Introduction
The Tephritidae, or true fruit flies, represent a highly diverse dipteran family of over 4000 globally distributed species (White & Elson-Harris, Reference White and Elson-Harris1992). The majority attack the seed-bearing organs of plants, with many species inflicting significant economic losses to commercial horticulture through fruit destruction (White, Reference White1996).
As with any pest insect, management of tephritids relies on accurate identification of the target species (Walter, Reference Walter2003); however, several important tephritid pest species remain difficult to identify as they belong to cryptic, sibling complexes. Sibling species can potentially vary in important traits, such as host use and pest status, geographic distribution and seasonal phenology, yet their study and management remains confounded due to high levels of morphological similarity (Clarke et al., Reference Clarke, Allwood, Chinajariyawong, Drew, Hengsawad, Jirasurat, Krong, Kritsaneepaiboon and Vijaysegaran2001; Garros et al., Reference Garros, Van Bortel, Trung, Coosemans and Manguin2006; Barik et al., Reference Barik, Sahu and Swain2009). Importantly, mating behaviour will also vary among sibling species, which has implications for the sterile insect management technique for which the mating success of treated insects is critical (Dyck et al., Reference Dyck, Hendrichs and Robinson2005).
One tephritid species complex, the Bactrocera dorsalis complex, contains over 70 described species (Drew & Hancock, Reference Drew and Hancock1994), some of which are regarded as the most destructive horticultural pests within their native south-east Asian range (Clarke et al., Reference Clarke, Armstrong, Carmichael, Milne, Raghu, Roderick and Yeates2005). Moreover, the invasive movement of certain species within the complex, including the spread of B. carambolae Drew & Hancock into South America during the mid 1970s (van Sauers-Muller, Reference van Sauers-Muller1991) and the recent expansion of B. invadens Drew, Tsuruta & White into Africa (Drew et al., Reference Drew, Tsuruta and White2005; Khamis et al., Reference Khamis, Karam, Ekesi, De Meyer, Bonomi, Gomulski, Scolari, Gabrieli, Siciliano, Masiga, Kenya, Gasperi, Malacrida and Gugielmino2009) renders this group a truly global problem. Despite their economic importance, identification of some species of the complex is difficult and is of ongoing concern (Clarke et al., Reference Clarke, Armstrong, Carmichael, Milne, Raghu, Roderick and Yeates2005). This is a particular challenge as some of the most destructive pest species in the complex are also the most morphologically similar (=cryptic), especially B. papayae Drew & Hancock, B. carambolae, B. philippinensis Drew & Hancock and B. dorsalis s.s. (Hendel). Difficulties chiefly arise due to broad intraspecific morphological variation between members of the B. dorsalis complex, which can seriously confound their identification (Drew et al., Reference Drew, Raghu and Halcoop2008). Without supporting information such as geographical origin (itself not a true ‘taxonomic character’), unambiguous species identification of these four species is often impossible. Further, diagnostic characters used in taxonomic keys are sometimes described using subjective or relational terminology, such as ‘narrow’, ‘slightly expanded’ and ‘occasionally’ (Lawson et al., Reference Lawson, McGuire, Yeates, Drew and Clarke2003; Drew & Hancock, Reference Drew and Hancock1994), making definitive identifications more difficult. Traditional morphometric approaches based on the length of the aedeagus have achieved some success in delineating species within the B. dorsalis complex, particularly between the sympatric species of B. occipitalis and B. philippinensis and, with some success, B. papayae and B. carambolae (Iwahashi, Reference Iwahashi1999a,Reference Iwahashib; Drew et al., Reference Drew, Raghu and Halcoop2008). Unfortunately, however, as aedeagus length is a continuous character state, even this measure is not always definitive. Traditional wing morphometric studies assessing vein length can effectively separate some members of the complex but not all of the four critical pest taxa listed above (Adsavakulchai et al., Reference Adsavakulchai, Baimai and Prachyabrued1998).
Whereas some other morphologically cryptic insect complexes have been adequately resolved using alternative approaches to morphological taxonomy, such as molecular data and ecological information (e.g. Scheffer & Lewis, Reference Scheffer and Lewis2001), efforts to discriminate between certain populations or species in the B. dorsalis complex have failed to reach a similar consensus. For instance, while molecular approaches have achieved some headway in discriminating certain species of the B. dorsalis complex (Muraji & Nakahara, Reference Muraji and Nakahara2002; Naeole & Haymer, Reference Naeole and Haymer2003), there is yet to be found a consistent marker to adequately resolve species boundaries between the four abovementioned species, with the exception of B. carambolae for which molecular markers have been identified (Yong, Reference Yong1995; Armstrong & Cameron, Reference Armstrong, Cameron and Tan1998; Armstrong & Ball, Reference Armstrong and Ball2005). Similarly, studies of the constituents of the male pheromone gland reveal a high degree of similarity between most pest species of the complex, again with the exception of B. carambolae which possesses a distinct suite of pheromones compared to chemically similar B. dorsalis s.s., B. papayae and B. philippinensis (Fletcher & Kitching, Reference Fletcher and Kitching1995; Wee & Tan, Reference Wee and Tan2005).
As the biological species status of the four taxonomically described species in question (i.e. B. dorsalis s.s., B. papayae, B. carambolae and B. philippinensis) is not universally supported by alternate lines of evidence, there remain two possibilities, either (i) that the species as described are biologically valid, but appropriate diagnostic markers remain to be found, or (ii) diagnostic markers do not exist as these species are at such an early stage in their evolutionary divergence that sufficient ‘species criteria’ (sensu de Queiroz, Reference de Queiroz, Howard and Berlocher1998) are yet to accrue. Indeed, what we observe for these four species in question, i.e. a lack of consistent evidence of species boundaries, is precisely what we might expect given a case of a recent and rapid evolutionary radiation (Shaffer & Thomson, Reference Shaffer and Thomson2007), as is the case hypothesised for this tropical species complex (Clarke et al., Reference Clarke, Armstrong, Carmichael, Milne, Raghu, Roderick and Yeates2005). Given the need to seek resolution on the specific status of these flies for management and trade reasons, we are currently pursuing new methods for their discrimination by seeking to use tools suited to detecting fine-scale differences which may be expected following recent radiations (Shaffer & Thomson, Reference Shaffer and Thomson2007).
One such approach is shape analysis, the quantification of the relative positions of homologous structures, e.g. landmarks or outlines, and then comparing variation in shape data between individuals or groups, a technique known as geometric morphometrics (Rohlf & Marcus, Reference Rohlf and Marcus1993; Rohlf, Reference Rohlf1999). This technique has been applied across a range of taxonomic levels, from revealing cryptic species to discriminating intraspecific populations and to resolving relationships at the generic level (Gilchrist & Crisafulli, Reference Gilchrist and Crisafulli2006; Bouyer et al., Reference Bouyer, Ravel, Dujardin, De Meeüs, Vial, Thévenon, Guerrini, Sidibé and Solano2007; Marsteller et al., Reference Marsteller, Adams, Collyer and Condon2009; Michez et al., Reference Michez, Meulemeester, Rasmont, Nel and Patiny2009). While traditional morphometric approaches (i.e. linear measurements and their ratios) can be applied to shape analysis (Daly, Reference Daly1985), they are often used only to compare variation in size and proportion rather than to quantify shape itself (Drew et al., Reference Drew, Raghu and Halcoop2008). Geometric morphometric techniques, however, focus directly on shape variation and provide potentially greater taxonomic utility over traditional techniques, as shape is regarded as a more heritable character than size (Bitner-Mathé & Klaczko, Reference Bitner-Mathé and Klaczko1999; Birdsall et al., Reference Birdsall, Zimmerman, Teeter and Gibson2000; Dujardin et al., Reference Dujardin, Le Pont and Baylac2003). Further, continuous characters, such as shape, are hypothesised to be among the first to show differences following isolation events (Bouyer et al., Reference Bouyer, Ravel, Dujardin, De Meeüs, Vial, Thévenon, Guerrini, Sidibé and Solano2007). Therefore, sensitivity to small shape changes between populations may be particularly useful for detecting differences between groups of organisms following recent evolutionary radiations. As the B. dorsalis species complex is considered to contain taxa that have undergone a recent radiation (Clarke et al., Reference Clarke, Armstrong, Carmichael, Milne, Raghu, Roderick and Yeates2005), geometric morphometric analysis, therefore, may be sensitive enough to resolve differences should they exist.
The aim of this study, therefore, was to use geometric morphometric techniques to measure wing size and shape for previously identified specimens of B. dorsalis s.s., B. papayae, B. carambolae and B. philippinensis to determine: (i) whether wing size or shape is an effective discriminator between the currently defined species; and, if so, (ii) to what degree the species’ wings differ from each other; but, if not, (iii) should any of these species be suspected as conspecific based on morphometric shape data? Importantly, we specifically wanted to test the strength of this approach towards resolving specimens which had been a priori identified and labelled in the laboratory of R.A.I. Drew and colleagues (i.e. the original author/s of three of the four cryptic species) without further species identification undertaken by us. We believe that if wing shape data accurately reflected the species groupings as previously determined by the species’ original authors, this technique could also show promise as an effective technique to apply to new, unidentified specimens.
Materials and methods
Sample collection and curation
Individuals for four species from the B. dorsalis complex were chosen for analysis. These were: B. papayae, B. carambolae, B. philippinensis and B. dorsalis s.s. Additionally, one congeneric species from outside the B. dorsalis complex, B. tryoni (Froggatt), was included as an out-group to provide a relative measure of unambiguous interspecific variation.
Individuals were sourced primarily from the Queensland Department of Employment, Economic Development and Innovation (DEEDI) insect collection, located in Brisbane, Australia. Specimens from DEEDI were used for the following reasons: (i) in most cases the specific identities of individuals from the collection were determined by the original species’ authors (i.e. R.A.I. Drew and D. Hancock); and (ii) the specimens within the collection have been collected from across a broad spatial distribution, allowing us to account for that source of variation in our analysis. The only flies included in the study that were not sourced from the DEEDI insect collection were B. carambolae from Suriname, South America and B. tryoni from Brisbane, Australia. The Suriname B. carambolae were collected by A. van Sauers-Muller as pupae from infested carambola fruit (Averrhoa carambola) placed in the field at Paramaribo, Suriname during August 2009; and wild, cue-lure trap-caught males of B. tryoni were collected by MKS during November 2009 in Brisbane, Australia. Details of collection records for all flies used are given in table 1.
Table 1. Collection localities of B. dorsalis s.l. flies used in the current study.
N, number of individuals.
Identification of material from the DEEDI collection was based on determinations carried out by R.A.I. Drew when revising the B. dorsalis complex in the 1990s (Drew & Hancock, Reference Drew and Hancock1994). Material from Suriname is considered to be B. carambolae, based on identifications made at the time of the first invasion (van Sauers-Muller, Reference van Sauers-Muller1991). The out-group, B. tryoni, was identified by MKS. Only males were examined for all populations, as most available specimens had been collected using the male specific attractant, methyl eugenol. Where females were available (for example in the Suriname material), we chose to ignore them to avoid potentially sex biased results. We recognise the limitation of using one sex and, therefore, recommend that any expansion on this work should include both sexes where possible.
Between 54 and 59 individuals per species from at least three geographically distinct collection sites were randomly selected from available specimens for each in-group species (table 1). In some instances, flies from more than one trap and from more than one collection date were used; each group of flies from a collection site is termed a ‘sample’. Within the limits of the specimens available, samples were chosen to maximise the geographic distance between collection sites so as to increase the amount of potential intraspecific variation measured. Generally, one sample came from one site, the only major exception being B. carambolae from ‘TN’ (table 1). These flies were collected along ‘Transect North’, Peninsular Malaysia (Drew & Hancock, Reference Drew and Hancock1994). There are no precise site details for the collection sites along the transect; however, each TN site is described as being approximately 20 km apart along a 200–250 km distance between the extreme localities of Kuala Kangsar (TN1) and Kota Bharu (TN17), Peninsular Malaysia (Drew & Hancock, Reference Drew and Hancock1994). Samples in the present study were from TN6, 12 and 14 (table 1).
All specimens were assigned a cross-referenced code number between QUT 001 and QUT 284, which was affixed to both (i) the slide-mounted wing and (ii) the pinned or alcohol-preserved voucher specimen (n.b. not all slide-mounted wings were included in the analysis due to damage or obscured landmarks). Only the following DEEDI insect collection material had DEEDI database accession numbers: B. carambolae from Kuala Kangsar Malaysia (selected specimens between QDPC 0-141684 and QDPC 0-141777) and B. carambolae from the TN Malaysia transect (selected specimens between QDPC 0-141663 and QDPC 0-141742). All loaned material is due to be redeposited in the original collection at the expiry of the loan (2012), while remaining voucher specimens are maintained at the Queensland University of Technology, Brisbane, Australia.
Morphometric analysis
In most specimens, the right wing of each fly was removed and slide-mounted using Canada balsam. In some instances (19% of flies sampled), the right wing was either damaged or missing and so the left wing was used instead. As fluctuating asymmetry (differences between the left and right wing) does not occur in other Bactrocera species (Gilchrist & Crisafulli, Reference Gilchrist and Crisafulli2006), we believe that the occasional (and non-systematic across-samples) use of the left wing would not bias the analysis. Wings were photographed at 10×magnification using an AnMo Dino-Eye microscope eye-piece camera (model AM423B) mounted into a Leica MZ6 stereo-microscope and saved in bitmap format using the computer program DinoCapture v3.2.0.5 (produced by AnMo). Fifteen homologous Type 1 landmarks (Bookstein, Reference Bookstein1991) were selected for comparison, these being the following vein junctions, vein terminations or vein suture: (1) basal junction of veins of cell bm; (2) anterior-most point of the suture located towards the base of vein sc; (3) inner antero-distal corner of cell bc; (4) junction of veins A1 and CuA2; (5) junction of CuA1 and CuA2; (6) junction of vein CuA1 and dm-bm cross vein; (7) junction of vein M and dm-bm cross-vein; (8) junction of vein CuA1 and dm-cu; (9) junction of vein M and dm-cu; (10) junction of vein M and r-m cross-vein; (11) junction of vein R4+5 and r-m cross-vein; (12) junction of vein R1 and costal vein; (13) termination of vein M; (14) termination of vein R4+5; and (15) termination of vein R2+3 (fig. 1). Each landmark was digitised using the computer program tpsDig2 v2.12 (Rohlf, Reference Rohlf2008) for which x, y coordinates were generated and saved as a text file.
Fig. 1. Right-hand wing of a Bactrocera dorsalis s.s. individual showing each of the 15 landmarks used in the geometric morphometric analysis. Scale bar=1 mm. 1, basal junction of veins of cell bm; 2, anterior-most point of the suture located towards the base of vein sc; 3, inner antero-distal corner of cell bc; 4, junction of veins A1 and CuA2; 5, junction of CuA1 and CuA2; 6, junction of vein CuA1 and dm-bm cross vein; 7, junction of vein M and dm-bm cross-vein; 8, junction of vein CuA1 and dm-cu; 9, junction of vein M and dm-cu; 10, junction of vein M and r-m cross-vein; 11, junction of vein R4+5 and r-m cross-vein; 12, junction of vein R1 and costal vein; 13, termination of vein M; 14, termination of vein R4+5; 15, termination of vein R2+3.
Centroid size was used as the measure of size for each fly wing, calculated using the computer program morphologika2 v2.5 (O'Higgins & Jones, Reference O'Higgins and Jones2006). Centroid size is an isometric estimator of size, calculated as the square root of the summed squared distances of each landmark from the centre of the landmark configuration (i.e. the mean position of all co-ordinates). Tests for significant differences among samples for each a priori defined species (based on their original classification) were undertaken using a one-way ANOVA with a Tukey post-hoc test. Further, comparisons among species (with intra-specific groups combined) were also tested using a one-way ANOVA with a Tukey post-hoc test; however, the individuals of the B. carambolae sample from TN Malaysia were removed from this analysis (see reasons below).
Raw landmark co-ordinate data were imported into the program CoordGen6f (part of the IMP software series (Sheets, Reference Sheets2006)) and aligned using a generalised Procrustes analysis procedure, which is a process for removing non-shape variation (i.e. rotation, translation and scale) from the data (Rohlf, Reference Rohlf1999).
As wing size can also significantly influence the shape of the wing (allometry), we undertook a multivariate regression of the dependant variable (wing shape) on centroid size (independent variable) (e.g. Drake & Klingenberg, Reference Drake and Klingenberg2008) using the software package MorphoJ v 1.02E (Klingenberg, Reference Klingenberg2011). This approach uses the shape of each specimen as represented by a vector of landmark coordinates following Procrustes superimposition and then relates this to size. The statistical significance of the regression was tested by permutation tests (10,000 replicates) against the null hypothesis of independence.
Procrustes transformed co-ordinate data was then imported into the statistical analysis package SPSS v 17.0 for subsequent canonical variates analysis (CVA), and also the computer program CVAGen6j (Sheets, Reference Sheets2006) for group assignment tests. The latter test reassigns individuals to an a priori defined group following CVA (based on Mahalanobis distances between the individual and group centroids).
For CVA, individuals were a priori defined as members of one of 13 samples based on collection locality (i.e. three sites for each of the four in-group species, one site for the out-group B. tryoni). Following CVA, intraspecific groups (except for TN Malaysia B. carambolae, see Discussion) were combined according to species to form five groups, i.e. B. dorsalis s.s., B. papayae, B. carambolae, B. philippinensis and B. tryoni, for which the assignment test was undertaken to determine the number of individuals from any one sample being reassigned to their a priori taxonomically defined species.
Results
Wing size
Intraspecific wing size variation among sampled populations was significantly different for B. dorsalis s.s. (F(2, 56)=4.876, P<0.05) and B. carambolae (F(2, 52)=14.628, P<0.001) but not so for either B. papayae (F(2, 54)=0.559, P>0.05) or B. philippinensis (F(2, 51)=1.610, P>0.05) (fig. 2). The Tukey post hoc test for B. dorsalis s.s. revealed that the Thailand sample was significantly smaller than either of those from Taiwan or India, with the latter two samples not significantly different from each other. The sample of B. carambolae from ‘TN Malaysia’ was significantly larger than the B. carambolae samples from Kuala Kangsar and Suriname, which again were not significantly different from each other. The out-group B. tryoni possessed the smallest wings (fig. 2).
Fig. 2. Wing centroid size (mean±1 SE) for males from each sampled location of Bactrocera dorsalis s.l. and the out-group B. tryoni (TRY). DOR, B. dorsalis s.s.; CAR, B. carambolae; PAP, B. papayae; PHI, B. philippinensis. See table 1 for locality details. Shading was added to delineate different species. Samples sharing the same letter within a species are not statistically different from each other based on intra-specific one-way ANOVA with Tukey post hoc test (P>0.05).
For the combined samples of each in-group plus the out-group, wing size varied significantly across species (F(4, 222)=9.239 P<0.05). The Tukey post hoc test showed, however, that no species, including the out-group B. tryoni, was able to be discriminated from at least one other species (table 2). It is thus concluded that wing size is not a good discriminator of these taxa.
Table 2. Wing centroid size (mean±SE) for each of the five Bactrocera species used in this study.
Species which are not significantly different (P>0.05; based on the Tukey post hoc test following one-way ANOVA) occur within the same column. N, number of individuals.
Wing shape
Generalised Procrustes superimposition produced a new set of co-ordinate data for each of the individuals used in the study. Multiple regression of wing shape on centroid size revealed a significant association (P<0.0001), which is perhaps not unexpected considering the degree of variation in wing sizes among the specimens sampled (see above). However, despite the statistically significant association observed, wing size was predicted to account for only 5.8% of total shape variation. Therefore, despite the significance of the association, its influence is considered relatively minor.
Canonical variates analysis on the Procrustes transformed data-set subsequently resulted in ten significant (α=0.05) canonical variates (CVs) (table 3), of which the first three explained 71.1% of the variation (CV1 Wilks’ λ=0.001, χ2=1552.726, P<0.001; CV2 Wilks’ λ=0.005, χ2=1207.279, P<0.001; CV3 Wilks’ λ=0.019, χ2=882.518, P<0.001). For each of the ten canonical variates, functions at group centroids were also generated (table 4), the first three of which are graphically represented for each of the 13 sample groups (fig. 3). Based on the graphical presentation of the first three canonical variates (i.e. fig. 3), samples cluster into species groups with the exception of B. carambolae ‘TN Malaysia’, which falls distant from the remaining B. carambolae samples.
Fig. 3. Plot of function centroids of the first three canonical variates produced from CVA on Procrustes coordinate data based on 15 wing landmarks from all individuals sampled for B. philippinensis, B. papayae, B. dorsalis s.s., B. carambolae and B. tryoni.
Table 3. Details of the ten significant canonical variates produced following CVA on 15 Procrustes transformed landmark data for all B. dorsalis s.l. and B. tryoni specimens used in the current study.
Table 4. Functions for group centroids for all significant canonical variates produced from CVA on 15 Procrustes transformed landmark data for all B. dorsalis s.l. and B. tryoni specimens used in the current study.
Success rate for the reassignment of individuals from any one sample to their a priori defined species ranged from 81% (B. papayae) to 100% for B. tryoni (table 5). Note that the number of B. carambolae used for reassignment was reduced to 36 due to the exclusion of B. carambolae samples from ‘TN Malaysia’. This sample group was removed from analysis after the initial canonical variates as post hoc morphological examination of the specimens by this paper's authors and R.A.I. Drew (one of the authors of the species B. carambolae) confirmed that this group of specimens likely represents a new species other than B. carambolae, therefore indicating a misclassification when the samples were originally identified.
Table 5. Reassignment of Bactrocera individuals to their a priori defined species, based on Mahalanobis distances to group centroids following canonical variates analysis on Procrustes transformed wing landmark data.
Rows, a priori groupings; columns, number of individuals reassigned to a priori group. % correct calculated based on the number of individuals reassigned to their a priori species group. N, number of individuals.
Discussion
Variation in wing size
Wing centroid size for each of the B. dorsalis s.l. populations studied revealed that some species (i.e. B. dorsalis s.s. and B. carambolae) contained one group significantly different in size from the remaining sample groups (fig. 2). Importantly, wing centroid size does not effectively discriminate one species from the other as there is no consistent significant difference in wing size between the species studied here (table 2).
Body size in tephritids is considered to be heavily influenced by environmental variables such as temperature and larval food source (Hooper, Reference Hooper1978; Krainacker et al., Reference Krainacker, Carey and Vargas1987). As a result, it is not surprising to record variation among samples within species. Such variation in intraspecific sizes between populations may explain the difference in wing sizes observed between groups of B. dorsalis s.s. seen in this study, particularly as each of the groups have been collected from geographically isolated, and presumably environmentally different, locations (i.e. Thailand, Taiwan and India). In contrast, wing centroid size among sample groups of both B. papayae and B. philippinensis was not significantly different. The collection localities for each of these species were, relative to those of B. dorsalis s.s., geographically close, especially for the latter species (table 1). Each sample was, therefore, likely to have been exposed to similar environmental conditions, possibly resulting in adults of similar size.
The situation of the sample of B. carambolae from ‘TN Malaysia’ is different. This group of flies has wings significantly larger than either of the other two samples of B. carambolae studied (fig. 2), including the group from Kuala Kangsar, which is, similarly to ‘TN Malaysia’, located in Peninsular Malaysia. Consequently, unlike the situation of B. dorsalis s.s., for which each sample group was collected from geographically isolated locations, it is less likely that such variation in wing size between two sample groups from Peninsular Malaysia can be explained due to different environmental conditions experienced during immature development. As stated previously, the difference in size between the ‘TN Malaysia’ sample and the remaining B. carambolae is likely due to this group of specimens representing a different species to B. carambolae, as determined following post hoc morphological examination (R.A.I. Drew, Griffith University, personal communications). This conclusion is further supported by our results based on wing shape data (fig. 3). The reason for the original misidentification remains unclear; however, individuals from this sample superficially resemble B. carambolae, particularly with regard to the expanded costal band on the wing. However, the ‘TN Malaysia’ specimens possess a costal band that is much broader and extends well beyond vein R2+3 and almost to vein R4+5 (not characteristic of B. carambolae for which the band only slightly overlaps vein R2+3 (Drew & Hancock, Reference Drew and Hancock1994)).
We conclude that while wing centroid size may reveal differences between populations of any given species (or indeed reveal a misdiagnosed population), it lacks utility as a diagnostic tool for effectively discriminating between the four morphologically cryptic species studied here.
Variation in wing shape
While wing size data may have limited utility in separating the four pest species of the complex, wing shape data appears to have potential for discriminating between B. dorsalis s.s., B. papayae, B. carambolae and B. philippinensis. This is demonstrated by the close proximity of group function centroids between conspecific populations (fig. 3) and the relatively high success rate for the correct reassignment of individuals to their a priori species (table 5). Importantly, this intraspecific conservation in wing-shape remains despite large geographic distances between some of the populations sampled, e.g. B. dorsalis s.s. is represented by samples from Taiwan, northern Thailand and India (a distance of over 4500 km between the extremes). Additionally, the high similarity in wing shape between the Malaysian and South American populations of B. carambolae (fig. 3 and table 5) further supports a view that wing shape is highly conserved within species regardless of the geographic proximity of regional populations. An important aspect to bear in mind is, however, the allometric effect of wing size on shape. As our results reveal the relatively minor overall influence of wing size on shape (5.8%), this association is statistically significant and should be considered in future studies.
While emerging as a potentially effective species level diagnostic tool, shape analysis also reinforces the very close morphological similarity between B. papayae to both B. dorsalis and B. carambolae. Of the 11 B. papayae individuals assigned to another species, six were reassigned as B. dorsalis and three to B. carambolae (the remaining two specimens placed with B. philippinensis) (see table 5). Such close affinity among these species with respect to shape data may reflect the close biological relationship between these taxa which may support them being considered conspecific; this is impossible to confirm, however, without further supportive biological information. Nevertheless, observations have been made previously regarding various levels of successful interspecific mating among these species under field-cage conditions (McInnis et al., Reference McInnis, Rendan, Jang, Sauers-Muller, Sugayama and Malavasi1999); and, therefore, their biological species limits remain unresolved.
Mechanisms driving wing shape variation for B. dorsalis species are unknown. Wing shape may play a functional role in the mating systems of B. dorsalis complex flies, as it does for species of the neotropical tephritid genus Blepharoneura, for which wing shape may influence audible signals produced during courtship (Marsteller et al., Reference Marsteller, Adams, Collyer and Condon2009). Further studies into the mating systems of B. dorsalis complex flies are required to determine if wing shape plays a role in producing different sounds among species and, more importantly, if sound is used as a key component in their specific mate recognition systems (sensu Paterson, Reference Paterson and Vrba1985).
The use of wing shape information is not, in isolation, an argument to confirm or refute species limits but rather one line of evidence to be used with other data sets (e.g. mating data and genetic studies). Shape variation has been documented for intraspecific systems, such as for Glossina palpalis gambiensis Vanderplank (Diptera: Glossinidae) populations occurring over geographical gradients in Africa (Bouyer et al., Reference Bouyer, Ravel, Dujardin, De Meeüs, Vial, Thévenon, Guerrini, Sidibé and Solano2007), conspecific strains of B. tryoni experiencing different environmental conditions during development (e.g. laboratory versus wild) (Gilchrist & Crisafulli, Reference Gilchrist and Crisafulli2006) and for other cryptic complexes within the Tephritidae (Kitthawee & Dujardin, Reference Kitthawee and Dujardin2010). Therefore, fine-scale shape variation between populations of B. dorsalis s.l. may simply be evidence of divergences occurring during the early stages of evolutionary radiation, at a stage which is concomitant with inter-population, rather than interspecific, differences. Such differences may be unresolvable using techniques such as molecular analysis, hence the current lack of molecular markers to discriminate between these species (with the exception of B. carambolae). Indeed, the high degree of intraspecific and overlapping morphological variation among the species studied here, combined with pheromone and cross-mating data suggesting hybridisation between some of these species (McInnis et al., Reference McInnis, Rendan, Jang, Sauers-Muller, Sugayama and Malavasi1999; Wee & Tan, Reference Wee and Tan2005), further emphasises the possibility that these taxa represent species in the early stages of divergence rather than distinct and reproductively isolated biological species.
In conclusion, geometric morphometric analyses demonstrate a capacity to resolve fine-scale differences in the B. dorsalis complex. Shape analysis using these techniques provides a promising opportunity to quickly and relatively easily identify morphologically cryptic pest species of the complex and has the potential for use as a rapid identification tool given a broader, more comprehensive dataset. More importantly, however, further behavioural and ecological research is required to confirm or refute whether the currently defined taxa studied here represent sound biological species, following which more reliable identification tools can be developed.
Acknowledgements
The authors conducted this review through CRC for National Plant Biosecurity project 20115 and would like to acknowledge the support of the Australian Government's Cooperative Research Centres Program. We thank the Department of Employment, Economic Development and Innovation, Longpocket, Brisbane, Australia for the loan of collection material and Alies van Sauers-Muller for supplying B. carambolae specimens from Paramaribo, Suriname. Thanks also to Karen Armstrong, Stephen Cameron and Ania Deutscher for helpful comments on the manuscript. Professor R.A.I. Drew, Griffith University, kindly shared his expertise on the B. dorsalis complex and helped confirm the misidentification of the ‘odd’ B. carambolae from ‘TN Malaysia’.
