DNA barcoding of Chironomidae from the Lake Skadar region: Reference library and a comparative analysis of the European fauna

The main aim of this study was to fill a gap in barcoding data of the European Chironomidae of the Balkan region, developing and testing the efficiency of a reference DNA barcode library for ancient Lake Skadar basin species (Montenegro/Albania), a region of Europe never before subjected to barcoding studies on Chironomidae. Another aim was to test the efficiency of DNA barcoding for the identification of European Chironomidae, including the estimation of optimal identification thresholds, using >12,000 barcodes.

For the Balkan region, only 271 COI barcodes of Chironomidae (3 from Croatia and 268 from Bulgaria) belonging to 40 taxa have been registered in BOLD prior to our study. In this region, among the most suitable areas for Chironomidae, we found the Lake Skadar basin, the largest lake in the Balkan Peninsula, located in the Zeta-Skadar valley (south-west part of the Balkan Peninsula) (Pešić et al., 2019). This is a shallow lacustrine ecosystem, of which two thirds (229 km 2 ) of its surface belongs to Montenegro and about one third (142 km 2 ) to Albania. The lake is approximately 44 km long and 14 km wide, with a surface area that fluctuates seasonally between 370 and 530 km 2 . The Lake Skadar basin contains a very young lake that originated in the mid-sub-Atlantic stage of the Holocene, fed by a geologically old spring system, which has its origins in the Pliocene Pešić et al., 2019).
The main aim of this study is to develop a comprehensive barcode reference library for the molecular identification of aquatic and terrestrial Chironomidae from the Lake Skadar area (Albania and Montenegro), through a DNA barcoding approach, targeting a fragment of ~658 bp of the COI gene. Moreover, taking advantage of publicly available COI sequences of European Chironomidae, we aim to estimate the general efficiency of DNA barcoding for the molecular identification of species within this family and to develop nucleotide distance thresholds K E Y W O R D S DNA barcoding efficiency, Europe, integrative taxonomy, Lake Skadar, non-biting midges to improve molecular identification for the family and for the different subfamilies of non-biting midges present in Europe.

| Sample collection and identification
Chironomidae specimens were collected from 72 sampling sites located in the Lake Skadar area (Montenegro and Albania) during four sampling campaigns: spring and fall 2014, summer 2015 and summer 2018 (Figure 1). Aquatic immature stages were collected from streams and rivers using a standard kick sampling procedure, from the bottom of the lake and from macrophytes, using a benthic dredge. Imagines were collected by sweeping through littoral vegetation using an entomological hand net and through light trapping (in open areas, a 500W white light was placed in front of a white screen of 2 m × 3 m, ensuring that the light was visible for a long distance above the lake surface). Sampling sites were selected to maximize the representativeness of habitats present in the area (i.e. open lake, shallow lake, springs, sublacustrine springs, river inflow and terrestrial parts). The collected specimens were removed from bulk samples and placed into absolute ethanol, then stored at −20°C.
Morphological identification of specimens took place prior to DNA extraction. Adult males were selected from each vial and morphologically identified using appropriate keys (Giłka, 2011;Langton & Pinder, 2007;Widerholm, 1989). The larvae and pupae were not morphologically identified; only molecular identification was performed after the development of the adult barcode library.

| DNA extraction, PCR and sequencing
For most specimens, DNA was extracted using the GeneMATRIX Tissue DNA Purification Kit, provided by EURx (Gdańsk, Poland), following the manufacturer's protocol. In the case of larvae, DNA was extracted from the whole body after tissue homogenization using sterile pestles. For adults, DNA was extracted from one/two legs (specimens >2 mm) or from the entire body (specimens <2 mm). The DNA extract was then used as a template for amplification through PCR, using the standard 658 bp barcode region of COI and the primer pairs LCO1490/HCO2198 (Folmer et al., 1994) and C_LepFolF/C_LepFolR (Folmer et al., 1994;Hebert et al., 2004). PCR was carried out in a final volume of 11 µl, containing DreamTaq reaction buffer (DreamTaq DNA polymerase, 2× DreamTaq buffer, dATP, dCTP, dGTP and dTTP, 0.4 mM each and 4 mM MgCl 2 ; Thermo Scientific Inc.), 5 µmol each primer, ultrapure water and the DNA template. The following PCR thermal protocol was adopted: 94°C for 1 min followed by five cycles of 30 s at 94°C, 1 min 30 s at 45°C and 1 min at 72°C; 36 cycles of 94°C for 30 s, 51°C for 1 min 30 s and 72°C for 1 min; with the final extension of 5 min at 72°C. Successful amplification was verified by agarose gel electrophoresis. The PCR products were purified using a mixture of FastAP (1 U/µl, ThermoFisher Scientific) and Exonuclease I (20 U/µL ThermoFisher Scientific). Positive amplicons were directly sequenced using the Sanger method and the marker-specific primers by Macrogen Europe (Amsterdam, the Netherlands) (Sanger & Coulson, 1975). For a further 85 specimens, DNA was extracted, and the barcode region was amplified and sequenced at the Centre for Biodiversity Genomics, Guelph, Canada, following the standard protocols of the centre (Porco et al., 2010). All electropherograms obtained were edited and primers removed using Geneious Pro 11 (Biomatters Ltd., Auckland, New Zealand; Kearse et al., 2012). The presence of an open reading frame was verified using Geneious software. Consensus sequences were deposited in BOLD (Dataset "DS-CHBAL-Chironomidae of the Skadar Lake basin," https://doi.org/10.5883/DS-CHBAL) and are available on GenBank (BankIt2349074, Accession numbers MT534631 to MT535400). Adult voucher specimens are deposited at the University of Lodz, Poland, and at the University of Milan, Italy, under P. Gadawski's and B. Rossaro's curation respectively.

| DNA barcode identification efficiency for Chironomidae
The reference library for the molecular identification of Chironomidae from the Lake Skadar area developed in this study was first examined using BOLD tools (Ratnasingham & Hebert, 2013). The distribution of barcode clusters [Barcode Index Numbers (BINs)] was performed F I G U R E 1 Sampling sites in the Lake Skadar area. Sampling sites are represented with green circles, the number of individuals from which DNA barcodes were developed is reported in the circles. Map source: OpenStreetMap contributors based on the Open Data Commons Open Database License by the Barcode of Life Data System v4. BIN was considered homogeneous when consisting of records of one species and heterogeneous when there was more than one. The number of haplotypes was estimated using the BOLD data systems' accumulation curve tool (Ratnasingham & Hebert, 2007;Villesen, 2007). The Refined Single Linkage (RESL) algorithm, a staged clustering process that employs single linkage clustering implemented in BOLD, was used to assign Chironomidae barcodes to operational taxonomic units.
Subsequently, newly developed sequences were aligned at the codon level using MUSCLE (Edgar, 2004) and a pairwise nucleotide distance matrix calculated using the library ape (Paradis & Schliep, 2018) in R version 3.5.2 (R Core Team, 2020), adopting Kimura's two-parameter (K2P) as a nucleotide substitution model (Kimura, 1980). This model is routinely used to calculate the distances between DNA barcodes (Nishimaki & Sato, 2019). All of the following analyses were performed only on the barcode sequences obtained from adult Chironomidae. A K2P nucleotide distance matrix was developed and used for the calculation of the optimal threshold for the molecular identification of species present in the dataset using threshold optimization analysis of the R library spider v1.4-2 (Brown et al., 2012). This method, firstly proposed by Meyer and Paulay (2005) allows the estimation of a threshold value that minimizes the cumulative identification error (the number of false positives (FP), conspecifics with a value of nucleotide divergence higher than the threshold value, plus the number of false negatives (FN), i.e., heterospecifics with a value of nucleotide divergence lower than the threshold value). The optimal threshold (OT) obtained was then used in the Best Close Match function of the R library spider v1.4-2 (Brown et al., 2012) to evaluate the efficiency of the newly developed dataset for the identification of Chironomidae in the Lake Skadar area.
The Best Close Match method compares each sequence of the dataset with the others included in the same dataset and checks if the best matches (pairs of sequences with the lowest values of nucleotide distance) are between sequences of organisms morphologically identified as the same species. The sum of the correct matches obtained thus allowed us to estimate the efficiency of DNA barcoding regarding our dataset.
To test the general efficiency of barcoding for Chironomidae, all publicly available COI-5P sequences of European Chironomidae available in BOLD (Ratnasingham & Hebert, 2007) were retrieved (date of download: 13.04.2021) and filtered to create an accurate species-level dataset, excluding all sequences missing species-level identification or proper scientific names and shorter than 350 base pairs (bp). This procedure was performed using an ad hoc R script.
This dataset, which also included the barcodes for the Chironomidae of Lake Skadar developed in this study, was analysed with the same procedure as previously described, to estimate the barcoding efficiency and the OT. Furthermore, the dataset was used to estimate and compare barcoding optimal thresholds and the efficiency of barcoding for the molecular identification of European Chironomidae subfamilies. For this purpose, the whole set of sequences was arranged into sub-datasets, each including barcodes of one of the following subfamilies: Chironominae, Diamesinae, Orthocladiinae, Podonominae, Prodiamesinae, Tanypodinae and Telmatogetoninae.
Minimum spanning haplotype networks (Bandelt et al., 1999) were reconstructed using PopART (Leigh & Bryant, 2015) for genera including species that posed difficulties during morphological identification, and for those for which incongruences between morphological and molecular species identification were recovered.

| Identification of Chironomidae larvae
To perform the molecular identification of barcodes generated from larvae and pupae, two different molecular identification methods were applied and compared (adults' barcodes developed in this study were deposited in BOLD before the molecular identification of larvae). (i) The BOLD taxonomy browser was used, and the congruence of BIN assignment was evaluated. Sequences were matched against barcodes registered in the BOLD COI database (already including barcodes of adults developed in this study), each one was assigned to a BIN which was first inspected for concordance (all the barcodes composing it had to belong to a single morphospecies).
In the case of concordance, the respective species identification was assigned to the developed barcode. When the BIN consisted of sequences of two or more species within the same genus, only genus-level identification was given to the barcode. Moreover, when the newly developed barcode formed a new BIN, identification was given to the family level. (ii) The optimal threshold for the identification of Chironomidae estimated in this study was used as a clustering threshold. A dataset comprising unidentified larval sequences plus all identified sequences at the species level present in BOLD was assembled and those sequences were clustered at the nucleotide distance value corresponding to the optimal threshold. When the derived clusters were composed of one or more unidentified larval sequences plus one or more sequences retrieved from BOLD belonging to the same species, the identification of the latter was assigned to the larval sequences. When the clusters included sequences of two or more species of the same genus, only genus-level identification was assigned to the larvae. Finally, when the clusters comprised only larval sequences, taxonomic assignment to the family level was given to those sequences. Clustering analysis was performed using the R library spider v1.4-2 (Brown et al., 2012).

| DNA barcode library for Lake Skadar Chironomidae
In this study, a total of 770 COI barcode sequences of Chironomidae were recorded from the Lake Skadar area (744 specimens collected in Montenegro and 26 in Albania), representing the first barcodes of Chironomidae from these two countries available on BOLD.
Almost half of the sequences were obtained from larvae (403 sequences), the rest were from adult males (367 sequences). In this study, we tested two approaches to the molecular identification of Chironomidae larvae from Lake Skadar that gave concordant results. Specifically, the identification of 93.5% of the sequences (377/403 sequences) provided identical results using both approaches, but using the BIN congruence method, 22 more sequences were assigned to the species level (Table S1). According to the results of the latter method, 263 out of 403 barcodes were assigned to 52 species; 74 barcodes were identified only to the genus level, were assigned to the following seven genera: Chironomus (17) Adult males were morphologically identified as belonging to 77 species, 37 individuals were identified only to the genus level or to a higher taxonomic rank (23 and 14 individuals respectively). The BIN and RESL (for operational taxonomic units) analyses of adult males assigned sequences to 98 BINs and 102 OTUs. Some species were collected only at the adult or pre-imaginal stages; therefore, the total dataset obtained for the Chironomidae of Lake Skadar includes barcodes from 97 different species (Table S2). Analysing the entire dataset (barcodes of adults and larvae) through BIN and RESL clustering analyses, the 770 barcodes were assigned to 159 different BINs and 167 OTUs. During the dataset publishing process, 65 BINs were new for BOLD (39.1%, 201 sequences), 100 of the remaining were concordant (60.2%) and one was discordant (0.6%). Seventyone BINs were each represented by a single sequence (42.7%). One hundred and one BINs (60.8%) were already present in the database.

| Chironomidae diversity of Lake Skadar
Within the Lake Skadar area, 97 species, assigned to 44 genera, were recorded, with a mean number of collected specimens per species of 6 (1-29). The most common genus was Chironomus, for which 114 barcode sequences were generated (95 barcodes identified as belonging to 11 species), followed by Polypedilum (60 barcodes, 51 identified at the species level to nine species), and Cricotopus and Procladius (50 and 49 barcodes, eight and two species respectively).

95.1%
Note: N of individuals-number of individuals whose barcode sequences were included in the analysed dataset; Incorrect-incorrect identifications detected by Best Close Match analysis; No id-number of sequences having no match with conspecifics, detected by Best Close Match analysis, due to a pairwise nucleotide distance higher than the adopted OT; Correct-correct identifications detected by Best Close Match analysis, plus singletons; OT-optimal threshold value estimated for the dataset; Cumulative error-cumulative error related to the OT estimation; Barcoding efficiency-barcoding efficiency estimated for the dataset, values are reported as a percentage.

| Optimal thresholds for molecular identification and barcoding efficiency
In this study, the identification efficiency of the de novo barcode ref- were estimated (4.1% mean intraspecific distance, Figure 2).
The estimated OT values ranged from 0.3% (Diamesinae) to 4.4%  (Table   S4). For the latter subfamily, the lowest barcoding efficiency within the Chironomidae was thus 87.5%. An efficiency of 100% was obtained for the smallest datasets, that is, Telmatogetoninae and Podonominae.

| Haplotype network analysis
A haplotype network was inferred for each of the following genera: Cricotopus, Chironomus, Diamesa, Micropsectra and Polypedilum.

F I G U R E 2
Inter-and intraspecific K2P distances for the developed datasets: European Chironomidae, Lake Skadar adult males, seven Chironomidae subfamiles. Green-intraspecific distance, orange-interspecific distance, OT-optimal threshold; N-number of sequences; n-number of species; BEbarcoding efficiency. Summary statistics on intra and interspecific distances (Table S5) Based on the haplotype network of Cricotopus, the sequences of this genus were clustered into two groups-the first consisting of four species and the other of seven species (Figure 3; Table S5). Only C.
sylvestris and C. glacialis presented shared haplotypes (Figure 3). The genus Chironomus was represented by 24 species in the haplotype network (Table S5). Two groups of species-C. pseudothummi-C. riparius and C. heteropilicornis-C. pilicornis-were found to share haplotypes ( Figure 4). The haplotype network for Diamesa consisted of 15 species (Table S5), and also within this genus, two groups of species had shared haplotypes-D. bohemani-D. zernyi and D. tonsa-D.
hyperborea ( Figure 5). In the haplotype network of the Micropsectra (25 species; Table S5) (Figure 6), no shared haplotypes were observed. Finally, the Polypedilum haplotype network obtained from 19 species (Table S5) included only one group sharing haplotypes-P.

| DNA barcoding efficiency and molecular identification thresholds
The present work led to the development of the first comprehensive barcode library for the molecular identification of Chironomidae of the Lake Skadar region, including 97 species, that is 45.3% of the regional Chironomidae fauna (Gadawski et al., 2022). barcoding as a tool for the identification of Chironomidae species (Ekrem et al., 2010Lin et al., 2015). The potential value of this method is also supported by the number of sequences obtained from unidentified larvae that were successfully identified to the species level (>65%). Interestingly, both the approaches adopted in this study for the molecular identification of larvae (assessing the congruence of BINs to which sequences were assigned using the BOLD taxonomy browser and the clustering analysis using the optimal threshold for Chironomidae) gave highly similar results. Few inconsistencies were However, we also conclude that the large proportion of larvae that were not assigned with species names (35%) indicates that a greater sampling effort should be applied to further expansion of the ref- and, consequently, also the optimal threshold value for the correct identification of these queries.
We also conclude that using specific thresholds for the identification of species belonging to different Chironomidae subfamilies is associated with lower identification errors than using a general threshold for the identification of all species within the family.

F I G U R E 3
Minimum spanning haplotype network for the genus Cricotopus. Each colour represents one species. The diameter of the circle is proportional to the abundance of the haplotypes. Numbers of substitutions between haplotypes are reported within parentheses Specifically, the optimal thresholds estimated for each European Chironomidae subfamily resulted in a total cumulative error lower than that related to the more general family threshold (total CE for subfamilies = 1011 with respect to family CE = 1079). This result suggests that using taxon-specific identification thresholds improves molecular identification results (Lin et al., 2015;Magoga et al., 2018).
Interestingly, when we analysed the intraspecific variability of species belonging to the different Chironomidae subfamilies, we found that some Tanypodinae species are characterized by unusually high intraspecific distances (e.g., Monopelopia tenuicalcar, mean intraspecific distance of 9.8%; Conchapelopia pallidula mean intraspecific distance of 8.1%; Guttipelopia guttipennis mean intraspecific distance of 7.6%; Natarsia punctata, mean intraspecific distance of 5.8%; and Ablabesmyia longistyla, mean intraspecific distance of 5.8%), which suggests the possible presence of cryptic diversity within these species. These high intraspecific distances also explain the lack of a barcoding gap in the distribution of nucleotide distances of this subfamily (Table S3). It is generally well known that DNA barcoding F I G U R E 4 Minimum spanning haplotype network for the genus Chironomus. Each colour represents one species. The diameter of the circle is proportional to the abundance of the haplotypes. Numbers of substitutions between haplotypes are reported within parentheses efficiency can be influenced by many factors, for example existence of superspecies or cryptic species within the studied group, occasional interspecific hybridization, introgression and intervention of bacterial endosymbionts changing pathways of mtDNA inheritance (Magoga et al., 2018). In addition to biological reasons, also

| Chironomidae of the Lake Skadar region
All the specimens collected in the Lake Skadar region and morphologically identified were assigned to species already listed in the European fauna. Further in-depth studies, based on molecular F I G U R E 6 Minimum spanning haplotype network for the genus Micropsectra. Each colour represents one species. The diameter of the circle is proportional to the abundance of the haplotypes. Numbers of substitutions between haplotypes are reported within parentheses data, should focus on the taxonomy of the local chironomid fauna to reveal the actual level of endemism, which could be high, even if that is not indicated by the most recent species checklist from the Lake Skadar basin (Gadawski et al., 2022). The progress of knowledge on chironomids suggests that many species have large distribution ranges. Thus, the presence of endemic cryptic species in the Lake Skadar area cannot be excluded (Giłka et al., 2013;Rossaro et al., 2019). Quite recently, an endemic chironomid species was described from Croatia (Micropsectra uva Giłka et al., 2013) and its presence in Lake Skadar cannot be excluded. The Lake Skadar basin, together with its large karst spring system, is definitely ancient, having originated more than 2.5 million years ago, and having been isolated for most of that time . Some of the karst spring connections are likely to be more than 100 Ka old and to have remained in prolonged isolation, increasing the probability of the emergence of divergent and locally endemic phylogenetic lineages or even species (Jabłońska et al., 2020). Since some larvae collected in this study were not F I G U R E 7 Minimum spanning haplotype network for the genus Polypedilum. Each colour represents one species. The diameter of the circle is proportional to the abundance of the haplotypes. Numbers of substitutions between haplotypes are reported within parentheses identified to the species level and their sequences were clustered in OTUs and BIN unassigned to species of BOLD, it is expected that undescribed or currently non-barcoded species may be present within the sampled material, possibly including endemics. For the barcode dataset developed using Lake Skadar adults Chironomidae as vouchers, the 1.5% mean intraspecific nucleotide distance was estimated. This value is in line with those found in previous barcoding studies on Chironomidae (Song et al., 2018) and lower than those usually estimated for species sampled on a larger geographical scale (Lin et al., 2015). Anyway, some species having an unusually high intraspecific diversity were recorded in this study, for example, Limnophyes minimus, for which a 17.4% maximum intraspecific distance was estimated. These distance values are similar to those found for Chironomidae species potentially hiding cryptic diversity (Lin et al., 2015;Song et al., 2018), further suggesting undescribed species could be present within the Lake Skadar area.
Among the 97 species recorded in the Lake Skadar area, 57 species were previously reported in the literature for the region (Gadawski et al., 2022). The seven records discussed hereafter are doubtful and need to be thoroughly investigated. Bryophaenocladius nigrus Albu, 1964, is known so far only from Romania, Finland and Germany, and this record would be the southernmost occurrence of this species. Chironomus (Chironomus) curabilis is currently known only from Russia (Belyanina et al., 1990). Chironomus (Lobochironomus) pseudomendax was initially identified as C. mendax by Wülker in Ryser et al. (1985). However, specimens from Sweden and Russia were subsequently identified as C. mendax, Storå, and this name was finally replaced by C. pseudomendax by Wülker Considering general chironomid diversity of Europe, we can observe that ~1300 species have been recorded so far, 599 species barcoded and publicly available in BOLD represent 46% of the European chironomid fauna (Saether & Spies, 2013 sylvestris. In the BOLD system, all barcodes of C. glacialis are clustered in one BIN, including some C. sylvestris barcodes (BOLD: AAA5299). Hirvenoja (1973) and Brodin et al. (2013) hypothesized that C. glacialis could be a junior synonym of C. sylvestris. Our findings support this hypothesis, but also a further scenario should be considered, that is, these two species could be recently evolved and maybe have not yet accumulated sufficient mutations to be separated by DNA barcodes.
Cricotopus glacialis is restricted to northern Europe (Gresens et al., 2012), and in BOLD it is reported from continental Norway (seven sequences), 11 sequences from the Svalbard archipelago, and three individuals from Iceland. Moreover, some Chironomidae individuals collected in the Italian Alps were assigned provisionally to C. glacialis, but a comprehensive analysis of their morphological characters could not fully confirm this identification (Bruno Rossaro det.). For this reason, the species was not included in the Italian fauna (Rossaro et al., 2019). Individuals of C. sylvestris, whose barcodes are published in the BOLD system, were collected in Norway (21 individuals   were found to group with others of Cricotopus relucens Hirvenoja, 1973 (PGBAL249-19).

| The genus Psectrocladius
Within the genus Psectrocladius, incongruences between morphological and molecular identification were recorded for the species Psectrocladius octomaculatus Wülker (Kieffer, 1956) (two sequences of individuals collected in Sweden) and Psectrocladius psilopterus (Kieffer, 1906) (four sequences belonging to individuals collected in Finland, Norway and Iceland). These species were found to share haplotypes and their sequences are assigned to the same BIN (BOLD: AAX8186). Morphologically, they can be separated both as adults and pupae (Wülker, 1956), but a high level of taxonomic expertise is required. The detected inconsistency could thus be indicative of species misidentifications, or, on the other hand, the lack of genetic distinctiveness could suggest these taxa are the same species and that they should be synonymized. Another possible scenario may involve the hybridization/introgression phenomena.

| The genus Chironomus
The analyses performed in this study identified two groups of species: Chironomus pseudothummi-C. riparius and C. heteropilicornis-C. pilicornis that share haplotypes (Figure 4). Chironomus pseudothummi, C. (Chironomus) lugubris and C. riparius Meigen, 1808 are difficult to separate on the basis of the morphology of adult males (Lindeberg & Wiederholm, 1979). Recently, morphological and molecular characters allowing accurate identification of C. lugubris were presented by Stur and Ekrem (2020). Chironomus pseudothummi and C. riparius can indeed be identified on the basis of pupal exuviae (Langton & Visser, 2003). Chironomus. lugubris has a North-Central European distribution, even though Rossaro et al.
(2019) preliminary identified some exuviae found in Italy as belonging to this species. The species was also reported from Switzerland and Hungary (Rossaro et al., 2019). The nine specimens whose barcodes are present in BOLD were collected from Spitsbergen (Norway) and assigned to one BIN (BOLD: AAB4581). Chironomus pseudothummi Strenzke, 1959 is distributed in Central and Northern Europe. It was not previously reported from the Mediterranean area, but it has been collected from the River Adda (Italy) (Rossaro et al., 2019). The species has been now reported also from the Lake Skadar basin and its occurrence is supported by molecular data The sequences of this species are assigned to three BINs in BOLD (BOLD: AAA7263 including 140 records, and BOLD: AAU4044 and BOLD: AAI4307 having one and two records respectively). This situation could be explained by a high level of intraspecific variability due to the wide species distribution.
Chironomus heteropilicornis Wülker (1996) (Martin, 2017;Wülker, 1996). The barcode data available in BOLD for C. heteropilicornis were developed from specimens collected from Norway (five barcodes) and those of C. pilicornis from Sweden (two barcodes). The sequences of C. heteropilicornis and C. pilicornis are assigned to one BIN (BOLD: ACX5781), which could indicate species misidentification or that both species are in fact one species and should be synonymized. Another possible scenario suggests the involvement of hybridization/introgression processes. Both species can be easily confused with C. plumosus in morphological identification, but they are clearly separated on the basis of molecular data.
Another species morphologically very similar to C. plumosus is Chironomus (Chironomus) curabilis Belyanina et al. (1990). Both molecular information and karyotype analysis (Polukonova et al., 2009) facilitate easier determination of these two species than using morphological characters. Chironomus plumosus is usually recognized as a widely distributed species in Europe, but is not well represented in the BOLD database. A total of 27 barcodes are currently available, 14 of which belong to specimens collected in Montenegro. Among the others, 11 were collected in Sweden and one in Poland. All these sequences are grouped within 18 haplotypes and assigned to one BIN-BOLD: AAU2239-but it is not homogeneous and includes also sequences of C. usenicus (8), C. tentans (two sequences) and Cladotanytarsus atridorsum (one sequence), which are probably the result of misidentification.

| The genus Diamesa
Our results also reveal two groups of species sharing COI haplotypes: Diamesa bohemani-D. zernyi and D. tonsa-D. hyperborea. It has previously been suggested that Diamesa zernyi Edwards, 1933, andDiamesa bohemani, Goetghebuer, 1932, are the same taxon and should be regarded as synonyms (Lencioni et al., 2021) and our results support that conclusion.  (Lencioni et al., 2021).
Among the other Diamesa species analysed in this study, an interesting case is represented by Diamesa insignipes Kieffer, 1908, for which 238 sequences were analysed, including 79 unique haplotypes. All these barcodes form one BIN (BOLD: ACR0046), and cluster together in the haplotype network of Diamesa genus ( Figure 5), with an average haplotype divergence of two nucleotide substitutions. According to BOLD metadata, all the specimens from which these barcodes were developed were collected at a single sampling site in Germany (between Bonn and Koblenz).

| The genus Micropsectra
In this study, the species Micropsectra apposita (Walker, 1856) and Micropsectra contracta, Reiss, 1965 were treated as synonyms (the latter as a junior synonym), following Saether and Spies (2013) and earlier literature (e.g. Giłka, 2011) which report a lack of morphological characters to reliably separate these two species. Our results, as well as the data on ecological preferences (the present authors' datasets), support their synonymy. Micropsectra junci is morphologically very similar to M. apposita (Walker, 1856), but relatively easily separable using the key of Giłka (2011), as both species occur in the same type of habitat (Giłka, 2001). Among the barcodes mined from BOLD, M. apposita (=M.

| The genus Polypedilum
From analyses of the barcode data of Polypedilum (Tripodura) tridens Freeman, 1955, and Polypedilum (Tripodura) bicrenatum Kieffer, 1921, shared haplotypes between these species were found. These are considered as distinct species, but they are very similar morphologically. In the literature, P. tridens is reported only from Africa. Within the barcodes we analysed, four P. tridens from Europe were present (three specimens from Germany and one from Norway), while P.
bicrenatum was represented by six barcodes, all from Sweden. All these sequences were assigned to one BIN (BOLD: ACV3478). It let us hypothesize about these species' synonymy, but further taxonomic investigation is necessary to confirm it.

| CON CLUS IONS
• Our study provides a reference library including 770 COI DNA barcodes (97 species) of Chironomidae from the Lake Skadar basin area. This represents 45.3% of the chironomid species recorded so far from Montenegro. Chironomids of this region have never been barcoded before.
• Some of the species detected in this study had not been previously recorded in the Lake Skadar area. Since only one individual was collected for each of them, further investigations in the area are needed to confirm the status of these species.
• The de novo DNA barcode dataset resulted in a high identification efficiency (98.6%).
• In this study, the usefulness of DNA barcoding for the molecular identification of Chironomidae was reaffirmed. The general identification efficiency estimated for European Chironomidae resulted in 95.8%, when a threshold of 1.6% was used as a limit between intra/interspecific levels. Moreover, the method enabled us to identify >65% of the larvae included in this study to species level.
This highlights the value of this method as a water bioindicator monitoring tool.
• Comparisons between optimal thresholds estimated at different taxonomic levels on Chironomids, viz. family and subfamily, have underlined the importance of using taxon-specific thresholds to increase the efficacy of molecular identification.
• The integration of morphological and molecular evidence supported five cases of synonymy representing genera: Cricotopus, Psectrocladius, Chironomus, Diamesa and Micropsectra within the Chironomidae family.
• Some cases of barcoding failure were observed for groups of closely related species or for species whose identification require a high level of taxonomic expertise, such as those belonging to