedomorphs with larger genomes (data not shown; Table 1).

These observations indicate that either genome sizes have expanded at a constant rate since the origin of a family (Martin and Gordon, 1995), thus resulting in larger genomes in older families; or, conversely, the appearance of younger families in the geological record have coincided with decreases in genome size (Kraaijeveld, 2010); or both processes might be impacting the mode and tempo of genome size evolution (Petrov, 2002 and Petrov et al., 2000). We conclude that smaller genome sizes are associated with families of more recent evolutionary divergence across all salamander taxa examined so far, and propose that similar trends between genome size and phylogenetic age might apply to other vertebrate lineages.

In order to determine whether there might be a correlation between the difference in genome size and the difference in nucleotide substitution rate, we chose an approach employed by Duchene and Bromham (Duchene and Bromham, 2013), which involves identifying phylogenetically independent sister pairs that share a common ancestor. The number of synonymous and non-synonymous substitutions is then determined from the respective branch lengths leading to each species in the sister pair. Because each species in the pair had exactly the same amount of time to accumulate the substitutions, the difference between branch lengths can be used to determine relative rates of substitution since the two species diverged. This approach has the advantages that phylogenetic non-independence is taken into account and divergence times are not required to assess differences in evolutionary rates.

We found that of the six genes available for this type of phylogenetic analysis only two, rag1 and pomc, were adequately represented in the relevant databases (including both gene sequence and genome size). The other genes examined did not yield a statistically reliable number of sister pairs for species with known C-values. Checking for phylogenetic saturation, we found that the pomc gene is saturated in this data set while rag1 is not (supplementary material, Fig. 2S). Hence, rag1, which is a very slowly evolving gene compared to the others, constituted the most reliable candidate gene to probe and assess the effect of genome size on substitution rate in the different sister pairs.

Using a Wilcoxon rank sign test to compare branch lengths (dS) between the larger and smaller genomes in the sister pairs (Fig. 3), we found a significant association between genome size and synonymous substitution rates in rag1 (sample size: 13; W: 21; P-value: 0.032), suggesting that in larger genomes the rag1 gene tends to have slower rates of synonymous substitution. The small number of sister pairs with a dN that was greater than zero (or where the difference was greater than two significant figures) did not allow us to measure the significance of a non-synonymous substitution effect.

A similar analysis was performed on the concatenated set of genes; but no significant correlation was found, suggesting that different genes have different sensitivities to genome size. This result is expected given that mutation/substitution rates vary considerably over the genome, and depend on chromosome context and chromatin organization, e.g., heterochromatin versus euchromatin (Herrick, 2011 and Schuster-Bockler and Lehner, 2012). To our knowledge, this observation represents the first evidence of a gene that exhibits some mutational sensitivity to genome size.

We next examined the relationship between average branch length and genome size at the genus level within Plethodontidae. Fig. 3 reveals phylogenetic relationships between species and genera belonging to the major tribes that comprise the Plethodontidae minus the tribe Ensatinini (Vieites et al., 2011). We compared four clades corresponding to the genera Bolitoglossa, Eurycea, Plethodon, and Hydromantes, Desmognathus and Aneides. The Plethodon are further divided into two clades corresponding to Eastern and Western Plethodon. For each clade, we determined the average C-value and the average branch length (dS) leading to each species from the root node of the respective clade (Table 1).

The results of this analysis revealed no association between average C-value and average branch length among the Plethodontidae, indicating that C-value and rate of evolution are independent, on average, in this family of salamander (Table 1). At the species level, in contrast, sister pairs in the genera Bolitoglossa and Hydromantes, both of which are composed of species with large genomes (> 40 pg), exhibit an unambiguous negative association between genome size and rates of evolution (Fig. 3; Table 1; supplementary material, Fig. 3S).

Similarly, sister pair analysis between the Western and Eastern Plethodon reveals that species with larger genomes in the Western clade tend to have slower substitution rates when paired with species from the Eastern clade (supplementary material, Fig. 4S). Together, these observations suggest that either substitution rates tend to decrease as genome size increases within a sister pair; or, conversely, they increase with decreasing genome size.

The phylogenetic tree based on the species examined here confirms a deep split between the Plethodontidae and the older urodela families, which diverged from each other about 130 million years ago (Mya) (Marjanović and Laurin, 2007, Marjanović and Laurin, 2014 and Mueller, 2006). A recent hypothesis proposes that the dominant mode of evolution involves two regimes: •

bursts of innovation are accompanied by an expansion in C-values;

Genera of Plethodontidae appear in the geological record about 38 Mya (Elmer et al., 2013 and Martin and Gordon, 1995), and have undergone a number of well documented adaptive radiations, for example the genera Plethodon in North America and Bolitoglossa in Central and South America (Elmer et al., 2013 and Vieites et al., 2007). In contrast, the obligate neotene families we examined are of older evolutionary origin (80 to 150 Mya) (Marjanović and Laurin, 2007, Marjanović and Laurin, 2014 and Martin and Gordon, 1995), and appear to have undergone prolonged periods of evolutionary stasis (Gao and Shubin, 2003). This suggests that the Plethodontidae have undergone a mode of speciation and genome size evolution that is distinct from the other salamander families.

To assess any differences in the mode of evolution between the Plethodontidae and the obligate/facultative neotene families, we measured evolutionary rates as determined by dS/Mya, and found that salamander species include some of the slowest evolving vertebrates (supplementary material, Fig. 5S). Interestingly, no significant difference in the average evolutionary rate was found between the Plethodontidae and the older obligate neotenes (Fig. 4), suggesting that urodela might be approaching a minimum permissible rate of evolution in vertebrates (Sung et al., 2012). Evolutionary rates of the Plethodontidae, for example, “bottom out” at a dS/Mya of 0.001—the lowest observed rates in this study (Fig. 4).

Remarkably, the Plethodontidae exhibit a larger variance in their rates of evolution, on the order of ten-fold compared to less than two-fold found in the aquatic paedomorphs (Table 1). This observation is consistent with the wide variation in genome sizes in Plethodontidae (5X; Fig. 1B). Hence, substitution rates and genome sizes in obligate neotene families appear to be more evolutionarily conserved than in the Plethodontidae. Together, these observations suggest that within-family variations in genome size are closely associated with variations in rates of substitution and evolution, and thus salamander families with higher levels of genome size diversity tend to have a wider range of evolutionary rates.

Here, we have investigated the relationship between genome size, genetic variation and evolutionary duration. Our principal findings concern: •

the positive correlation between genome size and phylogenetic age is now extended to genera inside the family Plethodontidae;

Other studies have also provided tentative evidence that large genomes in both plants and animals are associated with lower levels of genetic variation and rates of evolution (Buschiazzo et al., 2012 and Pierce and Mitton, 1980). In plants, for example, gymnosperms, which have significantly larger genomes than angiosperms, are evolving at much slower rates (Buschiazzo et al., 2012). Population genetic effects independent of cell physiology and other molecular traits, however, can apparently account for many of these observations (Larson, 1981 and Shaffer and Breden, 1989).

The results obtained from this study also revealed a weak but significant negative association between genome size and phylogenetic branch lengths in salamanders: shorter branch lengths in sister pairs tend to be associated with the species having the larger genome. Our analysis, moreover, provided preliminary evidence that changes in genome size between sister pairs are positively associated with differences in substitution rates: to the extent which genomes in a sister pair diverge from each other in size, they also tend to diverge in their respective rates of substitution (supplementary material, Fig. 5S).

Our findings, however, appear to be restricted to sister pairs alone, since a comparison between clade averaged rates of substitution in rag1 did not reveal an association between genome size and substitution rate (Table 1). Indeed, we found that despite widely varying life-history traits, different salamander families are on average evolving at similar rates (Fig. 4; Table 1), a finding that contrasts with the comparatively lower levels of heterozygosity reported in obligate paedomorphs (Larson, 1981, Parker and Kreitman, 1982, Pierce and Mitton, 1980 and Shaffer and Breden, 1989). An overall trend nevertheless emerges from the sister pair analysis presented here: rates of evolution (dS/Mya) tend to decrease as genome size increases (supplementary material, Fig. 6S). Additional studies, however, are needed to confirm this observation in urodels and to extend it to other amphibians and vertebrates.

Our analysis revealed a number of interesting findings concerning the relationship between genome size and substitution rates, notably among the Plethodontidae. We found, for example, that the clade including members of the genus Bolitoglossa, which have among the largest genomes within the Plethodontidae, are among the fastest evolving salamander species in this family. Bolitoglossa species with the larger genomes in sister pairs were found, however, to be evolving more slowly, suggesting that slower rates of substitution at the species level are associated with larger genomes within this salamander genus (supplementary material, Fig. 3S). The species B. platydactyla is a notable exception, and is evolving faster than the other Bolitoglossa species examined here, although it has the largest genome (supplementary material, Fig. 3S).

The clade including species belonging to the genus Eurycea revealed a similarly complex relationship between genome size and substitution rate in the Plethodontidae. These species all have similar genome sizes and are evolving at similar rates (data not shown). The species G. porphyriticus and Pseudotriton ruber, however, are evolving more slowly compared to the Eurycea congeners (not shown). When sister pairs are averaged over all species of the clade comprising Eurycea, we found that genome size and branch length nevertheless tend to be negatively associated within this genus (supplementary material, Fig. 3S).

An examination of the genus Plethodon revealed that, independently of genome size, these species are evolving more slowly than species in either the Eurycea or Bolitoglossa genera (supplementary material, Fig. 3S). We also found that species in the clade comprising the Western Plethodon (average C-value: 43.4 pg) are evolving more slowly than the clade comprising the Eastern Plethodon (average C-value: 24.6 pg), again supporting a potential negative association between genome size and substitution rates (supplementary material, Fig. 4S). The sister pair P. vandykei (69 pg) and P. larselli (48 pg), however, represent another exception to the general trend, and underscores the importance of other lineage-specific influences on substitution rates.

Finally, the clade including the genera Aneides (42–44 pg), Desmognathus (14–18 pg) and Hydromantes (42–72 pg) likewise reveals a highly heterogeneous relationship between genome size and rag1 evolutionary rates in Plethodontidae (supplementary material, Fig. 3S). In this clade, H. genei and H. italicus are the fastest evolving species while the Desmognathinae, which have the smallest genomes, are the slowest evolving. The Hydromantes sister pair, in contrast, exhibits a clear negative association between genome size and substitution rate in rag1 that is not found in either the Desmognathus or Aneides sister pairs. Indeed, sister pairs within the latter two genera do not display as large a difference in genome size compared to the Hydromantes sister pair, suggesting that differences in genome size must exceed a certain threshold before an unambiguous association between C-value and substitution rate can be detected (Fig. 3).

Earlier studies reported an apparent correlation between nuclear DNA content and evolutionary duration (phylogenetic age) in salamanders (Fig. 2) (Martin and Gordon, 1995). The authors interpreted the trend as evidence that genome size increases depending on how long a species has been an obligate neotene, and proposed that the rate of junk DNA accumulation could be used as a possible second molecular clock (accumulating 0.63 pg/Mya). This interpretation is consistent with observations that the ancestral urodel genome was much smaller and comparable in size to extant mammalian genomes (approximately 3 pg), indicating a massive expansion in salamander C-values since their time of origin (Organ et al., 2011).

Alternatively, the trend reproduced and extended here might suggest that, subsequent to long periods of expansion, reductions in genome size accompany evolution and speciation in salamanders (Kraaijeveld, 2010). We note, for example, the higher levels of diversity in evolutionary rates in the Plethodontidae (supplementary material, Figs. 3S and 5S), which is consistent with the corresponding tribes and genera radiating more recently into more heterogeneous terrestrial niches (Nevo and Beiles, 1991). Additionally, the Plethodontidae have the widest and most diverse range of genome sizes among salamander families, and include the smallest existing urodela genomes (Fig. 1B). Based on these and other observations, we propose that elevated levels of molecular diversity, both in terms of genome size and mutation rates, tend to be associated with smaller genomes and more recent adaptive radiations.

Due to a variety of evolutionary constraints, organisms with large genomes appear to be under selective pressure to evolve more efficient DNA replication and repair processes that minimize mutation rates to a level set by genetic drift (Massey, 2008 and Sung et al., 2013). How organisms achieve an optimal balance between mutation (predominantly DNA replication errors) and DNA repair rates is unclear at the molecular level, but compartmentalizing mutation rates within the genome between early and late replicating DNA provides one plausible, though not complete, explanation [for a more detailed discussion see (Herrick, 2011)] (Wintersberger, 2000).

Based on the considerations discussed above and preliminary observations on the effect of genome size on nuclear substitution rates, we propose a hypothesis according to which speciation events in salamanders are associated with contractions in genome size that are indirectly yet mechanistically related to increases in mutation and substitution rates in nuclear genes (Herrick, 2011). Accordingly, genome reduction (expansion) and reorganization during speciation might result in changes in the DNA replication and repair programs that impact rates of mutation and evolution. Eukaryotes with larger genomes, for example, rely more heavily on error prone DNA repair pathways, which operate preferentially in late S and G2/M phases, compared to other eukaryotes with smaller genomes (Farlow et al., 2011).

A corollary to that proposal implies that species with larger genomes should display correspondingly higher mutation and divergence rates in the latest replicating sequences (such as microsatellite DNA and TEs) compared to species with smaller genomes. Conversely, sequences replicating earliest in S-phase (such as housekeeping genes) should display correspondingly lower mutation rates in species with larger genomes compared to those with smaller genomes (Herrick, 2011). Some evidence suggests that this indeed is the case (Chen et al., 2010, Lang and Murray, 2011, Stamatoyannopoulos et al., 2009 and Weber et al., 2012). TE rich regions, tend to replicate late in the vertebrate S-phase, and therefore they experience higher rates of mutation and genetic erosion (Chen et al., 2010, Stamatoyannopoulos et al., 2009 and Weber et al., 2012).

Higher rates of mutational erosion and genetic extinction can explain why in lungfish and salamanders, which both have exceptionally low levels of genetic variation, TE content appears to be under-represented in the respective species’ genomes compared to other species with lower C-values (Metcalfe and Casane, 2014). The lower than expected proportion of TEs in large genomes can be explained by their inactivation and decay over time (Metcalfe and Casane, 2014) and/or through a genome size dependent reliance on error prone DNA repair pathways operating in late replicating DNA (Diamant et al., 2012 and Mao et al., 2008). Consequently, a large proportion of the lungfish and salamander genomes are expected to be comprised of late replicating “fossilized” TEs that have experienced correspondingly higher rates of mutation and substitution. Whether or not mutation rates between early and late replicating DNA are anti-correlated in a genome size dependent manner remains, however, to be demonstrated.

The findings presented here suggest that the relationship between mutation rate, rates of evolution and genome size in vertebrates and other eukaryotes warrants further investigation. Genome size, is a proxy variable for a number of different physiological, molecular and life-history traits such as cell cycle duration, metabolic rate and developmental time. It will be interesting to investigate how these variables interact with each other dynamically to modify and shape the overall architecture of the genome in different species during the course of evolution.