Haplotype network, Structure, and TESS
Structure analysis of SNP data suggested two optimal values of K ,
indicating substructure in the data. ΔK supported K = 2
(Fig. S1), which grouped a broadly northern S. hesperis cluster
(specimens from British Columbia, Alberta, Montana, South Dakota,
California, and southeastern Colorado) with S. atlantis, and a
southern S. hesperis cluster (specimens from Utah, Idaho,
southern Montana, southwestern Colorado, New Mexico, and Arizona) withS. zerene (Fig. 2a). LnPr(K ) supported K = 6 and
ΔK additionally had a small peak at K = 6. This resolvedS. zerene, S. atlantis , and northern S. hesperis as
distinct clusters, and further separated southern S.
hesperis into three geographically-defined clusters: 1. “central”
(specimens from northern Utah, Idaho, southern Montana, and southwestern
Colorado); 2. southern Utah; and 3. New Mexico and Arizona. Two of the
northern New Mexico S. h. dorothea specimens and all the ArizonaS. h. nausicaa specimens appeared to be mixtures between the
southern New Mexico and central populations. Similarly, the southwestern
Colorado S. h. electa specimens were intermediate between the
central and southern Utah populations, and the Idaho S. h. violaspecimens were intermediate between the central and northern populations
(Fig. 2b). Substructure analysis of S. zerene indicated genetic
differences in S. zerene zerene from California that were
not shared by any S. zerene specimens sampled from Nevada, Utah,
Idaho, Montana and Alberta. Substructure analysis of the northernS. hesperis cluster did not indicate additional geographic
substructuring; both LnPr(K ) and ΔK indicated an optimalK of 2, however this did not produce any meaningful sub-structure
in the data that corresponded to sampling locality, and given that
ΔK cannot estimate K = 1, we suggest that K = 1 is
a more meaningful result.
TESS supported K = 5. This analysis was largely congruent
with the K = 6 Structure results, except that it lumped the
central and southern Utah populations into a single cluster (Fig. 2a).K = 6 had a similar DIC score to that of the K = 5 TESS
results, but failed to add any meaningful geographic substructuring
(Fig. S1), further supporting K = 5 as optimal for this analysis.
Both TESS and Structure indicated a few likely hybrids: one S.
hesperis sampled from southern Utah shared ancestry with the northern
New Mexico S. hesperis population, our single specimen ofS. hesperis sampled from California (putatively S. h.
irene ) was admixed with S. zerene , and one S. zerene from
California shared ancestry with the southern Utah S. hesperispopulation.
The minimum spanning haplotype network depicted distinct S.
atlantis and S. hesperis clusters, however there was very little
haplotype variation within either a priori species (Fig. 2c). In
all cases there were only one or two nucleotide differences between the
“distinct” specimens and the major haplotype group for each species.
For S. atlantis, this haplotype variation largely correlated to
sampling location - eastern S. atlantis canadensis (dos Passos,
1935) specimens sampled in Ontario and Quebec and the S. atlantis
sorocko Kondla & Spomer, 1998 specimens from Colorado were marginally
distinct from S. atlantis hollandi (Chermock & Chermock, 1940)
sampled in Alberta and Manitoba; one S. atlantis hollandi from
Alberta and one from Manitoba were minimally different from the major
“hollandi ” haplotype. This geographic pattern was not observed
in S. hesperis, and almost all the specimens sampled had
identical haplotypes regardless of sampling location. Comparatively,S. zerene had much more haplotype diversity in the minimum
spanning network, with three distinct haplogroups. One group consisted
of the S. zerene specimens sampled from Alberta, Idaho, Montana,
and Utah and was the most distinct from S. hesperis and S.
atlantis , and a second group containing S. zerene zerene from
California was intermediate between the S. hesperis and S.
atlantis haplogroups. Interestingly, the third S. zerenehaplotype that was found in S. zerene gunderi sampled from Nevada
was identical to the major S. hesperis haplotype (Fig. 2c), but
these individuals did not appear admixed with S. hesperis in the
Structure or TESS analyses of SNPs; the single likely S. zerenehybrid was instead from California, and had a haplotype consistent with
the rest of the Californian S. zerene specimens.