The rph2 Gene Is Responsible for High Level Resistance to Phosphine in Independent Field Strains of Rhyzopertha dominica

The lesser grain borer Rhyzopertha dominica (F.) is one of the most destructive insect pests of stored grain. This pest has been controlled successfully by fumigation with phosphine for the last several decades, though strong resistance to phosphine in many countries has raised concern about the long term usefulness of this control method. Previous genetic analysis of strongly resistant (SR) R. dominica from three widely geographically dispersed regions of Australia, Queensland (SRQLD), New South Wales (SRNSW) and South Australia (SRSA), revealed a resistance allele in the rph1 gene in all three strains. The present study confirms that the rph1 gene contributes to resistance in a fourth strongly resistant strain, SR2QLD, also from Queensland. The previously described rph2 gene, which interacts synergistically with rph1 gene, confers strong resistance on SRQLD and SRNSW. We now provide strong circumstantial evidence that weak alleles of rph2, together with rph1, contribute to the strong resistance phenotypes of SRSA and SR2QLD. To test the notion that rph1 and rph2 are solely responsible for the strong resistance phenotype of all resistant R. dominica, we created a strain derived by hybridising the four strongly resistant lines. Following repeated selection for survival at extreme rates of phosphine exposure, we found only slightly enhanced resistance. This suggests that a single sequence of genetic changes was responsible for the development of resistance in these insects.


Introduction
Phosphine (PH 3 ) is the most economically viable fumigant for the control of insect pests of stored grain, making it the major method of control worldwide [1,2,3,4,5]. Low level resistance to phosphine was first reported in an FAO global survey undertaken in the 1970s [6]. Widespread high level resistance to phosphine in recent years now threatens the continued use of this chemical [7]. The lesser grain borer Rhyzopertha dominica (F.) is one of the most destructive pests of stored grains and high levels of phosphine resistance have been reported from several countries, such as Bangladesh [8], Brazil [9,10], India [11,12], China [13,14], and the Philippines [15]. In Australia, strongly resistant R. dominica was first detected in southern Queensland in 1997 [16], followed by detection of another strongly resistant strain 300 km to the north in 1998 (Collins, unpublished). Emergence of strongly phosphine resistant strains of R. dominica was recently reported in New South Wales [17] and in South Australia (Wallbank, personal communication).
Genetic analysis of the initial strongly resistant strain from southern Queensland SR QLD , (elsewhere referred to as QRD569), identified two major genes responsible for resistance [3,18].
Molecular genetic analysis demonstrated that the strong resistance of SR QLD was provided by synergistic interaction between the two genes, rph1 and rph2 [19]. Schlipalius et al. [19] also showed that rph1 is the major gene responsible for the weak resistance phenotype of the strain WR QLD , previously referred to as QRD369 [3].
Comparative genetic analysis [20] revealed that two major genes are responsible for the strong resistance phenotype of insects from more recent resistance outbreaks in New South Wales (SR NSW ) and South Australia (SR SA ). Results of complementation analysis involving the weakly resistant strain WR QLD suggest that the gene responsible for the weak resistance phenotype, rph1, contributes to resistance in both SR NSW and SR SA . Interaction between rph1 and a second gene(s) is likely responsible for the strong resistance phenotype exhibited by these strains. However, it was not determined whether or not any genes, other than rph1, were shared between the two strains. To date, the genetics of resistance in the second strong resistance strain collected from southern Queensland (referred to as SR2 QLD in the remainder of this report) has not been determined. Thus, it was not known whether the rph1 and rph2 genes originally identified in SR QLD also were responsible for resistance in SR2 QLD . This paper reports the genetic analysis of SR2 QLD as well as three additional strains that had been found previously to contain a resistance allele at the rph1 locus. These four strongly resistant strains, SR QLD , SR2 QLD , SR SA , and SR NSW , originate from four widely separated geographical regions of Australia. They have not previously been analysed to determine whether the rph2 gene originally identified in SR QLD also contributes to their resistance phenotype. Knowing the genetic basis of resistance in these strains will allow us to predict how resistance will develop in the field. We also have created a laboratory strain that contains the resistance alleles from each of the four strongly resistant strains. Our results are consistent with the same two genes contributing to resistance in each of the four resistance outbreaks. Not surprisingly, crossing these strains and reselecting for high level resistance does not result in a major increase in the resistance level.

Insect Strains
Four strongly resistant strains (SR NSW , SR SA , SR QLD and SR2 QLD ) and a weakly resistant strain (WR QLD ) were employed in this study. Two strains (SR NSW , SR SA ) were collected from Merriwagga in south-western New South Wales (1999) and Port Adelaide in South Australia (2000), respectively, and were assigned the following collection reference numbers: NNRD2864 and NSRD3075 [17]. Strains SR QLD and SR2 QLD were collected from Millmerran (1997) and Wandoan (1998) in southern Queensland and were assigned collection reference numbers QRD569 and QRD676, respectively. The weakly resistant strain (WR QLD ) was also collected from Millmerran (1997) and was assigned reference number QRD369. The approximate distance between the geographic origins of any two strongly resistant strains was 250-1500 km. All resistant strains were selected with phosphine for at least five generations to promote homozygosity at resistance loci. All strains were cultured on whole wheat and maintained at 30uC and 55% relative humidity.

Complementation and allelic relationship analysis
Complementation analysis was carried out to determine whether the rph1 gene that controls weak resistance to phosphine in WR QLD and contributes to strong resistance in SR QLD [19], also contributes to strong resistance in the second strongly resistant strain from Queensland, SR2 QLD . The test involved crossing WR QLD and SR2 QLD and determining the resistance phenotype of the F 1 and F 2 progeny as previously described for SR QLD [3]. In addition, crosses between all four strongly resistant strains were made to determine the relationships between their respective resistance alleles. Crosses representing all six pairwise strain combinations were produced from the four strongly resistant strains: SR NSW xSR SA , SR QLD xSR NSW , SR QLD xSR SA , SR QLD xSR2 QLD , SR NSW xSR2 QLD and SR2 QLD xSR SA . F 1 and F 2 progenies were generated from each cross and the response to phosphine exposure of individuals within these progenies were examined to determine the relationships between the resistance genes of the parental strains.

Interaction between resistance genes when combined in a single strain
We also determined whether a strain could be selected with enhanced resistance to phosphine from a population containing all resistance genes from each of the four independently derived strongly resistant strains employed in this study. To combine all the resistance genes from the four strains in a single population, we employed two different crossing strategies; combined crosses and double crosses.

Combined crosses
Combined crosses provided the easiest way to introduce all resistance genes into a single population. Initially, two pairwise crosses were made from the four parental strains, SR QLD xSR SA and SR NSW xSR2 QLD , by mating an adult virgin female with an adult male on kibbled grain inside a plastic capsule. Five identical matings were set up for each of the two crosses. After two weeks, the adults were removed to fresh kibbled grain and the old grain containing the progeny from each set of five identical matings was pooled in a single plastic cup. This cup was topped up with fresh kibbled grain and the F 1 progeny were allowed to mature. Fifty mature (1-2 weeks post eclosion) F 1 progeny from each of the two crosses were combined and allowed to mate freely for two weeks to produce the combined progeny. This is referred to as the combined crosses (CC). Collectively, the progeny of the combined crosses would carry all resistance alleles from the four parental strains. The resulting F 1 progeny of the combined crosses were, again, allowed to mate freely for two weeks to produce the F 2 generation. Approximately 600 adult F 2 progeny were allowed to mate freely to produce an F 3 generation. Selections for phosphine resistance commenced in the F 3 generation at a concentration of 0.5 mg/L phosphine. The survivors of this selection were retained to produce subsequent generations. The second and third selections were performed on the F 5 and F 7 generations, both at 1.0 mg/L phosphine. A parallel population of F 7 individuals that had previously been subjected to two rounds of selection were also tested for their resistance phenotype. A second mortality response curve was also performed in the F 9 generation, by which time the line had already been subjected to three rounds of phosphine selection.

Double Hybrid Crosses
The second approach was to set up defined crosses in each of two generations to establish a doubly hybrid strain. Initially, two single crosses were made, SR QLD 6SR NSW and SR2 QLD 6SR SA . Subsequently, F 1 individuals of the two single crosses were mated to produce double cross progeny. This ensured that the parental strains contributed equally to the genotypes present in the resulting progeny. These crosses involved the same four parental strains used for the combined crosses (CC) but in different pairwise combinations in case this influenced the likelihood of selecting specific resistance genotypes.
The initial crosses were carried out as reciprocal single pair crosses between virgin individuals, for a total of ten crosses for each pair of strains. The parents from the initial crosses were pooled on fresh kibbled grain to establish the two single cross strains (SR QLD 6SR NSW ) and (SR2 QLD 6SR SA ). Reciprocal crosses were then set up between progeny of the initial crosses to create 20 double crosses, each of which produced progeny derived from all four parental strains. The adults were removed and the grain containing the eggs of 20 double cross mating pairs was combined in a bottle containing ,200 g whole grain. The resulting adult (1-3 week post eclosion) F 1 double cross progeny were transferred to new grain and left to mass cross for two weeks to produce an F 2 generation. The F 3 generation was produced through mass crosses of approximately 600 F 2 adults. Successive generations were produced by mass crossing progeny of the previous generation. Selection for phosphine resistance was carried out on adults of the F 3 , F 5 and F 7 generations. Progeny of both the initial single crosses and subsequent double crosses were selected at 0.5 mg/L phosphine in the F 3 generation, but 1.0 mg/L phosphine in the F 5 and F 7 generations. Phosphine resistance was quantified in the F 7 and F 9 generations.

Phosphine fumigation
Phosphine resistance of the parental strains and the progeny of the crosses was measured by exposing insects to phosphine fumigation for 48 hours [3] at a range of phosphine concentrations (0.01-2.0 mg/L). Phosphine was generated in a collection tube containing aluminium phosphide introduced into a 5% sulphuric acid solution [21]. Phosphine concentration was determined by gas chromatography [Varian (Mount Waverley, Victoria, Australia) aerograph model 90-P] utilising dichlorofluoromethane (Refrigerant F24; Lovelock Luke, Mayne, Queensland, Australia) as the carrier gas and a gas density balance detector.
Adult beetles (1-3 weeks post eclosion) were confined within small plastic cups (50 beetles per cup) containing 5 g whole grain.
The cups were placed inside gas-tight desiccators and phosphine was injected into the desiccators through a septum. The insects were exposed to phosphine for 48 hours at 25uC and 70% r.h. then held for 14 days at 25uC and 55% r.h. when end-point mortality was assessed. A minimum of 100 insects was fumigated at each phosphine concentration.

Data analysis
Mortality data for each strain or hybrid line were subjected to log-concentration/probit-regression analysis [22]. Mortality data were first corrected for control mortality (#10%) based on Abbott's formula [23]. The probit analysis was carried out using the GenStat7 statistical package [24]. The goodness-of-fit to the log-dose/probit mortality line was determined by a chi-square test.
In the goodness-of-fit calculation, at doses where the expected response was less than one, the number of observed responses was combined with the value for an adjacent dose and the degrees of freedom for the chi-square analysis were adjusted accordingly. The resistance factors for both single cross and double cross progenies were calculated by dividing the LC 50 of each progeny generation by the mean LC 50 of the parental strains.

Complementation and allelic relationship analysis
The resistance phenotype of the four strongly resistant R. dominica strains that have been characterised in Australia differ by at most three fold. The rank order of resistance is SR SA > SR2 QLD ,SR QLD >SR NSW . Previous work suggested that evolution of resistance was constrained in the order in which genetic changes could occur [19]; that two genes contributed to resistance in multiple instances and that at least one resistance gene was common to multiple highly resistant strains [20]. The following comparative genetic experiments use complementation tests and gene stacking to reveal previously unknown resistance mechanisms or previously unidentified resistance genes.

The rph1 gene contributes to resistance in SR2 QLD
The second strongly resistant strain from Queensland, SR2 QLD , is distinct from SR QLD that was analysed previously [18] ( Schlipalius et al. 2002). SR2 QLD was crossed with the weakly resistant strain, WR QLD , which is homozygous for the resistance allele of the rph1 gene. Probit analysis revealed that the responses of parental strains SR2 QLD and WR QLD as well as their F 1 progeny were linear ( Figure 1). The relevant chi-square values of Figure 1. Resistance response of F 1 hybrids and F 2 progeny of a cross between a weakly resistant and a strongly resistant R. dominica strain from Queensland. Results are presented as log-dose mortality of the F 1 hybrids and the F 2 progeny with reference curves of the parental strains, WR QLD (Weak R-Strain) and SR2 QLD (R-Strain). Phosphine exposure was for 48 hours at 25uC and 70% r.h. doi:10.1371/journal.pone.0034027.g001 the response data were 1.407 (df = 7, p = 0.985) and 4.95 (df = 7, p = 0.666) for SR2 QLD and WR QLD respectively, and 2.717 (df = 5, p = 0.744) for the F 1 .
The F 1 progeny of this cross would be expected to exhibit a resistance phenotype at least as strong as the weakly resistant strain, WR QLD , if the rph1 gene contributes to resistance in SR2 QLD . If this gene does not contribute to resistance in SR2 QLD , the hybrid progeny would be heterozygous for the incompletely recessive resistance alleles and nearly completely sensitive to phosphine. The F 1 hybrids are more resistant than WR QLD , as expected if the rph1 gene contributes to resistance in both parental strains ( Figure 1).

Genetic complementation analysis of SR NSW 6SR SA
A resistance allele of the rph1 gene was previously found to contribute to the phosphine resistance phenotype of both SR NSW and SR SA via complementation analysis with the weakly resistant strain WR QLD [20]. In the present study, SR NSW and SR SA were crossed and their hybrid progeny analysed to confirm lack of complementation at rph1, as well as to determine relationships between resistance alleles at other loci. Probit analysis of both parental strains and their F 1 progeny revealed linear responses ( Figure 2, Table 1), indicating that the strains are homogenous with respect to their resistance phenotypes.
The previous report that both strains contain resistance alleles of rph1 suggests that the hybrid progeny should at least show a level of resistance equivalent to that of the reference strain, WR QLD , which is homozygous resistant at rph1. The F 1 hybrid progeny actually show a much higher level of resistance than seen for WR QLD (Figure 2), a level of resistance equivalent to that of the strongly resistant parental strain SR SA . As both strains are known to carry a semi-dominant resistance allele at a second locus [20], the results could be explained as an additive semi-dominant effect of two distinct genes (in addition to homozygosity at rph1). However, the resistance phenotype due to the second gene in each strain differs by about two fold. Thus, the results could also be explained as two distinct alleles in the same gene that, in combination, provide a level of resistance dictated by the weaker of the two alleles.
Interpretation of the F 2 results is complicated by the fact that there is only a two fold difference in the LC 50 between the parental strains, SR SA and SR NSW . The narrow distance between the parental mortality curves makes it difficult to identify a plateau in the F 2 response curve as an indicator of monogenic control of resistance (beyond that caused by the rph1 locus). Because the level of resistance due to the second gene differs between the strains three distinct F 2 genotypes would result, even if the second resistance factor in each strain was actually an allele of the same gene. This increases the difficulty of distinguishing a plateau in the curve. One thing that is clear is that none of the F 2 individuals approach the sensitivity of WR QLD as would be expected of 6.25% of the individuals if the second resistance factors were not allelic. The most reasonable conclusion is that the strong resistance phenotype exhibited by these strains is controlled by rph1 together with a second gene that is common to the two strains, though the strengths of the alleles of the second gene differ considerably.
Genetic complementation analysis of SR NSW 6SR QLD SR NSW was also crossed to SR QLD , a well-characterised strongly resistant strain and the first strongly resistant strain Figure 2. Resistance response of F 1 hybrids and F 2 progeny of a cross between strongly resistant R. dominica strains from South Australia and New South Wales. Results are presented as log-dose mortality of the F 1 hybrids and the F 2 progeny, together with reference curves of the parental strains, SR SA and SR NSW , and the weakly resistant strain from Queensland (WR QLD ). Phosphine exposure was for 48 hours at 25uC and 70% r.h. doi:10.1371/journal.pone.0034027.g002 collected in Australia. Responses of SR NSW and SR QLD , as well as their F 1 progeny were linear ( Figure 3, Table 1). Response curves of the parental strains are almost overlapping, indicating a very similar level of resistance between the two strains, with SR NSW only 1.2 times the resistance of SR QLD . The hybrid of this cross is slightly more sensitive to phosphine than is either parent, though it is still much more resistant (.10-fold) than the weakly resistant strain WR QLD . Both parents are known to be homozygous for a recessive resistance allele at rph1 [3,20] as well as homozygous for a second gene that is weakly semi-dominant.
The semi-dominance can not explain the very high level of resistance in the F 1 progeny rather, the resistance is equivalent to the synergistic action between the rph1 and rph2 genes previously reported for SR QLD [19]. Thus, the strong resistance exhibited by the F 1 progeny of this cross suggests that, as with SR QLD , rph1 and rph2 are responsible for resistance in SR NSW . The lack of a minor effect resistance factor that is unique to one or the other parental strain could completely explain the very minor decrease in resistance of the hybrid progeny relative to their parents. As with the F 1 progeny, a small proportion (5%-25%) of the F 2 progeny are more sensitive than the parental strains. Most of the F 2 progeny, on the other hand, show a resistance level between those of the parental strains ( Figure 3). The F 2 progeny would be expected to form a plateau at ,75% mortality if a single gene in addition to rph1, controls resistance in the stronger resistant parent SR NSW . However, as the response curves of both parental strains are overlapping, it is impossible to determine whether a plateau occurs at this point. Interpretation of the F 2 results is further complicated by the weak incomplete recessivity of rph2 in SR QLD [3,18] and a second resistance gene (possibly rph2) in SR NSW [20]. On balance, the F 2 results are completely consistent with the F 1 data, and indicate that a resistance allele of rph2 does indeed contribute to resistance in SR NSW . The results also suggest that the previously noted minor, dominant resistance factor from one of the parents is likely the product of a single gene.
Genetic complementation analysis of SR QLD 6SR SA As was done previously with the strongly resistant strain from New South Wales, the strongly resistant strain from South Australia, SR SA , was also crossed with its Queensland counterpart, SR QLD , which is homozygous for both phosphine resistance genes rph1 and rph2. Probit analysis of the response to phosphine of the parental strains and F 1 progeny of this cross indicate a homogeneous response (Figure 4). Unlike the previous SR QLD 6SR NSW cross in which the parental strains were nearly equally resistant, the resistance phenotype of the SR SA strain is less than half that of the strain from Queensland, SR QLD . The F 1 progeny exhibited a slightly more resistant phenotype than the parental strain SR SA . The response of the F 1 is as would be expected from non-complementation at the rph1 locus, indicating that each of the two parental strains is homozygous for a resistance allele of the rph1 gene. This is consistent with a previous detailed analysis in which the rph1 gene was found to contribute to resistance in each of SR SA , SR NSW [20] and SR QLD [18]. The additional resistance is presumably due to additional factors from the parental strains. Because of the comparative weakness of the strong resistance phenotype of SR SA , it is not possible to determine whether the resistance phenotype of the F 1 is due to noncomplementation at the rph2 locus or an additive effect of the incompletely recessive allele previously attributed to the second locus [20].
A small proportion of the F 2 progeny (15%-40%) are more sensitive than each of the parental strains as well as the F 1 progeny. This is similar to the observation of a minor effect resistance allele in the previous SR QLD 6SR NSW cross. The F 2 response curve demonstrates no clear indication of a plateau at ,75% mortality. As rph1 and rph2 genes are responsible for the strong resistance of SR QLD , the lack of a plateau may either indicate that an rph2 resistance allele is not present in SR SA or that interpretation of the resistance phenotype is clouded by incomplete recessivity of rph2 together with the phenotype of the minor effect gene. The data indicate that strong resistance of SR SA is controlled by rph1 plus a second major effect gene that may simply be a weak allele of rph2. While the data are consistent with this explanation, they do not rule out alternative explanations for the second major effect gene. Genetic complementation analysis of SR QLD 6SR2 QLD As with SR NSW and SR SA , we also crossed SR QLD with SR2 QLD , a second strongly resistant strain collected from Queensland. Probit analysis revealed a linear response curve to phosphine exposure for each parental strain as well as for the F 1 progeny ( Figure 5), suggesting genetic homogeneity of each strain as well as the hybrid progeny. The resistance of the two parental strains differs by only 1.6 fold, with SR QLD being more resistant.
The F 1 progeny are highly resistant to phosphine exposureslightly more than SR2 QLD . This indicates that the second Queensland strain, SR2 QLD , is not able to complement the original strongly resistant strain from Queensland, SR QLD , at either rph1 or rph2. Thus, rph1 and rph2 are responsible for the strong resistance phenotypes of each strain. As with the preceding analyses, a small fraction of the progeny is unusually sensitive to phosphine exposure. This would appear to result from the lack of a minor dominant resistance factor contributed by one of the two parental strains. The fact that this minor, dominant factor has been apparent in crosses between SR QLD and each of three independent strains, strongly implicates SR QLD as the source of this additional resistance factor, a possibility that was first noted in Collins [3].

Genetic complementation analysis of SR2 QLD 6SR NSW
A cross between SR2 QLD and SR NSW was also made to determine allelic relationships between phosphine resistance genes in the two strains. Probit analysis revealed that the parental strains and the F 1 progeny all show a homogeneous response to phosphine, indicated by linear phosphine resistance response curves ( Figure 6). As with the cross with the F 1 progeny of SR2 QLD 6SR QLD , the F 1 progeny of SR2 QLD 6SR NSW have a phenotype that is intermediate between the two strains. This indicates that resistance alleles of rph1 and rph2 are present in both strains.
The F 2 progeny are mostly more resistant than the parental strain SR2 QLD , though none is as resistant as SR NSW (Figure 6). The F 1 and F 2 data together with results from crosses with the reference strain, SR QLD suggest that the strong resistance phenotypes of SR2 QLD and SR NSW are due to a resistance alleles of the rph1 and rph2 gene with the possibility of additional genes of minor effect.

Genetic complementation analysis of SR SA 6SR2 QLD
We also crossed the second resistant strain from Queensland, SR2 QLD with the strain from South Australia, SR SA . Each of the parental strains and the F 1 hybrid progeny showed linear response curves ( Figure 7, Table 1). The resistance level of strain SR2 QLD is only 1.26 fold higher than strain SR SA , demonstrated by the very close proximity of their response curves. The resistance phenotype of the F 1 progeny of this cross was similar to that of the parental strain SR SA . The F 2 progeny were intermediate between the two parental strains (Figure 7). This indicates that alleles at the same loci, rph1 and probably rph2, contribute to resistance in both SR SA and SR2 QLD strains.

Complex genetic crosses
The most reasonable explanation of the preceding genetic results is that all four highly resistant strains simply carry alternative alleles of the same two resistance genes. As demonstrated with the pairwise crosses, hybrids between such strains in the field should not lead to a dramatic increase in resistance to phosphine. The crosses that follow were designed to test the effect of combining all four resistance genotypes in a single strain and subjecting it to strong selection. The goal was to determine the resistance phenotype that would result from genetic combinations that had not been tested in the pairwise crosses. This includes testing the effects of resistance factors other than rph1 and rph2 after repeated selection for homozygosity of recessive alleles. The first of two strategies that were used consisted of setting up two pairwise crosses between strains that differed most strongly in their resistance phenotypes. The F 1 progeny were then pooled to establish a ''combined cross'' strain. The second strategy consisted of setting up two pairwise crosses between strains that were most similar in their resistance phenotypes. This was followed by setting up a cross between hybrid F 1 s from the initial crosses to establish a ''double hybrid'' strain.

Combined crosses (CC)
The mortality response data of each of the four parental strains used to establish the line with the combined genotype were subjected to probit analysis as presented in Table 1. The LC 50 values of the parental strains are non-overlapping according to their 95% fiducial limits, indicating that the response phenotypes were unique. The strains that were initially crossed exhibited an approximately 2-fold difference in resistance levels ( Table 1).
The combined strain was produced in two steps. Initially two single crosses were produced, SR SA 6SR QLD and SR2 QLD 6 SR NSW . The progeny of these two crosses were then combined to establish an F 2 generation. Selection for resistance was carried out at 0.5 mg/L phosphine in the F 3 generation, 1.0 mg/L in the F 5 generation and 1.0 mg/L in the F 7 generation. Phosphine resistance of the combined cross was tested at the F 7 and F 9 generations, after having been selected for phosphine resistance two and three times, respectively. The F 7 generation had a level of resistance 1.4 fold that of SR NSW and in the F 9 generation resistance had increased to 1.9 fold. Taken together, after 9 cycles of breeding and three selections with phosphine, the progeny of the combined crosses showed an increase in resistance less than two fold higher than that of the most strongly resistant parental strain SR NSW . This level of resistance is no more that previously observed for the minor effect gene contributed by SR QLD .
Each of the response curves of the F 7 and F 9 progeny was linear suggesting that the line was quickly driven to genetic homogeneity by as few as two rounds of selection (Figure 8). The slopes of the response curves of the three weakest parental strains were nearly parallel (Figure 8), whereas the slope of SR NSW more closely matched those of the F 7 and F 9 progenies of the combined cross ( Figure 8). This suggests that a genetic factor from SR NSW has been selected in the progeny. Figure 6. Resistance response of F 1 hybrids and F 2 progeny of a cross between strongly resistant R. dominica strains from Queensland and New South Wales. Results are presented as log-dose mortality of the F 1 hybrids and the F 2 progeny with reference curves of the parental strains, SR2 QLD and SR NSW , and the weakly resistant strain from Queensland (WR QLD ). Phosphine exposure was for 48 hours at 25uC and 70% r.h. doi:10.1371/journal.pone.0034027.g006

Double Crosses (DC)
We also combined all four strong resistance genotypes using an alternative strategy in which the two most strongly resistant strains, SR NSW and SR QLD , were crossed as were SR SA and SR2 QLD . A double hybrid line was then produced by crossing individual offspring from each of the two single crosses. Selection for resistance was carried out in the F3, F5 and F7 generations as described for the combined crosses except that the single cross lines were also subjected to selection. The parental, single-cross and double-cross strains all exhibited homogeneous response curves with the exception of the single cross (SR SA 6SR2 QLD ) at the F 9 generation which gave a significant x 2 result (P,0.05) with a heterogeneity factor of 2.39 ( Table 2).
All progenies of the single crosses exhibited a significantly higher level of resistance than the parental strains from which they were derived. Similarly, the F 9 generation of the double cross was significantly more resistant than any of the parental lines, including the single cross lines from which the double cross lines were derived. Even the resistance level of the F 7 progeny of the least resistant of the two single crosses SR SA 6SR2 QLD is essentially equivalent to that of the strongest resistant strain SR NSW given the overlapping LC 50 (95% Fiducial Limit) (Table 2, Figure 9). These results are tabulated as resistance factors (ratios of LC 50 values) in Table 3.
The F 9 (selected 3 times) progeny of each single and double cross had a resistance factor from 1.5 to 2.1 times that of the most resistant parental strain from which it was derived (shown in bold in Table 3). The resistance factor of the double cross strain increased between the F 7 and F 9 generations by 30%. Interestingly, this was simply the sum of the increases of the two single cross strains (12% and 18%). The small changes and similarity between the strains indicates that the genetic interactions are simply additive as would be expected if sensitive alleles of minor effect genes were being progressively eliminated.

Allelic relationships between phosphine resistance genes
Due to the growing threat of resistance across the world, understanding the genetics of phosphine resistance will provide globally significant insight into effective phosphine resistance management strategies. The present work is a continuation of previous research in which we found that the rph1 gene contributed to the strong resistance phenotype in three of the strains that have been re-examined in this study [20]. We also demonstrate that an allele of rph1 contributes to resistance in a fourth strongly resistant strain, SR2 QLD . In addition to rph1, a second genetic factor that is semi-dominant contributes to the strong resistance phenotype of all four strains SR2 QLD , SR QLD , SR NSW and SR SA . This second gene was previously characterised in SR QLD and is named rph2.
Crosses between the strongly resistant strains provided additional insight into relationships between phosphine resistance alleles from the four resistant strains. These results confirm that the rph1 gene contributes to resistance in each of SR NSW and SR SA [20], SR QLD [18,19] and SR2 QLD . The results also suggest that the incompletely recessive synergistic factor first noted in SR QLD [3] is a synergistic resistance factor in each of the four strains. Schlipalius et al. [19] proposed that the evolution of resistance was constrained by the fact that the rph2 gene is relatively insignificant as a resistance locus in the absence of the resistance allele at rph1. We now provide Figure 8. Resistance response of selected F 7 and F 9 progenies of a pooled hybrid (combined crosses) progeny of two single crosses between the strong resistant R. dominica strains, SR QLD 6SR SA and SR NSW 6SR2 QLD . Results are presented as log-dose mortality of the F 7 and the F 9 progenies with reference curves of the parental strains, SR QLD , SR2 QLD , SR NSW and SRSA. Phosphine exposure was for 48 hours at 25uC and 70% r.h. doi:10.1371/journal.pone.0034027.g008 evidence that the evolution of resistance in independent outbreaks is constrained to just two major genes in R. dominica. In addition, a minor effect, dominant resistance factor was identified that contributes about 2-fold resistance. It is interesting to note that strong resistance toward phosphine in Tribolium castaneum is also due to two synergistically interacting genetic factors, suggesting that elements of resistance may be shared between species as well [25].
Combining resistance genotypes to establish maximal resistance levels We also combined resistance genotypes from all four strongly resistant strains to determine if enhanced resistance to phosphine could be produced. We employed both mass-combined and defined double-hybrid crossing strategies to combine all resistance alleles in a single population. The progenies of both masscombined and double-hybrid crosses exhibited increased levels of resistance relative to the parental strains after either two or three rounds of selection for phosphine resistance. The highest resistance levels obtained from the selected progenies of both mass-combined and double-hybrid crosses were 1.8 and 1.9-fold higher than the most resistant parental strain, SR NSW . This level of increase in resistance can most likely be attributed to genes of minor effect contributed by the genetic backgrounds of the parental strains.
The present study confirms and extends our previous understanding that the rph1 gene is a common contributor to resistance. We also present strong evidence that rph2 together with rph1 explains nearly all of the strong resistance phenotype. A few additional resistance factors appear to contribute to the resistance of the strains that we have investigated in this paper. The effect of such minor genes is to increase resistance about 2-fold beyond the Figure 9. Resistance response of selected F 7 and F 9 progenies of double crosses between the strong resistant R. dominica strains from Australia. Results are presented as log-dose mortality of the F 7 and the F 9 progenies with reference curves of the F 7 and F 9 progenies the single crosses, SR QLD 6SR NSW and SR SA 6SR2 QLD , and the parental strains, SR QLD , SR2 QLD , SR NSW and SRSA. Phosphine exposure was for 48 hours at 25uC and 70% r.h. doi:10.1371/journal.pone.0034027.g009 Table 3. Resistance factor of phosphine selected F 7 and F 9 progenies of both single and double crosses relative to their respective parental strains/lines. level of the most strongly resistant parental strain. The overarching hypothesis from this work is that limited genetic mechanisms are responsible for all strong resistance outbreaks in R. dominica and possibly other species as well. These results, however, do not rule out the possibility that novel resistance genes may eventually be isolated from R. dominica or other species in Australia or elsewhere. In this regard, the extremely high level of phosphine resistance recently observed in Cryptolestes ferrugineus is a prime candidate for further study [26].
We have now demonstrated that four strongly phosphine resistant strains of R. dominica each carry alleles of rph1 and rph2 that are responsible for nearly the entire resistance phenotype. This finding is quite remarkable given that phosphine is a very small and reactive molecule that could potentially have many target sites within a cell. Nevertheless, our findings suggest that the job of monitoring and managing resistance will be much more manageable than could have been the case if the genetic basis of resistance was more complex. It remains to be determined whether the same genetic basis of resistance extends to other species. If this is the case, it will allow development of a universal marker that will be useful in efficient monitoring and management of phosphine resistance.
The fact that homozygosity of rph1 alone confers only weak resistance to each of the four strains, suggests that as with SR QLD , the strong resistance phenotype is due to a synergistic interaction between rph1 and rph2. This means that the phosphine resistance problem can be alleviated by strategies that disrupt resistance caused by either rph1 or rph2, as such strategies need not target both mechanisms to be effective.
Development of diagnostic markers for monitoring resistance will be greatly facilitated by cloning of the resistance gene. Cloning of the gene will also allow comparative genetic analysis of resistance between species. Identification of the gene will also facilitate detailed genetic studies into the mode of action and mechanisms of resistance toward phosphine. This type of work is made more valuable by the outcome of the current study, which suggests that the resistance genes that we have identified may define the extent of the problem that will be faced by the grains industry.