Comparing geographic variation of non-coding nuclear DNA polymorphisms, which presumably are neutral to natural selection, with geographic variation of allozymes is potentially a good way to detect the effects of selection on allozyme polymorphisms. A previous study of four anonymous nuclear markers in the American oyster, Crassostrea virginica, found dramatic differences in allele frequency between the Gulf of Mexico and the Atlantic Ocean. In contrast, 14 allozyme polymorphisms were fairly uniform in frequency between the two areas. This led to the conclusion that all of the allozyme polymorphisms were kept uniform in frequency by balancing selection. To test the robustness of this pattern, six additional anonymous nuclear DNA polymorphisms were surveyed in oysters from Panacea, Florida and Charleston, South Carolina, on the Gulf and Atlantic coasts respectively. Unlike the previously studied DNA markers, the six DNA polymorphisms examined here show geographic variation that is not significantly greater than that of allozymes. The reason for the discrepancy between the two sets of DNA polymorphisms is unclear.
There are several recent comparisons of geographic variation in different classes of polymorphisms. At the alcohol dehydrogenase locus in Drosophila melanogaster, the allozyme polymorphism exhibits greater geographic variation along the east coast of the United States than 22 restriction site polymorphisms in the locus (Berry and Kreitman 1993), which is further evidence that the cline in allozyme allele frequency is caused by selection. In the Atlantic cod, Gadus morhua, the average FST of 11 anonymous DNA polymorphisms is significantly greater than that of 10 allozyme polymorphisms (Pogson, Mesa and Boutilier 1995), which is consistent with widespread balancing selection on the allozyme polymorphisms. DNA variation in a portion of the sodium channel gene of the tobacco budworm, Heliothis virescens, exhibited greater geographic variation than 13 allozyme polymorphisms, consistent with the hypothesis that the sodium channel gene is affected by insecticide selection that varies geographically in intensity (Taylor, Shen and Kreitman 1995).
One of the most dramatic examples of contrasting amounts of geographic variation as evidence of selection is from the American oyster, Crassostrea virginica. Allele frequencies at 14 polymorphic allozyme loci ("polymorphic" being defined as the most common allele having a frequency less than 0.95) exhibited little geographic variation in a series of samples from Texas to Massachusetts (Buroker 1983). This was not surprising, given that oyster larvae are planktonic for two to three weeks (Loosanoff and Engle 1940), which could result in extensive gene flow. Mitochondrial DNA restriction fragment polymorphisms exhibited large differences in frequency between the Gulf of Mexico and Atlantic Ocean (Reeb and Avise 1990), but this could be ascribed to the smaller population size of maternally inherited mitochondrial genomes, and to the fact that differentiating selection anywhere on the mitochondrial genome would cause all mitochondrial polymorphisms to differentiate due to hitchhiking. However, four restriction site polymorphisms from anonymous nuclear loci also exhibited large differences in allele frequency between the Gulf of Mexico and the Atlantic Ocean (Karl and Avise 1992). A statistical analysis of the data (McDonald 1994) showed that all four anonymous nuclear polymorphisms exhibited greater FST' than all 14 allozyme polymorphisms, which is significantly different from the pattern expected if all the polymorphisms were affected only by neutral processes (P<0.001, Mann-Whitney U-test). The original data of Karl and Avise (1992) are complicated by the failure of some alleles to amplify, but revised primers which amplify alleles from almost all individuals still exhibit large geographic variation in the restriction site polymorphisms (Hare, Karl and Avise 1996).
There is no a priori reason to suspect that anonymous nuclear DNA polymorphisms would be subject to differentiating selection. Therefore, a high amount of geographic variation observed between the Gulf of Mexico and the Atlantic Ocean would mean that there are two populations which have been small enough, for long enough, with little enough migration that they could differentiate due to random drift. If that is the case, the allozyme polymorphisms must be kept similar in allele frequency, in the face of isolation and random drift, by balancing selection (Karl and Avise 1992). This conclusion is startling; not only does it require that there is balancing selection on all of the 14 allozyme polymorphisms, it also requires that for every allozyme polymorphism, selection favors similar allele frequencies in two environments, the Gulf of Mexico and the Atlantic Ocean, which differ in temperature and perhaps many other environmental variables.
The only explanation other than balancing selection on allozyme polymorphisms would be that the markers used in Karl and Avise (1992) did not accurately represent the amount of geographic variation of neutral markers in oysters. The amount of geographic variation of neutral markers between two populations is expected to be highly skewed, with some polymorphisms having much greater FST than the mean (McDonald 1994). Thus it could be that Karl and Avise (1992) happened to study four polymorphisms from one end of the expected distribution of FST. It is also possible that the anonymous markers are affected by differentiating selection, although it is difficult to imagine a biological basis for selection on anonymous DNA polymorphisms that would not also affect allozymes. Some kind of technical artifact is also a possibility. Hare, Karl and Avise (1996) report artifacts when using two of the original sets of Karl and Avise (1992) primers, but they were able to design new primers that eliminated the artifacts. Because both the technique and the interpretation of Karl and Avise (1992) are so important, we have tested the robustness of the pattern they observed by obtaining data on geographic variation at additional anonymous nuclear DNA polymorphisms.
In March, 1994 and March, 1995, oysters collected at Panacea, Florida were purchased from Gulf Specimen Co. In April, 1994 and April, 1995, oysters were purchased from commercial shellfish companies in Charleston, South Carolina. A portion of the adductor muscle was dissected out, and DNA was extracted using a CTAB protocol (Winnepenninckx, Backeljau, and de Wachter 1993) modified for smaller volumes of tissue.
Anonymous nuclear DNA fragments were identified using a modification of the SCAR technique (Paran and Michelmore 1993). First, PCR was carried out using primers originally designed to amplify the locus encoding glucose-6-phosphate isomerase (Gpi) in Drosophila melanogaster. These primers were not in areas of the Gpi gene in D. melanogaster that are highly conserved across phyla, and where two primers were used they were not in the correct orientation to amplify Gpi, so there is no reason to consider these primers anything other than random with regards to oyster DNA sequence. PCR was run at low stringency (high magnesium and low annealing temperature), so that fragments were amplified at places where the primers annealed imperfectly.
The fragments which were amplified with the random primers were then sequenced. Some fragments were amplified using two different primers, one at each end of the fragment. These fragments were gel purified (Weichenhan 1991) and sequenced using dye-terminator cycle sequencing, with the PCR primers as sequencing primers. Other fragments were amplified with just one primer. Because these fragments had the same primer at each end, they could not be sequenced directly. Instead, they were digested with a four-cutter restriction enzyme that cut the fragment into two smaller fragments, then each fragment was gel purified and sequenced.
Once a fragment amplified with random primers was sequenced, PCR primers were designed to match the oyster sequence. These primers were located just inside the original random primers. Primers were designed for eight fragments; sequences from each fragment, annotated with the primers used to amplify the fragments, have been deposited in GenBank (accession numbers U60756 through U60761).
For each anonymous fragment, an initial search for polymorphisms was conducted on a small number of oysters. The variation identified in each fragment was then surveyed in 60 individuals from each location.
Fragment CVJ6 was initially surveyed in four oysters from Panacea, Florida, and fragment CVL3 was surveyed in five oysters from each location. These two fragments exhibited presence/absence variation: each oyster either had a PCR product clearly visible on agarose minigels, or it did not. This presence/absence variation was consistent across a variety of annealing temperatures and magnesium concentrations. These two fragments were then surveyed in all 60 oysters from each location at about the same time, and with the same DNA preparations, as other fragments that amplified from all individuals; this acted as a control for the quality of the DNA.
Fragment CVJ5 was initially surveyed in four oysters from Panacea, and this revealed size variation that was visible on agarose minigels. Fragment CVJ5 had two alleles, allele L, which was 272 bp long (including 40 bp of primers), and allele S, which was 88 bp shorter. Fragment CVL1 was initially surveyed in five oysters from Panacea, and it also exhibited size variation. The most common allele (L) of fragment CVL1 was 344 bp long (including 46 bp of primers). Other alleles were S, which was sequenced and found to be 39 bp shorter than L; M, which was intermediate in size between L and S; and X, which was larger than L. Fragment CVL1 could not be amplified, despite repeated attempts, from five individuals from Panacea and five individuals from Charleston.
Fragments CVB1 and CVB2 did not exhibit size or presence/absence variation, so PCR products from a few individuals were sequenced and polymorphic restriction sites were identified in the sequences. Fragment CVB1 was sequenced in four Panacea and five Charleston oysters, which revealed a polymorphic AccI site. The two alleles were called "C" and "U" for cut and uncut, respectively. Fragment CVB2 was sequenced in 17 Panacea and 15 Charleston oysters, and it had two polymorphic Tsp509I sites, site m at 68 bp and site e at 104 bp from one end of the 155-bp fragment. All of the four possible haplotypes (mCeC, mCeU, mUeC, mUeU) and ten possible genotypes were observed. No banding patterns that would require the presence of three of the four haplotypes were observed. The two CVB2 polymorphisms exhibited significant linkage disequilibrium with each other at Panacea (G-test, P=0.03) but not Charleston (P=0.19). The direction of disequilibrium was the same in both samples, with fewer mCeC and mUeU than expected.
Fragments CVB3 and CVL2 had no length or presence/absence variation, and sequencing of 18 individuals from each location did not reveal any polymorphic restriction sites. These fragments therefore will not be considered further.
The CV7 restriction site polymorphism originally identified by Karl and Avise (1992) was surveyed in the oysters from both locations, using both the original CV7 primers of Karl and Avise (1992) and the CV7.7 primers of Hare, Karl and Avise (1996) for PCR amplification, followed by digestion with Hinf I.
For the presence/absence fragments (CVL3 and CVJ6), FST was calculated according to Lynch and Milligan (1994). For the other four fragments, FST' was calculated (Nei 1986). For fragment CVL1, which has multiple size variants, all alleles except the most common one were pooled. Polymorphisms with different numbers of alleles have different expected distributions of FST', but pooling all but the most common allele makes the expected distributions similar (McDonald 1994). The same pooling was done with the published allozyme data. For fragment CVB2, it would be incorrect to consider the two restriction site polymorphisms to be independent. FST' was therefore calculated separately for each site and then averaged. Because the distribution of FST resulting from random drift of two populations is highly non-normal (McDonald 1994), the Mann-Whitney U-test (Sokal and Rohlf 1981, pp. 433-435) was used to compare FST between allozymes and DNA.
Genotype frequencies for fragments CVB1, CVB2 and CVL1 did not deviate significantly from the expected Hardy-Weinberg proportions (table 1). Fragment CVJ5 had a significant deficit of heterozygotes in the Panacea sample, but not in the Charleston sample. The general agreement with Hardy-Weinberg proportions is consistent with the fragments being single-copy nuclear DNA from oysters, because amplification of multiple-copy DNA, of mitochondrial DNA, or of DNA from a parasite could yield dramatic deviations from Hardy-Weinberg proportions. One fragment, CVL1, did not amplify from all individuals and exhibited a nearly significant deficit of heterozygotes at Panacea, which suggests that there is a CVL1 allele that does not amplify. Because the number of individuals from which CVL1 did not amplify was the same in Panacea and Charleston, the unamplifiable allele does not appear to differ greatly in frequency between the two locations.
DNA that was amplified with the CV7 primers of Karl and Avise (1992) and then digested with Hinf I exhibited banding patterns on gels which were clearly quite different in frequency between Panacea and Charleston, but which were difficult to interpret. Some individuals had restriction fragments whose total size was greater than twice that of the uncut PCR product, which is inconsistent with amplification of single-copy DNA. Hare, Karl and Avise (1996) also had difficulties with these primers, getting amplification of more than one PCR product from single individuals and heterozygote deficits caused by failure to amplify some alleles. Digestion with Hinf I of PCR products amplified by the CV7.7 primers of Hare, Karl and Avise (1996) yielded patterns that were consistent with amplification of single-copy DNA. There was a significant deficit of heterozygotes in the Panacea sample, but not in the Charleston sample. CV7.7 could not be amplified from two individuals from Charleston.
The six anonymous fragments exhibited little to moderate amounts of geographic variation in allele frequencies, as measured by estimators of Wright's FST (table 2). Three fragments, including one of each kind of polymorphism (presence/absence, length, and restriction site), had FST near those of the least differentiated allozymes, and three fragments had FST near those of the most differentiated allozymes. The geographic variation of these DNA polymorphisms was not significantly different from the variation of the 14 allozyme polymorphisms of Buroker (1983) (Mann-Whitney U-test, U=0.21, P=0.84). All four DNA polymorphisms of Karl and Avise (1992) had larger FST than all 14 allozyme polymorphisms and all six of the anonymous fragments presented here. The four Karl and Avise (1992) polymorphisms had significantly greater geographic variation than allozymes (U=2.97, P=0.003) and the six fragments presented here (U=2.56, P=0.01).
Four anonymous nuclear DNA polymorphisms that were previously surveyed in oysters exhibited much greater geographic variation between the Gulf of Mexico and the Atlantic coast than did allozymes, leading to the conclusion that balancing selection was keeping the allozyme polymorphisms uniform in frequency in the face of random drift (Karl and Avise 1992). However, the six anonymous nuclear DNA polymorphisms reported here do not exhibit greater geographic variation than allozyme polymorphisms. Here we consider a number of possible explanations for the discrepancy between the two sets of DNA polymorphisms.
Because oysters are commercially important, they are sometimes transplanted from one location to another. If Gulf oysters are genetically distinct from Atlantic oysters, as found by Karl and Avise (1992), one possible explanation for our markers not being different in frequency could be that our Charleston oysters were, unbeknownst to us, descended from imported Gulf oysters (or alternatively that our Panacea oysters were descended from imported Atlantic oysters). If this were the case, then the markers of Karl and Avise (1992) would also be similar in frequency in our two samples of oysters. We therefore surveyed fragment CV7 of Karl and Avise (1992), which they found to have large geographic variation between the Gulf and Atlantic coasts, in our samples from the two locations. With the original primers of Karl and Avise (1992), the results were difficult to interpret but clearly showed a large difference between the two locations. With the improved CV7.7 primers of Hare, Karl and Avise (1996), the polymorphic restriction site was dramatically different in frequency between our two samples (table 2). If our samples had actually been from only one of two genetically differentiated populations, they should have exhibited similar allele frequencies for fragment CV7.7, so this explanation seems unlikely.
The fragments originally amplified by Karl and Avise (1992) failed to amplify from some alleles (Hare, Karl and Avise 1996), which means that the original survey misidentified as homozygotes some individuals which were heterozygous for an amplifiable allele and a null allele. This artifact led to large apparent deficits of heterozygotes. In addition, the original primers sometimes amplified more than one region, as shown by restriction digest band sizes that added up to more than twice the fragment size. New primers which amplify shorter portions of the original fragments do not exhibit heterozygote deficits or anomalous band patterns and still show large geographic variation (Hare, Karl and Avise 1996), so these artifacts do not seem a likely explanation for the difference between their results and ours.
Techniques based on PCR amplification of anonymous DNA regions run the risk of amplifying DNA from organisms other than the species being studied, and the tissues of oysters could easily contain a diverse community of bacterial and eukaryotic symbionts. Amplification of a fragment from a symbiont seems particularly plausible for presence/absence variation, such as observed for fragments CVL3 and CVJ6, and the data from these fragments (and all PCR-based presence/absence variation, such as RAPDs) should be viewed with some caution. However, it is difficult to see how a fragment amplified from microorganisms would give anything like the Hardy-Weinberg proportions seen at the co-dominant markers used here and the improved versions of the Karl and Avise (1992) markers used by Hare, Karl and Avise (1996).
The strategies used for initially searching for polymorphisms are slightly different between Karl and Avise (1992) and here. Karl and Avise (1993) amplified fragments from three individuals from each of three geographically distant locations. They found 11 restriction site polymorphisms and four length polymorphisms in five fragments (Karl and Avise 1993), and chose one restriction site polymorphism from each of four fragments for further research (Karl and Avise 1992). For three of the six fragments surveyed here, polymorphism was identified in an initial sample from only one location, not both. This runs the slight risk that a polymorphism which is fixed for one allele in the initial location will be overlooked. On the other hand, initially surveying multiple locations risks the possibility that polymorphisms whose patterns of geographic variation fit a preconceived notion might subconsciously be the ones chosen for further research. However, if the 15 polymorphisms initially identified by Karl and Avise (1993) had a similar distribution of FST as the 14 allozyme polymorphisms (Buroker 1983) and the six DNA polymorphisms presented here, even a conscious choice of the four most differentiated polymorphisms would be unlikely to yield the dramatic differentiation found by Karl and Avise (1992), so the slight difference in search strategies would seem to be unimportant.
The molecular techniques used for identifying anonymous regions are somewhat different between Karl and Avise (1992) and this study. Their anonymous DNA fragments were obtained by probing genomic libraries for low copy number regions, sequencing the clones, and designing PCR primers based on this sequence. Our anonymous fragments were obtained by performing PCR with essentially random primers under low stringency conditions, sequencing the amplified fragments, and designing PCR primers based on this sequence. Since none of the artifacts listed above seems particularly likely as a cause of the great difference in geographic variation between their polymorphisms and ours, we are left with two possibilities. One is that there is some other, unknown artifact which causes errors in either their data or ours. The other possible conclusion is that their technique preferentially identifies regions of the oyster genome which are geographically differentiated, while our technique identifies regions of the genome which are not differentiated geographically. Why such similar techniques should yield such different results remains a complete mystery, but one which must be addressed before conclusions about the role of balancing selection in maintaining uniform allozyme allele frequencies in oysters can be accepted.
We thank M. P. Hare for copies of manuscripts prior to publication, P. M. Gaffney for much valuable advice about oysters, and V. Dilegge and A. Hunt for their insightful and detailed comments on the manuscript.
Berry, A., and M. Kreitman. 1993. Molecular analysis of an allozyme cline: alcohol dehydrogenase in Drosophila melanogaster on the east coast of North America. Genetics 134:869-893.
Buroker, N. E. 1983. Population genetics of the American oyster Crassostrea virginica along the Atlantic coast and the Gulf of Mexico. Marine Biol. 75:99-112.
Hare, M. P, S. A. Karl, and J. C. Avise. 1996. Anonymous nuclear DNA markers in the American oyster and their implications for the heterozygote deficiency phenomenon in marine bivalves. Molec. Biol. Evol. 13:334-345.
Karl, S. A. and J. C. Avise. 1992. Balancing selection at allozyme loci in oysters: implications from nuclear RFLPs. Science 256:100-102.
Karl, S. A., and J. C. Avise. 1993. PCR-based assays of Mendelian polymorphisms from anonymous single-copy nuclear DNA: techniques and applications for population genetics. Molec. Biol. Evol. 10:342-361.
Loosanoff, V.L., and J. B. Engle. 1940. Spawning and setting of oysters in Long Island Sound in 1937, and discussion of the method for predicting the intensity and time of oyster setting. Bulletin of the Bureau of Fisheries 49:217-255.
Lynch, M., and B. G. Milligan. 1994. Analysis of population genetic structure with RAPD markers. Molec. Ecol. 3:91-99.
McDonald, J. H. 1994. Detecting natural selection by comparing geographic variation in protein and DNA polymorphisms. Pp. 88-100 in B. Golding, ed. Non-neutral evolution: theories and molecular data. Chapman and Hall, New York.
Nei, M. 1986. Definition and estimation of fixation indices. Evolution 40:643-645.
Paran, I., and R. W. Michelmore. 1993. Development of reliable PCR-based markers linked to downy mildew resistance genes in lettuce. Theor. Appl. Genet. 85:985-993.
Pogson, G. H., K. A. Mesa, and R. G. Boutilier. 1995. Genetic population structure and gene flow in the Atlantic cod Gadus morhua: a comparison of allozyme and nuclear RFLP loci. Genetics 139:375-385.
Reeb, C. A., and J. C. Avise. 1990. A genetic discontinuity in a continuously distributed species: mitochondrial DNA in the American oyster, Crassostrea virginica. Genetics 124:397-406.
Sokal, R. R. and F. J. Rohlf. 1981. Biometry. W.H. Freeman, San Francisco.
Taylor, M. J. F, Y. Shen, and M. E. Kreitman. 1995. A population genetic test of selection at the molecular level. Science 270: 1497-1499.
Weichenhan, D. 1991. Fast recovery of DNA from agarose gels by centrifugation through blotting paper. Trends Genet. 7:109.
Winnepenninckx, B., T. Backeljau, and R. de Wachter. 1993. Extraction of high molecular weight DNA from molluscs. Trends Genet. 9:407.
Count Locus Genotype FL SC CVJ6 PP+Pa 16 35 aa 44 25 CVL3 PP+Pa 14 9 aa 46 51 CVJ5 LL 14 7 LS 21 29 SS 25 24 P 0.03 0.74 CVL1 LL 41 52 LM 5 0 LS 5 2 LX 3 0 MX 1 0 SS 0 1 P 0.97 0.07 CVB1 UU 36 35 UC 23 21 CC 1 4 P 0.20 0.68 CVB2 mUeU/mUeU 8 3 mUeU/mCeU 10 2 mUeU/mUeC 9 22 mUeU/mCeC 5 2 mCeU/mCeU 6 2 mCeU/mUeC 14 6 mCeU/mCeC 2 0 mUeC/mUeC 2 15 mUeC/mCeC 3 5 mCeC/mCeC 1 3 P(m) 0.49 0.07 P(e) 0.48 0.60 CV7.7 UU 4 49 UC 9 9 CC 47 0 P 0.01 0.38
Note.---For CVL3 and CVJ6, P indicates that the fragment is present, and a that it is absent. For CVJ5 and CVL1, allele names indicate the relative lengths of the PCR products, with S, M, L and X representing short, medium, long, and extra-long. For CVB1, CVB2, and CV7.7, C indicates a restriction site that cuts, and U indicates the uncut site. Sample size was 60 individuals for each location, except that CVL1 amplified successfully from only 55 individuals from each location, and CV7.7 amplified successfully from only 58 individuals from Charleston. P, probability of obtaining the observed deviation from Hardy-Weinberg proportions by chance (G--test with the Williams correction for continuity, except for CVL1 from Charleston, where Fisher's exact test was used).
Allele frequency Locus FL SC Fst CVJ5 0.608 0.642 -0.006 6Pgd 0.88 0.87 -0.005 CVB1 0.758 0.792 -0.005 Pgi 0.69 0.73 -0.002 CVL3 0.876 0.922 0.003 Est-3 0.55 0.63 0.004 Lap-2 0.66 0.73 0.006 Pgm-1 0.68 0.58 0.015 Aat-2 0.60 0.70 0.016 Adk-1 0.71 0.61 0.016 Sdh 0.93 0.86 0.024 Acp-3 0.50 0.66 0.041 Pgm-2 0.76 0.89 0.044 Lap-1 0.53 0.69 0.049 CVL1 0.864 0.973 0.053 Mpi-2 0.49 0.67 0.058 Ap-1 0.36 0.19 0.066 CVJ6 0.857 0.647 0.095 CVB2m 0.583 0.792 0.116 CVB2e 0.650 0.367 0.116 Est-1 0.78 0.48 0.163 CV19 0.671 0.304 0.224 CV195 0.914 0.548 0.278 CV7 0.597 0.045 0.508 CV32 1.000 0.409 0.583 CV7.7(HKA) 0.792 0.087 0.664 CV7.7(MVG) 0.858 0.078 0.757
Note.---Polymorphisms are listed from smallest to largest FST. The frequency of the most common allele in Panacea, Florida is shown. For CVL3 and CVJ6, FST was calculated according to Lynch and Milligan (1994); for all other polymorphisms, O(F,Dfo4()–)ST´ (Nei 1986) is calculated after pooling all alleles other than the most common allele in Panacea. The FST' for CVB2 is the average FST´ for the two polymorphic restriction sites in this fragment. DNA polymorphisms shown in bold type are from this study, and those italics are from Karl and Avise (1992); allozymes (Buroker 1983) are in plain type. CV7.7(HKA) is data from Hare, Karl and Avise (1996); CV7.7(MVG) is data on the same polymorphism from the present study.
Last Updated: October 19, 1998
URL of this document: http://udel.edu/~mcdonald/oyster.html