Sme et al.: Identification of Eleginus gracilis by means of microsatellite markers 
65 
♦ Chukchi SC BGulf SC ■ Nawaga •tomcod 
▲ Pacific cod ■ pollock • Arctic cod 
Figure 1 
Results of principle component (PC) analyses. (A) Allele composition (a 
correlation matrix) and (B) allele frequency profiles (a covariance ma¬ 
trix) of microsatellite data from saffron cod ( Eleginus gracilis [SC]) col¬ 
lected in the Chukchi Sea (Chukchi SC) and Gulf of Alaska (Gulf SC) in 
2011 and 2013, navaga [E. nawaga ) collected in the Barents Sea in 2013, 
Pacific tomcod ( Microgadus proximus) collected in Puget Sound during 
1997-1999 and in Prince William Sound in 2012, Pacific cod ( Gadus mac- 
rocephalus) collected in Puget Sound and Unimak Pass in 2013, walleye 
pollock (G. chalcogrammus ) collected in the southeastern Bering Sea in 
2015, and Arctic cod iBoreogadus saida) collected in the Chukchi Sea in 
2012. The symbols '+’ and ‘x’ denote individuals provided in saffron cod 
collections that were later re-identified as Arctic cod and Pacific tomcod, 
respectively. 
all of which were significant (adjusted 
pairwise homogeneity tests P<0.0001). 
The estimate of G'g T between the two 
E. gracilis collections was smaller 
than values between all other gadid 
pairs; whereas the estimate of jD chord 
was smaller than that of all but three 
of the comparisons of gadids, even 
though different suites of microsatel¬ 
lite loci were used. To provide a com¬ 
parison of the extent of divergence be¬ 
tween the two E. gracilis collections, 
values of G' S t and -D c hord were estimat¬ 
ed for the species pair Sebastes aleu- 
tianus and S. melanostictus from data 
in Gharrett et al. (2005), G'st=0.551 
and D chord =:0.064. The estimate of G' S t 
between the E. gracilis pair was lower 
(0.313) but the estimate of jD chord was 
higher (0.078) than that between S. 
aleutianus and S. melanostictus , pre¬ 
sumably because different algorithms 
were applied; D chord has a geometric 
basis and G'gT is based on ratios of 
heterozygosities adjusted to account 
for the amount of genetic variation ob¬ 
served at each locus (Hedrick, 2005). 
Individual-based PCA of allelic 
compositions (a correlation matrix) 
and allele frequency profiles (a cova¬ 
riance matrix) produced both species- 
and collection-specific clusters (Fig. 1). 
The plot of the first and second com¬ 
ponents of the correlation-based PCA 
separated individual species more 
clearly, but separation of the two E. 
gracilis collections was not as strong. 
The covariance-based PCA clearly sep¬ 
arated the two E. gracilis collections, 
but the other species were not separat¬ 
ed quite as well. The first five compo¬ 
nents of the correlation-based analysis 
accounted for 10.6% and the first two 
components accounted for 5.1% of the 
overall variation in allelic composition. 
In contrast, the first five components 
of the covariance-based PCA accounted 
for 24.3% and the first two for 14.1% 
of the overall variation in allelic fre¬ 
quencies. Nevertheless, sufficient vari¬ 
ation existed to separate these species 
and the two collections of E. gracilis. 
A series of 4 tests was needed to es¬ 
timate assignments of individuals because not all loci 
could be used for all species groups (Suppl. Table 2) 
(online only). The tests were the following: 1) all individu¬ 
als were assigned on the basis of the three loci all 
groups had in common—Elgrl4, Elgr23, and Elgr31; 2) 
the individuals scored in 1) as Chukchi Sea E. gracilis 
(CSC), GOA E. gracilis (GSC), E. nawaga (NAW), M. 
proximus (PTC), and G. macrocephalus (PCO) were as¬ 
signed on the basis of Elgr7, Elgrll, Elgrl3, Elgrl4, 
Elgr23, and Elgr31; 3) the individuals scored in 2) 
as CSC, GSC, or NAW were tested at Elgr7, Elgrll, 
Elgrl3, Elgrl4, Elgr23, Elgr31, Elgr44, and Elgr45; 
and 4) the individuals scored in 1) as PTC, PCO, G. 
chalcogrammus (WPO), or B. saida (ACO) were tested 
