|Home | About | Journals | Submit | Contact Us | Français|
We report our characterization of the genetic variation within and differentiation among wild-collected populations of the red flour beetle, Tribolium castaneum, using microsatellite loci identified from its genome sequence. We find that global differentiation, estimated as the average FST across all loci and between all population pairs, is 0.180 (95% confidence intervals of 0.142 and 0.218). A majority of our pairwise population comparisons (>70%) were significant even though this species is considered an excellent colonizer by virtue of its pest status. Regional genetic variation between Tribolium populations is 2–3 times that observed in the fruit fly, Drosophila melanogaster. There was a weak positive correlation between genetic distance [FST/(1 − FST)] and geographic distance [ln(km)]; pairs of populations with the highest degree of genetic differentiation (FST > 0.29) have been shown to exhibit significant postzygotic reproductive isolation when crossed in previous studies. We discuss the possibility that local extinction and kin-structured colonization have increased the level of genetic differentiation between Tribolium populations.
The red flour beetle, Tribolium castaneum, is a cosmopolitan human commensal and a common inhabitant of stored products, including cereals, baking flour, and livestock feed (Park 1962). Experimental crosses between 22 pairs of T. castaneum populations showed evidence of nascent reproduction incompatibilities, uncorrelated with geographic distance (Demuth and Wade 2007a, 2007b). In this study, we characterize the extent of the genetic differentiation among wild-collected populations of T. castaneum, some shared with the Demuth and Wade (2007a, 2007b) studies. We use microsatellite loci and investigate to what extent, if any, genetic differentiation is correlated with the observed degree of reproductive isolation. The earlier work (Demuth and Wade 2007a) found a weak, positive correlation with genetic distance using mitochondrial markers. Because our microsatellite markers are nuclear rather than mitochondrial and are much more variable, they allow us a more robust estimate of genetic distance between populations on a shorter time scale and a second, independent test for an association between reproductive isolation and genetic distance.
Genetic differentiation can occur in multiple ways, including both neutral differentiation in the absence of gene flow and adaptive differentiation in response to selection in different environments. It is the latter that is believed to result most often in reproductive incompatibilities between genetically divergent populations (Charlesworth et al. 1987). In the absence of the homogenizing effects of gene flow, 2 populations may take separate evolutionary trajectories in adaptive response to local environments and, if dispersal is geographically restricted, this differentiation manifests itself as isolation by distance. Divergent selection in differing environments may cause genetic differentiation by eliminating an allele from 1 population in 1 environment while fixing it in another population in the other environment. The biology of T. castaneum makes it difficult to predict the amounts of divergent selection and gene flow. Unassisted, Tribolium are notoriously poor migrators with 1 study showing that T. castaneum can only stay air borne for a maximum of 20 s (Good 1933). Although T. castaneum is found wherever grains are stored, it is unable to persist on uncracked grain, relegating its habitat to human stores of processed grain or cohabitation with other, often larger, boring insects. With the advent of agriculture and the subsequent human population expansion, flour beetles were afforded amiable habitats across the globe. Stored processed grain is an abundant yet relatively homogenous medium that is marketed worldwide. The environmental homogeneity of processed grain makes divergence by differential selection relatively unlikely and the likelihood of widespread human dispersal through the grain trade appears to ensure a relatively high level of gene flow. Both factors mitigate genetic differentiation between populations. Nevertheless, evidence from interpopulation hybridization indicates some genetic isolation among local populations with significant measurable effects on hybrid viability and fecundity.
The purpose of this study is to examine the population genetics of T. castaneum using microsatellite markers identified from its genome sequence (Tribolium Genome Sequencing Consortium 2008). We characterize the extent to which globally distributed populations are genetically differentiated and whether the observed differentiation is associated with geographic distance and/or intrinsic reproductive isolation.
All samples originated from collections of more than 50 adults. Each laboratory stock was established and maintained at a population size of >200 individuals on standard medium (20:1, flour:brewer's yeast, by weight) less than 24 h darkness, 28 °C, and approximately 70% relative humidity. Collection date and geographic location are given in Table 1. The populations, AdMO, BlIN(1), and JzM, were collected at grocery stores from flour or processed products (like pancake mix) intended for human consumption. The DxMO, Bh-I, and Go-IN populations were collected at granaries, where whole grains are processed into finely ground flour. Populations BlIN(2), BaE, DesT, and LiP were collected from markets and BoFL was collected from a pet store. The populations, RcMS and WLIN, were collected from livestock feed. The Chicago Standard Mixture (cSM) population is a laboratory stock created by mass mating the 4r strains of this species that Park used in his classic competition experiments (Park 1962; Park et al. 1964; for a detailed description of cSM, see Wade 1976). The founding stock from which the 4 Park strains were derived originated in Brazil. (It may well be that populations collected from different venues [e.g., granaries vs. livestock feed] experience different degrees of gene flow and may be isolated from one another; but, absent genetic studies, this would be a difficult avenue to pursue and we do not do that here.)
Genomic DNA was isolated using a cetyltrimethyl ammonium bromide extraction protocol (Doyle JJ and Doyle JL 1987). Genotypes were determined using 15 nuclear microsatellite loci (Table 2) taken from Demuth et al. (2007). Microsatellite loci were selected based on genetic variability and distribution across the genome (Tribolium Genome Sequencing Consortium 2008). There is at least 1 microsatellite on 8 of the 10 chromosomes, with all but 1 positioned within introns or intergenic regions (see Appendix B). Fragment sizes were scored electronically using an ABI 3730 DNA analyzer and genemapper version 4.0 (Applied Biosystems, Foster City, CA) as well as by electrophoresis through a 4.5% agarose gel (Nuseive) and called independently by Douglas W. Drury and Ashley L. Siniard.
Observed (Hobs) and expected (Hexp) heterozygosity were estimated using Arlequin (Schneider et al. 2000) for each population-by-locus combination and then averaged over all loci to get population estimates. Deviations from Hardy–Weinberg equilibrium (HW) were assessed using the method of Guo and Thompson (1992) for each locus-population combination using a Markov chain of 100000 steps and 1000 dememorization steps.
Null alleles occur when markers consistently fail to be detected in certain populations when there is 1) inconsistency in template DNA, 2) divergence in primer-binding sites, or 3) an overwhelming size discrepancy between alleles preferentially favoring the smaller allele. The occurrence of undetectable genotypes results in individuals with no genotypic information at a locus (homozygote nulls) and an overabundance of scored homozygotes (heterozygote nulls). The failure to account for null alleles results in an underestimation of within-population genetic diversity and thus an overestimation of differentiation between populations (Avise and Dakin 2004). We estimated the potential frequency of null alleles with the expectation maximization algorithm of Dempster et al. (1977) using FST Refined Estimation by Excluding Null Alleles (FreeNA) (Chapuis and Estoup 2007). We use the mean expected heterozygosity as a measure of genetic variability within-populations because it is robust to the presence of null alleles. To verify that null alleles were not occurring because of the inability of the fluorescently labeled adapter primer to be incorporated into the polymerase chain reactionproduct, we confirmed the electronically scored allele sizes by running product out on a 4.5% agarose gel. Allele calling was performed independently by Ashley L. Siniard and Douglas W. Drury and checked for conformation with electronic annotation.
We evaluated the genetic differentiation among populations and among loci by calculating locus-specific and population pairwise FST's excluding null alleles ([ENA] algorithm as implemented in FreeNA). Pairwise differentiation was also evaluated using Fisher's Exact tests implemented in GENEPOP (Raymond and Rousset 1995). We investigated isolation by distance with Mantel tests implemented in GENETIX (10000 permutations), using a semimatrix of ratios, FST/(1 − FST) with ENA corrected FST's, and a semimatrix of ln transformed geographic distance (Belkhir et al. 2004). We used the Bonferroni correction of 0.05 divided by the number of tests to correct for the multiplicity of comparisons (see Table 3 and Appendix A). We used R to plot the regression of FST/(1 − FST) on ln(km) and generate 95% confidence bands and prediction intervals (Ihaka and Gentleman 1996).
Pairwise Cavalli-Sforza and Edwards’ chord distance measures (1967), DC, which are robust to the presence of null alleles, were calculated in POPULATION (Langella 1999) from the genotype data (including null alleles algorithm implemented in FreeNA (Chapuis and Estoup 2007). The resulting distance matrix was used to reconstruct the phylogenic relationships using the neighbor joining method. Node stability was assessed by 10000 bootstrap replications (over loci). The tree was visualized using TREEVIEW (Page 1996).
A total of 57 alleles were observed at the 15 loci. The number of alleles per locus ranged from 2 alleles for loci 34G3 and 32H3 to 6 alleles for locus LG9F3. Observed heterozygosity ranged from 0 for alleles fixed within a population to 0.833, most individuals were heterozygotes. Of 196 locus-by-population combinations, only 1% (2) show a significant deviation from HW after Bonferroni correction for multiple comparisons (P < [0.05/196] = 0.00025). All populations contained at least 1 individual that failed to amplify at 1 locus. Locus 34E3 failed to amplify from every individual of the cSM population, raising the possibility that it could be fixed for a null allele. Overall, the mean frequency of segregating null alleles in each population ranged from 0.21 to 0.52. Expected heterozygosities varied significantly among populations from a low of 0.21 (standard error of the mean, SEM ± 0.07) to a high of 0.53 (SEM ± 0.09).
Most pairs of populations differed significantly at most loci with Bonferroni correction for multiple comparisons (P < [0.05/91] = 0.0005; 70 of 91 comparisons, including 6 cases with borderline significance 0.0001 < P < 0.0005). Pairwise FST ranged from 0.0289 to 0.353 (WlIN/BlIN(2) and BhI/cSM comparisons, respectively; Table 3). Global FST, calculated as the average pairwise FST for all loci and population pairs, was 0.180 with a 95% confidence interval of 0.142–0.218. Figure 1 depicts the relationships among populations using the Cavalli-Sforza and Edwards chord distances. Based on this tree, there is no obvious clustering of populations. This suggests that we are not characterizing previously, undescribed cryptic species, although populations from disparate ends of the clade show significant postzygotic reproductive isolation when crossed (see Discussion; Demuth and Wade 2007a, 2007b).
Although positive, there was no significant correlation between geographic distance ln(km) and genetic distance (Mantel test, R2 = 0.09, P > 0.05). A plot of genetic distance versus geographic distance (Figure 2) reveals that 5 population comparisons (AdMO × BhI, BoFl × BhI, BaE × BhI, GoIN × BhI, and AdMO × DxMO) lie outside the 95% prediction intervals. The Indian population represents the majority (4 of 5) of significant deviations from the expected genetic/geographic distance relationship. Evidence presented elsewhere (Drury DW, in preparation; Demuth and Wade 2007a, 2007b) supports the inference that the Indian beetle populations are undergoing a speciation event relative to the remainder of the species distribution (Thomson and Labonne 1998; Thomson and Beeman 1999). The only comparison not involving BhI to lie beyond the 95% predictive interval is the comparison with the second smallest interpopulation geographic proximity, AdMo × DxMO. Although AdMO and DxMO were collected only 25 miles apart, their deviation could be explained by venue of collection. The DxMO population was collected from a granary where local farmers bring their products to be milled into livestock feed. The AdMO population was found residing among bags of commercial pastry flour. Therefore, the true origin of the AdMO population could reside at or near any of the product distribution centers, unlike the case of the DxMO where transmission is assumed only to be among local farms.
The data show that the species T. castaneum exhibits a fair amount of genetic structure from the perspective of essentially neutral microsatellite markers. The value of FST, averaged over markers and populations, lies between the confidence limits of 0.142 and 0.218. Considering only the North American populations, the average pairwise FST among the 4 Indiana populations (BlIN(1), BlIN(2), GoIN, and WLIN) is 0.089 and that between the 2 Missouri populations (DxMO and AdMO) is 0.224; the average genetic differentiation between the 2 states is intermediate, 0.127. If we compare Africa (DesT), Central America (JzM), South America (LiP), and North American (DxMO), we find the average pairwise FST to be 0.090. Thus, there is as much or more regional variation among the Indiana or Missouri populations as there is among populations much more geographically distant from one another. This level of regional genetic variation in the red flour beetle is 2–3 times that observed in the fruit fly, Drosophila melanogaster, where the average FST across Australia is 0.054 and across the United States is 0.036 (Turner et al. 2008). Our global FST estimate of 0.180 is within the range of what has been seen in the Coleoptera; in a survey of genetic differentiation using allozymes, among populations in 7 beetle species, McCauley and Eanes (1987) found FST to range between 0.030 and 0.154. McCauley and Eanes (1987) also surveyed many other non-Coleopteran insects and found FST to range from 0.003 to 0.380. More recent beetle studies have found a range of genetic differentiation among populations of agricultural beetle pests, including the Western corn rootworm (Coleoptera: Chrysomelidae; FST = 0.006; Kim and Sappington 2005) and the Boll weevil (Coleoptera: Curculionidae; FST =0.241; Kim and Sappington 2006). Substantial heterogeneity in the degree of genetic differentiation also observed among species within the same genera, for example, bark beetle species (Dendroctonus ponderosae: FST = 0.030, Mock et al. 2007; Dendroctonus mexicanus: FST = 0.104, Zuniga et al. 2006) and Carabid beetle species (Agonum elongatulum: FST = 0.003, Platynus tenuicollis: FST = 0.27; Liebherr 1988).
We point out that the laboratory strain, cSM, has reduced levels of heterozygosity relative to the majority of the wild populations surveyed, ranking 13th of the total of 14 surveyed (Table 1). This confirms the earlier finding of Wade (1990, 1991) using quantitative genetic methods that natural populations of T. castaneum were more variable phenotypically and genetically for rate of population increase than the laboratory hybrid strain-d, cSM. A majority of our pairwise population comparisons, 70%, indicate significant genetic differentiation even though Tribolium beetles are considered excellent colonizers as per their “pest” status. The processes of local extinction and recolonization can increase the genetic differentiation variation among groups (McCauley and Wade 1980, Wade 1982; Wade and McCauley 1984), but the effect depends critically on the mode of colonization (Slatkin 1985; Wade and McCauley 1988; Whitlock and McCauley 1990). If a majority of the existing populations contribute migrants to newly colonized populations, then colonization represents increased gene flow, and the among-population genetic diversity decreases. However, when colonists are derived from only a single population or from a small group of relatives within a population (kin-structured colonization; Wade and McCauley 1988; Whitlock and McCauley 1990; Wade et al. 1994), migrants founding a new population can share high levels of genetic similarity and an array of new populations, each descended from such related colonizing propagules, will exhibit high levels of genetic differentiation one from another. Our genetic evidence suggests that, despite being good colonizers, Tribolium populations are likely to be founded by genetically related beetles from a single source. A single female can store the sperm of multiple males for 2–3 months and on finding a suitable substrate can produce upwards of 200 offspring. This facet of Tribolium biology, along with extremely poor long-distance flight ability, lends credence to the propagule model of colony foundation (Wade and McCauley 1984).
Classical genetic studies with flour beetles have revealed heterogeneous distributions of cryptic phenotypes. One of these phenotypes is associated with maternally acting selfish genes, termed MEDEA, which spreads by killing embryos that fail to inherit the element (Beeman et al. 1992). MEDEA elements are distributed heterogenously across the globe but are not found in India or Australia (Beeman and Friesen 1999). In a survey of the southern United States, 29 of 50 populations collected and tested contained MEDEA but the factor was less frequent in beetles collected below 33°N latitude (Beeman 2003). However, MEDEA is completely absent from Indian populations. MEDEA containing beetles are completely reproductively incompatible when females are M+ and males are from Indian populations (Thomson and Labonne 1998; Thomson and Beeman 1999). This incipient reproductive isolation is the first stage in the conversion of genetic variation within a species into permanent variation between 2 separate species.
We detected a positive but nonsignificant correlation between genetic distance [FST/(1 − FST)] and geographic distance (ln(km); Figure 2). Interpopulation differentiation in these beetles, on a global scale, is likely dictated by human traffic. Unlike most nonpest organisms, flour beetles can traverse an ocean by occupying a vessel's food stores. Therefore, we would expect population differentiation on a geographic scale to be determined by the frequency of human transport between locations, the number of populations contributing to new colonies, and the extent to which gene flow is prevented by intrinsic barriers, that is, reproductive isolation.
Among these population pair comparisons, 4 are particularly noteworthy. The 4 comparisons identified statistically as outliers (see shaded and numbered points in Figure 2), correspond to comparisons between the Indian population and 3 North American and 1 Ecuador population. Using the regression line in Figure 2, populations genetically differentiated to this degree should be more than 65 × 106 km apart. Demuth and Wade (2007a) and Thomson and Beeman (1999) found high levels of postzygotic reproductive isolation between the Indian and Ecuadorian and Indian and North American strains. Thus, the high degree of genetic differentiation (FST > 0.29) is coincident with significant postzygotic isolation.
National Institutes of Health (5R01GM65414-4 to M.J.W.).
We thank J. P. Demuth, T. E. Cruickshank, A. N. Brothers, and Y. Brandvain for their helpful comments on the manuscript and B. T. Furomoto and M. E. Whitesell for assistance in the laboratory.
|No of alleles||3||2||3||3||3||3||3||4||3||3||2||3||3||3|
|No of alleles||3||1||2||3||1||3||3||2||3||2||2||2||3||1|
|No of alleles||3||2||1||3||3||2||3||3||3||4||1||3||3||2|
|No of alleles||3||1||1||2||3||NA||2||4||2||2||2||3||2||4|
|No of alleles||3||3||1||3||3||2||4||2||3||4||2||3||3||3|
|No of alleles||1||3||1||1||1||1||2||1||1||1||1||1||1||1|
|No of alleles||2||2||2||2||2||2||2||2||2||3||1||2||2||2|
|No of alleles||2||2||2||1||1||1||2||1||1||2||1||1||1||1|
|No of alleles||2||2||1||3||1||1||2||2||2||1||1||2||2||3|
|No of alleles||2||2||1||3||2||3||3||1||2||2||2||2||2|
|No of alleles||2||2||2||2||2||1||2||1||1||2||2||2||2||2|
|No of alleles||2||1||2||2||3||2||2||3||3||2||2||2||1||2|
|No of alleles||2||2||2||2||2||1||2||2||2||3||1||2||2||2|
|No of alleles||1||2||1||2||2||2||2||2||2||2||2||2||2||2|
|No of alleles||1||1||2||1||1||1||2||1||2||3||1||1||1||1|
M, populations monomorphic for allele type.
NA, no. of alleles amplified.
NS, not significantly different from Hardy–Weinberg expectation with Bonferroni correction for multiplicity of comparisons (0.00025 = [0.05/196]).
|Locus||Linkage group||Genomic context||Motif||Number of repeats||Chromosomal positiona|
|32A3||1 (x)||Intron 1 of “Dual oxidase”||tat||12||805 638|
|32D7||3||Intron 2 of “Glutamine:fructose-6-phosphate aminotransferase 1”||taa||17||302 244|
|32E7||4||Intron 17 of “CG1516 Isoform E”||ttta||5||121 805|
|32F3||4||Intron 4 of a conserved hypothetical protein||a||13||33 750|
|32H7||5||Intron 8 of “rho-associated protein kinase 1”||a||14||1 445 763|
|32H3||5||Intron 7 of “neural-cadherin precursor”||aat||8||713 852|
|34E3||8||Intron 5 of “CG1486 isoform A”||tta||6||698 301|
|LG9B7||9||Intron 1 of “Teashirt”||t||14||66 579|
|LG9F3||9||Intron 8 of “vacuolar protein sorting 13D”||a||13||765 219|
|34H3||10||Second exon of “CG10936”||Tct||7||416 449|
Information based on Tribolium castaneum genome NCBI Build 2.1.