SEARCH

SEARCH BY CITATION

Keywords:

  • disease;
  • elongation factor;
  • marine;
  • overdominance;
  • Pisaster

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

When individuals have higher evolutionary fitness because of being heterozygous at a given gene region, it is known as overdominance. Although overdominant selection could represent an important mechanism for maintaining genetic variation within species, the prevalence of this mode of selection appears to be relatively low. Identification of cases of true single-locus heterozygote advantage are thus useful reference points in our overall understanding of how various forms of balancing selection influence and maintain genetic variation in natural populations. Here we report the apparent long-term maintenance of diversity via overdominant selection with homozygous lethality at an elongation factor locus in the sea star Pisaster ochraceus. Observing this pattern in a gene with such major effects on protein assembly indicates that overdominant selection could be a more prevalent factor in maintaining allelic diversity in the wild than previously recognized.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

A large body of research is dedicated to understanding how genetic variation is maintained (David, 1998; Hunt et al., 2007). Although the phenomenon of heterosis, or ‘hybrid vigour’, is well documented in model systems for promoting variation in quantitative traits (Mitchell-Olds, 1995), a general description of extrinsic factors that promote polymorphism at a single locus remains elusive (Fry, 2004). Given the role of genetic drift in reducing diversity in finite populations, individual polymorphisms should either be transient on evolutionary time scales or actively maintained by selection. Both frequency-dependent and spatially variable selection can maintain stable or balanced polymorphism (Cain, 1954; Gardner & Palmer, 1998; Hedrick, 2006). Less certain is the prevalence of overdominant selection (Hull, 1952; Dobzhansky, 1955; Parsons & Bodmer, 1961; Crow, 1987).

Following the characterization of overdominant selection on haemoglobin alleles in sub-Saharan Africa (Allison, 1964; Hexter & Wiesenfeld, 1968), population geneticists considered overdominance to be a generally important mechanism in maintaining variation in nature (Dobzhansky, 1955). However, with few cases emerging to document heterozygote advantage (Fry, 2004; Gemmell & Slate, 2006) and the growing acceptance of neutral theory as an explanation for polymorphism, overdominance is often considered to exert only a marginal influence on such variation. Nevertheless, theory argues that even a few overdominant loci in a genome may play important roles in evolution (Lewontin et al., 1978; Turelli & Ginzburg, 1983; Peters et al., 2003) including effects on reproductive or life history evolution (Omholt et al., 2000).

Marine species that rely on free spawning of gametes for reproduction potentially face extremely high variance in reproductive success (Hedgecock, 1994; Boudry et al., 2002; Turner et al., 2002). Such conditions may be optimal for evaluating the importance of genetic load on natural systems (Launey & Hedgecock, 2001) by permitting the maintenance of deleterious alleles at high frequency. Such species are often notable for harbouring relatively low levels of genetic variation despite enormous population sizes, suggesting that although selection will play a strong deterministic role in molecular evolution, stochastic events associated with reproductive success may ‘shelter’ deleterious mutants. Although there may be some distinction between strategies that lead to high variance in reproductive success and inbreeding per se, Addison & Hart (2005) noted significantly higher levels of the inbreeding coefficient Fis among marine invertebrates with free-spawned gametes, and inbreeding is associated with the maintenance of deleterious mutations (Frankham et al., 2002). Some species with large, well-mixed populations may in fact be quite prone to effective inbreeding and genetic load (David, 1998; Hoarau et al., 2005).

Pisaster ochraceus is a sea star long known for its critical role as ‘keystone predator’ in the marine intertidal (Paine, 1974). However, little genetic work has been done on this species despite this background. There are few genomic resources for this species, with all previous genetic studies employing allozyme data (Stickle et al., 1992b), AFLP data (M.S. Pankey, unpublished) or mitochondrial DNA (Harley et al., 2006). These studies have indicated that the range of P. ochraceus behaves as a panmictic population with relatively low diversity for an abundant free-spawning species (π = 0.46, about half of what has been reported for other echinoderms with the same reproductive life history; Foltz et al., 2004). We have identified a length-variant nuclear intron polymorphism conserved throughout the geographic range of P. ochraceus. The polymorphism is found in the eukaryotic elongation factor 1-alpha (eEF1a) gene, and preliminary observations indicated the insertion mutant acts as a recessive lethal. Here we analyse data from natural populations and laboratory crosses to examine the potential for overdominant maintenance of variation in natural populations.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

An initial screen of 350 individuals for variation at the eEF1a locus was associated with a phylogeographic survey of P. ochraceus (Harley et al., 2006). Individuals were collected from 17 sites ranging from La Jolla, CA to Juneau, AK (n = 12–30 per site) between August 2003 and July 2004 (Table 1). From each individual, genomic DNA was isolated from ∼3 tube feet following the Puregene protocol (Gentra Systems, Minneapolis, MN, USA). Species-specific primers were designed across the intron–exon boundary based on sequence obtained using the EF1 and EF2 primers described by Palumbi (1996). Using Genbank reference sequence AB070232 (for the confamilial sea star Asterias amurensis) to identify intron boundaries, forward primer (5′-aggctgccgataccttcaa-3′) begins at site 4163 in the exon and the reverse primer (5′-gctagtatctgtttctgtgtgactgc-3′) is situated in the intron beginning at site 4310. The 6-bp insert appears consistently at site 4239, which is 47 bp downstream of the exon–intron boundary. EEF1α genotypes were determined by scoring products of 20 μL PCR reactions [containing 1× PCR buffer, 0.5 U Taq polymerase (Promega, Madison, WI, USA), 1.8 mm MgCl2, 0.25 μm each dNTP (Bioline, Taunton, MA, USA), and 0.3 μm each primer, under standard profile with 50 °C annealing temperature] on 2% agarose. PCR products from n = 8 individuals were cloned (Invitrogen, Carlsbad, CA, USA) and sequenced to characterize the insertion and verify its location.

Table 1.   Observed genotype frequencies of eEF1a(II) across geographic range of Pisaster ochraceus.
Regionn II/II II/IIins IIins/IIinsFreq (II/IIins)Freq (IIins)χ2d.f. = 1 (P-value)Reject HWE?
  1. Genotypes scored on 2% agarose gels reveal only individuals homozygous for absence of intronic 6 bp insert (II/II) and heterozygous for insert (II/IIins), never homozygous for insert (IIins/IIins). Frequencies of each allele are geographically homogeneous; insert allele frequency under binomial distribution is consistent with a frequency of ∼0.24 at each site without significant deviation. Pearson’s chi-squared test indicates that most of the six populations deviate significantly from Hardy–Weinberg equilibrium, whereas across all six regional populations the deficiency of insertion homozygotes is highly significant. The equilibrium expectation if IIins/IIins genotype confers lethality [Hom = p2/(1 − q2), Het = 2pq/(1−q2) = 1 − Hom] is also rejected.

  2. Sites: Juneau, AK: St James Bay. Puget Sound: Alki Beach, Dabob Bay, Fidalgo Is. WA coast: Callam Bay, Kit Peak, Tatoosh Is., Cape Alava, Ruby Beach. OR Coast: Ft Stevens, Cape Arago, Cape Blanco. Northern CA: Indian Beach (Humboldt), Bodega Bay. Southern CA: Hopkins Marine Station, Coal Oil Point, La Jolla.

Juneau, AK (expected)1913 (13.4)6 (5.1) 0 (0.5)0.320.160.670 (0.41)No
Puget Sound (expected)6228 (33)34 (24.4) 0 (4.5)0.540.279.034 (< 0.01)Yes
WA coast (expected)10057 (60.8)43 (34.3) 0 (4.84)0.430.227.284 (< 0.01)Yes
OR (expected)5126 (28.7)25 (19.1) 0 (3.2)0.490.255.276 (0.02)Yes
CA (North) (expected)3315 (17.6)18 (13) 0 (2.4)0.550.274.707 (0.03)Yes
CA (South) (expected)7337 (41)36 (27.4) 0 (4.6)0.490.257.689 (< 0.01)Yes
Overall338176162 00.480.2433.53 (< 0.0001)Yes
Overall expected (lethality) 195.2 (207)123.3 (131)19.5 (0)  11.97 (< 0.001)Yes

Mature sea stars were obtained in April 2006 and genotyped. Gonads were dissected out with forceps after making a small incision with bone shears. Gonads were prepared following suggestions by (Strathmann, 1987a). Two male eEF1a heterozygotes were each crossed with three female heterozygotes and one female homozygote. Eggs were fertilized in 125 mL flasks in 100 mL distilled Instant Ocean (31 ppt) at 16 °C. Larval development was assessed daily under a dissecting scope. Between 3 and 5 days after fertilization (a time period chosen based on the plateau in developmental size at 450–500 μm persisting from approx. 72 h until several weeks later), 100–200 larvae from each cross were placed individually in 5 μL of detergent buffer for DNA extraction as described by Shiurba & Nandi (1979), and genotyped as described above.

A larger fragment of the EF1a gene was amplified using universal primers (Palumbi, 1996). To obtain the most accurate sequence data, reactions used AccuPrime High Fidelity polymerase (Invitrogen) with AccuPrime Buffer II. Reaction conditions were otherwise as before, with an annealing temperature of 50°. Amplicons from 20 individuals were pooled and cloned via Topo4.1 TA (Invitrogen) vector; from each pooled set of reactions 48 clones were picked, cultured and sequenced in both directions. We did this for two sets of pooled reactions, one from a ‘northern’ population (Alki Beach, WA) and one from a ‘southern’ population (Bodega Bay, CA).

Universal-primer reactions were also repeated with high-fidelity polymerase for six individuals that had been previously genotyped using our intron-specific primers. These reactions were cloned with the same methods and 32 clones were picked and sequenced from each individual. These data were used to more precisely analyse patterns of linkage disequilibrium and recombination and to confirm the single-locus status of this marker. Using a reported error rate for the Accuprime high-fidelity polymerase of 0.14%, we calculated the ‘effective bases sequenced’ for each individual as the number of clones sequenced times the mean length of the fragment. An expectation for the number of ‘error’ mutations is then the ‘effective’ length times the error rate; singleton mutations (only found once in the data set for an individual) were counted and compared with this expectation. Substitutions with low PHRED (Ewing & Green, 1998) quality scores (< 20) were not scored.

Polymorphic sites, including the insertion documented here, were examined for linkage disequilibrium using DnaSP v.4.10.4 (Rozas & Rozas, 1999). The insertion region was recoded as an additional nucleotide polymorphism for the purpose of this test; all polymorphic sites in 757 bp sequences generated from the pooled reactions using the universal EF1 primers (and spanning the intron) were included. Significance was examined with Fisher’s exact test, with Bonferroni correction. With the data pooled from multiple individuals, we estimated the number of segregating sites (S) within the population of mutant ins alleles and the population of wild-type alleles separately, as well as the mean pairwise nucleotide differences among sequences (π). Diversity levels were compared for the two populations and Tajima’s D was calculated to determine whether the allelic variation associated with the ins mutation indicated the age of the allele and/or the recent pattern of frequency change for this allele. Estimates of the selection coefficient were determined using observed allele frequencies as representation of an equilibrium frequency in an overdominant system, s1/(s1 + s2), where one selection coefficient is 1 (the lethal homozygote).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Our survey recovered two length variant alleles in a short fragment of eEF1a. Both variants were present at high frequency throughout the species’ range (Table 1). Although individuals heterozygous for the length variant were well represented in the survey, one of the two homozygous genotypes was absent in the 338 individuals scored, although 20 were expected under Hardy–Weinberg equilibrium. Sequence data reveal that heterozygotes carry an allele (IIins) containing a 6-bp insertion within an intron located ∼47 bp downstream of the intron–exon boundary at the 3′-end of exon II (Fig. 1). The region does not match any of the canonical splice-site sequences (Black, 2003), but failure to excise the intron could lead to a damaged or altered-function copy of the eEF1a transcript.

image

Figure 1.  (a) Location of screened region, sitting on upstream region of an intron separating exons II and III of the eEF1a gene (cd03693.1, pfam03144.12). Amino acid sequence shown from data to indicate homology. (b) Evidence of diploid genotype from genomic DNA on 2% agarose gel for four individuals from Bodega Bay, California; the third lane is a homozygote, whereas the other lanes are heterozygous for the IIins allele.

Download figure to PowerPoint

Subsequent crossing experiments confirmed that this homozygous condition (IIins/IIins) confers lethality in early development (< 5 days, Table 2). Larvae from six heterozygote (II/IIins × II/IIins) crosses and two mixed (II/II × II/IIins) crosses were genotyped within 120 h following fertilization. The genotype ratios of the progeny were considered under normal Mendelian ratios as well as under a ‘lethality’ model that assumes the IIins homozygote never survives. Although all crosses significantly deviated from normal Mendelian ratios, they matched the prediction of the modified lethality model in five of six heterozygote crosses. However, summing across all heterozygote crosses, there is a small yet significant excess of larval II/II homozygotes (< 0.001; < 0.05 if cross d in Table 2 excluded). Using observed population frequencies of the IIins allele (Table 1), the selection coefficient against the nonlethal homozygote appears to range from 0.2 to 0.37. A similar result was obtained from simulated populations, in which selection coefficients of at least 0.21 were required to consistently maintain polymorphism in this system (J. P. Wares, unpublished).

Table 2.   F1 genotype ratios.
 Observed II/II :  het : IIins/IIinsExpected (Mendelian)χ2 (P)Expected (Lethality)χ2 (P)
  1. Experimental crosses confirm that the IIins homozygote cannot survive up to 5 days past fertilization. Larvae from six heterozygote (II/IIins × II/IIins) crosses and two mixed (II/II × II/IIins) crosses were genotyped up to 120 h post-fertilization. The genotype ratios were considered under normal Mendelian ratios as well as the expected progeny ratio if the IIins homozygote never survives; significant deviations after Bonferroni correction are noted by asterisk.

Heterozygote crosses
 a58 : 79 : 034.25 : 68.5 : 34.2552.3 (< 10−12)*45.7 : 91.3 : 0 5.0 (< 0.08)
 b39 : 62 : 025.25 : 50.5 : 25.2535.4 (< 10−8)*33.7 : 67.3 : 0 1.3 (< 0.53)
 c37 : 81 : 029.5 : 59 : 29.539.6 (< 10−9)*39.3 : 78.7 : 0 0.2 (< 0.90)
 d90 : 91 : 045.75 : 90.5 : 45.7589.5 (< 10−20)*60.3 : 120.7 : 021.9 (< 0.00002)*
 e63 : 100 : 040.75 : 81.5 : 40.7557.1 (< 10−13)*54.3 : 108.7 : 0 2.1 (< 0.35)
 f76 : 115 : 047.75 : 95.5 : 47.7568.5 (< 10−15)*63.7 : 127.3 : 0 3.6 (< 0.17)
Homozygote crosses
 g27 : 20 : 023.5 : 23.5 : 0 1.0 (< 0.59)  
 h23 : 48 : 035.5 : 35.5 : 0 8.8 (< 0.012)*  

Within the gene fragment surveyed, linkage decay takes place over very short (200–300 bp) reaches of this gene fragment, with estimated |D| = 0.0342–0.0585 kbp. Examining linkage disequilibrium at this locus across all clones, we found only two sites were in significant LD with the insertion polymorphism, both of which are within the intron region about 250–350 bp downstream. In the set of gene trees generated from allelic diversity in the pooled populations (using high-fidelity polymerase), there are no clear patterns of relationship among individuals from distinct geographic regions, or among alleles from the two allelic classes (although one clade is predominantly ‘wild type’ and the other is mostly the IIins allele; results not shown). Parsimony analysis of these pooled data suggests that most substitutions are not homoplasious, but the insertion region appears to recombine with allele classes defined by other polymorphic sites. The potential for PCR-mediated recombination (Cronn et al., 2002) becomes apparent when larger samples of sequenced clones (again, using high-fidelity polymerase) from six individuals (three individuals scored as homozygotes and three as heterozygotes) are considered; for each individual, the minimum number of detected recombination events is 1 and the recombination parameter R is estimated at 2.5 (0.0028 between adjacent sites). The most apparent case of recombination is a single individual that was genotyped as a homozygote (and has no sequenced insertion) but carries an allele similar to the other ‘wild-type’ alleles as well as an allele that harbours the distinguishing SNPs of the IIins allele. In the three heterozygotes, the recombination tracts are clearly apparent as the two allele classes otherwise differ by at least seven nucleotide positions and their distinction is supported by > 90% bootstrap support (tree not shown). Sequence logos representing sequence diversity and divergence in the pooled fragments with and without the insertion are shown in Fig. 2.

image

Figure 2.  Sequence logos (Schneider & Stephens, 1990) generated by WebLogo v.2.8.2 (Crooks et al., 2004) represent the wild type (top, = 64) and IIins allele (bottom, = 70), with sequence position and sites of strong frequency differences indicated with stars (indel at sites 378–383 in alignment shown). All constant and parsimony uninformative sites removed from consideration. Phylogenetic reconstruction indicates considerable homoplasy in these data, but resolve two divergent clades (not shown).

Download figure to PowerPoint

Determination of gene copy number under conditions of high rates of PCR-mediated recombination and/or mutagenesis is difficult, but our data appear to support amplification of only a single gene region. From each of the three homozygotes, between 19 and 31 clones were obtained. In each case two sequenced alleles could be identified, but across all cloned sequences as many as 15 ‘singleton’ SNPs (found in only a single sequence) were recovered (approx. 34.5 kbp effectively sequenced, with expectation from published error rates of up to 48 artefactual mutations). Of the three heterozygotes cloned, two clearly showed only two alleles, with the insertion and wild-type allele differing by up to 12 substitutions (depending on how mutations counted). Similarly, there were as many as 17 singleton mutations (with expectation of up to 48 for an effective 34.5 kbp sequenced per individual). In the third heterozygote, these same characteristics and alleles were found but three of 15 ‘insertion’ clones shared three SNPs not found in either of the other two allele classes. These SNPs were not found in any of the other presumed alleles from the other five individuals cloned, suggesting further artefact.

In addition to the substantial divergence between the two allele classes, II and IIins (dA = 0.0094 ± 0.0019), we evaluated patterns of diversity within each class. From a data set developed from pooled tissue samples from multiple geographic populations, we found no significance in diversity between the two allele classes or the geographic locations (Puget Sound, two sites; northern California, two sites) they came from (data not shown). Comparing the two allele classes, Tajima’s D is −1.354 (n.s.) for the wild-type allele class, and 0.419 (n.s.) for the insertion class. Overall D for the entire data set is −0.870 (n.s.), although the area about 250–350 bp downstream of the insertion – where most of the diagnostic substitutions between the two classes are harboured – does have significantly high (< 0.05) D (ranging from 2.26 to 2.47) in a 100-bp sliding window analysis with 25 bp steps.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Overdominance is one of several ‘balancing selection’ hypotheses that could account for the maintenance of the homozygous-lethal insertion allele (Gemmell & Slate, 2006; Hedrick, 2006). For example, previous work has shown that strong selective clines may operate in coastal systems (Karl & Avise, 1992; Gardner & Palmer, 1998; Sotka et al., 2004), and spatial variation in selection pressure could control this polymorphism. Analytical and numerical estimates for the strength of selection (s) in this case are > 0.2, relatively strong but consistent with estimates of s in other natural systems (Kingsolver et al., 2001). However, previous surveys of this species suggest it is effectively panmictic across a large geographic range (Harley et al., 2006) and no populations reveal deviating frequencies of the lethal allele (Table 1), arguing against the likelihood of spatially variable selection. Temporal variation is also possible but difficult to detect (Hedrick, 2006); with a lethal allele this scenario would still require overdominance at least some of the time, and even well-studied cases of apparent overdominance may be conditional. Additionally, meiotic drive could skew allele frequencies (Lyttle, 1991) but the Mendelian ratios of surviving genotypes in our crosses (i.e. 1 : 1 ratio in homozygote–heterozygote crosses indicate normal frequency of syngamy) generally refute this possibility (Table 2). Although a small but persistent homozygote excess is recovered in the heterozygote crosses, this may be attributed to a number of artefacts including the potential for scoring bias on agarose gels, or a laboratory environment that does not reflect the natural selective environment. It is possible that selection on either viable genotype varies across life stages, and that our results could vary in longer experimental trials. If homozygotes are advantaged in the earliest stages of development, there still appears to be a net fitness advantage to heterozygotes at this locus that is consistent with the high frequency of heterozygotes (and the lethal allele) in natural populations.

We remain cautious about interpretation of this system as a single-copy phenomenon. Although linkage decay is quite rapid in our gene fragment and extensive cloning data from both genotypic classes largely conforms with expectations for a single-copy marker, the eEF1a gene itself may be multi-copy (Silar et al., 2001). Our genotypic analysis of wild and larval populations used a species-specific set of primers with one primer placed in the highly variable intron region, rather than universal primers employed in our cloning and sequencing efforts. In addition, a multi-copy system cannot easily account for the observed 1 : 2 : 0 ratio of offspring genotypes in heterozygous crosses; in the case of multiple gene copies we would expect to observe the insertion homozygote (although at low frequency) and a far higher excess of heterozygotes in the large sample of larvae generated through crosses.

Considering the apparent homozygous lethality of this system, overdominance remains a probable explanation although more work must be done to characterize the gene and specific associations between segregating polymorphisms and fitness (Gemmell & Slate, 2006). Epistatically mediated overdominance seems improbable because of the complete absence of viable IIins homozygotes (Kojima, 1959; Omholt et al., 2000) in the wild. Likewise, associative overdominance (physical linkage to another segregating polymorphism; Ohta, 1971; Shaw & Chang, 2006) is not evident given both the high rates of linkage decay across the sequenced region and the additional expected heterozygote excess for a multi-copy system. Recurrent or recombination-driven mutation is a possibility that is difficult to fully evaluate given the high rate of apparent PCR-mediated artefact in our sequence data; however, the considerable divergence among the two allele classes indicates that the polymorphism has long persisted in this large natural population.

Although theory predicts high frequency of deleterious alleles in large, out breeding populations under mutation-selection balance (Glémin, 2005), it is unusual to find such a population harbouring high numbers of a particular lethal allele at one locus. Cases of artificial selection (Gemmell & Slate, 2006) and inbreeding (Krüger et al., 2001; Frankham et al., 2002) are possible mechanisms for generating high frequencies of significantly deleterious alleles and/or phenotypes. High levels of inbreeding may seem unlikely for Pisaster, yet similar species with free-spawned gametes often exhibit surprisingly high inbreeding coefficients (Addison & Hart, 2005), and high variance in reproductive success among free-spawning males can cause strong deviations from HWE in the wild (Purser, 1966). Stickle et al. (1992a) presented data from eight allozyme loci suggesting a relatively high inbreeding coefficient (Fis = 0.088) and an overall deficiency of heterozygotes in three populations of P. ochraceus. If high variance in reproductive success generates a lower effective population size [and estimates of θ(π) = 0.46 from mitochondrial sequence data suggest Ne is on the order of 105; Harley et al., 2006], the stochastic effects of drift could reduce genetic diversity even at loci under balancing selection. However, Robertson (1962) calculated that heterozygote advantage can mitigate the effects of drift in populations of moderate effective population size given the deleterious allele frequency > 0.2 (mean frequency of IIins is approximately 0.23; Table 1).

The eEF1a product is responsible for binding aminoacyl-tRNAs to ribosomes (Riis et al., 1990) among other functions (Ejiri, 2002), making complete loss of function a lethal condition, yet mutants have been shown to exacerbate or rescue deleterious phenotypes caused by other loci (Tapio & Isaksson, 1990; Kinzy & Woolford, 1995). We do not yet know if the insertion analysed here has an effect on protein translation or expression because of incorrect or alternative splicing. Differential copy number has been shown to have effects on lifespan in flies (Stearns, 1993), but these were ‘addition’ experiments. Preliminary quantitative RT-PCR expression analysis of viable homozygous and heterozygous individuals suggests greater variance in eEF1a expression among individuals than among genotype classes (data not shown); this locus is considered a poor candidate for expression quantification assays because of variable expression across different life stages as well (Yan & Liou, 2006). Intron evolution (e.g. gain and loss) is also unusually rapid at the eEF1a locus, particularly in deuterostomes (Wada et al., 2002).

We conclude that heterozygotes for the mutant insertion at this locus have increased fitness, but have little information with which to identify the selective force maintaining this polymorphism. It is not yet known if the fitness difference is due to decreased viability in the homozygotes or increased fecundity in the heterozygotes. Presumably, if the heterozygote fitness advantage is due to functional hemizygosity at this locus, the lethal allele may persist through heterozygous individuals by acting either as a pseudogene or as a novel functional or regulatory alternative (i.e. alternative splicing). Such characterization will require assessing polymorphism between the two allele classes across a larger genome region as well as obtaining evidence of an alternative eEF1a transcript. It is also not clear how a deleterious form of such an important functional gene is maintained at the haploid gametic stage of this free-spawning species (see Grosberg & Strathmann, 1998), although Silar et al. (2001) note that while deletion of this gene is lethal, it is not required for gamete formation or fertilization.

Historically, cases of overdominance are easier to establish in model agricultural or experimental systems, where the selective agent – often disease – is quickly identified (although the mechanism may be unknown even for ‘classic’ cases; Gemmell & Slate, 2006). The strength of the selective agent aids identification (e.g. haemoglobin genotype and vulnerability to malaria, Hexter & Wiesenfeld, 1968; cystic fibrosis and cholera susceptibility, Gabriel et al., 1994; warfarin-resistant rats, Greaves et al., 1977 and congenital disorders of glycosylation, Freeze & Westphal, 2001). For the case of P. ochraceus, identifying the selective agent acting on the eEF1a locus will be difficult because of their long generation time (Strathmann, 1987b). Nevertheless, we believe this system represents a good opportunity to advance our understanding of marine disease ecology (Mydlarz et al., 2006), particularly given the recent publication of the first echinoderm genome (Sodergren, 2006). Previous studies have established a high incidence of necrotic disease (Eckert et al., 1999) and protozoan parasitism (Leighton et al., 1991) in this species. Considering recent echinoderm epidemics and widespread die-offs on the west coast of North America (Becker, 2006), disease could exert a substantial selective force on Pisaster. Given the prevalence of pathogen resistance in earlier studies of overdominance, we believe this to be a probable explanation for the maintenance of the described eEF1a polymorphism.

Although few such cases have been identified (Mestres et al., 2001; Mead et al., 2003; Wilder & Hammer, 2003), overdominance has often been credited for the variation seen in some of the first proteins characterized in diverse taxa, including humans (Hexter & Wiesenfeld, 1968), pigeons (Frelinger, 1972), and – in this study – sea stars. Overall, genes under balancing selection (including overdominance) are typically more difficult to detect in genome screening than those under directional selection (Beaumont & Balding, 2004), although some remarkable discoveries regarding overdominance are currently being made in this approach (Alonso et al., 2008). We are left to consider how much of the allelic variation observed in natural systems is transient (i.e. caused by neutral forces; Barrett & Schluter, 2007) and how much has persisted for its ecological or physiological relevance via selection (Hahn, 2008).

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We appreciate the expertise and insights provided by D. Hall, R. Grosberg, B. Cameron, D. Promislow, M. Turelli, W. Anderson, K. Dyer, and members of the Wares Lab at the University of Georgia. We appreciate assistance by C. Harley, E. Sanford, and A. Holloway in collecting specimens. Insightful comments by two anonymous reviewers greatly improved the manuscript. The University of Georgia Research Foundation funded this research.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References