Phylogeographic Divergence
Our recent phylogenomic dataset sampled 59 individuals and populations
and estimated four geographic lineages (Pyron et al. 2022), but with
fewer samples outside of the Blue Ridge mountains and including only 233
loci. Previous mitochondrial datasets indicated substantial population
divergence in the species (Beamer and Lamb 2020; Pyron et al. 2020),
with uncorrected ‘p’ distances in the COI barcode locus of
1.4–6.3%. Here, we recover three lineages rather than four,
corresponding to two major mountain ranges (Cohutta and Nantahala) and
the expansive Piedmont and associated populations. Accounting for IBD
when estimating population structure in conStruct suggests only two
major clusters (Piedmont versus montane). Overall, lineages of Seepage
Salamander are deeply divergent (~1.2Ma,
~4.7% mitochondrially), but exhibit extensive admixture
both at their contact zones and deep into the core geographic range of
each set of populations. The between-lineage Fst values border on “meaningful” significance (usually considered
~0.15) for population differentiation (Hartl and Clark
2007), but generally fall outside or on the low end of the “grey zone”
of genomic divergence (Fst >0.2,Da >0.01) indicative of speciation
across a wide variety of animals (Roux et al. 2016). Recent migration
rates inferred from Fst and estimated by GADMA
are >1 migrant/generation between all three lineages,
suggesting relatively high and constant rates of gene flow through time.
The overall relationship between geographic and genetic distance between
individuals, populations, and lineages matches the classic expectation
of a “Case IV” scenario resulting from a lack of regional equilibrium
(Hutchison and Templeton 1999). This pattern is typically driven by
differing scale-dependent effects of gene flow and drift influencing
population structure (Twyford et al. 2020), where migration has greater
impacts over shorter distances and drift predominates past a threshold
determined in part by habitat connectivity and migration range (van
Strien et al. 2015). The lack of between-lineage IBD likely reflects the
long-term impact of mountains and refugia in structuring local
populations, explaining the presence of very close yet very different
populations in the Blue Ridge. This also rejects IBA, as the strength of
between-population IBD is less than within-population, indicating no
acceleration of genetic divergence due to local adaptation in the Blue
Ridge and Piedmont (Freedman et al. 2023). Correspondingly, the
relatively recent expansion of the Piedmont lineages during the late
Pleistocene explains the weakness of IBD within the most widely
distributed lineage, as “Case I” dynamics (i.e., “pure” IBD) require
relatively long timescales to become apparent. This again underscores
the dynamic nature of these processes, and the attendant capacity of
related patterns to shift over time.
The RDA results confirm the hypothesis that there is some apparent
degree of ecological adaptation related to temperature and precipitation
differences between montane and Piedmont populations (Miranda et al.
2023), but they are not abundant in the genome. Furthermore, there is a
non-zero but limited degree of correlated adaptive divergence along
ecological gradients and related phenotypic axes. Robustness in terms of
size, limb length, and head shape (length and width) is likely related
to both desiccation tolerance in warmer environments (Baken et al. 2020)
and differing relative pressures of terrestriality (Ledbetter and Bonett
2019). That these variables show some degree of correlated adaptive
differentiation between montane and Piedmont environments is not
surprising, but it does not result in overall phenotypic divergence
between lineages or apparently act to limit hybridization and gene flow.
We consider our limited number of predictors in both analyses to offer a
tradeoff between potential false positives and false negatives (Forester
et al. 2018). Our sampling is not detailed enough to perform
cluster-specific analyses (Carvalho et al. 2021), but a better-annotated
molecular dataset could facilitate this in the future to detect
microgeographic variation and adaptation.
The SDMs along with other natural-history observations (Harrison 1992;
Graham et al. 2012) reinforce the strong influence of Level IV
Ecoregions on the distribution of this species, particularly outside of
the Blue Ridge Mountains. Similar results were seen in the approximately
co-distributed congener Desmognathus cheaha (Pyron et al. 2023).
We interpret this to indicate floristic or other associations (edaphic,
hydrological, etc.) with seepage environments not captured by climatic
variation and liable to change dramatically over ecological timescales
(A. Lee-Yaw et al. 2022). Our fieldwork underscores the extreme
microhabitat specificity involved in locating this species, but we have
not yet pinpointed the exact determinants of what constitutes suitable
habitat. For this reason, it is difficult to test hypotheses about
historical distributions with paleoclimatic modeling given the lack of a
mechanistic model and historical data layers that incorporate the
relevant factors, which are still mostly unknown. Similarly, we did not
consider modern estimated effective migration surfaces (Petkova et al.
2016) to be particularly relevant for investigating these historical
processes, which shift rapidly through space and time.
Consequently, we hypothesize that structure in this species arises
primarily from stabilizing rather than diversifying ecomorphological
selection. This results from a high degree of ecological specificity to
a highly specific microhabitat consisting of headwater seepages, moist
leaf litter, and various moss species. These salamanders are almost
never observed outside of this exact ecosystem, typically within the
range of a few meters around spring heads or ravine streams. Yet,
increases in suitable habitat during periods of cooling and expansion
out of refugia appear to drive rapid geographic occupancy across a large
expanse of the Piedmont and associated ecoregions, which is quickly
fragmented during interglacial periods. This adaptive stasis and limited
dispersal both drives IBD in geographically proximate areas but
contributes to lineage cohesion across climatic cycles when these
locally adapted lineages are brought into contact out of climatically
proximate refugia during glacial cycles. Similar processes appear to be
operating in other salamander systems in Mexico (Velo-Antón et al.
2013), suggesting analogous montane processes linking tropical and
temperate dynamics in landscape genetics and lineage formation. This is
essentially an extension of Janzen’s hypothesis (Janzen 1967; Muñoz and
Bodensteiner 2019; Wishingrad and Thomson 2023), wherein behavior,
ecology, and phenotype interact to drive local adaptation, constrain
ecomorphological divergence, and promote lineage cohesion. Interestingly
in this case, these processes ultimately appear to foster
phylogeographic diversification while constraining speciation
trajectories.