66 
Fishery Bulletin 11 6(1) 
Table 3 
Estimates of pairwise chord distances (.D chord ; above the diagonal) and standardized genetic differentiation measure 
(G'st. below the diagonal) for saffron cod (Eleginus gracilis) sampled in the Chukchi Sea and Gulf of Alaska (GOA) in 
2011 and 2013 and for navaga (E. nawaga), Pacific tomcod ( Microgadus proximus ), Pacific cod (Gadus macrocephalus ), 
walleye pollock (G. chalcogrammus ), and Arctic cod ( Boreogadus saida ) sampled in the Pacific Rim and Arctic Ocean 
during 1997-2015. All estimates were significant (adjusted probabilities: P<0.0001). Values of average unbiased ex¬ 
pected heterozygosity {H e ) are indicated in italic type on the diagonal. 
Collection 
A 
B 
C 
D 
E 
F 
G 
A 
Chukchi Sea E. gracilis 
0.859 
0.078 a 
0.076 a 
0.138 a 
0.189 b 
0.218 d 
0.076 d 
B 
GOA E. gracilis 
0.313 
0.689 
0.130 a 
0.183 b 
0.245 b 
0.296° 
0.095 e 
C 
E. nawaga 
0.414 
0.680 
0.863 
0.093 b 
0.137° 
0.158 e 
0.069° 
D 
M. proximus 
0.603 
0.779 
0.565 
0.733 
0.182 b 
0.228 d 
0.088° 
E 
G. macrocephalus 
0.877 
0.963 
0.822 
0.721 
0.633 
0.204 d 
0.092° 
F 
G. chalcogrammus 
0.868 
0.893 
0.739 
0.582 
0.449 
0.933 
0.087° 
G 
B. saida 
0.599 
0.680 
0.584 
0.781 
0.681 
0.607 
0.766 
a 8 loci; 
b 7 loci; 
c 6 loci; 
d 5 loci; 
e 4 loci for both Z) c h or d and G'st estimates. 
at Elgrl4, Elgr23, Elgr31, and Elgr38. The results of 
3) and 4) assigned each individual to its own group, 
except 1 CSC (96.7% of the total) and 1 AGO (98.1% of 
the total) (Table 4). 
Previous molecular studies have recognized G. mac¬ 
rocephalus, G. chalcogrammmus, and B. saida as dis¬ 
tinct species (Coulson et ah, 2006, Carr et ah, 1999) 
but the systematic relationships among E. gracilis, E. 
nawaga, and M. proximus are still unresolved (Meck¬ 
lenburg et al., 2016). Differences in the allele frequen¬ 
cy profiles are easier to see in plots that include only 
those four groups (Table 2, Suppl. Fig. 2) (online only). 
The M. proximus and E. nagawa distributions clear¬ 
ly differ from those of the 2 E. gracilis collections at 
Elgr07 and Elgrll. The profiles for M. proximus and 
E. nagawa clearly differ from those for the 2 collec¬ 
tions of E. gracilis at Elgr07 and Elgrll. M. proximus 
also differs at ElgrlS and Elgr31 and has a substan¬ 
tially higher number of large alleles. The numbers of 
observed alleles (Table 2) in the collection of GOA E. 
gracilis are relatively lower than those of the others 
and several are more abundant (Suppl. Fig. 2) (online 
only), which is consistent with the somewhat lower het¬ 
erozygosity (Table 2) of the GOA E. gracilis. 
Discussion 
Eight of the nine microsatellites that were evaluated 
for two collections of E. gracilis and that amplified re¬ 
liably were variable (heterozygosities 0.537 to 0.933) 
and had no apparent homozygote excess, indicating low 
null allele frequencies. The single exception, Elgr38, 
amplified reliably for the Chukchi Sea collection of E. 
gracilis but not for the GOA collection. At the other loci, 
the two collections had similar allele size ranges but 
differed substantially in allele frequencies (G' ST =0.313, 
•D c hord=0-078, P<0.0001). The observed differences were 
similar to those between two cryptic rockfish species 
that had overlapping ranges, S. aleutianus and S. 
melanostictus, although they were estimated with dif¬ 
ferent suites of loci. In the PCA plots, individuals from 
the two collections of E. gracilis were mostly distinct 
from each other, particularly in the analysis of the co- 
variance matrix, which focuses on the allele frequen¬ 
cies rather than allele composition. It is also notable 
that the PCA analyses included frequency differences 
of the other gadids analyzed, and differences between 
the 2 collections of E. gracilis were evident against the 
background variation from other species. 
Assignment tests placed all but one saffron cod in 
the group from which it originated. Not all nine mi¬ 
crosatellite loci amplified reliably in all of the other 
gadid species analyzed and some had an excess of ho¬ 
mozygotes, most likely as a consequence of null alleles; 
those loci were not used for assignment tests. Never¬ 
theless, where comparisons were possible, all the other 
gadids differed in microsatellite composition (PcO.OOOl) 
from both collections of E. gracilis and each other. The 
correlation matrix-based PCA, in particular, clustered 
individuals according to species or geographic groups of 
species. The PCA analyses turned out to be valuable in 
analyzing a large set of samples of putative E. graci¬ 
lis because the analysis revealed outliers that, when 
compared with the clusters of other gadids, enabled de¬ 
tection of individuals misidentified as E. gracilis. Two 
