Introduction
Liposcelis decolor (Psocoptera: Liposcelididae) is an invasive psocid that is a major pest of stored grain and grain storage structures throughout Australia (Rees, Reference Rees1998, Reference Rees2004) and a minor pest of stored products internationally (Turner, Reference Turner1994). Although sexually reproducing adult L. decolor are typically only 1.5 mm in length and apterous, under favourable climatic conditions, infestations are rapid and so severe that they are measured as ‘hundreds of thousands of individuals per kilo’ (Rees, Reference Rees1994). Liposcelis infestations in Australian grain storage systems once represented less than 1% of invertebrate pest infestations requiring control in 1990, but increased to greater than 40% of infestations in 1996–97 (Rees, Reference Rees1998). Since then, resistance to chemical control in L. entomophila and the severity of L. decolor infestations in South Australia has prompted their reclassification as major pests of stored grain in Australia (Nayak et al., Reference Nayak, Collins and Reid1998).
Long distance dispersal is not uncommon among Liposcelis species and can occur either through air currents and drifts (i.e. aerial plankton: Broadhead, Reference Broadhead1950) or via the commercial trade of commodities internationally (Broadhead, Reference Broadhead1954a,Reference Broadheadb) and in Australia (Stanaway et al., Reference Stanaway, Zalucki, Gillespie and Rodriguez2001). It has recently been revealed that the cosmopolitan invasive psocid L. bostrychophila, despite also being apterous and small in size, is capable of unrestricted geographic dispersal in excess of 15,000 km through human-facilitated transport (Mikac & Clarke, Reference Mikac and Clarke2006). The Australian grain storage system can be viewed as a series of networks of suitable habitat patches. The unintentional and unrestricted movement of individuals within these networks (via trucks, trains and ships) may be a key factor in the persistence of Liposcelis species there. Artificial dispersal and the resulting gene flow can increase the size and genetic diversity of local populations (e.g. Kolbe et al., Reference Kolbe, Glor, Rodríguez Schettino, Chamizo Lara, Larson and Losos2004) and facilitate the establishment of new populations.
In other invasive insect species, the elucidation of genetic structure and variation within and among geographically distant populations has provided insights into their migration pathways and gene flow (e.g. medfly Ceritis capitata: Bonizzoni et al., Reference Bonizzoni, Zheng, Guglielmino, Haymer, Gasperi, Gomulski and Malacrida2001). Through the use of microsatellite markers, such genetic insights can aid in the development of integrated approaches to management and control as demonstrated for the diamondback moth Plutella xylostella in Australia (Endersby et al., Reference Endersby, McKechnie, Ridland and Weeks2006), and western corn rootworm Diabrotica virgifera virgifera in the US (Kim & Sappington, Reference Kim and Sappington2005). The aim of this paper is to investigate the genetic structure of Australian L. decolor populations (and a single international population for comparison) to better understand dispersal patterns among these populations using microsatellite markers.
Materials and methods
Sample collection and processing
L. decolor infestations were sampled in South Australia at the bulk grain storage sites within the Port facilities of Ardrossan, Port Giles, Thevenard, Wallaroo and at the inland regional storage site Redhill (fig. 1). At each site, 50 crevice traps (150 mm×150 mm) were systematically placed at 50 m intervals on each level of the facility, such that the whole facility was sampled. L. decolor samples were then pooled per location and randomly chosen individuals were subjected to genetic analyses. L. decolor were also sampled from a single population in Manhattan, Kansas, USA. A single population from Kansas, USA was included in this study as a representative sample of a possible extreme in long distance dispersal with the Australian populations. Upon completion of the trapping period, individuals were removed, preserved in 100% ethanol, identified to species using the key of Mockford (Reference Mockford and Gorham1991) and stored at −20°C pending genetic analyses. Total genomic DNA was isolated from whole individual L. decolor from each population following methods described by Mikac (Reference Mikac2006). Six polymorphic (GATA)n microsatellite loci that had been identified from L. decolor DNA (Mikac, Reference Mikac2006) were used. PCR reactions were carried out in 12.5 μl volumes using 25 ng of DNA, 0.4 μM each primer, 0.2 mM each dNTP, 1× PCR buffer, 4 mM Mg2+, 1 M Betaine, 0.50× bovine serum albumin (BSA) and 1 unit of Taq (New England Biolabs). PCR cycling was performed on a Corbett Research Palm Cycler, using an initial denaturation step of 94°C for 3 min (mins) followed by 40 cycles of 94°C for 30 s, 45°C 30 s and 72°C 30 s, with a final extension temperature of 72°C for 5 min.
Fig. 1. Locations where Liposcelis decolor infestations were sampled within the South Australian export bulk grain storage sites at Thevenard, Wallaroo, Ardrossan and Port Giles and the regional storage site at Redhill.
The amplification products were run on a Corbett Research laser GelScan 2000 and visualized with CyberSafe using the following quality control practises. PCR assays were carried out using filter tips to eliminate the possibility of aerosol contamination from carryover products. Fragment size was inferred by comparison with a 50 base pair size ladder run in multiple lanes on each gel. A separate negative control was included in each set of PCR reactions as an indication of potential contamination. Multiple individuals from each population were genotyped twice to ensure correct scoring across gels.
Population structure and gene flow analysis in L. decolor
Allele sizing was performed using the ONEDScan gel documentation software (Scanalytics). Number of alleles and expected (HE) and observed (HO) heterozygosity and Weir & Cockerham's (Reference Weir and Cockerham1984) F IS (inbreeding co-efficient) per locus were estimated using FSTAT 2.9.3 (Goudet, Reference Goudet2001). MICRO-CHECKER (van Oosterhout et al., Reference van Oosterhout, Hutchinson, Willis and Shipley2004) was used to test for the presence of null alleles per locus and population. Tests of fit to Hardy-Weinberg equilibrium (HWE) and genotypic linkage disequilibrium were performed using Markov chain methods (10,000 dememorizations, 500 batches, 10,000 iterations) in GENEPOP 3.4 (Raymond & Rousset, Reference Raymond and Rousset1995a). The same method was used to test for significant differentiation (Fisher exact tests) among populations for all loci and population pairs. This test is considered relatively accurate for small sample sizes and low frequency alleles (Raymond & Rousset, Reference Raymond and Rousset1995b). Weir & Cockerham's (Reference Weir and Cockerham1984) F ST, estimated as θ, was calculated among population pairs using FSTAT version 2.9.3 (Goudet, Reference Goudet2001). Due to the presence of null alleles, further estimates of F ST were recalculated in FReeNA using the ENA (excluding null alleles) method described by Chapuis & Estoup (Reference Chapuis and Estoup2007). The ENA method corrects for a positive bias in F ST estimates when null alleles are present, thus providing a more accurate estimate of F ST in the presence of null alleles (Chapuis & Estoup, Reference Chapuis and Estoup2007). Testing for isolation by distance was undertaken in FSTAT version 2.9.3 (Goudet, Reference Goudet2001) using a Mantel test to determine if there was a significant positive correlation between matrices of genetic differentiation as estimated by F ST(1/FST−1) and the natural logarithm (ln) of geographic distance in kilometres among populations. Geographic distances between populations were based upon the distance between populations via transport corridors along roads, railroad tracks or across the sea. Evidence of recent bottlenecks was assessed for populations using BOTTLENECK 1.2 (Cornuet & Luikart, Reference Cornuet and Luikart1996). This analysis incorporated a stepwise mutation model (SMM: Kimura & Ohta, Reference Kimura and Otha1978) and a two-phase model (TPM) (Di Rienzo et al., Reference Di Rienzo, Peterson, Garza, Valdös, Slatkin and Freimer1994) in which 90% of the mutations follow the SMM, and 10% represent multistep changes (Estoup & Cornuet, Reference Estoup, Cornuet, Goldstein and Schlötterer1999). Wilcoxon sign-rank tests (Luikart et al., Reference Luikart, Allendorf, Cornuet and Sherwin1999) were used to determine whether deviations of observed heterozygosity relative to that expected at drift-mutation equilibrium were significant (P<0.05). A mode shift in allele frequency distribution was used as an indicator of a population bottleneck (Luikart et al., Reference Luikart, Allendorf, Cornuet and Sherwin1999).
Results
Genetic diversity
After correction for multiple tests (n=75), significant linkage disequilibrium was only detected for the paired loci Lip138/Lip169 and Lip130/Lip197 (P<0.05). However, this only involved one population for the paired loci Lip138/Lip169 and two populations for Lip130/Lip197. Therefore, these loci were retained in all further analyses. The loci Lip73, Lip130, Lip138 and Lip197 displayed significant heterozygote deficiencies in 3–4 populations. After correction for multiple tests (n=75), each of these loci were found to be out of HWE. For these same loci, MICRO-CHECKER indicated the possible presence of null alleles in one to five of the five populations for which there were adequate data. Loci Lip117 and Lip169 did not differ from expectations of HWE and no null alleles were indicated.
Among the sampled populations, the number of alleles and allelic richness was low (table 1). The number of observed alleles ranged from one to six per locus with an average of 3.2 alleles across the six loci (table 1). Among populations, the number alleles ranged from one (Ardrossan for Lip117 for which only three of 56 individuals repeatedly PCR amplified) to six alleles (Thevenard and Kansas, USA for Lip73) per locus to produce average values of 2.5–4.2 alleles per locus per population (table 1). Mean F IS values among populations ranged from 0.21 (Ardrossan) to 0.53 (Thevenard) (table 1). However, within populations, the F IS values per locus often varied considerably (e.g. 0–1; Port Giles and Wallaroo). Heterozygosity estimates (HO and HE) were low to moderate across most populations (table 1). Mean HO per population ranged from 0.26 (Thevenard) to 0.36 (Redhill), while HE ranged from 0.36 (Ardrossan) to 0.46 (Redhill) (table 1). In most cases (except Ardrossan for Lip73 and Port Giles for Lip130), if F IS values were >0.50 then HO values were low and significantly out of HWE (table 1).
Table 1. Number of alleles per locus, allelic richness and observed and expected heterozygosity at six microsatellite loci from six populations of Liposcelis decolor investigated: n, number of individuals scored per locus; # alleles, total number of alleles; r, allelic richness; F IS, Weir & Cockerham's (Reference Weir and Cockerham1984) inbreeding coefficient; HO, observed heterozygosity; HE, expected heterozygosity; *, significant deviation from Hardy-Weinberg equilibrium.
Population differentiation
Across all populations, θ estimates per locus ranged from 0.03–0.17, and across all loci θ was significantly different from zero (θ=0.12, 95% CI 0.08–0.15), suggesting overall population differentiation (table 1). For each pairwise combination of populations, θ ranged from 0.05–0.23; and all values were significantly different from zero after correction for multiple tests (n=15, P<0.05) (table 2). Likewise, ENA-corrected F ST estimates over all loci and populations were significantly different from zero (F ST=0.11, 95% CI 0.14–0.07). However, ENA-corrected F ST estimates in pairwise population tests were consistently lower than the estimated uncorrected θ, and none were significantly greater than zero (table 2).
Table 2. Pairwise estimates of Weir & Cockerham's (Reference Weir and Cockerham1984) θ (F ST), ENA-corrected F ST across all loci (Chapuis & Estoup, Reference Chapuis and Estoup2007) and approximate geographic distances (km). Values in bold were significant after corrections for multiple tests (n=15, P<0.05).
A Mantel test of isolation by distance (IBD) revealed no relationship between Slatkin's (Reference Slatkin1995) linearized F ST(F ST(1/F ST−1)) and the ln of geographic distance in kilometres for all populations (r2=0.12, P=0.18) and for Australian populations only (r2=0.19, P=0.44), suggesting that IBD does not describe the pattern of gene flow among populations. Similarly, ENA-corrected F ST data did not indicate an IBD effect for all populations (r2=0.06, P=0.40) and for Australian populations only (r2=0.005, P=0.83). Additional tests conducted for all populations, but restricted to loci without null alleles (i.e. Lip117 and Lip169) also revealed no IBD relationship (r2=0.05, P=0.38).
Tests for population bottlenecks using Wilcoxon tests were significant (P<0.05) for the Ardrossan, Port Giles, Wallaroo and Redhill populations. These populations all displayed mode shifts of allele distribution, a qualitative indication of recent bottleneck events.
Discussion
This study reveals a low to moderate level of genetic structure among distant L. decolor populations, thus supporting a hypothesis of long distance dispersal. This likely results from genetic exchange and successful colonisation among established populations within the Australian grain storage network. There was evidence of moderate levels of gene flow across different geographic scales, resulting in a lack of isolation by distance among populations within Australia and in comparison with a distant population (i.e. Kansas, USA). Human-mediated dispersal is the most effective means of dispersal for many insects such as Liposcelis species that are preadapted to an invasive lifestyle, due to their minute size, ubiquity, polyphagy and tolerance to a wide range of temperature and humidity levels (Broadhead, Reference Broadhead1954a,Reference Broadheadb; Turner, Reference Turner1994; Rees, Reference Rees1998; Mikac & Clarke, Reference Mikac and Clarke2006).
A surprising result was the relatively low level of genetic variation in L. decolor (table 1) given that the species reproduces sexually and infestations are known to occur as ‘hundreds of thousands of individuals per kilo’ (Rees, Reference Rees1994), thus suggesting very large effective population sizes. This is in contrast to moderate to high levels of genetic variation found in L. bostrychophila populations from the UK using allozymes (Ali & Turner, Reference Ali and Turner2001) and Australian and international populations of L. bostrychophila using randomly amplified polymorphic DNA (RAPDs) (Mikac & Clarke, Reference Mikac and Clarke2006). Nevertheless, using microsatellite markers, low levels of genetic variation have been found for other invasive insects with large effective population sizes, including the tephritid fruitflies Ceritis capitata (Casey & Burnell, Reference Casey and Burnell2001) and Bactrocera papayae (Shearman et al., Reference Shearman, Gilchrist, Crisafulli, Graham, Lange and Frommer2006). In L. decolor populations, low genetic diversity resulting from repeated chemical control in line with set management practices in individual grain storage facilities is expected as this process removes a significant portion of the individuals in a population (e.g. Pratt & Reuss, Reference Pratt and Reuss2004); our results suggest that these measures may be strong enough to reduce genetic diversity and induce a bottleneck on a population once control has ceased. It is not surprising, then, that significant bottlenecks were detected in the majority of populations investigated (i.e. Ardrossan, Port Giles, Wallaroo and Redhill). These bottlenecks may have contributed to the relatively low heterozygosity and to the moderate to high F IS values found for most populations (table 1), similar to findings of Colautti et al. (Reference Colautti, Manca, Viljanen, Ketelaars, Burgi, Macisaac and Heath2005) for the Eurasian spiny waterflea Bythotrephes longimanus. In the Argentine ant Linepithema humile, bottlenecks were responsible for an increase in their invasion success in their introduced range through the loss of alleles associated with intraspecific aggression (Tsutsui et al., Reference Tsutsui, Suarez, Holway and Case2000). It is possible that Liposcelis invasion success could in part be attributed to the cumulative loss of alleles that may have once hindered their invasion success, leaving individuals that are best able to survive and adapt to life in grain storage environments, as discussed by Ali & Turner (Reference Ali and Turner2001) for L. bostrychophila.
Alternatively, apparently low levels of genetic variation and significant deviation from HWE in this study could have been caused by the presence of null alleles or a Wahlund effect. Null alleles typically occur through PCR amplification failure of alleles at specific loci, which is thought to be caused by sub-optimal PCR conditions, degraded samples, insufficient DNA quantities or mutations in primer binding sites (Selkoe & Toonen, Reference Selkoe and Toonen2006), and some of these factors could have influenced our results. However, it is unlikely that significant deviation from HWE could have been exclusively caused by null alleles because the microsatellite loci cross-amplified in five other Liposcelis species (Mikac, Reference Mikac2006). Although relatively few problems were associated with RAPD PCR amplification of L. bostrycophila (Mikac & Clarke, Reference Mikac and Clarke2006), L. decolor proved to be extremely difficult to amplify at microsatellite loci (Mikac, Reference Mikac2006), thus suboptimal PCR conditions or DNA yield may have resulted in a proportion of the null alleles. Additionally, the microsatellite markers were developed (Mikac, Reference Mikac2006) from a single Australian population (i.e. Roseworthy, South Australia) that was not investigated in this study and that does not experience similar infestation levels or population fluctuations as the six populations investigated in this study. It has been noted that large effective population sizes typically result in higher frequencies of null alleles due to flanking sequence mutations (Chapuis & Estoup, Reference Chapuis and Estoup2007). Therefore, the large effective population sizes of the six populations investigated here may have contributed to the high frequency of null alleles detected, in comparison to initial investigations by Mikac (Reference Mikac2006) that did not observe null alleles using the same loci. Given that four of six L. decolor microsatellite loci were found to be out of HWE and were also shown by MICRO-CHECKER to possibly contain null alleles, we estimated F ST using loci with or without probable null alleles and used corrected ENA-F ST estimates (Chapius & Estoup, Reference Chapuis and Estoup2007). Overall estimates of F ST did not vary substantially as a result of probable null alleles, but may have lead to an overestimate of F ST values among population estimates of F ST (table 2). Thus, the probable presence of null alleles in the data set would have reduced the estimates of genetic diversity and possibly reduced estimates of gene flow (table 2). However, in the two loci (Lip117 and Lip169) for which there was no evidence of null alleles, all measures of genetic diversity were similarly low. In future studies of Liposcelis species using the microsatellites developed by Mikac (Reference Mikac2006), sequencing of the microsatellite flanking regions (to investigate possible flanking sequence mutations) should be conducted to further investigate the cause of suspected null alleles.
Tests for IBD were not significant, thus reinforcing the need to analyse the specifics of human-mediated transport of L. decolor over short and long distances within the Australian grain storage network. The unintentional movement of L. bostrychophila within the Australian grain storage network (via trucks, trains and ships) is a key factor in its persistence (Mikac & Clarke, Reference Mikac and Clarke2006) and potentially the persistence of L. decolor as well. Some of the estimates of genetic structure among the port storage sites investigated are consistent with the manner in which grain is shipped from South Australia for export. Among South Australian populations, the major port facility for central and western Eyre Peninsula at Thevenard, had one of the lowest levels of differentiation with Port Giles, the major port facility for southern Yorke Peninsula, despite their 800 km geographic separation (fig. 1, table 2). Indeed, these populations were less divergent than Ardrossan and Port Giles, which are 73 km apart, or Wallaroo and Port Giles, 140 km apart (fig. 1, table 2). However, differences in genetic structure between these later two examples were not unexpected since there is very limited transportation of grain between these sites. Port Adelaide (not sampled in this study) is the major grain port for the whole of eastern South Australia, and grain from most of Yorke Peninsula is directly transported to Port Adelaide. Genetic divergence among Australian populations and the single USA population from Kansas was moderate with one pairwise comparison showing little genetic differentiation (Redhill versus Kansas, USA; table 2). Grain commodity trading between the USA and Australia is not uncommon, and gene flow between these countries is possible and has probably occurred historically for L. decolor, as demonstrated by RAPDs for L. bostrychophila (i.e. gene flow among South Australian and Kansas, USA populations was found by Mikac & Clarke (Reference Mikac and Clarke2006)). Indeed, it has been speculated that the current distribution of Liposcelis pest species (i.e. L. bostrychophila, L. decolor and L. entomophila) in Australia was assisted by humans during the movement of grain throughout Australia (Rees, Reference Rees1994), particularly as the grain storage network is well connected by train, roads and seaports. Broadhead (Reference Broadhead1950) suggested that long distance dispersal was not uncommon among Liposcelis species, as they had been discovered as aerial plankton 300 m above ground. It is more likely, though, that the commercial trade of commodities is a more effective and reliable dispersal mechanism given that Liposcelis species were documented five decades previously from the UK in shipped commodities and shipping warehouses (Broadhead, Reference Broadhead1954a, Reference Broadheadb). Recently, Liposcelis species have been found in import shipping containers at Port Brisbane, Australia, proving further evidence of their assisted dispersal via human trade routes (Stanaway et al., Reference Stanaway, Zalucki, Gillespie and Rodriguez2001). Due to a relatively long life span (4–10 months for L. bostrychophila: Turner, Reference Turner1994), Liposcelis species can persist within the grain storage system for extended periods. Individuals can lie dormant until a suitable habitat and favourable abiotic conditions trigger heavy infestations, as frequently occurs in South Australia, particularly Thevenard (Rees, Reference Rees1994). Modern worldwide travel and transport is such that any continent's grain storage system can be reached within a short period of time, thus allowing Liposcelis species to disperse and establish populations with ease. This is supported by low to moderate levels of genetic differentiation among global populations of L. decolor and L. bostrychophila (Mikac & Clarke, Reference Mikac and Clarke2006) associated with global grain storage systems.
Acknowledgements
The authors would like to thank Dr Marion Hoehn, from the Helholtz Centre for Environmental Research-UFZ, Leipzig, Germany, for useful advice; and Dr Bruce Halliday, from the CSIRO Entomology, Canberra, Australia, and three anonymous reviewers for corrections and comments that significantly improved the manuscript. This research was part of Katarina Mikac's PhD thesis and was funded by the Australian Co-Operative Bulk Handlers, with support from the Institute for Applied Ecology and School of Resource, Environmental and Heritage and Sciences, University of Canberra.
