Molecular and Morphological Analyses Support Different Taxonomic Units for Asian and Australo-Paciﬁc Forms of Ischnura aurora (Odonata, Coenagrionidae)

: Despite the great technological progress that has aided taxonomical identiﬁcation, taxonomical issues remain for certain species found in remote and/or understudied geographical areas. The damselﬂy species Ischnura aurora has been the subject of a long-standing taxonomical debate, focused mainly on the existence of morphological and behavioural differences between Asian and Australo-Paciﬁc forms of this species that could justify their placement into two different species. Here, we carried out a comparative morphological analysis of specimens currently identiﬁed as I. rubilio from India and I. aurora from Asia and Oceania, combined with the analysis of mitochondrial and nuclear sequence data, both developed by us and available in public repositories. Our results split the Asian and Australo-Paciﬁc forms of I. aurora into two well-differentiated taxonomic units and, hence, different (albeit closely related) species, and support the speciﬁc status of I. rubilio . The results of our genetic analyses suggest the existence of a third (and even fourth) taxonomic unit, stressing the need to revise all available material belonging to the different I. aurora subspecies that have been described. Finally, we have identiﬁed several questionable DNA sequences currently available in public repositories, upon which previous conclusions about the phylogenetic position of I. rubilio are based. Our study stresses the importance of being able to link available DNA sequence data with voucher specimens as well as to carry out a careful examination of DNA sequence data prior to their inclusion in taxonomical studies.


Introduction
The science of taxonomy is central to many disciplines within biology, and hence being able to correctly identify species and to associate a scientific name with a particular organism is a prerequisite for ecological and conservation studies [1,2]. Until the development of alternative techniques to define species boundaries, the identification of species by means of morphological analysis was the only option available to taxonomists, even though the existence of "cryptic" species constituted a clear limitation. Taxonomy has shown great advances in recent years with the incorporation of technological advances (the most relevant being DNA sequencing) and the possibility of virtually accessing museum collections for specimen examination [3], which has led to integrative taxonomy, i.e., the study of variation in different types of datasets to delineate species boundaries more accurately [4]. Since 2003, when the DNA-based approach to taxonomy was first proposed [5][6][7], the so-called DNA barcode-the mitochondrial COI gene-has been widely used for species descriptions. However, the well-known limitations of mitochondrial DNA (e.g., incomplete lineage sorting, introgression, or the presence of nuclear copies of mitochondrial genes [8][9][10][11]) may lead to erroneous conclusions in species delimitation studies (see for example Papakostas et al. [12] or Ožana et al. [13]), and therefore the use of information species delimitation studies (see for example Papakostas et al. [12] or Ožana et al. [13]), and therefore the use of information from both mitochondrial and nuclear DNA markers is preferred, as it integrates information from two different genetic sources, hence allowing for a better support of taxa. Despite the acknowledged advantages of DNA sequences as a tool for species identification, species delineation is the crucial first step in taxonomy. To accurately delineate species boundaries, we need to be able to link a newly discovered species with its correct name, and describe its natural history, morphology, and behaviour [14].
Regardless of all the technological progress that has helped with taxonomic identification, the heterogeneous fieldwork effort carried out across the world may hamper the identification of specimens found in certain areas that, in some cases, host the highest diversity. Therefore, the combination of extensive fieldwork and modern technologies is necessary to clarify biodiversity, an essential matter nowadays due to the diversity loss produced because of climate change. Damselflies (Zygoptera: Odonata) constitute a good example to address taxonomic issues due to their high diversity in environments with difficult access and the high interspecific morphological similarities that exist in this insect group, sometimes only possible to unravel by genetic analysis [15]. Recent works have pointed to the genus Ischnura Charpentier, 1840, also known as forktails (Coenagrionidae), as a group with a non-resolved taxonomy, partly due to an incomplete knowledge of this genus in several areas of Asia [16,17]. Ischnura is a speciose genus with worldwide distribution, which has colonised many oceanic islands and shows great diversity in morphology, colouration, and behaviour. There are currently ca. 77 known species of Ischnura, which can be found in a range of diverse environments and spread across vast distribution areas [17][18][19]. One example of a species within this genus showing a wide geographical distribution and taxonomical issues is Ischnura aurora (Brauer, 1865). This species is a well-known migrant, passively dispersed throughout long distances as part of the aerial plankton [20,21] (pp. 396 and 409-411). Its ability to disperse over vast geographical areas explains its widespread distribution range that spans from the Pacific islands and Australia to the South-East Asiatic continent, India, Pakistan, and Iran [22,23] (Figure 1). There has been a long-standing debate about which name should be used when referring to individuals of this species. In 1858, the name Agrion delicatum was given by Hagen to a damselfly found in Sri Lanka, Bengal (currently India and Bangladesh) and There has been a long-standing debate about which name should be used when referring to individuals of this species. In 1858, the name Agrion delicatum was given by Hagen to a damselfly found in Sri Lanka, Bengal (currently India and Bangladesh) and Australia [24]. In his work, Hagen included only data on body and wing sizes, but failed to provide an actual description of the species, which has remained as a nomen nudum since then. Later, A. delicatum was transferred to the genus Ischnura and the species Ischnura delicata was described using material from Asia and Australia [25]. I. delicata was considered a senior synonym of Agrion (Ischnura) aurora, a very similar damselfly earlier described by Brauer [26] using material from Tahiti. In his description of I. delicata, Selys mentions a male of a "variety or juvenile" of the species from India, which he named Ischnura rubilio. Selys provided information on some key morphological characteristics of I. rubilio: "the basal articulation of the 4-6 segments not circled in black; the 8 segment entirely blue like the 9" [25] (p. 283). Even though this taxon is still awaiting formal description, it was considered a subspecies of I. aurora [27] until Kalkman et al. [28] raised it to specific status. Therefore, I. rubilio is currently recognised as a distinct species restricted to the Indian subcontinent, whereas I. delicata is considered a junior synonym of I. aurora. At the molecular level, there has been no consensus yet about the placement of I. rubilio within the genus Ischnura. A phylogenetic analysis of the genus by Dumont [23], based on nuclear (Internal Transcribed Spacer, ITS) and mitochondrial (Cytochrome Oxidase I, COI) DNA sequence data which included specimens of I. rubilio collected in Bhutan and India, concluded that this species was in fact distinct from I. aurora, being either basal to the genus Ischnura or a member of the pumilio group. However, a more recent phylogenetic analysis by Sánchez-Guillen et al. [18], based also on nuclear (ITS) and mitochondrial (Cytochrome B, CYTB and Cytochrome Oxidase II, COII) sequence data, has placed I. rubilio as a sister species of I. aurora.
Beyond this debate on taxonomic hierarchy, other authors have focussed on the existence of morphological and behavioural differences between the Asian and Australo-Pacific forms of I. aurora that would justify their placement into two different species [27,29]. According to these authors, males of the Asian forms of the species possess large postocular spots, a completely blue eighth abdominal segment, a more pronounced dorsal tubercle in the 10th abdominal segment, acute cerci in its superior region, and narrower and pointer paraprocts, while the females mate in the adult stage. Australo-Pacific forms of I. aurora, on the other hand, possess small and circular postocular spots, one-third of the eighth abdominal segment with a blue colour, less pronounced dorsal tubercle in the 10th abdominal segment, rounded cerci in its superior region, and females that mate only in the teneral stage. While Papazian et al. [27] and Rowe [29] agree on the existence of morphological differences, there is no consensus on which name should be given to each of these forms. Papazian et al. [27] considered the Asian forms from the Indian subcontinent as a subspecies of I. aurora, named I. aurora rubilio (according to Selys description above), and in support of their argument, Dumont stated that "I. aurora is rare or totally absent West of the Wallace Line" [23] (p. 307). On the other hand, Rowe [29] suggested the existence of Asian forms of I. aurora in South and East Asia, as well as in Sri Lanka and India. He stated that "this 'species' should therefore be cited as Ischnura delicata (Hagen in Selys 1876)" [29] (p. 189), and, also, he highlighted the complex taxonomy of the Asian forms of I. aurora, for which three further names exist (I. rubilio, I. amelia and I. bhimtalensis), which have been assumed at some point to be either subspecies or junior synonyms of I. aurora.
Therefore, both Papazian et al. [27] and Rowe [29] coincide in restricting the "true" Ischnura aurora to the Australo-Pacific geographic area, but they disagree in the taxonomic classification and distribution of the Asian forms of this species. Here, we carried out a comparative morphological analysis of specimens currently identified as I. rubilio from India and I. aurora from Asia and Oceania, focusing on those characteristics traditionally used to distinguish among the Asian and Australo-Pacific forms of I. aurora. We combine the morphological analysis of male and female specimens with the analysis of mitochondrial and nuclear sequence data, both developed by us and available at public repositories, with the aim of clarifying the taxonomic status of these taxa. Table 1. Information on Ischnura aurora and I. rubilio specimens used for DNA extraction and sequencing as described in the main text. For each specimen, we list the voucher ID, sex, and collection locality. n.a. indicates that no sequence could be obtained for a particular marker and/or specimen; Accession numbers labelled with an asterisk are those that correspond to sequences annotated as COI-like sequences (see Appendix A and Table 2).

Species
Specimen ID Sex Collection Locality GenBank Acc. Nos Total genomic DNA was extracted from individual legs using the GeneJet DNA extraction kit (ThermoFisher Scientific, Waltham, MA, USA), following the manufacturer's protocol. Fragments of the mitochondrial COI gene and the nuclear ITS were amplified using previously published primer pairs (COI-S0: TACCAATTATAATTGGAGGATTYGG/COI-AS0: CTTCTGGATGTCCAAARAATCA and ITS-F0: GGAAAGATGGCCAAACTTGA/ITS-28S-AS0: CCTCCGCTTATTAATATGCTTAAATTC [30]). PCR reactions were carried out at specific annealing temperatures (48 • C for COI and 52 • C for ITS) using the DreamTaq Green PCR Master Mix (ThermoFisher Scientific, Waltham, MA, USA). Before sequencing, PCR products were purified with shrimp alkaline phosphatase and exonuclease I (New England Biolabs, Ipswich, MA, USA) to remove unincorporated primers and dNTPs. Cleaned PCR products were sequenced bidirectionally using BigDye v.3.1 chemistry (Applied Biosystems, Foster City, CA, USA) at either the Macrogen Laboratories in Spain or at the CACTI genomics facility from the University of Vigo.

Genetic Analyses
DNA chromatograms were visually inspected, trimmed and automatically assembled using Geneious v. 9.1.8 (https://www.geneious.com/). Previously published COI and ITS sequences from specimens identified as I. aurora, I. rubilio or I. delicata; together with representative species of the Coenagrionidae genera, Ischnura, Aciagrion, Ceriagrion, Coenagrion, Enallagma, Erythromma, Mortonagrion and Pseudagrion, were downloaded from GenBank (https://www.ncbi.nlm.nih.gov/genbank/) and added to our datasets (see Appendix B Table A1). Prior to the genetic analyses, a quality control step was carried out to ensure that the sequences included in the final datasets were not derived from contaminations or, in the case of the mitochondrial DNA, also to rule out the amplification of paralogous copies of COI (nuclear mitochondrial DNA copies or numts [11,31] which were also recently described in odonates [13]). The quality control and data mining steps are described in detail in Appendix A.
After quality control, all sequences selected for inclusion in the final datasets were aligned using MAFFT [32], as implemented in Geneious v 9.1.8. Phylogenetic relationships were reconstructed using maximum likelihood (ML) and Bayesian inference (BI) approaches. ML analyses were carried out using IQTree v. 1.6.12 [33], with the best substitution model for each marker selected by ModelFinder [34]. Support of branches in the resulting tree was assessed by 10,000 ultrafast bootstrap replicates [35,36]. BI analyses were carried out using MrBayes 3.2.6 [37,38], as implemented in Geneious v 9.1.8. MCMC searches were run for 1.1 million generations, with default priors and with the GTR + I+G substitution model. Resulting phylogenetic trees were edited with FigTree v. 1.4.3. (http://tree.bio.ed. ac.uk/software/figtree/) and Inkscape v. 1.0 [39].
Genetic differentiation between Ischnura species (uncorrected p-distances) was estimated for each dataset in MEGA X [40] using the pairwise deletion option, which removes all ambiguous positions for each sequence pair. To further confirm the species delimitation within our datasets, we used the single-locus distance-based delimitation method implemented by the software Assemble Species by Automatic Partitioning (ASAP) [41]. Analyses were run at the ASAP web server (https://bioinfo.mnhn.fr/abi/public/asap/), using fasta files for each locus (COI and ITS) as the input files. The species delimitation analyses were carried out using only the sequences from I. aurora, I. delicata and I. rubilio, with the default options and with genetic distances computed under the Kimura (K80) model.

Morphological Analyses
For the morphological analyses, we examined a total of 31 ethanol-preserved individuals of Ischnura aurora (10 males and 9 females from China; 3 males and 2 females from Fiji; 3 males and 4 females from Australia; see Table A2), plus the four dried-preserved Ischnura rubilio specimens collected in India listed in Table 1. The aim of the morphological analysis was to determine whether the morphology of these specimens would corroborate the results of the genetic analyses and justify their placement as two different taxonomic units. Specimens were examined under an Olympus SZ60 stereoscopic microscope. Photographs of the individuals were taken using a Leica Flexacam C1 digital camera attached to the microscope at varying magnifications; and afterwards stacked and edited using the software GIMP v. 2.10 [42]. The male genital ligula of one individual of I. rubilio (MB-IrbkeM) and three I. aurora individuals (ACR2880 from China, ACR0818 from Australia and ACR5009 from Fiji) was dissected and observed under a scanning electron microscope (Philips XL30) at the CACTI microscopy service from the University of Vigo. The terminology used in the morphological descriptions follows that in Garrison et al. [43]. Abdominal segments are referred to as capital "S" plus the segment number. All measurements are given in millimetres.

Genetic Analyses
All sequences generated in this study were deposited in GenBank with accession numbers OM964914-OM964934 (see Table 1). The COI sequences obtained from the I. rubilio samples collected in India were not included in the mtDNA dataset, as they were likely "COI-like" sequences or numts, rather than orthologous copies of the mitochondrial COI gene ( Table 2; see also Appendix A). Regarding the sequences retrieved from GenBank, a total of thirteen sequences belonging to I. aurora, I. delicata and I. rubilio were also excluded from the datasets after the quality control steps carried out as described in Appendix A. Three I. aurora COI sequences were excluded because they corresponded to the 3 -end of the COI gene (Table 2 and Appendix A Figure A4). Five COI sequences belonging to I. aurora (N = 2) and I. delicata (N = 3) were discarded because they showed ambiguity-coded bases, which are not expected in a mitochondrial coding gene ( Table 2). Three sequences belonging to I. aurora (N = 1) and I. rubilio (N = 2) were excluded from the datasets because they were likely the product of specimen misidentification or misplacement of individuals in the laboratory at the time of DNA extraction ( Table 2; see also Appendix A). Finally, two COI sequences of I. rubilio were excluded from the mtDNA dataset because they showed several features consistent with these being "COI-like" sequences or numts. It is important to note that three of the I. rubilio sequences that were excluded from our datasets were those included in the phylogeny of the genus Ischnura by Dumont [23]. Table 2. Sequences of Ischnura aurora, I. delicata and I. rubilio that were excluded from the final genetic analyses after the quality control steps that are described in detail in the Appendix A, including the I. rubilio "COI-like" sequences generated in this study. After the quality control steps, the final number of sequences included in each dataset was 102 for COI (451 bp-long alignment) and 66 for ITS (639 bp-long alignment). The difference in size between both datasets stems mostly from the fact that the COI gene (and more specifically, the Folmer region) is the marker of choice in barcoding studies, hence the higher number of sequences from this marker that are available at public repositories.
The obtained phylogenetic trees were congruent between BI and ML methods and between nuclear and mitochondrial DNA markers, with strong support for the split of Ischnura aurora in two clades: the Australo-Pacific and the Asian clade (see Figures 2 and 3). The Asian clade included the individuals of I. aurora from China that were sequenced by us, together with I. aurora specimens from Thailand and individuals identified as I. delicata and I. rubilio collected in Pakistan and India, respectively, whose sequences were obtained from GenBank. For the ITS marker, the Asian clade also included the I. rubilio specimen from Kerala (India) sequenced by us. The Australo-Pacific clade included all I. aurora individuals from Australia and Fiji sequenced by us, plus all individuals identified as I. aurora whose sequences were downloaded from GenBank, and which were collected in the Australo-Pacific distribution area of the species (i.e., Australia, Japan, Samoa, Tonga, French Polynesia, Fiji, Wallis and Futuna, Guam, and New Guinea; see Figures 2 and 3, Table A1). For the COI dataset, there were three exceptions to this pattern: the first one was an individual identified as I. aurora from Kerala in India (GenBank No KR149808) that falls within the Australo-Pacific clade (see Figure 2 and Table A1). The other exceptions corresponded with two other individuals identified as I. aurora from India (GenBank No MT511656) and New Guinea (GenBank No MH449994) which also fall outside their expected clades in the phylogenetic analyses: both individuals appear as basal to the rest of the Australo-Pacific I. aurora (see Figure 2 and Table A1). and 3, Table A1). For the COI dataset, there were three exceptions to this pattern: the first one was an individual identified as I. aurora from Kerala in India (GenBank No KR149808) that falls within the Australo-Pacific clade (see Figure 2 and Table A1). The other exceptions corresponded with two other individuals identified as I. aurora from India (GenBank No MT511656) and New Guinea (GenBank No MH449994) which also fall outside their expected clades in the phylogenetic analyses: both individuals appear as basal to the rest of the Australo-Pacific I. aurora (see Figure 2 and Table A1).

Figure 2.
Tree representing the phylogenetic relationships among the Ischnura species analysed in this study, using mitochondrial DNA (COI) sequence data. Numbers above and below branches represent Bayesian posterior probability and maximum likelihood bootstrap values, respectively. Clades are labelled according to the ASAP proposed species delimitation. The star within the Australo-Pacific clade indicates the I. aurora individual from Kerala (India) retrieved from GenBank, while the specimens in red are the two I. aurora from New Guinea and India that are identified as different taxonomic units by ASAP (see also Table A1 and Figure A7a).

Figure 2.
Tree representing the phylogenetic relationships among the Ischnura species analysed in this study, using mitochondrial DNA (COI) sequence data. Numbers above and below branches represent Bayesian posterior probability and maximum likelihood bootstrap values, respectively. Clades are labelled according to the ASAP proposed species delimitation. The star within the Australo-Pacific clade indicates the I. aurora individual from Kerala (India) retrieved from GenBank, while the specimens in red are the two I. aurora from New Guinea and India that are identified as different taxonomic units by ASAP (see also Table A1 and Figure A7a).
The ASAP species delimitation analyses were congruent with the results of the phylogenetic analyses: for both the nuclear and mitochondrial DNA datasets, partitions with the best ASAP score split the Australo-Pacific and continental Asian forms of I. aurora in two different taxonomic units (see Figure A7), the latter also including the specimens currently under the name I. rubilio and I. delicata. For the COI marker, two sequences were identified as belonging to different taxonomic units in the analysis. The first sequence was that of a specimen identified as I. aurora collected at Baliem River in New Guinea with accession no MH449994, which was identified as a different species in the partition with the best ASAP score. The second one was that of an I. aurora specimen from India with accession no MT511656, which according to the partition with the second best ASAP score would correspond to a fourth species (see Table A1 and Figure A7a).   also Table A1 and Figure A7b).
The ASAP species delimitation analyses were congruent with the results of the phylogenetic analyses: for both the nuclear and mitochondrial DNA datasets, partitions with the best ASAP score split the Australo-Pacific and continental Asian forms of I. aurora in two different taxonomic units (see Figure A7), the latter also including the specimens currently under the name I. rubilio and I. delicata. For the COI marker, two sequences were identified as belonging to different taxonomic units in the analysis. The first sequence was that of a specimen identified as I. aurora collected at Baliem River in New Guinea with accession no MH449994, which was identified as a different species in the partition with the best ASAP score. The second one was that of an I. aurora specimen from India with accession no MT511656, which according to the partition with the second best ASAP score would correspond to a fourth species (see Table A1 and Figure A7a).
The results presented above were further supported by the genetic distances: the average p-distance between Asian and Australo-Pacific I. aurora was 2% for the ITS Figure 3. Tree representing the phylogenetic relationships among the Ischnura species analysed in this study, using nuclear DNA (ITS) sequence data. Numbers above and below branches represent Bayesian posterior probability and maximum likelihood bootstrap values, respectively. Clades are labelled according to the ASAP proposed species delimitation (see also Table A1 and Figure A7b).
The results presented above were further supported by the genetic distances: the average p-distance between Asian and Australo-Pacific I. aurora was 2% for the ITS marker and 3.1% for the COI marker. Similar values were found between the specimens labelled as I. delicata and I. rubilio and the Australo-Pacific I. aurora, whereas genetic differentiation between these two species and the Asian I. aurora was nearly zero in all cases (see Table A2). For the COI marker, genetic distances between the two identified clades and the I. aurora sequences with GenBank numbers MH449994 and MT511656 from New Guinea and India, respectively, were comparable to the distances found between the two clades (~2% in all cases; see Table A2).
labelled as I. delicata and I. rubilio and the Australo-Pacific I. aurora, whereas genetic differentiation between these two species and the Asian I. aurora was nearly zero in all cases (see Table A2). For the COI marker, genetic distances between the two identified clades and the I. aurora sequences with GenBank numbers MH449994 and MT511656 from New Guinea and India, respectively, were comparable to the distances found between the two clades (~2% in all cases; see Table A2).
Thorax: Males of both I. rubilio and I. aurora from China showed the posterior dorsolateral part of the pronotum (greenish) to be less elevated but the dorsal part (black) to be more expanded towards the mesothorax compared to that of I. aurora males from Australia and Fiji ( Figure 5). The posterior dorso-lateral elevations of the pronotum of females of both I. rubilio and I. aurora from China (greenish-brownish) are connected or nearly connected by an arch of the same colour (although this character shows high interindividual variability), while both elevations are restricted to the lateral border in the I. aurora females from Australia and Fiji ( Figure 6). Additionally, in females of I. rubilio and I. aurora from China, the dorsal expansion towards the mesothorax is wider in than in females from Australia and Fiji ( Figure 6). Mesostigmal male protuberances and female plates show similarities between I. rubilio and I. aurora from all studied populations. Some Australian specimens show completely black mesostigmal protuberances, while tips are greenish in the other specimens. Females of Ischnura rubilio and I. aurora from China showed a narrower black humeral stripe and a less expanded black colouration in the junction between the metepisternum and metepimeron (see Figure 4d-f).  Legs: No differences in leg colouration were found between the examined specime Wings: Cell number was variable within and between all the studied populations Abdomen: Ischnura rubilio and I. aurora males from China show a less expanded bla Legs: No differences in leg colouration were found between the examined specimens. Wings: Cell number was variable within and between all the studied populations. Abdomen: Ischnura rubilio and I. aurora males from China show a less expanded black colouration on the dorsal part of S2 than that observed in I. aurora from China and Fiji. In the former, the apical stripe on S2 expands into a mid-dorsal stripe that reaches to approximately half of the segment, whereas in the I. aurora specimens from Australia and Fiji, this mid-dorsal stripe is much wider and may reach up to the end of S2 in some individuals (Figure 7). The basal articulation of S2 appears to be circled in black in the I. aurora individuals from Australia and Fiji, whereas it has a lighter colour in I. rubilio and I. aurora from China ( Figure 7). S3 to S6 are citron yellow in males of I. rubilio and I. aurora collected in China and yellow-orange in males of I. aurora from Australia and Fiji, except for the black intersegment areas, which appear also of a lighter colour in I. rubilio and I. aurora from China ( Figure 4). S6 is dorsally black in its posterior part and S7 is dorsally black in all examined individuals.  Ischnura rubilio and I. aurora males from China show blue colouration in the dorsal and lateral part of S8 (except its anterior dorsal part, which shows a delta-shaped black colour) and S9, and laterally in the S10 (which vary between individuals) ( Figure 7). Some individuals show blue spots in the lateral part of S7. Ischnura aurora males from Australia and Fiji show blue colour in the dorsal and lateral part of the S9 but only in the posterior dorso-lateral part of the S8 (1/3 or less of the segment length; see Figure 7). Individuals from Australia show certain variability in these colourations, showing black colour interrupting the blue bands of S8 and S9 but also showing blue colour in the lateral part of the S10. The dorsal tubercle of S10 is more pronounced in I. rubilio and I. aurora males from China than in I. aurora males from Australia and Fiji ( Figure 7). All females examined showed a vulvar spine, except for one examined female from Australia. No differences were found in the length of the female genital valves.
Annal appendages: Ischnura rubilio and I. aurora males from China show more acute dorsal expression in their cerci and narrower paraprocts than males of I. aurora from Australia and Fiji (Figure 8). I. aurora from Fiji show a marked depression in the middle of the cerci under a lateral view (albeit this character shows high variability; Figure 8k), while the cerci in the rest of the specimens showed a straight border. Ischnura rubilio and I. aurora males from China show blue colouration in the dorsal and lateral part of S8 (except its anterior dorsal part, which shows a delta-shaped black colour) and S9, and laterally in the S10 (which vary between individuals) (Figure 7). Some individuals show blue spots in the lateral part of S7. Ischnura aurora males from Australia and Fiji show blue colour in the dorsal and lateral part of the S9 but only in the posterior dorso-lateral part of the S8 (1/3 or less of the segment length; see Figure 7). Individuals from Australia show certain variability in these colourations, showing black colour interrupting the blue bands of S8 and S9 but also showing blue colour in the lateral part of the S10. The dorsal tubercle of S10 is more pronounced in I. rubilio and I. aurora males from China than in I. aurora males from Australia and Fiji (Figure 7). All females examined showed a vulvar spine, except for one examined female from Australia. No differences were found in the length of the female genital valves.
Annal appendages: Ischnura rubilio and I. aurora males from China show more acute dorsal expression in their cerci and narrower paraprocts than males of I. aurora from Australia and Fiji (Figure 8 Images (a-f) are partly reproduced from Sanmartín-Villar et al. [14] with permission from the copyright holder.

Male genital ligula:
No differences were observed between the male ligula of the analysed specimens (see Figure 9).  Images (a-f) are partly reproduced from Sanmartín-Villar et al. [14] with permission from the copyright holder.

Male genital ligula:
No differences were observed between the male ligula of the analysed specimens (see Figure 9). Images (a-f) are partly reproduced from Sanmartín-Villar et al. [14] with permission from the copyright holder.

Male genital ligula:
No differences were observed between the male ligula of the analysed specimens (see Figure 9).

Discussion
The results of our genetic analyses indicate that the samples currently identified as I. aurora found within the Asian distribution area of this species all belong to a clade that also includes specimens currently under the names I rubilio and I. delicata. This clade appears to be closely related to that comprising the I. aurora individuals from the Australo-Pacific distribution area of the species. This split into two well-differentiated clades is supported by high bootstrap and posterior probability values for both nuclear and mitochondrial DNA (see Figures 2 and 3). Genetic distances between I. rubilio, I. delicata and I. aurora from China were nearly zero in all cases, whereas genetic distances between the Asian and the Australo-Pacific clade were~3% for COI and~2% for ITS (see Table A3). These distances found between both clades are similar to those found between other closely related species within the genus Ischnura (e.g., I. elegans/I. saharensis, Table A3; I. elegans/I. graelsii [47], or between closely related species in other Coenagrionidae genera (e.g., Paracercion [48]). The species delimitation analyses have provided further support for the placement of I. aurora from Asia within the same group as I. rubilio and I delicata, separated from the Australo-Pacific I. aurora (see Figure A7).
In agreement with the results of our molecular analyses, the morphological examination of material from Asia and Oceania points also to a closer relationship between the specimens currently under the name I. aurora collected in China and I. rubilio from India, than to I. aurora from Australia and Fiji. These morphological similarities also include the characteristics described above, which consistently differentiate I. aurora from China and I. rubilio from India from I. aurora from Australia and Fiji: in males, the less elevated but more expanded posterior part of the pronotum found in I. rubilio and I. aurora from China; and in females, the connectivity between the lateral elevations and the wider expansion, differing from the same structure in I. aurora females from Australia and Fiji (see Figures 5 and 6).
Beyond these morphological differences in the shape of the pronotum, all Asian and Australo-Pacific specimens examined by us differ mostly in their colouration pattern: Ischnura rubilio shows the same colour pattern as I. aurora from China, which differs from that observed in I. aurora from Australia and Fiji. Males of I. rubilio and I. aurora from China show a more expanded blue colouration on S8-S10. The abdominal blue colouration is highly variable in some Asian Ischnura [16] and similar between species (for instance, I. praematura Sanmartín-Villar & Zhang, 2022 shows an intermediate colouration between I. aurora from China and from the Australo-Pacific forms [17]), which might represent one of the main factors hindering the identification of these species. Nevertheless, we have found that all examined specimens from I. rubilio and I. aurora from China share the same colour pattern, which is different from that of I. aurora specimens collected in Australia and Fiji. These observations agree with those of Papazian et al. [27] and Rowe [29] regarding their examination of Asian and Australo-Pacific I. aurora; and with Selys in his description of I. rubilio [25]. Other differences that we have found between the Asian and Australo-Pacific material, and which have also been pointed out by some or all of these authors, include: (i) differences in the colour pattern of S1-S2 [27]; (ii) the paler basal articulations of the abdominal segments found in both I. rubilio and I. aurora from China when compared to Australo-Pacific I. aurora [25]; (iii) a more prominent dorsal tubercle in S10 in I. rubilio and I. aurora from China than in I. aurora from Australia and Fiji [29]; and (iv) the differences found in the anal appendages between the Asian and Australo-Pacific material that we have examined [29].
Contrary to what has been previously reported, we did not find differences in the postocular spots between I. aurora and I. rubilio. Papazian et al. [27] suggested that I. rubilio showed larger postocular spots and pointed out that they were "sometimes confluent across the occiput" [27] (p. 60). Even though we have observed that the occipital greenbrownish colour in some females of I. aurora from China reaches the postocular spots, we highlight the different colouration of the postocular spots (blue) and the occipital region (green-brownish) and hence, no confluence between the postocular spots in the examined specimens. Considering the high variability that this character shows in Ischnura [49], including the species we examined here; and given that we found no consistent differences across the examined specimens, we conclude that postocular spots cannot be used as a trait to distinguish between I. rubilio and I. aurora.
Our morphological analysis of Asian and Australo-Pacific material allowed us to find some morphological differences in the females, which have not been addressed previously. Thus, we found differences in the thoracic colouration pattern (the black colour is less expanded in females of I. rubilio and I. aurora from China; compared to females of I. aurora from Australia and Fiji; Figure 4), and morphological differences in the posterior part of the pronotum that consistently separate females of I. rubilio and I. aurora from China from Australo-Pacific I. aurora females (see Figure 6).
Overall, the results of our analyses support the split of the Asian and Australo-Pacific forms of I. aurora into two well-differentiated taxonomic units and therefore different species, further supporting the specific status of I. rubilio [28]. The results of our genetic analyses agree with those of Sánchez-Guillén et al. [18] in placing I. rubilio within the aurora clade, but contradict those by Dumont [23], who found I. rubilio to be either basal to the Ischnura group or a member of the pumilio clade. The quality control of the DNA sequence data carried out by us here led to the exclusion of several questionable sequences from our datasets, among which were those belonging to I. rubilio from Dumont's work. These sequences are likely the product of both specimen misplacement during laboratory work and sequencing of non-orthologous copies of the mitochondrial COI gene (see Table 2), which led to the wrong conclusion about the phylogenetic position of this species within the Ischnura clade [23]. Similarly questionable, "COI-like" sequences were recently identified in Odonata [13], and therefore we encourage research colleagues to carefully examine the sequences they use in their studies in a similar way as we have done here or has been described elsewhere e.g., [50] to avoid the potential amplification of numts, which might have important consequences when addressing taxonomical questions [11,13,31].
Although wind-driven dispersal could explain the exchange of individuals between Asia and Oceania (for instance, the individual from Kerala, India with accession no KR149808 that falls within the Australo-Pacific clade; see Figure 2), long-distance dispersal may at the same time be limited by the physical barrier constituted by the sea, which will in turn restrict the contact between the identified clades and lead to the observed genetic differentiation. The observed morphological differences in the shape of the pronotum of both males and females, along with the differences in the S10 dorsal tubercle and male anal appendages found between both Asian and Australo-Pacific species might not be enough to prevent successful hybridization, which could in turn result in a morphological and/or molecular cline along the species' contact zone, with individuals showing intermediate characters between I. rubilio and I. aurora. There are several examples of hybridization between closely related Ischnura species [51,52]. However, as it is assumed that reproduction occurs between mature individuals in I. rubilio and I. aurora Asian forms but between mature males and teneral females in I. aurora from the Australo-Pacific forms [29,53], even if individuals from both species may occasionally come into contact, mating would be restricted due to these behavioural incompatibilities. Without access to the actual specimen from which the DNA sequence was obtained, we cannot establish whether the individual from Kerala that falls within the Australo-Pacific clade is a migrant individual or a hybrid.
The recent discovery of other Asian species in which teneral females mate [16] stresses the need for additional ethological studies to unravel species barriers and species population dynamics. Additionally, the presence of the elaborated mesostigmal protuberances in males but not in females suggests that these do not play a role in male-female assembly (for instance as happens in the European Ischnura species; see [54]) but in male-male encounters, which poses an interesting topic for future studies about the function of such structures.
Finally, the results of our genetic analyses point also to the existence of at least a third (or even a fourth) taxonomic unit identified from our COI dataset. These correspond to a specimen from India and another specimen from Baliem River in New Guinea, both labelled as I. aurora. Interestingly, Baliem River is the same locality in which the subspecies I. aurora viduata was first described [55]. These results stress the need to revise all available material belonging to the numerous I. aurora subspecies described as well as to be able to correctly link available DNA sequence data with voucher specimens.
We provide below an identification key for the separation between I. rubilio and I. aurora:  (Figure 5a-d); paraproct in ventral view with an accessory medial lobe in addition to the basal one; height of S10 tubercle in posterior view subequal to the width separating each tubercle (Figure 8c,f); black apical stripe in S2 expanding into a mid-dorsal stripe that reaches approximately half of the segment (Figure 7a,b); S8 entirely blue dorsally (Figure 8a

•
Genetic analyses have showed that specimens currently under the names of Ischnura rubilio and I. delicata belong to a clade that also includes the I. aurora found within the Asian distribution area of this species. • All the I. aurora found within the Australo-Pacific distribution area cluster together in a separate clade. Species delimitation analyses have identified these two clades as different taxonomic units.

•
Concordant with the results of the genetic analyses, the morphology of the I. aurora collected in China is closer to I. rubilio than to I. aurora from Australia and Fiji. • Given these results, we confirm the status of I. rubilio as a valid species and provide an identification key for its separation from I. aurora. • Genetic analyses point also to the existence of at least a third taxonomic unit within the aurora clade, which stress the need to revise all available material belonging to the numerous subspecies of I. aurora that have been described. Funding: This work was funded by grants CGL2014-53140-P and PGC2018-096656-B-I00 to ACR from MCIN/AEI/10.13039/501100011033 and from "ERDF A way of making Europe", by the "European Union".

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: DNA sequences generated in this study have all been submitted to GenBank (https://www.ncbi.nlm.nih.gov/genbank/). Additional data and methodology are available in Appendices A and B. All the Ischnura specimens examined for morphological analyses and/or used for genetics are deposited in the collections of Adolfo Cordero-Rivera and Matjas Bedjanič.

Acknowledgments:
We thank Matjas Bedjanič who kindly provided us access to specimens from his personal collection. We also thank Rosser W. Garrison, who helped us with the diagnostic key included in the main text and provided valuable comments that helped us to improve our manuscript.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A. Quality Control (QC) and Curation of GenBank Sequence Data
All COI and ITS sequences deposited in GenBank that were labelled as belonging to the species Ischnura rubilio, Ischnura aurora and Ischnura delicata were downloaded from the GenBank nucleotide database (https://www.ncbi.nlm.nih.gov/genbank/), through Geneious v. 9.1.8 (https://www.geneious.com). These were added to the sequences obtained by us from samples of I. aurora from China, Australia and Fiji and samples of I. rubilio from India, as described in the Materials and Methods section of the manuscript. Additionally, sequences belonging to different species within the genus Ischnura and other closely related genera (Aciagrion, Ceriagrion, Coenagrion, Enallagma, Erythromma, Mortonagrion and Pseudagrion) were also downloaded from GenBank and added to each dataset. Sequences were aligned in MAFFT [32] as implemented in Geneious v. 9.1.8, with a gap open penalty of 3. Alignments were visually inspected, and tails trimmed manually before phylogenetic tree reconstruction. Phylogenetic relationships were reconstructed through maximum likelihood (ML) using FastTree 2.1.11 [56] also implemented in Geneious v. 9.1.8, with the following options: use of GTR model and optimization of the 20 Gamma Likelihood.

Appendix A.1. ITS Dataset
A total of 69 sequences were included in the preliminary ITS dataset. After inspection of the alignment, the sequence with accession number MZ809355, belonging to an Ischnura rubilio specimen from India, was excluded from further analyses because it corresponded with a partial region of the 18S ribosomal RNA gene, located upstream the region that we sequenced for our study. The tree obtained after removing this sequence and trimming the alignment tails is showed in Figure A1 below. Two of the sequences included in the dataset were placed in the tree far from their expected clades. The first one was the Ischnura aurora sequence with GenBank accession number FN356100, which was placed outside of the aurora clade. Instead, this sequence appears to be closely related to I. nursei within the clade that includes I. elegans, I. senegalensis, I. heterosticta, I. evansi, I. saharensis and I. abyssinica (see Figure A1). The second sequence was that of Ischnura rubilio with GenBank accession number MH447434, which falls outside of the Ischnura clade, and it is placed as a sister species to Aciagrion migratum (see Figure A1). Both sequences were used as queries in a BLAST search against the nr database. Searches were carried out using the MegaBLAST program implemented in Geneious v. 9.1.8 with default options.
The sequence FN356100 resulted in very similar matches with other Ischnura ITS sequences in the nr database, but none with I. aurora. The sequence MH447434 resulted in matches with Aciagrion migratum, Proischnura subfurcata and several Ischnura species, but none with I. aurora. The results of the phylogenetic analysis and the BLAST searches suggest that the specimens from which these sequences were obtained could have been misidentified and/or misplaced at the time of DNA extraction, and hence these sequences were removed from our datasets and excluded from the final analyses.
placed as a sister species to Aciagrion migratum (see Figure A1). Both sequences were used as queries in a BLAST search against the nr database. Searches were carried out using the MegaBLAST program implemented in Geneious v. 9.1.8 with default options.
The sequence FN356100 resulted in very similar matches with other Ischnura ITS sequences in the nr database, but none with I. aurora. The sequence MH447434 resulted in matches with Aciagrion migratum, Proischnura subfurcata and several Ischnura species, but none with I. aurora. The results of the phylogenetic analysis and the BLAST searches suggest that the specimens from which these sequences were obtained could have been misidentified and/or misplaced at the time of DNA extraction, and hence these sequences were removed from our datasets and excluded from the final analyses. sequences in the alignment sequences appeared to be quite dissimilar to the rest of the sequences in the dataset, and which introduced a 6 bp-long gap in the alignment ( Figure  A2a). These were sequences with accession numbers EU219876, EU219877 and LC198680; all of them were labelled as Ischnura aurora. Translation into protein yielded nearly identical sequences without stop codons ( Figure A2b). BLAST searches were carried out also with MegaBLAST as described above using these three sequences as queries. The searches resulted in nearly the same hits for sequences EU219876 and EU219877, including several Ceriagrion and Ischnura species. The sequence LC198680 yielded matches with Ceriagrion and Ischnura species but also with Anisoptera genera such as Anax and Onychogomphus. Several of the hits obtained in these searches corresponded with sequences of Ischnura spp. all belonging to the same study.
The tree obtained from FastTree analysis is shown below. It can be seen how these three sequences are placed together in the same clade, at the end of a very long branch and far from the rest of the sequences included in the analysis (see Figure A3). BLAST searches were carried out also with MegaBLAST as described above using these three sequences as queries. The searches resulted in nearly the same hits for sequences EU219876 and EU219877, including several Ceriagrion and Ischnura species. The sequence LC198680 yielded matches with Ceriagrion and Ischnura species but also with Anisoptera genera such as Anax and Onychogomphus. Several of the hits obtained in these searches corresponded with sequences of Ischnura spp. all belonging to the same study.
The tree obtained from FastTree analysis is shown below. It can be seen how these three sequences are placed together in the same clade, at the end of a very long branch and far from the rest of the sequences included in the analysis (see Figure A3).
The alignment of these sequences with reference COI sequences from published complete mitochondrial genomes of Ischnura species (Ischnura elegans (MK951668), I. senegalensis (MT787567) and I. pumilio (KC878732)) revealed that they corresponded to the 3 region of the COI gene, located downstream from the Folmer region that we amplified in our study ( Figure A4). Therefore, these sequences were excluded from further analyses.
The COI sequence from Ischnura rubilio with accession number MH449992 falls within the Aciagrion clade in the FastTree ML tree (see Figure A3), similarly to the results obtained for the ITS sequence from this same specimen (accession number MH447434). The results of a MegaBLAST search using sequence MH449992 as a query resulted in hits with Aciagrion, Ischnura pumilio or Ischnura elegans, among others, which further supports our conclusion that this could be a case of misidentification and/or misplacement of individuals in the laboratory. Hence, this COI sequence was also removed from our dataset.
Two sequences downloaded from GenBank with accession numbers MH449981 and MW143324 labelled as Ischnura rubilio appear together in the ML tree, in a clade basal to Ischnura pumilio (see Figure A3). Both sequences can be translated into an amino acid sequence without stop codons, and both protein sequences were identical. MegaBLAST searches carried out using these two sequences as queries yielded hits with several genera of marine sponges (Characella, Poecillastra and Theonella among others). The sequences obtained by us from the Ischnura rubilio specimens from India (MB-IrbKeF, MB-IrbKeM, MB-IrbGir and MB-IrbTam; see Table 1) appear in the tree together in a clade separated from the rest of the sequences included in the analysis (see Figure A3). Translation of these sequences into the protein yielded nearly identical amino acid sequences without stop codons, and BLAST searches using these four sequences as queries returned hits against other Odonata sequences in GenBank, but the closest matches were all sequences belonging to Nesobasis spp., a genus in the Coenagrionidae family which is endemic to the Fiji archipelago. Diversity 2022, 14, x FOR PEER REVIEW 21 of 32 Figure A3. Maximum likelihood tree obtained with FastTree 2.1.11 for the COI dataset (117 sequences; 457 bp-long). The tree is rooted by the midpoint. Highlighted in red are the ten sequences downloaded from GenBank and the four sequences obtained in this study that were finally excluded from our dataset. The coloured dots on the nodes represent FastTree support values according to the legend on the left.
The alignment of these sequences with reference COI sequences from published complete mitochondrial genomes of Ischnura species (Ischnura elegans (MK951668), I. senegalensis (MT787567) and I. pumilio (KC878732)) revealed that they corresponded to the Figure A3. Maximum likelihood tree obtained with FastTree 2.1.11 for the COI dataset (117 sequences; 457 bp-long). The tree is rooted by the midpoint. Highlighted in red are the ten sequences downloaded from GenBank and the four sequences obtained in this study that were finally excluded from our dataset. The coloured dots on the nodes represent FastTree support values according to the legend on the left. 3′ region of the COI gene, located downstream from the Folmer region that we amplified in our study ( Figure A4). Therefore, these sequences were excluded from further analyses. Figure A4. Alignment of sequences EU219876, EU219877 and LC198680 (box a) against reference COI sequences from Ischnura species (first three sequences in the alignment). Box b shows sequences that correspond with the Folmer region used in barcode analysis, which was the target region of the primers used in our study.
The COI sequence from Ischnura rubilio with accession number MH449992 falls within the Aciagrion clade in the FastTree ML tree (see Figure A3), similarly to the results obtained for the ITS sequence from this same specimen (accession number MH447434). The results of a MegaBLAST search using sequence MH449992 as a query resulted in hits with Aciagrion, Ischnura pumilio or Ischnura elegans, among others, which further supports our conclusion that this could be a case of misidentification and/or misplacement of individuals in the laboratory. Hence, this COI sequence was also removed from our dataset.
Two sequences downloaded from GenBank with accession numbers MH449981 and MW143324 labelled as Ischnura rubilio appear together in the ML tree, in a clade basal to Ischnura pumilio (see Figure A3). Both sequences can be translated into an amino acid sequence without stop codons, and both protein sequences were identical. MegaBLAST searches carried out using these two sequences as queries yielded hits with several genera of marine sponges (Characella, Poecillastra and Theonella among others). The sequences obtained by us from the Ischnura rubilio specimens from India (MB-IrbKeF, MB-IrbKeM, MB-IrbGir and MB-IrbTam; see Table 1) appear in the tree together in a clade separated from the rest of the sequences included in the analysis (see Figure A3). Translation of these sequences into the protein yielded nearly identical amino acid sequences without stop codons, and BLAST searches using these four sequences as queries returned hits against other Odonata sequences in GenBank, but the closest matches were all sequences belonging to Nesobasis spp., a genus in the Coenagrionidae family which is endemic to the Fiji archipelago.
Even though the results of the BLAST searches for both sets of I. rubilio sequences shown mentioned above might initially lead to the conclusion that they could all be cases of sample contamination; this would in fact be quite unlikely. Regarding the COI sequences obtained by us from the I. rubilio specimens from India, none of the Nesobasis spp. sequences to which they are similar were obtained in our laboratory. Furthermore, the ITS sequence obtained from the individual MB-IrbKeM falls within the clade of I. aurora (see Figure A1), further supporting the fact that this is unlikely a case of contamination. In the case of sequences MH449981 and MW143324, these were produced by two independent laboratories/researchers, even at different times. Sequence MH449981 corresponds to the I. rubilio specimen from Kerala (India) included in the work by Dumont [23] and which was submitted to GenBank in 2018. Sequence MW143324 was submitted to GenBank in 2020 and belongs to an unpublished study on Odonata in rice ecosystems of India. Overall, the chances of getting identical (or very similar) sequences because of contamination under these circumstances are very low. Hence, we opted for a more indepth analysis of these sequences with the aim to determine whether they should be Figure A4. Alignment of sequences EU219876, EU219877 and LC198680 (box a) against reference COI sequences from Ischnura species (first three sequences in the alignment). Box b shows sequences that correspond with the Folmer region used in barcode analysis, which was the target region of the primers used in our study.
Even though the results of the BLAST searches for both sets of I. rubilio sequences shown mentioned above might initially lead to the conclusion that they could all be cases of sample contamination; this would in fact be quite unlikely. Regarding the COI sequences obtained by us from the I. rubilio specimens from India, none of the Nesobasis spp. sequences to which they are similar were obtained in our laboratory. Furthermore, the ITS sequence obtained from the individual MB-IrbKeM falls within the clade of I. aurora (see Figure A1), further supporting the fact that this is unlikely a case of contamination. In the case of sequences MH449981 and MW143324, these were produced by two independent laboratories/researchers, even at different times. Sequence MH449981 corresponds to the I. rubilio specimen from Kerala (India) included in the work by Dumont [23] and which was submitted to GenBank in 2018. Sequence MW143324 was submitted to GenBank in 2020 and belongs to an unpublished study on Odonata in rice ecosystems of India. Overall, the chances of getting identical (or very similar) sequences because of contamination under these circumstances are very low. Hence, we opted for a more in-depth analysis of these sequences with the aim to determine whether they should be flagged as "COI-like" sequences or nuclear mitochondrial copies (numts), rather than orthologous copies of the mitochondrial COI [11,31]. In fact, numts were recently identified in odonates of the genera Leucorrhinia and Calopteryx [13].
Examination of the chromatograms corresponding to the I. rubilio specimens sequenced in this study revealed the presence of some double peaks, which the program interpreted as heterozygous positions, and hence, ambiguity-coded bases in the final sequences. Additionally, the chromatograms of all these individuals showed some messy regions, which were difficult to read, and other regions that were more readable (see Figure A5), which is usually the result of the co-amplification of both numts and orthologous copies of the COI [29]. This, together with the results of the BLAST searches, suggests that these could be non-orthologous copies of the COI gene and, therefore, we labelled these sequences as "COI-like" for submission to GenBank, and excluded them from further analyses.
An examination of codon usage comparing the odd COI sequences of I. rubilio from GenBank (MH449981 and MW143324) with the apparently legitimate COI sequences of I. aurora (MT680688, MG517558 and MH449988) and the same COI region extracted from published complete mitochondrial genomes of other Ischnura species (I. elegans (KU958378 and MK951668), I. pumilio (KC878752) and I. senegalensis (MT787567)) is shown in Figure A6.
Amino acid composition within the selected COI region is identical among the reference Ischnura spp. included in the analysis (I. pumilio, I. elegans and I. senegalensis), and very similar to that of I. aurora (see Figure A6), whereas the two questionable sequences of I. rubilio show some differences in the frequency of some amino acids (e.g., F, G, I, M or S; see Figure A6). These differences in amino acid composition are not expected in a gene such as COI, even less in the Folmer region sequenced here, which is highly conserved across species and even across genera [11]. Even though the two I. rubilio sequences from GenBank showed no stop codons, the observed differences in amino acid composition between them and other Ischnura legitimate COI sequences, coupled with the fact that they appear at the end of a long branch (see Figure A3), all constitute "red flags" that make us question these as true mitochondrial COI sequences of I. rubilio, and therefore we removed them from our dataset. orthologous copies of the mitochondrial COI [11,31]. In fact, numts were recently identified in odonates of the genera Leucorrhinia and Calopteryx [13].
Examination of the chromatograms corresponding to the I. rubilio specimens sequenced in this study revealed the presence of some double peaks, which the program interpreted as heterozygous positions, and hence, ambiguity-coded bases in the final sequences. Additionally, the chromatograms of all these individuals showed some messy regions, which were difficult to read, and other regions that were more readable (see Figure A5), which is usually the result of the co-amplification of both numts and orthologous copies of the COI [29]. This, together with the results of the BLAST searches, suggests that these could be non-orthologous copies of the COI gene and, therefore, we labelled these sequences as "COI-like" for submission to GenBank, and excluded them from further analyses. Figure A5. Screenshot of sections of two chromatograms of the Ischnura rubilio COI sequences obtained in this study, showing the messy but in some places readable chromatograms. The red arrows point to the double peaks which were interpreted as ambiguity-coded bases by the software.
An examination of codon usage comparing the odd COI sequences of I. rubilio from GenBank (MH449981 and MW143324) with the apparently legitimate COI sequences of I. aurora (MT680688, MG517558 and MH449988) and the same COI region extracted from published complete mitochondrial genomes of other Ischnura species (I. elegans (KU958378 and MK951668), I. pumilio (KC878752) and I. senegalensis (MT787567)) is shown in Figure  A6. Amino acid composition within the selected COI region is identical among the reference Ischnura spp. included in the analysis (I. pumilio, I. elegans and I. senegalensis), and very similar to that of I. aurora (see Figure A6), whereas the two questionable sequences of I. rubilio show some differences in the frequency of some amino acids (e.g., F, G, I, M or S; see Figure A6). These differences in amino acid composition are not expected in a gene such as COI, even less in the Folmer region sequenced here, which is highly conserved across species and even across genera [11]. Even though the two I. rubilio sequences from GenBank showed no stop codons, the observed differences in amino acid composition between them and other Ischnura legitimate COI sequences, coupled with the fact that they appear at the end of a long branch (see Figure A3), all constitute "red flags" Figure A6. Codon usage analysis for the questionable COI sequences of Ischnura rubilio downloaded from GenBank (MH449981 and MW143324; black dotted line with circles) and possibly true COI sequences of Ischnura aurora (orange line with triangles) and reference COI sequences extracted from published complete mitogenomes of several Ischnura species (blue line with triangles). The codon abbreviations for amino acids on the X axis are standard. The table below the X axis shows the frequency of each amino acid for each set of sequences included in the analysis. Table A1. Accession numbers for the COI and ITS sequences selected for genetic analyses after quality control as described in Appendix A. For Ischnura aurora, I. delicata and I. rubilio, we list the information on the clade to which each individual belongs according to the genetic analyses. Highlighted in bold are the individuals that fall outside their expected clades according to the genetic and ASAP analyses (see main text and Figure A7).

Species
Clade Collection Locality Data GenBank Acc. Nos   Figure A7. Graphical output of ASAP, showing the results of the species delimitation analysis using the Ischnura aurora; I. delicata and I. rubilio COI sequences and (b) the I. aurora and I. rubilio ITS sequences. Each node of the dendrogram is coloured depending on its probability of being a panmictic species: darker colours for nodes that may be split into smaller groups. The tables show the best partitions found by ASAP in each case, scored and sorted by their p-value (the smallest pvalue has rank 1) and their rank of relative barcode gap width (the largest gap has rank 1). The asapscore is the average of both ranks: the smaller the asap-score, the better. The predicted number of groups in each partition, together with their asap-score, are also indicated in the dendrograms, with the red star indicating the best partition according to the analysis in each case. Australo-Pacific and Asian clades are highlighted with the same colours used in Figure 2 in the main text. In (a), the individual highlighted in red inside the Australo-Pacific clade corresponds to a specimen of I. aurora from India. The individuals highlighted in red at the bottom of the dendrogram correspond to the Figure A7. Graphical output of ASAP, showing the results of the species delimitation analysis using the Ischnura aurora; I. delicata and I. rubilio COI sequences and (b) the I. aurora and I. rubilio ITS sequences. Each node of the dendrogram is coloured depending on its probability of being a panmictic species: darker colours for nodes that may be split into smaller groups. The tables show the best partitions found by ASAP in each case, scored and sorted by their p-value (the smallest p-value has rank 1) and their rank of relative barcode gap width (the largest gap has rank 1). The asap-score is the average of both ranks: the smaller the asap-score, the better. The predicted number of groups in each partition, together with their asap-score, are also indicated in the dendrograms, with the red star indicating the best partition according to the analysis in each case. Australo-Pacific and Asian clades are highlighted with the same colours used in Figure 2 in the main text. In (a), the individual highlighted in red inside the Australo-Pacific clade corresponds to a specimen of I. aurora from India. The individuals highlighted in red at the bottom of the dendrogram correspond to the specimens of I. aurora from India and New Guinea with COI sequences that are~2% divergent from the rest of the sequences included in the analyses and identified as belonging to a different species in the ASAP analysis.