Introduction

The phylum Nematoda comprises one of the largest and most diverse groups of animals. Most species are found in oceanic, freshwater and soil ecosystems, and only a few are pathogens of animals and plants1. Plant-parasitic nematodes (PPNs) have a diverse morphology and parasitic habits2. PPNs are distributed between the classes Chromadorea and Enoplea within very restricted orders (Rhabditida, Dorylaimida and Triplonchida)3. The order Dorylaimida, which belongs to Enoplea, includes several genera of PPNs in the family Longidoridae (Australodorus, Longidoroides, Longidorus, Paralongidorus, Paraxiphidorus, Xiphidorus and Xiphinema)3. These nematodes are of particular scientific and economic interest because they directly damage the roots of the host plant and some are vectors of several plant viruses (genus Nepovirus) that cause severe damage to a wide variety of crops4. Because of its great morphological diversity, the genus Xiphinema has been divided into two species groups5,6,7,8: (i) the Xiphinema americanum group, which comprises a complex of approximately 60 species, and (ii) the Xiphinema non-americanum group, which comprises a complex of more than 200 species. The traditional identification of these species by morphology and morphometric studies is very difficult because of their high intra-specific morphological variability, which can lead to considerable overlap of many characteristics and ambiguous interpretation6, 9. For this reason, new approaches are needed to ensure accurate species identification. Recently, numerous species from Longidoridae (44.4%) were molecularly characterized by ribosomal RNA genes (rDNA), i.e. partial 18S, ITS regions, or the D2 and D3 expansion segments of the 28S rRNA gene, as well as by the protein-coding mitochondrial gene cytochrome c oxidase subunit 1 (coxI), constituting a useful tool for species identification and the establishment of phylogenetic relationships within PPNs6, 10,11,12,13,14. Several studies conducted with 18S rRNA gene sequences11, 15, 16 did not provide taxonomic clarity among Longidoridae, since this gene seems to evolve too slowly to be useful as an appropriate marker for phylogenetic studies at the species level. The ITS region, D2–D3 of 28S rDNA sequences, and the coxI gene could be considered good markers for species identification. However, due to molecular variability in the ITS region, it appears better suited for species identification than for phylogenetic analysis17. Additionally, recent studies showed that mtDNA genes evolve much more quickly than rRNA genes, revealing low intra-specific and high inter-specific molecular variability for Longidoridae12, 16, 18,19,20,21. Therefore, it seems to be the most promising marker to relieve taxonomic confusion within this group. The coxI gene is frequently used as an efficient marker for species identification in the animal kingdom and may also be used to estimate species richness, particularly in understudied faunas22.

Therefore, the objectives of this research were to evaluate the variability of the mitochondrial marker gene coxI and partial sequence of the 28S rRNA gene within Longidoridae, as well as their usefulness as markers for barcoding and for reconstructing the phylogeny of the group.

Results and Discussion

coxI amplification in Longidoridae

A total of 136 new accessions belonging to 82 species for coxI were obtained for the first time in this study (Tables 1 and S1). Taxon coverage (species/genus species) of 11.9%, 8.3%, and 1.5% was achieved for Xiphinema, Longidorus and Paralongidorus, respectively. PCR amplification and sequencing for the partial coxI were carried out by combining several primers (Table 1). The best set of primers were COIF/XIPHR221, followed by JB3/JB423, COIF/COIR and COIF/XIPHR121. These sets of primers amplified a single fragment of approximately 500 bp. We did not find amplification of pseudogenes using these sets of primers. However, we did not perform a systematic analysis of primer amplification, as we started with the combination COIF/XIPHR2 in the majority of the studied samples; this combination was reported to be efficient in previous studies21. All new partial coxI sequences were obtained using voucher specimens identified by integrative taxonomy, with the combination of morphological characteristics and unequivocal molecular markers from the same individual nematode, viz. the D2–D3 region (Tables 1 and S1) and ITS1 in some cases.

Table 1 Taxa sampled for dagger and needle nematodes species of the family Longidoridae and sequences of cytochrome c oxidase subunit 1 (coxI) used in this study. Species identifications were based on morphology and barcoding using D2–D3 expansion segments of 28S rDNA1.
Full size table

mtDNA and rDNA molecular variability

To our knowledge, the present study is the largest survey ever conducted for Longidoridae mtDNA and rDNA molecular variability. It covers 44 species (268 sequences), 112 species (577 sequences) and 64 species (252 sequences) for partial coxI, D2–D3 and ITS respectively, with more than one sequence per species as available in GenBank or obtained in this study (Tables S2–S4). However, some genera of Longidoridae were underrepresented (e.g., Paralongidorus and Xiphidorus) (Table S1).

For the partial coxI gene, 14 species (101 sequences) from the X. americanum group were studied, of which 7 showed a percent similarity lower than 90%: X. americanum (78.82%), X. brevicolle ‘complex’ (76.67%), X. californicum (89.83%), X. incognitum (86.61%), X. rivesi (70.94%), X. peruvianum (79.71%) and Xiphinema sp. 1 (82.66%). In the X. non-americanum group, intra-specific molecular variability of coxI was analysed in 18 species (89 sequences), but only two species within this group showed similarity values lower than 90%: X. adenohystherum (88.40%) and X. italiae (69.73%). The intra-specific molecular variability detected in 11 studied Longidorus species (52 sequences) was high; 4 of them showed a percentage of similarity below 85%: L. magnus (78.70%), L. orientalis (78.78%), L. poessneckensis (84.62%), and L. vineacola (68.91%). Finally, only one species from the genus Paralongidorus with available partial coxI sequences was found—Paralongidorus paramaximus—with 99% similarity between the three sequences analysed.

The majority of sequence variability in all the studied genera appears at the third codon position, as for L. helveticus, which showed a sequence similarity of 92.66% with all variations at silent sites24, or L. poessneckensis, which showed an 81% sequence similarity with all molecular variability at silent sites, except for two nucleotides that caused changes in the amino acid sequence25. In the majority of the studied cases, mean Kimura 2-parameter distance (K2P) values did not exceed the interspecific distance mean, except for 5 species from the X. americanum group: X. americanum, X. brevicolle ‘complex’, X. peruvianum, X. rivesi, and Xiphinema sp. 1. However, these species comprise species complexes that must be further studied, as recently proposed by Orlando et al., because some of them may have been misidentified26. In contrast, intra-specific molecular variability detected in X. italiae and X. adenohystherum was accurate and correct. In both cases, these species were identified by integrative taxonomic approaches, and molecular analyses were performed using the same DNA extraction of single individuals for different markers (D2–D3 and coxI). Integrative identification of the X. non-americanum group is apparently less difficult due to more taxonomically informative traits (e.g., uterine differentiation) and the higher number of species molecularly studied. Similarly, Longidorus spp. with higher intra-specific variability were clearly delineated in this study (viz. L. vineacola and L. magnus) and previous studies (viz. L. orientalis 27, L. poessneckensis 25 and L. helveticus 24), using integrative taxonomy and the combination of unequivocal molecular markers (D2–D3 and partial coxI) from single individuals. Our results suggest that intra-specific variation in the partial coxI gene may be higher than expected. However, more species and more populations should be studied in the future to clarify the real molecular variability among species within Longidoridae.

In contrast, the D2–D3 region showed low intra-specific molecular variability, since no similarity value below 95% was detected for any of the studied species (except X. americanum, with 94.65% similarity), even though there are more sequences from this region than for the partial coxI (112 species for D2–D3 vs 43 species for coxI) (Table S3). However, this lower intra-specific molecular variability may confound species identification, especially within the X. americanum group, where seven species showed molecular similarity values of 99% (X. rivesi, X. santos, X. citricolum, X. americanum, X. thornei, X. pacificum and X. georgianum) (data not shown). High inter-specific similarity values were detected in the other species—L. wicuolea and L. silvestris or X. pseudocoxi and X. globosum—which showed a similarity value of 97%. Hence, in these species, this marker could not provide clear species identification, and other sequences and integrative taxonomic approaches must be applied28.

The ITS1 maker showed low intra-specific molecular variability in the majority of the species studied; only some species showed a significantly low similarity (below 90%), such as X. brasiliense (89%), X. inaequale (80%), X. chambersi (87%), and L. biformis (85%). Unfortunately, because no data were available to confirm that these cases were misidentifications, further research is needed to confirm this high molecular variability. ITS sequences have been a prominent choice for species identification because this region is one of the most variable nuclear loci, and the availability of universal primers that work with most nematodes29 has contributed to its extensive use (Table S4). However, the high length and sequence variability between Longidoridae species complicates the construction of a plausible alignment of this region. Thus, this region appears to be better for species delimitation than for phylogenetic studies17, 29.

Maximum intra- and minimum inter-specific distances for each coxI and D2–D3 sequences are shown in Fig. 1, which shows that higher molecular variability for K2P distance was associated with partial coxI than with D2–D3 region for intra- and inter-specific comparisons. As discussed above, the range of intra- and inter-specific distances in the X. americanum group was minimal for the D2–D3 region. Importantly, the difference between intra- and inter-specific distances in the X. non-americanum is large and non-overlapping. The intra-specific variability in coxI is largely attributable to X. italiae in this group.

Figure 1
figure 1

Intra- and inter-specific distance (K2P) for D2–D3 region and coxI markers for different groups of species within Longidoridae. Distances calculated using the biggest distance for intra-specific variability for each individual (sequence) among the sequences for the same species and the smallest distance among species for each individual. The box shows the third (Q3) and first (Q1) quartile range of the data and the median. Whiskers indicate minimum and maximum values of the data. Data falling outside the box and whiskers (circle) range are plotted but considered outliers.

Full size image

Barcoding

To evaluate how well various barcoding tools perform for Longidoridae, we analyzed datasets for species that had been previously identified using integrative taxonomy and in addition data for Longidoridae deposited in GenBank. Three software packages were tested: Weka, Spider and phylogenetic trees topology based on MrBayes. We included and excluded the X. americanum group to understand the effect of these close-related species in our analysis. Our results suggest that DNA barcoding could be a powerful tool for the majority of species in Longidoridae using several approaches: (a) supervised machine learning methods; (b) distance threshold methods and (c) monophyly for species with more than two sequences in phylogenetic trees. However, barcoding results were highly dependent on the selected molecular marker and the technique used (Tables 2 and 3). Both mitochondrial and ribosomal sequences have been used as barcoding regions for nematodes in studies with smaller sample sizes and a larger phylogenetic range30, 31. Since our sequences were all derived from single vouchered specimens and are of high quality because we sequenced PCR products from both ends, the present reference database could also be a valuable tool for validating field collections32. The marker could also be used for soil nematode metabarcoding33, 34. The majority of our sequences for partial coxI are 400 bp long, which is in the range of appropriate size suggested by iBOL data quality: length of finished sequence must be >75% of approved marker length (e.g., 500 bp for coxI), with an expectation of 2X coverage (http://ibol.org/about-us/how-ibol-works/). With this sequence, we could clearly re-identify the majority of species, except for closely related species in the X. americanum group or species that were probably misidentified. The D2–D3 marker showed considerable sequence similarity in the X. americanum group, and for this reason two datasets were studied—one with all sequences and other excluding these sequences—to check the validity for the X. non-americanum-group species (Tables 2 and 3).

Table 2 Accuracies (% correctly identified sequences from the test dataset) for barcoding in Longidoridae using the program Weka v.3.8.0. The datasets included all sequences of accessions that were identified to the species level and was divided into 80% as train set and 20% as test.
Full size table
Table 3 Accuracies for barcoding in Longidoridae using SPIDER package and tree-based comparison for monophyly using Bayesian inference.
Full size table

The coxI and D2–D3 markers performed differently depending on the barcoding techniques used. The learning methods implemented in the Weka package achieved similar results for the coxI marker, ranging from 78.43% to 88.24% (Table 2). The performance of classification by machine learning was not strongly influenced by the presence of X. americanum-group sequences (384 vs. 560 sequences in D2–D3) (Table 2). The Bayesian-based method naïve Bayes classifier35 did not perform well with the D2–D3 data including or excluding the X. americanum group (36.03 and 36.84% of sequences assigned to correct species). The best classifier was the iterative classifier optimizer36 with 94.59 to 96.05% of sequences assigned to the correct species, followed by the decision tree C4.5 (J48)37 and the rule-based RIPPER (Jrip)38.

Using the Spider package, the Near Neighbour method showed very good accuracy for coxI, with almost 100% of correct identifications. Best Close Match performed less well. For both methods, the exclusion of the X. americanum group increased accuracy (Table 3). These results showed the potential for barcoding with these software packages for the majority of our species using both markers. In the case of MrBayes, phylogenetic analysis for species with more than one sequence showed that 92.9% of our species presented a monophyletic position in the tree for coxI. This performance was similar for the D2–D3 marker when both including (90.1%) and excluding the X. americanum group in Longidoridae (100%) (Table 3).

The knowledge of intra- and inter-specific molecular variability is important to detect misidentifications or cryptic speciation in different nematodes groups. Approximately a quarter of the sequences for coxI and D2–D3 region including X. americanum group showed a larger intra-specific than inter-specific molecular diversity; while an approximately 10% of the sequences was for D2–D3 region excluding X. americanum group (Table 3). Even with these differences, the performance was good and probably these molecular differences included the important molecular variability of some species, low intra-specific variability in others (species from the X. americanum group), poorly corrected sequences from chromatograms or sequences from PCR cloning products and, in some cases, incorrect identifications deposited in GenBank. Using an experimental script provided by the R package Spider, we were able to calculate the approximate optimal molecular differences for barcoding, which were 6.36% for coxI and 2.87% and 2.04% for D2–D3 when including the X. americanum group or excluding it, respectively (Table 3). Although this script is experimental and should be used with caution, our integrative taxonomic identifications in Longidoridae support these values9, 20, 28, 39, 40.

Phylogeny of Longidoridae using nuclear and mitochondrial sequence data

The phylogeny obtained using the coxI fragment (583 sequences) showed a monophyletic clade for the X. non-americanum-group species and a clade for Paralongidorus and Longidorus species, while the X. americanum group was paraphyletic (Fig. 2). However, all clades were weakly supported (<0.95 Bayesian probability values (BPP)). The phylogenies at the species level relationship generally supported the phylogenetic relationships among groups of species in Xiphinema more than in Longidorus reported in former papers (Fig. S1)6, 9, 11, 28, 39, 40. Nevertheless, in this wider analysis, we could not clearly determine groupings such as X. brevicolle ‘complex’ (nested among X. diffusum, X. taylori, and X. incognitum), and some entries for X. rivesi (from different geographical locations) following the corrections performed by Orlando et al. for the X. americanum group (Fig. S1), as one X. rivesi sequence (AM086697) was considered as X. floridae (AM086696)26. In addition, Xiphinema sp. 5 studied by Orlando et al.26 nested inside Longidorus. However, when BLASTn was performed on GenBank, this sequence matched as a Xiphinema sp. The separation among species was remarkable, with the exception of a few species in the X. americanum group, using a phylogenetic approach. The base saturation (third nucleotide position in each codon) and the short fragment used in this study could be responsible for this lack of phylogenetic resolution at the genus level and between X. americanum and X. non-americanum group inside the genus Xiphinema. Additionally, different mutation rates in the mitochondrial genome and the wide evolutionary differences within these studied groups could complicate the phylogeny. A dataset excluding the third codon position did not improve the phylogeny, and in fact made it worse because of the low phylogenetic signal (Fig. S2). Probably, a possible improvement in the phylogenetic relationships among genera in Nematoda could be obtained using full mitochondrial genomes41, 42.

Figure 2
figure 2

Phylogenetic relationships within Longidoridae. Bayesian 50% majority rule consensus tree as inferred from analysis of the partial coxI sequence alignment under a TrN + I + G model. Posterior probabilities more than 0.70 are given for appropriate clades.

Full size image

The phylogeny of nuclear ribosomal marker (D2–D3) based on 1085 sequences of Longidoridae showed a similar pattern of separation among genera (Figs 3 and S3) after corrections for some misidentified species (X. cretense and X. diversicaudatum)43, 44. However, here, the separation for some species was better than in the coxI tree, since the X. non-americanum-group species and Longidorus-Paralongidorus (with the exception of L. laevicapitatus) were clearly separated into two well-supported clades (Figs 3 and S3). However, the X. americanum group formed a clade that is, however, weakly supported (≤0.90 BPP). As in the analysis with coxI, the genus Paralongidorus was nested among the Longidorus spp. clade. Xiphinema americanum s. s. species formed a low supported clade (0.77) (Fig. S3). As mentioned before, this group of species showed low nucleotide variability, probably because of a short speciation time among these species. Paralongidorus species formed a well-supported clade (1.00 BPP) inside Longidorus, with the exception of P. bikanerensis. This phylogeny is similar to others for Longidoridae9, 39, 45,46,47. Longer sequences probably need to be added in order to address this problem of deep resolution, but major clades have been clearly resolved using a more slowly evolving gene such as 18S. Recently, the sequencing of four additional mitogenomes of Longidoridae supported a similar phylogenetic pattern of Paralongidorus being most closely related to Longidorus, both associated with the Xiphinema species48.

Figure 3
figure 3

Phylogenetic relationships within Longidoridae. Bayesian 50% majority rule consensus tree as inferred from analysis of the D2–D3 region alignment under a GTR + I + G model. Posterior probabilities more than 0.70 are given for appropriate clades.

Full size image

Conclusions

This is the first broad study of the variability of molecular markers used for phylogenetic relationships and the identification of Longidoridae. This research significantly increases the number of coxI sequences available for Longidoridae using integrative taxonomic approaches with voucher specimens and the combination of several unequivocal molecular markers (coxI, D2–D3 region and ITS1, in some cases) from one individual nematode. The ITS1 region showed promise for barcoding and species identification because of the clear molecular variability among species. However, difficulties with obtaining an unequivocal alignment limit its usefulness beyond BLASTn-like searches. In addition, we revealed problems for species delimitation in Longidoridae, as well as phylogenetic relationships using coxI and D2–D3 regions. However, in shallow phylogenetic relationships (close to the external branches of the tree) or for a restricted number of species, these markers gave good results. Several barcoding methods showed the utility of coxI and D2–D3 for species identification, except for some species in the X. americanum group (for which more studies are necessary for longer sequences or different markers). Our results suggest that the use of more than one molecular marker is essential for the correct identification of Longidoridae unless integrative taxonomical approaches are employed.

Material and Methods

Samples and nematode extraction

Nematode soil samples were collected from 2007 to 2016, mainly in Spain but also in Greece, Japan, the USA, Russia and Italy, from the rhizosphere of a wide variety of plants, including both agriculture and natural ecosystems (Tables 1 and S1). At each site, several subsampling points were randomly selected for soil sampling in an area of 5 m2. Soil samples were collected with a shovel discarding the upper 5-cm top soil profile from a 5- to 40-cm depth, in the close vicinity of active roots. To obtain a representative soil sample per site, all subsample soils were thoroughly mixed before nematode extraction. Nematodes from the soil were extracted from a 500-cm3 sub-sample using the magnesium sulphate centrifugal-flotation method49. The extracted nematodes were identified by selecting adult nematode specimens belonging to Longidoridae. Nematodes were fixed in 4% formaldehyde, processed with glycerin50, and identified by morphological traits to the genus or species level. Some additional nematodes from the same morphotype were not fixed and were used for molecular studies from each site.

DNA extraction and PCR conditions

For molecular analyses, to avoid complications from mixed species populations in the same sample, at least two live nematodes from each sample were temporarily mounted on a drop of 1 M NaCl containing glass beads (to avoid crushing the nematode). Here, diagnostic morphological characteristics were observed and measurements were taken to confirm species identity. The slides were dismantled and DNA was extracted. Nematode DNA was extracted from single individuals and PCR assays were conducted as described by Castillo et al.51. The portion of the partial coxI gene was amplified, as described by Lazarova et al.21 using the primers COIF (5′-GATTTTTTGGKCATCCWGARG-3′), COIR (5′-CWACATAATAAGTATCATG-3′), XIPHR1 (5′-ACAATTCCAGTTAATCCTCCTACC-3′) or XIPHR2 (5′-GTACATAATGAAAATGTGCCAC-3′) and as Bowles et al.23 using primers JB3 (5′-TTTTTTGGGCATCCTGAGGTTTAT-3′) and JB4 (5′-TAAAGAAAGAACATAATGAAAATG-3′). PCR cycle conditions for mtDNA were as described by Lazarova et al.: 1 cycle of 94 °C for 1 min, 50 °C for a further 1 min and 72 °C for 2 min. This was followed by 40 cycles of 94 °C for 1 min, 45 °C for 1 min and 72 °C for 2 min. The PCR was completed with a final extension phase of 94 °C for 1 min, 45 °C for 1 min and 72 °C for 5 min21. The D2–D3 region was obtained using a protocol and primers described in Archidona-Yuste et al.9, 39. PCR products were purified after amplification using ExoSAP-IT (Affmetrix, USB products) and used for direct sequencing in both directions. The resulting products were run on a DNA multicapillary sequencer (Model 3130XL genetic analyser; Applied Biosystems, Foster City, CA, USA), using the BigDye Terminator Sequencing Kit v.3.1 (Applied Biosystems, Foster City, CA, USA), at the Stab Vida sequencing facilities (Caparica, Portugal). The newly obtained sequences were submitted to the GenBank database under accession numbers indicated on the phylogenetic trees and Tables 1 and S1.

Nucleotide variability analyses

A total of 577, 257, and 261 sequences from 112, 65 and 44 species of Longidoridae were used to calculate the intra- and inter-specific molecular variability of 28S, ITS1 and coxI, respectively. For intra-specific molecular variability, one dataset from each species with more than one available sequence (Tables S2–S4) was created and aligned using MAFFT v. 7.252 with default parameters. Then, pairwise divergence among taxa were computed as a percentage of sequence similarity, singletons sites and parsimony informative sites using the program MEGA v. 7.053 (Tables S2–S4). Additionally, for coxI, p-distance was calculated for each codon position. For inter-specific molecular variability, four datasets were created, including sequences from the X. non-americanum group, X. americanum group, Longidorus spp. and Paralongidorus spp. Nucleotide variability indices were calculated in the same way as the intra-specific molecular variability after grouping the different species in each dataset (MEGA v.7.0). “Spider” package54 with R version 3.1.1 freeware (R Core Development Team; CRAN, http://cran.r-project.org)55 generates two statistics for each sequence (individual) in the dataset: the furthest intra-specific distance among its own species and the closest, non-conspecific (i.e., inter-specific distance). These data were used to create Fig. 1 among makers and species groups.

Barcoding analyses

Species without clear taxonomic status (X. brevicolle) and sequences considered misidentifications using several phylogenetic analyses9, 26, 39, 43, 44, as well as sequences with less than 300 bp in the D2–D3 fragment, were excluded from the analysis. Two datasets were used, corresponding to the coxI and D2–D3 regions. Several barcoding methods were used to test the utility of these molecular markers for species identification: (i) supervised machine learning methods to classify species following the method explained by Weitschek et al.56 using the Weka machine learning software55, which includes a collection of supervised classification methods. Jrip, J48, and naïve Bayes were used as supervised classification methods. The dataset included all species identified with all molecular variability using a test option for the dataset with a percentage split of 80% train set of sequences and 20% as test sequences, this option is allowed in Weka v.3.8.057 using the following Weka classifiers: (1) the rule-based RIPPER (Jrip)38; (2) the decision tree C4.5 (J48)37; (3) the Iterative Classifier Optimizer57; and (4) the Bayesian-based method naïve Bayes35. (ii) Tests of barcoding “best close match”58, nearest-neighbour identification59, and a standard threshold cut-off for species separation was determined using the function “localMinima” (this function determines possible thresholds from the distance matrix for an alignment) using a dataset for both the coxI and D2–D3 regions (including and excluding the X. americanum group) using the indications and principal functions implemented in the “spider” package54 with R version 3.1.1 freeware (R Core Development Team; CRAN, http://cran.r-project.org)55. Additionally, iii) phylogenetic trees conducted using MrBayes were analysed for species monophyly and species congruence for species with more than one available sequence. For this analysis, species not forming a monophyletic clade were considered not well identified, and the number of divergent sequences was annotated.

ITS1 sequences were excluded from all analyses because of the high divergence degree and difficulties with regard to phylogenies and correct alignments. However, a molecular variability table was considered in order to elucidate the molecular diversity of this marker in Longidoridae.

Phylogenetics analyses

Nucleotide data sets consisted of the partial coxI fragments for barcoding species in Longidoridae and of protein coding fragments. Outgroup taxa were Heterodera elachista and Rotylenchus striaticeps. The newly obtained and published sequences for each gene were aligned using MAFFT v. 7.252 with default parameters. Sequence alignments were manually edited using BioEdit57. Phylogenetic analyses of the sequence data sets were performed based on Bayesian inference (BI) using MrBayes 3.1.260. The best fitting model of DNA evolution was obtained using jModelTest v. 2.1.761 with the Akaike Information Criterion (AIC). The Akaike-supported model, the base frequency, the proportion of invariable sites, and the gamma distribution shape parameters and substitution rates in the AIC were then used in phylogenetic analyses. BI analysis under a Tamura-Nei with a proportion of invariable sites and a gamma-shaped distribution (TrN + I + G) model for coxI mtDNA was run for 4 × 106 generations, while for the first and second nucleotide for each codon a transversion model with a proportion of invariable sites and a gamma-shaped distribution (TVM + I + G) was used, with 10 × 106 generations. The general time reversible model with a proportion of invariable sites and a gamma-shaped distribution (GTR + I + G) using 10 × 106 generations was used for the D2–D3 maker. The Markov chains were sampled at intervals of 100 generations. Two runs were performed for each analysis. After discarding burn-in samples and evaluating convergence, the remaining samples were retained for further analyses. The topologies were used to generate a 50% majority rule consensus tree. Posterior probabilities (PP) are given in appropriate clades. Trees were visualized using TreeView62 and FigTree v1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).