Comparative Analysis of Molecular Identication and Wing Morphometrics of Forensically Important Blow Flies in China

Background: Correct species identication is the most crucial step in applying entomological evidence to estimate the postmortem interval (PMI) since death of decomposed corpses. Wing morphometrics have been proposed in species classication as an alternative method of traditional morphology and molecular approaches. However, so far, this method has not been applied to the identication of Chinese Calliphoridae and few studies compare the two identication methods. Methods: We used landmark-based geometric morphometrics of wings to identify nine medically and forensically important blow y species of China. 270 specimens representing nine species and eight genera were sampled, 18 landmarks on the right wing were measured and analyzed using canonical variates analysis and discriminant function analysis. Then, a cross-validation test was used to evaluate reliability of the method. Moreover, in order to further assess the validity of this method, molecular identication is used for comparative analysis. Eighty sequences of cytochrome c oxidase subunit I (COI) of Calliphoridae isolated from different countries were downloaded from Genbank, including the data previously submitted by our team. Results: Different species and genera can be well separated through morphometric analysis with an overall classication accuracy of 80~100%, but discrimination between sexes was less effective. The results indicated that the discriminative eciency of the two methods is almost identical. Conclusions: Wing morphometrics can be used as a complementary method of molecular identication for the geographical location and gender identication of certain species as a simple and cheap method. comparison of three all three dendrogram,

. Many studies have proved the robustness of COI as a marker for y species discrimination [10].
Nevertheless, there is no signi cant difference in the usage of short fragments or even the entire sequence of COI [11]. DNA barcoding is based on a standardized small segment fragment sequence of cytochrome oxidase subunit 1 (COI) called "DNA barcode" and is widely used in public databases, such as GenBank and Bold (life data barcode). Thus, DNA barcoding is also widely used to identify y species [12]. In addition, many studies have proved the effectiveness of COI barcodes for the identi cation of many species of Calliphoridae [13,14]. Meanwhile, in GenBank, there are abundant molecular data about Calliphora, sequences of forensic important Calliphora have been published and uploaded online from different regions of the world, such as South Africa, the Caribbean Region, Southern European, the United States and Korea [15,16,17,18], making it possible to systematically study the sequence of Calliphora around the world. However, the molecular identi cation is a destructive approach for sample materials.
Sequencing has relative higher cost and sequence analysis might be problematic in identi cation of closely related species [19,20]. Therefore, other effective alternative methods were needed.
Besides DNA identi cation, geometric morphometrics is a method by using a series of reference points or landmarks that make the morphological comparison of any object structure possible, removing irrelevant information such as the position and orientation of the specimens, considering only their shape [21]. It is a quantitative study of the size, shape and shape changes of species and their covariation with other biotic or abiotic factors [22]. As an alternative approach, the use of landmark-based geometric morphometric analysis of insect wings has shown to be a valuable tool in many studies following early work by Brown [23]. Wing morphometrics has been increasingly popular in studies of Diptera, demonstrating the species identi cation value in Syrphidae [24] and forensically important families like Muscidae [25] and Sarcophagidae [26]. It has been proved that wing morphometrics can be used for the identi cation of necrophagous Calliphoridae from Thailand [27] and Europe [28], and distinguish different genus and species of Calliphoridae, Cochliomyia [29], Lucilia [23,30] and Chrysomya [29,31]. In addition, geometric morphometric analysis can be used to distinguish variability between geographical populations [29,32], seasonal morphs [33] or sexes [26,34] among the same species, and recently it is used to identify Piophilidae (Diptera) for forensic purpose [35]. All of these shows that the analysis of wing morphometric is a very useful identi cation method.
Up to now, morphometric analysis has not been widely used in the species of Chinese Calliphoridae. In this study, we apply wing morphometric for the identi cation of necrophagous ies in China at genus and species levels, evaluating the effect of allometric growth on species identi cation. As female ies are more common and di cult to identify in the forensic eld, the differences between male and female wings of nine species of ies were also investigated. In addition, we compared gene sequences of Calliphora from different cities of China with those from different countries and evaluate the reliability of the two types of methods for the identi cation of forensically relevant species of Calliphoridae. Therefore, this study examined the potential of these tools for specimen identi cation and concluded that wing morphometrics could be a simple and user-friendly method for research and practical application of forensic entomology.

Specimen collection
For wing morphometrics analysis, materials for the present study were collected from different places of China. Adult ies, both males and females, were collected directly with a handheld y net or lured to slightly decomposed pig cadavers placed in a eld. All specimens were identi ed according to Gregor et al. and then preserved in ethanol. A total of 270 specimens representing 8 genera and 9 species were used in this study: Triceratopyga calliphoroides (Rohdendorf, 1931)  Molecular Analysis COI gene sequences are obtained from Genbank, and the length was set more than 650 bp to contain the region of COI barcode. Then, alignment and phylogenetic were performed in MEGA 10.0.5 (Center for Evolutionary Medicine and Informatics) [36] yielding a total data set of 80 COI sequences (Table S1).
Consensus sequences were aligned using ClustalW under default parameters. The neighbor-joining (NJ) tree construction and genetic divergences calculation were conducted using the uncorrected pairwise distance model in MEGA version 10.0.5 [36]. The constructed tree was generated with the Tamura 3parameter model, and 500 bootstrap replicates were used to assess the reliability of the tree [37]. Genetic distances were calculated by the Kimura 2-parameter (K2P) model in MEGA version 10.0.5 [36].

Wing Morphometrics
Shape variation of genera and species At the genera level, the CVA revealed seven canonical variates, and the rst two canonical variates ( Fig.   S1) explained 75% of the total variation (CV1 = 52, CV2 = 23). The scatter diagram from CV1 and CV2 ( Fig. S1) showed three clearly separated clusters. The rst cluster were specimens of genera Hemipyrellia, which was form a distinct group separated from other genera. The second cluster was genera Achoetandrus and Chrysomya. The third cluster included specimens of genera Lucilia, Calliphora, Aldrichina, Triceratopyga and Protophormia. The Mahalanobis distances acquired from the pairwise comparisons of all eight genera indicated highly signi cant differences (permutation test with 10,000 rounds in MorphoJ: P < 0.0001), ranging from 4.0299 (Calliphora and Aldrichina) to 29.9547 (Achoetandrus and Protophormia) ( Table S2). The percentages of correctly classi ed specimens obtained from the cross-validation test ranged from 80-100.0% (Table S3).

Shape Variation
Canonical variate analysis (CVA) was used to analyze the most important shape characteristics between groups (genera or species) using MorphoJ software version 1.06 [40]. The CVA was used to magnify interspeci c variation and minimize intraspeci c variation. Mahalanobis distances were computed from DFA (discriminant function) as a statistical measure to evaluate the distance between groups, assessing the similarity between different populations. Permutation tests (10,000 replications) with Mahalanobis distances and Procrustes distances were used to test the statistical signi cance of pairwise differences between species and sexes. A cross-validated classi cation in discriminant function analysis (DFA) was used to evaluate the accuracy of species identi cation [40].

Allometry
Before the allometric analysis, a Procrustes ANOVA was used to test species and sex differences in allometry across the whole data set to evaluate the ability of allometries to identify species and sex (signi cant sex × species interaction). Then, sex-speci c multivariate regressions of shape on size were calculated for every species separately. The sex-dependent effect and species-dependent effect were analyzed with a permutation test of 10,000 rounds in MorphoJ software [40]. Besides, we also examined the results after removing allometry on species and sex discrimination. A cross-validation test of correctly classi ed specimens at the species level is conducted to compare the effects of species discrimination before and after removal of allometric effect.
Sexual Shape Dimorphism (sshd) The DFA revealed no signi cant differences in wing shape between males and females. The percentage of correct classi cation of the males ranged from 60% (He. ligurriens) to 86.7% (L. sericata) while the percentage of correct classi cation of the females ranged from 46.7% (Pr. terraenovae) to 100% (C. vomitoria) ( Table 8).

Phenetic Relationships Among Blow Flies
As shown in Fig. 4, the UPGMA dendrogram of nine species were divided into two distinct groups, with Ch. ru facies and Ch. Megacephala always in a separate group, other species in the same subfamilies cluster into a cluster except Pr. terraenovae. The comparison of three pictures shows that the identi cation e ciency with male samples is better than that with female samples. In all three dendrogram, C. vomitoria was wrongly assigned to the branch of He. ligurriens and L. sericata.

Phylogenetic analysis
The sequence of nal alignment contained 111 variable sites, revealed a strong AT bias, with the average nucleotide compositions of A (30.0%), T (37.9%), C (17.4%), and G (14.7%), respectively. At the species level, the nine esh y species constituted their own monophyletic clusters with very strong supportive values. The NJ tree analyzed with the Tamura 3-parameter model showed that phylogenetic analyses of COI sequences yielded a tree of two distinct clades. The rst clade contains two branches. The rst branch consists of C. vicina, Tr. calliphoroides, C. Vomitoria and A. grahami while the second branch contains He. ligurriens and L. sericata. The second clade also includes two branches, one branch is Pro. terraenovae, the other branch containing Ch. ru facies and Ch. megacephala. Although the samples come from different places, they can still form a cluster (Fig. 2) .

Allometric Effects In Species And Sexes
The result of Procrustes ANOVA showed that allometries differed in species and sexes (signi cant sex × species interaction in Table 5), and sex has no signi cant in uence on centroid size. The results of allometric analysis show that allometry accounted for 3.02% of the total shape variation, 2.49% of the female wings; for the male wings among nine species, allometry explained 2.16% of the total shape variation. The analysis of wing shape and size of each species revealed allometry in most species except for C. vicina, L. sericata and Pr. terraenovae (permutation test with 10,000 rounds in MorphoJ: P > 0.05). The wing shape variation was found to be signi cantly between sexes in A. grahami, Ch. megacephala and Ch. ru facies (permutation test with 10,000 rounds in MorphoJ: P < 0.0001, P < 0.01, and P < 0.05) ( Table 6). The percentage of correctly classi ed specimens at the species level in cross-validation test after removing the allometric effect is shown in Table 7. Table 5 Results of Procrustes ANOVA on centroid size and shape. Table 6 The predicted percentage in each blow y species and between sexes of each species.

Discussion
In this study, we use only wing shape rather than size for analysis since previous studies have demonstrated that wing size cannot be used to distinguish species [27]. Compared with size, the shape of wings is less susceptible to environmental factors and has been proved to be a more stable feature [42,43], which can provide very rich information for the phylogeny and evolution of organisms [44,45]. The results suggested that wing morphometrics can be useful for identi cation genera and species of necrophagous Calliphoridae in China, which is consistent with previous studies in other countries [29,38].
Except He. ligurriens, Ch. megacephala and Ch. ru facies, overlapping classi cation was observed among some species, but the percentage of overall correct classi cation rate at the level of genus and species is high, ranging from 80-100% in genus and from 76.7-100% in species.
In morphological research, multivariate regression used to estimate allometry effects of different species is an important step in shape analysis since allometric growth can affect taxonomic research and sexual dimorphism [27]. Procrustes ANOVA has showed that allometries differed in species or sexes, and both species and sex have a signi cant effect on wing shape. The results in this study showed wing size explains part of the variation in wing shape among species, within a species, and between sexes in most species. This study con rmed that size correction by using residuals from the regression of shape on size was a necessary step in species identi cation based on shape [27]. Besides, the correct identi cation rate of species before and after the removal of allometric growth were compared in this study. We found that the identi cation accuracy after the removal of allometric growth was slightly lower than that keeping allometric growth. But allometry accounted for only 3.02% of the total shape variation, revealing that the allometry effects seem only a very small factor on the shape variation of species. However, to ensure the accuracy of identi cation, we suggest that, in the future laboratory research or practical application, the allometric growth should be removed rst.
The differences between sexes were insigni cant in all species (P > 0.05), which is slightly different from other studies. Signi cant difference was found between sexes of C. vomitoria, but not of Ch. Megacephala and Pr. Terraenovae collected in Europe [28]while differences between sexes existed in blow ies collected in Thailand, namely Ch. Megacephala and He. Ligurriens [27], indicating the morphology of Calliphoridae might differ in various locality of origin around the world. Therefore, the geographical populations can be identi ed according to morphometric analysis [38]. It is worth noting that to compare the effectiveness of morphometric analysis in male and female identi cation for certain species, the same number of males and females were used in each species to avoid the bias caused by different numbers. There were high accuracy rates of sex identi cation in some species like C. vomitoria, C. vicina and L. sericata, and the correct recognition rate of female C. vomitoria up to 100, indicating the wing shape can be used to identify sex for certain species, but more samples are needed to con rm this conclusion.
According to taxonomic criteria, these species belong to three subfamilies. The genus Lucilia and Hemipyrellia belong to Luciliinae; Calliphora, Triceratopyga and Aldrichina belong to Calliphorinae; Protophormia (phormiini), Achoetandrus and Chrysomya (Chrysomyini) belong to Chrysomyinae. The results of molecular analysis clearly divided into three clusters with Luciliinae and Calliphorinae in a large branch, Chrysomyinae in a small branch, and phormiini and Chrysomyini separated from others. In addition, phylogenetic trees show that the same species from different countries cluster in one branch.
The interspeci c distance of less than 1% indicated less difference between species collected from different countries or different places in China. Distance analysis of intraspeci c variation between species showed that high percentage of intraspeci c divergence between calliphorid species ranged from 4.3-13.6%, demonstrating that these species can be well distinguished by molecular methods. The interspeci c gaps between C. vicina and C. vomitori as well as C. vicina and T. calliphoroides are relatively low as they belong to the same subfamily, which proved the effectiveness of molecular species identi cation. However, as seen in Fig. 4, no matter female or male species were used in wing morphometrics, C. vomitori and L. sericata belonging to different subfamilies always formed in one branch. In contrast, through molecular methods, all species can be accurately identi ed to corresponding subfamilies, suggesting that the accuracy of molecular identi cation is higher than that of wing morphometrics in these species. Besides, the integrity of the wing must be ensured in wing morphometrics analysis.
The comparison results show that both molecular analysis and wing morphometrics can be used as important methods to distinguish species, but for certain species, wing morphometrics may give a wrong result. Meanwhile, molecular identi cation cannot distinguish geographical populations for some species [32]. Therefore, we recommend classical molecular identi cation for species identi cation to achieve better accuracy, and wing morphometrics can be used as a supplementary tool to identify geographical origin and sex. In the future, more species would be included to.

Conclusion
Our study results show that molecular analysis is more e cient for species identi cation, and wing morphometrics can be used as a complementary method combined with molecular methods or other methods to ensure the accuracy of identi cation. However, wing morphometrics has some advantages in identifying geographic population and sex of certain species. As a simpler and cheaper method, it is quite convenient for non-taxonomists to perform the identi cation after basic training [31,35]. In the future research, larger sample size and other identi cation methods could be used to further test this conclusion and to achieve an optimal identi cation method set.