Local-Scale DNA Barcoding of Afrotropical Hoverflies (Diptera: Syrphidae): A Case Study of the Eastern Free State of South Africa

Simple Summary Hoverflies are regarded as the second most important pollinators after bees. They also provide important environmental services including the biodegradation of organic wastes, as well as the predation of pests. Hoverflies are usually divided into several groups or regions including the Holarctic, the Oriental, the Australasian and the Afrotropical. The latter is considered one of the most diverse groups but is still poorly studied due to the unavailability of complete and detailed identification keys for numerous genera and/or species. Published taxonomy studies on hoverflies in South Africa were published in the 1980s. This study aimed to investigate the barcoding of hoverfly species found in the Free State province of South Africa in order to ascertain their taxonomy and establish their genetic richness and differentiation. From 78 specimens of hoverflies sampled in the eastern Free State of South Africa, DNA barcodes helped to confirm the taxonomy of 15 hoverfly species from nine genera. With the barcodes generated in this study, the identification of Afrotropical species can be improved, but about 40% of the known species cannot be identified using the available identification keys. Abstract The Afrotropical hoverflies remain an understudied group of hoverflies. One of the reasons for the lack of studies on this group resides in the difficulties to delimit the species using the available identification keys. DNA barcoding has been found useful in such cases of taxonomical uncertainty. Here, we present a molecular study of hoverfly species from the eastern Free State of South Africa using the mitochondrial cytochrome-c oxidase subunit I gene (COI). The identification of 78 specimens was achieved through three analytical approaches: genetic distances analysis, species delimitation models and phylogenetic reconstructions. In this study, 15 nominal species from nine genera were recorded. Of these species, five had not been previously reported to occur in South Africa, namely, Betasyrphus inflaticornis Bezzi, 1915, Mesembrius strigilatus Bezzi, 1912, Eristalinus tabanoides Jaennicke, 1876, Eristalinus vicarians Bezzi, 1915 and Eristalinus fuscicornis Karsch, 1887. Intra- and interspecific variations were found and were congruent between neighbour-joining and maximum likelihood analyses, except for the genus Allograpta Osten Sacken, 1875, where identification seemed problematic, with a relatively high (1.56%) intraspecific LogDet distance observed in Allograpta nasuta Macquart, 1842. Within the 78 specimens analysed, the assembled species by automatic partitioning (ASAP) estimated the presence of 14–17 species, while the Poisson tree processes based on the MPTP and SPTP models estimated 15 and 16 species. The three models showed similar results (10 species) for the Eristalinae subfamily, while for the Syrphinae subfamily, 5 and 6 species were suggested through MPTP and SPTP, respectively. Our results highlight the necessity of using different species delimitation models in DNA barcoding for species diagnoses.


Introduction
More than 6000 species of hoverflies have been recorded worldwide, with approximately 600 species from the Afrotropical region distributed into three subfamilies, namely, Eristalinae, Microdontinae and Syrphinae [1][2][3].Adult hoverflies (Syrphidae family) are often regarded as one of the most important pollinator groups of the Diptera order after bees [4][5][6].This important ecosystem service that they provide in both natural and agricultural land uses makes hoverflies a key model group for impact surveys of agricultural and horticultural processes as well as pest management [5,[7][8][9].
One critical issue for these applications is that the understanding of the response of hoverflies to landscape management in relation to their diversity can be challenged by the limitation of the usable identification keys for Afrotropical species [2].This has been particularly highlighted in the genus Eumerus Meigen, 1822, which has been shown to be too vague for species identification through the routine use of morphological taxonomy tools [10].Most of these Afrotropical species were described centuries ago, and their morphological diagnoses have been used in various studies without thorough revisions to encompass as much species diversity as found within the region [11,12].Although morphological diagnoses have their limitations, the method is very useful when used with molecular techniques [13].This indicates the need to use an informative and effective molecular technique to assist in morphological identification as a prerequisite for efficient conservation [13].
DNA barcoding [14] is currently the most used molecular taxonomy tool that provides the possibility of distinguishing animal species by sequencing a fragment of the 5 end of the COI (cytochrome c oxidase subunit 1) mitochondrial gene.In addition to the detection of sequence variation within and between taxa, the use of this approach holds many other advantages in the study of Afrotropical hoverflies.It is recommended for those taxa that deserve further taxonomic investigations due to ambiguous morphology [15].It links the unknown sex to species for which only one of the sexes is known, thus reducing bias in species-level data [15].Moreover, DNA barcoding can be used to relate the developmental stages of species for which the reproductive ecology is poorly understood [15].Afrotropical hoverflies constitute a large and diverse group [6,16], but few studies have focused on the genetic diversity of this group [11,17,18].It is estimated that a major part of the existing species of Afrotropical hoverflies has not been described to date [2,[16][17][18][19][20].The occurrence of cryptic species may also complicate species identification.In their study, Jordaens et al. [17] reported the possibility of the occurrence of cryptic species in Afrotropical hoverflies and observed that differentiating intra-from interspecific divergence among subfamilies using distance analyses was better defined at the genus level.That study remains the most important contribution to the study of Afrotropical hoverflies both from a morphological and a DNA barcoding perspective.More recently, many researchers have made substantial contributions to species delimitations and phylogenetic inferences in the Merodon Meigen, 1803 [21,22], Sphaerophoria Le Peletier & Serville, 1828 [20] and Eumerus [18,19] genera of Afrotropical hoverflies.
In South Africa, hoverfly species have been studied based on their external morphology, and the country harbors species from three out of the four hoverfly subfamilies, namely, Syrphinae, Microdontinae and Eristalinae [23].The need for molecular studies on these beneficial flies is critical to conclusively establish species diversity and genetic differentiation between subfamilies.This study aimed to conduct a DNA barcoding investigation of hoverfly species from selected areas of the eastern Free State of South Africa in order to ascertain their taxonomy and establish their genetic richness and differentiation.

Field Sampling, DNA Extraction and Sequencing
Specimens of hoverflies were collected using entomological nets.The sampling ensured the inclusion of as much species diversity as possible from the study area, which encompassed the towns of Harrismith (28   49 00 E), the Golden Gate Highlands National Park (28 • 30 22 S 28 • 37 0 E) and the farming lands near the village of Verkykerskop (27.918 • S 29.281 • E) in the Maluti-A-Phufong and Phumelela municipalities (Figure 1).Collected specimens were sorted into morphotypes using the descriptions and illustrations in Whittington [2], Jordaens et al. [17] and Scholtz and Holm [23], preserved in 99% ethanol and deposited in the museum of the Department of Zoology and Entomology of the University of the Free State-QwaQwa Campus.

Field Sampling, DNA Extraction and Sequencing
Specimens of hoverflies were collected using entomological nets.The sampling ensured the inclusion of as much species diversity as possible from the study area, which encompassed the towns of Harrismith (28°17′0″ S 29°08′0″ E) and Phuthaditjhaba (28°32′00″ S 28°49′00″ E), the Golden Gate Highlands National Park (28°30′22″ S 28°37′0″ E) and the farming lands near the village of Verkykerskop (27.918° S 29.281° E) in the Maluti-A-Phufong and Phumelela municipalities (Figure 1).Collected specimens were sorted into morphotypes using the descriptions and illustrations in Whittington [2], Jordaens et al. [17] and Scholtz and Holm [23], preserved in 99% ethanol and deposited in the museum of the Department of Zoology and Entomology of the University of the Free State-QwaQwa Campus.Total genomic DNA was extracted from two legs of each individual following the method described by Mengual et al. [24].This was done following the instructions of the manufacturer of the NucleoSpin ® Tissue DNA Extraction kit (Macherey-Nagel, Düren, Germany).Tissue samples were pre-lysed in the presence of Proteinase K at 37 °C for 10 h with vortexing every 1 h.Thereafter, buffer B3 from the NucleoSpin ® Tissue DNA Extraction kit was added to the lysed samples and incubated at 70 °C for 10 min.DNA was eluted using 100 µ L buffer BE at room temperature.The eluted DNA was stored at −35 °C until PCR could be performed.A negative control (blank) without tissue was included in each set of DNA extractions.A total of 108 specimens were processed for DNA extraction.
Each reaction tube included the following components: 1-3 µ L DNA extract, 1 µ L of each primer (with an initial concentration of 10 μM), 9.5 μL nuclease-free water and 12.5 μL OneTaq ® 2X Master Mix (Inqaba Biotechnical Industries, Pretoria, Gauteng, South Africa) with Standard Buffer.OneTaq ® 2X Master Mix is a ready-to-use PCR master mix that also contains agarose gel loading buffer and tracking dyes.The universal primers were used to amplify 574 bp of the COI gene (Table 1).Amplifications were performed in 0.2 Total genomic DNA was extracted from two legs of each individual following the method described by Mengual et al. [24].This was done following the instructions of the manufacturer of the NucleoSpin ® Tissue DNA Extraction kit (Macherey-Nagel, Düren, Germany).Tissue samples were pre-lysed in the presence of Proteinase K at 37 • C for 10 h with vortexing every 1 h.Thereafter, buffer B3 from the NucleoSpin ® Tissue DNA Extraction kit was added to the lysed samples and incubated at 70 • C for 10 min.DNA was eluted using 100 µL buffer BE at room temperature.The eluted DNA was stored at −35 • C until PCR could be performed.A negative control (blank) without tissue was included in each set of DNA extractions.A total of 108 specimens were processed for DNA extraction.
Each reaction tube included the following components: 1-3 µL DNA extract, 1 µL of each primer (with an initial concentration of 10 µM), 9.5 µL nuclease-free water and 12.5 µL OneTaq ® 2X Master Mix (Inqaba Biotechnical Industries, Pretoria, Gauteng, South Africa) with Standard Buffer.OneTaq ® 2X Master Mix is a ready-to-use PCR master mix that also contains agarose gel loading buffer and tracking dyes.The universal primers were used to amplify 574 bp of the COI gene (Table 1).Amplifications were performed in 0.2 mL thin-walled PCR tubes using the Multigene Optimax thermocycler (Labnet International, Inc, New York, NY, USA).PCR conditions included an initial denaturation at 95 • C for 2 min, followed by 29 cycles for 30 s, denaturation at 94 • C for 30 s, annealing at 49 • C for 2 min and extension at 72 • C, followed by a final extension for 8 min at 72 • C [24].Negative controls consisting of the PCR reaction mix excluding the DNA extract were also used to control for DNA contamination.Successful amplification was checked based on size and products (bands) by running the PCR product (2 µL) on an agarose gel (0.5 g SeaKem ® LE Agarose (Lonza Rockland, Inc., Rockland, ME, USA), Lonza in 50 mL TAE buffer, 1.5% w/v) stained with 2 µL ethidium bromide.Specifically, 2 µL of the PCR product was loaded and run for 30 min at 100 V and 350 Ma.The gel was visualized using the Gene genius Bioimaging System and SynGene software (Gensnap v6.00.22.1., Synoptics, Ltd., (Synoptics, Ltd., Cambridge, UK).The samples were then shipped to Inqaba biotec TM (Hatfield, Pretoria, South Africa) for one-directional DNA sequencing with the LCO1490 primer.Sequencing reactions were performed using the ABI v3.1 BigDye ® kit (Applied Biosystems, Waltham, MA, USA).Purified sequences were run on an ABI 3500 XL Genetic analyzer, POP7TM (ThermoScientific, Waltham, MA, USA).

Sequencing Analysis
The sequences (574 bp) were screened, cleaned and edited using SeqMan Ultra (DNAS-TAR Inc., Madison, WI, USA).After low-quality sequences were removed, 78 were retained out of the 108 sequences generated.These sequences were deposited in Genbank (see the Table A1 in Appendix A for the accession numbers).They were tentatively identified using the BOLD (Barcode of Life Data Systems) identification system and compared to published COI sequences of Afrotropical hoverflies deposited in Genbank by Jordaens et al. [17] and other sources.All sequences were queried through Blast search (BLAST v.2.9) using blastn in the nr/nt database in NCBI's Genbank.Sequences were translated to protein using the invertebrate mitochondrial code in MEGA X [26].The sequences were aligned in MUSCLE and trimmed in length in order to ensure equal alignment lengths.Sequences harbouring gaps marked as (−) were taken into account using the pairwise deletion option in MEGA X [26].The final alignments were improved manually and saved in Fasta and Newick formats.Three analytical approaches were used to analyse the sequences.

Distance Analyses and Genetic Diversity
Pairwise distances between nucleotide sequences were calculated as absolute distances [27] using the Kimura 2-parameter model and as LogDet transformations since base composition differed between the sequences [28].Intraspecific variation and interspecific divergence were assessed in species that are represented by at least two conspecific reference sequences in the dataset [29].In addition, DNA polymorphism parameters such as the number of haplotypes (h), haplotype diversity (Hd), nucleotide diversity (π), number of segregating sites and the k-mean number of pairwise differences were calculated in DnaSP v5 [30].

Species Delimitation
The final aligned unique haplotype sequences were used for the analysis of species delimitation.Three species delimitation models were applied in order to retrieve the correct species partitioning, namely, the maximum likelihood solution (MPTP), the heuristic solution (SPTP) grouped in the Poisson tree process (PTP) [31] and the assemble species by automatic partitioning (ASAP) [32].PTP usually uses a phylogenetic tree as input and delimits species using non-ultrametric phylogenies [31].The maximum likelihood tree was generated using the RAxML software package implemented on a web server "RAxMLBlackbox" (http://embnet.vital-it.ch/raxml-bb/,accessed on 28 November 2021) [33], with a GTR + G substitution model applied in the analyses.The maximum likelihood tree was used as the input file.The analysis was run after the outgroup was removed.ASAP was performed using the command line version downloaded from https://bioinfo.mnhn.fr/abi/public/asap(accessed on 19 November 2021) with default parameters.The calculation methods of Jukes-Cantor (JC69) and Kimura (K80) TS/TV and p-distance methods were tested and compared.Two runs were conducted, one with the whole dataset and the other with independent datasets selected to represent species from the two subfamilies present, i.e., Eristalinae and Syrphinae.

Phylogenetic Analyses
Phylogenetic reconstructions were performed using the neighbour-joining (NJ) [34] and maximum likelihood (ML) approaches [35].Analyses were run independently with all generated sequences (n = 78) and the unique sequences (n = 35).In addition to the sequences generated in this study, thirty-nine (39) supplementary sequences of hoverflies were downloaded from Genbank.Identical sequences were removed using DAMBE v.7 [36].The NJ tree was estimated in MEGAX using the Kimura 2-parameter model of substitution (K2P distance), with 1000 bootstrap replicates.In addition, the dataset was subjected to Bayesian information criterion following the Akaike information criterion corrected (AICc) value and maximum likelihood value (lnL) to select the best model of evolution.Initially, the dataset was partitioned in MEGA X using the find best DNA/protein model option.Confirmation was performed using the web program FindModel (http://hcv.lanl.gov/content/hcv-db/findmodel/findmodel.html,accessed on 19 November 2021), developed from MODELTEST [37].MEGA X selected the GTR + G + I (first position), GTR + G (second position) and GTR + I (third position) model.FindModel selected the GTR + G (first position), TN93 + G (second position) and HKY + G (third position) models.The ML analysis was performed by taking into consideration the most appropriate models of evolution.Then, the 35 unique sequences along with 39 Genbank sequences were used to construct the ML tree.The South American hoverfly Alipumilio avispas Vockeroth, 1964 (AY261709) was used as an outgroup.

Distance Analyses and Genetic Diversity
DNA sequence analyses indicated that all of the final sequences selected were of good quality.Only 4 out of 78 (5%) sequences studied had missing data.The 78 hoverfly COI sequences represented 15 nominal species from nine genera (Table A1 in Appendix A).More than one conspecific sequence was available for 10 of the species, while the other 5 species were represented by only one barcode.The mean number of barcodes per species was 5.2, with 66.7% of species represented by at least two conspecific sequences (Table A1 in Appendix A).Two distinct COI sequences could not be identified by comparison with the sequences of Jordaens et al. [17].However, an analysis of previously published COI sequences from Pu et al. [38] confirmed that the two distinct sequences were from Eupeodes corollae Fabricius, 1794.The results from the Blast search revealed that a few COI sequences from the species included in this study are available in Genbank from sources other than Jordaens et al. [17].
No distance divergence was found within the Eristalinus vicarians Bezzi, 1915 species.There was no positive relationship between the intraspecific divergence and the number of specimens.When considering the divergences between genera, we observed lower divergence between the genera Phytomia and Eristalinus (≤10.41%)compared to the relatively high divergence between the other genera (≥14.88%)(Tables 4 and 5).The aligned sequences consisted of 35 unique haplotypes (H1-H35).Nucleotide diversity index (π = 0.01423) and the average number of nucleotide differences (k = 8.16) for A. nasuta revealed that this species was indeed the most divergent of the local species included in the present study.The majority of species were represented by low frequencies, ranging from 0.00070 to 0.00581, while E. corollae, represented by only two specimens, showed relatively high nucleotide diversity (0.00348).

Species Delimitation
We first analysed the ability of the three delimitation models (ASAP, MPTP and SPTP) to determine as accurately as possible the correct number of hoverfly species in our dataset.For each calculation method, ASAP produced 10 partitions with different species numbers; hence, only the partitions with the best ASAP score are presented here (Table 6).The partition with the best ASAP score revealed 14 species in our dataset.However, the partition with the second-best ASAP score exhibited 16 species in our dataset, which was around (±6%) of the expected number of species.The use of the model based on threshold distances led to an estimate of 14 hoverflies species (confidence interval: 14-17) for our whole dataset.MPTP and SPTP were not congruent in validating the 14 species as in ASAP.In contrast, for the Syrphinae dataset, both models revealed 5 species (MPTP model) and 6 species (SPTP model), while they both revealed 10 species for the Eristalinae dataset, bringing the total number of species to 15 (MPTP model) and 16 (SPTP model).The SPTP model agreed with the results of the ASAP analysis partitioning with the second-best ASAP score, which also identified 10 and 6 species for the Eristalinae and Syrphinae subfamilies, respectively.

Phylogenetic Analyses
The neighbour-joining (NJ) analysis of the COI sequences revealed that all species formed distinct clusters in the tree (Figure 2).All nine species represented by more than two haplotypes formed clusters with 100% bootstrap values, except A. nasuta where the combined bootstrap value of the clusters was 95% (Figure 2).However, a Genbank sequence putatively identified as Allograpta fuscotibialis (Accession number: MT449479) grouped closely with sequences of Allograpta nasuta identified with the help of the published work of Jordaens et al. [17].The distances between the two taxa ranged from 2.86 to 3.08%, and the grouping of that putative A. fuscotibialis sequence with the rest of the known A. nasuta sequences was supported by a bootstrap value of 96 (Figure 3).In the maximum likelihood (ML) analysis with species with more than two haplotypes, four nominal species (44.44%) were supported by high bootstrap values (>98%).A low (78%) bootstrap value supported the grouping of E. quinquelineatus sequences with the putative barcode of E. quinquelineatus from Genbank (Accession number: KR831005) (Figure 3).The barcodes of the genus Allograpta formed two clusters, and one of these included a Genbank barcode of Allograpta nasuta (Accession number: KR830850).The ML tree suggested a close relationship between Allograpta nasuta and Allograpta fuscotibialis followed by Mesembrius strigilatus and Eristalis tenax, Syritta pipiens Linnaeus, 1758 and Syritta flaviventris Macquart, 1842 and Phytomia erratica Bezzi, 1912 and Phytomia incisa Wiedemann, 1830.The divergence within the genus Eristalinus was observed in the tree, as three distinct clusters were formed for the species in this genus (Figure 3).

Discussion
The COI gene has been used successfully for species delimitation in Afrotropical hoverfly genera such as Merodon Meigen, 1803 [21,22], Sphaerophoria Le Peletier & Serville, 1828 [20], Eumerus Meigen, 1803 [18,19], Toxomerus Fabricius, 1798 [39], Allograpta Osten Sacken, 1875, Asarkina Macquart 1842, Episyrphus Matsumura, 1917 and Exallandra Vockeroth, 1969 [24].Using the COI barcode, Jordaens et al. [17] identified 90 species from 24 genera collected in Ghana, Togo, Benin and Nigeria.Even though species estimation from the Afrotropics is approximately 600 species, their study helped to improve species identification in this understudied group of hoverflies.Our study is the first local attempt using DNA barcoding to ascertain the taxonomy and establish the genetic richness and differentiation of species from the Free State province of South Africa.The three analytical approaches used in the present study conformed to morphotype assignment, except for the Allograpta genus where doubt still exists with regard to the number of species represented within the genus.In modern taxonomy, the use of DNA barcoding in combination with traditional taxonomy has led to the creation of reference libraries (of major groups of living organisms), which are useful for congruent species delimitation as in the case of Afrotropical hoverflies (e.g., Mengual et al. [20]).
In the present study, the different species delimitation methods tested resulted in various numbers of species, thus demonstrating the challenge of applying adequate criteria for distinguishing between species.The method named assemble species by automatic partitioning (ASAP), developed by Puillandre et al. [32], estimated a range of 14 to 17, while the Poisson tree processes based on the MPTP and SPTP [31] models estimated 15 and 16 species.In spite of the difference in species numbers, the results from the three models were closer when we divided the whole dataset into two subsets according to the subfamilies (Eristalinae and Syrphinae).The two subsets differed in terms of species, sequence and haplotype numbers.The number of sequences for each dataset was 66 and 12 for the Eristalinae and the Syrphinae datasets, respectively.The MPTP and SPTP models recovered the same species number for the Eristalinae dataset, confirming the usefulness of models based on the phylogenetic species concept (PSC) for species delimitation in this group.The PSC proposed by Cracraft [40] defined a species as "the smallest diagnosable cluster of individual organisms within which there is a parental pattern of ancestry and descent".Although it has been recently criticised because it oversplits the number of taxa [41][42][43], the use of such a concept in conjunction with another model based on genetic distance calculations appears to be useful for determining the true number of species.
In the present study, the COI divergence between A. nasuta and A. fuscotibialis was low (mean K2P = 2.86%) and comparable to that found between other Afrotropical hoverfly species in the genus Melanostoma (M.cf.floripeta Speiser, 1910 and M. cf.bituberculatum Loew, 1858, mean K2P = 2.95%) and the genus Microdon (Microdon brevicornis Loew, 1858 and Archimicrodon sp., Hull, 1937 mean K2P = 2.84%) [14].From the study of Jordaens et al. [17], divergences between species from monophyletic genera were even lower for the genus Allobaccha (A. euryptera Bezzi, 1908 and A. picta Wiedemann, 1830, mean K2P = 0.51%) and the genus Rhingia (R. caerulescens Loew, 1858 and R. semicaerulea Austen, 1893, mean K2P = 1.66%), whereas high intraspecific divergences were observed in Graptomyza triangulifera Bigot, 1883 (mean K2P = 2.13%), Asarkina ericetorum Fabricius, 1781 (mean K2P = 2.63%), Phytomia natalensis Macquart, 1850 (mean K2P = 2.93%), Allobaccha picta Wiedemann, 1830 (mean K2P = 3.07%), Syritta bulbus Walker, 1849 (mean K2P = 4.57%) and Polybiomyia divisa Walker, 1857 (mean K2P = 5.63%).One of the explanations provided by the authors is that this may reflect geographical structuring or evolutionary history and that speciation events were recent in these lineages; consequently, the COI sequences had not yet accumulated many mutations.Thus, they argued that this could signal the occurrence of cryptic species.In our case, the values for intraspecific variations were consistently lower than those for interspecific divergence.This is in accordance with the findings from other studies carried out in restricted geographical zones on other taxa including earthworms [44] and bivalves [45].Regarding the two taxa of Allograpta (Table 2), without deep morphological investigation, it is difficult to use the present distances to sort out the ambiguity between them.The use of additional molecular markers (e.g., nuclear 28S and 18S ribosomal RNA genes) could also provide more insights into speciation dynamics in these organisms [20].The use of additional markers can increase the power of taxonomic resolution or provide the opportunity to test different hypotheses because different markers would potentially have different evolutionary histories as they evolve differently (slow and fast rates of mutation).For instance, the 14 species recovered with the ASAP model could correspond to a relatively low rate of speciation, but when the speciation rate increases it becomes difficult for the model to delimit species by using pairwise genetic differences.Indeed, a shortcoming of this model is that individuals that experienced recent speciation events with faster speciation rates (>1) could be wrongfully considered as the same species [32].Also, the fact that certain taxa, such as E. fuscicornis, P. erratica, S. flaviventris, S. pipiens and Asarkina sp.Macquart, 1842 (Table A1 in Appendix A) were represented by a limited number of sequences might have also influenced species delimitation.Ahrens et al. [46] discussed the issue of singletons (lone representative sequence of a species) in DNA-based species delimitation studies and argued that a high proportion of singletons may compromise the estimation of intra-versus interspecies evolutionary processes.From their study on approximately 100 species of beetles (Scarabaeidae family), 48.5% of species were singletons, which produced poor results when different delimitation approaches were used among which were the Poisson tree processes (PTP), generalized mixed yule coalescent (GMYC) and automatic barcode gap discovery (ABGD).Given our relatively small dataset, the same limitation might have applied as a result of the overall percentage of singletons, which stood at 33.33%.
Based on Blast analysis, all of our sequences showed similarity to published sequences in Genbank within a range of 92.48-100%.All relevant Genbank sequences included in the dataset were genetically distant, with 0 to 7.85% divergence among the 15 putative Afrotropical species.Taxa clustered together with high bootstrap values irrespective of their number of sequences in the dataset.The neighbour-joining tree exhibited 100% bootstrap values for all clusters, except in A. nasuta where a 95% value was observed.In the case of A. nasuta and A. fuscotibialis, the occurrence of high intraspecific molecular variability for the COI marker (as shown in Table 3) indicates that this marker may not be adequate for distinguishing between species within the Allograpta genus.This was further supported by the fact that both taxa were conflated in the tree with significant bootstrap support values.Throughout the literature, different cases of inconsistency between morphological and molecular data have occurred (e.g., for Allograpta lineages, Mengual et al. [24]).
The data generated from this study support the identification of Afrotropical hoverflies for 15 nominal species from nine genera.Moreover, based on the work of Whittington [2]

Figure 1 .
Figure 1.Study area in the Maluti-A-Phufong and Phumelela municipalities (Free State province of South Africa).Red circles represent sampling locations.

Figure 1 .
Figure 1.Study area in the Maluti-A-Phufong and Phumelela municipalities (Free State province of South Africa).Red circles represent sampling locations.

17 Figure 2 .
Figure 2. Neighbour-joining tree depicting the degree of relationship between the Afrotropical hoverfly species produced from the COI gene.Bootstrap support values higher than 70 are reported on the tree.Information on specimens is given in the Table A1 in Appendix A. Alipumilio avispas (AY261709) was included as the outgroup.

Figure 2 .
Figure 2. Neighbour-joining tree depicting the degree of relationship between the Afrotropical hoverfly species produced from the COI gene.Bootstrap support values higher than 70 are reported on the tree.Information on specimens is given in the Table A1 in Appendix A. Alipumilio avispas (AY261709) was included as the outgroup.

Figure 3 .
Figure 3. Maximum likelihood tree based on the general time-reversible model.Bootstrap support values higher than 70 are reported on the tree.The twelve green bars represent specimens included in species delimitation models (ASAP, MPTP and SPTP).The thirteen black bars indicate published sequences (and accession numbers) downloaded from either Genbank or Bold.Alipumilio avispas (AY261709) was included as the outgroup.

Figure 3 .
Figure 3. Maximum likelihood tree based on the general time-reversible model.Bootstrap support values higher than 70 are reported on the tree.The twelve green bars represent specimens included in species delimitation models (ASAP, MPTP and SPTP).The thirteen black bars indicate published sequences (and accession numbers) downloaded from either Genbank or Bold.Alipumilio avispas (AY261709) was included as the outgroup.

Table 1 .
Primers used for amplifying the COI gene in this study.

Table 2 .
Mean pairwise interspecific distance (%) between the Afrotropical hoverfly species sampled in the study area.Kimura 2-parameter distances are shown in the lower left matrix, and LogDet distances are shown in the upper right matrix.

Table 3 .
Mean intraspecific variation (%) for each species sampled in the study area.Only species represented by more than one specimen were used for the analysis.n = number of specimens.Standard errors are presented in parentheses.

Table 4 .
Mean variation (%) within each genus.Only genera represented by more than one species were used for the analysis.n = number of species.

Table 6 .
Results of the analysis of the sequence dataset based on delimitation models.
, who assessed the faunal diversity of Afrotropical Syrphidae, 196 Syrphidae species occurring in South Africa were compiled into a checklist (see details here: https://www.syrphidae.com/checklist.php?country=ZA, accessed on 17 January 2023.Our work reveals that at least 15 different species belonging to this family of Diptera are found in our study area.Interestingly, five of these species, namely, Betasyrphus inflaticornis,