G Model ACTROP-3330; No. of Pages 7
ARTICLE IN PRESS Acta Tropica xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Acta Tropica journal homepage: www.elsevier.com/locate/actatropica
Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens M.J.R. Hall a,∗ , N. MacLeod b , A.H. Wardhana a,c a
Life Sciences Department, Natural History Museum, London SW7 5BD, UK Earth Sciences Department, Natural History Museum, London SW7 5BD, UK c Parasitology Department, Indonesian Research Centre for Veterinary Science (Balai Besar Penelitian Veteriner), Bogor, Indonesia b
a r t i c l e
i n f o
Article history: Received 16 December 2013 Received in revised form 26 March 2014 Accepted 29 March 2014 Available online xxx Keywords: Chrysomya bezziana Old World screwworm fly Traumatic myiasis Wing morphometrics Museum collections
a b s t r a c t The Old World screwworm (OWS) fly, Chrysomya bezziana (Diptera: Calliphoridae), is a major economic and welfare problem for humans and animals in the Old World tropics. Using a bootstrapped log likelihood ratio test of the output of Procrustes principal components and canonical variates analyses for a small sample of museum specimens from which 19 2D wing landmarks had been collected: (1) a consistent and statistically significant difference exists between landmark configurations derived from wings of pinned specimens and those removed from the body and mounted on slides; (2) a highly statistically significant sexual dimorphism in wing morphometry was identified; and (3) a highly statistically significant difference in wing morphometry between populations of the OWS fly from Africa (Tanzania, South Africa Sudan, Zaire, Zimbabwe,) and Asia (Sumba, Indonesia) exists. These results show that wing orientation and gender must be considered when conducting morphometric investigations of OWS fly wings. The latter result is also consistent with results from previous molecular and morphological studies, which indicate there are two distinct genetic lineages within this species. Wing morphometry holds great promise as a practical tool to aid in identification of the geographical origin of introductions of this important pest species, by providing diagnostic markers to distinguish geographical populations and complement molecular diagnostics. © 2014 M.J.R. Hall. Published by Elsevier B.V. All rights reserved.
1. Introduction The Old World screwworm (OWS) fly, Chrysomya bezziana (Diptera: Calliphoridae), is a major livestock pest in the Old World tropics. Its larvae are obligate parasites of mammals causing traumatic cutaneous myiasis, a human, animal welfare and economic problem (Spradbery, 1994). Based on studies of mitochondrial and nuclear DNA, this species has been shown to comprise two main genetic-geographic lineages, one in sub-Saharan Africa and the other ranging across Asia, from the Middle East Gulf region to Indonesia (Hall et al., 2001; Wardhana et al., 2012a). In addition to this genetic evidence, analysis of a suite of morphological characters supports recognition of these lineages (Hall et al., 2001; Wardhana et al., 2012b). The morphological work followed on from earlier, qualitative indications of morphologically distinct
∗ Corresponding author. Tel.: +44 0 20 7942 5715; fax: +44 0 20 7942 5054. E-mail address: [email protected] (M.J.R. Hall).
geographical populations of the OWS fly (Colless in Spradbery, 1991). Morphometric studies have become increasingly popular in studies of medically important insects (Dujardin, 2011). To date, no use has been made of wing morphometrics to compare geographical populations of the OWS fly. However, following early work by Brown (1979), other studies have demonstrated the value of wing morphometrics for interspecific comparison of species of Calliphoridae; for example, to distinguish Cochliomyia hominivorax (New World screwworm fly) from Cochliomyia macellaria (Lyra et al., 2009), and to distinguish Chrysomya albiceps from Chrysomya megacephala (Vásquez and Liria, 2012). Studies of other medically important insects have also shown that wing morphometrics has the potential to identify intraspecific differences between geographically separated populations (e.g. Solano et al., 1999; Bouyer et al., 2007; Gómez-Palacio et al., 2012). It is our future objective to investigate intraspecific differences between OWS fly populations, to determine whether wing morphometrics can be a complementary or surrogate tool to molecular markers of populations and, thereby, have value in determining
http://dx.doi.org/10.1016/j.actatropica.2014.03.023 0001-706X/© 2014 M.J.R. Hall. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
2
the origin of specimens in recent introductions (Ready et al., 2009). However, before initiating such studies it is necessary to address, briefly, more basic questions. The OWS fly has long been recognised to occur in relatively low densities and is difficult to trap as adults in large numbers (Spradbery, 1994; Ulrech et al., 2012). Therefore, in order to obtain sufficient specimens for morphometric analysis it can be of considerable value to use museum collections of adult OWS flies to provide a source of morphological characters (e.g. Hall et al., 2001). Ideally, study of museum specimens should be nondestructive and measurement of the wings made while they are attached to the specimen. However, the wings of preserved and/or pinned specimens can be difficult to orient appropriately for imaging and are susceptible to deformation. As a result, measurements made on attached wings might not be comparable to measurements obtained from wings that have been removed and flattened mechanically after mounting on slides. To resolve this issue the preliminary study reported here had three objectives: (1) to determine whether there is a consistent and structured difference between the morphometric configurations of landmarks collected from wings on flies and those collected from flattened, slide mounted wings; (2) to determine if there was sexual dimorphism in wing morphology that would require a separate analysis of the sexes; and (3) to undertake a preliminary morphometric comparison of wings from populations of flies from the two OWS fly geographical lineages, Africa and Asia. 2. Materials and methods 2.1. Selection of specimens and capture of images The specimens selected for study of wings both on and removed from flies were 22 pinned adult African females from the collections of the Natal Museum and the National Collection of Insects, Pretoria, both in South Africa, and from The Natural History Museum, London. These included specimens almost 100 years old and of historical significance (Fig. 1), from Sudan (x1), Tanzania (x1), Zaire (x1), Zimbabwe (x16) and South Africa (x3). The sample sizes are small, but reflect the low numbers of adults of this pest species found typically in museum collections due to their rarity in trap catches (Spradbery, 1994; Ulrech et al., 2012). Only females were selected to avoid potential problems of sex-linked variation (Lyra et al., 2009). Identification to species was based on the diagnostic morphological characters described by Hall (2008). All specimens were captured as adults or were reared after collection as mature larvae from wounds. None was obtained from laboratory cultures. Specimens on pins were oriented by eye under a Leica M165C binocular microscope so that the majority of the wing was in the horizontal plane. Images were captured with a Leica DFC295 camera and Leica Application Suite version 3.5.0 software. After
Fig. 1. Adult female of C. bezziana reared from a larva collected from an ear wound on a donkey, Zanzibar, Tanzania, 25 November 1915. Wings are displayed in an apparently good condition for imaging prior to morphometric analysis.
capture of images from these pinned specimens, they were relaxed for 24 h in a closed Perspex box with damp tissues to make them less brittle during handling. The wings were removed near to their base, retaining the basicosta on the wing, using a combination of fine forceps and a fine scalpel blade. These amputated wings were then placed under a cover slip on a glass slide in Euparal mountant. This was done by firstly placing a drop of Euparal onto the slide, then placing the wing onto the drop, adding a thin layer of Euparal to the wing and then placing the cover slip on top. The mountant layer was kept as thin as possible to maximise wing flattening. All mounted wing slides were placed into an oven (56 ◦ C) overnight to clear air bubbles before imaging with the same system as used for attached wings. To compare the African specimens with a population from the Asian geographical lineage of the OWS fly, 13 adult females were sampled from a culture established at the Parasitology Department, Indonesian Research Centre for Veterinary Science, Bogor, Indonesia after collection on the island of Sumba. Wings from these specimens were removed and mounted on slides for imaging using the same procedure described above. A comparison of sex differences was made between these same Sumba females and 13 Sumba males from the same culture. 2.2. Wing morphometric character coding On each wing, the positions of 19 landmarks (Fig. 2) were recorded in the same order after importing the digital images
Fig. 2. Wing of C. bezziana showing the 19 landmarks used in the morphometric analysis.
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
3
into tpsDig1 software (http://life.bio.sunysb.edu/morph/). This set of landmark coordinate points were located at the base of the wing, the intersections of wing veins with each other, and the intersection of wing veins with the wing margin. These included the 15 landmarks of Lyra et al. (2009) plus four additional landmarks, mainly near to the wing base. The scaled x,y coordinates of each landmark were then concatenated into three datafiles reflecting the comparisons described above and analysed in Wolfram MathematicaTM notebook routines (available from NM on request: [email protected]) that implement the procedures discussed in MacLeod (2009, 2010) for analysis.
2.3. Statistical analysis Analysis of these data involved: (1) determination of centroid size values using all landmark data (Bookstein, 1986; MacLeod, 2008); (2) projection of all geometric data into a Procrustes shape space (Rohlf and Slice, 1990; MacLeod, 2009); (3) summarization of the major directions of shape variation via Procrustes principal component analysis (PCA); and (4) analysis of the set of PCA scores (=projections of configuration positions onto the PC axes) necessary to represent at least 95 percent of the observed shape information via canonical variates analysis (CVA) (Davis, 2002; MacLeod, 2005, 2007) to achieve maximum separations of group centroids relative to within-group dispersions. The statistical significance of the between-groups was assessed by calculating the log-likelihood ratio index (see Manly, 2004) for the projected positions of individuals in these two a priori-defined datasets and comparing that to a bootstrap-generated distribution of log-likelihood ratio index generated for 1000 pseudoreplicate datasets, each of which was composed of two a priori-defined datasets of identical size to the study dataset but whose members were drawn randomly from the pooled investigation dataset. Accordingly, the CVA of each pseudoreplicate dataset, and subsequent calculation of the between-groups log-likelihood ratio index, provides an indication of the range and frequency of values this index would be expected to assume if it were used to characterise differences between samples drawn from a single population whose variable variances and covariance structure are identical to those of the investigation dataset.
3. Results 3.1. Comparison of wings on flies with those removed and slide-mounted The projection of the landmark configuration geometries for the ‘on fly’ and ‘on slide’ datasets into the subspace formed by the first two PC axes is shown in Fig. 3A. Together these two axes summarize 43 percent of the observed shape variation and individually they represent the two most consistent, but mutually independent, trends in shape variation present in these data. Projected configuration positions that lie close to one another in this space indicate shapes that are broadly similar, while those located at some remove from one another indicate shapes that are broadly different. While the shape fields defined by these groups do exhibit different tendencies overall, it is clear that there are broad areas of overlap between the two groups within this space. By itself this overlap does not indicate how distinct the shapes of the two groups are, as Fig. 3A only shows two aspects of 38 dimensional space. Nevertheless it does indicate that, for this dataset, between-groups shape distinctions are not contributing a large component of variation to the structure of shape variation within the dataset as a whole. Despite this level of group overlap, there is a surprisingly
Fig. 3. Results of the comparison of wing landmark geometries for ‘on fly’ (grey points/bars, n = 22) and ‘on slide’ (white points/bars, n = 22) wing samples of C. bezziana. (A) Projections of Procrustes-aligned landmark configurations on the first two principal components (=relative warps) of the shape covariance matrix. (B) Projections of pooled sample PCA scores for the first 14 principal components (=the number needed to represent 95 percent of the observed shape variation) onto the discriminant axis that best separates group centroids. (C) Results of a robust statistical test of the between-groups distinctions shown in (B). The histogram shows the empirical distribution of 1000 CVA analyses of bootstrapped pseudoreplicate datasets of the pooled groups PCA scores formulated according to the procedure described in Section 2.
large amount of between-groups separation evident by the separation of group centroids on this plot. A comprehensive view of the true level of shape distinction existing between these groups requires further analysis, specifically the CVA of PCA scores on 14 PCA axes required to represent at least 95 percent of the shape variation observed in this sample. As there are only two groups present in the sample, only a single CVA axis is required to achieve maximal separation between group centroids (Fig. 3B). In this plot it can be seen that, with the exception of three ‘on fly’ specimens that fall into the distribution of ‘on slide’ landmark configurations, the two groups are wholly separable. As these data have been derived from the same wings, the only reasonable conclusion that can be drawn from these data is that inconsistencies in orientation and deformation of the wings
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7 4
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
Fig. 4. Examples of wings of African (left; Zimbabwe) and Asian (right; Sumba, Indonesia) lineages of C. bezziana (female specimens).
measured when they are attached to the fly bodies is introducing a substantial, consistent, and well structured source of measurement error into the assessment of wing geometry, at least as this is represented by the location of wing vein landmarks. While the graphical distinctions between the ‘on fly’ and ‘on slide’ datasets are clear from Fig. 3B, the statistical significance of these differences needs to be assessed quantitatively. Mitteroecker and Bookstein (2011) have questioned the use of CVA in morphometric contexts as it will always try to force groups apart and in some cases (e.g., for underdetermined data matrices) apparently profound differences can be manufactured by the method even for random data. However, resampling methods (e.g., bootstrapping, jackknifing, see Manly, 1997) can be used to correct the misperception of group difference via the construction of valid probability distributions for CVA-related statistical indices derived from various null models against which the observed values of these indices can be compared. A 1000 pseudoreplicate bootstrapped test of the betweengroups distinction for these data as summarized by the log-likelihood ratio () index produces a frequency distribution for the expected values of the index calculated from CVA results under the null hypothesis of no difference between groups other than random sampling differences (Fig. 3C). This distribution is analogous to the 2 distribution for this statistic which is the more standard manner of assessing its significance for the analysis of raw, measured variables. Since the observed value of the index for these data is 42.25, reference of this value to the bootstrapped distribution of hypothetical values indicates that there is less than a 0.1 percent probability of such a large separation between group centroids arising as a result of a CVA analysis due to random sampling differences. Accordingly, we conclude that for this, albeit small, dataset the group differences observed in Fig. 3B are significant statistically. 3.2. Comparison of wings from two geographical populations The wings of African specimens had a more strongly infuscated base than those of Asian specimens (Fig. 4). Using the PCA-CVAlog-likelihood ratio test as above, the comparison of disarticulated, slide mounted wings of females demonstrated that there was no overlap between the PCA scores for these two geographical regions along PC-1 (Fig. 5A). This result indicates that, for this dataset, between-group wing-vein shape differentiation is the dominant factor controlling shape variation in the character complex. When the information included in all 13 PCA axes required to represent 95 percent of the observed shape variation was taken into consideration (Fig. 5B) the large and consistent degree of wing vein shape differentiation is evident. In terms of statistical significance, the highest index value of the randomized LLR index obtained from
1000 bootstrapped pseudoreplicate CVA analyses (Fig. 5C) was
ARTICLE IN PRESS Acta Tropica xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Acta Tropica journal homepage: www.elsevier.com/locate/actatropica
Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens M.J.R. Hall a,∗ , N. MacLeod b , A.H. Wardhana a,c a
Life Sciences Department, Natural History Museum, London SW7 5BD, UK Earth Sciences Department, Natural History Museum, London SW7 5BD, UK c Parasitology Department, Indonesian Research Centre for Veterinary Science (Balai Besar Penelitian Veteriner), Bogor, Indonesia b
a r t i c l e
i n f o
Article history: Received 16 December 2013 Received in revised form 26 March 2014 Accepted 29 March 2014 Available online xxx Keywords: Chrysomya bezziana Old World screwworm fly Traumatic myiasis Wing morphometrics Museum collections
a b s t r a c t The Old World screwworm (OWS) fly, Chrysomya bezziana (Diptera: Calliphoridae), is a major economic and welfare problem for humans and animals in the Old World tropics. Using a bootstrapped log likelihood ratio test of the output of Procrustes principal components and canonical variates analyses for a small sample of museum specimens from which 19 2D wing landmarks had been collected: (1) a consistent and statistically significant difference exists between landmark configurations derived from wings of pinned specimens and those removed from the body and mounted on slides; (2) a highly statistically significant sexual dimorphism in wing morphometry was identified; and (3) a highly statistically significant difference in wing morphometry between populations of the OWS fly from Africa (Tanzania, South Africa Sudan, Zaire, Zimbabwe,) and Asia (Sumba, Indonesia) exists. These results show that wing orientation and gender must be considered when conducting morphometric investigations of OWS fly wings. The latter result is also consistent with results from previous molecular and morphological studies, which indicate there are two distinct genetic lineages within this species. Wing morphometry holds great promise as a practical tool to aid in identification of the geographical origin of introductions of this important pest species, by providing diagnostic markers to distinguish geographical populations and complement molecular diagnostics. © 2014 M.J.R. Hall. Published by Elsevier B.V. All rights reserved.
1. Introduction The Old World screwworm (OWS) fly, Chrysomya bezziana (Diptera: Calliphoridae), is a major livestock pest in the Old World tropics. Its larvae are obligate parasites of mammals causing traumatic cutaneous myiasis, a human, animal welfare and economic problem (Spradbery, 1994). Based on studies of mitochondrial and nuclear DNA, this species has been shown to comprise two main genetic-geographic lineages, one in sub-Saharan Africa and the other ranging across Asia, from the Middle East Gulf region to Indonesia (Hall et al., 2001; Wardhana et al., 2012a). In addition to this genetic evidence, analysis of a suite of morphological characters supports recognition of these lineages (Hall et al., 2001; Wardhana et al., 2012b). The morphological work followed on from earlier, qualitative indications of morphologically distinct
∗ Corresponding author. Tel.: +44 0 20 7942 5715; fax: +44 0 20 7942 5054. E-mail address: [email protected] (M.J.R. Hall).
geographical populations of the OWS fly (Colless in Spradbery, 1991). Morphometric studies have become increasingly popular in studies of medically important insects (Dujardin, 2011). To date, no use has been made of wing morphometrics to compare geographical populations of the OWS fly. However, following early work by Brown (1979), other studies have demonstrated the value of wing morphometrics for interspecific comparison of species of Calliphoridae; for example, to distinguish Cochliomyia hominivorax (New World screwworm fly) from Cochliomyia macellaria (Lyra et al., 2009), and to distinguish Chrysomya albiceps from Chrysomya megacephala (Vásquez and Liria, 2012). Studies of other medically important insects have also shown that wing morphometrics has the potential to identify intraspecific differences between geographically separated populations (e.g. Solano et al., 1999; Bouyer et al., 2007; Gómez-Palacio et al., 2012). It is our future objective to investigate intraspecific differences between OWS fly populations, to determine whether wing morphometrics can be a complementary or surrogate tool to molecular markers of populations and, thereby, have value in determining
http://dx.doi.org/10.1016/j.actatropica.2014.03.023 0001-706X/© 2014 M.J.R. Hall. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
2
the origin of specimens in recent introductions (Ready et al., 2009). However, before initiating such studies it is necessary to address, briefly, more basic questions. The OWS fly has long been recognised to occur in relatively low densities and is difficult to trap as adults in large numbers (Spradbery, 1994; Ulrech et al., 2012). Therefore, in order to obtain sufficient specimens for morphometric analysis it can be of considerable value to use museum collections of adult OWS flies to provide a source of morphological characters (e.g. Hall et al., 2001). Ideally, study of museum specimens should be nondestructive and measurement of the wings made while they are attached to the specimen. However, the wings of preserved and/or pinned specimens can be difficult to orient appropriately for imaging and are susceptible to deformation. As a result, measurements made on attached wings might not be comparable to measurements obtained from wings that have been removed and flattened mechanically after mounting on slides. To resolve this issue the preliminary study reported here had three objectives: (1) to determine whether there is a consistent and structured difference between the morphometric configurations of landmarks collected from wings on flies and those collected from flattened, slide mounted wings; (2) to determine if there was sexual dimorphism in wing morphology that would require a separate analysis of the sexes; and (3) to undertake a preliminary morphometric comparison of wings from populations of flies from the two OWS fly geographical lineages, Africa and Asia. 2. Materials and methods 2.1. Selection of specimens and capture of images The specimens selected for study of wings both on and removed from flies were 22 pinned adult African females from the collections of the Natal Museum and the National Collection of Insects, Pretoria, both in South Africa, and from The Natural History Museum, London. These included specimens almost 100 years old and of historical significance (Fig. 1), from Sudan (x1), Tanzania (x1), Zaire (x1), Zimbabwe (x16) and South Africa (x3). The sample sizes are small, but reflect the low numbers of adults of this pest species found typically in museum collections due to their rarity in trap catches (Spradbery, 1994; Ulrech et al., 2012). Only females were selected to avoid potential problems of sex-linked variation (Lyra et al., 2009). Identification to species was based on the diagnostic morphological characters described by Hall (2008). All specimens were captured as adults or were reared after collection as mature larvae from wounds. None was obtained from laboratory cultures. Specimens on pins were oriented by eye under a Leica M165C binocular microscope so that the majority of the wing was in the horizontal plane. Images were captured with a Leica DFC295 camera and Leica Application Suite version 3.5.0 software. After
Fig. 1. Adult female of C. bezziana reared from a larva collected from an ear wound on a donkey, Zanzibar, Tanzania, 25 November 1915. Wings are displayed in an apparently good condition for imaging prior to morphometric analysis.
capture of images from these pinned specimens, they were relaxed for 24 h in a closed Perspex box with damp tissues to make them less brittle during handling. The wings were removed near to their base, retaining the basicosta on the wing, using a combination of fine forceps and a fine scalpel blade. These amputated wings were then placed under a cover slip on a glass slide in Euparal mountant. This was done by firstly placing a drop of Euparal onto the slide, then placing the wing onto the drop, adding a thin layer of Euparal to the wing and then placing the cover slip on top. The mountant layer was kept as thin as possible to maximise wing flattening. All mounted wing slides were placed into an oven (56 ◦ C) overnight to clear air bubbles before imaging with the same system as used for attached wings. To compare the African specimens with a population from the Asian geographical lineage of the OWS fly, 13 adult females were sampled from a culture established at the Parasitology Department, Indonesian Research Centre for Veterinary Science, Bogor, Indonesia after collection on the island of Sumba. Wings from these specimens were removed and mounted on slides for imaging using the same procedure described above. A comparison of sex differences was made between these same Sumba females and 13 Sumba males from the same culture. 2.2. Wing morphometric character coding On each wing, the positions of 19 landmarks (Fig. 2) were recorded in the same order after importing the digital images
Fig. 2. Wing of C. bezziana showing the 19 landmarks used in the morphometric analysis.
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
3
into tpsDig1 software (http://life.bio.sunysb.edu/morph/). This set of landmark coordinate points were located at the base of the wing, the intersections of wing veins with each other, and the intersection of wing veins with the wing margin. These included the 15 landmarks of Lyra et al. (2009) plus four additional landmarks, mainly near to the wing base. The scaled x,y coordinates of each landmark were then concatenated into three datafiles reflecting the comparisons described above and analysed in Wolfram MathematicaTM notebook routines (available from NM on request: [email protected]) that implement the procedures discussed in MacLeod (2009, 2010) for analysis.
2.3. Statistical analysis Analysis of these data involved: (1) determination of centroid size values using all landmark data (Bookstein, 1986; MacLeod, 2008); (2) projection of all geometric data into a Procrustes shape space (Rohlf and Slice, 1990; MacLeod, 2009); (3) summarization of the major directions of shape variation via Procrustes principal component analysis (PCA); and (4) analysis of the set of PCA scores (=projections of configuration positions onto the PC axes) necessary to represent at least 95 percent of the observed shape information via canonical variates analysis (CVA) (Davis, 2002; MacLeod, 2005, 2007) to achieve maximum separations of group centroids relative to within-group dispersions. The statistical significance of the between-groups was assessed by calculating the log-likelihood ratio index (see Manly, 2004) for the projected positions of individuals in these two a priori-defined datasets and comparing that to a bootstrap-generated distribution of log-likelihood ratio index generated for 1000 pseudoreplicate datasets, each of which was composed of two a priori-defined datasets of identical size to the study dataset but whose members were drawn randomly from the pooled investigation dataset. Accordingly, the CVA of each pseudoreplicate dataset, and subsequent calculation of the between-groups log-likelihood ratio index, provides an indication of the range and frequency of values this index would be expected to assume if it were used to characterise differences between samples drawn from a single population whose variable variances and covariance structure are identical to those of the investigation dataset.
3. Results 3.1. Comparison of wings on flies with those removed and slide-mounted The projection of the landmark configuration geometries for the ‘on fly’ and ‘on slide’ datasets into the subspace formed by the first two PC axes is shown in Fig. 3A. Together these two axes summarize 43 percent of the observed shape variation and individually they represent the two most consistent, but mutually independent, trends in shape variation present in these data. Projected configuration positions that lie close to one another in this space indicate shapes that are broadly similar, while those located at some remove from one another indicate shapes that are broadly different. While the shape fields defined by these groups do exhibit different tendencies overall, it is clear that there are broad areas of overlap between the two groups within this space. By itself this overlap does not indicate how distinct the shapes of the two groups are, as Fig. 3A only shows two aspects of 38 dimensional space. Nevertheless it does indicate that, for this dataset, between-groups shape distinctions are not contributing a large component of variation to the structure of shape variation within the dataset as a whole. Despite this level of group overlap, there is a surprisingly
Fig. 3. Results of the comparison of wing landmark geometries for ‘on fly’ (grey points/bars, n = 22) and ‘on slide’ (white points/bars, n = 22) wing samples of C. bezziana. (A) Projections of Procrustes-aligned landmark configurations on the first two principal components (=relative warps) of the shape covariance matrix. (B) Projections of pooled sample PCA scores for the first 14 principal components (=the number needed to represent 95 percent of the observed shape variation) onto the discriminant axis that best separates group centroids. (C) Results of a robust statistical test of the between-groups distinctions shown in (B). The histogram shows the empirical distribution of 1000 CVA analyses of bootstrapped pseudoreplicate datasets of the pooled groups PCA scores formulated according to the procedure described in Section 2.
large amount of between-groups separation evident by the separation of group centroids on this plot. A comprehensive view of the true level of shape distinction existing between these groups requires further analysis, specifically the CVA of PCA scores on 14 PCA axes required to represent at least 95 percent of the shape variation observed in this sample. As there are only two groups present in the sample, only a single CVA axis is required to achieve maximal separation between group centroids (Fig. 3B). In this plot it can be seen that, with the exception of three ‘on fly’ specimens that fall into the distribution of ‘on slide’ landmark configurations, the two groups are wholly separable. As these data have been derived from the same wings, the only reasonable conclusion that can be drawn from these data is that inconsistencies in orientation and deformation of the wings
Please cite this article in press as: Hall, M.J.R., et al., Use of wing morphometrics to identify populations of the Old World screwworm fly, Chrysomya bezziana (Diptera: Calliphoridae): A preliminary study of the utility of museum specimens. Acta Trop. (2014), http://dx.doi.org/10.1016/j.actatropica.2014.03.023
G Model ACTROP-3330; No. of Pages 7 4
ARTICLE IN PRESS M.J.R. Hall et al. / Acta Tropica xxx (2014) xxx–xxx
Fig. 4. Examples of wings of African (left; Zimbabwe) and Asian (right; Sumba, Indonesia) lineages of C. bezziana (female specimens).
measured when they are attached to the fly bodies is introducing a substantial, consistent, and well structured source of measurement error into the assessment of wing geometry, at least as this is represented by the location of wing vein landmarks. While the graphical distinctions between the ‘on fly’ and ‘on slide’ datasets are clear from Fig. 3B, the statistical significance of these differences needs to be assessed quantitatively. Mitteroecker and Bookstein (2011) have questioned the use of CVA in morphometric contexts as it will always try to force groups apart and in some cases (e.g., for underdetermined data matrices) apparently profound differences can be manufactured by the method even for random data. However, resampling methods (e.g., bootstrapping, jackknifing, see Manly, 1997) can be used to correct the misperception of group difference via the construction of valid probability distributions for CVA-related statistical indices derived from various null models against which the observed values of these indices can be compared. A 1000 pseudoreplicate bootstrapped test of the betweengroups distinction for these data as summarized by the log-likelihood ratio () index produces a frequency distribution for the expected values of the index calculated from CVA results under the null hypothesis of no difference between groups other than random sampling differences (Fig. 3C). This distribution is analogous to the 2 distribution for this statistic which is the more standard manner of assessing its significance for the analysis of raw, measured variables. Since the observed value of the index for these data is 42.25, reference of this value to the bootstrapped distribution of hypothetical values indicates that there is less than a 0.1 percent probability of such a large separation between group centroids arising as a result of a CVA analysis due to random sampling differences. Accordingly, we conclude that for this, albeit small, dataset the group differences observed in Fig. 3B are significant statistically. 3.2. Comparison of wings from two geographical populations The wings of African specimens had a more strongly infuscated base than those of Asian specimens (Fig. 4). Using the PCA-CVAlog-likelihood ratio test as above, the comparison of disarticulated, slide mounted wings of females demonstrated that there was no overlap between the PCA scores for these two geographical regions along PC-1 (Fig. 5A). This result indicates that, for this dataset, between-group wing-vein shape differentiation is the dominant factor controlling shape variation in the character complex. When the information included in all 13 PCA axes required to represent 95 percent of the observed shape variation was taken into consideration (Fig. 5B) the large and consistent degree of wing vein shape differentiation is evident. In terms of statistical significance, the highest index value of the randomized LLR index obtained from
1000 bootstrapped pseudoreplicate CVA analyses (Fig. 5C) was
