Introduction
True fruit flies (Diptera: Tephritidae) are serious pests of horticulture worldwide, due to their broad larval host range, cosmopolitan distribution and high invasive capacity (McPheron & Steck, Reference McPheron and Steck1996; Malacrida et al., Reference Malacrida, Gomulski, Bonizzoni, Bertin, Gasperi and Gugliclmino2007). The major fruit fly pest in Australia is the Queensland fruit fly, Bactrocera tryoni (Froggatt). Native to Queensland, this species spread south in the 1860s (Drew, Reference Drew1989) and is now found along the entire east coast of Australia (Meats, Reference Meats1981) where it infests almost all horticultural crops (Drew, Reference Drew, Drew, Hooper and Bateman1978). Genetic structuring in these east coast populations have been studied in Yu et al. (Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001), Meats et al. (Reference Meats, Clift and Robson2003) and Sved et al. (Reference Sved, Yu, Dominiak and Gilchrist2003).
In addition to the east coast populations, Australia has a second endemic fruit fly region, the northwestern region comprising the Northern Territory and Western Australia (fig. 1). The east and northwest regions are separated by an arid zone straddling the border between the Northern Territory and Queensland (Drew, Reference Drew1989).
Fig. 1. Map of Australia showing the sampling sites in Queensland, Western Australia and the Northern Territory. Numbers correspond to locations given in table 1. Circles indicate the regions used in the analysis. The arid region separating Queensland and the Northern Territory is shaded.
Until recently, the northwest region was considered to be relatively free from serious fruit fly pests. The main species described in the region, B. aquilonis (May), a sister species to B. tryoni, was not considered to be a pest (Drew, Reference Drew1989), having been reared from just four commercial crops: peach, guava and two species of citrus (Allwood & Angeles, Reference Allwood and Angeles1979). However, in 1985, a large increase in the range of preferred hosts was reported for B. aquilonis in the Darwin area (Smith et al., Reference Smith, Chin, Allwood and Collins1988). Collections of fruit before and after this date showed that 40 cultivated fruit species, including mangoes and star fruit, were now being infested. Pest flies were soon found in other major towns in the Northern Territory and, in 2000, new outbreaks of pest flies were reported in the horticultural regions of Kununurra (834 km southwest of Darwin: S. Smith, personal communication).
The cause of the increase in infestation of commercial crops since 1985 is unknown and is complicated by the fact that B. tryoni and B. aquilonis are morphologically very similar. The first specimens of B. aquilonis, collected in Darwin in 1961, were identified as B. tryoni (May, Reference May1963), before being described as a separate species based on their paler colouration and small differences in markings on the frons, wings, thorax and abdomen (May, Reference May1965). The characters used to differentiate the species have since been questioned as being inconsistent and within the range of natural variation seen in B. tryoni. Nevertheless, the species status of B. aquilonis was maintained after a study by Drew & Lambert (Reference Drew and Lambert1986). They found minor differences in the ovipositor, male genitalia and egg chorion. However, these differences are not easily observed, and classification is generally based on trapping location. B. tryoni and B. aquilonis produce viable and fertile hybrids in the laboratory (Drew & Lambert, Reference Drew and Lambert1986).
The morphological similarity between the two species has led to speculation that the recent outbreaks of pest flies in northwestern Australia may be the result of an invasion of B. tryoni or that a distinct new strain of B. aquilonis has adapted to utilize commercial crops as hosts (Wang et al., Reference Wang, Yu, Raphael and Gilchrist2003; Raphael et al., Reference Raphael, Whyard, Shearman, An and Frommer2004). A third possibility is that a hybridisation event between invading B. tryoni and B. aquilonis has occurred (Osborne et al., Reference Osborne, Meats, Frommer, Sved, Drew and Robson1997; Morrow et al., Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000; Yu et al., Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001).
Previous studies have been inconclusive in identifying the cause of outbreaks in the northwest. An Australia-wide trapping survey conducted in early 1994 failed to find any flies unambiguously identified by morphology as B. aquilonis, suggesting that B. tryoni may have invaded the northwest (Osborne et al., Reference Osborne, Meats, Frommer, Sved, Drew and Robson1997). However, studies using microsatellites have shown significant differences between east coast and northwest flies. Yu et al. (Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001) studied flies collected over five years in Darwin and Queensland and showed significant differentiation at the six microsatellite loci analysed. Wang et al. (Reference Wang, Yu, Raphael and Gilchrist2003), using 30 microsatellite markers, found that east coast B. tryoni and samples from Darwin and Kununurra were all equally genetically differentiated from each other (F ST=0.01–0.02). That degree of differentiation was smaller than that measured between east coast samples of B. tryoni and another sibling species, B. neohumeralis (F ST~0.08). In contrast, mitochondrial DNA sequencing revealed sequence differences between east coast flies and a sample from Kununurra in the northwest, but not between B. tryoni and B. neohumeralis (Morrow et al., Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). The aim of the present study was to determine whether a comprehensive DNA microsatellite survey could differentiate populations of Bactrocera in the northwest region.
Materials and methods
Sampling
Four groups of flies were used in the analysis (fig. 1, table 1):
1. For use as reference samples, B. tryoni was collected in Queensland where it is endemic. Flies were collected between 1999 and 2003 from both the east coast and towns in the arid west near the border with the Northern Territory.
2. NT: These flies were collected across the Northern Territory by one of the authors (ECC) in March 2002 and October 2003, except for the Darwin 1999 and Gove (site 17) samples that were collected by the NT Department of Primary Industries, Fisheries and Mines. Traps were set up to catch flies in urban and natural areas along transects from Kakadu to Darwin (152 km), Darwin to Katherine (314 km) and Katherine to Kununurra (514 km). Traps were set in bush areas near creeks and rivers as flies are thought to follow water courses (Meats, Reference Meats1981). Where possible, traps were set near Kakadu plum trees (Terminalia ferdinandiana), the favoured host of B. aquilonis (Smith et al., Reference Smith, Chin, Allwood and Collins1988).
3. KUN: The third group consisted of outbreaks of flies trapped in Kununurra, a developing horticultural region. Since 2000, large numbers of B. aquilonis have been trapped on an extensive trapping grid maintained by the Department of Agriculture and Food, Western Australia.
4. WA: Flies trapped by the WA department of Agriculture and Food in 2001 from Broome and Derby, isolated towns on the coast of Western Australia. B. tryoni had previously been eradicated from WA (in 1990) and these two populations are presumed to have been recently founded.
Table 1. Descriptive statistics for each sample of flies used in the study. Sample size averaged over loci (n), number of alleles averaged over loci (A), proportion of low frequency alleles over all loci (low freq alleles), number of alleles unique to that sample (unique alleles), reduction in allelic richness per locus (allelic richness) with significance given (*, P<0.05; **, P<0.01;***, P<0.001; ref, reference sample), expected heterozygozity (He), observed heterozygozity (Ho) and the inbreeding coefficient (f). For each sample the data is combined over 14 loci except samples 1 and 32 (13 loci) and samples 39 and 40 (11 loci).
Standard cue-lure traps were used to collect the samples (White & Elson-Harris, Reference White and Elson-Harris1992; Osborne et al., Reference Osborne, Meats, Frommer, Sved, Drew and Robson1997). Since females are not attracted to cue lure, all samples consisted of males only. The Queensland samples (QLD, samples 1–16) were each from single traps, except for samples 1–4 which were combined from numerous traps along the east coast, between Brisbane and Cairns. Yu et al. (Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001) and Wang et al. (Reference Wang, Yu, Raphael and Gilchrist2003) have previously shown that there is no detectable population structuring along the entire Queensland coast.
The 15 Northern Territory samples consisted of flies caught in up to four traps within 70 km of each other in a ten-day period (table 1) with no more than 20 flies from any single trap being analysed. The one exception was the Timber Creek sample, which was a combination of 18 traps set up between Katherine and Kununurra (390 km). The samples from Kununurra (samples 32–38) were combined over four-month periods from traps placed throughout the town. The validity of grouping traps into samples was checked using Hardy-Weinberg tests.
Microsatellites
Total DNA was extracted from each fly using the Chelex method (Walsh et al., Reference Walsh, Metzger and Higuchi1991), and was typed using standard fluorescent PCR methods as described in Yu et al. (Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001) at 14 polymorphic loci: Bt10, Bt11, Bt14, Bt15, Bt17, Bt32 (Kinnear et al., Reference Kinnear, Bariana, Sved and Frommer1998, renamed in Yu et al., Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001), Bt1.1a, Bt2.6a, Bt2.6b, Bt6.12a, Bt4.1a, Bt8.12a, Bt8.6a (Wang et al., Reference Wang, Yu, Raphael and Gilchrist2003), Bp78 (Shearman et al., Reference Shearman, Crisafulli, Gilchrist, Graham, Lange and Frommer2006). Four of the samples had been used in a previous study and had been scored at fewer loci than were used in the current study; sample 1, 13 loci (no Bp78); sample 32, 13 loci (no Bt8.12a); and samples 39 and 40, 11 loci (no Bt15, Bt17 or Bt2.6b).
Each locus and sample was tested for departure from Hardy-Weinberg equilibrium and linkage disequilibrium in Genpop 3.3 (Raymond & Rousset, Reference Raymond and Rousset1995), correcting for multiple comparisons using the sequential Bonferroni test.
Summary statistics for each sample were calculated using the GDA software (Lewis & Zaykin, Reference Lewis and Zaykin2002). The number of alleles, proportion of low frequency alleles (those with an allele frequency less than 0.1) and number of unique alleles were counted. Allelic richness was estimated taking the Qld 2000 sample (sample 1) as the reference population. Expected allele numbers for the given sample sizes were calculated by rarefaction (Hurlbert, Reference Hurlbert1971; Leberg, Reference Leberg2002). The significance of any changes from the expected values was tested using a resampling technique; 5000 resamples of the test sample size were drawn from the reference population, and the number with fewer alleles than the test sample was counted. Results were combined over loci (Sokal & Rohlf, Reference Sokal and Rohlf1995) and are presented as the average reduction in allele number per locus. These calculations were performed in Excel Visual Basic.
Differentiation between samples
F ST values were calculated between each sample pair, and their significance was tested using the exact test procedure in FSTAT 2.9.3 (Goudet, Reference Goudet2002) with 5000 permutations. Population differentiation was also tested using chi-squared tests summed over loci (Yu et al., Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001). A hierarchical analysis of molecular variance (AMOVA) was performed in ARLEQUIN 2.0 (Schneider et al., Reference Schneider, Roessli and Excoffier2000) to determine how much of the variation present in the complete data-set was due to between-region variation compared to within-region and within-sample variation. To calculate the significance of the variance components, 5000 resamples were used.
Spatial structuring was investigated by comparing genetic distance with geographic distance using the isolation by distance procedure implemented in GENEPOP ON THE WEB (Raymond & Rousset, Reference Raymond and Rousset1995; Rousset, Reference Rousset2000) with 5000 permutations. In order to remove sample pairs with a geographic distance of 0, only one sample from each location was used. Samples from 2002 or, if 2002 was not available, 2001 were included. To visualize the genetic structuring, correspondence analysis using all samples was conducted in JMP (SAS Institute, 1994).
To further investigate population structure, the Bayesian clustering program STRUCTURE 2.1 (Pritchard et al., Reference Pritchard, Stephens and Donnelly2000; Falush et al., Reference Falush, Stephens and Pritchard2003) was used. STRUCTURE assigns individuals to clusters which are in Hardy-Weinberg and linkage equilibrium based on a user-defined number of clusters (K). Highly repeatable results were obtained with 50,000 iterations after a burn-in of 50,000 iterations and without prior population information.
Estimates of the rate of migration between the QLD and NT were calculated using the program BAYESASS 2.1 (Wilson & Rannala, Reference Wilson and Rannala2003), which uses a Bayesian multilocus method to estimate the recent rate of migration expressed as a proportion of the recipient population.
Mitochondrial sequence analysis
Mitochondrial sequences from the cytochrome b (cytb) (Kocher et al., Reference Kocher, Thomas, Meyer, Edwards, Pääbo, Villablanca and Wilson1989) and cytochrome oxidase II (COII) (Simon et al., Reference Simon, Frati, Beckenbach, Crespi, Liu and Flook1994) genes were amplified in individual flies from the northwest and Queensland as described in Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). PCR products were purified using the Wizard® SV gel and PCR clean-up system (Promega, USA) before being sequenced. Sequences were aligned using SEQUENCHER™ 4.1 (Gene Codes) and the number of mitotypes (mitochondrial haplotypes) scored.
Five flies from Darwin 2002, Katherine 2002 and Kununurra 2000 and one fly from Kakadu 2002 and Kununurra 2001 were compared to a sample of five flies from west Queensland. Comparisons were made between these sequences and those from the previous study by Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000).
Results
Summary statistics
A total of 1159 flies from 40 samples were scored at 11 to 14 microsatellite loci (table 1). The number of alleles at each locus ranged from 5 to 41, with heterozygozities between 0.30 and 0.88. After correcting for multiple comparisons (table size=692, including all samples and all loci scored), two loci showed deviations from Hardy-Weinberg equilibrium in a small number of samples. Bt15 was not in equilibrium in seven samples (samples 1, 4, 12, 14, 32, 36 and 38) and Bt32 in five samples (samples 4, 18, 21, 32 and 36). In a previous study by Yu et al. (Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001), Bt15 was shown to have a low frequency of null alleles in Queensland populations. Consequently, Bt15 and Bt32 were excluded from those analyses which assume equilibrium. One other locus, 2.6b, showed an excess of homozygotes in sample 24. Two pairs of loci showed significant linkage disequilibrium in 1–2 samples; loci 2.6b and 2.6a in samples 32 and 35 and 2.6b and 8.6a in sample 32. These samples are both outbreak samples from Kununurra. Excluding these loci did not change the results presented; thus, they were included in the analyses.
Descriptive statistics were uniform across the range of samples (table 1). The number of alleles and unique alleles were, as expected, greater in the large samples from Queensland. Allelic richness decreased most in samples from Kununurra and WA, with the most westerly samples being the most depleted. Several of the northwest samples had alleles not found in Queensland.
Differentiation between samples
Within the Northern Territory samples, genetic differentiation was minimal with only one of 105 pairwise F ST estimates significant (P<0.001). Those two samples were separated by four years and 350 km (samples 20 and 31; table 2). Similar spatial and temporal homogeneity was shown among the Queensland samples. Differentiation between the Northern Territory and Queensland regions was greater, with the overall F ST=0.015, P<0.05; however, most pairwise comparisons were non-significant.
Table 2. Results of genetic differentiation tests between all pairs of fly samples. The significance of chi-squared tests of heterogeneity is given below the diagonal by asterisks (NS, not significant; *, P<0.05; **, P<0.01; ***, P<0.001). Values above the diagonal are FST estimates. Significant values are in bold and underlined. All tests were corrected for multiple comparisons. Sample numbers and names correspond to those in table 1. Boxed results include samples from the same region.
In contrast, the Kununurra samples showed significant heterogeneity across years (table 2) with corresponding variation in genetic distance to the Queensland and Northern Territory samples. Despite being only 165 km apart, the two WA samples were highly differentiated: F ST=0.10, P<0.001. Sample 39 (Derby) was significantly differentiated from all other samples, while sample 40 (Broome) was similar to the Northern Territory samples.
A Mantel test conducted by comparing geographic distance with F ST/(1–F ST) was highly significant (P<0.001; fig. 2). Sample 39 (Derby) was not included in this analysis since it showed strong founder effects. When the Queensland samples were excluded, this regression was no longer significant (P=0.6).
Fig. 2. Geographic distance (as ln(geographic distance)) plotted against genetic distance (as F ST/(1–F ST)) calculated between samples of flies. Only one sample from each location was included to eliminate comparisons with a geographic distance=0. QLD refers to comparisons made between samples within this region, similarly for NT. All other comparisons were made between regions (▪, QLD; X, NT; ▵, NT-QLD; •, KUN-QLD; ○, KUN-NT; –, QLD-Broome; +, NT-Broome; ◊, KUN-Broome).
A correspondence analysis separated the samples into two main groupings (fig. 3). One consisted of the Queensland samples and the other of the Northern Territory, Kununurra and Broome samples. The groupings were separated by the first axis, which explained 11% of the variation in all the data. The Derby sample was separated from all other samples on both the first and second axes (8.6% of variation). AMOVA showed that 99.2% of the variation was within the samples (P<0.001) with no significant variation either between regions (i.e. Queensland versus northwest samples, excluding Derby) or between samples within regions. The rate of migration, m, estimated to be occurring from Queensland into the Northern Territory was 0.012 (standard deviation=0.008).
Fig. 3. Correspondence analysis of all fly samples plotted on the first and second axes (% of variation explained by each axis). Samples from the different regions are indicated by different shapes (○, KUN; X, NT; ▪, QLD; ◆, WA). Of the two WA samples (Derby and Broome), the Derby sample (sample 39) was separated from all others.
Clustering analysis, using the program STRUCTURE, included all samples. Although the number of clusters (K) was varied from K=1 to 8, the posterior probabilities did not sharply plateau (fig. 4a). At K=2, the Queensland samples were incompletely separated from the Northern Territory samples (fig. 4b). Of the Queensland individuals, 59% were assigned to cluster 1 with probability greater than 0.9. At this level, 4% of the Queensland samples were assigned to cluster 2 while 37% were not assigned to any cluster. Of the northwest samples, 38% were assigned to cluster 2 while 6% were assigned to cluster 1 and 56% remained unassigned. Reanalysis with the Queensland samples removed revealed no further groupings among the northwest samples, although at K=4 a possible cluster was found that included some of the Kununurra samples.
Fig. 4. Results of Bayesian clustering analysis between all QLD (1–16), NT (17–31), KUN (32–38) and WA (39–40) samples: (a) posterior probabilities of K, the number of clusters, and (b) the probability at K=2 with which each individual is assigned to each of the two clusters. The sample numbers refer to those in table 1. For larger samples, the sample number is listed repeatedly to indicate which individuals belong to which sample.
mtDNA sequencing
Mitochodrial DNA sequences from the cytb region of 22 individuals (GenBank accession numbers GQ121835–GQ121856) were combined with the sequence data from Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000) to give a total of 41 samples from Queensland and 26 from the northwest (including Kununurra) (fig. 5). Over the 240 bp sequenced, 28 mitotypes were observed. The most common mitotype found throughout the northwest (5/26) was also found in Queensland (4/41). Only one other mitotype, the consensus sequence, was shared between the regions. Sixteen samples had a unique mitotype. Previous analysis of flies from Kununurra had found that site 166 was fixed for G (Morrow et al., Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). In the current study, again, all flies from Kununurra have this G, as do some but not all the NT samples (7/11).
Fig. 5. Polymorphic sites in the sequenced segment of the cytb gene. Samples are from Darwin (Dwn), Kakadu (Kak), Katherine (Kth), Kununurra (Kun00, Kun01) and Queensland (QLD). Samples marked with a * are from Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). Nucleotides are numbered from the same starting point as Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). The figure shows the polymorphic sites, their position and the common and alternative base at each site.
From the COII region, 500 bp were sequenced in 16 individuals from the northwest and four from Queensland (Genbank accession numbers GQ121857–GQ121876); 176 bp overlapped with the sequences of Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). A total of 22 mitotypes were present in the combined data-set (fig. 6) with two mitotypes, including the consensus sequence, shared between the two regions. Base 246, which was fixed for G in the three Kununurra flies sequenced by Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000) was either a G or an A in the individuals sampled here and was present in one of the Queensland samples. The results here show no obvious separation between the Queensland and northwest samples.
Fig. 6. Polymorphic sites in the sequenced segment of the COII gene. Samples are from Darwin (Dwn), Kakadu (Kak), Katherine (Kth), Kununurra (Kun00, Kun01) and Queensland (QLD). Samples marked with a * are from Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). Nucleotides are numbered from the same starting point as Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000). The figure shows the polymorphic sites, their position and the common and alternative base at each site.
Discussion
Genetic differentiation
The genetic distance between the Northern Territory and Queensland samples was small (F ST=0.015), consistent with low levels of migration from Queensland into the Northern Territory. This degree of differentiation is within the range of genetic distances observed in eastern Australia between various populations of B. tryoni (F ST=0.002–0.08: Sved et al., Reference Sved, Yu, Dominiak and Gilchrist2003, Wang et al., Reference Wang, Yu, Raphael and Gilchrist2003). This finding suggests that the two Northern Territory and Queensland samples have only recently separated and/or that there has been continuing gene flow since separation.
Within the Northern Territory, a high level of homogeneity was found despite the inclusion of flies from urban, horticultural and natural areas. Samples were not only homogeneous over time but also showed no evidence of spatial structuring. The only significant difference between localities occurred between the Darwin sample from 1999 and the combined sample from traps around Timber Creek in 2003. This may have been due to the relatively low population densities around Timber Creek. Similarly, no structuring among samples was evident using the Bayesian clustering analysis or the correspondence analysis.
There was no evidence of isolation by distance in the Northern Territory. These findings are contrary to the proposition that B. tryoni is present in Darwin but B. aquilonis is present outside that urban area (Smith et al., Reference Smith, Chin, Allwood and Collins1988; Osborne et al., Reference Osborne, Meats, Frommer, Sved, Drew and Robson1997; Morrow et al., Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000; Wang et al., Reference Wang, Yu, Raphael and Gilchrist2003). These results suggest that just one strain is present in the Northern Territory. A similarly extensive and stable population has already been shown to exist in Queensland (Yu et al., Reference Yu, Frommer, Robson, Meats, Shearman and Sved2001).
The samples of flies from Kununurra varied between years and showed differing levels of differentiation from the Northern Territory samples. In clustering analyses, they grouped mainly with the Northern Territory samples, suggesting they originate from the Northern Territory. Trapping in the drier areas between the Northern Territory and Kununurra regions yielded very few or no flies. Thus, it is likely that the Kununurra population is isolated but with occasional migration from the Northern Territory region. For the WA samples, those from Derby showed evidence of founder effects in contrast to the Broome sample, which grouped closely with the Northern Territory flies. A similar fragmentation of B. tryoni populations was found in the drier areas of eastern Australia, where B. tryoni is restricted to urban areas in which artificial irrigation provides necessary moisture (Dominiak et al., Reference Dominiak, Mavi and Nicol2006). These populations are also often short lived and subject to strong founder effects (Meats et al., Reference Meats, Clift and Robson2003; Sved et al., Reference Sved, Yu, Dominiak and Gilchrist2003).
Mitochondrial DNA
Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000) carried out an analysis of two mitochondrial sequences (cytb and COII) in B. tryoni and a limited number of flies from Kununurra classified as B. aquilonis. They found evidence of fixed differences between the two species. However, using a wider sampling regime, our sequence data provided no evidence for the separation of northwest samples from the Queensland samples. The potential fixed difference between northwest and Queensland flies found by Morrow et al. (Reference Morrow, Scott, Congdon, Yeates, Frommer and Sved2000) was shown not to be fixed in the present study.
Species status of B. aquilonis
Despite the morphological differences found in previous studies (Drew & Lambert, Reference Drew and Lambert1986), the genetic data do not support the separation of the eastern and northwestern flies into two species. The level of genetic differentiation is similar to that observed between geographically close populations of other tephritid species (Mun et al., Reference Mun, Bohonak and Roderick2003; Bonizzoni et al., Reference Bonizzoni, Guglielmino, Smallridge, Gomulski and Malacrida2004; Augustinos et al., Reference Augustinos, Mamuris, Stratikopoulos, D'Amelio, Zacharopoulou and Mathiopoulos2005; Nardi et al., Reference Nardi, Carapelli, Dallai, Roderick and Frati2005). Compared with previous studies in B. tryoni, the genetic distance found here between the Northern Territory and Queensland flies (F ST=0.015) is larger than that between endemic east coast populations of B. tryoni (average F ST=0.002) but smaller than that found between populations of B. tryoni in individual towns further south (average F ST=0.03–0.08: Sved et al., Reference Sved, Yu, Dominiak and Gilchrist2003). The genetic distance between B. tryoni and B. neohumeralis (another species pair that produce viable hybrids) is F ST=0.085 (Wang et al., Reference Wang, Yu, Raphael and Gilchrist2003). Based on genetic distances alone, the Northern Territory and Queensland flies appear more likely to be allopatric populations of the same species.
It seems extremely unlikely that our sampling did not pick up flies that previously would have been classified as B. aquilonis. However, if we did not, then B. aquilonis must represent a very small proportion of the flies in this region. The sample from Darwin in 1999 used in this study had been morphologically identified by R. Drew (personal communication) as B. aquilonis. That group could not be distinguished from the other samples using molecular markers.
It is possible that B. aquilonis existed in the past and has now been replaced by B. tryoni. However, our sampling covered urban, horticultural and bush areas and no evidence of distinct populations was found in any habitat. If B. tryoni has displaced B. aquilonis, then replacement has occurred in all habitats in less than 30 years. This seems unlikely since trappings occurred in National Parks, large areas of intact natural habitat.
An alternative possibility is that B. aquilonis is not a separate species. Under this scenario, the flies first trapped in the Northern Territory in 1961 (May, Reference May1963) would have been B. tryoni, and all morphological differences are within the natural range of variation of B. tryoni. One of the main differences used to differentiate B. tryoni and B. aquilonis, apart from distribution, is pest status, with B. aquilonis considered benign. However, we suggest that the change in pest status may be due to a large increase in population size following the expansion of the horticultural industry in the Northern Territory in the 1980s. Annual fruit production increased from 545 tonnes in 1981 to 1946 tonnes in 1983. This included an increase in mango production, a favoured host of B. tryoni, from an average of seven tonnes between 1977 and 1981 to 150 tonnes between 1983 and 1986, with most new plantings occurring in the Darwin region (Northern Territory Department of Primary Production, 1972–1991). This increase corresponds with the change in behaviour observed by Smith et al. (Reference Smith, Chin, Allwood and Collins1988) of an increase in pest fruit fly activity in the Darwin region from 1985. The rise of the ‘new’ pest species in northwestern Australia may only reflect a large increase in population size resulting from increased host availability.
Wang et al. (Reference Wang, Yu, Raphael and Gilchrist2003) also discussed the question of whether B. tryoni and B. aquilonis were separate species. Their less comprehensive survey also found the two species to be minimally genetically differentiated. For the purposes of species definitions, they concluded that since the two groups are allopatric, it makes little difference whether they are considered one or two species. From a quarantine perspective, the genetic similarity between B. tryoni and the northwestern species would suggest that control and disinfestation protocols should be similar in both regions. Quarantine practices appropriate to B. tryoni outbreaks have been in place in the northwest region for a number of years (Yonow & Sutherst, Reference Yonow and Sutherst1998).
Acknowledgements
We thank Stuart Smith and John Moulden for their expertise and assistance and Dan Crisafulli and Libby Cameron for help with the field collections. The NT Department of Primary Industry, Fisheries and Mines and the WA department of Agriculture and Food provided some fly samples. This work was supported by grants from the Australian Research Council and the Kununurra growers committee.
