L.), sampled between 1985 and 1999 from a spawningarea in the Trondheimsfjord, Norway, were assayed for the nuclear-encoded locus
Pan
I
(
pantophysin). The majority of samples were taken during the spawning season at twonearby sampling locations at depths of 100 and 60 m, respectively. Genetic analysisrevealed significant effects of cohort, sex and sampling location on allele frequencies at
Pan
I. The contribution of each of these three factors to the total among-sample diversity(
F
ST
) of 8.01% at
Pan
I was estimated to be 3.78, 2.55 and 1.68%, respectively. Sign testsrevealed a significant excess of heterozygotes at both sampling locations in females; asignificant excess of heterozygotes in males was observed at one of the localities. Mutation,genetic drift and immigration do not appear to contribute significantly to the observedgenetic heterogeneity at
Pan
I, leaving natural selection as the main explanatory factor forthe Hardy–Weinberg imbalance. The dynamics of the selection at
Pan
I appear to becomplex. Analysis of age, sex and cohort proved crucial to disentangle putative explanatoryfactors from their secondary effects.
Received 20 May 2003; revision received 29 August 2003; accepted 29 August 2003
Introduction
Allozymes, single-copy nuclear DNA, mitochondrial DNA(mtDNA) and microsatellites are genetic markers that havebeen frequently used to study population genetic structures.These different classes of genetic markers may differ con-siderably in their rates and mechanisms of evolution andthere are several reports of discordance among classes ofmarkers in describing population structure (Pogson
et al
.1995; Turan
et al
. 1998; Buonaccorsi
et al
. 2001).Microsatellites are evolving comparatively fast accord-
ing to a stepwise mutation model. Such a mutation patternhas the potential to result in homoplasy (Goldstein &Schlötterer 1999). In addition, there are several reports ofnull alleles in microsatellites, making information of allelefrequencies of less potential in phylogeographical studiesbut they are, on the other hand, considered as sensitivemarkers for detecting shallow genetic divergences (Shaw
et al
. 1999).
Mitochondrial DNA is effectively one single locus andmainly maternally inherited, resulting in four-fold lowereffective population size compared with nuclear loci.Consequently, mtDNA is evolving faster with respect togenetic drift and only the maternal side of the evolution isdetected, which can be an important issue if, for example,males and females have different migration pattern (Karl
et al
. 1992).The special features of mtDNA (Clayton 1992; Wolstenholm
1992; Saccone 1994) and microsatellites (O’Connel & Wright1997; Li
et al
. 2002), described above, are well known. Whatis more critical and more difficult to deal with is naturalselection that might act on a single locus. Natural selectioncan potentially act in many different ways in time andspace and the selective agents are often unknown. Phylo-genetic and population genetic studies assume neutralityamong the molecular markers included, an assumptionthat is probably valid in most cases. There are, however,reports of loci that are affected by natural selection (Place& Powers 1979; Karpov & Novikov 1980; Karl & Avise1992; Mork & Giæver 1999). One such locus is pantophysin(
Pan
I) (Pogson 2001). Polymorphism at this nuclear-encoded
Correspondence: S. Karlsson. Fax: +47 73591597; E-mail:[email protected]
restriction fragment length polymorphism (RFLP) locus inAtlantic cod was described initially by Pogson
et al
. (1995),who used this genetic locus to study population structurein Atlantic cod throughout its distributional range. Thisstudy was followed by more recent studies involving codfrom Icelandic waters, the northeast Arctic and coastalcod in northern Norway (Fevolden & Pogson 1995, 1997;Jonsdottir
Fevolden (1998) presented field observations that suggestedthat natural selection might be acting upon
Pan
I geno-types. This was later supported by studies at the molecularlevel (Pogson 2001). Most recently, Pogson & Fevolden(2003) used nucleotide sequence variation in the
Pan
I
A
-allele group to explore further the genetic differentiation at
Pan
I between the Arctic and coastal cod complex of north-ern Norway. As pointed out by Pogson & Fevolden (2003),details of the dynamics of
Pan
I polymorphism in popula-tions of cod remain elusive.
In the present study, variation in
Pan
I genotypic compo-sition among groups was examined in a time series fromspawning locations of a cod stock in Trondheimsfjord,Norway. The time series includes samples broken down byage, sex, gonadal stage and body length and is representedby 26 cohorts (year classes). This time series enabled us toexamine temporal stability at the
Pan
I locus, as well as totest for association between genotypes at
Pan
I and vari-ation in phenotypic traits.
Materials and methods
The principal sampling location, Verrasundet, is a relativelynarrow side-fjord in the inner part of the Trondheimsfjordin Norway (Fig. 1). Verrasundet is a well-defined spawninglocality for cod (Dahl 1899; Sundnes 1980; Mork
et al
. 1985;Mork & Giæver 1999) and has two natural basins. The innerbasin (Verrabotn) has a maximum depth of 67 m and isseparated from the outer basin (Skalvik) by a thresholdat 15 m depth. The latter basin has a maximum depth of110 m and is separated from the main fjord by a 70-mthreshold. Cod spawn in March to May in both basins(Sundnes 1980).
Samples from spawning cod in both basins were col-lected over a 15-year period with a bottom trawl (35 mmstretched mesh in the cod-end) by the R/V ‘Harry BorthenI’ from Trondhjem Biological Station, NTNU (The Nor-wegian University of Science and Technology). From thetotal sample of 1295 cod, 1290 individuals were assayedfor genotypes at the pantophysin locus (
Pan
I) (Table 1)
.
Except for the sample taken in 1990, sex, age, gonadalstage and total body length were recorded for each indi-vidual. Age determination from otoliths was performed byEkli (1997 and unpublished) according to the breakage-and-side-illumination method of Rollefsen (1933). Determination
of gonadal maturation stage followed Sivertsen (1935):stage I, immature; stage II, maturing; stage III, running andstage IV, spent. Body length ranged from 14.0 to 99.0 cmand age ranged from 1 to 14 years. Gonadal stage ranged
Fig. 1 Map of the Trondheimsfjord showing the sidearmVerrasundet with the two trawling sampling locations forspawning cod, Skalvik and Verrabotn, indicated by vertical bars.
Table 1 Summary data
Season Locality NPanIA male
PanIA female
Females (%)
99 Spring Verrabotn 92 0.927 0.883 3399 Spring Skalvik 54 0.825 0.926 6398 Spring Verrabotn 73 0.940 0.700 2198 Spring Skalvik 27 0.714 0.875 7498 Autumn Verrabotn 50 0.879 0.971 3497 Spring Verrabotn 96 0.962 0.750 2096 Spring Verrabotn 24 0.806 0.833 7596 Spring Skalvik 72 0.875 0.750 8395 Spring Verrabotn 76 0.930 0.868 2595 Spring Skalvik 20 1.000 0.900 2594 Spring Verrabotn 91 0.943 0.792 1392 Spring Verrabotn 100 0.867 0.862 4091 Spring Verrabotn 72 0.873 0.808 1891 Spring Skalvik 8 1.000 090 Spring Verrabotn 96 — — —89 Spring Verrabotn 65 0.875 0.833 4089 Spring Skalvik 24 0.818 0.654 4986 Spring Skalvik 96 0.861 0.657 3985 Spring Verrabotn 100 0.958 1.000 485 Spring Skalvik 59 0.776 0.800 17Total 1295 0.903 0.824 31
N, sample size; PanIA, frequency of the A-allele at PanI, presented for each sex in each sampling year and sampling location. As the 1990 sample was not sexed, allele frequencies in sexes for this sample are missing.
from stage I to IV except for the samples taken in 1995 (noindividuals of gonadal stage IV) and the autumn of 1998(no individuals with gonadal stage III).
Samples of blood, white muscle, gills, heart, liver andkidney were stored frozen or in ethanol and DNA wasextracted from one of these tissue types. Approximately70 mg of tissue were minced with a glass rod in 2-mLplastic tubes in 0.75 mL reaction buffer (0.1
m
Tris, pH 7.8,0.005
m
EDTA, 0.5% sodium dodecyl sulphate) and6
µ
L (20 mg/mL) proteinase K. The reaction mixturewas incubated overnight at 50
°
C, followed by phenol :chloroform extraction of DNA as described in Sambrook
et al
. (1989). The extraction step was, however, performedonly once and a higher proportion (2 : 3) of chloroformwas added. The mixture was centrifuged at 5000
g
for15 min.
The polymerase chain reaction (PCR) amplification of
Pan
I was as described by Fevolden & Pogson (1997) exceptthat the PCR protocol closely followed that describedby Galvin
et al
. (1995). To a total reaction volume of 20
µ
L,containing 1
×
reaction buffer IV (Advanced Biotechnolo-gies™), 3.75
µ
m
MgCl
2
, 0.25 m
m
dNTP, 0.0705
µ
m
forwardprimer (B), 0.0720
µ
m
reverse primer (Syn 7) and 1 U of
Taq
polymerase, 2 ng
−
5
µ
g of template were added. The PCRreactions were carried out in a Hybaid® Omni-E thermalcycler programmed for 30 cycles of denaturation at 94
°
C(30 s), primer annealing at 55
°
C (30 s), primer extension at72
°
C (30 s) followed by a 7-min extension period at 72
°
C.The polymorphism at
Pan
I was detected by digestion ofthe PCR product with the restriction enzyme
Dra
I. Amastermix (8
µ
L) that contained 2.92
×
restriction enzymebuffer and 6.67 U of
Dra
I (Gibco BRL®) was added to 10
µ
Lof PCR product and incubated at 37
°
C for 90 min. Diges-tion was stopped by adding 1
µ
L of EDTA (0.5
m
) and 4
µ
Lof loading dye (0.25% bromophenol blue in 30% glycerol).
Digestion fragments corresponding to alleles A andB were separated in 2% submarine agarose gels (1
×
TBEbuffer) at 9.80 V/cm for 20 min and detected by ethid-ium bromide fluorescence. The electrophoresis apparatus(Electro-Fast System; AB-0679; Advanced Biotechnologies)was adjusted to allow for a separation distance of 2 cm.Genotyping at
Pan
I was straightforward. As a control ofgenotyping reproducibility, the 1996 sample was geno-typed both at Trondhjem Biological Station (TBS) and byÓlöf Bartels Jónsdóttir (Marine Research Institute, Reykjavik,Iceland) with complete concordance of genotypes in all 96individuals.
Statistical analyses of continuous variables were performedby one-way analysis of variance (
anova
) as implementedin
statgraphic
Plus 2.1 (STSC, Inc.). For discrete variables,
χ
2
and sign tests were used as appropriate. In situationswith low expected values (i.e. more than 25% of cells withexpected numbers < 5), the traditional
χ
2
tests were replacedby the Monte Carlo-based exact test (1000 iterations) of
Zaykin & Pudovkin (1993). Probabilities estimated fromindividual exact tests were combined into an overall pro-bability by the ‘omnibus’ test of Fisher (1954). Results fromsign tests are noted in the form of
b
(
x; n, p
) where
b
is bino-mial distribution and
x
is the number of events from
n
trials,each with a likelihood of
p
. In order to compare bodylength independently of age, individual body lengths weremultiplied by a factor derived from the relationship (linearregression in a double logarithmic plot) between age andmean length in the total material. Estimates of the fixationindex
θ
, the
F
ST
analogue of Weir & Cockerham (1984) andtests of conformance to Hardy–Weinberg genotypic expecta-tions were carried out with version 2 of
. (1982). Grouping the material into cohort, locationand sex resulted in 86 test groups (cf. Table 2). As
negst
allows a maximum of 70 groups, we reduced the materialfrom 86 to 70 groups by excluding the 16 smallest groups(each containing one to two individuals) from the nestedgene diversity analysis.
A difference in allele frequency between sexes at repro-duction is expected to result in an excess of heterozygotesamong offspring as compared with Hardy–Weinberg ex-pectations (Hedrick 1985; Falconer & Mackay 1996). Thefrequency of heterozygous offspring (qAB) in this situationis expected to be:
qAB = (qA
��
×
qB
��
) + (qB
��
×
qA
��
) (1)
where qA and qB are the parental frequencies of A and Balleles at a two-allele locus such as
Pan
I. Equation (1) wasused to adjust genotypic proportions in the offspring(cohorts) compared with those expected under Hardy–Weinberg equilibrium (HWE). Deviation from this expectedheterozygosity was tested by sign tests within cohort,location and sex. However, because all samples were notsexed, and because not all cohorts or their parents wererepresented, there were only two to seven test groupswithin the same cohort, location and sex. To increase thenumber of test groups, we used the expected increase inheterozygosity based on the overall sexual frequenciescompared with the Hardy–Weinberg ratio on each location.Adding this quantity to the Hardy–Weinberg hetero-zygosity in each sex on each location in each cohortincreased the number of test groups and thereby enabledmeaningful sign tests.
Results
Sample information and number of individuals genotypedare summarized in Table 1. Results from analysis of variationat
Pan
I are presented first as observed in annual samplesand then after sorting individuals by cohort (year classes).
I in all annual samplesfrom Verrasundet were in HWE (range
P
= 0.06 to 1.0).However, an excess of heterozygotes was observed in mostsamples. In females, the heterozygote excess was observedin 11 of 12 samples and this trend was statistically sig-nificant by a sign test [
b
(11; 12, 0.5),
P
= 0.006]. In males,eight of 12 samples showed heterozygote excess but thistrend was not statistically significant [
b
(8; 12, 0.5),
P
=0.388]. The total material had an excess of heterozygotes of0.009 compared with Hardy–Weinberg.
Allele frequencies differed significantly among sam-pling years when sexes were pooled ( = 73.01,
P
≈
0) aswell as in females ( = 25.84,
P
< 0.005,
F
ST
= 0.0273)and males ( = 22.56,
P = 0.0125, FST = 0.0085). Theoverall PanIA allele frequency (all individuals included)was 0.881 and the observed variance among samplingyears was 2.3 × 10−3. In 10 of 11 sampling years, males hadhigher frequencies of PanIA than females (b(10; 11, 0.5),P = 0.012). The overall PanIA allele frequency (total materials)was 0.824 in females and 0.903 in males; this difference wasstatistically significant ( = 59, P ≈ 0).
Subdivision of annual samples into sampling location(Verrabotn and Skalvik) revealed a significantly higherPanIA allele frequency in cod from Verrabotn as comparedwith cod from Skalvik (pooled sexes, = 27.32, P ≈ 0;males, = 18.50, P ≈ 0; females, = 3.06, P = 0.08).Significant differences in PanI allele frequency amongannual samples were found at both sampling locations formales (Monte Carlo test, P = 0.002, Verrabotn; P = 0.049,Skalvik) but only in Skalvik for females (Monte Carlo test,P = 0.545, Verrabotn; P = 0.002, Skalvik). Sexually matureindividuals (gonadal stage II–IV) had a significantly higherallele frequency of PanIA than immatures (total material,
= 28.40, P ≈ 0). In mature males (P = 0.001), but not inmature females (P = 0.77), the allele frequency varied sig-nificantly among individuals of different age.
No significant effect of PanI genotype on gonadal stagein sexually mature individuals (stage II–IV) was found(Monte Carlo test, P = 0.574, females; P = 0.098, males).However, the age of mature individuals (stage II–IV) dif-fered significantly between the AA and AB genotypes ofPanI in males (P ≈ 0) but not in females (P = 0.39). No sig-nificant effect of PanI genotype on average standardizedbody length was found (males, F = 2.57[2,815], P = 0.077 andfemales, F = 2.35[2,1192], P = 0.097).
Analysis by cohort
Analyses of annual samples (above) established thatvariation at PanI was not independent of sampling year,location, sex and fish age. However, because age, sex andrelative sizes of samples from the two sampling locations
varied among sampling years, interactions among thesefactors may have confounded statistical analysis. To dis-entangle these various effects, further analyses wereperformed on cohorts rather than sample year (cf Table 2).
Homogeneity of allele frequencies within the samecohort over time (pooled locations) was tested via exacttests followed by the omnibus test. Allele frequencies werehomogeneous in both males (P = 0.13) and females (P =0.87). Many of the age groups were, however, representedby very few individuals and the statistical test power wascorrespondingly low.
Effect of cohort within locations and sex. Subdividing the databy cohort, sex and sampling location led to subgroups withvery few individuals and hence to use of exact tests(Zaykin & Pudovkin 1993). The effect of cohort on PanIA
allele frequency was highly significant when the data weresubdivided by sex within locations (Table 3, Fig. 2a–d).
Effect of sampling location within cohort and sex. To testwhether there was a significant difference in PanIA allele
χ [ ]112
χ [ ]102
χ [ ]102
χ [ ]12
χ [ ]12
χ [ ]12 χ [ ]1
2
χ [ ]12
Table 2 Sample sizes and sex distributions in cohorts
frequency between the two sampling locations, we usedtwo approaches: χ2 test (Table 3) and a sign test (Table 4).χ2 tests based on the largest samples (i.e. those withexpected number smaller than 5 in less than 25% of cells,i.e. in males cohorts 77, 78, 84, 85, 86, 91 and 94 and infemales cohorts 83, 84, 85, 86, 89, 90, 91 and 94) showed asignificant effect of sampling location in males ( = 18.30,
P = 0.01) but not females ( = 12.02, P = 0.15). Sign tests,however, revealed no significant trend either in males [b(13;19, 0.5), P = 0.17] or in females [b(11; 16, 0.5), P = 0.21].
Effect of sex within cohort and location. To test whether therewas a significant difference in PanIA allele frequencybetween sexes, regardless of the effect of cohort and loca-tion, we used a χ2 test (Table 3) and a sign test (Table 4). χ2
tests performed on those cohorts where the expectednumbers in test cells fulfilled the assumptions of the test(Verrabotn, cohorts 84, 85, 86, 87, 89, 90, 91, 93 and 96;Skalvik, cohorts 83, 84, 85 and 91) revealed significantdifferences in PanIA allele frequency between sexes(Table 3). Sign tests revealed a significant trend of higherPanIA allele frequency in males as compared with femalesin Verrabotn [b(14; 17, 0.5), P = 0.012] but not in Skalvik[b(12; 17, 0.5), P = 0.144].
Hierarchical gene diversity analysis. Hierarchical gene diver-sity analysis was used to estimate relative contributions ofcohort, location and sex to allele frequency variation atPanI. The data were subdivided into groups containingindividuals within the same cohort, location and sex (cf.Table 2). The results showed a total FST of 8.01%, of whichcohort stood for 3.78%, location 1.68% and sex for 2.55%.
Heterozygosity. Different allele frequencies between malesand females in a spawning population are expected toresult in an excess of heterozygotes among offspring as
Table 3 χ2 tests for homogeneity of PanIA allele frequency incohorts, locations and sexes
Homogeneity test Results of χ2 test
Results of Monte Carlo tests (1000 iterations)
Among cohorts
�� Verrabotn = 865.3, P ≈ 0 P ≈ 0
�� Skalvik = 64.7, P ≈ 0 P ≈ 0
�� Verrabotn = 185.4, P ≈ 0 P ≈ 0
�� Skalvik = 133.0, P ≈ 0 P ≈ 0
Between locations*
�� = 18.3, P = 0.01 —
�� = 12.0, P = 0.15 —
Between sexes*
Verrabotn = 21.7, P = 0.01 —
Skalvik = 11.1, P = 0.02 —
*Represents χ2 -values summed over cohorts.
χ [ ]182
χ [ ]172
χ [ ]192
χ [ ]202
χ [ ]72
χ [ ]82
χ [ ]92
χ [ ]42
Fig. 2 PanIA allele frequencies (qA) bycohort and sex in (a) Verrabotn (VB) and(b) Skalvik and by cohort and location for(c) males and (d) females.
compared with the proportions at HWE. We estimatedexpected heterozygosity for each cohort based on equationI. As described in the Materials and methods, we increasedthe number of test groups by using the expected increasein heterozygosity (3.6 × 10−3 in Verrabotn and 7.6 × 10−6 inSkalvik) based on the overall sexual frequencies comparedwith the Hardy–Weinberg ratio on each location. Addingthis quantity to the Hardy–Weinberg heterozygosity ineach sex on each location in each cohort resulted in 14–18groups in the four tests. Significant heterozygote excesswas found in all tests except for males in Verrabotn(Table 4).
The most critical assumptions for these estimates are: (i)the sexual differences in allele frequency are the same eachgeneration and (ii) an equal probability of mating successfor all males and females. These two assumptions are prob-ably violated to some extent leading to an over- or under-estimate of expected heterozygosity among the offspring.The sexual difference in allele frequency must, on the otherhand, be relatively large in order to result in a noticeableincrease in heterozygosity among the offspring and conse-quently the errors of the estimates will be correspondinglysmall.
Phenotypic effects. The non-random association betweenPanI genotypes and various phenotypic traits (i.e. age ofmature individuals and gonadal stage), which was observedin the pooled data, was no longer significant in analysesbased on cohorts (sign tests). This could either be due tolower statistical power in analyses within cohorts or toconfounding effects from analyses of nonhomogeneousgroups of the pooled data.
Discussion
An important issue is the highly skewed and variable sexratio in our samples. One possible explanation for thiscould be that the fishing gear used was selective due todifferent behaviour between males and females. It is
possible that males and females are not evenly distributedin the sampling area and/or that they are not aggregatedat the same depth. Other studies have shown that sex ratioin a sample depends on the fishing gear used (Sundnes1991). In general, there were fewer females represented,resulting in lower statistical power in tests performed onfemales as compared with males. Similarly, there werefewer individuals represented from Skalvik comparedwith Verrabotn. The statistical analyses in this study werechallenged by small sample sizes when divided intocohort, location and sex. These small sample sizes havenaturally contributed to the observed variation in PanIallele frequencies. However, as in all statistical tests, thevariances are included and only contribute to decrease thestatistical power.
The four evolutionary forces that can potentially affectallele frequencies in a Mendelian population are mutation,genetic drift, gene flow and natural selection. As anextremely high mutation rate would be necessary toexplain our observations, we restrict our discussion to thethree latter evolutionary forces relative to the observedvariation in allele frequencies at PanI.
Genetic drift
If genetic drift alone was responsible for the fluctuations inallele frequencies at PanI observed in Verrasundet, thistwo-allele polymorphism would be expected to go rapidlyto fixation. From the observed variance in allele frequencyat PanIA (2.3 × 10−3) and the overall allele frequency ob-served at PanIA (p) the corresponding effective populationsize would be estimated from the following equation(Bhattacharyya & Johnson 1977):
SE = √(p(1 − p))/2Ne
where SE is the square root of the variance. Thecorresponding effective population size is 23 individuals.Such a small effective population size, in the absence of
Test Outcome P
in males 13 of 19 0.17 in females 11 of 16 0.21
in Verrabotn 14 of 17 0.012 in Skalvik 12 of 17 0.144
Excess of heterozygotes Skalvik males 15 (15) of 17 0.002 (0.002)Excess of heterozygotes Skalvik females 15 (15) of 16 0.0005 (0.0005)Excess of heterozygotes Verrabotn males 14 (10) of 18 0.03 (0.814)Excess of heterozygotes Verrabotn females 13 (13) of 14 0.002 (0.002)
PanIA, frequency of the A-allele at PanI. The test of heterozygote excess shows two results, the one in parenthesis is after correction for the effect on offspring of different allele frequency between the sexes in spawning cod (cf. Materials and methods).
PanI PanVerrabotnA
SkalvikAI >
PanI PanVerrabotnA
SkalvikAI >
PanI PanI�� ��A A >
PanI PanI�� ��A A >
Table 4 Sign tests and two-way binomialprobabilities over cohorts for various fea-tures of the PanI locus
S E L E C T I O N - I N D U C E D V A R I A T I O N I N C O D 3271
other evolutionary forces, would theoretically bring thePanIA-allele to fixation by genetic drift in about 26generations (Kimura 1970). Moreover, all selectivelyneutral loci should be affected to a similar extent. Previousstudies, however, have given no indication that cod inVerrasundet are characterized by low genetic variability ascompared with other populations of cod (Mork et al. 1985).Furthermore, a preliminary study of the hypervariablemicrosatellite locus GMO132 among 646 cod fromVerrasundet (Karlsson, unpublished) revealed a similarnumber of alleles (n = 30) as the study material of Bentzenet al. (1996) (n = 672, 24 alleles) which included cod fromthe northwest Atlantic shelf and Barents Sea. The highallelic diversity at GMO132 in this stock could not havebeen maintained if the effective population size was lowenough to explain by genetic drift the intercohort allelefrequency variation in PanI. Moreover, using a multiallelicversion of the simulation software developed by Mork(1994) and the observed allele frequencies at GMO132(which were very skewed with many low-frequencyalleles), we found that the 30 alleles observed would bereduced to on average seven alleles (range 5–9 in 20iterations) after only four generations (our sampling timespan) if Ne was 23 individuals. In this context it is, however,important to mention the possibility of a low Ne : N ratio,due to large variance in reproductive success, which couldbe expected in a highly fecund species (Hedgecock 1994;Hauser et al. 2002). A low Ne : N ratio does not necessarilymean that the effective population size is low because codand other gadoid species can exist in tremendous numbersof individuals on spawning grounds. Several studies haveshown that populations of marine species have highergenetic variation as compared with anadromous or freshwater species (Gyllensten 1985; Ward et al. 1994; Dewoody& Avise 2000). This general pattern is probably due tolarger effective population sizes in marine species but alsoto less restrictions to gene flow in the marine environment.One explanation for the generally high level of allelicvariability and heterozygosity reported for cod inVerrsundet at isozyme loci (Mork et al. 1985) and GMO132could be that loss of genetic variability due to genetic driftis balanced by reintroduction of alleles by migration (geneflow). Such a scenario cannot be excluded but appearsunlikely from our data of the large intercohort variationin PanI allele frequencies as there are no indications ofimmigration (see below).
Gene flow (immigration)
The observed heterozygote excess at PanI in this study isinconsistent with an immigration hypothesis. Immigrantswith different allele frequencies in sufficient numbers toexplain the annual variation in PanI allele frequenciesshould create a relatively strong Wahlund effect (i.e.
homozygote deficiency). The opposite was observed. Theexcess of heterozygotes in the total material was 0.009.
Moreover, tagging experiments have shown that, at sex-ual maturation, spawning cod display a cyclical migrationpattern within the fjord. After spawning in Verrasundet,cod disperse into the main fjord and remain there until thenext spawning season. Only a small proportion (∼1.5%) oftagged fish have been recaptured outside Trondheimsfjord5 years after release (Sundnes 1980). However, more cri-tical in the present context is the migration of cod into theTrondheimsfjord. Tagging experiments performed by theInstitute of Marine Research (IMR, Bergen, Norway) overmany decades have indicated that the Trondheimsfjordreceives none or very few individuals from coastal andoceanic areas outside the fjord (O.R. Godø, IMR, personalcommunication).
Natural selection
The exclusion of mutation, genetic drift and immigrationas major factors for the allele frequency variation at PanIamong cod in the Trondheimsfjord leaves natural selectionas the most likely explanatory force. Other studies havereached the same conclusion; two recent studies at themolecular level (Pogson 2001; Pogson & Fevolden 2003)also concluded that natural selection is acting at PanI. Inseveral independent studies (Fevolden & Pogson 1995;Pogson et al. 1995; Staub 1997) very high FST values havebeen reported for PanI (GM798) between cod populationscompared with other cDNA RFLP loci. As pointed out byLewontin & Krakauer 1973), FST values for a set of neutralgenetic markers are expected to belong to the samestatistical distribution and to show the same mean valuefor a given set of populations (but see Robertson 1975 andNei & Maruyama 1975). Beaumont & Nichols (1996)identified PanI (GM798) as one of two cDNA RFLPmarkers introduced by Pogson et al. (1995) that showedextremely high FST values. They concluded that naturalselection was responsible for the observations.
Different scenarios of natural selection can explain thepersistence of the polymorphism at the PanI locus. Themost striking observations made in the present study are:(i) significant variation in allele frequency among cohorts;(ii) significant difference in allele frequency between sexes;(iii) significant difference in allele frequency between adja-cent sampling locations and (iv) excess of heterozygotes,especially pronounced in females. These patterns of vari-ation may indicate different strength and/or types of selec-tion between the sexes. A stable equilibrium of the PanIallele frequencies could be achieved if: (i) selection is ofopposite direction in the two sexes (i.e. advantage of one ofthe homozygotes in one sex and advantage of the otherhomozygote in the other sex); (ii) selection is stabilizing(heterozygote advantage) in both sexes or (iii) selection is
directional in one of the sexes and stabilizing in the other(Hedrick 1985).
In addition to explaining allele frequency differencesbetween sexes, natural selection can also account for thevariation in allele frequencies between sampling years andcohorts. Cod year-class strength may vary by a factor of10–20 in the Trondheimsfjord (Ekli 1997). If we assume thatvariation in year-class strength is mediated by total naturalmortality, then the allele-changing effects of selection atPanI may vary between years accordingly.
Although the PanIB- and A-allele are distinguished byan A→G transition in a noncoding region (in the thirdintron; Fevolden & Pogson 1997), the two alleles are cou-pled to differences at the protein level, including fouramino acids, suggesting that the alleles encode for proteinswith considerably different physiological properties(Pogson 2001).
It is possible that the strength and direction of a selectioncan vary between years due to some unknown variableenvironmental factor. One such factor may possibly betemperature, as integral membrane proteins of synapticvesicles, like synaptophysin (or pantophysin), are espe-cially sensitive to changes in temperature (Prosser &Nelson 1981). Cod collected in the Barents Sea, Newfound-land, Balsfjord (northern Norway) and four sampling loca-tions south of Iceland show lower PanIA allele frequenciescompared with Trondheims fjord cod, while cod fromNova Scotia, the Celtic Sea and the North Sea show higherPanIA allele frequency (Jónsdóttir et al. 1999; Pogson 2001).These observations might imply that there is a correlationbetween PanIA allele frequency and temperature. A corre-lation between latitude (temperature) and allele frequen-cies has been shown for other loci, such as HbI (Sick 1965),Transferrines ( Jamieson & Turner 1978) and LDH (Place &Powers 1979), as well as for meristic characters such asnumber of fin rays and number of vertebrae (Schmidt1930).
We observed no heterogeneity of allele frequencieswithin the same cohort over time in adult cod. This mayimply that selection that leads to heterozygote excess anddifferences in allele frequency occurs during the earlylife history when there is a particularly high mortality.Controlled crossing experiments with full-sibs exposed todifferent environments (e.g. temperature) could perhapsshed some light on the role of various environmentalfactors.
In conclusion, the present study has demonstrated sig-nificant temporal instability in PanI allele frequenciesacross approximately four generations of Trondheimsfjordcod. Females have a significantly lower frequency of thePanIA allele than males and a significant excess of hetero-zygotes. We also observed a microgeographical hetero-geneity between two sampling locations separated by only6 km. Consideration of the patterns of PanI variation led to
the rejection of genetic drift, immigration and mutation asmain explanatory factors for the variation observed. Therewere strong indications that PanI in Trondheimsfjord codis affected by natural selection acting through differencesin genotypic performance. We found that a reasonableexplanation of the sum of observations made in this studyis that maintenance of the polymorphism at PanI involvesa complex, dynamic form of natural selection that variesbetween sexes and among years. Consequently, a pointestimate of PanI allele frequencies from a single sample isan unreliable characteristic of the cod population in Trond-heimsfjord because observed allele frequencies dependheavily on sampling year, age composition, microlocalityand sex ratio within the sample.
Acknowledgements
We would like to thank Prof. Svein Erik Fevolden at the Universityof Tromsø who kindly helped us with protocols for the laboratorywork. PhD Tony Ryan has been very helpful in sharing with us hisskills and knowledge of the various procedures/techniques in thelaboratory. The research was funded by the Norwegian ResearchCouncil (grant no. 145336/432). Prof. John Gold, Texas A&MUniversity (on Fulbright sabbatical at Trondhjem BiologicalStation, NTNU, Norway); Dr Peter Smith, NIWA, Wellington, NewZealand; Prof. Bjørn Ivar Honne, The Norwegian Crop ResearchInstitute, Kvithamar, Norway and two unknown referees gaveus valuable advice in the finalization of this paper.
References
Beaumont MA, Nichols RA (1996) Evaluating loci for use in thegenetic analysis of population structure. Proceedings of the RoyalSociety of London B, 263, 1619–1626.
Bentzen P, Taggart CT, Ruzzante DE, Cook D (1996) Microsatellitepolymorphism and the population structure of cod (Gadusmorhua) in the North West Atlantic. Canadian Journal of Fisheriesand Sciences, 53, 2706–2721.
Bhattacharyya GK, Johnson RA (1977) Statistical Concepts andMethods. John Wiley, New York, NY, USA.
Buonaccorsi VP, McDowell R, Graves JE (2001) Reconciling pat-tern of inter-ocean molecular variance from four classes ofmolecular markers in blue marlin (Makaira nigricans). MolecularEcology, 10, 1179–1196.
Chakraborty R, Haag M, Ryman N, Ståhl G (1982) Hierarchicalgene diversity analysis and its application to brown trout popu-lation data. Hereditas, 97, 17–21.
Clayton DA (1992) Transcription and replication of animal mito-chondrial DNAs. International Review of Cytology, 141, 217–232.
Dahl K (1899) Beretning om fiskeriundersøgelser i og om Trond-hjemsfjorden. Det Kongelige. Norske Videnskabers Selskabs Skrifter,10, 1–62.
Dewoody JA, Avise JC (2000) Microsatellite variation in marine,freshwater, and anadromous fishes compared with other animals.Journal of Fish Biology, 56, 461–473.
Ekli PH (1997) Årsklasseundersøkelser 1963–1994 hos torsk(Gadus morhua L.) i Trondheimsfjorden. Master of Science Thesisin Marine Biology, Trondhjem Biological Station, NorwegianUniversity of Science and Technology (In Norwegian).
S E L E C T I O N - I N D U C E D V A R I A T I O N I N C O D 3273
Falconer DS, Mackay TFC (1996) Small populations: I change ofGene Frequency under Sinip Litied Conditions. In: Introductionto Quantitative Genetics, 4th edn, pp. 61–63. Longman Group,Essex, England.
Fevolden SE, Pogson GH (1995) Differences in nuclear DNARFLPs between Norwegian coastal and the north-east Arcticpopulation of Atlantic cod. In: Ecology of Fjords and CoastalWaters (eds Skjoldal HR, Hopkins C, Erikstad KE, Leinaas HP),pp. 403–415. Elsevier Science, B.V., Amsterdam.
Fevolden SE, Pogson GH (1997) Genetic divergence at the synapto-physin (PanI) locus among Norwegian coastal and north-eastArctic populations of Atlantic cod. Journal of Fish Biology 51, 895–908.
Fisher RA (1954) Statistical Methods for Research Workers, 12th edn.Oliver & Boyd, Edinburgh.
Galvin P, McGregor D, Cross T (1995) A single locus PCR ampli-fied minisatellite region as a hypervariable genetic marker ingadoid species. Aquaculture, 137, 31–40.
Goldstein DB, Schlötterer C (eds) (1999) Microsatellites: Evolutionand Applications. Oxford University Press, Oxford.
Gyllensten U (1985) The genetic structure of fish: Differences inintraspecific distrubution of biochemical genetic variationbetween marine, anadromous, and freshwater species. Journal ofFish Biology, 26, 691–699.
Hauser L, Adcock GJ, Smith PJ, Ramirez JHB, Carvalho GR (2002)Loss of microsatellite diversity and low effective populationsize in an overexploited population of New Zealand snapper(Pagurus auratus). Proceedings of the National Academy of Sciencesof the United States of America, 99 (18), 11742–11747.
Hedgecock D (1994) Does variance in reproductive success limiteffective population sizes of marine organisms. In: Genetics andEvolution of Aquatic Organisms (ed. Beaumont AR), pp. 122–134.Chapman & Hall, London.
Hedrick PW (1985) Genetics of Populations. Jones and Bartlett, Bos-ton, MA, USA.
Jamieson A, Turner RJ (1978) The extended series of Tf alleles inAtlantic cod, Gadus morhua. In: Marine Organisms (eds BattagliaB, Beardmore J), pp. 699–729. Plenum Publishing, Lowestoff,England.
Jonsdottir ODB, Imsland AK, Danielsdottir AK, Thorsteinsson V,Nævdal G (1999) Genetic differentiation among Atlantic cod insouth and south-east Iceland waters: Synaptophysin (PanI) andhaemoglobin (HbI) variation. Journal of Fish Biology, 54, 1259–1274.
Karl SA, Avise JC (1992) Balancing selection at allozyme loci inoysters: implications from nuclear RFLPs. Science, 256, 100–102.
Karl SA, Bowen BW, Avise JC (1992) Global population geneticstructure and male-mediated gene flow in the green turtle(Chelonia mydas): RFLP analysis of anonymous nuclear loci.Genetics, 131, 163–173.
Karpov LK, Novikov GG (1980) The haemoglobin alloforms in cod(Gadus morhua L.), their functional characteristics and distribu-tion in the populations. (Translated from Russian.) Voprosyichtiologli, 20, 823–827.
Kimura M (1970) Stochastic processes in population genetics, withspecial reference to distribution of gene frequencies and prob-ability of gene fixation. In: Mathematical Topics in PopulationGenetics. Biomathematics (ed. Kojima K), pp. 178–209. Vol. 1.Springer-Verlag, Berlin.
Lewontin RC, Krakauer J (1973) Distribution of gene frequency asa test of the theory of the selective neutrality of polymorphisms.Genetics, 74, 175–195.
Li Y-C, Korol AB, Fahima T, Beiles A, Nevo E (2002) Microsatel-lites: genomic distribution, putative functions and mutationalmechanisms: a review. Molecular Ecology, 11, 2453–2465.
Mork J (1994) Straying and population genetic structure. Aqua-culture and Fisheries Management, 25 (Suppl. 2), 93–98.
Mork J, Giæver M (1999) Genetic structure of cod along thecoast of Norway: Results from isozyme studies. Sarsia, 84, 157–168.
Mork J, Ryman N, Utter F, Sundnes G (1985) Genetic variation inAtlantic cod (Gadus morhua) throughout its range. Canadian Journalof Fisheries and Aquatic Sciences, 42, 1580–1587.
Nei M, Maruyama T (1975) Lewontin-Krakauer test for neutralgenes. Genetics, 80, 394.
O’Connel M, Wright JM (1997) Microsatellite DNA in fish. Reviewsin Fish Biology and Fisheries, 7, 331–363.
Place AR, Powers DA (1979) Genetic variation and relative cata-lytic efficiencies: Lactate dehydrogenase B allozymes of Fundu-lus heteroclitus. Proceedings of the National Academy of Science, 76,2354–2358.
Pogson GH (2001) Nucleotide polymorphism and natural selec-tion at the pantophysin (PanI) locus in the Atlantic Cod, Gadusmorhua (L). Genetics, 157, 317–330.
Pogson GH, Fevolden SE (1998) DNA heterozygosity and growthrate in the Atlantic cod Gadus morhua (L). Evolution, 52 (3), 915–920.
Pogson GH, Fevolden SE (2003) Natural selection and the geneticdifferentiation of coastal and Arctic populations of the Atlanticcod in northern Norway: a test involving nucleotide sequencevariation at the pantophysin (PanI) locus. Molecular Ecology, 12,63–74.
Pogson GH, Mesa KA, Boutilier RG (1995) Genetic populationstructure and gene flow in the Atlantic Cod. Gadus morhua: acomparison of allozyme and nuclear RFLP loci. Genetics, 139,375–385.
Prosser CL, Nelson DO (1981) The role of nervous systems intemperature adaption of poikilotherms. Annals of Review ofPhysiology, 43, 281–300.
Raymond M, Rousset F (1995) Genepop (version 2.1): Populationgenetics software for exact tests and ecumenicism. Journal ofHeredity, 86, 248–249.
Robertson A (1975) Remarks on the Lewontin-Krakauer test.Genetics, 80, 395–396.
Rollefsen G (1933) The otoliths of the cod. Fiskeridirektoratetsskrifter Seric Havundersøkelser, 4 (3), 1–14.
Saccone C (1994) The evolution of mitochondrial DNA. CurrentOpinions in Gene Development, 4, 875–881.
Sambrook J, Fritsch EF, Maniatis T (1989) Molecular Cloning. A Lab-oratory Manual, 2nd edn. Cold Spring Harbor Laboratory Press,New York.
Schmidt J (1930) The Atlantic Cod (Gadus callarias L.) and localraces of the same. Meddelelser fra Carlsberg Laboratoriet, 18, 1–72.
Shaw PW, Turan C, Wright JM, O’Connel M, Carvalho GR (1999)Microsatellite DNA analysis of population structure in Atlanticherring (Clupea harengus), with direct comparison to allozymeand mtDNA RFLP analyses. Heredity, 83, 490–499.
Sick K (1965) Haemoglobin polymorphism of cod in the North seaand the North Atlantic ocean. HERE, 54, 3.
Sivertsen E (1935) Torskens gyting. Med særlig henblikk på denårlige cyklus i generasjonsorganenes tilstand. Fiskeridirektoratetsskrifter Seric Havundersøkelser, 4 (10), 1–29 (In Norwegian).
Staub WD (1997) The genetic stock composition of Atlantic codalong the Norwegian coast and its implication for management
of the coastal fishery. Master of Science Thesis in Marine Science.University of California, Santa Cruz, USA.
Sundnes G (1980) Sluttrapport, Beistadfjordprosjektet, TBS rapport.Norwegian University of Science and Technology, Trondheim,Norway (In Norwegian).
Sundnes G (1991) Redskapskader på den norsk-arktiske torsketa-mmen (Gear damages on the Norwegian-Arctic cod stock). In:Det Konglige Norske Videnskapers Selskab, Forhandlinger, pp. 157–160. Tapir forlag, Trondheim, Norway.
Turan C, Carvalho GR, Mork J (1998) Molecular genetic analysis ofAtlanto-Scandian herring (Clupea harengus) populations usingallozymes and mitochondrial DNA markers. Journal of theMarine Biological Association UK, 78, 269–283.
Ward RD, Woodwark M, Skibinski DOF (1994) A comparison ofgenetic differentiation levels in marine, freshwater, and ana-dromous fishes. Journal of Fish Biology, 44, 213–232.
Weir BS, Cockerham C (1984) Estimating F-statistics for the ana-lysis of population structure. Evolution, 38 (6), 1358–1370.
Wolstenholm DR (1992) Animal mitochondrial DNA: structureand evolution. International Review of Cytology, 141, 173–216.
Zaykin DV, Pudovkin AI (1993) Two programs to estimate signi-ficance of: χ2 values using pseudo-probability tests. Journal ofHeredity, 84, 84–52.
The authors are part of the population genetics group atTrondhjem Biological Station, the Norwegian University ofScience and Technology. Their main research interests are thepopulation genetic structure of marine organisms and theevolutionary forces involved, with special focus on ecologicallyand economically important gadoid species in the north Atlantic.
