Introduction
Chironomids represent a dominant group of benthic macro-invertebrate populations and have been observed as one of the principal groups of aquatic organisms both in terms of number and distribution in sampling studies (Carr & Hiltunen, Reference Carr and Hiltunen1965). Chironomids are ecologically important due to their contribution to the food web, nutrient cycling, and pollutant accumulation; however, the adaptive ability of chironomids and ease of both natural and anthropogenic-mediated transport warrant a concern for their potential role as invasive pests, especially in recently disturbed environments (Failla et al., Reference Failla, Vasquez, Fujimoto and Ram2015). The terrestrial adult stage is short-lived and often characterized by swarms of mating adults, sometimes presenting a great nuisance to humans in environments where emergences occur (Ali et al., Reference Ali, Stanley and Majori1985, Tabaru et al., Reference Tabaru, Moriya and Ali1987, Iwakuma, Reference Iwakuma1992, Hirabayashi & Okino, Reference Hirabayashi and Okino1998). The ecological roles of chironomids have made sampling and subsequent species-level identification an important and useful biological tool for monitoring lake health (Langdon et al., Reference Langdon, Ruiz, Brodersen and Foster2006).
The identification of chironomid larvae to species level represents a challenging task for taxonomists. Identifying characteristics are more difficult to distinguish among the immature features of larval specimens, as compared with mature adults (Oliver, Reference Oliver1971). Adults possess more developed and specific features and thus are more amenable in establishing species level identifications (Ekrem et al., Reference Ekrem, Willassen and Stur2007). Larval keys usually identify chironomids only to genus level (Ram et al., Reference Ram, Banno, Gala, Gizicki and Kashian2014), and few chironomid keys exist that enable identification to species level. Identification of larvae to species level is usually accomplished by rearing larvae and collecting the pupae or adults (Inoue et al., Reference Inoue, Kimura, Denda, Tokioka, Tsushima, Amano and Hirabayashi2008). The fully developed pupae or adults are then morphologically identified in order to assign species level identifications for the corresponding larvae.
DNA barcoding has been instrumental in facilitating identification of cryptic larval chironomid species (Sharley et al., Reference Sharley, Pettigrove and Parsons2004, Pfenninger et al., Reference Pfenninger, Nowak, Kley, Steinke and Streit2007, Carew et al., Reference Carew, Marshall and Hoffmann2011). Studies combining the use of adult and larval DNA sequences have aided the species level identification of larvae within specific genera such as Cricotopus (Sinclair & Gresens, Reference Sinclair and Gresens2008), Cladopelma (Carew et al., Reference Carew, Pettigrove and Hoffmann2005), Procladius (Carew et al., Reference Carew, Marshall and Hoffmann2011) and Corynoneura (Silva & Wiedenbrug, Reference Silva and Wiedenbrug2014). However, these techniques have not been applied in describing the composition of diverse communities of chironomid larvae in the Great Lakes.
The present study specifically addresses the assessment of a benthic community from the standpoint of aquatic species monitoring and identification with its application in the Western Lake Erie region. Studies of larval and adult mitochondrial cytochrome c oxidase I (COI) genes have shown that DNA barcodes are an excellent tool provided that a comprehensive DNA barcode library at the species level is available. Such DNA libraries should contain a set of sequences that have been obtained from diverse larvae or adults that have been morphologically identified by expert taxonomists (Ekrem et al., Reference Ekrem, Willassen and Stur2007).
Although the public databases are quantitatively abundant, only a small number of sequences are useful in the Great Lakes region. Expansion of the reference databases in regions where chironomids are of interest would make species level identifications from molecular analysis more accessible and consistent. The present study uses newly determined sequences of taxonomically identifiable adult specimens to enrich the chironomid sequence database for Western Lake Erie and to improve identification of the diverse larval community in this region.
Materials and methods
Collection of larvae
Benthic organisms were collected from the Maumee Bay area of Western Lake Erie and the Toledo Harbor region of the Maumee River in the spring and summer of 2012, at 14 sites illustrated in fig. 1. On 2 and 30 May, 12 June, 12 July, 6 and 28 August, 2012 a total of 128 benthic samples were collected from 14 different sites . Benthic grab samples were obtained via a hand-operated bottom dredge (AMS, Ben Meadows, Janesville, WI), as previously described (Ram et al., Reference Ram, Banno, Gala, Gizicki and Kashian2014). Samples were collected from most sites on each date, washed on a 0.5 mm sieve to remove fine sediments while retaining organisms, and stored temporarily on ice in the field in 80% ethanol obtained by adding a fourfold volume of 100% ethanol (Fisher Science, Pittsburgh, PA). Upon delivery to the laboratory, samples were washed again on a 0.5 mm sieve and stored in 90% ethanol at 4 °C until shipment to EcoAnalysts, Inc (Moscow, ID) for sorting and morphological identification.
Fig. 1. Map of 14 benthic collection sites from 2012. Benthic communities of the Maumee Bay and Maumee River of the Western Lake Erie region (inset) were sampled at the lettered locations.
Morphological taxonomy of larvae
EcoAnalysts performed morphological identification of all organisms in each sample, sorting different taxa into separate vials. These animals included 2410 chironomid larvae that were mostly classified only to genus level by morphological features. Only a small proportion could be identified to species level, as described in results. Thousands of non-chironomid specimens were also observed and will be reported elsewhere.
DNA sequencing of larvae
Sorted, identified chironomid larvae were returned to the laboratory for molecular analysis. All specimens of rare taxa (those identified in each collection by Ecoanalysts fewer than five times) were sequenced. Among more common taxa, at least five specimens of each taxon randomly chosen from among the available specimens were sequenced from each collection. Due to particular interest in Cricotopus spp., known agricultural pests, all Cricotopus larvae that were collected were sequenced. The result of this selection, designed to obtain multiple representative sequences, whenever possible, from every taxon identified by Ecoanalysts, was that a total of 189 larvae out of the 2410 larvae collected were chosen for sequencing.
The full body, anterior aspect, and posterior aspect of chironomid larvae chosen for sequencing were photographed, and then sterile dissecting methods were used to obtain a small piece of tissue from the mid-portion of each selected specimen. Each dissected piece was placed in 30 μl of 100% ethanol in individual wells of a 96-well micro-plate and sent to the Canadian Center for DNA Barcoding (CCDB; Biodiversity Institute of Ontario, University of Guelph, Ontario, Canada) for sequencing of up to 658 base pairs of the mitochondrial COI gene using forward and reverse primers HCO2198 and LCO1490 (Folmer et al., Reference Folmer, Black, Hoeh, Lutz and Vrijenhoek1994, Hebert et al., Reference Hebert, Cywinska, Ball and deWaard2003a ). The anterior and posterior ends of each larva were retained as voucher specimens.
Collection of adults
Adults were collected from the Maumee Bay region, including from the hull and sides of the boat while the benthic collections already described were in process. Briefly, the flies were trapped via a hand-held vacuum cleaner and subsequently emptied into a series of 50 ml collection tubes containing isopropyl alcohol (pilot studies indicated that DNA was as readily obtained from specimens preserved in isopropyl as in ethyl alcohol). Adult flies were then sorted, and each individual was placed in its own vial containing 85% ethanol. Thirty-nine undamaged adults of diverse macroscopic characteristics were chosen for morphological and molecular analysis.
DNA sequencing of adult chironomids
For the first set of 20 specimens, two legs were detached from each adult chironomid and preserved in 90% ethanol for DNA analysis. DNA isolation was performed according to a Qiagen DNA spin-column_protocol (https://www.qiagen.com/us/resources/resourcedetail?id=21b4511b-4aaa-470a-a141-191ed91c54be&lang=en). Isolated DNA was amplified by polymerase chain reaction (PCR) using COI forward and reverse primers, HCO2198 and LCO1490 (Folmer et al., Reference Folmer, Black, Hoeh, Lutz and Vrijenhoek1994, Hebert et al., Reference Hebert, Cywinska, Ball and deWaard2003a ). The PCR product was purified and diluted with sterile PCR water to a concentration appropriate for sequencing. Genewiz (South Plainfield, NJ) sequenced the purified COI product in the forward and reverse directions. A consensus sequence was determined using DNAbaser software (DNA Baser Sequence Assembler v4.x 2014, Heracle BioSoft SRL, www.DnaBaser.com), evaluating chromatograms and aligning sequences from both directions. For the remaining 19 adults, two legs were removed from each fly and placed in a microplate that was sent to CCDB for COI sequencing, based on the protocol described above for larvae. Altogether, 37 of the 39 specimens belonged to the Chironomidae. The remaining two specimens, a culicid (Anopheles species) and a chaoborid (Chaborus punctipennis) were excluded from the analysis.
Morphological identification of adults
Morphological identification of the 39 specimens (minus the two legs used for molecular taxonomy) was completed at the Great Lakes Science Center (Ann Arbor, MI). The tissues were digested, and the remnants were mounted on slides to identify key morphological features. Initially, a specimen's size, color and shape were noted and then a pair of legs, wings and one antenna were mounted on a glass slide while the rest of the body was cleared of muscle tissue and then mounted. Keys by Townes (Reference Townes1945) and Cranston et al. (Reference Cranston, Dillon, Pinder and Reiss1989) were used to key adult specimens to genus. Species within a genus were identified using Townes (Reference Townes1945), Dendy & Sublette (Reference Dendy and Sublette1959), Roback (Reference Roback1971), (Epler, Reference Epler1988), Heyn (Reference Heyn1992), Saether (Reference Saether2009), and Saether (Reference Saether2011). When available, at least two specimens of each species were mounted and identified; additional specimens were studied for confirmation based on size, color and shape and returned to their labeled vials pending any possible need to mount and reconfirm based on DNA analysis.
Initial database search
All of our larvae and adult sequences were initially screened with the Barcode of Life Data Systems (BOLD) species level identification engine (http://www.boldsystems.org/index.php/IDS_OpenIdEngine) to determine if there were matches that differed by <3% in the BOLD database. If no match was found, the sequence was then subjected to a BLAST search of the National Center for Biotechnology Information (NCBI) GenBank database to confirm that no match existed in either database. If the result from BOLD included published results that also existed in the GenBank database, no further search was done. If a sequence had a database match differing by <3%, representative sequences from the matches were used as reference sequences in subsequent analysis.
Neighbor-joining analysis of larvae, adults and reference sequences
Neighbor-joining analysis was performed using MEGA software (Tamura et al., Reference Tamura, Peterson, Peterson, Stecher, Nei and Kumar2011). Pairwise differences of all larval sequences were calculated and analyzed graphically to determine natural groupings of sequences to define operational taxonomic units (OTUs) of sequences. COI barcode relationships were determined by constructing neighbor-joining trees and calculating pairwise differences using a maximum composite likelihood algorithm (Tamura et al., Reference Tamura, Peterson, Peterson, Stecher, Nei and Kumar2011).
In addition to the reference sequences chosen from the results of BOLD and GenBank searches, a database of useful species-level reference sequences was developed by downloading all sequences that registered as ‘chironomidae’ in the GenBank database and selecting quality sequences from among them. After aligning the potential reference sequences using Clustal W in MEGA, sequences that were too short or of poor quality (e.g., with multiple ‘N's, <75% contiguous, etc.) were removed from the analysis.
COI barcode relationships of larval sequences were inferred from these quality reference sequences, comparing 619 nucleotide positions. Since the analysis of average pairwise differences supported defining OTUs as clusters having no more than 3% pairwise differences within the OTU (see below), reference sequences that differed from the larval OTUs by more than 3% were removed and redundant reference sequences were eliminated. Subsequent neighbor-joining analysis defined OTUs as clusters having >97% identity. Any genetic grouping that is described in this paper as being the same species or OTU adhered to this standard. Sequences obtained from chironomid adults were then added to the analysis, creating a final tree with sequences from larval specimens, taxonomically identified adult specimens, and quality reference chironomid sequences from GenBank and BOLD .
Genus and species names, their authorities and years, and their family and subfamily identities were verified by reference to http://zipcodezoo.com/Key/ and through use of the NCBI taxonomy browser (http://www.ncbi.nlm.nih.gov/taxonomy/).
Results and conclusions
The numbers of larval and adult specimens that were classified by morphological criteria are summarized in table 1, along with a subset of specimens for which COI barcodes were sequenced. Altogether, 2410 larvae were identified morphologically by EcoAnalysts, revealing 23 genera, among which only 6 genera had specimens that could be identified to species level. Of the 2410 specimens, 189 were selected for sequencing, including 22 (11.6%) specimens that had been identified by EcoAnalysts to species (table 1). Out of 39 adult specimens subjected to morphological and molecular analysis, 37 were adult males, along with two other insect specimens. Of the 37 adult chironomid specimens 33 (89%) were identified to species level, comprising the 15 different species listed in table 1. COI barcode sequences were obtained from all 39 adult specimens. All of these newly identified sequences have been submitted to NCBI for inclusion in the GenBank database as accession numbers KP954634–KP954653 (adults), KR085203–KR085223 (adults) and KR085224–KR085412 (larvae).
Table 1. Species level morphological identification of adult and larval chironomids.
Pairwise differences of larval sequences
A histogram of pairwise differences of 189 larval sequences illustrated in fig. 2A shows structure that helped define OTUs for this study. Graphic analysis reveals 3 major peaks: (1) pairwise differences of <3% that we used in subsequent analysis to define OTUs; (2) a second peak between 3 and 6% pairwise differences. The pairs within this peak include four different genera (Cryptochironomus, Procladius, Microchironomus, and Dicrotendipes) for which the members of each pair had the same genus but whose species taxa had been designated only as ‘sp.’ by EcoAnalysts; and (3) pairwise differences that were >11%, peaking at ~20% as the most frequent pairwise difference. Pairs within this peak included both intragenus (e.g., Chironomus, 16%; Coelotanypus, 14%; Cryptochironomus, 14%; and Polypedilum 11%) and intergenus (e.g., Glyptotendipes and Procladius, 20%) differences.
Fig. 2. Histograms of pairwise distance values of (A) larval sequences and (B) all sequences (larvae, reference database and adults) combined.
Reference sequences
Out of over 2000 species level ‘chironomidae’ COI sequences downloaded from GenBank and BOLD, deletion of short or poor quality sequences left 1447 as reference database sequences for COI barcode analysis. Of these 1447 sequences, 11 non-redundant sequences differed by <3% from one or more larval and adult sequences and were used as reference sequences in constructing neighbor joining taxon-ID trees. Altogether, 250 larval, adult and reference database sequences were compared at 619 nucleotide positions. The histogram of pairwise differences retained its three-peak structure with the inclusion of the adult and reference database sequences (fig. 2B).
Neighbor-joining analysis larvae, adults and reference sequences
A neighbor-joining taxon-ID tree of all larval sequences contained 45 OTUs (fig. 3). Seven of these (15.6%) were supported by reference sequences from GenBank as shown in detail (fig. 4). OTUs in fig. 4 were identified as follows: Chironomus entis/plumosus (OTU 2), C. quinnitukqut/decorus (OTU 13), Micropsectra insignilobus (OTU 22), Paratanytarsus natvigi (OTU 23), Paratanytarsus grimmii (OTU 33), Cricotopus sylvestris (OTU 35) and Cricotopus bicinctus (OTU 36). OTUs 2, 23, 33, 35 and 36 were identified reliably based on close relationships to reference sequences from GenBank. However, OTU 13 grouped with both an adult and a reference sequence, the identification of which differed from each other and will be discussed below. For OTU 22, the genus of the reference sequences from GenBank differed from the genus of the larva morphologically identified by EcoAnalysts, a discrepancy that will also be discussed below. In OTU 2, two reference sequences enabled identification of both larvae and adults that had previously only been identified to genus level.
Fig. 3. Condensed neighbor – joining tree with maximum composite likelihood algorithm depicting 45 distinct larval operational taxonomic units (OTUs) based on mitochondrial cytochrome c oxidase I (COI) DNA sequences. Values at nodes correspond to a bootstrap 1000 test. Numbers in parentheses indicate the number of sequences within each branch. Numbers in blocks to the right of each branch correspond to the OTU numbers referred to throughout the rest of the paper. The analysis is based on 619 nucleotide positions in 189 larval sequences. The identifications are according to the highest taxonomic level achievable by EcoAnalysts. The scale represents fractional difference of sequence according to the length of the branch. For condensed branches, triangle height represents the number of sequences in the OTU, while width indicates the distance value corresponding to the lowest branch point.
Fig. 4. GenBank reference sequences identified seven larval operational taxonomic units (OTUs) to species level. Each subtree has its own scale in the bottom left corner. Numbers in boxes indicate the clade number that corresponds to the larval tree (fig. 3). For OTU 33, only the UKpi13 reference sequence of Paratanytarsus grimmii was within a 3% distance. The DBO3.5 reference sequence is included to indicate the scale.
Figure 5 shows 13 larval OTUs out of 45 (29%) that could be compared with sequences of identified adult chironomids, from which they differed by <3%. OTUs 4, 11, 15, 16, 37, 38, 39, 40 and 45 were identified reliably (Cryptochironmus digitatus, C. fulvus, Cladopelma viridulum, Glyptotendipes meridionalis, Procladius bellus, Ablabesmyia annulata, Tanypus stellatus, Coelotanypus scapularis, Procladius denticulatus, respectively). OTUs 1, 13, 21 and 25 (Chironomus crassicaudatus, C. decorus, Dicrotendipes lucifer, Stictochironmus devinctus, respectively) contain discrepancies when compared with adult and reference sequences. Of the 15 different chironomid species identified as adults (table 1), only Axarus festivus failed to have a corresponding sequence among the larvae. In addition, a Glyptotendipes senilis adult was taxonomically identified but not included in the phylogeny due to poor sequence quality and length. While 19 out of the 45 OTUs (42%) are identified, 26 out of the 45 still have no reference sequences or identified adults from which to assign the larvae with putative species identifications.
Fig. 5. Larval operational taxonomic units (OTUs) identified by adult sequences. Each subtree has its own scale in the bottom left corner. Numbers in boxes indicate the clade number that corresponds to the larval tree (fig. 3).
Based on COI sequence relationships, OTU 22 (fig. 4) shows a specimen classified by EcoAnalysts as a Cladotanytarsus sp. larva that is most likely to actually be M. insignilobus. In OTU 22, the claimed Cladotanytarsus larva falls within a large Micropsectra reference sequence clade, being most closely related to M. insignilobus as indicated by fig. 6. Furthermore, the sequence of this specimen differs from four other larvae also identified by EcoAnalysts as Cladotanytarsus sp. by a pairwise difference of ~20%. Contamination of this sample during sequencing by Microspectra DNA from another specimen is highly unlikely because no other Micropsectra larvae were collected or handled during the present study.
Fig. 6. A subtree extracted from the neighbor-joining taxon-ID tree constructed with GenBank reference and larval sequences before redundant sequences were removed. The relationship between the GenBank reference sequences and the sequenced larva in operational taxonomic unit (OTU) 22 that had been classified by EcoAnalysts as Cladotanytarsus sp. suggests that this specimen was most likely misclassified during morphological analysis, as its position within a large Micropsectra clade is evident.
Discussion
This study improves the reference databases of COI barcodes for chironomid larval identification by conducting a quality review of existing database sequences of chironomid COI barcodes and determining additional sequences from newly collected morphologically identifiable adult chironomids. This study also provides support, through its analysis of pairwise differences in COI barcodes, for using 97% identity as a natural amount of within-species variation defining chironomid OTUs. Sequences provide specific, reliably generated data for classifying specimens. Nevertheless, as will be discussed, the need for referencing the sequences to potentially ambiguous morphological identifications and to databases that may vary in sequence quality and taxonomic reliability means that ambiguities, inconsistencies, and errors may still occur, and care must be taken in using sequence data for identification.
Improvement of chironomid reference databases
A search for high quality reference sequences in GenBank to identify 45 OTUs of chironomid larvae specimens from Maumee River and the Maumee Bay area of Western Lake Erie revealed that only 15.5% of sequenced larvae OTUs had corresponding reference sequences in GenBank. In order to achieve even that degree of identification this study reviewed over 2000 chironomid sequences in GenBank to assure selection of only high quality sequences of sufficient length and adequate annotation. This study adds 33 new reference sequences to the public database, based on careful taxonomic identification and COI barcoding of adult specimens. Addition of the sequences from taxonomically identified adult specimens to these databases aided in identification of OTUs 1, 4, 11, 15, 16, 21, 25, 37, 38, 39, 40 and 45, which would not have been possible prior to this study. Nevertheless, more work needs to be done, as 26 of the 45 OTUs in the tree have yet to be identified. In addition, we have yet to collect larvae with sequences corresponding to the adult specimens of A. festivus. Absence of these corresponding larvae in our dataset could indicate that A. festivus larvae occupy different or harder substrate habitats than those we sampled. In any case, addition of A. festivus to the GenBank database will make future identifications of Axarus larvae possible.
Relationship of taxa to pairwise differences
Data in fig. 2 show gaps in the distribution at around 3% difference and between 6 and 10%. A threshold of 3% for species differentiation has been used for various animal groups (Hebert et al., Reference Hebert, Cywinska, Ball and deWaard2003a , Hebert et al., Reference Hebert, Ratnasingham and de Waard2003b , Sinclair & Gresens, Reference Sinclair and Gresens2008). In some cases, different species appear within the same COI cluster (e.g., OTU 24 and 25). Although these specimens remain within the confines of the 3% threshold designated by molecular morphology, the taxonomic designations do not agree. Evolutionary processes such as hybridization or rapid speciation where divergent mutations have not yet accumulated may explain this phenomenon (Michailova & Fischer, Reference Michailova and Fischer1986, Ekrem et al., Reference Ekrem, Willassen and Stur2007, Proviz, Reference Proviz2008, Abbott et al., Reference Abbott, Albach, Ansell, Arntzen, Baird, Bierne, Boughman, Brelsford, Buerkle, Buggs, Butlin, Dieckmann, Eroukhmanoff, Grill, Cahan, Hermansen, Hewitt, Hudson, Jiggins, Jones, Keller, Marczewski, Mallet, Martinez-Rodriguez, Möst, Mullen, Nichols, Nolte, Parisod, Pfennig, Rice, Ritchie, Seifert, Smadja, Stelkens, Szymura, Väinölä, Wolf and Zinner2013). The second peak in pairwise differences, between 3 and 6%, suggests that some chironomid groups may have a greater within species variation. The question also arises: could the specimens in the region between the two larger histogram peaks represent instances of incipient speciation? In most cases, in our data pairwise differences >11% clearly represent different species and usually different genera. However, cases where apparently the same species has pairwise differences this large might also represent cryptic species (Anderson et al., Reference Anderson, Stur and Ekrem2013), revealed by large intraspecific pairwise differences. Polypedilum halterale, represented by OTU 31 and 32, which differ by 11%, is one such example.
Ambiguities, inconsistencies and possible errors
Despite the care with which larvae and adults were sequenced and the adult specimens were identified, the results shown here exhibit several ambiguities, inconsistencies and possible database or identification errors. Examples of ambiguities in identifying larvae are OTU 13 and 21 (figs. 4 and 5). In OTU 13, a reference sequence and a sequenced adult were classified as different species within the genus Chironomus. For some Chironomus species, the COI barcode and morphological identification alone may be inadequate for establishing species level identifications (Ekrem et al., Reference Ekrem, Willassen and Stur2010, Proulx et al., Reference Proulx, Martin, Carew and Hare2013) For example, C. quinnitukqut is a part of the C. decorus group, and species within this group are often separated based on karyotype analysis (Martin et al., Reference Martin, Sublette and Caldwell2011). In OTU 21, Dicrotendipes simpsoni represents a species within the Dicrotendipes lucifer complex. Similarly, difficulty in assigning species level identifications to chironomids within the genus Cricotopus exists, specifically within the C. sylvestris species group (Gresens et al., Reference Gresens, Stur and Ekrem2012). Assignment of species level identifications within these types of difficult genera will be ambiguous unless additional markers are utilized, techniques such as karyotyping are incorporated, or morphological keys are improved.
Ambiguity is also present in OTU 2 (fig. 4) and OTU 24 (fig. 3). For OTU 2, identification as C. entis is evident, yet examination of a karyotype could place it as a closely related species, such as Chironomus borokensis Kerkis, Filippova, Shobanov, Gunderina & Kiknadze, 1988 (Proviz & Bazova, Reference Proviz and Bazova2013). Both of these taxa belong to the C. plumosus group, which contains sibling species differentiated by karyotype (Kiknadze et al., Reference Kiknadze, Butler, Golygina, Martin, Wulker, Sublette and Sublette2000, Golygina & Kiknadze, Reference Golygina and Kiknadze2012). For OTU 24, six larval specimens with identical sequences were classified as Cladotanytasus sp. (4 specimens), Stempellinella sp. (1 specimen), and Cryptochironomus sp. (1 specimen). Since this OTU is fully within a clade in which identification of most other specimens are in the tribe Tanytarsini, and Cryptochironomus is not classified as a member of this tribe, the Cryptochironomus identification is likely to be in error. Whether the correct designation of OTU 24 is Cladotanytarsus or Stempellinella remains ambiguous.
In some cases, the assignment of a larval taxon by qualified taxonomists (e.g., those who work for EcoAnalysts) even at the genus level is inconsistent with identifications derived from molecular data. In the present study, the morphological identification of some larval specimens in OTUs 1, 13, 21, 22 and 25 did not agree with the identification determined from GenBank reference or adult chironomid COI sequences. Morphological identification may be compromised by condition and maturity of specimens, preservation and inadequate reference materials available to taxonomists. It is also possible that after identification a specimen might get cross-contaminated by DNA from another specimen or a specimen or vial might be mislabeled, leading to such inconsistencies. Avoiding errors is essential to use this process effectively. To as great an extent possible, we consulted expert taxonomists and sterilized dissecting equipment with ethanol and flame between specimens-to-minimize the risk of sample-to-sample cross-contamination prior to subsequent barcoding.
Since mistakes in reference databases could also lead to ambiguities or incorrect identification, we consider here several instances in which reference databases may have errors. For example, in fig. 4, OTU 33 has two Paratanytarsus sp. larval sequences grouping with a 100% identical P. grimmii reference sequence; however, another P. grimmii reference sequence (label: DBO3.5) is >3% divergent and is suspected to be incorrectly labeled as P. grimmii. Taxon-ID trees of quality reference sequences constructed without the inclusion of adult sequences (figs. 7 and 8) reveals that the highly divergent ‘Paratanytarsus grimmii’ sequence is the only one out of 23 very closely related sequences with this designation, leading us to suspect that this identification is likely incorrect (fig. 7). Instead, the correct identity of this reference sequence is more likely to be P. kathleenae, in agreement with the other 22 specimens in this clade (fig. 8).
Fig. 7. The neighbor-joining taxon-ID tree with reference sequences only, which was constructed as an intermediate step in obtaining the final tree.
Fig. 8. Paratanytarsus sequence relationships for a part of the uncut reference sequence tree.
Despite the existence of such ambiguities, inconsistencies and possible errors, the clades of the taxon-ID tree of larvae in fig. 3 mostly show excellent congruence with previous morphological taxonomic classification to the family, subfamily, or tribe levels. Thus, the clades of OTUs 1–21 and 30–33 contain genera that are all classified as subfamily Chironominae in the tribe Chironomini; genera of OTUs 37–45 are all members of the subfamily Tanypodinae; and genera of OTUs 34–36 are all members of the subfamily Orthocladiinae. Only OTUs 22–29 represent an exception to this general congruence of molecular clades with known subfamilies and tribes: while the majority of these genera are classified as subfamily Chironominae of the tribe Tanytarsini, exceptions are one specimen in OTU 24 and 5 specimens of OTU 25, whose genera are classified as members of the tribe Chironomini. The mix of tribes within this clade may simply indicate the difficulty of determining larval morphology or it may reflect identification errors.
Significance of identifying chironomid larvae
Species level identification of chironomid larvae is useful due to the importance of larvae in aquatic food webs (Oliver, Reference Oliver1971, Armitage et al., Reference Armitage, Cranston and Pinder1995) and the negative impacts of some species as pests (Ali, Reference Ali1996, Broza et al., Reference Broza, Halpern, Gahanma and Inbar2003) and potential invaders (Failla et al., Reference Failla, Vasquez, Fujimoto and Ram2015). For example, OTU 33 (fig. 4) confirms the presence of P. grimmii, a parthenogenic nuisance species known for colonizing water treatment ponds and their associated equipment, such as granular activated carbon absorbers (Langton et al., Reference Langton, Cranston and Armitage1988, Olsen et al., Reference Olsen, Leadbeater, Callow, Holden and Bale2009). Also, OTUs 35 and 36 identify the presence of two species of Cricotopus that are both known to be colonizers and detrimental pests of rice fields in California (Clement et al., Reference Clement, Grigarick and Way1977).
In addition to identifying known nuisance species, this method has the potential to identify the presence of new species that are either previously undetected or invasive. For example, OTU 22 indicates the presence of M. insignilobus, a species associated with waters of low organic content that has a very limited record of detection in the Great Lakes, but whose distribution is well described in Northern Europe (Saether, Reference Saether1979, Ilyashuk & Ilyashuk, Reference Ilyashuk and Ilyashuk2001). To our knowledge, no previous records exist of this species in Lake Erie, although new species of the Micropsectra genus, such as M. subletteorum, a sister species of M. insignilobus, have recently been described in Minnesota via similar use of molecular and morphological methods (Anderson et al., Reference Anderson, Stur and Ekrem2013).
Future research needs
Despite the possibility of ambiguities, inconsistencies and reference database errors, we recommend the molecular barcode methods used in this study to identify chironomid larvae in future studies. Disagreements in molecular identifications exist, suggesting that developing a more comprehensive library of diverse genetic markers and employing additional identification techniques, such as karyotyping, may resolve some issues. Because some OTUs are only identified by one reference sequence, confirming their identity with barcodes of replicate reference specimens would be beneficial. In addition, improving the quality of existing databases is needed. To some degree, errors were avoided by sorting out sequences that had low quality scores or many ambiguous bases (N's, for example). COI is able to provide presumptive species-level identifications in many cases and in general is considered accurate and reliable (Silva et al., Reference Silva, Ekrem and Fonseca-Gessner2013, White et al., Reference White, Pilgrim, Boykin, Stein and Mazor2013). Nevertheless, COI does not provide as great a resolution as CAD or Cyt b genes (Ekrem et al., Reference Ekrem, Willassen and Stur2010, Carew et al., Reference Carew, Marshall and Hoffmann2011). The use and establishment of other DNA markers in the future could contribute significantly to the reference data and improve the field of DNA-based taxonomic identification.
The larvae and adults in this study were collected from just one small portion of Lake Erie and the Maumee River. For studies throughout the Great Lakes, reference databases ought to be established for specimens elsewhere in the region since reference sequences are likely to be region-specific. While we have expanded the number of reference sequences available for this relatively little studied portion of Western Lake Erie, if this study were performed in Scandinavia, a heavily studied region, the number of reference sequences that would match sequences of environmentally collected midge sequences would likely be higher. Establishing barcodes and analyzing the phylogeny is important in regions where chironomids have not been heavily studied in this manner (Bergsten et al., Reference Bergsten, Bilton, Fujisawa, Elliott, Monaghan, Balke, Hendrich, Geijer, Herrmann, Foster, Ribera, Nilsson, Barraclough and Vogler2012).
In order to improve the results obtained from the methods of chironomid identification used in this study, a more comprehensive collection and identification of adult flies within the Western Lake Erie region needs to be done. For future studies, obtaining a larger sample size and a more diverse assortment of adult flies from the region in question may increase the number of identifiable larval OTUs. Establishing quality reference sequences that are supported by professional taxonomists is integral to utilize this process. As more quality sequences are submitted to public databases species identifications based on molecular taxonomy will be more accessible.
Because chironomid larvae contribute significant biomass and diversity to aquatic ecology, it is important to have reliable methods of species level identification. The use of barcodes from adult midges collected from Western Lake Erie helped to resolve the species level identity of several larval clades collected in benthic samples. The present study validated the use of adults for further resolution of larval species identification. Our study enhances existing work in regions where chironomid populations are prominent and allows species level identifications to be more reliable, accurate and accessible. Establishing a comprehensive reference database of multiple DNA barcodes using reliable, cross-referenced adults identified by expert taxonomists, as was done in this study could potentially resolve problems of species level taxonomy of larvae in the family Chironomidae.
Acknowledgements
This project was supported by grants from the Environmental Protection Agency (Grant number GL00E00808-0) and funds from Wayne State University. We thank Tjorborn Ekrem for his advice and communication, as well as helpful reviewers for aiding in improving the presentation of our research. This article is contribution 1945 of the U.S. Geological Survey Great Lakes Science Center.
