In this paper, we review the history of Hydropsychinae genus-level classification and nomenclature and present new molecular evidence from mitochondrial cytochrome c oxidase subunit I (COI) and nuclear large subunit ribosomal ribonucleic acid (28S) markers supporting the monophyly of the genus Hydropsyche. Both molecular and morphological characters support a broad conservative definition of Hydropsyche. Caledopsyche, Hydatomanicus, and Occutanspsyche are synonymized with Hydropsyche. The following species groups are established: Hydropsyche bronta Group (generally corresponding with Ceratopsyche and Hydropsyche morosa and newae Groups), Hydropsyche colonica Group (generally corresponding with Orthopsyche), Hydropsyche instabilis Group (generally corresponding with Hydropsyche s.s.), and Hydropsyche naumanni Group (generally corresponding with Occutanspsyche). Molecular data recovered Hydromanicus as paraphyletic, and Cheumatopsyche and Potamyia as sister taxa. The genus names Plectropsyche and Streptopsyche are reinstated.
Hydropsychid caddisflies (Trichoptera:Hydropsychidae) are critical components of biomonitoring programs throughout their geographical range because of their high abundance and wide range of pollution tolerance values among species. However, the classification of genera within the subfamily Hydropsychinae has been subject to shifting generic nomenclature. Stable nomenclature systems are the foundation for biologists who use comparative biology to study the evolutionary history of freshwater bioindicators like caddisflies. When nomenclatural changes obscure or do not reflect monophyly, the power of phylogeny as a comparative framework diminishes. As taxonomic hypotheses change, it is important to preserve “cognitive value” and monophyly (Schefter 2005) with broad generic definitions across the global geographic range of a group, rather than relying on regional gaps in morphological and ecological characteristics that can lead to the elevation of regional species groups to nonmonophyletic genera.
Thus, our goal was to examine molecular and morphological characters to provide evidence supporting a stable genus-level nomenclature for the subfamily Hydropsychinae and the Hydropsyche sensu lato (s.l.) lineage. Historically, Hydropsyche s.l. has included: Abacaria, Aoteapsyche, Caledopsyche, Ceratopsyche, Herbertorossia, Hydatomanicus, Hydronema, Hydropsyche (Hydropsyche), Hydropsyche (Occutanspsyche), Mexipsyche, Orthopsyche, and Symphitopsyche. If monophyletic, Hydropsyche s.l. is one of the most speciose lineages in all of Trichoptera with >500 described species (Morse 2009). Its members are found in Holarctic, Oriental, Afrotropical, and Australasian (minus Australian) streams and rivers, and their larvae exhibit a wide range of pollution tolerances (Resh and Unzicker 1975, Lenat 1993, Lenat and Resh 2001). Despite the ecological importance of the group, the evolutionary history of Hydropsyche s.l. has been obscured by the lack of: 1) a universally accepted definition of the genus Hydropsyche, 2) knowledge of the larvae, pupae, and females of most species, and 3) support for the phylogenetic position of Hydropsyche within the subfamily Hydropsychinae (Geraci et al. 2005, Schefter 2005).
The history of Hydropsychinae generic classification has included much debate on the meaning of the names Hydropsyche, Symphitopsyche, and Ceratopsyche. The genus Hydropsyche Pictet, 1834, has been split into 10 genera and 3 subgenera (reviewed by Schefter 2005). Ulmer (1907, 1951, 1957), Mosely (1941), McFarlane (1976), and Ross and Unzicker (1977) all described new genera based on adult males whose genitalia differed from Hydropsyche sensu stricto (s.s.) males. In North America, the above genus names also corresponded to larval characters (Schuster and Etnier 1978, Schuster 1984, Schefter and Wiggins 1986), but larvae in most other parts of the world remain largely undescribed or unassociated, and hence their characters states unknown. Schmid (1979) broadly defined the genus Hydropsyche, while noting that, at the time, it was the most morphologically homogeneous genus in all Trichoptera and that splitting the genus amounted, in his opinion, to taxonomic inflation. Hydropsyche bronta Ross, 1938 (the type species of Ceratopsyche), exemplifies the differing opinions on Hydropsyche nomenclature: this species has had 3 generic, 1 subgeneric, and 3 species-group names in its history (Table 1). These nomenclatural debates have resulted in some authors using Ceratopsyche as a genus (Merritt et al. 2008) or subgenus name (Tian et al. 1996), whereas others have rejected it altogether and have referred instead to Hydropsyche species groups (Schefter and Wiggins 1986, Malicky and Chantaramongkol 2000, Mey 2003, Olah and Johanson 2008). Additional nomenclatural systems that have been used in major works on Hydropsyche s.l. across biogeographical regions are summarized in Table 2.
Table 1
Various classification schemes for the species originally described as Hydropsyche bronta Ross 1938.
Table 2
Varying taxonomic treatments of species previously classified in the genera Hydropsyche, Ceratopsyche, Mexipsyche, and Symphitopsyche. These classifications do not fully overlap because they do not encompass all species or species groups, but they do represent a significant portion of the fauna and terminology used. Mey (2003) used the term “clade” to refer to a subdivision within a species group and Olah and Johanson (2008) used the term “cluster” to refer to a subdivision within a species group.
The first hypothesis of phylogenetic relationships among Hydropsychinae genera (Fig. 1A) was based on male genitalic characters (Ross and Unzicker 1977), but it did not consider the world fauna or outgroup relationships. Other studies of Hydropsychinae classification have focused on immature (Schefter and Wiggins 1986, Wiggins 1996) or adult male stages (Schmid 1964, Mey 1998, 2003, 2005, Schmid 1998, Malicky and Chantaramongkol 2000, Olah and Johanson 2008). Based on a parsimony analysis of morphological characters from all life stages except eggs, Schefter (2005) suggested synonymizing 5 genera with Hydropsyche (Fig. 1B, node C), but did not change genus-level nomenclature designations. Olah and Johanson (2008) examined morphological characters in a comparative (but non-matrix-based) study of Hydropsychinae, and formally synonymized: 1) Ceratopsyche, Herbertorossia, Mexipsyche, and Symphitopsyche with Hydropsyche, 2) Aeoteapsyche with Orthopsyche, 3) Plectropsyche with Cheumatopsyche, 4) Streptopsyche with Calosopsyche, and 5) Hydatomanicus and Hydatopsyche with Hydromanicus. Thus, 2 different interpretations of similar morphological characters (Schefter 2005, Olah and Johanson 2008) suggested 2 different genus-level nomenclature systems for Hydropsychinae.
We used molecular data from fragments of the mitochondrial (mt) cytochrome c oxidase subunit 1 (mtCOI) and nuclear large subunit ribosomal ribonucleic acid (rRNA) (28S) genes to infer phylogenetic relationships within Hydropsychinae. Our objectives were to test the monophyly of the Hydropsyche (sensu Schefter 2005) lineage with multiple genes, and, in so doing, to examine outgroup relationships with reference to Hydropsyche. This approach allowed us to compare the deoxyribonucleic acid (DNA)-based topology with that inferred from parsimony analysis of morphological characters (Fig. 1B; Schefter 2005) and to test whether previously designated nomenclature systems communicated monophyly.
Methods
DNA sequences were obtained for as many representative species of Hydropsyche sensu Schefter (2005) as possible. Additional exemplars for DNA sequencing were chosen to maximize species group representation (Schefter and Wiggins 1986, Malicky and Chantaramongkol 2000, Mey 2003, Olah and Johanson 2008). Specimens were obtained from the Clemson University Arthropod Collection, Nanjing Agricultural University (China), Smithsonian's National Museum of Natural History, the University of Minnesota Insect Collection, and Dr. Hans Malicky. Additional DNA sequences for Hydropsychinae taxa were downloaded from GenBank (Kjer et al. 2001, Zhou et al. 2007). Our analysis consisted of 2 tiers. For the 1st tier, we sequenced the D2 variable region of the nuclear 28S rRNA gene and 657 base pairs (bp) of the COI gene region for 60 Hydropsychinae exemplars (D2COI data set; Appendix 1). These gene fragments were chosen because they trace species and genus boundaries reliably for hydropsychid caddisflies (Zhou et al. 2007) and provide phylogenetic signal at both shallow (COI) and deeper (28S) phylogenetic levels (Kjer et al. 2001, 2002). COI data were generated in collaboration with the Trichoptera Barcode of Life Campaign ( http://www.trichopterabol.org).
A 2nd -tier data set was assembled consisting of the COI gene sequence and the sequences for the D1, D2, and D3 regions of the 28S rRNA gene (28SCOI data set; Appendix 2) for 12 species representing all available Hydropsychinae genera (both currently and previously recognized genera). The D1 and D3 regions were included with D2 because they varied little in length, had fewer alignment-ambiguous nucleotides, and thus, potentially provided more characters at deeper levels of phylogeny. Three species of Hydromanicus were included that represented types 1 (Hydromanicus nr. truncatus Betten) and 2 (Hydromanicus nr. canaliculatus Li, Tian, and Dudgeon) from Schefter's (2005) analysis. Fresh specimens of Abacaria, Hydromanicus seychellensis Ulmer (Hydromanicus type 3, Schefter 2005), Hydronema, and Schmidopsyche were not available for DNA analysis. The D2 fragment from Calosopsyche continentalis Flint & Bueno-Soria or Cheumatopsyche hoogstraali (Ross) (referred to hereafter and in figures as Plectropsyche hoogstraali Ross to reflect the updated nomenclature described below) and the D1 and D3 fragments from Hydromanicus nr. canaliculatus Li, Tian, and Dudgeon or Hydropsyche instabilis (Curtis) could not be sequenced.
DNA extraction, polymerase chain reaction, sequencing
Genomic DNA was extracted from either 1 leg or from the entire animal using Qiagen DNeasy Kits (Qiagen, Hilden, Germany) and standard protocols. For some specimens the following modifications were made. An initial volume of 20 µL of Proteinase K was added to the Qiagen ATL buffer, and the legs or entire animals were incubated at 55°C for 24 to 48 h. An additional 20 µL of Proteinase K was added to the buffer every 24 h. Polymerase chain reaction (PCR) amplification of the 28S rRNA fragment was performed on 1 µL of genomic DNA from each species in 25-µL reactions according to the following recipe: 12.5 µL of Qiagen Taq PCR Master Mix, 5.0 µL Qiagen Q solution, 1.0 µL of each 10-µmol oligonucleotide primer, and 4.5 µL of double distilled (dd) H2O. The primers used were D1–UP ([5′-GGAGGAAAAGAAACTAACAAGGATT-3′] paired with D1–DN [5′–CAACTTTCCCTTACGGTACT–3′]; D2UP–4 [5′–GAGTTCAAGAGTACGTGAAACCG–3′] paired with D2DN–B [5′–CCTTGGTCCGTGTTTCAAGAC–3′]); and D3–UP [5′– ACCCGTCTTGAAACACGGAC–3′]) paired with D3–DN [5′–CTATCCTGAGGGAAACTTCGGA–3′]).
Purified PCR products were sequenced on an ABI 3730XL or 3130XL sequencer (Applied Biosystems, Foster City, California) using BigDye® Terminator v3.1 (Applied Biosystems) and standard reaction parameters. Each gene fragment was sequenced bidirectionally and then assembled as contig files using either Sequencher (v.4.0.5; Gene Codes Corporation, Ann Arbor, Michigan) or LaserGene (v.6; DNASTAR, Inc., Madison, Wisconsin). Acquisition of the COI barcode region was done at the Canadian Centre for DNA Barcoding, University of Guelph, Canada. Standard barcoding protocols were followed (Ivanova et al. 2006, deWaard et al. 2008). Genomic DNA was extracted using an AcroPrepTM 96 1-mL filter plate (PALL) with 3.0-µm glass fiber (Ivanova et al. 2006). DNA was eluted in 40 µL of distilled (d) H2O. Full-length COI barcodes were amplified using 2 primer sets: LepF1 ([5′–ATTCAACCAATCATAAAGATATTGG–3′]/LepR1 [5′–TAAACTTCTGGATGTCCAAAAAATCA–3′]) (Hebert et al. 2004) and LCO1490 ([5′–GGTCAACAAATCATAAAGATATTGG–3′])/HCO2198 [5′–TAAACTTCAGGGTGACCAAAAAATCA–3′]) (Folmer et al. 1994). Standard DNA barcoding protocols were followed for COI sequencing as described by deWaard et al. (2008) and Hajibabaei et al. (2005). COI barcodes and detailed specimen information were deposited in the Barcode of Life Data (BOLD) Systems (Ratnasingham and Hebert 2007) as part of the Trichoptera Barcode of Life Campaign ( http://www.trichopterabol.org).
Alignment
Edited 28S rRNA D1 and D3 sequences were aligned following the Trichoptera secondary structural model provided by Kjer et al. (2001). Alignment of hydrogen-bonded stems and stem-and-loop numbering for the D2 fragment followed models available at: http://hymenoptera.tamu.edu/rna/index.php (Gillespie 2004, Gillespie et al. 2005). Regions of expansion and contraction (REC) and regions of ambiguous alignment (RAA) were excluded from the analysis (aligned DNA data set available from http://rci.rutgers.edu/∼insects/pdata.htm).
Phylogenetic analyses
Bayesian analysis was performed with MrBayes (v.3.1.2; Ronquist and Huelsenbeck 2003). Gaps were coded as “–”, and missing data were coded as “?” for all analyses. The consensus tree produced in each analysis was rooted a posteriori with Calosopsyche parander (Botosaneanu) (referred to hereafter as Streptopsyche parander (Botosaneanu) and in figures to reflect the updated nomenclature described below) because this species was found to be the sister taxon to the rest of Hydropsychinae in previous analyses (Geraci 2007).
D2COI data set
The data were partitioned into 28S rRNA (424 nucleotides [nts] including gaps) and COI (657 nts), and 2 different model schemas were used. In the first Bayesian analysis, the general time reversible + time invariant + Γ (GTR+I+Γ) model was applied to both partitions, as recommended by MrModeltest (v.2.2; J. A. A. Nylander, Uppsala University, Uppsala, Sweden). A Mixed GTR–Codon model was used in a 2nd analysis, with the GTR model applied to the 28S rRNA partition and the Codon model to the COI partition. The GTR model had 6 Γ rate categories, whereas the Mixed model had 4 categories. Both analyses were run with default values for other model prior parameters; revmat, statefreq, shape, and Pinvar were unlinked. Each analysis had 4 Metropolis-coupled Markov Chain Monte Carlo (MCMC) chains (3 heated and 1 cold) that were run for 5 million generations (with 10% of the trees discarded as burn-in). GARLI (v.0.951; Zwickl 2006) was used to analyze the D2COI 60-taxon data set using the maximum likelihood criterion under the GTR model with default parameters. The most likely tree topology was rooted a posteriori with Streptopsyche parander (Botosaneanu) and right-ladderized using TreeView (v.1.6.6; Page 1996). HyPhy (v.0.99; Kosakovsky Pond et al. 2004) then was used to calculate nonsynonymous substitution rates for each branch. Likelihood parameters were optimized on the GARLI maximum likelihood topology based on the D2COI data set for 60 taxa. Parsimony analyses were done with PAUP (v.4.10b; Sinauer, Sunderland, Massachusetts; Swofford 1999). Heuristic searches with tree bisection and reconnection (TBR) branch swapping were done for each data set, and strict consensus trees were constructed for each analysis. Bootstrap analyses were run for each data set (10,000 replicates, stepwise addition).
28SCOI data set
The data were partitioned into 28S rRNA (979 nts including gaps) and COI (657 nts). Two analyses were run using MrBayes. The GTR model analysis applied the GTR+I+Γ model to both partitions. The Mixed model analysis applied the GTR+I+Γ model to the 28S rRNA partition, and the Codon model to the COI partition. The GTR model had 6 Γ rate categories in each analysis, whereas the Codon model had 4 Γ rate categories. Default values were used for all other parameters, and revmat, statefreq, shape, and Pinvar all were unlinked. Four Metropolis-coupled MCMC chains (3 heated and 1 cold) were run for 3 million generations (with 10% burn-in) for the GTR model analysis, and for 10 million generations (with 20% burn-in) for the Mixed model analysis. PHASE (v.2.0; Hudelot et al. 2003) also was used to analyze the 28SCOI data set in a Bayesian framework to enable partitioning of 28S rRNA stem-and-loop regions. The 28S rRNA data were partitioned into loops and hydrogen-bonded stems according to secondary structure, and the COI data were partitioned into codon positions. The RNA7+I+Γ model with 6 Γ categories was applied to hydrogen-bonded stems, the reverse + time invariant + Γ (REV+I+Γ) model with 6 Γ categories was applied to loops, and the YNH98 codon model was applied to the COI partition. The MCMC chains were run using a random start chain and model parameters for 1 million burn-in iterations and 10 million sampling iterations (sampling period = every 100 iterations).
Results
All analyses recovered a monophyletic Hydropsyche clade that subsumes these previously established genus-group names: Aoteapsyche, Caledopsyche, Ceratopsyche, Herbertorossia, Hydatomanicus, Hydropsyche (Hydropsyche), Hydropsyche (Occutanspsyche), Mexipsyche, and Orthopsyche. The consensus trees produced by both GTR and Mixed GTR/Codon model Bayesian analyses of the D2COI data set for 60 taxa recovered 100% posterior probability (p.p.) support for Hydropsyche (Fig. 2A, B). The topology of Hydropsyche derived from molecular data is congruent with the parsimony-derived topology inferred from morphology (Schefter 2005) except for the placement of the Caledopsyche exemplary species (Hydropsyche atalanta (Schefter & Ward) and H. CJG sp. NC2). Caledopsyche was not erected based on genitalic characters, but rather on a wing vein autapomorphy (Kimmins 1953). However, DNA characters support the placement of Caledopsyche species within Hydropsyche. Chinese species previously classified as Mexipsyche (Hydropsyche grahami A, C, and G Banks; H. furcula Tian & Li) did not form a monophyletic lineage, and Hydropsyche grahami Banks might contain a series of cryptic lineages, an observation supported by a larger data set of COI barcodes (XZ, unpublished data). Olah and Johanson (2008) synonymized Hydatomanicus with Hydromanicus, but DNA data supported with 100% p.p. Hydromanicus ovatus (previously Hydatomanicus ovatus) (Li, Tian, & Dudgeon) as belonging to Hydropsyche (Fig. 2A, B). This conclusion also is supported by larval morphology (Zhou 2007). The placement of Hydropsyche ovatus (Li, Tian, & Dudgeon) as the basal species within the H. instabilis Group in both Bayesian analyses (Fig. 2A, B) suggests that it belongs to that species group. Exemplars of H. (Hydrocheumatopsyche) Marlier were not available to us, so that subgenus distinction is retained here. DNA from Hydromanicus seychellensis Ulmer, African Symphitopsyche, or any Abacaria or Hydronema exemplars were unavailable to us, so the current nomenclature for those groups is maintained here.
Bayesian analyses of the 28SCOI data set also recovered 4 Hydropsyche sensu Schefter (2005) exemplars as a monophyletic clade with 100% p.p. support (Fig. 3A, B). Parsimony analyses supported Hydropsyche as monophyletic except when only COI nucleotides were used (data not shown). HyPhy analysis recovered a nonsynonymous substitution rate for the Hydropsyche branch (0.093) that was similar to that of the branches for other genera within Hydropsychinae (Fig. 4). The only branch that had a higher rate (0.125) was for Cheumatopsyche. Further examination of the translated amino acids revealed 2 unreversed substitutions at positions 136 and 172 (out of 219 total amino acids translated from the 657-nucleotide COI fragment; Fig. 4). These COI amino acid substitutions are synapomorphies for Hydropsyche, as defined here (see Genus diagnosis below).
Bayesian analyses for the 28SCOI data set recovered conflicting topologies with regard to relationships among Cheumatopsyche, Potamyia, and Hydromanicus s.s. (Schefter's type 2; Fig. 3A, B). The Mixed model analyses that accounted for COI codon position recovered the clade (Cheumatopsyche spp. + Potamyia spp.) as the sister taxon to Hydropsyche but with only 79% p.p. support (Fig. 3A). However, the GTR model analysis recovered (Hydromanicus canaliculatus + Hydromanicus melli) as sister to (Cheumatopsyche spp. + Potamyia spp.) with 70% p.p. support, but strongly (defined here as >95% p.p.) supported that clade as sister to Hydropsyche (Fig. 3B). The conflicting topologies recovered by the GTR vs Mixed models for the larger taxon sample (Figs 2A vs 3B) mirrored this uncertainty in outgroup relationships. Schefter (2005) recovered Hydronema as sister to Hydropsyche, as it was defined at the time. The sister taxon to Hydropsyche remains equivocal because specimens of Hydronema were not available for DNA extraction. However, Cheumatopsyche and Potamyia were recovered as sister taxa with 100% p.p. support (Figs 2A, B, 3A, B). This relationship is not congruent with the parsimony topology inferred from morphology (Fig. 1B) of Schefter (2005), but it produces the same genus-level nomenclature system.
The 28S rRNA and COI data also suggest that Hydromanicus (sensu Olah and Johanson 2008) is not monophyletic and needs formal taxonomic revision (Figs 2A, B, 4). The paraphyly of Hydromanicus also was supported by morphology data (Fig. 1B; Schefter 2005) and by examination of larval characters of some Chinese Hydromanicus. The larvae of Hydromanicus canaliculatus Li, Tian, & Dudgeon and H. melli (Ulmer) share synapomorphies (e.g., head glabrous, anterior margin of frontoclypeal apotome asymmetric, deeply excised, etc.), and both are distinctly different from those of Hydromanicus nr. truncatus Betten (Zhou 2007). Last, we revert to the previous nomenclature for Streptopsyche parander (Botosaneanu) and Plectropsyche hoogstraali Ross because our topology and resulting classification was congruent with that inferred from morphological characters in a parsimony framework (Schefter 2005). Neither study supported the synonymy of Streptopsyche with Calosopsyche or the synonymy of Plectropsyche with Cheumatopsyche (sensu Olah and Johanson 2008). Therefore, revised nomenclature is used in all figures and in Appendices 1 and 2 for clarity and ease of comparison to Schefter's (2005) topology.
Proposed classification
Hydropsychinae genera
Our analyses lead us to propose the following classification of the genera of Hydropsychinae:
Family Hydropsychidae Curtis, 1835
Subfamily Hydropsychinae Curtis, 1835
GENUS Abacaria Mosely, 1941
GENUS Calosopsyche Ross & Unzicker, 1977
GENUS Hydropsyche Pictet, 1834
SUBGENUS Hydropsyche Pictet, 1834
Hydropsyche bronta Group (generally corresponding with Ceratopsyche and H. morosa and newae Groups)
Hydropsyche colonica Group (generally corresponding with Orthopsyche)
Hydropsyche instabilis Group (generally corresponding with Hydropsyche s.s.)
Hydropsyche naumanni Group (generally corresponding with Occutanspsyche)
SUBGENUS Hydrocheumatopsyche 32Marlier, 1962
GENUS Cheumatopsyche Wallengren, 1891
SUBGENUS Abacarioides Marlier, 1961
SUBGENUS Achirocentra Marlier, 1961
SUBGENUS Cheumatopsyche Wallengren, 1891
SUBGENUS Cheumatopsychodes Marlier, 1961
SUBGENUS Ethiopsyche 32Marlier, 1962
GENUS +Electrodiplectrona Ulmer, 1912
GENUS Hydromanicus Brauer, 1865
Synonym GENUS Hydatopsyche Ulmer, 1926 (Olah and Johanson 2008:14)
GENUS Hydronema Martynov, 1914
GENUS +Palaehydropsyche Wichard, 1983
GENUS Plectropsyche Ross, 1947
GENUS Potamyia Banks, 1900
GENUS Schmidopsyche Olah & Schefter 2008
GENUS Streptopsyche Ross & Unzicker, 1977
Hydropsyche
Furthermore, our analyses lead us to recognize the following synonyms for Hydropsyche:
Genus Hydropsyche Pictet, 1834
Type species: Hydropsyche cinerea Pictet [subsequent designation Ross 1944:86, = Hydropsyche instabilis (Curtis, 1834)].
Synonym Aoteapsyche McFarlane, 1976, type species: Hydropsyche raruraru McFarlane (original designation); considered a synonym of Hydropsyche by Schefter 2005:148 (synonymized with Orthopsyche by Olah and Johanson 2008:164).
Synonym Caldra Navás, 1924, type species: Caldra nigra Navás (original designation); synonymized with Hydropsyche by Botosaneanu and Malicky 1978:344, synonymy not confirmed in this study.
Synonym Ceratopsyche Ross & Unzicker, 1977, type species: Hydropsyche bronta Ross (original designation); synonymized as a subgenus of Hydropsyche by Schefter et al. 1986:68, reduced to synonym of Hydropsyche by Olah and Johanson 2008:56.
Synonym Caledopsyche Kimmins, 1953, type species: Caledopsyche cheesmanae Kimmins (original designation); NEW SYNONYM.
Synonym Herbertorossia Ulmer, 1957, type species: Hydromanicus ungulatus Ulmer (original designation); synonymized with Hydropsyche by Schefter 2005:148 and Olah and Johanson 2008:56.
Synonym Hydatomanicus Ulmer 1951, type species: Hydromanicus verrucosus Ulmer (original designation); synonymized as a subgenus of Hydropsyche by Malicky and Chantaramongkol 2000:791–860 (considered a synonym of Hydromanicus by Olah and Johanson, 2008:14).
Synonym Mexipsyche Ross and Unzicker, 1977, type species: Mexipsyche dampfi Ross and Unzicker (original designation); synonymized with Hydropsyche by Schefter 2005:148 and Olah and Johanson 2008:56.
Synonym Occutanspsyche Li and Tian, 1989, type species: Hydropsyche polyacantha Li and Tian (original designation); described originally as a subgenus of Hydropsyche; reduced to NEW SYNONYM of Hydropsyche in this study.
Synonym Orthopsyche McFarlane, 1976, type species: Hydropsyche fimbriata McLachlan (original designation); synonymized with Hydropsyche by Schefter 2005:148.
Synonym Plesiopsyche Navás, 1931, type species: Plesiopsyche alluaudina Navás (original designation); synonym of Symphitopsyche according to Ross and Unzicker 1977:304–305, synonymy not confirmed in this study.
Synonym Symphitopsyche Ulmer, 1907, type species: Hydropsyche mauritiana McLachlan (monobasic); synonymized with Hydropsyche by Scott 1983:319, Schefter 2005:148, and Olah and Johanson 2008:56.
Genus diagnosis
The following synthetic diagnosis combines morphological characters described by Schefter (2005) and molecular characters from this study of the 28S rRNA D2 fragment and COI gene.
Adults (character numbers refer to those by Schefter 2005)
The pro-episternal setal wart (Character 4) is a synapomorphy for Hydropsyche (Schefter 2005). Other characters have been shown to vary in some Hydropsyche taxa (see Schefter 2005, for further discussion of each of these characters). Maxillae each has its 2nd maxillary segment subequal in length to its 3rd, and its 5th segment is subequal in length to its segments 1–4 combined (Characters 1 and 2) (Banks 1914, Ross 1944, Ulmer 1951). A tarsal setal bundle is present on each foretarsus of the male (Character 5) (Ulmer 1951). Each forewing has its crossvein cu at or near the thyridial nygma, not close to crossvein m–cu (Character 6) (Ross 1944, Ulmer 1951). Each hind wing has its crossvein m–cu present and conspicuous (Character 11). The dorsum of the head has 7 warts (Character 3). Posterior lobes are present on segments X–XI of the female (Character 40).
Larvae
The submentum is cleft (Character 46). The foretrochantin is biramous (Character 47), and a pair of large sclerites occurs in the intersegmental fold posterior to the prosternal plate in Hydropsyche species (Schuster and Etnier 1978, Morse and Holzenthal 2008), however, these character states also can be found in some Chinese Hydromanicus species (Zhou 2007).
DNA characters
All Hydropsyche species examined had an “A–C” bulge in stem 2–2′ of the 28S rRNA D2 fragment. The secondary structure of the D2 fragment for Hydropsyche instabilis (Curtis) is illustrated in Fig. 5. Hydropsyche also is characterized by 2 mtCOI amino acid changes: leucine to methionine at position 136 (out of 219 amino acids), and valine to isoleucine at position 172 (Fig. 4).
Distribution
Afrotropical (AT), Australasian (AU), East Palearctic (EP), Nearctic (NA), Neotropical (NT), Oriental (OL), West Palearctic (WP). The distribution of Hydropsyche within the Neotropical Region is limited and not yet fully known. Hydropsyche species have not been found in Australia, but are known from New Zealand, New Caledonia, Indonesia, and other islands in the Australasian Biogeographic Region.
Discussion
Our objectives were to test the monophyly of Hydropsyche (Schefter 2005) and its relationships to other hydropsychine genera based on DNA data. Both model-based and parsimony trees support a broad definition of Hydropsyche that is largely, but not entirely, congruent with that inferred from morphology data (Schefter 2005). Hydropsyche, as defined in our paper, has diagnostic morphological characters for both adults and larvae and from both the COI amino acid and 28S rRNA D2 data sets. However, the resolution and support beyond the Hydropsyche node is inconsistent (Fig. 2A, B). We support the use of species group names for groups whose members possess apomorphic, diagnostic morphological, behavioral, or ecological characters. Expanded taxon sampling from existing species groups and genera is needed to analyze the phylogenetic relationships within Hydropsyche. In particular, more sampling of the Nearctic and Oriental Hydropsyche bronta Group, Mexipsyche (generally corresponding to the Hydropsyche propinqua Group), and African Symphitopsyche species are needed to test the monophyly of those taxa and standardize species group terminology.
We suggest that the geographically restricted use of non-matrix-based interpretations of male genitalic structures to establish genera (e.g., Fig. 1A), in the absence of corroborating female, larval, or DNA characters, can contribute to nomenclatural instability. The link between genitalic diversity and stability of generic definitions in Hydropsychidae can be seen by comparing the taxonomic histories of Cheumatopsyche and Hydropsyche. Cheumatopsyche has not been split into multiple genera, and the description of Cheumatopsyche subgenera (Marlier 1961, 1962a, b) did not obscure the definition of the genus. If the Hydropsyche and Cheumatopsyche lineages have comparable distributions and species numbers (Table 3), why was Hydropsyche split in so many contested ways, whereas Cheumatopsyche was not? One possible reason is differing taxonomic philosophies of describers (i.e., lumpers vs splitters), but this reason is unlikely because many of the same authors described species from both groups. We suggest, in agreement with Schmid (1979), that perhaps too much emphasis was placed on phallic characters for defining higher-level taxa in Hydropsychidae systematics without considering that some of these structures might have evolved convergently or in parallel. Cheumatopsyche lacks the diversity in phallic structures that Hydropsyche species display (Schefter 2005, Korecki 2006). Furthermore, if only the North American fauna is considered, the distinction between the Hydropsyche and Ceratopsyche male genitalic forms is more pronounced because the species that display intermediate morphological forms (Fig. 2A, B) are not found in the Nearctic Region.
Table 3
Comparison of Hydropsyche and Cheumatopsyche species numbers, taxonomy, distribution, and phallic characters.
Phylogenetic relationships among Hydropsyche species groups might be illuminated by further examination of homology relationships among phallic morphology characters. However, insect male genitalia have been shown to be complex and subject to sexual selection (Eberhard 1985, 2004, Hosken et al. 2001, Hosken and Stockley 2004, House and Simmons 2005), and their evolution is driven by mating systems (Arnqvist 1998, Arnqvist et al. 2000) or coevolution via reproductive conflict (Cordoba-Aguilar 2002, Ronn et al. 2007). Relying on such potentially plastic characters to define or synonymize genera in the absence of corroborating evidence, or to infer phylogeny without firmly establishing homology among phallic characters, could lead to classification via functional analogy or convergent evolution instead of via shared ancestry. As DNA sequencing campaigns continue to assist in the association of life stages (Zhou et al. 2007, Zhou 2009), the immatures and females of more species will be described, and we will be able to use a combined evidence approach to revise species group relationships for Hydropsyche.
Our study demonstrates that revisionary taxonomy at the generic level is important to both basic phylogenetics and applied research. As we continue to gain appreciation for the value of combined data sets that include structural attributes from all life stages plus molecular characters from multiple genes, consistent generic definitions become increasingly important. Consistency is needed to avoid creating chimera taxa from unrelated species that happen to have the same genus name (e.g., H. canaliculatus + H. nr. truncatus) and using those chimera in combined evidence phylogenetic analyses. Both morphological and molecular data support Hydropsyche as a species-rich and widespread monophyletic lineage that is characterized by at least 2 synapomorphic amino acid changes in the mitochondrial COI genome and 1 secondary structural change in the nuclear 28S rRNA genome. The evolutionary history of Hydropsyche subgenera and species groups is long and complex (as evidenced by its wide geographic range; Appendix 1), probably with multiple colonization and extinction events at both local and global scales. Mey (2003) inferred that Southeast Asia was the center of taxonomic diversity for Hydropsyche, and that Hydropsyche species are relatively recent immigrants to the Afrotropics (Mey 2005). A revision of the World Hydropsyche species that examines morphological and molecular diversity across the entire geographic scope of the genus is needed to test such biogeographical and ecological hypotheses regarding the evolutionary history of species groups.
Last, our analysis provides a basic framework for future applied research on hydropsychid larvae. We know from decades of bioassessment data that larvae of North American Hydropsyche species display a particularly wide range of pollution tolerance values (Resh and Unzicker 1975, Lenat 1993, Lenat and Resh 2001), but we do not yet know why. Is phylogenetic signal inherent in this pattern: i.e., are sister species more likely to have similar tolerance values than are more distantly related species within the Hydropsychinae? Molecular approaches like DNA barcoding are being used to expedite larval–adult associations (Zhou et al. 2007), and biomonitoring programs continue to expand worldwide (Morse et al. 2007). Differences in larval physiology and behavior might explain why different Hydropsychidae species have different pollution tolerance values (Petersen and Petersen 1984, Vuori 1994, Vuori and Kukkonen 1996, Tessier et al. 2000a, b, c, d, Illes et al. 2001, Buchwalter and Luoma 2005, Buchwalter et al. 2008). Our ability to study the mechanisms driving those physiological and behavioral differences in a phylogenetic context will depend on taxonomists and ecologists in different parts of the world basing hydropsychine generic classification on monophyly and using the name Hydropsyche consistently.
Acknowledgments
The following people and institutions provided specimens for this analysis: R. Blahnik, D. Cartwright, F. deMoor, T. L. Erwin, O. S. Flint, L. Harvey, H. Hoang, R. Holzenthal, K. A. Johanson, N. Sangpradub, D. Sembel and colleagues, I. Stocks, T. Tsuruishi, J. Ward, A. Wells, B. Welter, L. Yang, the Bishop Museum, Nanjing Agricultural University (China), the National Museum of Natural History (Smithsonian Institution), the University of Minnesota Insect Collection, and the University of Sam Ratulangi (Sulawesi). We thank P. H. Adler and M. W. Turnbull for manuscript reviews, T. L. Erwin for valuable advice on both content and style, and J. Korecki for suggestions on genitalic terminology. M. W. Turnbull also provided laboratory facilities and technical guidance. This research was funded by National Science Foundation (NSF) grant DEB0316504 and DEB081686. The acquisition of DNA barcodes was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) grant and by grants from Genome Canada through the Ontario Genomics Institute to P. D. N. Hebert. This is Technical Contribution Number 5575 of the Clemson University Experiment Station.
Literature Cited
Appendices
Appendix 1
Exemplar specimens used for the analyses with the D2COI data set (mitochondrial cytochrome c oxidase subunit I [COI] and D2 region of nuclear large subunit ribosomal ribonucleic acid ([28S D2]), their collection locality, GenBank accession numbers, and BOLD (Barcode of Life Data System; Ratnasingham and Hebert 2007) sample identification (ID) numbers.
Appendix 2
Exemplar specimens used for the analyses with the 28SCOI data set (nuclear large subunit ribosomal ribonucleic acid (rRNA) [28S] regions D1, D2, and D3 and mitochondrial cytochrome c oxidase subunit I [COI]), their collection locality, GenBank accession numbers, and BOLD (Barcode of Life Data System; Ratnasingham and Hebert 2007) sample identification (ID) numbers. n.s. = not sequenced.