Gharrett et al.: Phylogeographic analysis of mitochondrial DNA variation in Oncorhynchus kisutch 
531 
ments were separated by electrophoresis through a 1.5% 
agarose gel (a mixture composed of one part Ultra Pure™ 
agarose [BRL Gibco, Grand, NY] and two parts Synergel™ 
[Diversified Biotech Inc., Boston, MA]) in 0.5xTBE buffer 
(TBE is 90 mM Tris-boric acid, and 2 mM EDTA, pH 
7.5). DNA in the gel was stained with ethidium bromide 
and photographed on an ultraviolet light transillumina- 
tor. Digests that produced fragments too small for detec- 
tion in agarose/Synergel™ gels were resolved in 12% poly- 
acrylamide (29:1 acrylamide:bisacrylamide; lxTBE) gels. 
DNA fragments separated in polyacrylamide gels were 
stained with SYBR Green 1 Nucleic Acid Stain™ (Molec- 
ular probes, Eugene, OR), which is more sensitive than 
ethidium bromide. Either a 1-kilobase (kb) ladder or Hae 
III digested (f>x 174 RF phage DNA (BRL Gibco, Grand, NY) 
was used as a molecular weight reference for estimating 
restriction fragment sizes. 
Restriction sites were inferred from fragment patterns 
that could be related to each other by the gain or loss of 
a single site. Composite haplotypes were constructed from 
restriction fragment patterns of all restriction enzymes 
across all mtDNA PCR regions. Using the rules of Castel- 
loe and Templeton (1994) to resolve ambiguities, we con- 
structed the single most probable parsimonious tree de- 
picting restriction site changes between haplotypes. Using 
REAP (McElroy et al., 1990), we estimated haplotype (nu- 
cleon) and nucleotide diversities within populations (Nei, 
1987) as well as average nucleotide divergences between 
populations. Nucleotide divergence between populations 
takes into account both the haplotype frequency differ- 
ences between populations and the nucleotide divergences 
between haplotypes (Nei and Tajima, 1983; Nei, 1987; Nei 
and Miller, 1990). Homogeneity of haplotype diversities 
among populations was tested by using the Monte-Carlo 
simulation in REAP (McElroy et al., 1990X10,000 iter- 
ations; Hedges, 1992) to establish probability levels for 
goodness-of-fit statistics (Roff and Bentzen, 1989). 
Populations were clustered from pair-wise nucleotide di- 
vergences by using the Fitch and Margoliash (1967) least- 
squares method (FITCH in PHYLIP; Felsenstein, 1995). 
For comparisons between populations, the precision of es- 
timates of nucleotide divergence depends on sample size. 
Therefore, stability of the topology was examined by boot- 
strapping (2000 iterations; Hedges, 1992) over individuals 
within each collection. A consensus tree (CONSENSE in 
PHYLIP; Felsenstein, 1995) that shows the stability of the 
topology, but not the branch lengths, was generated from 
the set of bootstrapped trees. 
The hierarchical structure of the expanded set of coho 
samples was analyzed by analysis of molecular variance 
(AM OVA; Excoffier et al., 1992) with Arlequin (Schneider 
et al., 1997). Collections were grouped geographically into 
four regions: Southeast Alaska, Southcentral Alaska, Ber- 
ing Sea, and Asia. With appropriate choices of divergence 
matrices, the analysis can examine the structure from 
haplotype frequencies (e.g. Weir, 1996) or from nucleotide 
diversities based on paths between haplotypes traced 
through a haplotype tree (Excoffier et al., 1992). Signifi- 
cance (P MC ) of d>-statistics (Excoffier et al., 1992) was es- 
timated from distributions of the statistics generated by 
17,000 permutations (Hedges, 1992) at the appropriate 
level of hierarchy. 
Nested clade analysis of geographical distributions of 
haplotypes and subclades (e.g. Templeton and Sing, 1993; 
Castelloe and Templeton, 1994; Templeton, 1998) was con- 
ducted with GEODIS 2.0 (Posada et al., 2000). 
Results 
Our general approach was to conduct a broad preliminary 
survey to obtain genome-wide information for mtDNA 
restriction site variation. Subsequently, we focused on the 
variable restriction sites and examined larger sample sizes 
and additional populations. From those results we con- 
structed a fine-scale mtDNA “gene tree” and analyzed the 
geographic distributions of mtDNA lineages to deduce the 
nature of the historical demographic changes that influ- 
enced present-day population structure. The approach also 
allowed us to determine the effects on estimates of molec- 
ular parameters that occur when analyses focus on vari- 
able restriction sites. 
Diversities of coho salmon mtDNA 
Ten coho salmon from each of seven drainages (Fig. 1) 
were analyzed to survey broadly the species’ mtDNA vari- 
ability using 12 restriction endonucleases (Appendix 1). 
The total number of restriction sites inferred from restric- 
tion fragment patterns was 298 (an average of 291.28 
per haplotype), which corresponds to 1284 nucleotides (an 
average of 1254.80), or a maximum of 7.73% (an average 
of 7.56%) of the coho salmon mtDNA genome (Table 2). 
Sixteen sites (1.2% of the total) were variable. Individu- 
ally, the regions averaged between 29 and 57 restriction 
sites, which correspond to a maximum of between 5.58% 
and 9.57% of the nucleotides in a region. Although the 
amplified regions had some overlaps (Table 1), no vari- 
able sites were observed in the areas of overlap; and no 
invariant sites were shared between regions, except possi- 
bly Dde I sites in the 408-bp overlap between A8/COIII and 
ND3/ND4 (Table 1). Because of the large total number of 
sites examined, a few overlapping Dde I sites would cause 
only a slight decrease in nucleotide diversity estimates and 
have little effect on nucleotide divergence estimates. 
Restriction site variation was observed in five of the sev- 
en PCR-amplified mtDNA regions for the 12 restriction 
endonucleases used. Between zero and five variable sites 
were observed per region. No variation was detected in the 
A8/COIII and ND3/ND4 regions. The largest number of 
variable sites (5) and the highest level of nucleotide diver- 
sity (5.99 substitutions per 1000 base pairs) were observed 
in the ND5/ND6 region (Table 2). 
Because there is no recombination between heterolo- 
gous mtDNA molecules, the composite haplotype is the ap- 
propriate unit to consider in genetic analysis (e.g. Avise, 
1989). Our preliminary survey discovered 11 haplotypes 
(Table 3). As a whole, the sample of 70 fish had a haplotype 
diversity of 0.803 and a nucleotide diversity of 1.70 substi- 
tutions per 1000 base pairs. Haplotype diversities within 
