Smith et al.: A comparison of three genetic methods for stock discrimination of Hoplostethus atlanticus 
807 
frequencies of two primer fragments (A16-1, E19-3, 
Table 4). As with the allozyme data, there were dif- 
ferences between all pairwise comparisons, with the 
exception of those samples taken at Ritchie Bank and 
Box (Table 2). 
The mtDNA data also showed a significant het- 
erogeneity in the total data set but demonstrated less 
genetic differentiation than the allozyme and RAPD 
data sets, with only two pairwise comparisons show- 
ing a significant difference (Table 2). However all 
three methods, which have measured different parts 
of the genome, gave similar results of low genetic 
exchange among the four populations (Table 3). None 
of the estimates of N g m are true estimates because 
our data sets are biased in favor of polymorphic mark- 
ers, which will tend to inflate the G ST estimate; such 
estimates of N e m (Table 2) can be used to compare 
only relative levels of geneflow between areas (Fer- 
guson, 1994). In this respect there is 8-10 times as 
much gene flow between Ritchie Bank and Box than 
between these sites and Waitaki, when measured 
with allozymes and RAPD’s, and 15-38 times as 
much with mtDNA (Table 3). Significant genetic dif- 
ferences between spawning groups provides evidence 
of genetic isolation, and thus the data reveal three 
genetic groups: 1) Puysegur, 2) Waitaki, and 3) 
Ritchie Bank and Box (Table 2). 
There are problems with RAPD analyses that may 
preclude them from use as stock markers for orange 
roughy. Because RAPD markers are dominant, a 
number of assumptions have to be made to analyze 
the data (Lynch and Milligan, 1994). Some of these 
assumptions, in particular that fragments with the 
same electrophoretic mobility are genetically identi- 
cal and that absent fragments represent the same 
DNA fragment, may not be valid. Fragments that ex- 
hibited weak staining activity were not scored, so that 
there is a subjective element when scoring RAPD gels. 
In the absence of breeding studies, the allelic na- 
ture of presence or absence of RAPD fragments may 
be suspect. Garcia and Benzie (1995) reported an 
extra RAPD fragment in prawn larvae that was ab- 
sent in adults, although they found Mendelian in- 
heritance of other RAPD markers. Unlike the other 
two genetic methods, there are no internal checks 
that can be used to fit RAPD phenotypes to a genetic 
model: with allozymes there is an expected gel phe- 
notype for each enzyme and all alleles are equally 
expressed; with mtDNA the size of the restricted frag- 
ments should add up to the size of the undigested 
fragment. 
Some primers produced weak fragments that were 
not repeatable in reamplifications. These weak frag- 
ments may be produced by excessive PCR cycles; Bell 
and DeMarini (1991) have shown that by increasing 
Table 4 
Heterogeneity x 2 tests and gene diversity ( G ST ) for seven 
random amplified polymorphic DNA (RAPD) primers in 
four populations of orange roughy. df = degrees of freedom; 
P = probability value; G ST = gene diversity. (* = significant 
at Bonferroni-modified P for multiple tests). 
Primer and 
fragment 
Y 2 
(3 df) 
P 
G st 
(3 df) 
P 
A14-1 
2.820 
0.422 
0.010 
0.415 
A14-2 
3.306 
0.347 
0.012 
0.341 
A14-3 
9.089 
0.028 
0.032 
0.031 
A15-1 
1.543 
0.672 
0.006 
0.671 
A15-2 
8.189 
0.042 
0.029 
0.044 
A15-3 
3.593 
0.309 
0.013 
0.299 
A15-4 
2.138 
0.544 
0.008 
0.540 
A16-1 
16.921 
<0.001* 
0.079 
<0.001* 
A16-2 
1.473 
0.688 
0.005 
0.685 
A16-3 
6.874 
0.076 
0.024 
0.087 
A16-4 
5.072 
0.167 
0.019 
0.156 
A16-5 
0.513 
0.916 
0.002 
0.915 
A17-1 
2.138 
0.544 
0.008 
0.540 
A17-3 
6.715 
0.082 
0.025 
0.073 
D15-1 
5.436 
0.143 
0.018 
0.178 
D15-2 
2.114 
0.549 
0.008 
0.513 
D15-3 
3.862 
0.277 
0.014 
0.280 
D15-4 
1.678 
0.642 
0.007 
0.615 
E19-1 
8.080 
0.044 
0.300 
0.043 
E19-2 
9.283 
0.026 
0.032 
0.033 
E19-3 
23.079 
<0.001* 
0.082 
<0.001* 
E19-4 
6.558 
0.087 
0.026 
0.071 
E19-5 
2.680 
0.444 
0.010 
0.423 
E19-6 
0.823 
0.844 
0.002 
0.887 
H17-1 
3.824 
0.281 
0.014 
0.268 
H17-2 
3.119 
0.374 
0.012 
0.343 
H17-3 
0.776 
0.855 
0.003 
0.863 
H17-4 
2.155 
0.541 
0.007 
0.604 
H17-5 
0.500 
0.919 
0.002 
0.913 
Total (87 df) 164.66 
<0.001 
0.019 
<0.001 
the number of PCR cycles above 30, nonspecific DNA 
products can be obtained. However, in our prelimi- 
nary amplifications in extracting DNA from frozen 
tissue samples, less than 40 cycles produced faint 
fragment patterns for most primers; thus 40 cycles 
were used as a standard. The RAPD technique has 
been shown to be very sensitive to changes in con- 
centration of primer, concentration of template, an- 
nealing temperature, and the concentration of mag- 
nesium ions, all of which can affect the number and 
intensity of bands (Devos and Gale, 1992; Ellsworth 
et al., 1993; Patwary et al., 1993; Penner et al., 1993). 
We sought to avoid these problems by standardizing 
DNA quantities prior to amplification, performing all 
amplifications on the same thermocycler, and using 
the same batch of chemicals. Tissue samples from 
the four spawning sites were collected and stored 
under similar conditions. 
