Relevant Patterns of Population Structure in Ambystoma macrodactylum in Space and Time

8.10.1 Regional Patterns of Range Expansion and Isolation

Seven glacial refugia resulting from glacial fragmentation are qualitatively identified from shallow history mtDNA divergence and are accentuated by haplotype distributions emanating from source areas along the coast (1) and interior (2). Refugia were identified (Thompson et al. unpublished) as: 1a) Haida Gwaii; 1b) Coastal/Cascade Mountains; 1c) Santa Cruz and Monterey Counties, California; 2a) Blue Mountains; 2b) Clearwater Mountains; 2c) Montana's Rocky Mountains; 2d) Salmon River Mountains. We use NCPA to describe the genealogy and spatial extent of post-glacial migration from these places for breeding salamander populations (Fig. 6).

Fig. 6. The Cordilleran Ice sheet (dark grey shading) forced populations into seven isolated refugia during the Pleistocene; these are identified by the stippled regions and associated numbers

During periods of glacial advance sea levels were lower than at present, and isostatic rebounding molded physical geography and shifted hydrological patterns (Pielou 1991; Clague and James 2002). In this dynamic setting multiple Pleistocene refugia occurred in the Pacific Northwest, but the Haida Gwaii refuge is most controversial (Warner and Mathewes 1982; O'Reilly et al. 1993; Byun et al. 1997; Soltis et al. 1997; Demboski et al. 1999; Stone and Cook 2000; Brunsfeld et al. 2001). The Haida Gwaii refugium is an ice-free terrestrial and freshwater environment that existed along Hecate Strait and the continental shelf, and extended to the northern parts of Vancouver Island. A precise size, time and type of ecological system that was supported are all unclear (Warner and Mathewes 1982; Josenhans et al. 1995; Mandryk et al. 2001). Coastal clades A-5 (Fig. 7a) and D-1 (Fig. 7b) suggest that the region harbored salamanders; Ambystoma gracile exhibits a similar north-south division along British Columbia's coastal mountains (Titus and Gains 1991). If A. macrodactylum or A. gracile persisted in this refugium during glaciation, then their descendants emigrated after the glaciers retreated to establish populations that contain the genetic signatures of this history. The possibility of a Haida Gwaii refugium raises an interesting biogeographic question, because it is well established that marine situations effectively isolate amphibians (Darlington 1957; Duellman and Trueb 1986; but see Anderson 1960), yet the Haida Gwaii clade, A-5 (Fig. 7b), spans Georgia

Strait and the continental Interior (clade A-5 overlaps with the ranges of A-4 and C-l. Fig. 7b).

Fig. 7. Nested clade designs from Fig. 5 are spatially overlain onto a map for (a) the cytochrome b genealogy and (b) the ISR genealogy. Numbers in small circles and nesting levels correspond to those in the network design (Fig. 5). An inset of Vancouver Island, in (a), isolates clade A-5 from others for visual clarity. For reasons of simplicity, only the nesting categories that have significant range differences are indicated, and private haplotypes (haplotypes that occur in only one population, sensu Hartl and Clark 1989) are also exempt in this regard. Several populations are labeled in the post-glacial regions of Alberta and British Columbia (circles containing letters a--c), which contain a mixture of disparate clades

Fig. 7. Nested clade designs from Fig. 5 are spatially overlain onto a map for (a) the cytochrome b genealogy and (b) the ISR genealogy. Numbers in small circles and nesting levels correspond to those in the network design (Fig. 5). An inset of Vancouver Island, in (a), isolates clade A-5 from others for visual clarity. For reasons of simplicity, only the nesting categories that have significant range differences are indicated, and private haplotypes (haplotypes that occur in only one population, sensu Hartl and Clark 1989) are also exempt in this regard. Several populations are labeled in the post-glacial regions of Alberta and British Columbia (circles containing letters a--c), which contain a mixture of disparate clades

The occurrence of identical haplotypes throughout these populations (H8 in cytochrome b, clade A-5, Figs. 5a, 7a) resulted from Holocene descendants that migrated into respective areas either via a terrestrial land bridge (Mandryk et al. 2001) or by enduring a marine journey; the latter hypothesis is conditional upon salinity levels that would need to have been sufficiently lowered by the influx of glacial meltwater to permit waif dispersal (Anderson 1960; Walker and Pellatt 2003).Deep genealogical sub-divisions are not expressed in areas further south of the glacial margins, unlike the situation for several other species. For example, there is a close genetic relationship between populations spanning the Klamath-Siskiyou biogeographic boundary (Bury and Pearl 1999), in contrast to the genealogical subdivisions observed for the arionid slug, Prophysaon coeruleum (Wilke and Duncan 2004). This suggests that, under certain conditions, salamanders may migrate more readily than certain other species. While California's populations are currently isolated (Baily 1948; Russell and Anderson 1956), the measured divergence between populations that cross this biogeographic boundary reflect recent genealogical connections. Subsequent climate change modified the intervening environments to isolate these populations (Thompson et al. unpublished).

Another salamander species, Taricha torosa, migrated southward across the Klamath-Siskiyou boundary and along the Sierra Nevada to finally arrive in California during the middle Miocene (Tan and Wake 1995). The Miocene, however, is too distant in time to account for the small degree of divergence measured between Ambystoma macrodactylum populations. Sierra Nevada's populations are geographically nearest, and genetically and phenotypically most similar to those of California (Dr. D. Wake personal communication; Dr. R. Sage unpublished). Hence, the phylogeographic pattern of A. macrodactylum matches those of Batrachoseps attenuatus and Ensatina eschscholtzii xanthoptica, and is hypothesized to have colonized Monterey Bay and Santa Cruz, California during a Pleistocene range expansion from the Sierra Nevada through the Central Valley (Stebbins 1949; Moritz et al. 1992; Jockusch and Wake 2002).

The interior mountain chains have been sampled more intensively and provide greater resolution about migration and endemic history. Oregon's clade, B-5 (Fig. 7b), occurs within a transverse mountain system that is identified as a refugium. This pattern represents an otherwise cryptic contributor to the biodiversity that is endemic to the dry central interior, and is corroborated by the distinctiveness of this clade. To the east of these mountain chains the Snake River is a barrier where it runs through Hell's Canyon along Idaho's western border (Baker 1983; Alt and Hyndman 1995). Several species, in addition to the long-toed salamander, point to Oregon's transverse mountain system as an important genetic refugium. In the eastern section of these mountains, the tailed frog, Ascaphus truei, has a distinctive genetic signature, and a spatial distribution that is comparable to that of Ambystoma macrodactylum; both species closely match the interior coniferous forest regions (Heusser 1983; Pauken and Metter 1971;

Nielson et al. 2001; Thompson et al. unpublished). The red-tailed chipmunk (Tamius ruficaudus) has a divergent genetic clade situated adjacent to Idaho's western border and is isolated by Palouse prairie in Washington (Good and Sullivan 2001). Flycatchers (Epidonax hammondii) similarly exhibit a unique pattern of allelic diversity within the Wallawa Mountains (Johnson and Marten 1989). Several factors contribute to the genetic isolation of populations to these mountains. First, Pleistocene environments were colder and drier in Washington (Heusser 1983; Walker and Pellatt 2003), making the conditions even more inhospitable and range marginal, and reducing migration rates. Secondly, meltwater and river flow rates were possibly severe enough to isolate populations during the Pleistocene, as glacial Lake Missoula released floods into the Snake and Columbia Rivers (Alt and Hyndman 1995).

The Clearwater refugium and the crest of the Bitterroot Mountains are two prominent features that have been implicated in the phylogeographic diversity of the interior (Brunsfeld et al. 2001). The Clearwater valley is deep enough to impose unique thermal conditions and has been implicated in explaining the endemism characteristic of this area (Daubenmire 1975; Brunsfeld et al. 2001; Good and Sullivan 2001). Ferguson (1961) suspected secondary contact, or a zone of morphological intergradation between Ambystoma macrodactylum columbianum and A. m. krausei in central Idaho that implicates the Clearwater populations and a crossing of the Bitteroot Mountains, but secondary contact has not been identified for populations in this area. This has resulted in a revision of the subspecies map (Fig. 2).

Lineages in Montana's Rocky Mountains, however, are mixed with lineages that came from the Salmon River Canyon refugium and migrated through the Bitterroot Mountains. The Salmon River Canyon houses the closest relatives of the coastal lineages, and a large proportion of the genetic diversity evolved in this area. Given the unique nature of this clade and that secondary contact in the Montana's Rocky Mountains has occurred, the mountain topography of the Salmon River Canyon may have only recently changed to permit a connection into Montana, but was otherwise isolated by the Bitteroot Mountains throughout the Pleistocene. This area developed throughout the Pleistocene, with the highest fluvial terraces of the Middle Fork dated at 0.4--1.1 mya (Meyer and Leidecker 1999). Salmon River Mountain's clade A-7 (Fig. 7a) exhibits a significant pattern that is reflective of restricted gene flow and isolation by distance (Table 1a), but this inference is peculiar in the context of qualitative and glacial biogeography. One population of this clade occurs in northern postglacial British Columbia, while a second is found along the coast in the Puget Trough (Fig. 7a), both of which are distant from the core of this clade. Because Bonneville floods, that ran westward through the Snake River Canyon during the late Pleistocene (15 ka), had incredible discharge through restricted valleys, waif dispersal by way of Pleistocene fluvial dynamics (Baker 1983; Waitt and Thorson 1983; Jarrett and Malde 1987) provides an explanation for a member of this clade being found along the coast. This Bonneville dispersal hypothesis was also proposed by Taylor (1985) in relation to contemporary and fossil mollusk distributions.

The Cordillera was colonized by migrants originating from the Clearwater refugium, Salmon River Mountains, and the Rocky Mountains of Montana (Thompson et al. unpublished) and likely traced along the interior route identified by Conroy and Cook (2000). Populations emigrating from the Clearwater refugium, however, contain a mixture of haplotypes nested below the A-step and may indicate a different type of genetic expansion coming from a larger and more stable population.

8.10.2 Molecular Demographics

Salamander populations that existed proximally south of the ice maximum were likely affected by peri-glacial climates and associated hydrology (Waitt and Thorson 1983; Thompson et al. unpublished). The demographic pattern evolved as a result of the ebb and flow of climate and paleogeography, because these factors influenced the associated migratory connections and carrying capacity of these populations (Knowles 2001). We observed that gene pools originating from peri-glacial environments expanded their ranges northward with the migration routes tracing through the same valleys that the Cordilleran ice sheet flowed through (Thompson et al. unpublished). Although we have an apriori expectation of range expansion for these lineages, the NCPA analysis did not necessarily infer this type of explanation a posteriori.

Inferences drawn from Templeton et al.'s (1995) NCPA require a combination of alternate methods for further testing and investigation of alternative causal factors (Masta et al. 2003; Morando et al. 2004). For example, the NCPA spatial statistics for the cytochrome b data set were unable to generate an inference of sudden range-expansion for the lineages emanating from the Salmon River refugium (Thompson et al. unpublished) because two samples of a haplotype in clade A-7 (Fig. 7a) are widely displaced from each other in directions away from the clade center; hence, the effective distance covered by this haplotype, although large, skews the center of the clades' distribution north-westward, and decreases the measured degree of spatial spread of the entire clade. The NCPA approach cannot infer range expansion in this particular instance, because the event is "... older than the coalescence time for the gene region being investigated" (Templeton et al. 1995: 772). In general, we could not test whether range expansion was responsible for genetic patterns for lineages that obviously expanded their ranges and transgressed into the post-glacial Cordillera. Therefore, we compared these results against the inferential capabilities of the mismatch distribution.

A mismatch distribution is the probability (Fi) that two neutral sequences, randomly selected from a population, have accumulated mutations at i nucleotide sites (Rogers and Harpending 1992; Excoffier 2004). The coalescent parameter 6 is used to calculate Fi, and is estimated by the observed mean of pairwise genetic differences (Avise et al. 1988; Rogers and Harpending 1992). If a population remains constant in size over successive generations (t), or experiences a temporary reduction in population size (i.e., a bottleneck), then Fi(t) quickly approaches an equilibrium state. The plotted distribution of the number of mutation differences versus Fi(t) has a smooth, L-shaped decline for populations that are in equilibrium. Most empirically observed mismatch distributions, however, do not decline so smoothly, but show waves in the distribution. Single modal waves are produced when populations are simulated to change greatly in effective size from an equilibrium state; this is called the sudden expansion model (Schneider et al. 2000). Waves occur in the distribution because genetic distances tend to increase as population size increases, but after a sudden expansion ". the mean pairwise difference increases much more rapidly than its standard deviation" (Rogers and Harpending 1992: 556). Hence, waves in the mismatch distribution provide a statistical approach for identifying historical instances of increased population size.

Mismatch distributions are commonly employed to identify global migration events that occurred after the Pleistocene, because gene flow increases with migration and the relative degree of connectivity among isolated or restricted populations. The genetics of once restricted populations changing into a state of panmixia is similar to the sudden expansion model because the effective population size increases, but the geographic pattern is affected differently. The geography of genealogical branching depends on the rate of migration (Rogers 1995; Branco et al. 2002; Ray et al. 2003; Morando et al. 2004; Thompson et al. unpublished). Mismatch distributions are ideally limited, however, to species that have high levels of interpopulation gene flow (Avise 2000; Excoffier 2004). Based on autosomal alleles, Ambystoma macrodactylum has a small effective size (Ne ~ 100) and exchanges less than one individual per generation among basins in the Bitterroot Mountains of eastern Idaho and western Montana (Funk et al. 1999; Tallmon et al. 2000). These demographic estimates might be even smaller for the mitochondria genome, because it has one-quarter the effective size of autosomal genes, and the females may be less likely to migrate (Sheppard 1977; Moore 1995; Thompson et al. unpublished). However, there are circumstances in vertebrates under which mitochondrially mediated gene flow is more fluid than chromosomal material of the heterogametic (XY : Male) sex (Anderson 2004). During times of glacial retreat, meltwater rendered hydrological connections more dynamic and could have increased the effective migration rate through waif dispersal along river valleys.

Given that the crest of uni-modal waves in the mismatch distribution is centered on the time of expansion x, and that populations return to an equilibrium after a period of reduced size (Rogers and Harpending 1992; Excoffier 2004), we expect modes to center around 14 ka, the time of abrupt climate change and glacial retreat (Clague and James 2002; Walker and Pellatt 2003). Demographic parameters, Fu's (1997) Fs test, Tajima's (1989) neutral model D-statistic (10,000 simulations), and mismatch distributions (1000 simulations of the sudden expansion model) were calculated in ARLEQUIN ver. 2.000 (Schneider et al. 2000). For these calculations, we selected a subset of the ISR and cytochrome b clades inferred by the NCPA to have experienced a significant period of isolation and/or range expansion. We also investigated lineages that are inferred by the biogeographic interpretation to have expanded into the northern Cordillera after the last glacial retreat.

Multi-modal and uni-modal waves occur in the mismatch distributions (Fig. 8), but multi-modal patterns are more prevalent in the deeper clades. None of the clades deviate significantly from Tajima's (1989) neutral model, but three clades are significant for Fu's (1997) Fs test (Table 2). Both of these tests give significant results for factors other than selection, such as population expansion, bottlenecks, and mutation rate heterogeneity among nucleotides. Tajima's (1989) test for genetic neutrality is overly conservative if effective population size has shifted or if the mutation rate is heterogeneous over the segment of DNA being investigated. This statistic does, however, give negative D values for models of sudden expansion (Tajima 1989; Aris-Brosou and Excoffier 1996; Schneider et al. 2000). It is important to compare these statistics for the two markers under study, because one may reveal information that the other does not. For example, according to the Log Likelihood scores of MODELTEST (Posada and Crandall 1998; a program that identifies the models that statistically fit each data set) the cytochrome b sequences exhibit a uniform mutation pattern, but the mutations of the ISR marker make Tajima's (1989) test unreliable because they vary considerably among sites (gamma shape parameter = 0.0891). Fu's Fs is thought more likely to show significance if range expansion occurred (Ford 2002; Morando et al. 2004).

Only the Salmon River Mountain clade, B-9 (Fig. 8b), is significant for Fs (Table 2b), expanded its range according to the NCPA spatial statistic inferences (Table 1b), and shows a modal wave in the mismatch distribution. Moreover, the mismatch Q1 value is much larger than that for other clades, and the model of sudden expansion is not rejected by the simulated data (Table 2b). Salmon River Mountain's clade A-6, identified by the cytochrome b marker, is directly comparable because the haplotypes belonging to it are physically linked to the haplotypes in the Salmon River Mountain's clade B-9, identified by the ISR marker (Fig. 8a). For this clade, Tajima's D value is positive, but it similarly has a mismatch Q1 value much larger than that of others, and a raggedness index that is consistent with the demographic-expansion model (Table 2a). However, does clade A-6 of the Salmon River Mountains (Fig. 8a) clearly exhibit a modal wave? This particular clade is unique because it is observed outside of the Salmon River Mountains (Fig. 7a) and it is genealogically distinct within this race.

These observations confirm our suspicions that the Salmon River Mountains are sampled insufficiently. Clade D-2 from the Clearwater Mountains (Fig. 8b) exhibits a modal wave, has a negative Tajima's D value, and fits a model of sudden expansion of considerable size (Table 2b). Moreover, Fu's Fs statistic is non-significant, as is expected of the ISR, a neutral genetic marker. These three clades are likely candidates in which a historical event of sudden expansion is genetically perceptible. The NCPA analysis did not infer a significant pattern for the Clearwater Mountain clade D-2, but agrees with the interpretation of expansion for the Salmon River Mountain clades. Continuous range expansion rather than long-distance colonization is inferred by the NCPA analysis because the Dn and/or I-t Dn values are not significantly reversed from the Dc values (Templeton 2004; Thompson et al. unpublished). Clade D-4, also from the Salmon River Mountains, has some characteristics of the shallower clades, because it has a modal wave in the distribution (Fig. 8b), but the stepwise expansion model is rejected by the sum of squared deviations test statistic (Table 2b). The recent effect of post-Pleistocene range dynamics, however, is noticeable in the consistent peak at low pairwise differences (Fig. 8). The estimated points of these peaks are identified by the unit of mutation time (t = 2ut) for the mismatch distributions that are not significantly rejected for the model of sudden expansion (Table 2).

Fig. 8. Mitochondrial mismatch distribution showing the number of nucleotide site differences for phylogeographically significant clades plotted for the (a) ISR and (b) cytochrome b genealogies. The solid lines with circles represent observed values, the broken lines with squares are values obtained by coalescent simulations, and the dots with triangles are values calculated according to a model of sudden expansion (see Rogers 1995, Schneider et al. 2000, Ray et al. 2003, and Excoffier 2004 for further details). The simulated and model distributions are closely matched in each plot. Multi-modal plots tend to occur at the deeper clade levels, and uni-modal waves tend to occur at the lower clade levels

Fig. 8. Mitochondrial mismatch distribution showing the number of nucleotide site differences for phylogeographically significant clades plotted for the (a) ISR and (b) cytochrome b genealogies. The solid lines with circles represent observed values, the broken lines with squares are values obtained by coalescent simulations, and the dots with triangles are values calculated according to a model of sudden expansion (see Rogers 1995, Schneider et al. 2000, Ray et al. 2003, and Excoffier 2004 for further details). The simulated and model distributions are closely matched in each plot. Multi-modal plots tend to occur at the deeper clade levels, and uni-modal waves tend to occur at the lower clade levels

Table 2. Summary statistics of the evolutionary demographics for the (a) cytochrome b and (b) ISR markers as obtained in ARLEQUIN (Schneider et al. 2000). Values of significance are included for Fu's (1997) Fs and Tajima's (1989) neutral model D statistics, plus modeled and simulated demographic parameters from the mismatch distribution plots o o o

CD O

Table 2. Summary statistics of the evolutionary demographics for the (a) cytochrome b and (b) ISR markers as obtained in ARLEQUIN (Schneider et al. 2000). Values of significance are included for Fu's (1997) Fs and Tajima's (1989) neutral model D statistics, plus modeled and simulated demographic parameters from the mismatch distribution plots

O

CM

01

m

oo

o

CD

m

O)

CM

CM

t—

t—

o

CO

CM

CO

03

t—

CD

CD

o

0

o

o

o

O

o

o

O

o

O)

oo

oo

oo

o

O

CM

m

o

m

o

CM

CM

o

m

O

01

t—

01

01

01

m

U1

oo

cm

CM

m

o

o

o

o

o

o

o

o

o

o

CD

* 00

oo

CD

01

CM

m

CO

o

CD

<v

CD

oo

t—

o

m

m

CM

CM

o

o

t—

cm

o

01

01

o

0

o

o

o

o

o

o

o

o

®

oil o

o

CM

CM

.0 0.

0.

o ®

o o

o o o

O

o o o

4 4.

0 0 0.

0 0 0.

3 5 .0

0 0 0.

0 0 0.

0 0 0.

o

o

O

o

o

0.

0.

0.

0.

0.

0.

a>

CO CO

O

5 5 .4

8 8 4.

o o o

0 8

CM

0 0 0.

8 6 .7

9 7 .5

4 2 .3

0)

O

5.

7.

3.

1.

1.

3.

0.

0.

0.

o

i-

1

.Q

oo oo

7 7.

CM CD

3 .9

6 9 3.

0 3 8.

7 6

00 CO O

5 5 6.

8 8 2.

CO

o

o

O

o

o

o

0.

0.

O

0.

0.

0.

> o ra

Fs

3.537

71 .3

CM CM

.9

.4

8 .8

9 4 .3

9 2 .4

7 4 .2

0 9 .0

01 .2

1.

3.

CM

o 1

i

i

i

1.

0.

0.

5 ®

6.781

4.354

1.915

3.464

1.067

1.000

0.142

0.051

0.500

0.400

0.667

Q

CO CO

CM

5

7 4 .3

.8

0 .71

8 2 .7

6 2

9 0 3.

7 .81

0 0 .0

o

o i

1.

1.

i

1

1.

0.

OT ®

.7

.31

.7

0 .5

91 .0

9 5 .2

7 3 2.

01 .3

0 8 4.

7 6 .6

5.

4.

1.

CM

1.

1.

0.

0.

0.

0.

0.

z

O)

4

5 5

8

31

4

7 2

9 3

6

5

3

TD lü

5 m CO

5 o

5

5 g o

5

5 01 CO

5 01 CO

5 O z

5

5

T-

ciJ

(N

(N

T-

to

T

T-

1-

uj

a

Ö

Ö

CO

<

<

<

<t

CO CO

CD CD

CO CD CO

CN CN CO

CN CN m CO

CO CN CO

"D J5

LV m

¿Ooooo

ro ii

■D

CQ

C

--

Q)

(A

Mo

U)

O

O

(D C

1 1 A

S^-D

<D

(D TO

W

CQ

LO

II

Ï

ra o

co

.£= CD

Û_

■D

"D V

C CX

Q)

J2

(D frt:

j S> TO ® ¡= ® Í ííi (A o E ¿ o i « 11 m f l I I

0 0

Post a comment