Gharrett et al Phylogeographic analysis of mitochondrial DNA variation in Oncorhynchus kisutch 
535 
Kamchatka River 
Crooked Creek 
Buskin River 
Little Susitna 
Berners River 
Hugh Smith 
Indian River 
Ford Arm 
Fish Creek 
Kanektok River 
Karluk River 
Eek River 
Delta Clearwater 
► Kamchatka River (R) 
► Crooked Creek (SC) 
► Buskin River (Kl) 
► Little Susitna (SC) 
► Berners River (SE) 
► Hugh Smith (SE) 
Indian River (SE) 
Ford Arm (SE) 
► Fish Creek (SE) 
► Kanektok River (BS) 
► Karluk River (Kl) 
► Eek River (BS) 
-► Yukon-Delta Clearwater (BS) 
T 
Average nucleotide divergence 
(substitutions per thousand) 
Figure 4 
A Fitch-Margoliash least-squares phenogram constructed from nucleotide divergences among coho 
salmon collections and a consensus tree (inset) derived from 2000 iterations (Hedges, 1992) of boot- 
strap resampling from the individuals within each collection. Geographic regions for the samples are 
Russia (R), Southcentral Alaska (SC), Kodiak Island (Kl), and Southeast Alaska (SE), and Bering 
Sea (BS), The consensus tree is unsealed and describes the confidence of the topology depicted in the 
phenogram. 
Discussion 
Coho salmon mtDNA variation 
We examined the coho salmon mtDNA genome using 
restriction analysis of seven PCR products and detected 
fragments as small as 25 bp, which made it possible to 
sample nearly all (over 97%). of the mitochondrial genome. 
We are unaware of other fish species that have been 
as thoroughly surveyed. Our estimates of species-level 
parameters of molecular evolution should be quite robust 
in comparisons with those for other taxa, including fish. 
Our results (Table 2) suggest that mtDNA variation is 
not evenly distributed throughout the genome and that fo- 
cusing analysis on variable sites overestimates nucleotide 
diversity estimates (see Tables 3 and 4). Many previous 
studies of Pacific salmon species have limited the portion 
of the mtDNA genome surveyed (e.g. chum salmon [Cro- 
nin et al.; 1993; Park et al., 1993; Seeb and Crane, 1999]; 
sockeye salmon [Bickham et al., 1995; Taylor et al. 1996]; 
chinook [Cronin et al., 1993]). Other studies have included 
a limited geographic range of samples and have restricted 
the portion of the mtDNA genome studied (e.g. sockeye 
[Burger et al., 1997]; chinook [Adams et al., 1994]) or sur- 
veyed selected variable sites (e.g. pink salmon [Brykov et 
al., 1996; Seeb et al., 1999]; chum salmon [Scribner et al., 
1998]; sockeye [Taylor et al., 1997]). Consequently, no pre- 
vious studies have produced results that are appropriate 
for a broad comparison with our residts. 
We evaluated our results in the context of other fish 
species by plotting the effective number of haplotypes ob- 
served (the number of haplotypes which, if equally abun- 
dant, would result in the observed haplotype diversity, n c , 
against the average nucleotide divergences between hap- 
lotypes for piscine species. For these comparisons, we used 
only data that were derived from 20 or more individuals 
and that surveyed the entire mtDNA genome (Fig. 5). On- 
ly two studies of Pacific salmon met those criteria; and our 
estimates of nucleotide diversity (0.4-4. 5 per 1000 nucle- 
otides) were generally lower than estimates for chinook 
salmon (6 haplotypes: 1.3-8. 1; Wilson et al., 1987) and 
chum salmon (2 haplotypes: 2.4; Thomas et al., 1986). 
Average nucleotide divergences between coho salmon 
haplotypes are quite low in relation to those of most oth- 
er species studied; whereas, the effective number of hap- 
lotypes is near the median. The effective number of (se- 
lectively neutral) haplotypes (n c ) is monotonically related 
to the product of effective population number (of females 
